Review of research on the influences on personal well-being and application to policy making

Final report for Defra Review of research on the influences on personal well-being and application to policy making Professor Paul Dolan*, Ms. Tessa...
Author: William Lloyd
28 downloads 3 Views 3MB Size
Final report for Defra

Review of research on the influences on personal well-being and application to policy making

Professor Paul Dolan*, Ms. Tessa Peasgood Tanaka Business School, Imperial College London Dr. Mathew White Centre for Well-being in Public Policy, University of Sheffield 24 August 2006

* Corresponding author: [email protected]

1

ACKNOWLEDGEMENTS We would like to thank participants in a stakeholder meeting that helped us identify some of the measures of well-being to include in the review and helped us develop further the assessment criteria in section 2. We would like to thank Jacque Mallender for chairing this meeting. We are grateful to the steering group and to Ed Diener for providing comments on sections 1-3 of this report. We also appreciate the working papers that were sent to us by our academic colleagues. Data from the British Household Panel Survey were provided through the Data Archive at the University of Essex. The data were originally collected by the ESRC Research Centre on Micro-social Change at the University of Essex. Neither the original collectors of the data nor the Archive bear any responsibility for the analyses or interpretations presented here. We would like to thank Julie Newton for helping to improve the final version of this report. Finally, a big thank you is owed to Isabella Earle, who was involved in funding this research and has provided us with excellent help and support throughout. Although Defra has commissioned and funded this study, the views expressed in it do not necessarily reflect Defra policy.

2

CONTENTS EXECUTIVE SUMMARY .........................................................................................7 1. BACKGROUND ....................................................................................................12 2. DEFINING AND MEASURING WELL-BEING...............................................14 2.1 Concepts of well-being ......................................................................................14 2.1.1 Objective lists..............................................................................................14 2.1.2 Preference satisfaction ................................................................................14 2.1.3 Flourishing accounts ...................................................................................15 2.1.4 Hedonic accounts ........................................................................................16 2.1.5 Evaluative accounts ....................................................................................16 2.1.6 Conclusion ..................................................................................................16 2.2 Operational definitions of well-being ................................................................16 2.2.1 Time frame..................................................................................................17 2.2.2 Reference standards ....................................................................................17 2.2.3 Sensitivity ...................................................................................................17 2.2.4 Reliability....................................................................................................17 2.2.5 Cardinality...................................................................................................18 2.2.6 Interpersonal comparability ........................................................................18 2.3 Policy evaluation................................................................................................18 3. REVIEW OF MEASURES OF PERSONAL WELL-BEING...........................19 3.1 Preference satisfaction accounts ........................................................................19 3.1.1 Income.........................................................................................................19 3.1.2 Quality-adjusted life years ..........................................................................21 3.2 Flourishing accounts ..........................................................................................22 3.2.1 Psychological well-being scale ...................................................................22 3.2.2 Orientation to happiness .............................................................................23 3.3. Hedonic accounts ..............................................................................................24 3.3.1 Positive and negative affect scale ...............................................................24 3.3.2 Affectometer 2 ............................................................................................25 3.3.3 Day reconstruction method .........................................................................25 3.4 Evaluative accounts ...........................................................................................26 3.4.1 Satisfaction with life scale ..........................................................................26 3.4.2 Personal well-being index...........................................................................27 3.4.3 Life satisfaction...........................................................................................28 3.5 Combined accounts............................................................................................29 3.5.1 Centre for Epidemiological Studies Depression Scale (CES-D) ................29 3.5.2 CASP-19 .....................................................................................................29 3.5.3 General Health Questionnaire.....................................................................30 3.6 Summary of review of measures........................................................................31 4. REVIEW OF THE FACTORS AFFECTING WELL-BEING .........................33 4.1 Review strategy..................................................................................................33 4.2 Factors associated with personal well-being......................................................34 4.2.1 Income.........................................................................................................34 4.2.1.1 Absolute income ..................................................................................35 4.2.1.2 Relative income ...................................................................................39 4.2.1.3 Wealth ..................................................................................................42 4.2.1.4 Debt......................................................................................................42 4.2.1.5 Expectations and perceptions...............................................................42 4.2.2 Personal characteristics (who we are, our genetic makeup) .......................43

3

4.2.2.1 Age.......................................................................................................44 4.2.2.2 Gender..................................................................................................44 4.2.2.3 Ethnicity...............................................................................................45 4.2.2.4 Personality............................................................................................45 4.2.2.5 Physical characteristics ........................................................................46 4.2.3 Socially developed characteristics (human and physical capital)...............46 4.2.3.1 Education .............................................................................................46 4.2.3.2 Health...................................................................................................48 4.2.3.3 Type of work........................................................................................48 4.2.3.4 Unemployment.....................................................................................49 4.2.4. How we spend our time (The work and activities we engage in)..............50 4.2.4.1 Hours worked.......................................................................................50 4.2.4.2 Commuting ..........................................................................................51 4.2.4.3 Housework ...........................................................................................51 4.2.4.4 Caring for others ..................................................................................52 4.2.4.5 Community involvement and volunteering .........................................52 4.2.4.6 Sleep.....................................................................................................53 4.2.4.7 Exercise................................................................................................53 4.2.4.8 Religious activities...............................................................................53 4.2.5 Attitudes and beliefs towards self/others/life..............................................54 4.2.5.1 Attitudes towards our circumstances ...................................................54 4.2.5.2 Trust .....................................................................................................55 4.2.5.3 Political persuasion and attitudes.........................................................55 4.2.5.4 Religion................................................................................................56 4.2.6. Relationships..............................................................................................56 4.2.6.1 Marriage and intimate relationship ......................................................56 4.2.6.2 Having children....................................................................................58 4.2.6.3 Seeing family and friends ....................................................................58 4.2.7 Wider economic, social and political environment (Where we live)..........59 4.2.7.1 Income inequality.................................................................................59 4.2.7.2 Unemployment rates ............................................................................60 4.2.7.3 Inflation................................................................................................60 4.2.7.4 Welfare system and public insurance ..................................................60 4.2.7.5 Degree of democracy ...........................................................................61 4.2.7.6 Climate and the natural environment ...................................................61 4.2.7.7 Safety and deprivation of the area .......................................................61 4.2.7.8 Urbanisation.........................................................................................61 4.3 Summary of existing evidence...........................................................................62 5. ANALYSIS OF THE BRITISH HOUSEHOLD PANEL SURVEY.................64 5.1 Well-being measures in the BHPS.....................................................................64 5.2 Comparing the measures....................................................................................65 6. CONCLUDING REMARKS ................................................................................70 6.1 Existing evidence ...............................................................................................70 6.2 Further research .................................................................................................74 6.2.1 Measuring well-being .................................................................................74 6.2.2 Methodological challenges .........................................................................74 6.2.3 Key factors for future research ...................................................................75 6.2.3.1 Income rank .........................................................................................75 6.2.3.2 Education .............................................................................................75 6.2.3.3 Social capital........................................................................................75

4

6.2.3.4 Other factors.........................................................................................76 6.2.4 Well-being in policy ...................................................................................76 GLOSSARY OF TERMS..........................................................................................77 SECTION 5 TABLES AND FIGURES ...................................................................81 5.1 The McClements equivalent income scale ........................................................81 5.2 Correlations between different well-being measures.........................................81 5.3 Average ranks in well-being across different groups ........................................82 5.4: Determinants of life satisfaction, inverse-GHQ, SF-6D, and CASP-19 using OLS and OLS fixed effects, BHPS various years....................................................85 REFERENCES...........................................................................................................88 APPENDIX A: DESCRIPTION OF MEASURES .................................................97 Table A.1: Affectometer 2 Scale .............................................................................97 Table A.2: CASP 19 Scale.......................................................................................98 Table A.3: Centre for Epidemiological Studies Depression (CES-D) Scale ...........99 Table A.4: Day Reconstruction Method (DRM) Scales ........................................100 Table A.5: General Health Questionnaire (GHQ) Scale........................................101 Table A.6: Happiness/life satisfaction Single-Item Measures...............................102 Table A.7: Orientations to Happiness (OTH) Scale ..............................................106 Table A.8: Personal Well-Being (PWI) Scale .......................................................107 Table A.9. Positive and Negative Affect Scales (PANAS) ...................................108 Table A.10: Psychological Well-Being Scales (PWBS)........................................109 Table A.11: Satisfaction With Life Scale (SWLS) ................................................110 APPENDIX B: SUMMARIES ................................................................................111 B.1. Income ............................................................................................................112 B.2. Personal characteristics – who we are, our genetic makeup ..........................126 Table B.2.1: Age ................................................................................................126 Table B.2.2: Gender...........................................................................................131 Table B.2.3: Ethnicity ........................................................................................135 Table B.2.4: Personality.....................................................................................138 Table B.2.5: Physical characteristics .................................................................139 3. Socially developed characteristics – our human and physical capital ...............140 Table B.3.1: Education ......................................................................................140 Table B.3.2: Health ............................................................................................145 Table B.3.3: Type of work.................................................................................149 Table B.3.4: Unemployment..............................................................................151 B.4. How we spend our time..................................................................................156 Table B.4.1: Hours worked................................................................................156 Table B.4.2: Commuting....................................................................................158 Table B.4.3: Housework ....................................................................................159 Table B.4.4: Caring for others ...........................................................................160 Table B.4.5: Community involvement and volunteering...................................161 Table B.4.6: Sleep..............................................................................................162 Table B.4.7: Exercise.........................................................................................163 Table B.4.8: Religious practice..........................................................................164 B.5. Attitudes and beliefs towards self/others/life .................................................166 Table B.5.1: Attitudes towards our circumstances ............................................166 Table B.5.2: Trust ..............................................................................................168 Table B.5.3: Political persuasion .......................................................................169 Table B.5.4: Religious beliefs............................................................................170 B.6. Relationships ..................................................................................................172

5

Table B.6.1: Marriage/intimate relations ...........................................................172 Table B.6.2: Having children.............................................................................180 Table B.6.3: Contact with family and friends....................................................183 B.7. Wider economic, social and political environment – where we live .............185 Table B.7.1: Income inequality..........................................................................185 Table B.7.2: Unemployment rates .....................................................................187 Table B.7.3: Inflation.........................................................................................188 Table B.7.4: Welfare and public insurance........................................................189 Table B.7.5: Democracy ....................................................................................190 Table B.7.6: Climate & quality of natural environment (pollution) ..................191 Table B.7.7: Security of local environment (crime rates/risk)...........................192 Table B.7.8: Urbanisation ..................................................................................193 APPENDIX C: INDIVIDUAL STUDY SUMMARIES…………………………194

6

Executive Summary

EXECUTIVE SUMMARY The UK Government’s Sustainable Development Strategy, ‘Securing the Future’, (2005) commits the Government to get a better understanding and focus on well-being (www.sustainable-development.gov.uk/publications). The specific requirement for this project was to “review the evidence relating to the causative factors associated with various concepts and components of well-being”, with a special focus on personal well-being. In order to do this, we: 1. Set out the various concepts or accounts of well-being and consider the important ways in which measures under each account of well-being may differ (Section 2) 2. Evaluate a range of measures according to the potential differences (Section 3) 3. Undertake an extensive review of the existing literature on the relationship between a range of economic, social and environmental factors associated with well-being as defined according to the different measures (Section 4) 4. Analyse the British Household Panel Survey to highlight the extent to which different measures of well-being produce similar results (Section 5) Concepts of well-being There is a need to move beyond objective lists of well-being that tend to focus on aggregate or social well-being, this review focuses on accounts and measures of personal well-being. We identify four main accounts of personal well-being: 1. Preference satisfaction – based on fulfilling our desires 2. Flourishing accounts – based on the satisfaction of certain psychological needs 3. Hedonic accounts- based on how we feel 4. Evaluative accounts- based on how we think and feel Because of the overlap between hedonic and evaluative accounts, we collectively refer to them as ‘subjective well-being’ (SWB). Review of measures Different policies may result from a focus on one account of well-being compared to another. Similarly, different implications may be drawn from the use of one type of measure within a given account. We therefore consider a number of ways in which the measures of well-being may differ within an account and across studies. The measures include income, life satisfaction questions and the General Health Questionnaire (GHQ), which investigates psychological health. The review considered whether the measures were sufficiently robust to provide evidence to determine which factors influence wellbeing. From the first stage of review, we conclude that the measures of preference satisfaction currently available (e.g. income) offer an incomplete picture of individual well-being. It is difficult to rule out any of the other measures from a theoretical basis, although the evidence suggests that measures of flourishing accounts of well-being may not be sufficiently reliable for use in a UK policy context.

7

Executive Summary

Review of factors associated with well-being Whether the choice of measure matters (or not) depends partly on whether different measures are correlated with, or caused by, a different set of economic, social and environmental factors. Our review of evidence in relation to this concentrates on large datasets where there is an opportunity to isolate the relationship between a given factor and well-being through controlling for other factors and, in some limited cases, to make inferences about whether that factor might cause changes in well-being. The search for relevant literature in economics and psychology identified a total of 153 papers. To enhance the clarity and transparency of our review, we detail each study reviewed and provide summaries of all the studies that have considered a given characteristic and its relationship to well-being. In evaluating the existing evidence, we give more weight to those studies that control for as many of the other factors as possible. Emphasis is also given to evidence from the UK. Most of the evidence relates to measures of SWB. The potential influences on well-being emerging from our review were categorised under seven broad headings: 1. Income – there is evidence of diminishing marginal returns to income i.e. the relationship with SWB gets weaker as income rises; relative income is shown to have significant negative relationship with SWB; there is some indication that individuals adapt to changes in income levels and that subjective assessments of financial position are important to SWB 2. Personal characteristics – age has a U shaped relationship with SWB, with SWB lowest around 35-50 and women tend to score lower on mental health measures than men but there is a wide degree of within-gender variance 3. Socially developed characteristics – the relationship between education and SWB is indeterminate, SWB is strongly related to health, particularly psychological health, and unemployment is highly detrimental to SWB, although the effect is moderated by living close to others who are unemployed 4. How we spend our time – the evidence suggests that more activity, be it formal (e.g. paid work), informal (e.g. some forms of volunteering), social (e.g. church attendance) or physical (e.g. taking walks) was generally associated with higher SWB and that commuting and informal care giving are detrimental to SWB. 5. Attitudes and beliefs – the perceptions of our circumstances matter to our SWB, the degree of trust in others seems to be positively correlated with SWB but the evidence is very limited, belief in god is positively associated with well-being 6. Relationships – seeing family and friends and an intimate relationship are associated with higher SWB and the breakdown of that relationship is strongly detrimental to well-being 7. Wider economic, social, political and natural environment – there is very little robust evidence on the relationship between well-being and the economic, social, political and natural environment we live in (inflation, unemployment rates, income inequality, crime etc.) largely because it is very difficult to control for the range of other variables that will affect these relationships

8

Executive Summary Analysis of the British Household Panel Survey To explore the relationship between a range of economic, social and environmental factors associated with wellbeing and investigate how this varies according to different wellbeing measures in a UK context, we conducted primary analysis on the British Household Panel Survey (BHPS). We focused on this dataset because it is a longitudinal survey of about 5,000 households (10,000 individuals) that has been running since 1991, and it contains data on a range of well-being measures, including income, annual household consumption, life satisfaction and GHQ. Analysis of the average well-being of certain key subgroups (those aged over 70, the disabled, single parents, males, commuters) shows important differences by measure, with the most striking differences between the income and consumption measures (preference satisfaction accounts) and the measures of SWB. However, our analysis also confirms many of the findings of the review: 1. Income – having a problem paying for accommodation, which is related in some respects to debt, was consistently associated with poorer well-being. 2. Health – having problems walking, regularly visiting a GP and poor self-rated health are associated with lower well-being, as is being a full-time carer. 3. Employment status – being unemployed and being long-term sick are also associated with lower well-being. 4. Relationships – being divorced or separated and living alone is associated with poorer well-being across all measures. 5. Talking frequently to neighbours was the only example where well-being was consistently positively associated with well-being across all measures. Recommendations There are a number of key issues for future research that emerge from this project. There are important questions about precisely which measures of well-being should be used, methodological challenges about how to interpret the evidence, some specific factors that should be explored further given their potential policy relevance, and important questions about the distribution of well-being across society and the general public’s views on well-being in policy. Measuring well-being Income is an incomplete measure of well-being as defined by the satisfaction of preferences and so we should be very careful about how we interpret the well-being effects of changes in income. Given the lack of reliable evidence on measures of flourishing, future research should consider the degree to which flourishing accounts of well-being produce different results from hedonic or evaluative accounts. Where there is a commitment to the routine assessment of SWB, a global measure of life satisfaction on a 0-10 scale should provide reliable information in many policy contexts. For those interested in understanding the relationship between how we allocate our time and our well-being, the day reconstruction method (DRM), which measures feelings as recalled from activities during the previous day, offers a promising avenue for future research.

9

Executive Summary

Methodological challenges One of the main findings of this research is that different results can be obtained according to how the factors are categorised and which other factors are controlled for in the analysis. Therefore, researchers and policy-makers and researchers need to be careful in interpreting both significant and non-significant findings. Future research interested in the impact of specific variables on well-being should systematically introduce different control variables and explore the impact this is having on the relationship between the main independent and dependent variable. Whilst all researchers will be limited by their data, as consensus is reached on the main causes of well-being, this should enable greater consistency across studies. One very firm conclusion that can be drawn from our review is that the existing evidence base is not quite as strong as some people may have suggested and there are some important avenues for future research that could be explored with the existing panel datasets. This, in addition to the lack of clear evidence on causality, makes it difficult for us to make clear policy recommendations at this stage. Nevertheless, our findings suggest researchers should at least be aware of the impact of income, relative income, health, personal and community relationships and employment status in their analysis. We are also able to make some clear recommendations about where future research into some of these and other policy relevant variables should be directed. Key factors for future research Income rank – The importance of income rank is just beginning to be recognised, and so more research is needed to understand how income rank impacts upon well-being, and how income comparisons work. This would include exploring to whom people compare themselves to and a greater understanding of precisely why and how reference incomes impact on well-being is also needed. Education – There is a need to investigate the ambiguous relationship between education and well-being. The coefficient on education is often responsive to the inclusion of other variables within the model and there is a suggestion that, like income, the benefits to education may be positional rather than absolute. The effect of social status and rank across a range of domains in life is therefore something that requires urgent attention. Social capital – The BHPS analysis suggests that social contact is positively associated with well-being and so future research is needed to understand the link between contact with friends, family and neighbours and well-being and, critically, the direction of causality in this relationship. Environmental factors – There is very little evidence in this regard and so future studies should be conducted that focus on the effects of green space, pollution etc.

10

Executive Summary Well-being in policy An important issue which has been largely ignored within the literature is the distribution of well-being across society and across different population sub-groups. There are many policy relevant questions which could be explored with the existing panel datasets, such as: how does the distribution of well-being differ according to different measures of well-being; how does the distribution of well-being compare to other distributions such as income; and how has the distribution of well-being changed over time? There is also the need for new data in relation to how the general public feel about using particular measures of personal well-being to inform public policy. This would be one way in which the largely normative question of the policy relevance of different measures of personal well-being could be addressed with empirical evidence. From the existing evidence, it appears that measures of SWB would produce different results to preference satisfaction measures, such as income. However, despite many caveats and uncertainties about how to interpret some of the existing evidence, it would seem that most measures of SWB would produce similar results to one another. Therefore, for those interested in the subjective assessment of an individual’s life, it might not matter too much which measure of SWB is used to assess the well-being of different population groups.

11

Background

1. BACKGROUND Policy-makers are increasingly interested in incorporating subjective well-being into their policy decisions, and in how well-being is affected by a range of economic, social and environmental factors. The UK Government’s Sustainable Development Strategy, ‘Securing the Future’ (HM Government, 2005), commits the Government to get a better understanding and focus on well-being.1 Against this background, the detailed requirements of the project specification were as follows: ‘This study will review the evidence relating to the causative factors associated with various concepts and components of well-being. It will … address the question: “What are the main influences on wellbeing: how far do these differ between, or coincide for, different definitions of wellbeing and why?”’ An extensive review of the relevant literature and primary analysis of the British Household Panel Survey (BHPS) are used to address these issues. Policy-makers are faced with a number of important obstacles before they can interpret the existing evidence. First, they are faced with an array of seemingly very different measures claiming to measure well-being and subjective well-being. For example, improvements in material circumstances would be sufficient for some to claim improved well-being whilst, for others, such changes are only important to the extent that they result in positive hedonic experience. Second, how the concepts are measured may vary from study to study even if the concept of well-being does not. For example, the time frame that respondents are asked to consider may vary. Third, some measures may lack practical usefulness within a policy context in that UK data may not be available or prohibitively expensive to collect, or the data that is available may lack policy relevance in that the data may not be perceived by policy makers or the public to be a suitable target for government policy. All of this means that, even if causality could be established (which it rarely can be conclusively), it may still be impossible to compare the results and conclusions of one study with those from another study. Therefore, in section 2, we set out the five main accounts of well-being (objective lists, flourishing, preference satisfaction, hedonic and evaluative) and consider six important ways in which the measures of personal well-being may differ across studies (including the time frame of assessment and the reference standards used). In addition, we consider the issues of practicality and policy relevance. In section 3, we review some of the more recognised and widely used measures of personal well-being under each of the accounts relevant to this project (that is, excluding the objective list account) and consider how the measures have been applied and whether they are potentially useful in a UK policy context. The measures reviewed are ones that have been widely used in the literature or that have been recommended for use for evaluative purposes, and the list of measures was drawn up in consultation with policy-makers across a range of government departments. Many of the measures, particularly those that cut across different accounts of well-being, are

1

See: http://www.sustainable-development.gov.uk/publications/pdf/strategy/SecFut_complete.pdf

12

Background ones that have been used in the health domain, which reflects the fact that the measurement of well-being – or quality of life – is more developed in this area. Once the measures have been categorised and evaluated, it will then be possible to more accurately assess whether different measures of well-being produce similar results; that is, whether they are affected by the same set of variables, whether they produce the same rankings of individuals and groups and whether they would have the same implications for policy. The review, detailed in section 4, considers whether the evidence on the influences on personal well-being is sufficiently robust to be taken on board by policy-makers. We provide summaries for all the characteristics reviewed as well as detailed accounts of all the papers reviewed so that our interpretation of the existing evidence is clear and transparent. Our focus is on research conducted on large datasets where there is an opportunity to isolate the impact of one factor upon wellbeing through controlling for other factors. In section 5, and building on the review, we present some fresh analysis of the BHPS in order to highlight the extent to which different measures of well-being can produce different results. The BHPS is an annual longitudinal survey of around 5,000 households, resulting in around 10,000 individual interviews. The BHPS contains data on a range of well-being measures and so it allows us to make direct comparisons between them In section 6, we make some specific recommendations about how policy-makers might use and develop the existing evidence base. In many cases, where findings are relatively consistent across different measures of well-being (e.g. regarding unemployment), policy-makers can have reasonable confidence in the data but in other cases (e.g. in relation to the effect of education or income), further research is needed before it becomes possible to draw clear inferences about which policies might impact upon well-being.

13

Defining and measuring well-being

2. DEFINING AND MEASURING WELL-BEING 2.1 Concepts of well-being There are essentially five basic approaches to defining well-being and we briefly set these out below. 2.1.1 Objective lists Objective list accounts do not provide a formal theory of well-being; rather they offer a list of attributes and characteristics which are taken to constitute well-being. The contents of the list vary but tend to include items such as economic resources, political freedom, good health, and the ability to read. In a similar vein, basic needs theorists assert that important and politically relevant aspects of well-being are whether essential human needs are met. Basic needs accounts offer neither a theory nor a measure of individual well-being, but rather suggest attributes which are essential means for creating well-being. Rawls (1971) developed an index of primary goods, which included rights, liberties and opportunities, income and wealth, opportunities and the bases of self-respect. Max Neef (1992) organised human needs into a matrix of nine fundamental categories (subsistence, protection, affection, understanding, participation, idleness, creation, identity and freedom), seeing each need arising at four different levels of activity: being, having, doing and interacting. More recently, Ryan and Deci’s selfdetermination theory (Ryan and Deci, 2000) posits three basic psychological needs (autonomy, competence and relatedness) which are necessary for individual wellbeing. The presence or absence of certain objective attributes may lead to more preferences being satisfied or better experiences in life but the value of these objective attributes is independent of these consequent effects. Therefore, the judgement about the contribution of various things (education, health etc.) towards well-being does not come from the individual but draws on theoretical and intuitive accounts of what is of value. Objective list accounts measure well-being in terms of objective outcomes such as literacy rates and suicide rates. Whilst some of these items may be measured from a subjective perspective, their importance to well-being is determined externally. 2.1.2 Preference satisfaction According to this account, an individual’s life goes better for her if she gets what she wants. In the simplest versions of this account, there are no constraints on what an individual can want; for example, she is entitled to want things that do not make her feel better off and she may want things that are not related to her own experience, such as desire for the truth (Parfit, 1984). All that matters for her well-being is whether a desire is met. More recent formulations of preference satisfaction require that preferences are informed in the sense that they are based on the considered use of

14

Defining and measuring well-being all relevant information and some accounts exclude certain ‘anti-social’ preferences, such as those related to envy, even when they are informed (Harsanyi, 1996). For the best part of the last one hundred years, most economists have adopted a preference satisfaction account of well-being. Through the market mechanism, a consumer expresses his willingness to pay for a range of goods and services, and what he is willing to pay reflects his preferences. All else equal, if an individual’s income increases, she is able to satisfy more of her preferences. It is not the income per se that makes her better off but, rather, the increase in choice that means she can satisfy more of her desires. It is not surprising, then, that most economics textbooks introduce a utility function in which utility (i.e. individual well-being) is increasing in income (and often increasing in income alone). Most public spending generates important benefits for which market data do not exist e.g. a range of health effects. Economists have developed methods for valuing these effects, which involve inferring a monetary value from observing consumer behaviour, or ‘revealed preference’ (Day, 2003). Where the value of benefits cannot be inferred from revealed preferences, ‘stated preference’ methods have been developed, which elicit monetary values through hypothetical choices presented to respondents (Jones-Lee, 2002). Attaching monetary values to changes in quality of life is particularly problematic, and so health economists have developed alternative preference-based methods for valuing health states, which express health on a scale between zero for death and one for full health (Dolan, 2000). These values can then be combined with duration to calculate quality adjusted-life years (QALYs). 2 2.1.3 Flourishing accounts Aristotle proposed a perfectionist version of well-being in which the well-being of an individual is judged by considering how close they are to reaching the potential of humankind. Aristole’s term for this, eudaimonia, has been translated as flourishing, happiness or well-being, and accounts have developed to include not just biological and species perfectionism, but also fulfilment at an individual level, including achievement of individual goals and realisation of an individual's true potential. Ryff and colleagues have developed a measurable account of flourishing in terms of psychological well-being (PWB). This is represented by six aspects of human potential: autonomy, personal growth, self-acceptance, life purpose, mastery and positive relatedness. These can all be seen as essential components of what it is to be a flourishing human being (Ryff and Keys, 1995). Some of the things on an objective list (freedom, education, nourishment, shelter etc.) are important only insofar as they help create an environment in which people can express higher psychological functioning. PWB is measured according to the extent to which people have autonomy etc. which is usually determined by responses to various self-reports. 2

QALYs are considered under the preference satisfaction account because the quality-adjustment weights for health states are cureently derived using preference-based methods that require respondents to make trade-offs between changes in health status and changes in the risk of death or life expectancy (Dolan, 2000). There is nothing in principal which precludes health state values from being based upon the SWB directly associated with those states, and there are strong grounds for arguing that this should indeed be the case (see Dolan and Kahneman, 2006).

15

Defining and measuring well-being

2.1.4 Hedonic accounts Hedonism takes the view that pleasure is the only thing that is good for us, and pain is the only thing that is bad (Bentham, 1789). It is problematic to identify a state as pleasurable or painful when there may not be a specific feeling which is common to all pleasurable or painful experiences. Therefore, recent theorists have considered alternative ways of knowing the extent of pleasure, such as the degree to which we wish feelings to be prolonged (Crisp, 2003) or the extent of pro-attitude towards the feelings, or our desires for particular feelings (Feldman, 2002). At an empirical level, researchers have tended to include a range of positive and negative feelings and emotions besides what we commonly take to be pleasure and pain. For example, the day reconstruction method is used to measure the frequency and intensity of positive and negative emotions over time (Kahneman et al., 2004a). 2.1.5 Evaluative accounts Sumner (1995) argues that preferences and feelings each focus on one aspect of how an individual’s life can be thought of as going well. An informed individual’s assessment of his life overall can incorporate each of these aspects. An evaluative account therefore reflects the weights we attach to our hedonic experiences alongside our assessments of how well life is going for us more generally. An individual’s assessment of their life has become to be understood in the literature as their subjective well-being (SWB), which is an umbrella term for how we think and feel about our lives (Diener et al., 1999). 2.1.6 Conclusion These theoretical differences are much more than different methods for measuring the same thing. Different conceptions of well-being adopt different perspectives from which to judge the individual, which makes direct comparisons between accounts potentially meaningless. Whether or not the measures produce different results in practice is of course one of the main questions for this research.

2.2 Operational definitions of well-being Different policies may result from a focus on one account of well-being as compared to another. In addition, it is possible that different implications may be drawn from the use of one type of measure within a given account. In order to highlight the potentially important – and policy relevant – differences between measures, we outline six ways in which measures within the same account of well-being may differ. In addition, the measurement of well-being can be based on self-reports or on external assessment (e.g. Michalos, 2004). However, most accounts and measures we review here rely on self-reports, so we only discuss the issue of self versus observer reports only where appropriate.

16

Defining and measuring well-being 2.2.1 Time frame The time frame over which well-being is measured may vary across studies. For instance, it is possible to look at current income in the current month or average income over the last 12 months (which could form part of an objective list or proxy for a preference satisfaction account). Similarly, positive affect can be examined over the course of a single day or an entire week (hedonic accounts). 2.2.2 Reference standards Different measures may use different explicit or implicit reference standards (Michalos, 1985). For example, two measures of an evaluative account may differ because one asks people to compare their life satisfaction with "others in your household" while another asks them to compare it with "the average UK citizen". In addition to their being potential differences in the standards of social comparison there are also potential differences in temporal comparisons ("compared to last year versus last month") and aspirational comparisons ("compared to what you expected versus what you dreamed of"). People will not necessary use these explicit reference standards in any systematic way (Klar and Giladi, 1999) and a variety of reference standards may be used even when not explicitly stated (Schwarz and Strack, 1999). 2.2.3 Sensitivity Results may also differ within accounts due to differences in the sensitivity of the measures. Sensitivity refers to the degree to which measures are able to distinguish between different states of well-being. A measure using a ten point scale is potentially more sensitive to change than a three point scale. A multi-item scale is generally more sensitive than a single-item scale. An insensitive measure will fail to detect important differences or changes in well-being. However, an overly sensitive measure may detect differences or changes that are not meaningful or relevant to the individual. The larger the sample size, the easier it is to find statistically significant differences even with relatively insensitive scales but greater sample size also increases the possibility of finding irrelevant differences, especially for multi-item, highly sensitive scales. Finally, sensitivity should be similar across the full range of the measure to avoid both ceiling and floor effects (the failure to distinguish differences at very high or low levels of well-being). 2.2.4 Reliability Measures may also be more or less reliable in terms of stability of assessment across people or over time (assuming no significant changes in the meantime). That is, it should have low measurement error. To some extent, reliability is traded-off against sensitivity. In general, less sensitive scales are more reliable.

17

Defining and measuring well-being

2.2.5 Cardinality Cardinality concerns the extent to which increments in a scale validly represent increments in the underlying concept. Cardinality is achieved when an increase on, say, a seven point life satisfaction scale from 2 to 6 reflects an increase in well-being twice as large as an increase from 2 to 4. If the scale is not cardinal, inferences can only be drawn in an ordinal fashion i.e. 6 is more than 4 but we do not know by how much. 2.2.6 Interpersonal comparability Interpersonal comparability refers to the degree to which the responses of different individuals can be meaningfully compared. Put simply, does person A's 7 on the scale reflect the same underlying level of well-being as person B's 7? Where there is evidence of agreement between self and observer ratings a certain degree of interpersonal comparability can be assumed since the respondent and the observer appear to be using the scale in a similar way (Lepper, 1998; Sandvik et al., 1993). When large scale comparisons between groups, rather than between individuals, are examined the issue of interpersonal comparability becomes less important. As long as there is no reason to suppose systematic variance in scale use as a function of group membership, any differences between groups are likely to be meaningful, irrespective of any variation in which the scales are used within groups.

2.3 Policy evaluation Measures of well-being also need to be practical. Observer assessments of the "flourishing" of a representative sample of the UK population over an entire year would involve considerable resources and, if a simpler (e.g. self-report) measure produces essentially the same results, it might be favoured for its cost-effectiveness. Other issues with respect to practicality are whether the data is currently collected in the UK and whether it has been used with representative samples of the population. A slightly different consideration is policy relevance. This considers whether a measure of well-being is credible and whether it is potentially useful or relevant to policymakers. The main issue is whether the measure is focused on aspects of well-being that are considered to be the jurisdiction of government (e.g. literacy rates) or whether it includes aspects that are widely viewed as being beyond the proper realm of policy (e.g. sexual satisfaction).

18

Review of measures of personal well-being

3. REVIEW OF MEASURES OF PERSONAL WELL-BEING In response to the need to move beyond objective accounts and measures of wellbeing which have dominated policy until now, the focus of this review is on measures of personal well-being, with specific reference to subjective well-being. Therefore, we do not review measures of objective lists since they generally focus on aggregate or social well-being (e.g. Human Development Index, Index of Economic Well-being etc.) and do not take account of how individuals think and feel about their lives. We use the term well-being as shorthand for personal well-being and use SWB where appropriate. We do review two preference satisfaction measures – income and QALYs – in this section for comparison purposes, since these are currently used by the UK government as measures of well-being and as direct inputs into resource allocation decisions. The measures reviewed are ones that have been widely used in the literature or that have been recommended for use for evaluative purposes. For the flourishing account, we review the Psychological Well-Being Scale (Ryff and Keys, 1995) and the Orientation to Happiness Scale (Peterson et al., 2005) as these are widely referred to in the academic literature. For the hedonic account, we have focused on the Positive And Negative Affect Scale (Watson et al., 1988) as this has also been widely used, The Affectometer 2 (Kammann and Flett, 1983), as this is being used in Scotland and the Day Reconstruction Method (Kahneman et al., 2004a) as this is being put forward as a method that can provide data for a national account of well-being (Kahneman and Krueger, 2006). Appendix A details each measure. The three evaluative measures we review are the Satisfaction with Life Scale (Diener et al., 1985), the Personal Well-being Index, which combines evaluations of a number of domains into an overall value (International Well-Being Group, 2005) and single item life satisfaction questions. Many surveys use such a measure although there is variance in the exact wording and response scales used and, since this may affect the results generated, we review the different wording of this question from many different studies. In addition, we review three measures that have been widely used and that cut across flourishing, hedonic and evaluative accounts; namely, the Center for Epidemiological Studies-Depression Scale (CES-D, Radloff, 1977), Quality of Life Scale (CASP-19, Hyde et al., 2003), and the General Health Questionnaire (Goldberg, 1978).

3.1 Preference satisfaction accounts 3.1.1 Income On the face of it, income would appear to be relatively straightforward to measure but there are, in fact, many questions about precisely what type of income we wish to measure. Most reports are of gross income but net income more accurately represents an individual’s choice set, and so we would ideally wish to adjust for income tax, national insurance, pensions contributions, council tax rebates, housing benefits etc. There are also important issues surrounding the use of household or individual 19

Review of measures of personal well-being income, and how best to allocate household income to the household members. An equal distribution of resources is usually assumed but there may be power imbalances within the household (Haddad and Kanbur, 1990). The needs of a household grow with each additional member but not proportionally i.e. the needs for housing space, electricity, etc. will be less than three times as high for a household with three members as compared to one. Economies of scale will vary according to the type of good consumed e.g. there are more potential economies of scale in heating than in food. Equivalence scales assign each type of household a value in proportion to its needs. A range of equivalence scales exist, which take into account the size of the household and the age of its members, but which also can produce quite different results (Atkinson et al., 1995). Moreover, if we assume that the benefits of additional members are at least as great as the costs incurred then nonequivalised income should give a closer approximation to total utility. If individuals are able to save (often in middle age) and dis-save (when young and old) in order to smooth their consumption over time, then annualised permanent income may provide a better measure of an individual’s ability to satisfy preferences at any point in time. Consumption measures could be used as a proxy for permanent income since our consumption patterns may more accurately reflect the income we will earn over our lifetime than our income in any one year will. Again, it is problematic to collect complete consumption data although total consumption can be estimated from details of consumption of some key items, such as food, housing costs and durable goods. Time frame Reference standards Sensitivity

Reliability Cardinality

Interpersonal comparability

Most reports are of annual income but annualised lifetime income, which incorporates wealth (savings, investment etc.) more accurately reflects ability to meet preferences. Each individual’s or household’s income is treated separately and so there are no reference standards applied. Income is potentially one of the most sensitive measures since it is possible to show very small changes. In practice, because of a general reluctance about reporting income, many studies will gather income data in broad ranges. There are problems with the reliability of self-reports, particularly from the self-employed and informal sector (Moore et al., 2000). If income as a measure of well-being is cardinal, then an increase in income of £20,000 should lead to an increase in well-being which is 10 times as great as an increase of £2,000. However, there are likely to be diminishing marginal returns to income i.e. as income increases, each additional pound brings less additional benefit to the individual. If we can adjust income by the (diminishing) marginal utility of income, then the adjusted measure of income may be cardinal. Most studies imply that the log of income would be approximately cardinal. This is the adjustment for distributional weights that is recommended in the Treasury Green Book. If income (or the log of income) can be treated cardinally, and if current income is the only factor that determines desire fulfilment, then it is possible to make interpersonal comparisons. 20

Review of measures of personal well-being Practicality Policy relevance

Notwithstanding problems associated with self-reports, income data has been routinely gathered in the UK for many decades, so it is certainly a practical measure. We are used to measuring income and using it to represent government performance, so it has a great deal of policy relevance.

3.1.2 Quality-adjusted life years To allow for comparisons across conditions, health states have been described using descriptive systems that describe health according to a number of dimensions, such as physical functioning, pain and mental health. Although commonly referred to as health states, there is little to distinguish some of these descriptive systems from measures of well-being. The two descriptive systems that allow the valuation for each health state to be expressed in a single index number between 0 (for death) and 1 (for full health) and that have been widely used in the UK are the EQ-5D and the SF-6D (Brazier et al., 2004). Health state valuation methods require a respondent to state the probability mix of full health and death that makes them indifferent between that gamble and the certainty of an intermediate health state – the standard gamble (SG) method – or else requires them to state the length of time in full health that they consider to be equivalent to a longer period of time in poor health – the time trade-off (TTO) method (Dolan, 2000). SG and TTO values for the EQ-5D and the SF-6D are currently being used to calculate the number of QALYs generated by new drugs and other treatments. The National Institute for Health and Clinical Excellence are now using QALYs to inform recommendations about whether new technologies should be reimbursed by the NHS. It is common for most QALY studies to ask patients to describe their own health and then to use values for those states that have been elicited from a representative sample of the general population. In this sense, the overall QALY figures are a combination of self-report (the description of health) and external assessment (the valuation of health). Because the public often fail to appreciate what life would really be like in health states they have little experience of and because preferences generally fail to appreciate future changes in preferences and well-being, this approach to allocating resources has been subject to increasing criticism (Dolan and Kahneman, 2006). Time frame

Reference standards

The EQ-5D asks respondents to describe their health today, whilst the SF-6D asks about the past four weeks. Valuations for health states can be elicited for any duration, although most studies have asked respondents to imagine being in the health states for between ten years and the rest of their lives (Dolan, 2000). Most studies do not explicitly ask respondents to make comparisons with other people but, when asked to imagine a health state different to his own, a respondent’s valuation may be related to how different that state is to his current state or to states that he has direct personal experience of (Dolan, 1999). Respondents may compare themselves to others of their age. 21

Review of measures of personal well-being Sensitivity

Reliability

Cardinality

Interpersonal comparability Practicality

Policy relevance

The EQ-5D and the SF-6D have both been found to distinguish between different patient groups in most of the studies that they have been used in to date (Brazier et al., 2004). SG and TTO valuations have also been shown to cover the full range of values between full health and dead but floor effects have been found for the EQ-5D and ceiling effects for the SF-6D (Brazier et al., 2004). The EQ-5D and SF-6D have shown good reliability over time (Brazier and Deverill, 1999). The SG and TTO have split-test correlation coefficients of around 0.80 and test-retest reliability correlation coefficients also around 0.80 (Dolan, 2000). SG valuations lie on a cardinal scale if individuals behave in such a way that they treat probabilities in a linear fashion (i.e. the change from a 1% to a 2% risk is the same as the change from 99% to 100%). However, there is evidence that people systematically overweight small probabilities and underweight large ones, and the degree to which they do this varies from one context to another, so it difficult to know how best to weight probabilities in order to achieve cardinality (Camerer, 1993). For a response to a TTO question to provide a cardinal value, it is necessary that there is no discounting of future utilities i.e. that each moment in time is given equal weight. It has proved difficult to estimate discount rates for health (Cairns, 1992), and so it is difficult to say what discount rate to adjust TTO values by. If valuations fail to satisfy cardinality, it is difficult to see how they will be interpersonally comparable. The EQ-5D and SF-6D have been administered on many different patient groups and it is relatively quick and cheap to collect data. It would appear that both the SG and the TTO are practical in that they have both been widely used in practice and most studies have reported high response rates and high levels of complete data (Dolan, 2000). NICE is currently using QALYs to inform its recommendations about whether the NHS should reimburse particular drugs and technologies, so the approach has some degree of policy relevance.

3.2 Flourishing accounts 3.2.1 Psychological well-being scale Ryff's Psychological Well-Being Scale (PWBS; Ryff, 1989) is based on earlier notions of a hierarchy of human needs (Maslow, 1954) and the scale was designed to manifest these ideas. The full scale consists of 120 items, 20 items on each of sixsubscales: self-acceptance; positive relations with others; autonomy; environmental mastery; purpose in life; personal growth. The overall score is supposed to represent a person's Psychological Well-Being. A shortened version including 3 items for each scale has been included in the Midlife in the US study (MIDUS). A lot of subsequent

22

Review of measures of personal well-being work has attempted to examine the degree to which these concepts are distinct from SWB but the overlaps seem to be considerable (e.g. Keyes et al., 2002). Time frame Reference standards Sensitivity

Reliability Cardinality Interpersonal comparability Practicality

Policy relevance

None specified None specified This should be high given the 6 sub-scales and items. For data from the MIDUS study, the means ranged from 15.9 to 17.8, with standard deviations from 4.1 to 3.2 (Keyes et al., 2002) but the full distributions were not reported (see also Ryff, 1989). Keyes et al. (2002) present relatively low levels of internal reliability, with Cronbach's alpha from 0.37 to 0.59. No evidence No evidence Not currently collected in the UK but the reduced scales would be relatively easy to collect and something like them is being collected in the next wave of the European Social Survey. The measure takes about five minutes or less to complete. The "lower" hedonic pleasures are principally associated with the poorly educated and the "higher" eudemonic pleasures are principally associated with the better educated. Thus, the policy relevance of the scale appears to depend, at least in part, on the acceptability of these underlying values for policy (see also Nettle, 2005).

3.2.2 Orientation to happiness The Orientations to Happiness Scale (Peterson et al., 2005) is an attempt to develop measures of Seligman's (2002) belief that the "good life" is made up of three related yet distinct aspects: the pleasurable life, the meaningful life, and the engaging life. The pleasurable life refers to SWB. The meaningful life refers to the notion that there is more to life than hedonism and is related to Ryff's (1989) ideas on psychological well-being. The engaging life is an attempt to build on Csikszentmihalyi's (1990) notion that beneficial experiences are actually achieved when people are in a state of "flow" - i.e. are engaged in complex tasks that stretch their abilities and provide challenges but not insurmountable obstacles. To our knowledge, the 18 item scale (6 items for each dimension) has only so far been used as part of the Authentic Happiness on-line surveys completed by several hundred people. Time frame Reference standards Sensitivity

None specified None specified This should be high given that there are six items for each scale. Means for the three scales ranged from 3.05 to 3.42 (SD 0.72 to 0.88) and the response distribution was more normal than for the satisfaction with life scale, or SWLS (Peterson et al., 2005).

23

Review of measures of personal well-being Reliability

Cardinality Interpersonal comparability Practicality Policy relevance

Internal reliability of all three scales was high (Chronbach’s α for pleasure=0.82, engagement=0.72 and meaning=0.82). However, all items are positive leaving the scales open to the possibility that a response bias towards saying "yes" or "no" will create artificially high inter-item reliability. No test-retest data has been reported. No evidence No evidence Data are relatively easy to collect over the internet although not currently collected in the UK. The measure takes about 10 minutes to complete. The main justification for three scales is that they add independent explanatory power when regressed onto the SWLS. However, it is not clear why we need this additional scale if the SWLS already incorporates all of these aspects of well-being.

3.3. Hedonic accounts 3.3.1 Positive and negative affect scale The Positive and Negative Affect Scale (PANAS) was developed by Watson et al., (1988) and was in many ways an extension of Bradburn's (1969) Affect Balance Scale. The basic idea is to get people to assess the degree to which they have experienced various emotional states – described by a list of adjectives – over a certain period of time e.g. people are asked to say how often they have felt cheerful, calm and nervous. One of the innovations of the PANAS was to examine different time frames, e.g. yesterday, last week, last month, last year and so on, with systematic differences being found. However, in large scale surveys usually only one time is used, such as the last two weeks or last month. Time frame Reference standards Sensitivity

Reliability

Cardinality Interpersonal comparability

In the MIDUS study, the time frame is the last 30 days. None specified Positive scale "moment" mean = 29.7 (SD = 7.9) and negative scale mean = 14.8 (SD = 5.4) (Watson et al., 1988). However, since the scales run from 10-50 (for 10 item scales), it seems that the positive scale has a more normal distribution and thus is more sensitive than the negative one which appears to be showing a floor effect. Watson et al. (1988) report high internal (alpha) reliability (0.86 to 0.90 for the negative scale and 0.84 to 0.87 for the positive scale). Test-rest reliability for negative scale = 0.48 and for positive scale = 0.58. No evidence Lepper (1998) reports self-other (spouse) correlations of 0.43 to 0.45 for positive affect, 0.33 to 0.43 for negative affect and 0.47 to 0.53 for affect balance.

24

Review of measures of personal well-being Practicality Policy relevance

Not currently used in the UK but takes only about five minutes to complete. This depends on whether policy should be concerned with hedonic outcomes.

3.3.2 Affectometer 2 Developed in New Zealand by Kammann and Flett (1983), the Affectometer 2 was originally developed with the aim of monitoring changes in well-being over time and the latest version consists of 40 items which the authors claim are related to ten "mnemonic qualities of happiness": Confluence, social interest, optimism, freedom, self esteem, energy, self efficacy, cheerfulness, social support and thought clarity. The selection of these categories does not, however, appear to be based on clear theoretical grounds. It has been included in a longitudinal study of middle-aged women in Melbourne, Australia (e.g. Dennerstein et al., 2002). Time frame Reference standards Sensitivity

Reliability

Cardinality Interpersonal comparability Practicality

Policy relevance

The ‘past two weeks’ Temporal comparisons about the past and predicted future in a number of items. The degree of sensitivity is hard to assess though the number of items suggests it should be high. Dennerstein, Lehert and Guthrie's (2002) longitudinal study does not show effects of hormonal changes on the measure suggesting it is either not sensitive enough to pick up important physiological changes or that these changes are not important for psychological well-being generally. Kammann and Flett (1983) report a high alpha for the entire scale of α = .95, suggesting good internal reliability. Moreover, Dennerstein, Lehert and Guthrie's (2002) longitudinal study reports strong correlations of r > .60 across the three time points, suggesting high reliability over time. No evidence No evidence The measure was included in the Scottish Health Education Population Survey (Mallam, Angle, Wimbush & Fraser, 2004) and the responses are currently being analysed by Ruth Tennant and Sarah Stewart-Brown at Warwick Medical School in terms of reliability and validity. Given that the general thrust is much the same as several shorter measures its practicality in terms of the length of time needed to fill it in is low. Since it focuses on both positive and negative mental states it may have some policy relevance.

3.3.3 Day reconstruction method The Day Reconstruction Method (DRM, Kahneman et al., 2004a,b) was developed to counter the practical problems of Experience Sampling Methods (ESM). ESM

25

Review of measures of personal well-being requires people to complete something like the PANAS 'On-line' by providing them with a palm top and "buzzing" them at random times throughout the day when they are supposed to record what they are doing and how they are feeling at that moment. Clearly, this is very labour intensive so the DRM simply asks people to write a diary of the main events or episodes of the day before (i.e. yesterday) and evaluate these according to similar criteria (I was feeling depressed, happy etc.). It has so far been used by Kahneman and colleagues with over 2000 women in the US and in France. Time frame Reference standards Sensitivity

Reliability Cardinality Interpersonal comparability Practicality Policy relevance

Yesterday None specified As it currently stands, people report roughly 12-14 episodes a day and each episode is evaluated in terms of 8 dimensions on a six point scale so potentially sensitivity is extremely high. Different dimensions and emotions could be used. The articles published to date do not report distributions of responses across the scale. No evidence The focus on time means there is an implied cardinality if time is linearly weighted. No evidence Not easy to put the full questionnaire in a standard survey as it takes up to two hours to complete. The measure focuses on activities and so policy relevance may depend on the nature of the activity. Arguments over breakfast might not be of concern to policy makers, but stress while commuting might be.

3.4 Evaluative accounts 3.4.1 Satisfaction with life scale The Satisfaction with Life Scale (SWLS) is a five-item scale developed by Diener et al. (1985) to assess global cognitive evaluations of life, rather than affective reactions, in the general population. A review of the measure by Pavot & Diener (1993) reports widespread use and good psychometric properties (see below). The SWLS has been used in a large International Student Survey, MIDUS and in the Authentic Happiness on-line surveys since 2003. The scale is really analogous to the single item life satisfaction measures discussed below but provides greater reliability for analysis than the single-item measures. The scale is also often used by researchers trying to validate their own scales such as Ryff's Psychological Well-Being scales suggesting it is widely considered to be an important measure of cognitive evaluations of well-being.

26

Review of measures of personal well-being Time frame Reference standards Sensitivity

Reliability

Cardinality Interpersonal comparability Practicality Policy relevance

None specified The measure explicitly asks for counterfactual comparisons to the “Ideal” and to living life over again, and a temporal reference to “so far”. There is also a version that asks respondents to consider themselves in the past, present and future (Pavot et al., 1998). Diener et al. (1985) report a mean of 23.5 (SD = 6.4), which suggests a skew towards the positive end and possible ceiling effects. Pavot and Diener (1993) examine sensitivity by reviewing studies that monitor changes in SWLS as a factor of life changes. They find predictable changes following both negative events, such as illness of spouse, and positive events, such as receiving therapy. All studies report good internal reliability, with Chronbach's α around 0.90 (Diener et al., 1985; Lepper, 1998; Peterson et al., 2005). Test re-test reliability is around 0.75 (Diener et al., 1985; Lepper, 1998). No evidence Lepper (1998) reports self and spouse correlations of around 0.50. Data is easy to collect and the SWLS takes about two minutes to complete, but most of the current data comes from the US. To the extent that our comparisons with our ideal selves are relevant for policy, the SWLS has policy relevance.

3.4.2 Personal well-being index The Personal Wellbeing Index was created from the Comprehensive Quality of Life Scale (ComQol, Cummins, 1993). The aim was that the various domains of life should be amenable to both objective and subjective measurement. The 11-point (010) choice was preferred as this maximises discriminative capacity and is simple to understand. Time frame Reference standards Sensitivity

Reliability Cardinality Interpersonal comparability Practicality

Present None specified There is the usual skew with more people being at the upper end of the scale. The International Wellbeing Group (2005) report on Australian data suggesting the normal range for individuals is 50100 points. However, 40-90 is the normal range for Chinese respondents. Cronbach alphas of 0.70-0.85 have been reported (International Well-being Group, 2005). Inter-domain correlations are moderate at round .30 to .55 and item-total correlations are at least .50. No evidence No evidence No UK data but only takes a few minutes to complete.

27

Review of measures of personal well-being Policy relevance

Is likely to appeal to policy makers because it clearly focuses on several specific domains and thus general changes in well-being can be attributed to changes in the specific domains through regression analyses. For instance, policy makers may be interested in a fall in general well-being due to poorer "health" or "safety" but not due to a lack of "personal achievement".

3.4.3 Life satisfaction Life satisfaction is said to be a more cognitive approach to measuring SWB than the affect measures such as PANAS. The measures we review here are all single item questions. Many of the single item life satisfaction questions ask about 'happiness' (e.g. General Social Survey) instead of life satisfaction. Since the responses to this alternative still require people to make a global assessment and since our review of the literature found very few instances where results appeared to deviate as a function of this wording difference, we group the two types of item together for present purposes. However, we do draw attention to those cases where the questions appear to offer different conclusions about the causal factors of well-being in Section 4. Other satisfaction questions are often used relating to specific domains (e.g. marital / job) but since we are only concerned with global well-being we do not review the papers that focus exclusively on these factors as the dependent variable. What we can say is that the relations between satisfaction on the domains and global satisfaction are generally positive (e.g. van Praag, Frijters & Ferrer-i-Carbonell, 2003). Time frame Reference standards Sensitivity

Reliability

Cardinality Interpersonal comparability Practicality Policy relevance

Either no time frame is usually specified, or the last year, or sometimes "at present". Reference standards are rarely used but Appendix A does contain some exceptions. Cummins & Gullone (2000) argue that an 11-point (0-10) EndDefined Response Scale is better than 7 point Likert scales or the 3-4 adjectival descriptor scales used by many surveys because adjectival descriptors are not separated by equal psychometric intervals. 0-10 optimizes respondent discriminative capacity and is simple to understand, but most respondents cluster around 6-8. Using panel data from the GSOEP spanning 17 years, Fujita & Deiner (2005) found average life satisfaction in the first five and last five years to be correlated r = .51, a very high level of test rerest reliability for such a time gap. Recent research suggests that assuming cardinality or ordinality of the answers to life satisfaction questions is relatively unimportant for the results (Ferrer-i-Carbonell and Fritjers, 2004). No evidence In general very practical and some of the measures are already collected in the UK. Only take a few seconds to complete. Life satisfaction is likely to include many things which are of policy relevance (e.g. health) but also some aspects of life which may not be of policy concern (e.g. hobbies). 28

Review of measures of personal well-being

3.5 Combined accounts 3.5.1 Centre for Epidemiological Studies Depression Scale (CES-D) The CES-D was designed by Radloff (1977) as a measure of depressive symptoms in the general population. It developed from a pool of items from previously validated depression scales. It measures current levels of depression focusing mainly on the affective component and includes positive as well as negative items (e.g. "I felt "happy"). Although it seems to be related to "negative" psychology, the positive items mean that it does consider both sides of the coin to some extent. The measure has been used in the National Survey of Households and Families (NSHF) in the US. Time frame

During the past week.

Reference standards Sensitivity

Statement 4 is a social comparison, but this item was not included in the NSHF. The scale ranges from 0-60 and several studies report use of the full range (Beekman et al., 1997; Parikh et al, 1988). Internal consistency coefficient in the original study was around 0.85 in a healthy sample and in patients (Radloff, 1977), and similar result have been found more recently (Hann et al., 1999). Test re-test reliability in this latter study was around 0.50. The scale is really only used for diagnosis, and was not designed to be cardinal. No evidence

Reliability

Cardinality Interpersonal comparability Practicality Policy relevance

The full version takes only a couple of minutes to complete. To reduce misery, depression and negative emotions is seen by some to be a more legitimate goal of policy than improving positive emotions (Kahneman, personal communication).

3.5.2 CASP-19 The CASP-19 (Hyde et al., 2003) was a measure originally developed to investigate the quality of life of older people in the UK, from a needs satisfaction perspective and the initials stand for the four dimensions of the scale: Competence, Autonomy, Selfrealisation and Pleasure. The measure was included in wave 11 of the BHPS. Time frame Reference standards Sensitivity

In general The domains of family, money and health are used to focus attention on limitations. A good distribution across the range although a negative skew (with nearly 20% scoring 55 or above) means there seems to be a ceiling effect. Factor analysis on the scale (Hyde et al., 2003) produced a single latent factor so their claim for four sub-domains is questionable.

29

Review of measures of personal well-being Reliability Cardinality Interpersonal comparability Practicality Policy relevance

Hyde et al. (2003) report moderate to high sub-domain internal reliability (α = 0.59 to 0.77). No evidence No evidence Inclusion in the BHPS makes data easy to collect in a UK context and takes less than five minutes to complete. It is primarily a measure designed to be used with older people and it is unclear whether it would prove useful among a general population sample e.g. item 1 (my age prevents me from doing the things I would like to) is likely to result in very different interpretations amongst the young and old.

3.5.3 General Health Questionnaire The General Health Questionnaire (GHQ) was developed to investigate psychological (rather than physical) health (e.g. Goldberg & Williams, 1988). The GHQ12 questions can also be thought of as a measure of an individual’s evaluation of their life, some of the questions tapping into evaluation (e.g. thinking of oneself as worthless). Simultaneously, many questions relate to current feelings, and cover a spectrum of positive and negative affect so could be thought of as proxying for positive and negative affect hence picking up well-being from a hedonic perspective. It originally contained 60-items but various shorter versions have been developed and now the 12 item scale is commonly used. Like the CES-D scale it asks about the frequency of recent symptoms. It has been used to assess levels of depression, anxiety, sleep disturbance and happiness in the general population. It has been used in many UK studies including all waves of the BHPS, the Health Survey for England and the National Child Development Survey, and the English Longitudinal Study of Aging. The GHQ-12 includes six positive and six negative states and a choice of four options for each in which the presence or intensity of the state over the last few weeks is related to its usual frequency or intensity, thereby creating a 36 point ‘Likert’ scale. Alternatively, the number of questions to which the individual responds in the worse two categories may be aggregated giving a 12 point ‘caseness’ score. The term ‘caseness’ refers to the use of cut off points to divide the population into cases and normals, where caseness expresses the probability that the respondent might be found to have a psychiatric illness, within each study the cut off point is chosen such that it best discriminates between two groups, the most common cut off point is a GHQ caseness score of 4 or more. Interpreting the GHQ responses is particularly problematic since the questions compare the last few weeks to ‘usual’, however, what individuals are taking as their reference point is not clear. A low score on the Likert (0-36) scale may represent lower well-being than a slightly higher score, which may be suggestive of greater stability. Consequently, the assumption of ordinality in the Likert scale is problematic. Time frame Reference standards

The ‘past few weeks’ Uses "Compared to usual" and explicitly mentions "health", which may focus attention on this domain when answering the questions. 30

Review of measures of personal well-being Sensitivity Reliability

Cardinality Interpersonal comparability Practicality Policy relevance

Shown to be sensitive to picking up depression (Goldberg and Williams, 1988). Satisfactory internal consistency (αs >.60) has been demonstrated by both split-half and Cronbach’s alpha analyses (Goldberg and Williams, 1988). Test-retest reliability is problematic due to the fact that the dysfunctions captured by the GHQ12 are supposed, by definition, to be transitory. Was designed to detect cases not be a cardinal measure. No evidence Has been measured in the UK, easy to collect further data and only takes only a couple of minutes to complete. The alleviation of mental health problems may be seen as more legitimate goal than ‘happiness maximisation’.

3.6 Summary of review of measures There are potentially important differences between the various measures of wellbeing which should not be overlooked when considering which measures to use in a policy setting, nor when interpreting evidence in relation to the factors associated with those measures. Our review has considered the following differences: •

Time frame – Many of the measures provide few temporal or comparative frames of reference. Instead, they ask respondents about their lives "in general". An exception is the time frame in the DRM, which asks specifically about yesterday.



Reference standards – The measure with the most explicit reference standards is the SWLS, which includes one temporal and two counterfactual comparisons (e.g. "close to my ideal"). In their review of context effects, Schwarz and Strack (1999) concluded that explicit reference standards can seriously affect the responses to such questions, thus making comparisons across different questions problematic. However, respondents will often use some standards of reference when answering questions (Parducci, 1984) and, if no explicit standards are offered, we have very little insight into those that people actually do use. Thus, it could be argued that at least some stated standards provide researchers with a clearer picture of what people base their responses on. Of course, focusing people onto a particular time, person, etc. may focus them on elements of their well-being that do not weigh so heavily in their overall experiences of life (Kahneman et al., 2004a).



Sensitivity – Many of the measures suffer from ceiling effects i.e. for most measures there is a negative skew with people bunching at the upper end of the scales. A possible limitation of life satisfaction questions is their relative stability over time, leaving policy makers with the dilemma of whether or not any real progress has been made. Despite this general stability, recent analysis using the GSOEP panel data reports systematic and lasting changes in reported life satisfaction as a function of life events such as widowhood and unemployment (e.g. Fujita and Diener, 2005). This suggests these items are sensitive to important external circumstances.

31

Review of measures of personal well-being



Reliability – The measures all show high to moderate internal and test-retest reliability, with the exception of the short-form of the Psychological Well-being Scale.



Cardinality – None of the measures fully address the issue of cardinality but it is encouraging that, in some contexts at least, the results do not differ much if the data are treated as ordinal or cardinal (Ferrer-i-Carbonell and Fritjers, 2004). There is more that needs to be done with the scales here



Interpersonal comparability – We cannot yet confidently claim that two identical responses to a measure by different individuals mean the same thing.



Practicality – In terms of the feasibility of using the measures in the UK, we noted above that all of the measures are relatively short (1 to 26 items). With the exception of the DRM, all can be completed in less than ten minutes. Many of them, especially the single items measures, would take considerably less time hence their use in large scale surveys. Additionally, some data are already available in the UK. This includes the CASP-19 and the GHQ, which have been included in the BHPS at some time. The single item measures used in the National Child Development Survey also offers uniquely British data and many of the international surveys such as the Eurobarometer, World Values Survey and the European Social Survey also have some data points from the UK.



Policy relevance – The issue of whether or not a measure has policy relevance is a difficult one and may change over time. One account of well-being, or a measure within it, could be considered most suitable at one point in time but could change subtly or even quite dramatically due to moral debate, political discourse, or changes in public preferences. It is unclear how and why some measures come to represent the status quo at a particular point in time and it is certainly not within the scope of this report to consider such matters in any detail. It is worth making two brief points that are relevant here. First, the measurement of personal wellbeing might itself increase its policy relevance. The greater use of subjective measures might be associated Psychological Heisenberg Principle (Diener and Seligman, 2004). The original Heisenberg Principle argues that the act of observing a phenomenon often results in changes to the phenomenon itself. Diener and Seligman argue that the systematic monitoring of personal well-being will have positive knock on effects because "what a society measures will in turn influence the things that it seeks". Second, and notwithstanding the dynamic nature of public preferences, we need to find out much more about what the public consider to be the legitimate focus of government policy in relation to well-being.

From the first stage of review, we conclude that the measures of preference satisfaction currently available (e.g. income) offer an incomplete picture of individual well-being. It is difficult to rule out any of the other measures from a theoretical basis, although the evidence suggests that measures of flourishing accounts of well-being may not be sufficiently reliable for use in a UK policy context.

32

Review of the factors affecting well-being

4. REVIEW OF THE FACTORS AFFECTING WELL-BEING 4.1 Review strategy This review aims to summarise the current knowledge on the economic, social and environmental factors which relate to changes in well-being. The literature on wellbeing is vast and we need to focus on the literature that provides the best evidence on the causes of well-being. Statistical data analysis techniques, data availability, computer processing power and theoretical understanding have progressed since the early work on the causes of well-being, and these developments have enabled researchers to begin to distinguish between real and spurious correlation and, to some extent at least, between correlation and causation. Our focus is on research conducted on large datasets where there is an opportunity to isolate the impact of one factor upon well-being through controlling for other factors. There were four strands to the search strategy. The first was to identify all published papers in economics journals published since 1990 via ‘Econlit’ using the search terms: ‘subjective well-being’, ‘psychological well-being’, ‘happiness’ and ‘life satisfaction’. The abstracts were read and assessed according the following criteria: the work reported original data analysis, the focus of the work was on understanding individual well-being, and the work has been peer reviewed. Whilst satisfaction with an aspect of life (job, health etc.) is important to understanding well-being, it is not a full measure of individual well-being, and are therefore not the focus of this work. The second strand of the literature search was to identify all published papers in psychology (identified via ‘Psychinfo’) that have studied well-being through the use of large, nationally representative datasets, using the same search terms as above plus specified datasets that have been identified via the search on economics journals. Third, we identify all reviews of the psychology literature on the causes and correlates of well-being published since 2000. Finally, we identify significant grey literature, which contains original data analysis, from the key economists working on subjective well-being. This search strategy identified a total of 153 papers. The papers were reviewed according to a template where consideration is given to the dataset used, the specific measure of well-being used, the statistical and econometric analysis conducted, and the findings relating to the correlates with, and causes of, well-being. As with bringing together a range of evidence, there is the danger of oversimplifying things. Therefore, in Appendix B, we provide summaries of all the studies that have considered a characteristic and its relationship to well-being and in Appendix C we detail each study that we have reviewed according to the template. Some studies provide more robust evidence than others and so we have categorised the evidence according to its quality. At the least robust and persuasive end of the spectrum lies correlation studies that do not control for other variables. The observed relationship between education and well-being, say, might then have more to do with the relationship between income (which is highly correlated with education) and wellbeing than with the relationship between education and well-being. Therefore, more robust evidence comes from those studies that control for as many factors as possible.

33

Review of the factors affecting well-being

A spurious correlation between two factors may still arise if they are both correlated with a third omitted factor (and often difficult to observe), such as personality. Therefore, more persuasive evidence on the relationship between education and wellbeing would be provided by studies that not only control for income etc. but that also control for any heterogeneity associated with individuals’ propensity to systematically report their well-being or other variables in particular ways (e.g. optimists may always score their more highly than pessimists). We therefore give greater weight to the limited but increasing number of studies that control for unobserved individual effects (this is usually done by fixed effects models using panel data which allow us to compare an individual to themselves at different points in time). Of course, even here, a correlation between education and well-being does not tell us the degree to which education causes well-being or vice versa. The best clues to causality are provided by panel data which allows us to follow the dynamics of, and changes in, events and well-being. Therefore, we give greater weight to findings which also have supporting studies to show the direction of causality.

4.2 Factors associated with personal well-being We have considered all the potential influences on well-being that have been identified in the literature. These fall under seven broad headings: 1) income3; 2) personal characteristics – who we are, our genetic makeup; 3) socially developed characteristics – our health and education; 4) how we spend our time - the work we do, and activities we engage in; 5) attitudes and beliefs towards self/others/life – how we interpret the world; 6) relationships – the way we connect with others; and 7) the wider economic, social and political environment – the place we live. Of course there is also a degree of overlap between some of the characteristics under different headings. For instance, while we place attendance at church under the “how we spend our time” heading, it is clearly related to beliefs about God which we examine under the “attitudes and belief” heading. Moreover, many of the characteristics may interact with one another and so we highlight any important interaction effects where the evidence is available. We also consider some of the problems associated with interpreting the evidence that is currently available. Throughout, we draw particular attention to different correlations associated with different measures of well-being. Emphasis is given to evidence from the UK and, where possible, we also consider some of the implications for policy that may follow from the evidence. 4.2.1 Income This is a very complex area where much research has been undertaken. Therefore, we begin by summarising the evidence on income and some of the policy implications that might follow from it before discussing the evidence in more detail. 3

Income can, of course, be used as a proxy for a measure according to the preference satisfaction account but the focus in what follows is on its relationship with well-being as defined according to the flourishing, hedonic and evaluative accounts.

34

Review of the factors affecting well-being

The majority of individual level studies find a positive association between absolute income and well-being (usually measured in terms of life satisfaction). The results generally suggest that there are diminishing returns to income i.e. the well-being hit of income falls as income rises. Some of this positive association is likely to be due to reverse causation, as indicated by studies which show higher well-being leading to higher future incomes, and some is likely to be due to unobserved individual characteristics, such as personality factors, as indicated by studies which find a reduced income effect after controlling for individual effects. Studies that have included relative income (defined in a range of different ways with a range of different reference groups) suggest well-being is strongly affected by relativities. This suggests that additional income may not increase well-being if those in the relevant comparison group also gain a similar increase in income. However, increases in income that result in increases in tax yield, which could be used to fund public services which may themselves enhance well-being. For a given income level, having high aspirations and expectations have a negative effect on SWB. Aspirations themselves appear to be driven in part by past incomes, implying adaptation to higher levels of income. The importance of aspirations reinforces findings that the perceptions of financial status have stronger predictive power than actual income. These findings imply that additional income for those who are not at low levels of income is unlikely to increase SWB in the long run if the additional income serves to increase expectations of necessary income. The policy implications of the relative income effect will depend in large part on the specific comparisons that individuals make. For example, if most income comparisons are upward looking towards to the wealthiest in the area or the reference group, and if the difference between personal income and comparison income is what matters to individuals, then higher taxes on the wealthiest may enhance overall well-being. On the other hand, if relative income effects operate through a desire to mimic consumption habits of those around us (as suggested by Duesenberry, 1949), then indirect taxation may be needed in order to capture the externalities of consumption (Layard, 2005). Therefore, more research is needed to understand the precise nature of the comparisons individuals make in relation to other people’s income. Similarly, more research is needed on the precise mechanisms through which expectations are formed before we could draw any firm policy recommendations from the evidence. 4.2.1.1 Absolute income It is important to be clear about what is meant by ‘income’, and how it is to be measured. When we ask whether income increases well-being, what we really want to know is what happens when an individual has more access to resources. However, access to resources is likely to be difficult to determine for a number of reasons. First, there are inaccuracies in measuring income, since it is notoriously difficult to get accurate income measures, particularly for some groups such as the self-employed. As a result, some surveys (such as the GSS in the US) only gather income categories. Second, in order to compare incomes across countries, it is necessary to use an exchange rate. Most studies do use purchasing power parity rates, which should equal the ratio of the two countries' price level of a fixed basket of goods and services, and which therefore attempts to give a rate of exchange based on what can be done with

35

Review of the factors affecting well-being the money in the different countries. However, this does not fully address all the problems of cross national comparisons of income. Third, net income (after taxation and subsidies) will get closer to an individual’s access to resources, but many datasets contain only gross income. The relationship between gross and net income may vary according to other variables, such as employment status or age, creating the possibility of systematic biases. Fourth, in situations where we know individual and household income, we still lack information on how income is distributed within the household. We will need to make some assumption about the resources an individual has command over, which could be based on a simple per capita basis, or an equivalence scale which accounts for the age of all members of the household. In addition to accounting for the need created by an additional member of a household, equivalence scales also aim to account for increasing economies of scale of household production. Fifth, an individual’s current access to resources will also depend upon accumulated wealth, and access to capital (whether they can borrow easily). For example, while at college an individual is likely to be spending more than their income and running up debts, and in retirement they are also likely to spend beyond their income and run down their savings. Consequently, there are likely to be systematic differences between current income and access to resources and the life stage. Finally, in addition to the resources owned by the individual, access to resources is also related to the publicly funded resources to which they have access. This would include free or subsidised education, health care, transport, parks etc. There may also be benefits arising from living in a high income country in addition to publicly funded facilities, such as privately funded charity. These problems should make us cautious about drawing simple conclusions about the relationship between income and well-being and, in particular, how access to resources affects well-being. Interpreting the results is made more complicated by the fact that different studies have addressed the issue of income differently. For example, some have looked at whether richer countries are happier, others at whether countries get happier as income rises, others at whether richer people are happier, and yet others at whether individuals get happier as they get richer. The answer to one of these questions cannot be used as evidence about another question. Indeed, the evidence on income and well-being has famously led to seemingly contradictory conclusions e.g. no change in well-being over time as income has risen but a positive relationship (in cross-sectional data) between income and well-being. This has become known as the Easterlin Paradox (after Richard Easterlin who was one of the first people to identify this paradox in 1974). It is therefore essential to consider each of the above perspectives, and search for convincing explanations for any apparent inconsistencies. Diener and Biswas-Diener’s in-depth review in 2002 found that most cross-national studies had a correlation between GDP per capita and measures of well-being (usually life satisfaction) of between 0.5 and 0.7, suggesting a strong and consistent correlation between national income and well-being, with higher income countries experiencing higher levels of SWB. Of course, bivariate correlations do not account for the many other ways in which countries may differ. Countries vary across a wide range of

36

Review of the factors affecting well-being factors e.g. the extent of human rights, democracy, political voice, quality of government, education and literacy, income inequality, access to public goods, culture etc. As income is likely to be related to many of these factors, it is possible that the income-well-being correlation is spurious. Helliwell (2003) and Inglehart & Klingemann (2000) find that the correlation with national income and life satisfaction drops considerably if quality of government and democracy is controlled for. However, if income is the cause of the higher levels of these other attributes (such as political freedom, education, sewage etc.) then it would be reasonable to consider the effect of income without controlling for these factors. If, however, they are due to a factor not related to income, such as technological progress, then the simple correlation will be inappropriate. Studies which give some national level controls find that national income has a significant positive (e.g. Inglehart & Klingemann, 2000) or non-significant (Helliwell, 2004) effect on average life satisfaction. The extent of the correlation depends on which countries are considered. A broader international sample, including low, middle and high income countries will find a stronger correlation than a mainly high income sample. Diener and Seligman (2004) find a correlation between GDP/capita and life satisfaction of only 0.08 for richer countries in the World Values Survey with a GDP/capita of over $10,000. Correlations between individual life satisfaction and national income tend to be positive, even controlling for own income. Using European data, both Di Tella et al. (2003) and Fayey & Smyth (2004) find that GDP/capita increases life satisfaction. Di Tella et al. (2003) estimate that an increase in GDP of $1000 (1985$) raises the proportion of those reporting they are very satisfied from 27.3% to 30.9%. However, Fayey & Smyth (2004) also find that the coefficients on the third and fourth GDP quartiles are fairly similar suggesting a plateau effect of GDP. Controlling for the level of income, the annual growth of national income has also been found to increase life satisfaction (e.g. Haller & Hadler, 2006), although this findings is not always robust (e.g. Di Tella et al., 2003). In the longer term, growth has not been found to have a strong relationship with income as the data suggest that average satisfaction levels within a given country tend to be highly stable over time, even during periods of significant economic growth, with much of the evidence drawing on the experiences of the US, UK and Japan (Easterlin, 2005a). As with cross country comparisons, many other variables change in addition to national income during the time periods concerned, including political factors such as the rights of women, national and international stability, crime levels and environmental factors. Since these factors may themselves be related to changes in income levels, it is not clear what is causing (or masking) any correlation between average income and average life satisfaction. Furthermore, aggregate data may disguise different trends within sub-groups. For example, Blanchflower & Oswald (2004) find a negative trend for life satisfaction in the US (1972-1998) for whites and women, but a rising trend for blacks and men despite few overall effects at the aggregate level. There seems fairly good evidence that within most countries at a point in time, when other factors are not controlled for, individual income and life satisfaction are

37

Review of the factors affecting well-being positively related (e.g. Easterlin, 2001). The relationship is stronger amongst low income groups (Biswas-Diener & Diener, 2001), and weakens as income levels rise (Cummins, 2004). Of course, other factors may be related to individual income, and so is necessary to explore the income-SWB relationship using multiple regression analysis. Most studies find a significant relationship between individual or household income and well-being. Many studies find a significant positive effect of income when entered in logs (which serves to reduce the gap between ever increasing amounts of income), suggesting a curvilinear effect which would imply diminishing marginal returns to income. However, most studies do not give details of testing for the appropriate form in which to enter income and the log transformation may have been chosen because of the lack of normality of the income distribution, as opposed to good evidence that the log transformation is the best representation of the impact of income on life satisfaction due to diminishing returns to income. Other studies explore the effect of income using income categories. However, their ability to explore the impact of income at the top end of the distribution is often limited by widening income bands and an open top band. Studies that have entered income as categories find income to be positive but give slightly mixed results on the shape of the relationship. Many find a consistently increasing relationship although some studies find middle incomes to have higher well-being than the highest incomes. A non-linear relationship between life satisfaction and income is also supported by Helliwell (2003). There are some notable exceptions to the generally positive effect of individual and household income on well-being. Using the BHPS, Clark and Oswald (1994) find no relationship between GHQ and personal income and Clark (2003b) finds these variables to be negatively related. Clark (2003b) suggests that this non-significant or negative relationship may be due to the costs of gaining the additional income, in terms of additional work hours and effort, and, perhaps supporting this, finds household income to have a positive effect. However, some studies fail to find a significant effect of even household income (e.g. Wildman and Jones, 2002). When exploring a pure income effect in models with many control variables we should not overlook the potential for indirect effects of income, such as the effect on housing tenure, health, education, safety of the environment, security, savings etc. Some of the benefits of an individual's income may even be experienced by other members of the family. The results at the national level suggest that income is more important at certain life stages. Gerlach & Stephan (1996) find that log of household income per capita increases life satisfaction for those under 49 but not 50s and over. Cummins et al. (2004) find that the youngest and oldest groups are less influenced by income than the middle aged (26-55) but Marks & Flemming (1999) find no significant interaction between age and income. Findings on the impact of income across the genders are mixed. For example, Blanchflower & Oswald (2004) find for the US that there is no significant difference between the effect on men and women but for the UK they find that being in the 2nd and 3rd income quartiles is more beneficial to women than men.

38

Review of the factors affecting well-being Studies in which a direct comparison can be made between different measures of well-being tend to find that life satisfaction is slightly more strongly related to income than overall happiness (Leigh & Woolfers, 2005; Lelkes, 2006), and that both are more strongly related to income than the GHQ (Clark, 2003b). The DRM study (Kahneman et al., 2004a) finds a significantly larger correlation between log household income and life satisfaction (r = .20) than overall affective experience throughout the day (r = .05; difference p < 0.001). This suggests that income is less important for day-to-day happiness than it is for one's evaluation of life. Further support for this conclusion comes from comparing the mean levels of 'enjoyment' (on a scale from 0-6) across activities engaged in at home for those with relatively low ($30,000) versus high incomes ($90,000): the mean difference in enjoyment scores between these two groups was very small. Whatever measure is used, there is still the issue of causality, and there is some evidence that higher well-being leads to higher income. For example, using Russian data Graham et al. (2004) find that controlling for equivalent household income in 1995 unexplained (residual) life satisfaction from 1995 has a positive effect on log equivalent income in 2000. A one point increase in unexplained life satisfaction in 1995 yields about 3% increase in income in 2000. Marks & Fleming (1999) find using Australian data that income change was predicted by an index of domain satisfactions. For respondents who were two standard deviations higher in life satisfaction, there was an 8 to 12 percent greater income increase at the next time period compared with the lower group (although this may be due to the fact that they knew they were about to earn more at the next time point and thus expected income was the cause of more life satisfaction). Diener et al. (2002) find that higher cheerfulness in first year at college correlated with higher income 19 years later. At a national level, Kenny (1999) argues the economic growth of nations is more rapid in happier countries. Overall, there appears to be reasonably robust evidence that individual or household income has a positive but non-linear effect on life satisfaction. Even studies that control for individual fixed effects, comparing an individual to themselves at different points of time (e.g. Frijters et al., 2004), show a consistent positive effect of income, although the size of this effect is often reduced. Unfortunately, since most studies are based on ordered logit or probit models, which express the likelihood of a given wellbeing based on the set of background characteristics, it is often difficult to compare the effect sizes across studies. 4.2.1.2 Relative income Income measures can also communicate the relative position of an individual in a social hierarchy. If studies are cross section then a positive coefficient on income cannot distinguish between the benefit of absolute income and the benefit arising from the status or rank position provided by an individual’s income. Controlling for fixed effects cannot fully address this, since those times when an individual’s income is above their average income may also be the times when they are higher up the social order. Some evidence for a ranking effect can be gained by panel data studies which control for real incomes, and find an additional effect of a ranking variable. For example, Alesina et al. (2004) use panel data to show that being in the top two income

39

Review of the factors affecting well-being quintiles increases happiness and life satisfaction in the US and in Europe even when controlling for household income. It is possible to test the relative income hypothesis by considering the impact of changes to the individual’s rank position within their relevant peer group. Holding an individual’s own income constant, an increase in reference group income will lower their rank position within that reference group. This effect can be explored by considering the sign and significance of reference income in a typical well-being function. Findings from studies using this type of model point to a relative income effect. For example, Luttmer (2004) finds using US data that the log of predicted average earnings in the individual's local area is negatively related to overall happiness. Those who socialise with their neighbours less than once a month, and hence would be less exposed to comparison incomes, are less affected by reference (neighbourhood) income. Luttmer finds that the coefficient on the log of household income is slightly bigger than that on the log of relative earnings, suggesting that some of the benefit to increases in own income would remain even if reference incomes and own income increased by the same percentage. Ferrer-i-Carbonnell (2005) finds using German data that the coefficient on log of reference income (this time defined by those of the same age, region and education level) is negative and similar to that for log of family income. Clark (2003b) using the BHPS found that reference group income (based on sex, region and year) has a significantly negative influence on life satisfaction for a group of full-time employed workers, although reference income is not significantly related to the GHQ. Graham & Felton (2006) find no significant effect for reference wealth in Latin America when reference wealth is taken as national averages, but find a significant negative effect once reference wealth is taken by city size and nationality is controlled for, suggesting relative income effects operate at a more localised level than nationally. However, for the US both McBride (2001) and Blanchflower & Oswald (2004) fail to find any significant effect of relative income. Senik (2004) finds contrasting results in Russian data. Here, average reference income (based on education, occupation, industry, work experience, gender and region), increases life satisfaction when controlling for household and individual income levels. The relative income effect has also been explored by considering the difference between an individual’s income and the income of a reference group. Dorn et al. (2005) adopt this structure and find support for a relative income hypothesis but they cannot rule out an additional absolute income effect. Ferrer-i-Carbonell (2005) also presents the German data in this structure, finding a positive effect of the difference between log of own income and log of reference income, significantly for West Germany but not East Germany. She concludes that increases in family income accompanied by identical increases in the income of the reference group do not lead to significant changes in well-being. An alternate structure is to model the ratio of household income to average reference income. If the ratio of own income to reference income increases, then the individual must be moving up the ranking, and vice versa. Hudson (2006) adopted this structure for an analysis of European data and found that the ratio of household income to average country income has a positive effect on life satisfaction. Using US data,

40

Review of the factors affecting well-being Blanchflower and Oswald (2004b) found that the ratio of individual income to state income per capita had a significant and positive effect on overall happiness even when they controlled for household income, household income per capita and regional house price index (to control for state level costs of living). This would suggest that declining reference group income increases overall happiness. They also find that the ratio of own income to the 5th income quintile has the largest effect, suggesting that people make upward comparisons. As noted by Easterlin (1995; 2001), if the relative income effect dominates the absolute income effect this would explain why cross section data show that wealthier individuals within a society are happier, but that average life satisfaction levels remain stagnant as all members become wealthier. However, positive correlations between average well-being and national income found in international cross section data, particularly in lower income country samples, requires either that comparisons of relative position are made across nations, or that an absolute income effect operates in many countries. Evidence on international relative income is limited. Fayey & Smyth (2004) argue that the significance of GDP quartile, holding income constant suggests that relatives between countries matters. Graham & Felton (2006) analyse Latin American responses to an economic ladder question in which people are asked to place themselves on a ladder where one stands for the poorest level of society and ten the richest and found that average country wealth increases responses suggesting individuals compare themselves at least in part to an international society. There is some evidence that reference income matters more to certain groups. Luttmer (2004) found that the effect of log of average earnings was strongest for those seeing more of their neighbours, those aged 30 to 60 and for married or divorced respondents, but similar between men and women, for those with and without children, and those renting or home owners. McBride (2001) suggests reference income matters less at lower incomes where absolute income is more important. However, this would seem to conflict with the findings of Ferrer-i-Carbonell (2005), which suggest reference income matters more to those below average incomes. Since the latter results are derived from a model which controls for the unobserved individual effects, they should be given greater weight. Crucial to interpreting the findings on the relationship between income and well-being is the assumption that use of the well-being scale itself is independent of income. If people respond to questions about happiness relative to a subjective happiness norm, which is income-dependent, then this undermines our ability to identify the relationship between happiness and income. For example, an 8/10 on a happiness scale may mean something different for a rich person to a poor person; the poor person may think that given their income levels they are satisfied with the way their lives are going. Evidence from fixed effects models, do not rely on the equivalent use of the wellbeing scale between different people, but they do rely on equivalent use of the scales for each individual over time, when incomes are changing. The extent to which people with changing incomes change their use of the scales is an area requiring further research. As it stands, there is good evidence that relative income effects are relatively large and significant, and these findings have potentially important implications for policy. Although the evidence cannot rule out an effect of absolute income, it does

41

Review of the factors affecting well-being suggest that the impact of absolute income is smaller than that of an individual's comparison income. Furthermore, there is some evidence that the nature of this comparison is upward looking. 4.2.1.3 Wealth Headey et al. (2004) also explore the effect of wealth on well-being in an attempt to overcome some of the limitations of using income as a measure of access to resources. They find that the log of net worth has a greater effect on life satisfaction than equivalised household income in Australia, Germany and Britain. A positive effect of log of wealth has also been found by Graham & Pettinato (2001) for life satisfaction in Latin America and when assets are entered linearly in Latin America (Graham & Felton, 1996) and the US (Martin & Westerhof, 2003). Brown et al. (2005) also find that, controlling for income (household or individual), annual savings are positively related to inverse-GHQ in Britain. These studies may suggest a well-being benefit from the presence of assets and savings or from the ability to save, which may reflect lower expenditure needs and consequently the ability of income to meet needs. 4.2.1.4 Debt Borooah (2005) found that needing to borrow money mid week increases the chances of being unhappy. Cummins et al. (2004) found that for all income groups those who cannot afford to pay off their credit card have a lower PWI score. Brown et al. (2005) controls for a wide range of variables and find that the presence of credit card debt reduces inverse-GHQ. However, it is still difficult to determine a clear relationship between debts and life satisfaction. Large debts may indicate a very good credit rating, and are potentially investments rather than a deficit between income and consumption. Secure debts, such as a mortgage, or debts for investments have not been found to impact negatively on life satisfaction (Brown et al., 2005, Cummins et al., 2004). 4.2.1.5 Expectations and perceptions Reference income may not simply be a matter of the actual rank one holds but the perception of one’s position in the social hierarchy, and how closely one’s perceived rank matches one’s expectations of social position i.e. whether one’s income matches one’s expected income. Expected income is likely to be determined in part by past income levels. Di Tella et al (2006) find that average income over the last five years is not significantly related to life satisfaction, (except for those with right wing political views), which is supportive of the idea that we adapt to changes in income. However, Graham et al (2004) finds income five years earlier increases life satisfaction. This suggests that the relationship between past income and current SWB is indeterminate: it may reduce SWB if it leads to unfulfilled expectations, but may also proxy for assets and previous wealth accumulation, which can increase SWB. One of the difficulties with exploring adaptation effects is that little is known of the time frame over which adaptation operates or how income expectations are formed. Incomes as far back as childhood may be influential. McBride (2001) finds that, controlling for current income, having a standard of living which is much worse than one’s parents at the same age has a negative effect on overall happiness. Parental

42

Review of the factors affecting well-being income is likely to feed into an individual’s expectations of their own income and, controlling for current income, those with higher expectations are likely to judge their current income less favourably which may consequently lower well-being. It is important to note that the evidence of the negative effect of expectations is from models in which income is controlled for. This therefore should not be interpreted as evidence that high expectations in themselves are bad, since they may be part of a dynamic process by which those with higher expectations have the motivation to gain higher incomes. There is some evidence that aspirations increase as income rises. For example, Stutzer (2004) estimates for Switzerland, that aspiration income increases by 4.2% following a 10% increase in income. Aspirations are also raised by having more rich people live in one’s community. Stutzer (2004) finds that the proportion of rich people in the community increases aspirations of those who have contact with neighbours more than others, confirming that the effect of the proportion of rich is unlikely to be due to cost differences. These shifts in aspirations may have negative effects if expectations rise above those which can be met by current income. For example, Graham & Pettinato (2002) identify a group of ‘frustrated achievers’, those who are unhappy despite a rapid growth in income because of rising aspirations. Similarly, Dockery (2004) finds that the log of hourly wage is negatively related to life satisfaction once satisfaction for pay is controlled for, which he explains in terms of expectations about wage increases. Once a subjective measure of financial assessment is included, income measures can become non-significant (Haller & Hadler, 2006 for international data, Johnson & Krueger, 2006 for the US, Wildman & Jones, 2002 on the GHQ in Britain). Furthermore, perceptions of change have also been shown to have an impact. Brown et al. (2005) find for the UK that perceptions of future deterioration and especially perceptions of deterioration from the previous year reduce psychological well-being (GHQ). An individual’s perception of their financial status can incorporate actual income, expectations of income based on past levels and the income or expenditure of those with whom the individual compares themselves to. Not surprisingly then, perceptions of financial status have stronger predictive power than actual income. Income, therefore, appears to matter for well-being, but in ways which are far more complex that simply looking at the annual income of the household. In addition to current income, consideration should be given to the role of savings, assets, debts and financial commitments and the ease with which they can be serviced, past levels of income and expenditure commitments and habits. There is also the need to consider more fully the effects of income rank and the income of those whom an individual compares themselves to, as well as expectations about future income and general attitudes towards income and material consumption. 4.2.2 Personal characteristics (who we are, our genetic makeup) Our personal characteristics, such as our age, gender and ethnicity have the potential to be important for our well-being, but such variables have often only been added as controls or to look at the interacting effect of other variables, like marriage for

43

Review of the factors affecting well-being different genders and age groups. Thus much of the evidence presented in this section is based directly on the results presented, rather than discussed, in the papers.. 4.2.2.1 Age The balance of evidence suggests a negative relationship between age and well-being across a range of single item happiness and life satisfaction questions. However, more sophisticated analyses find a positive relationship between age squared and life satisfaction (e.g. Blanchflower & Oswald, 2004; Ferrer-i-Carbonell & Gowdy, 2005), suggesting a U-shaped curve with higher levels of well-being at the younger and older age points and the lowest life satisfaction occurring in middle age, between about 3550 years, depending on the study. Clark (2003a) finds a similar pattern for the GHQ using the BHPS, showing a minimum around 36 years. The relationship is more complicated for negative measures (e.g. CES-D scores and negative affect). While there is some evidence that negative affect may increase, especially in very old age (over 70s, e.g. Greenfield & Marks, 2004), there is also evidence that age is associated with a decrease in depressive symptoms (e.g. Magdol, 2002). Thus, it seems that while older people may be experiencing more negative dayto-day emotions, these do not lead to more depressive symptoms. There is also evidence that unemployment (e.g. Clark & Oswald, 1994), lower income and a lack of a close relationship (Cummins et al., 2004) are harder during middle age. The main problem with the age variable is that most studies deal with different cohorts and it may be differences in cohorts rather than age per se that is being picked up. Longitudinal studies such as the GSOEP are able to tackle this to some extent (and report similar results) though these will need to be carried out over an even longer time span before we can control more fully for cohort effects. 4.2.2.2 Gender The balance of evidence for gender suggests that women tend to report higher scores for positive measures (e.g. happiness, Alesina et al., 2004) and negative measures (e.g. CES-D scores, Kim & McKenry, 2002; inverse GHQ scores in the BHPS, Clark & Oswald, 1994) of well-being. It is unclear whether this greater range is due to greater variance in actual emotional experiences or greater willingness to report emotional diversity. What we can say is that there also appears to be a large degree of variance. For instance, Baker and colleagues (Baker et al., 2004) found a relatively strong effect of gender for CES-D scores (unstandardised coefficient = .29) using the Americans’ Changing Lives data. However, the standard error was 0.18, suggesting that while women in general may express greater negative emotion, there is often no consistent pattern simply as a function of gender. The cross-gender differences are generally higher for the more negative measures such as CES-D though (e.g. Brown, 2000). A few studies report no gender differences (e.g. Louis & Zhao, 2002) even using the same datasets (e.g. GSS). This suggests that other correlates may also be more important than gender per se given that different studies have different control variables. Indeed, when specific subsets are examined, such as those who cannot work

44

Review of the factors affecting well-being due to health problems (Oswald & Powdthavee, 2005) or those who provide informal care for others (van den Berg & Ferrer-i-Carbonell, 2005), the gender effect often disappears. Nevertheless, there is also evidence suggesting that men may suffer more from part-time employment, unemployment and living alone than women (Bardasi & Francesconi, 2004; Cummins et al., 2004). There is also some suggestion that depressive symptoms may be higher among less educated women (Brown, 2002). If the aim of policy were to alleviate misery, then there may be some justification for targeting women as they tend to exhibit higher levels of depression. On the other hand, proponents of a positive psychology approach (e.g. Seligman, 2002) may advocate policies that aid men's experience and expression of positive emotions. Given that we can detect a difference between the measures, the next step might be closer inspection of whether there is a difference between the causes of personal wellbeing across genders for different measures. 4.2.2.3 Ethnicity In the US, the balance of evidence suggests that whites have higher well-being than African Americans on positive (e.g. happiness) and negative (e.g. CES-D) measures (e.g. Lee & Bulanda, 2005; Magdol, 2002; Thoits & Hewitt, 2001). In their analysis of the ACL data, Thoits & Hewitt (2001) find that controlling for prior levels of happiness and life satisfaction, race is the strongest predictor of current happiness and life satisfaction, stronger than other demographics such as age and gender as well as circumstances such as employment and marital status. There is also some suggestion that ethnicity may interact with age, since older respondents tend to show less differences as a function of ethnicity (Baker et al., 2005; Greenfield & Marks, 2004). Comparing whites to the category "Other" may not be particularly helpful since it is hard to interpret null effects such as those reported by Theodossiou (1998) using the BHPS. One reason is that some ethnicities, in particular Hispanics (Luttmer, 2004), tend to show higher levels of well-being than whites and thus the outcome of the comparison may depend upon the proportion of different ethnic groups within this "Other" category. The Theodossiou study was one of only a handful of European studies which included ethnicity so it remains unclear whether the effects found in the US generalise to Europe and to the UK in particular. Future analysis of UK datasets could pay more attention to the issue of ethnicity in order to address this issue. 4.2.2.4 Personality A considerable amount of psychological research has considered the relationship between personality and well-being (for a review see, DeNeve & Cooper, 1998). Much of the current understanding of the importance of personality for personal wellbeing is based upon studies using the Minnesota Twin Registry. These have shown that monozygotic twins (reared apart or separately), have much higher correlations of SWB than dizygotic twins (Lykken & Tellegen 1996), suggesting that SWB has a strongly genetic component. However, few studies have examined this relationship using large scale surveys of the kind included in our review. Using the WVS data, Helliwell (2006) found a very moderate relationship between personality and SWB once other factors such as social trust and religious beliefs were

45

Review of the factors affecting well-being controlled for. Similarly, Lee et al. (2001) find that depressive symptoms (CES-D) were not related to the degree of sociability, and a measure of extraversion (which is normally found to correlate positively with measures of life satisfaction). Self-esteem has been found to be negatively associated with depressive symptoms (CES-D scores; unstandardized coefficient = -.46, p < 0.001, Kim & McKenry, 2002). People higher in self-esteem seem less likely to suffer from depression. In addition, many of the sub-scales of the GHQ, which could also be interpreted as personality variables (e.g. self-worth), correlate positively with life satisfaction using the UK, BHPS data (Ferrer-i.Carbonell & Gowdy, 2005). Moreover, psychological problems during adolescence predict negative well-being (CES-D) using the UK NCDS data, but not life satisfaction (Flouri, 2004). Whilst there is some evidence that personality is related to well-being, it is not possible from the existing evidence to tell whether this effect is innate or whether it is due to attitudes that are developed in childhood and adulthood. The role for policy interventions would clearly be greater in the latter than in the former case. 4.2.2.5 Physical characteristics Only two studies in our review have examined physical characteristics. Using a simple cut off point of a Body Mass Index (BMI) of 30, Gerdtham and Johannesson (2001) found no difference in life satisfaction as a function of being above or below this point in a Swedish sample. In Australia, Cummins and colleagues (Cummins et al., 2004) found that their personal well-being index (PWI) scores were highest for people in the most common height range (160-169 cm) and lower for extremes at either end. They also found that PWI drops after a weight of 100kg and drops sharply at 120kg with scores dropping at BMI scores of 30. However, shortness and high BMIs were both associated with lower income in the Cummins et al. (2004) study, and thus further research needs to consider the relations of these variables controlling for income and other socio-economic factors. Moreover, given that physical characteristics are often thought to be a primary factor in a range of disorders associated with negative well-being (e.g. anorexia and bulimia), it is surprising that more research using large scale data sets has not been carried out to examine the relations between physical characteristics and well-being.

4.2.3 Socially developed characteristics (human and physical capital) 4.2.3.1 Education Some studies find a positive relationship between each additional level of education (e.g. Blanchflower & Oswald, 2004) and well-being, while others find that middle level education is related to the highest life satisfaction (e.g. Stutzer, 2004). However, there is some evidence that education has more of a positive impact on low income countries (Fahey and Smyth, 2004; Ferrer-i-Carbonell, 2005). It has also been shown that mother’s education can have a positive effect on an individual’s happiness, in which case the assessment of the impact of education from a society perspective may

46

Review of the factors affecting well-being need to include the impact that education has on the well-being of an individual’s family over time. There is some indication that measures of flourishing reveal a stronger relationship with education e.g. Ryff’s PWB measure was found to have a higher correlation with education than a combined SWB scale (Keyes et al., 2002). Studies exploring the relationship between education and mental health have found mixed results. Education is usually found to reduce depression in the US as measured by the CES-D (e.g. Baker et al., 2005), although some studies find no significant difference (e.g. Lee et al., 2001). In the UK, Flouri (2004) finds no significant relationship with the GHQ, and some studies find that education is associated with worse GHQ scores (e.g. Clark, 2003a). Education qualifications may be related to unobservable traits at the individual level, such as motivation, intelligence or family background and so ideally we should look to those studies which control for unobserved heterogeneity. However, fixed effects models can only pick up the effect of individuals completing their education or returning to education at a later date and most adult survey respondents are unlikely to change their education level during their time in a panel survey, and consequently fixed effects models are unlikely to find any significant effect for education (e.g. Meier & Stutzer, 2006). Furthermore, the well-being impact of education for those that do return to gain further education as adults is unlikely to be the same as that derived from those who initially remained in the education system for longer. In addition, the coefficient on education is often responsive to the inclusion of other variables within the model. Education is likely to be positively correlated with income and health, and if these are not controlled for we would expect the education coefficient to be more strongly positive. For example, the positive effect of education on overall happiness found by Blanchflower and Oswald (2004a) could be picking up a health effect since this is not controlled for. However, the inclusion of variables correlated with education as controls raises a further problem – if the correlation is due in part to a causal path from education to, say higher income, then fully controlling for income will underestimate the full contribution which education is making to well-being. To the extent that education has caused greater income and health, we would ideally wish to include this fact in the effect of education. The indirect effect of education on life satisfaction via health is explored by Bukenya et al (2003) on US data and Gerdtham and Johannesson (2001) on Swedish data. They both find that the positive coefficient on high school and attending college increases by about one third from the standard model, which suggests that this indirect effect is considerable. However, the indirect effect of education on income or health, may be caused by education establishing the relative position or social status of the individual. This relative position may result in higher well-being. If this is the case, then any boost in well-being from education would not be a consequence of education per se but of the sorting mechanism (or relative education) i.e. those with more education may gain a higher income not because they are more productive but because they have been sorted by the education system to be those at the top of the social order. Graham & Pettinato (2001) find that years of education increases overall happiness in Latin

47

Review of the factors affecting well-being America, but that the effect becomes non significant once social mobility and relative economic standing is included, which indicates that the benefits to education may be positional rather than absolute. Overall, then, the evidence on the relationship between education and well-being is very mixed and is one area where the choice of well-being measure makes a difference. In general, education and “flourishing” are positively associated with one another, and education and life satisfaction are less so. However, much depends on the assumptions that are made about how education is related to the other things (health, income etc.) that are also related to well-being. 4.2.3.2 Health Studies consistently show a strong relationship between well-being and both physical and psychological health, regardless of how well-being is measured. Mental health measures seem to be more responsive to health status than life satisfaction measures (Headey & Wooden, 2004). There is also some evidence that physical function is more strongly related to life satisfaction than it is to happiness (Michalos et al., 2000). Psychological health appears to be more highly correlated with subjective well-being than physical health but there is some circularity in showing that psychological health explains the variation in well-being when well-being is measured using a psychological diagnostic tool such as the GHQ or the CES-D. Some of the association may be caused by the impact that well-being has on health but the effect sizes of the health variables are substantial suggesting that even accounting for the impact of well-being on health, health is still impacting on wellbeing. Furthermore, specific conditions, such as heart attacks and strokes reduce wellbeing (Shields & Price, 2005), and the causality here is most likely to be from the health condition to well-being. Of course, a third factor (such as personality) may be related to both well-being and health, and this would make finding a significant relationship between health and well-being more likely. Studies using fixed effects models continue to show a strong effect of health on well-being but they are still unable to control for time variant unobservable variables, such as current mood. Oswald & Powdthavee (2005) present some evidence that individuals adapt to disability status, finding that the length of time an individual has experienced the disability reduces the negative impact of the disability. However, adaptation is far from complete. The fixed effects model, finds that disability reduces life satisfaction (on a 1 to 7 scale) by 0.596 points for those with no past disability, by 0.521 points after 1 year of disability, 0.447 points after two years and 0.372 after three years. An interpretation of adaptation requires that the scale is being consistently used throughout the time period, and is independent of health status. 4.2.3.3 Type of work There is insufficient evidence to draw clear conclusions about the impact of type of work on well-being. Given the amount of time people spend at work, this is an area that requires more investigation.

48

Review of the factors affecting well-being Some evidence from the UK suggests that casual work is detrimental to men’s mental health and women’s life satisfaction (Bardasi & Fracesconi, 2004), and that belonging to a union is beneficial to life satisfaction (Blanchflower & Oswald, 1998). There is a little more evidence on self-employment. Many European studies fail to find any significant difference between being employed and being self employed but Blanchflower & Oswald (1998) find a robust positive effect of self-employment using UK, International (ISSP) and US (GSS) data. Using US and European data, Alesina et al. (2004) find that the positive effect of self-employment is limited to the rich. Due to the lack of clarity surrounding the effect of self-employment, there is insufficient evidence to make any judgement about how the different measures of well-being interact with self-employment. 4.2.3.4 Unemployment Studies consistently show a large negative effect of individual unemployment on wellbeing, across a range of national and international data sets. Models which treat life satisfaction scales as a continuous variable, tend to find that the unemployed have around 5 to 15 % lower scores than the employed (e.g. Di Tella et al., 2001). Using ordered probit and logit models on European data, Lelkes (2005) found that unemployment reduces the probability of a high life satisfaction score (at least 8/10) by 19%, and a high overall happiness score by 15%. Lucas et al. (2004) use the GSOEP to show that people who are later unemployed do not start out with low life satisfaction, and when in the reaction phase (a year before, the period of unemployment and a year after) they experience more than half a point lower life satisfaction on a 0 to10 scale. This suggests any selection effects are minimal. Furthermore, controlling for individual heterogeneity using fixed effects models, again finds a strongly robust impact of unemployment (e.g. Ferrer-iCarbonell & Gowdy, 2005), although the effect size is often slightly reduced. Using the BHPS Oswald & Powdthavee (2005) find a reduction of 0.496 points on a 1 to 7 scale, falling to 0.334 once fixed effects are controlled for. Men have been found to suffer most from unemployment and some studies also find that the middle aged suffer more than the young or old. Those with higher education suffer more in Britain (Clark & Oswald, 1994), those with right wing political leanings in the US (Alesina et al., 2004) and those in high income countries (Fayey & Smyth, 2004). In the UK, Shields & Price (2005) find that the impact of unemployment on the GHQ is related to the extent of employment deprivation in the area, with the individual unemployment effect being neutralised in areas with employment deprivation of over 22%. Using the BHPS, Clark (2003a) finds that the negative effect of unemployment would be neutralised at a rate of 24%. The same study also finds that, for those working, having an unemployed partner is detrimental to well-being, but for the unemployed it is beneficial. There is some mixed evidence of adaptation to unemployment. Using the BHPS, Clark & Oswald (1994) find that the negative coefficient on unemployment reduces with the length of unemployment but, using the GSOEP, Lucas et al. (2004) find that individuals who are unemployed for more than a year experience a more negative reaction to unemployment, and previous unemployment experience does not reduce the harm of current unemployment. They also find that once unemployment has

49

Review of the factors affecting well-being ended, individuals do not return to their pre-unemployment levels of life satisfaction. Similarly Louis & Zhao (2002) find that any period of unemployment over the last 10 years has a negative impact on a combined general happiness scale. Using a fixed effects model, Wildman & Jones (2002) find that the negative unemployment coefficient for men, falls from 1.979 points (on a 0 to 36 GHQ Likert scale), to 0.989 once satisfaction with finances and expectations of future financial position is controlled for. Studies which have controlled for income and found a negative effect of unemployment have interpreted this as a non-financial loss. However, Wildman & Jones’s (2002) findings suggest that current income may not be the most suitable measure of financial position, and that some of the damage of unemployment arises due to increased concerns over future finances. It has been demonstrated that being unemployed has severe and long lasting negative consequences for life satisfaction and happiness, which cannot be explained only in terms of the loss in income. The effect can be moderated by living in household or area with other unemployed people. If personal well-being is a goal of government policy, then efforts should be directed towards alleviating some of the losses associated with periods of unemployment. 4.2.4. How we spend our time (The work and activities we engage in) Decisions related to how we use our time are very relevant for policy. Our individual decisions have implications for the provision of public services (e.g. exercise may reduce demand for NHS treatments) and for other people (e.g. the positive effects of volunteering and the negative effects of commuting in a car). Such decisions can be influenced by legislation and incentives and further consideration of the well-being effects of how we use our time may be an important part of how such policies are framed. Our review contains 32 papers which examined the relationships between well-being and time use. Most of the major headings of time use have been examined in these data sets including hours worked, commuting, housework, caring for others, sleep, exercise, and religious activities and exercise. Notably, there is no data on the relationship between the number of holidays and well-being. 4.2.4.1 Hours worked While the evidence is relatively clear that employment is better than unemployment, the relationship between the amount of work (e.g. number of hours worked) and wellbeing is less straightforward. Data from the German GSOEP suggests that life satisfaction rises as hours worked increases, controlling for individual fixed effects (Meier & Stutzer, 2006; Weinzierl, 2005). This supports evidence from the UK's NCDS, which suggests that part-time work is associated with lower life satisfaction among men than full-time work (Schoon et al., 2005). However, other studies report no differences between full-time and part-time work and life satisfaction and GHQ scores in the BHPS (Bardasi & Francesconi, 2004) or happiness in the GSS or ISSP (Blanchflower & Oswald, 2004a; 2005). Luttmer (2000) reports a negative relationship between the log of usual working hours and happiness using the NSFH data. Using the GSOEP, Meier & Stutzer (2006) find

50

Review of the factors affecting well-being an inverse U-shaped curve between life satisfaction and hours worked suggesting that well-being rises as hours worked rise but only up to a certain point before it then starts to drop as hours become excessive. Such a relationship has intuitive appeal, some work is good but too much work is bad, but given that only two of the studies we reviewed have considered these possibilities, further research is needed to confirm these findings. In particular, more research is needed to re-examine existing datasets to look not just at part-time versus full-time work but at the specific number of hours and the potential non-linear effects. There is also evidence that at least some work amongst older (typically retired) people seems to be associated with an increase in positive (e.g. happiness) and a decrease in negative (CES-D) well-being (Baker et al., 2005; Ritchey et al., 2001). Across all the studies, there is little consideration given to the type of work undertaken and this could also be an important moderating effect. The only study which examined emotions at work directly used the DRM approach (Kahneman et al., 2004) and found that positive emotions at work greatly exceeded negative ones, but that the ratio was smaller than for almost all other activities. Since working hours are a major policy discussion point it is perhaps surprising that more research examining the effects of working hours on well-being has not been carried out using the large-scale data sets we examine here, especially datasets that would be applicable in the UK context such as the BHPS. We would there recommend studies that examine the relationship between the number of hours worked with a range of other factors, such as type of work, combined number of hours worked and commuting, and the potential trade-offs between well-being at work and well-being at home. 4.2.4.2 Commuting The DRM data suggests, like work itself, that commuting is actually associated with more positive than negative emotions though the ratio is the lowest of all activities investigated (Kahneman et al., 2004). In Germany, Stutzer & Frey (2005) find lower life satisfaction with greater commuting time (using both normal and fixed effect models) and also find that this does not seem to result in greater well-being for other family members. Research to explore the consequences of different types of commuting may help minimise the loss caused by commuting. However, policies on household location will need to take into consideration a range of other factors (such as living in an urban area) to see the net effect of location on well-being. 4.2.4.3 Housework Only three of the studies we reviewed considered the effects of housework on wellbeing and none of them used UK data. Again, the DRM study (Kahneman et al., 2004) found a higher ratio of positive to negative emotions during housework although there is no discussion regarding potential variance as a function of different types of housework such as washing up versus cleaning the toilet. Using the NSFH data, Magdol (2002) finds a positive relationship between the amount of housework and depressive symptoms (CES-D scale) among women. Using a subset of the same data, Lee et al. (2001) find no relationship between amount of

51

Review of the factors affecting well-being housework and depressive symptoms among a sample of older people (aged 65yrs and over). However, these authors did find that the more this sample disliked housework the more depressive symptoms they exhibited. In other words, what seemed to matter most was not how much but the degree to which this activity was voluntary or not. This may be one area that is not considered to be relevant to policy intervention. 4.2.4.4 Caring for others The evidence from the few studies that examined the effects of amount of time engaged in informal care-giving suggests that more care is associated with worse GHQ scores (Hirst, 2003;2005), lower happiness (van den Berg & Ferrer-i-Carbonell, 2005) and more depressive symptoms (Marks et al., 2002). The effects are especially strong for close kin as opposed to non-kin (Marks et al., 2002), which may be due to more hours of kin care-giving or because of greater emotional attachment. The transition into and out of care-giving has also been explored. Not surprisingly, that transition into care-giving is associated with a range of negative well-being outcomes (Hirst, 2005; Marks et al., 2002). Using BHPS data, Hirst (2005) found that women's GHQ scores were also negatively affected by the transition out of a high load care-giving role. At this stage, it is unclear whether this is due to the loss of a defined role or because the person being cared for either had to leave the house to receive more professional care (indicating a worsening of the state), or maybe even passed away. However, the losses are such that they might be given greater prominence in debates about informal care in health and social care policy. 4.2.4.5 Community involvement and volunteering Of the eight studies that included measures of community involvement/volunteering, only two found a clear positive relationship between these activities and well-being (Helliwell, 2003; Pilcher, 2006). Specifically, both studies found a positive relationship between life satisfaction and membership in (non-church) organisations. When the personality variable of extraversion was added to the model, the effect disappeared. Although Helliwell and Putnam (2004) do report a positive relationship between average organisational membership rates and life satisfaction and happiness using the WVS data, this is not replicated in their analysis of the ESS data. In terms of volunteering, Hadler and Haller (2006) found no relationship between volunteering and happiness or life satisfaction (across 34 countries using the WVS data). Furthermore, although Thoits and Hewitt (2001) did find a positive relationship, it also seemed to be the case that happier people tended to do more voluntary work, to some extent undermining the argument that volunteering is the cause of greater wellbeing. Finally, Greenfield & Marks (2004) found that among a sub-set of older people, volunteering was associated with more positive, but not less negative affect. There was also some evidence of greater purpose in life among this sample, suggesting the need for further work examining Psychological Well-Being effects, although again this was qualified by the degree to which the person had other meaningful roles in life. Therefore, while some observers have claimed that greater community involvement is a win-win situation, providing better outcomes for the community at large and making

52

Review of the factors affecting well-being those involved feel better about themselves, the evidence we review here suggests more caution is needed. Although evidence for the UK is lacking, if the pattern here is consistent with elsewhere it seems that taking part in more community and volunteer activities is not necessarily the panacea for well-being that some have suggested it to be. Nevertheless, the potential for a win-win situation does seem to be there if we can identify exactly which community or voluntary activities are associated with the most positive effects, when in the life cycle these effects are most likely to occur, who in particular is likely to benefit most from them and so on. 4.2.4.6 Sleep Our review contains only two studies that investigated the effects of sleep and they suggest that poorer sleep leads to both poorer day-to-day emotional well-being (Kahneman et al., 2004) and lower global levels of life satisfaction (UK, BHPS data examined by Ferrer-i-Carbonelli & Gowdy, 2005). Indeed, the DRM data suggest that poor sleep may be a particular problem because it affects the quality of many of our other day-to-day activities. One problem with interpretation of this data is, of course, causality – poor sleep may lead to worse well-being outcomes and worse well-being, stress etc. may lead to sleep related problems. 4.2.4.7 Exercise There is evidence that even simple types of exercise such as gardening (Ferrer-iCarbonell & Gowdy, 2005) may be associated with higher life satisfaction and that this may be especially important for the over 60s (Baker et al., 2005). The amount of time engaged in physical activity among the over 60s was also negatively associated with depressive symptoms (Baker et al., 2005). Although a review of the broader literature on exercise and well-being has recently appeared (Biddle & Ekkekrakis, 2005), little use has been made of large data-sets and thus there seems to be an important gap in research here. Given that exercise may not only help to reduce a number of negative outcomes (e.g. weight gain and depressive symptoms), but also promote a range of positive ones (e.g. higher levels of happiness and life satisfaction) it would seem to have high policy potential. 4.2.4.8 Religious activities Compared to many of the other activities people engage in, religious activity has received quite a lot of attention, with 10 studies considering how the amount of time spent in such activities influences well-being. In this section, we simply consider the effects of time spent in these activities rather than the effect of religious beliefs, which are dealt with below. The evidence across these studies is fairly consistent and suggests that regular engagement in religious activities is positively related to happiness (e.g. Cohen, 2002; Ferris, 2000), life satisfaction (e.g. Clark & Lelkes, 2005; Hayo, 2004) and positive emotions (Kahneman et al., 2004) and negatively associated with depressive symptoms (e.g. Lee et al., 2001). While some studies only examine whether or not the person actually attends church, others examine different amounts of time spent in these activities. Using WVS data, Helliwell (2003) finds higher life satisfaction to be associated with church attendance of once or more a week. A similar finding is found in Eastern Europe (Hayo, 2004)

53

Review of the factors affecting well-being though less frequent attendance did not result in higher levels of life satisfaction than no attendance. Contrary to this latter finding, and using ESS data, Clark & Lelkes (2005) report that church attendance of at least once a month is enough to have an effect on life satisfaction. However, since attendance of once a week or more is included within 'at least once a month', the significant effect may be due to weekly attendance rather than less frequent attendance. None of the studies we reviewed considered UK data specifically. Of particular interest were findings suggesting that the effects of religious activities were relatively comparable across different religious denominations (e.g. Catholics and Protestants, e.g. Cohen, 2002). Thus it may be that going to church and the associated social interactions are more important than the particular beliefs. However, the DRM study (Kahneman et al., 2004) suggests that praying is also associated with high amounts of positive emotion. Since praying is, in many religions at least, a largely solitary exercise it seems that some of the benefits to happiness of practicing religious activities are not simply due to more social interactions. Finally, there is some evidence to suggest that religious attendance reduces the effect of income on happiness, especially for African Americans (Dehejia et al., 2005). Hence religious African Americans may to some extent be ‘insured’ against the happiness effect of drops in income. These results raise interesting issues about the appropriateness or otherwise of government support for religious activities. The consequences for those who may not attend church, and thus feel excluded, will need to be taken into account in any welfare analysis. 4.2.5 Attitudes and beliefs towards self/others/life The majority of factors examined in our review are objective in the sense that they describe people's objective circumstances (e.g. income, marital status, education level). In this section, we consider the impact on well-being of factors that are more subjective or psychological, such as attitudes towards the self, other people, society and issues of faith. Although there is an enormous psychological literature examining the relationship of various attitudes, beliefs and coping strategies with well-being, much of this has used specific clinical or convenience samples rather than large-scale or representative samples. The evidence we review relates to four subjective categories in particular, attitudes towards one's objective circumstances, trust in others, political attitudes and religious beliefs. 4.2.5.1 Attitudes towards our circumstances The evidence suggests that perceptions of our circumstances can be very important predictors of life satisfaction and other self-report measures of well-being. One domain that has been relatively extensively researched is financial satisfaction. As might be expected, poorer perceptions of one's current financial situation are usually associated with lower life satisfaction (e.g. Graham & Pettinato, 2001; Hayo & Seifert, 2003; Louis & Zhao, 2002). There is also evidence suggesting that perceptions of change in financial circumstances, as opposed to current circumstances, may also be important for well-

54

Review of the factors affecting well-being being. Using the BHPS, Brown and colleagues (Brown et al., 2005) find lower GHQ scores when people perceived their current financial situation to be worse than last year and when next year's situation is predicted to be even worse (see also Wildman & Jones, 2002). Similar findings are reported when perceptions of job security are examined (e.g. Dockery, 2003; Graham & Pettinato, 2001). Importantly, perceptions of financial circumstances appear to fully mediate the effects of objective circumstances (Johnson & Kruger, 2006) suggesting they have a more direct influence on global life satisfaction. Although such a mediating role of domain perceptions has also been found for some other domains (e.g. personal relationships), it does not appear to be universal. Martin & Westerhof (2003) find that objective health seems to have a direct effect on life satisfaction irrespective of satisfaction with health. Being ill may thus have a negative effect on well-being even if one does not fully realise it. Nevertheless, when only domain perceptions are used to predict global life satisfaction, the explanatory power of such models is generally much larger than when simply the objective circumstances are entered as predictors (van Praag et al., 2003). Of course, it must be borne in mind that both life satisfaction and domain satisfaction are subjective reports and thus share a great deal of covariance. However, the results suggest that policies which act upon objective circumstances may not have the desired effect unless attitudes perceptions, expectation and attitudes are also considered. 4.2.5.2 Trust The evidence is relatively clear from the few studies that have looked at trust and the effects are relatively large. Using WVS and ESS data, Helliwell (2003; 2006; Helliwell & Putnam, 2004) has found that social trust (trust in most other people) is associated with higher life satisfaction and happiness, and a lower probability of suicide. Moreover, trust in key public institutions such as the police, the legal system and government is also associated with higher life satisfaction (Helliwell & Putnam, 2004; Hudson, 2006), as are beliefs about the wrongness to cheat on one's taxes (Helliwell, 2003). However, there are few reliable UK data and, as with much else, establishing causality is a real problem. 4.2.5.3 Political persuasion and attitudes There was evidence that, in general, preferences for democracy and pro-market values are associated with higher life satisfaction in Latin America and Russia (Graham & Pettinato, 2001). However, it may be that these attitudes are held by people who have benefited more from these systems. There was also evidence that unemployment had worse effects on the happiness of “right wingers” in the US but that inequality was worse for the life satisfaction of “left wingers” in Europe (Alesina et al., 2004). Both findings have intuitive appeal. That inequality was less a problem for left-wingers in the US may reflect different beliefs about social mobility than in Europe. Using BHPS data, Ferrer-i-Carbonell & Gowdy (2005) consider environmental attitudes and find that people have higher life satisfaction if they care about animal extinction, but lower life satisfaction if they care about the ozone layer. Such results lend support to the idea that further suggests that the impact of external circumstances is dependent upon perceptions and attitudes.

55

Review of the factors affecting well-being

4.2.5.4 Religion Again, the evidence here supports the idea that our beliefs affect our well-being, with religious people generally being happier than non-religious people, irrespective of their faith. Taking perhaps the broadest approach, Helliwell (2003; 2006) reviews WVS data and finds that belief in a God is associated higher levels of life satisfaction and a lower probability of suicide than no belief. Other research replicates the life satisfaction finding (e.g. Clark & Lelkes, 2005) and also finds positive effects on depressive symptoms (e.g. Baker et al., 2005) and GHQ scores (using BHPS data, Flouri, 2004). However, the effects seem to be stronger in the US than in Europe (Helliwell & Putnam, 2004) and are sometimes not found at all (Smith, 2003). It seems to make relatively little difference which religion one belongs to (Christian, Judaism, Hinduism, Buddhism etc). Reviewing data in the World Database of Happiness, Rehdanz and Maddison (2003) found that the average happiness of different countries was not affected by the proportion of the population with different religious beliefs. More specifically, Ferris (2002) found no differences in happiness in the US as a function of whether respondents were Jewish, Catholic or Protestant (see also Cohen, 2002). This study did, however, suggest that there may be important differences as a function of denominations within Christianity. For instance, Evangelists and Fundamentalists report being happier than more traditional doctrines, although it is unclear whether these denominations make people happier, or attract happier people to start with. There also tend to be wide variances in well-being scores within the same religions suggesting that individual differences are important and it would be unwise to talk simply about all Catholics, all Jews etc (Haller & Hadler, 2006). For instance, within religions there are differences in the strength of people's beliefs, the degree to which they use God to help cope with difficulties and their degree of spirituality, all of which have been found to be associated with different levels of SWB (Cohen, 2002). Stronger religious beliefs may also "insure" people against a loss of income or employment (UK data, Clark & Lelkes, 2005) since religious people's well-being (especially Catholics) drops as little as half as non-religious people following these negative shocks. Nevertheless, some negative shocks may be hard to deal with in a religious context. For instance, there is evidence that divorced women in the UK gain little in terms of life satisfaction from greater religiosity (Clark & Lelkes, 2005).

4.2.6. Relationships Here we consider the relationship between well-being and personal relationships, including marriage, having children and seeing family and friends. 4.2.6.1 Marriage and intimate relationship Generally speaking, being alone appears to be worse for well-being than being part of a partnership. This emerges using both positive (happiness/life satisfaction/SWLS)

56

Review of the factors affecting well-being and negative (e.g. CES-D) measures of well-being. Regular sex was also associated with more positive well-being and since the effects were strongest when this was with the same partner, it seems that being in a caring relationship is important for wellbeing rather than simply being in a string of less close relationships (Blanchflower & Oswald, 2004a). Although there is some variation across studies, is seems that being married is associated with the highest level of well-being and being separated is associated with the lowest level of well-being, lower even than being divorced or widowed (e.g. Helliwell, 2003). There is evidence that the amount of well-being associated with being an unmarried cohabitor depends on the degree to which the relationship is perceived to be stable (Brown, 2000). As opposed to unstable unmarried partnerships, stable ones are associated with similar levels of well-being as married partnerships. Therefore, the evidence again suggests that objective circumstances do not always have direct effects on well-being and that it is important how these experiences are perceived. A number of studies have considered gender differences and most find similar effects for men and women (e.g. Frey & Stutzer, 2000). Using BHPS data, Wildman and Jones (2002) report that while men and women appear to suffer equally following widowhood, divorce and separation, single women may actually have higher well-being than married women. There is also evidence that parental divorce negatively effects well-being in adulthood (e.g. Blanchflower & Oswald, 2004a) but this was not found in all studies (Louis & Zhao, 2000). Moreover, subsequent remarriage of a parent who has suffered widowhood seems to be associated with lower levels of later well-being than subsequent remarriage of a parent due to divorce (Biblarz & Gottainer, 2000). Other longitudinal evidence suggests some selection effects with people who become divorced being less happy even before being married (e.g. Lucas, 2005). These studies also show how well-being tends to drop in the period leading up to divorce or widowhood and takes a number of years to stabilise again, and that it may never reach original baseline levels. However, as with much of the evidence reported here, there are widespread individual differences in the rate and degree of adaptation to the new state. Some people recover fairly quickly, others appear to never fully recover. On a more positive note, finding someone new is often associated with a return to something like original levels of well-being. Although we have tried to summarise the balance of evidence, the actual results are often more complicated. Using BHPS, data Flouri (2004), reports that "having a partner" (versus "not having a partner") is associated with lower GHQ scores. However, Clark (2003b) reports that there is no significant difference between being married and being single, when controlling for income. One possible explanation for these findings is that "not having a partner" can also include being separated and widowed, both of which Clark does find have a negative effect on GHQ scores. Thus, Flouri's finding may actually be showing not that "having a partner" is better than "not having a partner" but instead that "losing a partner" is worse than "having a partner". While this may be true, the picture is complicated further by the fact that when Clark (2003) controlled for employment status but not income, being married did have a significant positive effect on GHQ scores compared to being single. Speculation as to

57

Review of the factors affecting well-being why this may be the case is beyond the scope of the current report, but this discussion highlights the complicated nature of determining causes of well-being and the fragile nature of some of these effects if they are so readily influenced by the other variables in the analyses. In general, stable and secure intimate relationships are beneficial for well-being and the dissolution of relationships is damaging, at least in the short to medium term. These effects are of considerable magnitude. Therefore, an argument can be made for greater assistance in encouraging individuals to develop the skills necessary to establish beneficial long lasting relationships and minimise the harm caused by their dissolution. 4.2.6.2 Having children The evidence with regard to the well-being effects of having children is mixed and differs across measure and country. Overall, 13 studies report no effects, 14 report negative effects of having children, three report positive effects and two report mixed effects depending on the type of well-being measure. On closer inspection, it seems that children generally affect well-being more negatively for single parents (e.g. Frey & Stutzer, 2000), when the children are over 3yrs (e.g. Shields & Price, 2005), if the family has recently moved (e.g. Magdol, 2002), or if the child is sick and needs more than average care (Marks et al., 2002). In other words, if other circumstances are relatively negative, children seem to be an additional challenge to well-being. Nevertheless, there is also evidence that the well-being of women (but not men) is compromised (i.e. lower CSE-D scores) if they had wanted children but didn't end up having them (Koropeckyj-Cox, 2002). Although policy may not wish to provide incentives for or against having children, the absence of a clear positive effect of children merits further investigation. 4.2.6.3 Seeing family and friends Looking at social networks generally, there is evidence that better networks and more time spent socialising is associated with higher levels of life satisfaction, greater happiness and more positive CASP-19 scores. However, there are also studies which find no effects e.g. Baker et al. (2005), which may be because prior levels of life satisfaction and CES-D scores were controlled for. This suggests that people who tend to engage in more social activities are generally happier and thus cause and effect is not clear. As elsewhere, further research is needed using large-scale datasets to examine selection effects of social contact and to see whether personality variables may interact with the returns of social contact on well-being. One of the few studies that differentiates between contact with family and friends finds a significant positive effect on life satisfaction only for contact with family but not friends, though the effect was small (Martin & Westhof, 2003). A second study which differentiated between family and friends used the DRM and suggests that both are associated with a higher proportion of positive to negative affect compared to other activities, including caring for one's children (Kahneman et al., 2004). However, two studies also suggest there may be circumstances where greater contact with others is not indicative of better well-being. Martin and Westhof (2003) also

58

Review of the factors affecting well-being report that global satisfaction is lower when contact involves care from friends and family. A second situation where more contact with family might not be a good thing is if an adult is still living at home with their parents (Pilcher, 2006). Again, cause and effect are unclear here. Does remaining at home lower one's life satisfaction or are people with lower life satisfaction less likely to leave home? The answer to this question would seem to have potentially important policy implications in the sense that policy makers may be able to use this as an indicator of people who might need more support. For instance, it would be interesting to see what effects on well-being result from difficulties starting on the property ladder if it forces people to remain at home with their parents. It would appear that, overall, socialising with family and friends is positively associated with well-being, although the direction of causality is yet to be established. This has implications for government policies which encourage a geographically mobile labour force thereby weakening networks of family and friends. The benefits arising from close contact with family could be taken into consideration in areas such as social housing allocation. 4.2.7 Wider economic, social and political environment (Where we live) Where we live, the environment around us, the political, social and economic system of the country we live in. These are all factors that can only be changed at the individual level if we move geographical location. 4.2.7.1 Income inequality The evidence on the impact of income inequality on well-being is mixed. Looking at international data using the World Values Survey, Fayey & Smyth (2004) find that inequality reduces life satisfaction, whereas Haller & Hadler (2006) find that inequality increases life satisfaction. One explanation for these contrasting findings using international data may be that the inclusion of particular countries can be influential on the results (Bjornskov, 2003). Specifically, the relatively happy Latin American countries tend to have fairly unequal income distributions, and relatively unhappy former-Communist countries tend to have fairly equal income distributions. The results from European data are also mixed. Using Eurobarometer data, O’Connel (2004) finds that income inequality is positively related to life satisfaction, whereas Alesina et al. (2004), who use more controls and a longer time span, find that inequality reduces life satisfaction, particularly for those with left wing political leanings and the poor. For Britain, Clark (2003b) finds that income inequality in one’s reference group (based on gender, region and wave) increases life satisfaction, particularly for those under 40, those on below average incomes and those with greater increase in income over the last three years. This is the only study in which a comparison of different measures of well-being is possible, finding that the impact on GHQ is mostly insignificant. It is therefore difficult to find an interpretation compatible with this mixed evidence. It is worth noting that the impact of income inequality may be indirect, for example, via the effect it has on health. Furthermore, it is likely to vary depending on the how

59

Review of the factors affecting well-being the inequality is interpreted. Clark (2003b) and Alesina et al. (2004) explain findings for the UK and US in terms of income equality communicating messages of opportunity. What will be communicated through income inequality is likely to vary according to perceptions of mobility. Where mobility is perceived to be lower, such as Europe and Germany, inequality is found to have a negative impact. 4.2.7.2 Unemployment rates A number of studies have found that national unemployment rates are negatively related to happiness and life satisfaction. Di Tella et al. (2001) suggest that their finding of a negative relationship between life satisfaction and unemployment rates may be due to the fear of personal unemployment which arises from higher national unemployment rates. However, Alesina et al. (2004) fail to find a significant effect of the unemployment rate using the same Eurobarometer data. If the unemployment rate is positively correlated with income inequality in Europe (Cysne, 2004) then the fact that this is the only study using the Eurobarometer data to control for income inequality casts some doubt on the interpretation of this unemployment rate effect. Therefore, more research is needed to gain greater understanding on the extent of the well-being losses from a higher unemployment rate. 4.2.7.3 Inflation Investigating the impact of inflation is limited to comparisons across countries over time. Within the same country it would be impossible to isolate an inflation effect from any other time effects. Using aggregate data, both cross section (Bjornskov, 2003) and time series (Woolfers, 2003) failed to find a significant effect of inflation on life satisfaction. However, controlling for individual personal characteristics, inflation has been found to have a consistent negative effect (e.g. Alesina et al., 2004). The inflation impact is worst for those with right wing political leanings (Alesina et al., 2004). In addition, a volatile inflation rate also reduces life satisfaction (Woolfers, 2003). Many studies have a limited number of macro variables, which opens the possibility that other important variables are not adequately controlled for. For example, inflation may correlate with income inequality or lack of trust. The relative harm caused by inflation and unemployment has been estimated in some studies however, this varies from 1.6:1 (Di Tella et al., 2001), 2.9:1 (Di Tella et al., 2003) up as high as to 5:1 (Woolfers, 2003). Hence a percentage increase in unemployment is more damaging than a percentage increase in inflation and macroeconomic policy might wish to take this into account. 4.2.7.4 Welfare system and public insurance Evidence on the impact of the welfare state is limited. Veenhoven (2000) finds no correlation between welfare expenditure and average happiness or average life satisfaction. However, Di Tella et al. (2003) analyse individual level European data and find that a higher benefit replacement rate (using the OECD index of (pre-tax) replacement rates i.e. unemployment benefit entitlements divided by an estimate of the expected wage) increases life satisfaction for both the unemployed and the employed. Since the replacement rate does not automatically change in line with the business cycle, it is a preferable measure to use.

60

Review of the factors affecting well-being

4.2.7.5 Degree of democracy The Swiss federal structure gives variation in political institutions and in direct popular rights between 26 Cantons (which deal with education, welfare, and police for example). Thus, it is particularly useful for studying the effects of political institutions. Frey & Stutzer (2000) find that extended individual participation in the form of initiatives and referenda, and of decentralised (federal) government structures raises life satisfaction. This is generated not just by the outcomes of democracy but also by the political process itself. However, Dorn et al. (2005), cast some doubt on the robustness of these findings, showing that controlling for language group, the democracy index used by Frey and Stutzer (2000) is only significant in their survey data at 10%, and using a new, more representative survey, the Swiss Household Panel, it is insignificant. Using international data needs to address the high correlation between income and democracy, however, when controlling for income (Inglehart and Klingemann, 2000) and language group (Dorn et al., 2005) a positive link between democracy and life satisfaction is still found. 4.2.7.6 Climate and the natural environment Current evidence of the impact of pollution and environmental factors on well-being is very limited. Welsch (2002) notes the difficulty of isolating any effect of pollution due to the high negative correlation between income and pollution. However, he does provide evidence that suggests that pollution, as measured by nitrogen dioxide, has a detrimental impact on overall happiness (Welsch 2002; 2003). Ferrer-i-Carbonell & Gowdy (2005) find that environmental problems where one lives reduce life satisfaction but although income is controlled for in this model, this could still be picking up socio-economic status and household wealth. However, the airport noise investigated by van Praag & Baarsma (2005) is likely to be less correlated with socioeconomic factors, and shows a clear detrimental impact upon life satisfaction. Rehdanz & Maddison (2003) show that extreme weather is detrimental to happiness. 4.2.7.7 Safety and deprivation of the area Controlling for ones own income, the evidence suggests that living in an unsafe or deprived area is detrimental to life satisfaction and mental health (Ferrer-i-Carbonell, 2005; Lelkes, 2005; Shields & Price, 2005; Wiggins et al., 2004). These results suggest that further research is required to explore the effectiveness of policies aimed at addressing the negative externalities imposed upon people living in such areas. 4.2.7.8 Urbanisation There is some evidence across a range of geographical locations that living in large cities is detrimental to life satisfaction and living in rural areas is beneficial (e.g. Hudson, 2006 for Europe). However, some results are non-significant and population density was not found to effect happiness (Rehdanz & Maddison, 2003), or mental health (Shields & Price, 2005), or the Satisfaction with Life Scale (Peterson et al., 2005). It is important to note that many of these studies control for income, at least to some extent, and since incomes are likely to be lower in rural areas, this may give a deceptive appearance of greater rural well-being. For the purposes of geographical

61

Review of the factors affecting well-being equity, the bivariate relationship between geographical area and well-being may be more helpful than a pure location effect. More research is needed to explore the source of the benefit of living in less urban areas and to explore the effectiveness of recreating these in an urban environment.

4.3 Summary of existing evidence It is problematic to summarise the complexities of the relationships between economic, social and environmental factors and well-being as there is a danger of over-simplifying findings, which then leave themselves open to misinterpretation. This is why the summary sheets and tables in Appendices B and C are useful: others can draw their own conclusions about the balance of evidence. Unambiguous findings are rare but the summaries below represent our own interpretations of the evidence. Most of the measures where robust evidence is available are those that tap into SWB (i.e. hedonic and evaluative accounts) in some form or other (e.g. happiness and life satisfaction questions, SWLS, GHQ and CES-D), so our summaries refer primarily to the relationship between a given factor and SWB. Income • Absolute income – general increases in income, particularly for high earners, are unlikely to increase SWB • Relative income – shown to have a significant negative relationship with SWB • Wealth – having savings may be positively related to SWB • Debt – evidence that this is associated with low levels of SWB • Expectations and perceptions – some indication that individuals adapt to changes in income levels, especially when their past income was high and when they have high expectations, and subjective assessments of financial position may be more important to SWB than actual income Personal characteristics • Age – a U shaped relationship with SWB, with SWB lowest around 35-50. • Gender – women tend to score lower on mental health measures than men but there is a wide degree of within-gender variance • Ethnicity – US studies show non-whites to have lower SWB but limited evidence for the UK • Personality – studies of twins have shown that genetic inheritance may partly determine how we assess our lives but there is not much data from large surveys • Physical characteristics – limited evidence Socially developed characteristics • Education – the relationship to SWB is indeterminate • Health – SWB is strongly related to health, particularly psychological health. • Type of work – limited evidence • Unemployment – highly detrimental to SWB, although the effect is moderated by living close to others who are unemployed

62

Review of the factors affecting well-being How we spend our time • Hours worked – limited evidence • Commuting – generally associated with lower levels of life satisfaction and mood • Housework – limited evidence • Caring for others – those who are informal care givers for long periods have consistently lower levels of SWB than non-carers • Community involvement and volunteering – a positive correlation generally exists between SWB and participation in the community but certainly not in all studies • Sleep – limited evidence • Exercise – limited evidence • Religious practice – church attendance is associated with higher SWB, largely irrespective of the particular religion Attitudes and beliefs towards self/others/life • Attitudes towards our circumstances – may be an important determinant of SWB • Trust – the degree of trust in others seems to be positively correlated with life satisfaction but the evidence is very limited • Political persuasion – limited evidence • Religious beliefs – belief in god is associated with higher SWB Relationships • Marriage – being in an intimate relationship (but not necessarily married) is associated with higher levels of SWB, and the dissolution of the relationship is detrimental to SWB • Having children – the effect is indeterminate • Seeing family and friends – positively associated with SWB Economic, social and political environment • Income inequality – the effect is indeterminate • Unemployment rates – limited evidence • Inflation – limited evidence • Welfare and public insurance – limited evidence • Democracy – limited evidence • Climate and quality of natural environment – limited evidence • Security of local environment (crime rates/risk) – living in an unsafe area is associated with lower life satisfaction and mental health • Urbanisation – some evidence that SWB is lower in more densely populated areas

63

Analysis of the BHPS

5. ANALYSIS OF THE BRITISH HOUSEHOLD PANEL SURVEY In this section, we conduct primary analysis of the BHPS that highlights the extent to which different measures of well-being produce similar results and whether the choice of well-being measure makes more of a difference for some population sub-groups than for others. We focused on this dataset because it is a longitudinal survey of about that has been running for a number of years and it contains data on a range of wellbeing measures. Specifically, the BHPS is a longitudinal survey of a nationally representative sample of the population of England, Wales and Scotland (South of the Caledonian Canal, since 1991), and the United Kingdom, including Northern Ireland (since 1999). It is designed as an annual survey of around 5,000 households, resulting in around 10,000 individual interviews. If respondents split-off from original households, all adult members of their new households are also interviewed. Children are interviewed once they reach the age of 16. Interviewing begins in September and continues into the spring, with many questions relating to the year to September.

5.1 Well-being measures in the BHPS The BHPS contains data on a range of well-being measures and so it allows us to make direct comparisons between them. The measures we consider are 1. Gross annual income – in all waves 2. Annual household consumption on some key items: food, utilities, housing costs, child care, private medical insurance, meals out, going out, and some durable goods – in all waves 3. SF-6D – in wave 9 4. Life satisfaction overall – in waves 6-10, 12-13. 5. GHQ12 – in all waves. 6. CASP-19 – in wave 11. Measures [1-3] correspond to a preference satisfaction account, [4] to an evaluative account and [5-6] to a combined account, as defined in section 2. Note that there are no pure flourishing and hedonic measures although there are elements of them in the CASP-19 and the GHQ12, respectively. As we noted in Section 4, the choice of income measures is not a neutral decision. However, the focus here is not on the magnitude of the differences between income measures but rather on the magnitude of differences between well-being measures. We therefore use only equivalised household income, using McClements (1977) equivalence scale (Table 5.1). This is the most commonly used equivalence scale in the UK and is recommended for use by the Office National Statistics. The more people there are in a household, the less income there is to go around and thus the lower the equivalised income. However, each additional person uses up fewer additional resources (e.g. rent, heat, light etc.) and so each additional person costs only a fraction of the first. Imagine two families who each earn £40,000 per year. One is a married couple with no children and the other a married couple with two children (aged 6 and 9) and a grandparent at home. Since the basic unit of comparison on 64

Analysis of the BHPS McClements (1977) scale is two adults, the married couple with no children has a score of 1.0, so £40,000/ 1.0 = £40,000 (equivalised income). However, for the other family, their £40,000 is multiplied by 1.0 (for the couple) + 0.21 (for the 6 year old) + 0.23 (for the 9 year old) + 0.42 (for the grandparent) = 1.86. Thus, in total their equivalised income is £21,505. As noted in Section 3, within a preference satisfaction account of well-being, consumption is arguably a better indication of the extent of preferences which have been met rather than income. Within the BHPS data is available on consumption on some key items: food, utilities (gas, electric, coal), housing costs (net rent or mortgage after rebates), child care, private medical insurance, meals out, going out (leisure expenditure), and some durable goods. If total consumption of these items correlates highly with consumption of non-reported items, the total consumption measure would be reasonable. Consumption is equivalised in the same way as income. As also discussed in Section 3, an alternative means of measuring preference satisfaction is through health state valuation and QALYs. In Wave 9, all respondents were asked the 36 questions comprising the short form medical health outcomes survey. These responses can be combined with public preference data to generate values for the SF-6D that lie on a scale from 0 (equivalent to death) to 1 (full health).

5.2 Comparing the measures Table 5.2 presents the correlations between the various well-being measures. Broadly speaking, correlations below about 0.30 suggest a weak relationship between variables (i.e. how a person scores on one measure will not help predict how they will score on another). Correlations between 0.30 and 0.50 suggest a moderate relationship and above 0.50 a strong relationship (i.e. a high score on one scale will be associated with a high score on another). The results suggest that consumption and income (i.e. measures of preference satisfaction) are not as strongly correlated as might have been expected, which suggests that people have quite different savings/consumption rates. The subjective measures (GHQ, CASP & life satisfaction) correlate relatively well with each other, suggesting that they are tapping into similar underlying constructs. In contrast, the subjective measures do not correlate very well with income or consumption. The SF-6D measure that is used in constructing QALYs correlates more strongly with the psychological measure of health (i.e. the GHQ) than it does with life satisfaction, which is to be expected as both are oriented towards an evaluation of health states rather than global life evaluations. Given these differences across measures, we are interested in knowing whether there are particular situations and subgroups of people who score well using one measurement instrument but poorly using another. The measures do not have clear cut off points which determine whether someone is experiencing high or low well-being, and so all we can say is the relative position of an individual’s score on each scale according to the scores of other people. We can then compare the ranking of different groups across the different measures. Each respondent in each time period is assigned their rank according to each measure.

65

Analysis of the BHPS For illustrative purposes, we consider the differences in rank for the six well-being measures across the following groups: 1. Gender: Men compared to women 2. Age: Over 70s compared to under 70s 3. Education: Those with high education (a degree or equivalent) compared to others 4. Health: Those with poor health (registered, or consider themselves to be, disabled) compared to those not in poor health defined in this way 5. Parenthood: Single parents compared to others 6. Commuting: Those who commute to work for at least 30 minutes each way compared to those with shorter commuting times Figure 5.3 summarises the results. Scores above (below) 0.5 show higher (lower) well-being for the relevant group. These results should be treated with caution since they are based on univariate analyses (i.e. the correlation between well-being and a single factor) but they do illustrate how some groups appear to do well on some measures and less well on others. For example, older people have relatively high life satisfaction despite scoring lower than 0.5 on the other measures, while the opposite is true for those who commute more than an hour a day. Not surprisingly, perhaps, it is uniformly bad to be disabled and, to a lesser extent, to be a single parent. As would be expected, the disabled rank low by all well-being measures. However, the SF-6D ranking is notably lower. The SF-6D measure is specifically health focused, and so it does not give weight to other dimensions of life which may be influential within the other measures. Similarly, single parents score poorly on all measures. There is an interesting divergence in this group between the income rank and the relative higher consumption rank. To fully explore the differences in the factors which are related to the different measures of well-being requires the use of multivariate analysis in which many factors entered into the analysis simultaneously. Since the GHQ Caseness and life satisfaction questions were both asked in waves 6 through to 13 (excluding 11), we can compare the influences on these two measures of well-being, making full use of the panel nature of the BHPS.4 4

The relationship between well-being (WB) and individual characteristics can be represented as the well-being for the individual (i) at time period (t) as a function of a vector of individual characteristics (including external circumstances) and time dummies (X) for individual (i) at time period (t) and an unobserved, individual effect ( α i ) plus an error term ( uit ).

WBit = β X it + α i + u it

(1)

where uit is assumed to be independent and identically distributed over individuals and time with mean zero and variance σ2. The individual effect ( α i ) in this model, which can be thought of as something like an individual’s genetic happiness level, shifts the level of well-being; a happy disposition, or tendency to answer survey questions in a favourable manner, will shift the level of well-being upwards in each time period. In order to get estimates of β, which are the parameters of interest, it is necessary either to estimate the unobserved, individual effect ( α i ) or remove it from the model. The latter can be done using a fixed effects approach, which takes the difference of each variable from its individual level mean, thereby removing the individual effect and comparing each individual to themselves at different time periods. However, this method is problematic when looking at the impact of variables which do not change much over time at the individual level (e.g. education level), and can say nothing about those variables which do not change within individuals (e.g. sex and ethnicity at birth).

66

Analysis of the BHPS Given uncertainty about whether the measures can be treated cardinally (see Section 3), it may be more appropriate to model life satisfaction and the GHQ using ordinal techniques. However, it has been shown that the results from these models are very similar to those from fixed effects ordinary least-squares (OLS) models, which simply seek to minimise the residual sum of squares in a regression model (Ferrer-iCarbonell & Frijters, 2004). Given the advantage of the OLS methods in terms of ease of interpreting the coefficients, we use fixed effects OLS here. For consistency, we also use OLS for analysis of SF-6D and CASP-19. In each case, the well-being measures were re-scaled to a 0-1 scale to aid comparison. In all cases, we include as independent variables those factors which have been shown to influence well-being from our review in Section 4, and which are available within the dataset. To help interpretation of the data in Table 5.4, consider the effects on life satisfaction of talking to neighbours at least once a week. The relationship between these variables is +0.032. This means that people who talk to their neighbours regularly have life satisfaction scores roughly 3% higher than those who do not. So, if an individual who is not friendly with the neighbours has a score of 8.0, the person who is equivalent on all other dimension but who is friends with their neighbours would score 8.24. The results in Table 5.4 show that the determinants of life satisfaction are very similar to the determinants of the GHQ, with in most cases the significance and coefficient size being broadly similar. With regards to the effect of income, a problem paying for accommodation, which is related in some respects to debt, was consistently associated with poorer well-being. In terms of health, having problems walking, giving high amounts of care and regularly visiting one's GP are all associated with lower wellbeing. In terms of employment status, being unemployed and being long-term sick are also associated with lower well-being as are attitudes towards one's health. In terms of relationships, being divorced or separated and living alone is associated with poorer well-being across all measures. Finally, talking frequently to one's neighbours was the only example where well-being was consistently positively associated with well-being in all instances. By and large, these findings are all consistent with conclusions that emerged from our review of the literature in Section 4. Given that the present findings used different measures of well-being and both cross-sectional and longitudinal data suggest that they are especially robust and policy makers – particularly in the UK – can have relative confidence in them. However, there are also several noticeable differences between measures. In some instances, the differences are in terms of magnitude of effects with some measures showing a significant effect (always in the same direction) and others showing no significant effects. In other cases, the data suggest different conclusions, with some measures suggesting the characteristic improves well-being and others suggesting that it worsens well-being. For example, equivalent income is positively associated with life satisfaction and CASP-19, not associated with the SF-6D and only positively associated with the GHQ in Wave 11. These findings seem to suggest that more money makes people feel better about their lives but not about their physical or mental health, and highlight how there can sometimes be important differences within an account of well-being (in this case within the preference satisfaction account).

67

Analysis of the BHPS

Age is also sometimes negatively associated with well-being (life satisfaction in Wave 9 and CASP-19) but often not significantly associated at all. Similarly, being a widow who lived alone was generally associated with lower well-being but not for CASP-19. This is particularly interesting because CASP-19 was developed for an older cohort, who is more likely to be widowed, so perhaps it is better at emphasising the positives of old age and widowhood than the other measures. CASP-19 also found a consistent negative effect of children whereas the other measures found either no relation or rare (and very small) effects. Again, this suggests that the CASP-19 is problematic for general population use and highlights the need for further development of measures designed to tap into flourishing accounts of well-being. Despite these differences there are key factors which have a consistent relationship with well-being regardless of which measure is used. The following significantly affect well-being in a negative ways under all accounts: being unemployed, being long term sick, not seeing friends and family or not talking to neighbours regularly, being divorced or separated and living alone, poor self-rated health, problems walking, high number GP visits, caring for someone full-time, and problems paying for accommodation. Whilst regression analysis assumes a causal relationship, it does not prove that this is so, and causality needs to be judged from other evidence. Considering each of these factors in turn, we can say that being unemployed does cause reductions in wellbeing: fixed effects models have shown the lower well-being of the unemployed is not due to personality characteristics and the well-being loss from unemployment is similar whether unemployment is voluntary or involuntary (Winkelmann & Winkelmann, 1998). It is plausible that being long-term sick would reduce well-being but low well-being may mean that people lack motivation to come off long-term sickness. It is highly plausible that contact with family, friends and neighbours act as a boost to well-being but it is also possible that periods where one is experiencing lower wellbeing are those times where one lacks motivation to socialise. A possibility for future work to address this causality issue is to explore voluntary and involuntary (such as after moving to a new location) differences in contact with friends and family. It may also be possible to consider differences which are exogenous to current mood but which are related to talking to neighbours, such as house type and location. Whilst it may be that lower well-being increases the likelihood of divorce, this path seems unlikely to be the sole cause of the substantial separation and divorce effect. Assessing the direction of causality in the association between the range of selfassessed health variables and well-being is problematic because it is hard to know where the distinction between health and well-being should be drawn, particularly when health contains psychological health. Even for GP visits, it could be that low well-being is what drives the contact with the GP. Low well-being is unlikely to be the cause of carer workload. However, there may be other unobserved variables, related to carer work load which are being picked up in this relationship e.g. the presence of a sick household member. Ideally studies are needed which control for this. Problems paying for accommodation may arise

68

Analysis of the BHPS following a period of low well-being in which an individual looses control over their finances. However, it would seem more likely that difficulties meeting financial commitments cause lower well-being. This discussion highlights the fact that the direction of causality can be inferred from some of the existing evidence but much of it is based on intuition. We plan to make further use of the panel nature of the BHPS to investigate some of the issues around causality more fully, and would urge others to do likewise in the few other panels currently available (e.g. the GSOEP).

69

Concluding remarks

6. CONCLUDING REMARKS One of the main purposes of this review was to identify the influences on personal well-being. We have tackled this by examining the different accounts of well-being, the various measures of well-being associated with these different definitions and findings emerging from large-scale survey data using these measures. Although this approach does not allow us to firmly establish causality, and therefore say unambiguously what the influences on well-being are, it does provide a means of assessing the robustness with which certain factors are associated with well-being. If a certain characteristic consistently affects well-being in a certain direction, irrespective of the concept and measures of well-being used, this is highly suggestive of an important relationship between the two.

6.1 Existing evidence One of our main aims has been to identify whether and where there is consistency across different measures and whether the measures being used are picking up on the kind of influences we would expect them to. In our discussion of well-being measures in Section 3, we paid particular attention to internal features of the measures, such as time frame and comparison standards. Now that we have reviewed the evidence we are also able to say something about the degree to which the measures produce similar results (convergent validity) and the degree to which the measures distinguish between important determinants of well-being (discriminant validity). One of the most encouraging findings of our review and primary analysis of the BHPS data is that different measures of well-being often produce very similar results (convergent validity). Most studies that use more than one measure on the same data find remarkably similar co-variants between the single life satisfaction question and the single happiness question (e.g. van den Berg & Ferrer-i-Carbonell, 2005 for the Netherlands). Thoits and Hewitt (2001), using the US Americans’ Changing Lives study, found that life satisfaction, overall happiness and the CES-D depression scale were also generally related to the same variables. Therefore, it seems that many of the methodological issues surrounding the specific measures that we discussed in Section 3 such as the time frame, use of reference standards, number of scale points and so on, made relatively little difference in many contexts. A further source of confidence in many of the measures of well-being we review is that they are picking up differences in objective circumstances that we would expect to find (discriminant validity). For instance, even simple one-item happiness and life satisfaction questions are showing significant differences between those who are employed versus unemployed, single versus living with a partner, those who live in a state with good versus poor quality of governance, and so on. Importantly, these measures are not asking about happiness in specific domains (work, relationships etc) which we would expect to pick up on such differences. Rather, they are global assessments which are based on a myriad of factors and it is therefore very encouraging that the effects of different circumstances are being picked up even against the backdrop of all the competing influences on subjective well-being.

70

Concluding remarks The results from the analysis of the BHPS support and supplement the findings from the review. They generally support the finding that both absolute income and relative income are important, but the relationships are not straightforward and more work is needed to understand complexities of income and rank and the perception of income and rank. The inclusion of variables identifying those who have had problems paying for accommodation supports those findings that suggest that debt and problems with paying the bills is detrimental to well-being. This may indicate that the non-linearities in the income relationship are not fully captured by the log transformation, and, where household income is such that the household is under specific financial pressure, life satisfaction is reduced. The findings on divorce and separation support the consistent finding that these are difficult times for people. The detrimental effect of poor subjective health and of more objective measures (problems walking and GP visits) is in line with the importance of objective and subjective health found in the review. The detrimental effect of unemployment is well supported by the BHPS data, and supplemented by the identification of the additional detrimental effect of having an employment status of long term sick (controlling for health status). Many studies do not specifically classify long term sick as separate from being out of the labour force but these are a significant and politically relevant group. The findings on the (negative) impact of adopting caring roles and the (positive) effect of seeing friends and family are consistent with those identified in the literature review. Despite the many encouraging findings regarding convergent and discriminate validity we also find instances where different conclusions appear to emerge depending on which theoretical account of well-being, which specific measures within accounts and which statistical analyses were used. One issue we were particularly keen to investigate was whether flourishing accounts produce different results from hedonic or evaluative accounts. Unfortunately, very few studies in our review included flourishing account measures. One exception is Keyes et al. (2002) who, using data from the MIDUS study find that the correlation between PWB and education is higher than the correlation between life satisfaction and education (0.20 versus 0.07). This seems to indicate that any positive effects of education are more important for well-being in terms of flourishing than in terms of global life evaluations, and this makes some intuitive sense. This finding is supported by the BHPS analysis which finds that well-being increases with education when measured by the CASP-19, but decreases or is not significantly related to well-being when measured using life satisfaction, the SF-6D or the GHQ. There are issues with the reliability of the PWB and the CASP-19, as discussed in Section 3. There is also some evidence of differences across measures within an account of wellbeing. For example, where both positive and negative measures of well-being have been used in the same study there are a number of instances where these hedonic measures suggest different conclusions. Women tend to score higher on positive measures but also higher on negative measures (e.g. Baker et al., 2005 and Ritchey et al., 2001 using the Americans’ Changing Lives study and Helliwell & Putman, 2004 using the World Values Survey) though this findings is not always found (e.g. Greenfield & Marks, 2004). Using MIDUS data, Greenfield and Marks (2004) find that income and volunteering both tend to increase positive affect but not reduce negative affect. By contrast, getting older is associated with a reduction in negative

71

Concluding remarks affect but no increase in positive affect. Finally, psychological problems during adolescence have been found to predict negative but not positive measures of wellbeing (Flouri, 2004, using UK, NCDS data). Diener et al. (1999) argue that it is important to measure positive and negative affect separately as they are separate constructs with separate biological processes and are affected by different circumstances rather than simply being two ends of the same spectrum. We would also urge that future studies incorporate measures of both positive and negative aspects of well-being since this may increase sensitivity by distinguishing circumstances that aid (or hinder) either the establishment of positive experiences or the reduction of negative ones. Given that policy may be particularly interested in one or other of these, it is important to establish where the effects are associated with changes in positive as compared to negative aspects of well-being.. In relation to accounts of SWB, we find a number of differences between “happiness” and “life satisfaction” as a function of precise wording in the limited number of studies where both were included. Using WVS data Helliwell and Putnam (2004), for instance, find that those over 55 are less happy than the under 25s but not less satisfied with their lives. They also find that religion is associated with greater happiness but not life satisfaction and that education reduces life satisfaction, but is not significantly related to overall happiness. Also using the WVS, Haller & Hadler (2006) find that being a housewife increases overall happiness but not life satisfaction, whereas having children, being a student, political freedom, and social expenditure increases life satisfaction but not happiness. Helliwell (2004) conjectures the life satisfaction question “triggers answers that are more reflective of ones whole life experiences than one’s current circumstances or mood”. Further research could test this hypothesis by asking people to think aloud when responding to happiness and satisfaction questions to see whether they do indeed focus attention on different aspects of life. However, as with many of the findings in this report, it is important not to overstate the differences. While there are occasions when the choice of well-being measures (dependent variable) has some influence on the findings, the differences across most subjective measures of wellbeing that have been used in large scale datasets are generally small. It is therefore important to explore other reasons for the variability in the findings. For example, results and conclusions differ according to the form in which the independent variables are entered into the well-being function. For example, recent studies have shown convincingly that there is a non-linear relationship between life satisfaction and age. Comparisons against studies which restrict age to be linear are therefore of little value. There may be other factors which have a non-linear relationship with well-being that have been under-explored (e.g. education, health). A useful avenue for future research is to focus on those attributes which have been shown to impact on well-being in certain studies and explore in more detail the shape of that relationship. Another source of apparent discrepancy between results arises from the use of different categorisation of variables and choice of reference category. For example, marital status, employment status and education can all be categorised in slightly different ways (e.g. treating married and those cohabiting together or separately) and

72

Concluding remarks the choice of reference category can vary (e.g. the employment category may be compared to those employed, those out of the labour market or those employed and self-employed). Therefore, care is needed in making comparisons between studies. Different findings may also arise due to the inclusion of different control variables. We noted earlier for instance that Clark (2003b) found that compared to being single, being married resulted in significantly better GHQ scores when controlling for employment status but not when controlling for income. Findings like this alert researchers and policy makers alike to the need to be careful in interpreting both significant and non-significant findings without considering what else is being controlled for in any given analysis. Future research interested in the impact of specific variables on well-being should systematically introduce different control variables and explore the impact this is having on the relationship between the main independent and dependent variable. Whilst all researchers will be limited by their data, as consensus is reached on the main causes of well-being, this should enable greater consistency across researchers. Our findings suggest researchers should at least be aware of the impact of income, relative income, health, personal and community relationships and employment status. It has been shown by Ferrer-i-Carbonell and Frijters (2004) that controlling for the individual unobserved effects can impact on the findings of what influences wellbeing. Most studies reviewed here which compare different models find that controlling for fixed effects weakens the coefficient sizes and the significance level. Fixed effects models are unsuited to exploring the impact of variables with little or no variability within people (e.g. education), so whilst being theoretically superior in some respects, the evidence cannot be taken as evidence of non-significance of such variables. It is an ongoing challenge for researchers to develop methods to best explore the impact of variables with minimal individual life time variance yet which are also likely to be correlated with the unobserved individual effect. Finally, it is not always clear what should be controlled for in any well-being model. Consider the case of expectations. If we control for income and other variables, we generally find that high expectations are detrimental to well-being (Graham & Pettinato’s (2002) ‘frustrated achievers’). But should we really be controlling for income, since people with high expectations generally earn more money, perhaps precisely because they have high expectations. In some circumstances, researchers and policy-makers may gain important information from bivariate analysis. In summary, the different measures of SWB generally produce similar results but these are different from those generated by preference satisfaction, such as income. As we indicated in Section 3, there are some question marks about the measures of flourishing, such as the PWB, and there may be a need to treat the results from such measures with a little more caution than those from measures of SWB, such as the global life satisfaction question. There is also some doubt about the usefulness of the CASP-19 in a general population (as compared to an older) sample. One very firm conclusion to draw is that the existing evidence base is not quite as strong as some people may have suggested and there are some important avenues for future research to which we now turn.

73

Concluding remarks

6.2 Further research There are a number of key issues for future research that emerge from this project. There are important questions about precisely which measures of well-being should be used, methodological challenges about how to interpret the evidence, some specific factors that should be explored further given their potential policy relevance, and important questions about the distribution of well-being across society and the general public’s views on well-being in policy. 6.2.1 Measuring well-being Income is an incomplete measure of well-being as defined by the satisfaction of preferences and so we should be very careful about how we interpret the well-being effects of changes in income. For those committed to a flourishing account of wellbeing, given some of the problems with the reliability and interpretation of the results from the existing measures of psychological well-being, we would suggest the development of new measures. Given the lack of reliable evidence on measures of flourishing, future research should also consider the degree to which flourishing accounts of well-being produce different results from hedonic or evaluative accounts. Where there is a commitment to the routine assessment of SWB, a global measure of life satisfaction on a 0-10 scale should provide reliable information in many policy contexts. For those interested in understanding the relationship between how we allocate our time and our well-being, the day reconstruction method (DRM), which measures feelings as recalled from activities during the previous day, offers a promising avenue for future research. It also provides useful data for those wishing to calculate QALY-type indices which show the profile of well-being over relatively short periods. However, as it stands, the approach is open to the criticism that policymakers and the public are likely to give some weight to considerations besides feelings, such as the evaluative components of our moment-to-moment experiences. Therefore, it would be worth exploring ways in which the more evaluative aspects of our lives can be incorporated into the well-being of different activities. 6.2.2 Methodological challenges One of the main findings of this research is that different results can be obtained according to how the factors are categorised and which other factors are controlled for in the analysis. Therefore, researchers and policy-makers and researchers need to be careful in interpreting both significant and non-significant findings. Future research interested in the impact of specific variables on well-being should systematically introduce different control variables and explore the impact this is having on the relationship between the main independent and dependent variable. Whilst all researchers will be limited by their data, as consensus is reached on the main causes of well-being, this should enable greater consistency across studies. Further investigation of existing data should consider more fully the household as a unit of analysis and the potential trade-offs in well-being across family members. For example, commuting might be bad for the commuter but bring benefits (from higher 74

Concluding remarks income and a nicer house) to other household members. To date most well-being research has focused on the well-being of the individual. However, there are good reasons for also focusing on well-being at the household level. Magdol (2002), for instance, finds higher levels of depressive symptoms amongst women who feel they have sacrificed their careers for their partners. Whether or not any increases in their partner's well-being is sufficient compensation for this loss remains to be explored. One very firm conclusion that can be drawn from our review is that the existing evidence base is not quite as strong as some people may have suggested and there are some important avenues for future research that could be explored with the existing panel datasets. This, in addition to the lack of clear evidence on causality, makes it difficult for us to make clear policy recommendations at this stage. Nevertheless, our findings suggest researchers should at least be aware of the impact of income, relative income, health, personal and community relationships and employment status in their analysis. We are also able to make some clear recommendations about where future research into some of these and other policy relevant variables should be directed. 6.2.3 Key factors for future research 6.2.3.1 Income rank The importance of income rank and perception of income rank are just beginning to be recognised. More research is needed to understand how income rank impacts upon well-being, and how income comparisons work. This would include exploring to whom people compare themselves. A greater understanding of precisely why and how reference incomes impact on well-being is also needed, for example, is it driven by an ordering effect or by the distance between an individuals income and the income of those around them; does it operate by making the individual dissatisfied with their own income or does it create a pressure on individuals for them to overspend and put themselves under financial pressure? Recent evidence suggests that comparisons are upward looking (Ferrer-i-Carbonell, 2005) and that it is the income of the top income group which dominates the reference income (Blanchflower and Oswald, 2004b). This finding requires validating in a range of different settings, and on UK data. 6.2.3.2 Education Another policy relevant relationship is the one between education and well-being. However, the evidence currently available is ambiguous. Some studies find a positive relationship between each additional level of education and life satisfaction while others find that middle level education is related to the highest satisfaction. The coefficient on education is often responsive to the inclusion of other variables within the model and there is a suggestion that, like income, the benefits to education may be positional rather than absolute. The effect of social status and rank across a range of domains in life is therefore something that requires urgent attention. 6.2.3.3 Social capital The BHPS analysis suggests that social contact is positively associated with wellbeing. This is an important finding, and suggests the role of social capital and contact

75

Concluding remarks with local community has been under explored within the literature, particularly within fixed effects models. Future research is needed to understand the link between contact with friends, family and neighbours and well-being and critically the direction of causality in this relationship. Unlike many variables, there is unlikely to be a time delay in the causal pathways between social contact and well-being, which complicates any investigation into the direction of causality. One option to explore may be the indirect modelling of contact with neighbours via such variables as housing type and neighbour contact in the area. Such approaches could be used to establish causality more generally. 6.2.3.4 Other factors There is the need for more research into the effect of macroeconomic indicators (e.g. unemployment and inflation rates, benefits levels etc.) on well-being. In particular, given the increased prominence given to problems associated with environmental and climate change, there is the need to consider more fully the effects of environmental factors (such as air pollution) on well-being. Where possible, the effects of crime and the fear of crime on well-being should also be explored more fully. 6.2.4 Well-being in policy The usefulness of any measure of well-being in policy will depend in part on how it addresses distributional issues and on the degree to which it are considered to be a legitimate focus of government. The distribution of well-being across society and across different population sub-groups has largely been ignored in the literature to date. There are many policy relevant questions which could be explored with the existing panel datasets, such as: how does the distribution of well-being differ according to different measures of well-being; how does the distribution of well-being compare to other distributions such as income; and how has the distribution of wellbeing changed over time? As well as analysis of existing datasets, there is also the need for new data, both in relation to the determinants of well-being, and also in relation to how the general public feel about using particular measures of personal well-being to inform public policy. This would be one way in which the largely normative question of policy relevance could be addressed with empirical evidence. It would also be interesting to see how different policy-makers across different areas and levels of decision-making view the different concepts of well-being as they are applied to policy. In conclusion, it appears from the existing evidence that measures of SWB would produce different results to preference satisfaction measures, such as income. However, despite many caveats and uncertainties about how to interpret some of the existing evidence, it would seem that most measures of SWB would produce similar results to one another. Therefore, for those interested in the subjective assessment of an individual’s life, it might not matter too much which measure of SWB is used to assess the well-being of different population groups.

76

Glossary

GLOSSARY OF TERMS ACL -

American's Changing Lives. A US based general population survey (see Table A6).

Affectometer 2 -

A 40 item self-report questionnaire developed in New Zealand to measure positive and negative affective states and experiences.

BHPS -

British Household Panel Survey. A longitudinal panel survey running since 1991 and using face to face interviews with all adult members of a household. Study began with 5,000 households and has expanded over time (see Table A6)

CASP-19 -

A 19 item self-report questionnaire designed to measure wellbeing in elderly populations. The initials stand for the four supposed dimensions of the scale: Competence, Autonomy, Self-realization & Pleasure.

CES-D -

Centre for Epidemiological Studies Depression. A 20 item selfreport questionnaire examining depressive symptoms (see Table A3)

Chonbach's -

A statistical measure of the inter-correlation between the items on a Alpha scale. The higher the alpha, the greater the likelihood that the items are reliably measuring the same underlying concept. It is generally agreed that for a scale to be deemed reliable the alpha must be at least 0.6.

DRM -

Day Reconstruction Method. A methodology asking people to reconstruct their previous day and comment on what they were doing, who they were with and how they felt. (see Table A4)

Econlit -

Economic Literature. An electronic database containing references to the most important publications in economics and related disciplines.

EQ-5D -

A descriptive systems for health states that defines health in five dimensions (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) and that allows the valuation for each health state to be expressed in a single index number between 0 and 1

ESS / EVS -

European Social Survey / European Values Survey. A nationally representative general population survey carried out in over 20 European countries.

77

Glossary Eudaimonia -

A definition of well-being based on Aristotelian notions of developing human potential rather than experiencing positive affect.

Eurobarometer -

An interview based general population survey carried out in 12 different countries across Europe.

Fixed Effects -

Method of regression analysis which makes inferences about the coefficients conditional on the individual effects in the sample. This is equivalent to generating dummy variables for each individual and including them in the regression to control for these unobserved but fixed effects

GHQ / GHQ-12 -

General Health Questionnaire / General Health Questionnaire 12 item Scale. A self-report questionnaire which, despite the name, asks about mental rather than physical health (see Table A5). There are various versions of the GHQ, the most commonly used is the GHQ12, which has 12 questions. This may be scored either on a 0 to 36 scale (known as GHQ12 Likert) or on a 0 to 12 scale (known as GHQ Caseness).

GSOEP -

German Socio-Economic Panel Survey. A longitudinal panel survey using face to face interviews with an entire sample of over 24,000 respondents.

GSS -

General Social Survey. Although several countries carry out General Social Surveys, the one we refer to here is primarily the one from the United States (see Table A6).

HHPS -

Hungarian Household Panel Survey (run by the Tarki Institute) National panel survey based in Hungary (see Table A6)

HILDA -

Household Income Labour Dynamics in Australia. A household-based panel study which began in 2001, based general population survey (see Table A6).

ISSP -

International Social Survey Program. Annual collaboration between survey researchers conducting carried out in over 40 countries worldwide.

MIDUS -

Midlife in the US. A self-report survey carried out in the US.

NCDS -

National Child Development Survey. A British survey which monitors a cohort of people born in Britain between the 3rd and 9th March 1958 every seven years.

NICE -

National Institute for Health and Clinical Excellence. An independent UK organisation responsible for providing national guidance on the promotion of good health and the prevention and treatment of ill health.

78

Glossary

NSFH -

National Survey of Households and Familes. A US based representative survey of English and Spanish speaking households.

OLS -

Ordinary Least Squares. A method for estimating the coefficients in a regression model based on minimizing the residual sum of squares.

OTH -

Orientations to Happiness Scale. An 18 item self-report questionnaire designed to examine Seligman's (2002) concept of Authentic Happiness.

PANAS -

Positive and Negative Affect Scale. Self report questionnaire which asks people to say how often they have experienced a range of different emotional states over a recent time period.

Psychinfo -

Psychological Information. An electronic database containing references to the most important publications in psychology and related disciplines.

PWBS -

Psychological Well-Being Scales. A self-report questionnaire developed by Carol Ryff (1989) to examine Eudaimonic aspects of well-being.

PWI -

Personal Well-being Index. An Australian self-report questionnaire designed to examine satisfaction with life in general as well as with specific life domains (e.g. health, work etc.).

QALY -

Quality Adjusted Life Year. A measure that combines length of life and the value of quality of life into a single number, where one QALY equals one year in full health.

RLMS -

Russian Longitudinal Panel Survey. A Russian based self-report survey with an almost representative sample.

SCBS -

Social Capital Benchmark Survey. A US based survey with representative samples for certain regions.

SF-6D

A descriptive system for health states that defines health according to six dimensions (physical functioning, role limitations, social functioning, pain, mental health and vitality) and that allows the valuation for each health state to be expressed in a single index number between 0 and 1.

SLLS / LNU -

Swedish Level of Living Survey. A Swedish based general population survey running between 1968-1991.

79

Glossary SHSP -

Swiss Household Panel Survey. A Swiss based household survey running since 1999.

SG -

Standard Gamble. Respondents are presented with a choice between 1) an intermediate health state and 2) a gamble between full health and death. The probability of death is varied until they are indifferent between options 1 and 2.

Split-test correlation- Randomly divide all items that purport to measure the same construct into two sets, take the correlation between the two sets. If they are measuring the same latent construct the correlation should be high. SWB -

Subjective Well-Being. Widely defined as a composite measure of how people think and feel about their lives based on selfreported affect and life satisfaction.

SWLS -

Satisfaction With Life Scale. 5 item self report questionnaire to measure subjective satisfaction with life.

Test-retest -

A statistical approach which tests people on two separate occasions reliability to examine the degree to which their responses are consistent over time.

TTO -

Time Trade Off. Respondents state the length of time in full health that they consider to be equivalent to a longer period of time in poor health.

Wave -

Within panel data sets (such as the BHP) each consecutive data collection round (which may be annually or less frequently) is known as a ‘wave’. Wave 1 is the first year data is collected, wave 2, the second year and so on.

WVS -

World Values Survey. A self-report survey carried out in many countries around the world. TABLES AND FIGURES

80

Section 5 tables and figures

SECTION 5 TABLES AND FIGURES 5.1 The McClements equivalent income scale Type of household member

Equivalence value 1.00 0.42 0.36 (/ adult) 0.61 0.46 0.42 0.36 (/ adult) 0.36 0.27 0.25 0.23 0.21 0.18 0.09

Married head of household (i.e. a married couple of 2 adults) 1st additional adult 2nd (or more) additional adult Single head of household (i.e. 1 adult) 1st additional adult 2nd additional adult 3rd (or more) additional adult Child aged: 16-18 13-15 11-12 8-10 5-7 2-4 Under 2

5.2 Correlations between different well-being measures Equivalised income Equivalised income Equivalised consumption SF-6D Life satisfaction GHQ CASP-19

Equivalised consumption

SF-6D

Life satisfaction overall

GHQ Caseness

1 0.3941 (n= 65637) 0.1331 (n= 15073) 0.0618 (n= 89484) 0.0497 (n=1.5e+05) 0.1879 (n= 16862)

1 0.1369 (n=6452) 0.0245 (n= 48172) 0.0164 (n= 63488) 0.1348 (n=15006)

81

1 0.3796 (n= 14673) 0.5102 (n= 14618) Na

1 0.5054 (n= 88473) Na

1 0.5008 (n= 16722)

Section 5 tables and figures

5.3 Average ranks in well-being across different groups Each respondent in each time period is assigned their rank according to where they fall in the distribution on each measure of well-being, using a scale of 0 for the lowest and 1 for the highest of that time period. We then consider sub-groups of the data, and see whether their average ranking is different for the different well-being measures. Average ranking of males by well-being measures 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Equivalent income

Equivalent consumption

SF-6d

Life satisfaction overall

Reversed GHQ Caseness

CASP-19

Reversed GHQ Caseness

CASP-19

Average ranking of those aged over 70 by well-being measures 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Equivalent income

Equivalent consumption

SF-6d

82

Life satisfaction overall

Section 5 tables and figures Average ranking of those with a degree by well-being measures 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Equivalent income

Equivalent consumption

SF-6d

Life satisfaction overall

Reversed GHQ Caseness

CASP-19

Average ranking of those disabled by well-being measures 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Equivalent income

Equivalent consumption

SF-6d

83

Life satisfaction overall

Reversed GHQ Caseness

CASP-19

Section 5 tables and figures Average ranking of single parents by well-being measures 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Equivalent income

Equivalent consumption

SF-6d

Life satisfaction overall

Reversed GHQ Caseness

CASP-19

Average ranking of those who commute for at least an hour by well-being measures 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Equivalent income

Equivalent consumption

SF-6d

84

Life satisfaction overall

Reversed GHQ Caseness

CASP-19

Section 5 tables and figures

5.4: Determinants of life satisfaction, inverse-GHQ, SF-6D, and CASP-19 using OLS and OLS fixed effects, BHPS various years

Cross-sectional samples Life satisfaction

Wave 9 Inverse GHQ

Sf-6

Income Log equivalent hh income Reference income Problems paying for accommodation

0.0150*** -0.0001

-0.0006 -0.0004

0.0022 -0.0001

-0.0718***

-0.0691***

Personal characteristics Age Age2 Male

-0.0043*** 0.0053*** -0.0122**

Education (highest qualification) No qualifications (reference) O levels, commercial or equivalent A levels Degree Still studying Health Problems walking Hospital stay in last year Six or more GP visits

Longitudinal sample (fixed effects)

Wave 11 Inverse GHQ CASP-19

Waves 6-10 & 12-13 Life satisfaction Inverse GHQ

0.0079* -0.0006

0.0110*** 0.0005**

0.0035** 0.0000

-0.0233***

-0.0805***

-0.0459***

-0.0374***

-0.0015 0.0023* 0.0347***

-0.0004 -0.0002 0.0224***

0.0005 0.0002 0.0338***

-0.0010* 0.0005 -0.0108***

-0.0065 -0.0001 (dropped)

-0.0033 -0.0003 (dropped)

-0.0204*** -0.0332*** -0.0320*** -0.0365

-0.0135* -0.0319*** -0.0309*** 0.0097

-0.0015 -0.0022 -0.0072* -0.0082

-0.0081 -0.0161 -0.0222*** -0.0127

0.0099*** 0.0137*** 0.0208*** -0.0214

-0.0060 0.0055 -0.0005 -0.0328

-0.0030 0.0110 0.0078 -0.0388

-0.0455*** -0.0019 -0.0220***

-0.0441*** -0.0174* -0.0638***

-0.1223*** -0.0238*** -0.0512***

-0.0604*** -0.0101 -0.0493***

-0.0488*** 0.0023 -0.0067*

-0.0286*** -0.0025 -0.0103***

85

0.0042 -0.0004** -0.0493***

-0.0454*** -0.0133*** -0.0298***

Section 5 tables and figures

Employment status Employed (reference) Long term sick Retired Unemployed Maternity leave Self employed Family carer Student Gov. training Other activity How we spend our time Active in religious orgs Gives care >50hrs pw Attitudes towards health Health poor or very poor Self-rated health fair Self-rated health good (ref) Self-rated health excellent Intimate relations Married (reference) Never married live alone Never married live with others Cohabiting Divorced or separated live alone Divorced or separated live with others Widow live alone Widow live with others

-0.0462*** 0.0363*** -0.0638*** 0.0247 0.0058 0.0040 0.0194 0.0131 -0.0444

-0.0501*** 0.0105 -0.0893*** -0.0401 -0.0026 -0.0058 -0.0381** -0.0054 -0.0326

-0.0582*** 0.0000 -0.0183*** -0.0298 -0.0017 -0.0050 -0.0008 -0.0101 -0.0477***

-0.0728*** 0.0176 -0.0571*** -0.0078 0.0013 -0.0149 -0.0211 -0.0654 -0.0073

-0.0511*** 0.0137** -0.0335*** 0.0257 0.0217*** -0.0101* 0.0155** 0.0302 0.0165

-0.0457*** 0.0050 -0.0386*** 0.0409*** -0.0005 -0.0145** 0.0175** -0.0222 -0.0092

-0.0696*** -0.0021 -0.0793*** -0.0427** -0.0183** -0.0239*** 0.0005 -0.0498 -0.0187

0.0202** -0.0286*

-0.0017 -0.0562***

-0.0044 -0.0179**

0.0069 -0.0614***

0.0201*** -0.0466***

-0.0326***

-0.0320***

-0.2209*** -0.0919***

-0.2565*** -0.1034***

-0.1681*** -0.0914***

-0.2061*** -0.0654***

-0.0908*** -0.0414***

-0.0792*** -0.0374***

-0.1490*** -0.0479***

0.0626***

0.0456***

0.0465***

0.0311***

0.0440***

0.0201***

0.0254***

-0.0619***

-0.0207

-0.0142**

0.0048

-0.0066

-0.0183*

-0.0087

-0.0333*** -0.0154*

0.0006 -0.0205*

-0.0023 -0.0013

0.0331*** 0.0020

-0.0015 0.0039

-0.0219** 0.0017

-0.0153 -0.0036

-0.0888***

-0.0337**

-0.0153***

-0.0307**

-0.0253***

-0.0360***

-0.0494***

-0.0969*** -0.0650*** -0.0658***

-0.0653*** -0.0435*** -0.0688***

-0.0106 -0.0217*** -0.0224**

-0.0362* -0.0211* -0.0064

-0.0273*** -0.0033 -0.0028

-0.0542*** -0.0323*** -0.0400**

-0.0848*** -0.0842*** -0.0818***

86

Section 5 tables and figures

Having children No children in the household (reference) Children at home in the household Children at school in the household Children at home and school in the household Friends and family See friends or family at least weekly Talk to neighbours at least weekly

0.0021

0.0077

0.0130**

0.0053

-0.0331***

-0.0047

0.0024

-0.0100

0.0001

-0.0046

-0.0293***

-0.0016

0.0042

-0.0191

0.0059

0.0101

0.0005

-0.0372***

-0.0107

0.0063

0.0211***

0.0215**

0.0066*

0.0265***

0.0260***

0.0047

0.0070

0.0320***

0.0165**

0.0065*

0.0317***

0.0168***

0.0093***

0.0071*

-0.0178**

Year dummies included Constant Yes Yes Yes Observations 10360 10360 10360 R2 within / Adj. R2 0.1991 0.1755 0.5390 * p < 0.05, ** p < 0.01, *** p < 0.001 i.e. these indicate different degrees of significance

87

Yes 12231 0.1816

Yes 12231 0.2877

Yes Yes 57834 0.0383

Yes Yes 57834 0.0587

References

REFERENCES Alesina, A., Di Tella, R & MacCulloch, R. (2004) Inequality and happiness: are Europeans and Americans different? Journal of Public Economics, 88, 2009– 42 Atkinson, A. B., Rainwater, L., and Smeeding, T. (1995) Income Distribution in OECD Countries: the Evidence from the Luxembourg Income Study (LIS). OECD, Paris Baker, L. A., Cahalin, L. P., Gerst, K, and Burr, J. A. (2005) Productive activities and subjective well-being among older adults: The influence of number of activities and time commitment. Social Indicators Research. 73, 431-58 Bardasi, E. & Francesconi, M. (2004) The impact of atypical employment on individual well-being: Evidence from a panel of British workers. Social Science and Medicine, 58, 1671-1688 Beekman, A.T., Deeg DJ, van Limbeek J, et al. (1997) Criterion validity of the Center for Epidemiologic Studies Depression Scale (CES-D): results from a community-based sample of older subjects in the Netherlands. Psychological Medicine, 27, 231–5 Bentham, J. (1789/1996) An introduction to the principles of morals and legislation, 1996 edition, edited by J.H. Burns and H.L.A. Hart. Oxford: Clarendon Press Biblarz, T.J. & Gottainer, G. (2000) Family structure and children's success: A comparison of widowed and divorced single-mother families. Journal of Marriage and the Family, 62, 533-548 Biddle, S.J.H. & Ekkekakis, P. (2005) Physicall active lifestyles and well-being. In F.A. Huppert, N. Baylis & B. Keverne (Eds.) The Science of Well-Being (pp. 141-168). Oxford: Oxford University Press Biswas-Diener, R. and Diener E. (2001) Making the Best of a Bad Situation: Satisfaction in the Slums of Calcutta, Social Indicators Research, 55(3), 32952 Bjornskov, C. (2003) The Happy Few: Cross-Country Evidence on Social Capital and Life Satisfaction. Kyklos, 56(1), 3-16 Blanchflower, D. G. and Oswald, A. J. (1998) What makes an Entrepreneur? Journal of Labor Economics, 16 (1), 26-60 Blanchflower, D. G, and Oswald A. J. (2004a) Money, sex and happiness: an empirical study. Scand. J. of Economics, 106(3), 393-415 Blanchflower, D. G. and Oswald A. J. (2004b) Well-being over time in Britain and the USA. Journal of Public Economics, 88, 1359-86 Blanchflower, D. G., and Oswald A. J. (2005) Happiness and the Human Development Index: The Paradox of Australia. The Australian Economic Review, 38(3), 307–18 Borooah, V. K. (2005) How to assess happiness? A tale of three measures. Applied Economics Letters, 12, 191-94 Bradburn, N.M. (1969) The Structure of Psychological Well-Being. Chicago: Aldine Brazier, J. and Deverill, M. (1999) A checklist for judging preference-based measures of health related quality of life: Learning from psychometrics, Health Economics, 8, (1) 41-51

88

References Brazier, J.E., Tsuchiya, A., Roberts, J. and Busschbach, J. (2004) A comparison of the EQ-5D and the SF-6D across seven patient groups. Health Economics, 13(9), 873-884 Brown, S.L. (2000) The effect of union type on psychological well-being: Depression among cohabitors versus marrieds. Journal of Health and Social Behaviour, 41, 241-255 Brown, S., Taylor K.B., and Wheatley Price, S. (2005) Debt and distress: Evaluating the psychological cost of credit. Journal of Economic Psychology, 26, 642-66 Bukenya, J. O, Gebremedhin, T. G. and Schaeffer, P. V. (2003) Analysis of Rural Quality of Life and Health: A Spatial Approach. Economic Development Quarterly, 17(3), 280-93 Cairns, J. (1992) Future discounting: health, wealth and time preference, Project Appraisal 7(1), 31-40 Camerer, C. (1993) Individual decision-making, in Handbook of Experimental Economics, Kagel J and Roth A, Princeton University Press Clark, A. (2003a) Unemployment as a Social Norm: Psychological Evidence from Panel Data. Journal of Labor Economics, 21(2), 323-51 Clark, A. (2003b) Inequality-aversion and income mobility: A direct test. June 2003 Clark, A. and Lelkes, O. (2005) Deliver us from Evil: Religion as Insurance. PSE, Paris, November Clark, A. and Oswald, A. (1994) Unhappiness and Unemployment. The Economic Journal, 104 (424), 648-59 Cohen, A. (2002) The importance of spirituality in well-being for Jews and Christians. Journal of Happiness Studies, 3, 287-310 Cowell, F.A., Gardiner, K. A. (2000) Welfare Weights. OFT Economic Research Paper Office of Fair Trading 202 Crisp, R. (2003) Hedonism reconsidered. Paper for Minnesota Interdisciplinary Workshop on Well-Being, 24-26 October 2003 Csikszentmihalyi, M. (1990) Flow: The Psychology of Optimal Experience. New York: Harper Collins Cummins, R.A. (1993) The Comprehensive Quality of Life Scale: Adult. 4th Edition (ComQol-A4), School of Psychology, Deakin University, Melbourne Cummins, R.A. & Gullone, E. (2000). Why we should not use 5-point Likert scales: The case for subjective quality of life measurement. Proceedings, Second International Conference on Quality of Life in Cities (pp. 74-93). Singapore: National University of Singapore. http://acqol.deakin.edu.au/instruments/PWI/2000QOL_in_CitiesLikert_Scales-ss.doc Cummins, R.A., Eckersley, R., Kai, L.S., Okerstrom, E., Davern, M. and Woener, J. (2004), Australian Unity Well-being Index Survey 11, report 11.0, The Wellbeing of Australians – Personal Financial Debt, Australian Centre on Quality of Life Cysne, R.P. (2004) On the positive correlation between income inequality and unemployment, Graduate School of Economics, Getulio Vargas Foundation (Brazil), Working Paper, No 561 Day, B. (2003). What price peace? A comprehensive approach to the specification and estimation of hedonic housing price models, CSERGE Working Paper 8, University of East Anglia Dehejia, R., DeLeire T., and Luttmer, E. F. P., (2005) Insuring consumption and happiness through religious organizations. RWP05-047, KSG Faculty

89

References Research Working Paper Series, John F. Kennedy School of Government, Harvard University DeNeve K.M. & Cooper H. (1998) The happy personality: A meta-analysis of 137 personality traits and subjective well-being. Psychological Bulletin, 124,197229 Dennerstein, L., Lehert, P., & Guthrie, J. (2002) The effects of the menopausal transition and biopsychosocial factors on well-being. Archives of Women's Mental Health, 5, 15-22 Diener, E. & Biswas-Diener, R. (2002) Will money increase subjective well-being? A literature review and guide to needed research. Social Indicators Research, 57, 119-69 Diener, E., Emmons, R.A., Larson, R.J., & Griffin, S. (1985) The satisfaction with life scale. Journal of Personality Assessment, 49, 71-75 Diener, E., Lucas, R.E., Oishi, S. & Suh, E. (2002) Looking up and looking down: Weighting good and bad information in life satisfaction judgements, Personality and Social Psychology Bulletin, 28(4), 437-445 Diener, E. Oishi, S. & Lucas, R. E. (2003) Personality, culture and subjective wellbeing: Emotional and cognitive evaluations of life. Annual Review of Psychology, 54, 403-25 Diener, E. and Seligman, M. E. P. (2002) Very happy people. Psychological Science, 13(1), 81-4 Diener, E. and Seligman, M. E. P. (2004) Beyond money. Toward and economy of well-being. Psychological Science in the Public Interest, 5(1) Diener, E. Suh, E.M., Lucas, R.E. & Smith, H.L. (1999) Subjective well-being: Three decades of progress. Psychological review, 125, 276-302 Di Teller, R, Haisken-De New, J. and MacCulloch, R. (2005) Happiness Adaptation to Income and to Status in an Individual Panel. Di Teller, R., MacCulloch, R. and Oswald, A. (2001) Preferences over inflation and unemployment. Evidence from surveys of happiness. The American Economic Review, 91(1), 335-341 Di Teller, R., MacCulloch, R. and Oswald, A. (2003) The macroeconomics of happiness. The Review of Economics and Statistics, 85(4), 809-27 Dockery, A.M. (2003) Happiness, life satisfaction and the role of work: Evidence from two Australian surveys, Working Paper No. 03.10, School of Economics and Finance, Curtin Business School, Curtin University of Technology Dolan, P. (1999) Whose preferences count?, Medical Decision Making, 19, 482-486 Dolan, P. (2000) The measurement of health-related quality of life for use in resource allocation decisions in health care. In A.J. Culver and J.P. Newhouse (Eds.) Handbook of Health Economics (Vol, 1, pp. 1724-1760). Amsterdam: Elsevier Science B.V Dolan, P. and Kahneman, D. (2006) Interpretations of utility and their implications for the valuation of health, resubmitted to Economic Journal Dorn, D., Fischer J.A., Kirchgassner, G., Souta-Poza, A. (2005) Is it culture of democracy? The impact of democracy, income and culture on happiness. University of St Gallen, Department of Economics, Discussion paper no. 2005-12 Duesenberry, J. S. (1949) Income, saving and the theory of consumer behaviour. Cambridge, Harvard University Press Easterlin, R. (1995) Will raising the incomes of all increase the happiness of all, Journal of Economic Behaviour and Organization, 27, 35-47

90

References Easterlin, R. (2001) Income and happiness: towards a unified theory. The Economic Journal, 111, 465-84 Easterlin, R. (2005a) A puzzle for adaptive theory, Journal of Economic Behavior and Organization, 56, 513-521 Easterlin, R. (2005b) Diminishing marginal utility of income? Caveat emptor. Social Indicators Research, 70, 243-55 Fahey, T. and Smyth, E. (2004) Do subjective indicators measure welfare? Evidence from 33 European societies. European Societies, 6(1), 5-27 Feldman, F. (2002) The Good Life: A Defence of Attitudinal Hedonism. Philosophy and Phenomenological Research, 65, 604-628 Ferrer-i-Carbonell, A. (2005) Income and well-being: an empirical analysis of the comparison income effect. Journal of Public Economics, 89, 997-1019 Ferrer-i-Carbonell, A. and Frijters, P. (2004) How important is methodology for the estimates of the determinants of happiness. The Economic Journal, 114, 641659 Ferrer-i-Carbonell, A. and Gowdy, J. M. (2005) Environmental awareness and happiness. Rensselaer Working Papers in Economics, Number 0503, April 2005 Ferrer-i-Carbonell, A. and van Praag, B. M. S. (2002) The subjective costs of health losses due to chronic diseases. An alternative model for monetary appraisal. Health Economics, 11(8), 709-22 Ferriss, A. L. (2002) Religion and the Quality of Life. Journal of Happiness Studies, 3(3), 199-215 Flouri, E. (2004) Subjective well-being in midlife: The role of involvement of and closeness to parents in childhood. Journal of Happiness Studies, 5, 335-35 Frijters, P., Haisken-DeNew, J. P. and Shields M. (2004) Money does matter! Evidence from increasing real income and life satisfaction in East Germany following reunification. The American Economic Review, 94(3), 730-739 Frey, B. S., Luechinger, S. and Stutzer, A. (2004) Valuing Public Goods: The Life Satisfaction Approach. IEW Working Paper No. 184, University of Zurich. Frey, B. S. and Stutzer, A. (2000) Happiness, Economy and Institutions. The Economic Journal, 110, 918-38 Fujita, F. and Diener, E. (2005) Life Satisfaction Set Point: Stability and Change. Journal of Personality and Social Psychology, 88(1), 158–64 Gerdtham, U.G. and Johannesson, M. (2001) The relationship between happiness, health, and socio-economic factors: results based on Swedish microdata. Journal of Socio-Economics 30, 553-557 and Gerdtham, U.G. and Johannesson, M (1997) Working Paper No 207, Working Paper Series in Economics and Finance, Stockholm School of Economics Gerlach, K. and Stephan, G. (1996) A paper on unhappiness and unemployment in Germany. Economic Letters, 52(3), 325-30 Goldberg, D.P. et al. (1978) Manual of the General Health Questionnaire. Windsor, England, NFER Publishing Goldberg, D.P. and Williams, P. (1988) Users' Guide to the General Health Questionnaire. NFER-Nelson, Windsor Goodchild, M.E. & Duncan-Jones, P. (1985) Chronicity and the General Health Questionnaire. The British Journal of Psychiatry, 146, 55-61 Graham, C., Eggers, A., and Sukhtankar S. (2004) Does happiness pay? An exploration based on panel data from Russia. Journal of Economic Behavior and Organization, 55(3), 319-42

91

References Graham, C. and Felton, A. (2006) Inequality and Happiness: Insights from Latin America. Journal of Economic Inequality, 4(1), 107-122 Graham, C. and Pettinato, S. (2001) Happiness, markets and democracy: Latin America in Comparative Perspective. Journal of Happiness Studies, 2, 237-68 Greenfield, E.A. and Marks, N.F. (2004) Formal volunteering as a protective factor for older adult's psychological well-being. Journals of Gerontology: Series B: Psychological Sciences and Social Sciences, 59B, 258-64 Haddad L., and Kanbur R. (1990) How Serious is the Neglect of Intra-Household Inequality?. The Economic Journal, 100:866-881 Haller, M. and Hadler, M. (2006) How social relations and structures can produce happiness and unhappiness: an international comparative analysis. Social Indicators Research, 75, 169-216 Hann D, Winter K, Jacobsen P. (1999) Measurement of depressive symptoms in cancer patients: evaluation of the Center for Epidemiological Studies Depression Scale (CES-D). Journal of Psychosomatic Research, 46, 437-43 Harsanyi, J. C. (1996) Utilities, preferences, and substantive goods. Social Choice and Welfare, 14, 129-145 Hayo, B. (2004) Happiness in Eastern Europe. Mimeo, Philipps University Marburg, Number 12/2004 Hayo, B. and Seifert, W. (2003) Subjective economic well-being in Eastern Europe. Journal of Economic Psychology, 24, 329-48 Headey, B. & Headey, D. (2003) German reunification: Welfare gains and losses East and West. Social Indicators Research, 64, 107-38 Headey, B. and Wooden, M. (2004) The effects of wealth and income on subjective well-being and ill-being. The Economic Record, 80 Special Issue, S24-S33 Headey, B., Muffels, R., and Wooden, M. (2004) Money doesn’t buy happiness… or does it? A reconsideration based on the combined effects of wealth, income and consumption. IZA DP 1218, Institute for the Study of Labour, July 2004 Helliwell, J. F. (2003) How’s life? Combining individual and national variables to explain subjective well-being. Economic Modelling, 20, 331-360 Helliwell, J. F. (2006) Well-being, social capital and public policy: What’s new?, Economic Journal, 116, C34-C45 Helliwell, J. F. and Putnam, R. (2004) The social context of well-being. Phil Trans R Soc Lond, 359, 1435-1446 Hirst, M. (2003) Caring-related inequalities in psychological distress in Britain during the 1990s. Journal of Public Health Medicine, 25, 346-43 Hirst, M. (2005) Carer distress: A prospective, population-based study. Social Science & Medicine, 61, 697-708 HM Government (2005). Securing the Future - UK Government sustainable development strategy,. Defra, London. Hudson, J. (2006) Institutional trust and subjective well-being across the EU. Kyklos, 59(1), 43-62 Huta, V., Park, N., Peterson, C., Seligman, M.E.P., Pursuing Pleasure Versus Eudaimonia: Links with different aspects of well-being Hyde M., Wiggins R.D., Higgs P. and Blane D.B. (2003) A measure of quality of life in early old age: the theory, development and properties of a needs satisfaction model (CASP-19), Aging & Mental Health, 7(3), 186–194 Inglehart, R. and Klingemann H-D. (2000) Genes, Culture, Democracy and Happiness. In: Culture and Subjective Wellbeing, Diener, E. and Suh, E. M. (eds), MIT Press, Cambridge, Massachusetts

92

References International Wellbeing Group (2005) Personal Wellbeing Index. Melbourne: Australian Centre on Quality of Life, Deakin University Jones-Lee, M. (2002) An anomaly in the valuation of transport safety. Applied Health Economics and Health Policy, 1.1, 7-8 Johnson, W. and Krueger, R.F. (2006) How money buys happiness: Genetic and environmental processes linking finances and life satisfaction. Journal of Personality and Social Psychology, 90, 680-91 Kafka, G.J., and Kozma, A. (2002) The construct validity of Ryff's Scales of Psychological Well-Being (SPWB) and their relationship to measures of subjective well-being. Social Indicators Research, 57, 171-90 Kahneman, D. and Krueger, A.B. (2006) Developments in the measurement of subjective well-being, Journal of Economic Perspectives, in press Kahneman, D., Krueger, A.B., Schkade,D.A., Schwarz, N. and Stone, A.A., 2004a, A Survey Method for Characterizing Daily Life Experience: The Day Reconstruction Method. Science, 306(5702), p. 1776-80 Kahneman, D., Krueger, A.B., Schkade, D.A., Schwarz, N., & Stone, A.A. (2004b) A survey method for characterising daily life experience: The day reconstruction method (DRM). American Economic Review, 94(2), 429-434 Kammann, R. & Flett, R. (1983) Affectometer 2: a scale to measure current level of general happiness. Australian Journal of Psychology, 82, 1007-1022 Kenny, C. (2005) Does development make you happy? Subjective wellbeing and economic growth in developing countries. Social Indicators Research, 73, 199-219 Keyes, C.L.M. (2000) Subjective change and its consequences for emotional wellbeing. Motivation & Emotion, 24, 67-84 Keyes, C.L.M., Schmotkin, D., and Ryff, C.D. (2002) Optimizing well-being: The empirical encounter of two traditions. Journal of Personality and Social Psychology, 82, 1007-22 Kim, H.K. and McKenry, P.C. (2002) The relationship between marriage and psychological well-being: A longitudinal analysis. Journal of Family Issues, 23, 885-911 Klar, Y. & Giladi, E.E. (1999) Are most people happier than their peers, or are they just happy?. Personality and Social Psychology Bulletin, 25, 585-594 Koropeckyj-Cox, T. (2002) Beyond parental status: Psychological well-being in middle and old age. Journal of Marriage and the Family, 64, 957-71 Layard, R. (2005) Happiness. Lessons from a new science, Allen Lane, Penguin Books Lee, G.R. and Bulanda, J.R. (2005) Change and consistency in the relation of marital status to personal happiness. Marriage and Family Review, 38, 69-84 Lee, G.R., DeMaris, A., Bavin, S. and Sullivan, R. (2001) Gender differences in the depressive effect of widowhood in later life. Journals of Gerontology: Series B: Psychological Sciences and Social Sciences, 56B, 56-61 Lelkes, O. (2006) Knowing what is good for you. Empirical analysis of personal preferences and the “objective good”. The Journal of Socio-Economics, 35, 285-307 Leigh, A. and Wolfers, J. (2005) Happiness and the Human Development Index: Australia is not a paradox, 7 December draft Lepper, H.S. (1998) Use of other-reports to validate subjective well-being measures. Social Indicators Research, 44, 367-379

93

References Louis, V.V. & Zhao, S. (2002) Effects of family structure, family SES, and adulthood experiences on life satisfaction. Journal of Family Issues, 23, 986-1005 Lucas, R.E. (2005) Time does not heal all wounds: A longitudinal study of reaction and adaptation to divorce. Psychological Science, 16, 945-50 Lucas, R.E., Clark, A.E., Georgellis, Y., and Diener, E. (2003) Reexamining adaptation and the set point model of happiness: Reactions to changes in marital status, Journal of Personality and Social Psychology, 84 (3), 527-39 Lucas, R.E., Clark, A.E., Georgellis, Y. and Diener, E. (2004) Unemployment alters the set point for life satisfaction. Psychological Science, 15 (1), 8-13 Lucas, R.E. and Fujita, F. (2000) Factors influencing the relation between extraversion and pleasant affect. Journal of Personality and Social Psychology, 79, 1039-56 Luttmer, E.F.P. (2004) Neighbors as Negatives: Relative Earnings and Well-Being. July 22 Macdonald, M. and Douthitt, R.A. (1992) Consumption Theories and Consumers’ Assessments of Subjective Well-Being, 26(2), 243-61 Lykken, D. and Tellegen, A. (1996) Happiness is a stochastic phenomenon, Psychological Science 7(3), 186-189 Magdol, L. (2002) Is moving gendered? The effects of residential mobility on the psychological well-being of men and women. Sex Roles, 47, 553-60 Mallam, S., Angle, H., Wimbush, E. and Fraser, E. (2004) Health Education Survey 1996-2003. NHS Health Scotland Marks, G.N. and Flemming, N. (1999) Influences and consequences of well-being among Australian young people: 1980-1995. Social Indicators Research, 46, 301-23 Marks, N.F., Lambert, J.D. and Choi, H. (2002) Transitions to caregiving, gender, and psychological well-being: A prospective US national study. Journal of Marriage and the Family, 64, 657-67 Maslow, A. (1954) Motivation and Personality. London: Harper & Row Martin, M. & Westerhof, G.J. (2003) Do you have to have them or should you believe you have them? Resources, their appraisal, and well-being in adulthood. Journal of Adult Development, 10, 99-112 Max-Neef, M. A. (1992) Development and human needs. Real life economics: understanding wealth creation. P. Ekins and M. Max-Neef. London, Routledge: 197-214 McBride, M. (2001) Relative income effects on subjective well-being in the crosssection. Journal of Economic Behaviour and Organization, 45, 251-78 McClements, L. D. (1977) Equivalence scales for children. Journal of Public Economics 8(2), 191-210 Meier, S. and Stutzer, A. (2006) Is Volunteering Rewarding in Itself? March 27 Michalos, A.C. (1985) Multiple discrepancies theory (MDT). Social Indicators Research, 16, 347-413 Michalos, A.C. (2004) Social indicators research and health-related quality of life research. Social Indicators Research, 65, 27–72 Michalos, A.C., Zumbo, B.D., Hubley, A. (2000) Health and the Quality of Life. Social Indicators Research, 51, 245-286 Moore, J.C., Stinson, L.L., and Welniak, E.J. (2000) Income Measurement Error in Surveys: A Review. Journal of Official Statistics, 16(4), 331-61 Nettle, D. (2005) Happiness: The Science Behind Your Smile. Oxford: Oxford University Press

94

References O’Connell, M. (2004) Fairly satisfied: Economic equality, wealth and satisfaction. Journal of Economic Psychology, 25, 297-305 Oswald, A. and Powdthavee, N. (2005) Does happiness adapt? A longitudinal study of disability with implications for economists and judges, September 30 Parfit, D., (1984) Reasons and Persons. Oxford: Oxford University Press Parikh, R.M., Eden, D.T., Price, T.R., Robinson, R.G. (1988) The sensitivity and specificity of the Center for Epidemiologic Studies Depression Scale in screening for post-stroke depression. International Journal of Psychiatry and Medicine, 18, 169-81 Parducci, A. (1984) Value Judgements: Toward a relational theory of happiness. In J. R. Eiser (ed.). Attitudinal Judgement. New York: Springer-Verlag Pavot, W. & Diener, E. (1993) Review of the satisfaction with life scale. Psychological Assessment, 5, 164-172 Pavot, W., Diener, E. & Suh, E.M. (1998) The temporal satisfaction with life scale. Journal of Personality Assessment, 70, 340-354 Peterson, C., Park, N. & Seligman, M.E. (2005) Orientations to happiness and life satisfaction: The full life versus the empty life. Journal of Happiness Studies, 6, 25-41 Pichler, F. (2006) Subjective quality of life of young Europeans. Feeling happy but who knows why? Social Indicators Research, 75, 419-444 Radloff, L. S., (1977) The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401 Rawls, J. (1971) A Theory of Justice. Cambridge MA: Harvard University Press Rehdanz, K. and Maddison, D. (2003) Climate and Happiness, Centre for Marine and Climate Research, Hamburg, Working Paper FNU-20, April 2003 Ritchey, L. H., Ritchey, P.N., and Dietz, B.E. (2001) Clarifying the measurement of activity. Activities, Ageing & Adaptation, 26, 1-21 Ryan, R. M., & Deci, E. L. (2000) Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American Psychologist, 55, 68-78 Ryff, C.D. (1989) Happiness is everything, or is it? Explorations on the meaning of psychological well-being. Journal of Personality and Social Psychology, 57, 1069-1081 Ryff, C.D. & Keyes, C.L.M. (1995) The structure of psychological well-being revisited. Journal of Personality Social Psychology, 69, 719-729 Sandvik, E., Diener, E. & Seidlitz, L. (1993) Subjective well-being: The convergence and stability of self-report and non-self-report measures. Journal of Personality, 61, 317-42 Schwarz, N. & Strack, F. (1999) Reports of subjective well-being: Judgemental processes and their methodological implications. In Kahneman, D., Diener, E. & Schwarz, N. (Eds.). Well-being. The foundations of hedonic psychology. New York. Russell Sage Foundation Seligman, M.E.P. (2002) Authentic Happiness. New York: Free Press. Schoon, I., Hansson, L., and Salmela-Aro, K. (2005) Combining work and family life: Life satisfaction among married and divorced men and women in Estonia, Finland and the UK. European Psychologist, 10, 309-19 Senik, C. (2004) When information dominates comparison Learning from Russian subjective panel data. Journal of Public Economics, 88, 2099–123

95

References Shields, M. and Wheatley Price, S. (2005) Exploring the economic and social determinants of psychological well-being and perceived social support in England. Journal Royal Statistical Society Part 3, 513-37 Smith, K. (2003) Individual welfare in the Soviet Union. Social Indicators Research, 64, 75-105 Stutzer, A. (2004) The role of income aspirations in individual happiness. Journal of Economic Behaviour and Organisation, 54, 89-109 Stutzer, A. and Frey, B.S. (2005) Stress That Doesn’t Pay: The Commuting Paradox. University of Zurich, August 16 Sumner, L.W. (1995) The subjectivity of welfare. Ethics, 105,4, 764-790 Theodossiou, I. (1998) The effects of low-pay and unemployment on psychological well-being: a logistic regression approach. Journal of Health Economics, 17, 85-104 Thoits P.A., and Hewitt, L.N. (2001) Volunteering work and well-being. Journal of Health and Social Behaviour, 42, 115-31 van den Berg, B. and Ferrer-i-Carbonell, A. (2005) The well-being of informal caregivers: A monetary valuation of informal care, paper for iHEA 2005, Barcelona, Spain van Praag, B.M.S. (2003) The anatomy of subjective well-being, Journal of Economic Behaviour and Organization, 51:29-49 van Praag, B.M.S. and Baarsma, B.E. (2005) Using happiness surveys to value intangibles: The case of airport noise. The Economic Journal, 115, 224-46 van Praag, B.M.S., Frijters, P. and Ferrer-i-Carbonell, A. (2003) The anatomy of subjective well-being. Journal of Economic Behaviour and Organization, 51, 29-49 Watson, D., Clark, L.A. & Tellegen, A. (1988) Development and validation of brief measures of positive and negative affect: The PANAS scale. Journal of Personality and Social Psychology, 54, 1063-1070 Weinzierl, M. (2005) Estimating a relative utility function, Harvard University, July 8 draft Welsch, H. (2002) Preferences over prosperity and pollution: Environmental valuation based on happiness studies. Kyklos, 55, 473-494 Welsch, H. (2003) Environment and happiness. Valuation of air pollution in ten European countries. German Institute for Economic Research, DIW Berlin, Discussion Papers Number 356, July Wiggins, R.D., Higgs, P.F.D, Hyde, M. & Blane, D.B. (2004) Quality of life in the third age: Key predictors of the CASP-19 measure. Ageing & Society, 24, 639708 Wildman, J. and Jones, A. (2002) Is it absolute income or relative deprivation that leads to poor psychological well being? A test based on individual-level longitudinal data Wolfers, J. (2003) Is business cycle volatility costly? Evidence from surveys of subjective well-being. International Finance, 6(1), 1-26

96

Appendix A: Description of Measures

APPENDIX A: DESCRIPTION OF MEASURES The tables in this appendix present more details about the specific measures reviewed in Section 3. For ease of referencing, we present them in alphabetical order rather than the order in which they appear in the text (which is related to type of measure).

Table A.1: Affectometer 2 Scale Measure:

Affectometer 2

Instructions

"Please say how often you have thought or felt each of the following over the past two weeks"

Response scale

5 point response scale: 0 = ‘not at all’ , 1 = ‘occasionally’, 2 = ‘some of the time’, 3 = ‘often’ or 4 = ‘all of the time’.

Items

40 items: 20 statements & 20 adjectives. Half are positive and half negative. Positive statements: My life is on the right track My future looks good I like myself I can handle any problems that come up I feel loved and trusted I feel close to people around me I can do whatever I want to do I have energy to spare I have been smiling and laughing a lot I have been thinking clearly and creatively Negative statements: I wish I could change some part of my life I feel as if the best years of my life are over I feel like a failure I have been left alone when I don’t want to be I have lost interest in other people and don’t care about them My life seems stuck in a rut I can’t be bothered doing anything Nothing seems very much fun anymore My thoughts have been going round in useless circles Positive adjectives: Satisfied, Optimistic, Useful, Confident, Understood, Loving, Free-and-easy, Enthusiastic, Good natured, Clear headed. Negative adjectives: Discontented, Hopeless, Insignificant, Helpless, Lonely, Withdrawn, Tense, Depressed, Impatient, Confused.

Scoring

Two sub-scales are calculated, one based on the negative items and one on the positive items. Overall well-being is calculated by subtracting the negative from the positive items.

97

Appendix A: Description of Measures

Table A.2: CASP 19 Scale Measure:

Instructions

Items

Response scale

Scoring

CASP 19

“Here is a list of statements that people have used to describe their lives or how they feel. We would like to know how often, if at all, do you think this applies to you:” (in the CAPS, not CASP order) Control My age prevents me from doing the things I would like to.* I feel that what happens to me is out my control.* I feel free to plan for the future. I feel left out of things.* Autonomy I can do the things I want to do. Family responsibilities prevent me from doing things I want to do.* I feel that I can please myself what I can do. My health stops me from doing the things I want to do.* Shortage of money stops me from doing the things I want to do.* Pleasure I look forward to each day. I feel that my life has meaning. I enjoy the things that I do. I enjoy being in the company of others. On balance, I look back on my life with a sense of happiness. Self-realization I feel full of energy these days. I choose to do things that I have never done before. I feel satisfied with the way my life has turned out. I feel that life is full of opportunities. I feel that the future looks good for me. 1 Often 2 Sometimes 3 Not often 4 Never * items = reversed Coded Sum of all items (0 to 57) = “Quality of life in old age.”

98

Appendix A: Description of Measures

Table A.3: Centre for Epidemiological Studies Depression (CES-D) Scale Measure:

Instructions

Items

Response scale

Scoring

Centre for Epidemiological Studies Depression Scale (CES-D) Below is a list of some of the ways you may have felt or behaved. Please indicate how often you have felt this way during the past week by checking the appropriate space. 1. I was bothered by things that usually don't bother me. N 2. I did not feel like eating; my appetite was poor. N 3. I felt that I could not shake off the blues even with help from my family or friends. N 4. I felt that I was just as good as other people.* 5. I had trouble keeping my mind on what I was doing. N 6. I felt depressed. N 7. I felt that everything I did was an effort. N 8. I felt hopeful about the future.* 9. I thought my life had been a failure. 10. I felt fearful. N 11. My sleep was restless. N 12. I was happy.* 13. I talked less than usual. N 14. I felt lonely. N 15. People were unfriendly. 16. I enjoyed life.* 17. I had crying spells. 18. I felt sad. N 19. I felt that people disliked me. 20. I could not get going. N (0) Rarely or none of the time (Less than 1 day) (1) Some of a Little of the Time (1-2 days) (2) Occasionally or a Moderate Amount of the Time (3-4 days) (3) Most or All of the Time (5-7 days) * Items reverse coded. . Sum of 20 items i.e. 0-60. 22 or higher indicates probable Major Depression 15-21 indicates need for more assessment and treatment for Mild to Moderate Depression 15 or less is not indicative of depression About 20% of the general population will score > 15. N

= Items included in NSHF. The NSHF used only 12 items and the response scale was 0-7 (i.e. no. of days last week.). There are also shorter versions with 8 or 4 items.

99

Appendix A: Description of Measures

Table A.4: Day Reconstruction Method (DRM) Scales Measure:

Day Reconstruction Method (DRM)

Instructions

The overall instructions are to write a diary about "yesterday" and evaluate the various episodes (roughly about an hour long) in terms of emotions felt. "How did you feel during this episode? Please rate each feeling on the scale given."

Response scale

0 "not at all" 6 "Very much"

Items

Impatient for it to end*; Happy; Frustrated/annoyed*; Depressed/Blue*; Worried/anxious*, Enjoying myself; Tired*; Stressed*

Scoring

* = negative items. Episodes where the mean of negative items is higher than positive items reflects an instant of "bad time". The number of bad time episodes during an entire day is used to construct a "U-Index". The more time spent in negative states the lower the wellbeing.

100

Appendix A: Description of Measures

Table A.5: General Health Questionnaire (GHQ) Scale Measure:

General Health Questionnaire (GHQ)

Instructions

"We would like to know how your health has been in general, over the past few weeks. Please answer the following questions by circling the number that best applies to you. Have you recently…"

Response scale

There are three basic response formats and the items using each format are shown below. Some response scales also use specific words such as 'able' in particular response scales (e.g. 'Much less able than usual').

Items

Been able to concentrate on whatever you are doing? - (Better/same/Less /Much less) than usual. Lost much sleep over worry? (R) Felt constantly under strain? (R) Felt that you couldn't overcome your difficulties? (R) Been feeling unhappy and depressed? (R) Been losing self-confidence in yourself? (R) Been thinking of yourself as a worthless person? (R) - (Not at all/No more/Rather more /Much more) than usual. Felt that you were playing a useful part in things? Felt capable of making decisions about things? Been able to enjoy your normal day-to-day activities? Been able to face up to your problems? Been feeling reasonably happy, all things considered? - (More so/Same as/Less/Much less) than usual.

Scoring

R (Reverse coded negative items) Full Scale = 0 - 36. The threshold (in terms of the number of symptoms identified) for determining ‘caseness’ varies according to the aims of the individual study. However, for the GHQ12 the threshold for identification of cases usually varies between four and six (Goldberg & Williams, 1988).

101

Appendix A: Description of Measures

Table A.6: Happiness/life satisfaction Single-Item Measures Survey

American's Changing Lives (ACL)

Details

US Multistage stratified area probability sample. Response rate (67%) Wave 1 1986 - 3,617 Wave 2 1989 - 2,867 Wave 3 1994

Australian (HILDA)

National probability interview sample. Wave 1 2001 had 7,682 households with 13,969 successful interviews.

Questions

Response scale in order of presentation. (Most have a ‘don’t know’ option).

Comments

"Now thinking about your life as a whole. How satisfied are you with it? Are you....?"

"Completely satisfied" "Very satisfied" "Somewhat satisfied" "Not at all satisfied"

Note positive bias of response scale compared to Euro & Latino barometers

"Strongly agree" "Agree" "Disagree" "Strongly disagree"

Counterfactual comparison

"Very happy" "Pretty happy" "Not too happy"

General Social survey question

“The more satisfied you are, the higher the number you should pick. The less satisfied you are, the lower the number.”

No explicit anchors used here.

(Wave 1) "My life could be happier than it is right now" (Wave 2) "Taking all things together, how would say things are these days? Would you say you were....?"

“All things considered, how satisfied are you with your life?“

Wave 2 interviews with 13,041, 12,000 from wave 1

British Household Panel Survey (BHPS)

Began in 1991 and is a multipurpose study following the same representative sample of individuals. It is household-based, interviewing every adult member of sampled households. Wave 1 consists of some 5,500 households & 10,300 individuals. Samples from Wales, Scotland Northern Ireland added later.

“0 – 10” "How satisfied are you with your life overall?"

1 = "not satisfied at all " 7 = "completely satisfied"

Would you say that you are more satisfied with life, less Satisfied, or feel about the same as you did a year ago?

"More satisfied" "Less satisfied " "About the same"

102

UK data

Temporal comparison

Appendix A: Description of Measures

Canadian General Social Survey (CGSS)

Established in 1985, conducts telephone surveys from a sample selected across the 10 provinces. Until 1998, the sample size was about 10,000. This was increased in 1999 to 25,000.

Introduced in Cycle 12.

"Very happy?" "Somewhat happy?" "Somewhat unhappy?" "Very unhappy?"

Note the use of a no opinion option but no middle neither/nor.

“I am going to ask you to rate certain areas of your life. Please rate your feelings about them” (Including) a) Your life as a whole right now? b) Your self-esteem

"Very satisfied" "Somewhat satisfied" "Somewhat dissatisfied" "Very dissatisfied"

Note the one item self-esteem questions.

Presently, would you describe yourself as...

Eurobarometer

300,000 people in 12 European countries. Interviews are one to one in people’s homes.

"On the whole, are you ........ or......... with the life you lead?"

"Very satisfied" (4) "Fairly satisfied" (3) "Not very satisfied" (2) "Not at all satisfied" (1)

European Social (Values) Survey (ESS)

Nationally representative cross section in over 20 European countries.

"All things considered, how satisfied are you with your life as a whole nowadays?"

1 “Dissatisfied” - 10 “Satisfied”

German SocioEconomic Panel Survey (GSOEP)

Hungarian Household Panel Survey (Run by Tarki)(HHPS)

Households selected using multistage random sampling, all members of household asked to participate. Annual face to face interviews. The entire sample is over 24,000 respondents who participated in a least one of the 1 to 15 waves.

Between 1991-1997, a nationwide sample of 2600 households was surveyed on a yearly basis.

UK data

(Translation issue) "How satisfied are you at present with your life as a whole?" "How happy are you at present with your life as a whole?"

Please tell me to what extent you are satisfied with each of the following parts of your life. (Including) a) Your career, the way your life has worked out. b) your standard of living

103

0 ("Completely dissatisfied") 10 ("Completely satisfied")

Some say "how satisfied" (e.g. van Praag et al., 2003).

0 ("Completely unhappy") 10 ("Completely happy")

Some: "how happy" (e.g. Winkelmann, 2004).

0 ("Not at all satisfied") - 10 ("Fully satisfied")

Note how career is part of the way life in general has worked out. Do people using this survey make this clear?

Appendix A: Description of Measures Annual programme of crossnational collaboration on surveys covering topics important for social science research. 41 member countries.

"If you were to consider your life in general these days, how happy or unhappy would you say you are, on the whole?"

4 "very happy" 3 " fairly happy" 2" not very happy" 1"not at all happy"

17spanish speaking countries (1997-2000) 1000 interviews per country by MORI, not nationally representative in all countries.

"How satisfied are you with your life?"

"Not at all " (1) "Somewhat " (2) "Satisfied" (3) "Very" (4)

Note how the first anchor is negative unlike GSS, Eurobarometer

Midlife in the US (MIDUS)

US National probability sample, random digit dialling. English speaking. Over sampling of 6574 years.

"Please rate your life overall these days on a scale from 0 to 10 where 0 is the worst possible life overall and 10 is the best possible life overall."

0 = Worst possible life overall 10 = Best possible life overall

(Cantril's self-anchoring scale)

Minnesota Twin Registry

A monitoring of twins born in the US from 1936 to 1955

"Taking the good with the bad, how happy and contented are you on average now, compared with other people?"

International Social Survey Programme (ISSP)

Latinobarometer

National Child Development Survey (NCDS), US

Cohort of people born in Britain, from 03/03/58 to 09/03/58. Most recent data 2000, age 42. Of the initial 17,414 individuals 11,419 in 2000.

“How satisfied are you with your life so far?”

1 ("Lowest 5%") 2 ("Lowest 30 %") 3 ("Middle") 4 ("Upper 30%") 5 ("Top 5%") 0 ("Completely dissatisfied") - 10 ("Completely satisfied")

“How satisfied were you with your life 5 years ago?” “How satisfied do you expect to be with your life in 5 years time?”

National Survey of Families and Households (NSFH), US

Representative sample living in English/Spanish speaking homes. 1987-88 (Wave 1) 19921994 (Wave 2 ) 10,000 in panel

“Taking things all together, How would you say things are these days?"

104

1 ("Very unhappy") 7 ("Very happy")

Social comparison

Appendix A: Description of Measures

Personal Well-being Index (PWI), Australia

Social Capital Community Benchmark Survey (SCBS), US

International Wellbeing Group (2005). National sample of 3,000 r and representative samples in 40 communities nationwide (across 29 states) covering an additional 26,200 respondents.

“Thinking about your own life and personal circumstances, how satisfied are you with your life as a whole?”

“All things considered, would you say you are …. “

0 "Completely dissatisfied" 5 "Neutral" 10 "Completely satisfied" 1 Very happy 2 Happy 3 Not very Happy 4 Not happy at all

Russian Longitudinal Monitoring Survey (RLMS)

An approximate probability sample (restricted by practical limitations) with households in 20 regions in Russia 1995-1998. Phase 1 = 6,334 households (17,154 individuals)

e.g. Wave 8 “To what extent are you satisfied with your life in general at the present time?”

1 Fully satisfied 2 Rather satisfied 3 Both yes and no 4 Less than satisfied 5 Not at all satisfied

Swedish Level of Living Survey (LNU)

Taken several times between 1968-1991. In 1991 there were 6, 773 individuals

"We have now been through a lot of questions about your living conditions in different areas. How do you yourself view your own conditions? By and large, do you think that your situation is:"

1 very good 2 rather good 3 neither good nor bad 4 rather bad 5 very bad

Swiss Household Panel Survey (SHPS)

1999-2004, roughly 4,000 households and 7,000 individuals per year.

“In general how satisfied are you with your life?”

0 ("not at all satisfied ") - 10 ("completely satisfied")

US General Social Survey (USGSS)

30,000 individuals over period 1972 to 1994

"Taken all together how would you say things are these days? Would you say you are ...?" "All things considered, how satisfied are you with your life as a whole these days?" (2005)

"Very happy" (3) "Pretty happy" (2) "Not too happy" (1) 1 “dissatisfied” 10 “satisfied”

World Values Survey (WVS)

Grew out of the European Values Survey group (EVS). Nationally representative UK samples of around 1,000 individuals in 1998 & 1999 collected by Mori and Gallup

“Taken all things together, would you say you are…”

105

Technically this is not part of the PWI (see above) but is used to investigate the relative importance of each domain on total well-being.

1 Very Happy 2 Quite happy 3 not very happy 4 not at all happy

One of the more "objective" measures in that it asks about living conditions - rather than feelings etc.

Appendix A: Description of Measures

Table A.7: Orientations to Happiness (OTH) Scale Measure:

Orientations to Happiness Scale

Instructions

"Please say how much you feel each of the following statement applies to you personally...."

Response scale

1 "Very much unlike me" to 5 "Very much like me"

Items

Life of meaning o "My life serves a higher purpose." o "In choosing what to do, I always take into account whether it will benefit other people." o "I have a responsibility to make the world a better place." o "My life has a lasting meaning." o "What I do matters to society." o "I have spent a lot of time thinking about what life means and how I fit into its big picture." Life of Pleasure o "Life is too short to postpone the pleasures it can provide." o "I go out of my way to feel euphoric." o "In choosing what to do, I always take into account whether it will be pleasurable." o "I agree with this statement: "Life is too short - eat dessert first." o "I love to do the things that excite my senses." o "For me, the good life is the pleasurable life"

Scoring

Life of engagement o "Regardless of what I am doing, time passes very quickly." o "I seek out situations that challenge my skills and abilities." o "Whether at work or play, I am usually "in a zone" and not conscious of myself." o "I am always very absorbed in what I do." o "In choosing what to do, I always take into account whether I can lose myself in it." o "I am rarely distracted by what is going on around me." Scale scores = average of 6 items.

106

Appendix A: Description of Measures

Table A.8: Personal Well-Being (PWI) Scale Measure:

Personal Well-Being Index (PWI) International Wellbeing Group (2005).

Instructions

Items

"The following questions ask how satisfied you feel, on a scale from zero to 10. Zero means you feel completely dissatisfied. 10 means you feel completely satisfied. And the middle of the scale is 5, which means you feel neutral, neither satisfied nor dissatisfied.” “How satisfied are you with…… ?” 1) your standard of living? 2) your health? 3) what you are achieving in life ? 4) your personal relationships? 5) how safe you feel? 6) feeling part of your community? 7) your future security?

Response scale

0 "Completely dissatisfied" 5 "Neutral" 10 "Completely satisfied"

Scoring

For our purposes the most important score is the aggregate of all 7 domains (The PWI). For ease of interpretation the means are multiplied by ten to give a 0-100 scale.

107

Appendix A: Description of Measures

Table A.9. Positive and Negative Affect Scales (PANAS) Measure:

Positive & negative affect scales

Instructions

"How much of the time during the past 30 days have you felt...."

Response scale

(1) "none of the time" to (5) "all of the time"

Items

Positive: Cheerful, in good spirits, extremely happy, calm and peaceful, satisfied and full of life Negative: So sad nothing could cheer me up, nervous, restless or fidgety, hopeless, that everything was an effort, worthless

Scoring

Summed scores for each scale (6-30).

108

Appendix A: Description of Measures

Table A.10: Psychological Well-Being Scales (PWBS) Measure:

Psychological Well-Being Scales

Instructions

"To what extent do you think the following statements are true of you?"

Response scale

1 (strongly disagree) - 6 (strongly agree)

Items

Scoring

Autonomy o I tend to be influenced by people with strong opinions.* o I have confidence in my opinions, even if they are different from the way most other people think. o I judge myself by what I think is important, not by the values of what others think is important. Positive Relations with Others o Maintaining close relationships has been difficult and frustrating for me.* o I have not experienced many warm and trusting relationships with others.* o People would describe me as a giving person, willing to share my time with others. Purpose in Life o I live life one day at a time and don’t really think about the future.* o Some people wander aimlessly through life, but I am not one of them. o I sometimes feel as if I've done all there is to do in life.* Self-Acceptance o I like most parts of my personality. o When I look at the story of my life, I am pleased how things have turned out. o In many ways, I feel disappointed about my achievements in life.* Environmental Mastery o The demands of everyday life often get me down.* o In general, I feel I am in charge of the situation in which I live. o I am quite good at managing the many responsibilities of my daily life. Personal Growth o I gave up trying to make big improvements or changes in my life a long time ago.* o I think it is important to have new experiences that challenge how you think about yourself and the world. o For me, life has been a continuous process of learning, changing, and growth. * Reverse coded negative items Summed scores for all multi-item scales.

109

Appendix A: Description of Measures

Table A.11: Satisfaction With Life Scale (SWLS) Measure:

Satisfaction With Life Scale

Instructions

"Below are five statements that you may agree or disagree with. Using the scale below indicate your agreement with each item by placing the appropriate number on the line preceding that item. Please be open and honest in your responding."

Response scale

1 (strongly disagree) - 7 scale (strongly agree)

Items

o o o o o

Scoring

35 - 31 Extremely satisfied; 26 - 30 Satisfied; 21 - 25 Slightly satisfied; 20 Neutral; 15 - 19 Slightly dissatisfied; 10 - 14 Dissatisfied 5 - 9 Extremely dissatisfied

_ _ _ _ _

In most ways my life is close to my ideal. The conditions of my life are excellent. I am satisfied with my life. So far I have gotten the important things I want in life. If I could live my life over, I would change almost nothing.

110

Appendix B: Summaries

APPENDIX B: SUMMARIES Here are the summaries of the detailed reviews of each of the 153 studies in the review, listed in the order that the categories appear in the main text. B.1. Income - B.1 Income B.2. Personal characteristics – who we are, our genetic makeup - B.2.1 Age - B.2.2 Gender - B.2.3 Ethnicity - B.2.4 Personality - B.2.5 Physical Characteristics B.3. Socially developed characteristics – our human and physical capital - B.3.1 Education - B.3.2 Health - B.3.3 Type of work - B.3.4 Unemployment B.4. How we spend our time - the work we do, and activities we engage in - B.4.1 Hours worked - B.4.2 Commuting - B.4.3.Housework - B.4.4 Caring for others - B.4.5 Community involvement and volunteering - B.4.6 Sleep - B.4.7 Exercise - B.4.8 Religious practice B.5. Attitudes and beliefs towards self/others/life – how we interpret the world - B.5.1 Attitudes towards our circumstances - B.5.2 Trust - B.5.3 Political persuasion - B.5.4 Religious beliefs B.6. Relationships – the way we connect with others - B.6.1 Marriage and intimate relationship - B.6.2 Having children - B.6.3 Seeing family and friends B.7. Economic, social and political environment – where we live - B.7.1 Income inequality - B.7.2 Unemployment rates - B.7.3 Inflation - B.7.4 Welfare and public insurance - B.7.5 Democracy - B.7.6 Climate and quality of natural environment - B.7.7 Security of local environment (crime rates/risk) - B.7.8 Urbanisation

111

Appendix B: Summaries

B.1. Income Table B.1: Income

Correlations and bivariate relationships between income and well-being Study Abbott et al (2004)

Country (data) India

Findings No relationship was found between income group and life satisfaction or positive and negative feelings. Correlation between income and life satisfaction of 0.45.

Biswas-Diener and Diener (2000)

India, Slums in Calcutta, n=83

Blanchflower and Oswald (2004)

UK (Eurobarometer), 1973-1998 & US (GSS), 1972-1998

Diener and Biswas- International (WVS II), Diener (2002), 19 countries Review

Diener and Seligman (2004)

International (WVS)

Christoph & Noll (2003)

Europe (Eurobarometer) 19912000 International, mostly from the US

Cummins (2000), Review

112

For the UK, from 1973-1998, controlling for demographic characteristics, the trend in life satisfaction and overall happiness is very slightly positive. For the US, from 1972-1998, controlling for demographic characteristics, the trend in overall happiness is not significant once marriage and employment variables are included. The trend is negative and significant for whites and for women, and positive and significant for blacks and men. Correlations between income and SWB ranged from 0.5 to 0.7. Life satisfaction is above neutral for 86% of the cases in high income countries compared with 69% in low income countries. Mean positive affect was 80% in high income countries compared with 56% for low income countries. Mean negative affect was 9% in high income countries compared with 27% for low income countries. “Thus, it can be said that higher income corresponds to modest differences in happiness, but it substantially reduces the risk of the rarer experience of unhappiness” (Diener and BiswasDiener 2002: 127) For nations with a GDP/capita over $10,000 the correlation between life satisfaction and GDP/capita was 0.08. Life satisfaction is higher in the highest income quintile than the lowest income quintile. The mean correlation between income and SWB (measured by subjective quality of life, life satisfaction, and overall happiness) for 9 low

Appendix B: Summaries

Cummins et al (2004)

Australia (Australian Unity Index), 2004

Easterlin (1995)

US (GSS), Europe, Japan.

Easterlin (2001)

US (GSS), 1994

113

income samples, published from 1973 to 1998, was 0.257 (SD 0.125) and the mean correlation of income and SWB (measured by positive affect, life satisfaction, satisfaction with standard of living, and overall happiness) for 24 average income samples, published from 1970 to 1999, was 0.135 (SD 0.071). Using a more subjective income measure where individuals categorised their own income according to a measure of relative income, and excluding studies with obvious confounders (such as unemployment and disability), gives a mean SWB for the poor (from 11 studies) of 61.9 (SD 7.3), for the medium income (from 11 studies) of 72.0 (SD 5.3) and the rich (from 6 studies) of 79.8 (2.8), which gives significant differences. Youngest and oldest groups are less influenced by income than middle age groups (aged 26-55). A stronger relationship between an index of domain satisfactions (PWI) and income is found for sole parents and those living alone and those who are separated. The relationship is also stronger for those at home full time or home-carers, compared to those who are working. The variability of PWI is also higher for those on low incomes. As income increases the frequency of reporting sad events decreases and frequency of happy events increases, up to an income of about AUS$60,000. Those who can not afford to pay off their credit card have lower PWI. Of those in the highest income group (> AUS$ 150,000) 13% fail to pay their credit card off, their PWI is 74.1 versus 77.2 for those who do pay off their card in this income group. In the US, from 1972 to 1991, there was a slight declining trend in overall happiness at a time when GDP/capita more than doubled. In Japan, from 1958 to 1987, SWB was stable despite per capita income growing 5 fold. In Europe, from 1973 to 1989, there was a minimal change in happiness levels despite GDP/capita rising by 50%. The lowest income group (< $10,000) had 16% who report they were very happy compared with 44% of the highest income group (> $75,000). Following birth cohorts for 24 years shows stable happiness despite rising income up to retirement and declining thereafter. However, those with high education, which is taken as a proxy for higher income, consistently have higher mean

Appendix B: Summaries

Easterlin (2005)

International (WVS)

Hagerty and Veenhoven (2003)

International (WDH), 1958-1996, 21 countries

Headey & Headey (2003)

Germany (GSOEP), 1997

Kenny (2005)

Tanzania

Leigh & Wolfers (2005)

International (WVS), 78 countries

happiness throughout their life span. For Japan, 1958-1987 and US 1941-59 and 19722000 overall happiness was not significantly related to the log of GNP/capita. Six countries experienced a positive significant correlation with GDP/capita and life satisfaction. One (Belgium) had a significantly negative correlation. On average a 1% increase in GDP/capita leads to a 0.004 increase (on a 10 point scale) for high income countries, 0.006 for middle income countries and 0.010 for low income countries. In West Germany life satisfaction (0 to 10 scale) is lower in the poorest income quintile (6.1) than for the richest quintile (7.4). In East Germany life satisfaction was generally lower, and lowest in the poorest income quintile (5.8), compared with (6.7) the highest. The percentage of respondents thinking themselves as very poor, poor, average, rich, or very rich was not related to the wealth of the area. Coefficient of 0.8 for log of GDP (t=8.3) on life satisfaction, and 0.08 (t=2.5) on overall happiness.

Correlations between income and well-being, controlling for other variables but not unobserved individual effects Study Alesina et al (2004)

Country (data) US (GSS), 19811996 & Europe (Eurobarometer), 1975-1992

Van den Berg & Ferrer-i-Carbonell (2005)

Netherlands, 2001/02

Bjornskov (2003)

International (WVS), 32 countries, 1993

Blanchflower & Oswald (2004)

Britain (Eurobarometer) & US (GSS)

Findings Household income is positively related to overall happiness in the US and life satisfaction in Europe. Controlling for income, being in the top two income quintiles also increase happiness/life satisfaction in both countries over the time period suggesting some role for relative income effects. Log of net family income is positively related to overall happiness, significantly so for an ordered logit model on responses to a 1 to 5 scale, but nonsignificantly for OLS model on responses to a 0-10 visual analogue scale of happiness at this moment. Controlling for income inequality, inflation and social capital, GNI per capita is not significant, but economic growth has a positive, significant relationship with life satisfaction. For GB, income quintiles are positively and monotonically related to life satisfaction and overall happiness. The coefficients on the 2nd and 3rd income quartiles are greater for women than men. For the US, household income per capita (estimated from mid points of categorical bands) has a positive 114

Appendix B: Summaries

Blanchflower & Oswald (2004)

US (GSS)

Borooah (2005)

Ireland (Survey West Belfast)

Brown et al (2005)

Britain (BHPS), 1995 & 2000

Bukenya et al (2003)

US (Survey West Virginia)

relationship with overall happiness, and the coefficients are similar for men and women and blacks and whites. The log of household income is also significant. The log of state income per capita is non significant, however, the ratio of household income per capita over state income per capita is significant even when household income per capita and a regional house price index (to control for state level costs of living), are included in the model. The largest relative effect comes from a ratio of own income to the state income level of the 5th income quintile, suggesting people use upward comparisons. Family income is positively related to overall happiness. Needing to borrow money mid week increases the chances of being unhappy and of having thoughts of self-harm but not tranquilizer use. However, income is not controlled for in this model therefore this may be picking up an income effect. Individual monthly income and annual savings are positively related to inverse GHQ. Using household income, only savings are positively related. Individual credit card debt is negatively related. The presence of individual (household) debt reduces the probability of scoring the maximum on the GHQ12 score by over 6% (nearly 5%). However, secured debt in the form of a mortgage, investments, windfall income and the log of house value does not effect GHQ12. There is a strong statistical association between an individual’s financial expectations and their levels of psychological well-being. Those with pessimistic views of their relative financial position (that it is worse than last year, and will get worse next year) report significantly lower GHQ12 scores than otherwise equivalent household heads. The association with psychological well-being is more than twice as strong for those who view their current financial position as worse than one year previously (M.E. -0.16) as compared to those are pessimist about future financial conditions (M.E.-0.075). A 10% increase in the level of outstanding credit (i.e. an additional £115.35) would reduce the probability, of a household head with otherwise mean characteristics, reporting a maximum GHQ12 score by 0.092. To eliminate this effect monthly labour income would have to rise by £64.30 or nearly 7%. Higher income quartiles report higher life satisfaction. However, a reduced form model instrumenting for health finds a higher coefficient on the 2nd quartile, 115

Appendix B: Summaries

Clark (2003)

Britain (BHPS), 1991-1998

Clark (2003)

Britain (BHPS), 1991, 2001

Clark & Lelkes (2005) Clark & Oswald (1994) Dehejia et al (2005)

Europe (ESS)20022003 Britain (BHPS)

Di Tella et al (2001)

Europe (Eurobarometer), 12 countries, 19751991

Di Tella et al (2003)

Europe (Eurobarometer) 1975-1992, US (GSS) 1972-1994

Dockery (undated)

Australia (LSAY & HILDA)

Dorn et al (2005)

Switzerland (Leu

US (NSFH), 1987/88 and 1992/94

followed by 3rd then 4th. The coefficient on the 4th quartile in an ordered probit model, falls from 0.750 to 0.005 in the reduced form. Income is negatively related to inverse GHQ, although partner and household income are positively related. This may be due to increased own income being related to increased hours worked which are not controlled for. Own income is positively related to life satisfaction, but negatively to inverse GHQ for a group of full time employees. Mean income of a reference group (defined by same sex, region and wave) is negatively related to life satisfaction, although not significantly so for the inverse GHQ. Income quartiles are positively, and monotonically, related to life satisfaction. No robust effect of income on inverse GHQ. The log of household income is positively related to overall happiness, and a change in the log of household income is also positively related to changes in overall happiness. Income quartiles are positively and monotonically related to life satisfaction. Being in the second income quartile compared to the first increases life satisfaction (on 4 point scale) by 0.12 points (0.2 and 0.3 for the 3rd and 4th income quartile, respectively). For Europe, income quartiles are positively and monotonically related to life satisfaction (4 point scale). The income coefficients for overall happiness (3 point scale) are very similar. For the UK only, the income coefficients are considerably larger. For the US the income coefficients are a similar size to Europe. For Europe they also explore the impact of GDP per capita which has a significantly positive effect on life satisfaction. An increase in GDP of $1000 (1985$) raises the proportion of very satisfied from 27.3% to 30.9%. Lagged GDP/capita (previous 1 and 2 years) is not significantly related to life satisfaction. However, the authors note the potential for the trended nature of GDP to create bias within this model, therefore they look to a differences model. The growth of GDP is significant only when country specific time effects are included. For HILDA, assessing one’s financial position as very comfortable or prosperous compared to just getting along, increases life satisfaction. Assessing it as poor or very poor decreases life satisfaction. For Switzerland, the difference between own income 116

Appendix B: Summaries

Easterlin (2005) Fayey and Smyth (2004)

Ferrer-i-Carbonell (2005) Frey and Stutzer (2000)

Gardner & Oswald (2005)

Gerdtham & Johannesson (2001)

1992), SHP, and reference income (which they define as 40% of International (ISSP) the average income from a Canton) increases overall happiness, but the level of average income in the Canton does not have a significant effect. However, whilst this points to a relative income, rather than absolute income effect, they can not reject a null hypothesis of equivalence between the difference and reference coefficients (β4 and β2), which would imply that reference income is having no effect. The square of the difference term is negative suggesting diminishing returns to income above the reference level. The finding for Switzerland is replicated on international data. International Across countries at once point in time overall (WVS), 1994 happiness is significantly related to log GNP/capita. Europe (EVS) Controlling for social class individual income increases life satisfaction. Controlling for individual income GDP/capita also increases life satisfaction. The coefficients are similar between the third and fourth quartiles (1.367 and 1.423) suggesting some plateauing of the income effect. Similarly, growth in GDP also increases life satisfaction. GDP/capita explains the majority of the variation in life satisfaction at the country level, suggesting a role for relative status between countries. Running the model separately for GDP/capita quartiles finds the individual income coefficient to be greater for low GDP countries (0.188, 0.145, 0.089, 0.090) for the GDP quartiles, respectively. Britain (BHPS) Log of family income is positively related to life satisfaction. Switzerland Income increases life satisfaction. People in the highest income category were 6.8% more likely to report complete satisfaction with life, compared with 9.2% more likely in the second income bracket, 5% more likely in the third and 2.9% more likely in the fourth. Britain (BHPS) Modelling the change of GHQ finds no significant effect for the year of lottery win or the year after, due to high standard errors. However, comparing 2 years before the win with 2 years after, finds that those with no win and those with a small win have a slight worsening of GHQ (rise of 0.19 and 0.18 points), however, for those with a large win GHQ drops by 1.22 points. Sweden (SLLS) Compared to being in lowest income quartile being in the fourth income quartile increases life satisfaction (life as a source of satisfaction). In a reduced form model, instrumenting for health, the income quartile coefficients for the third and fourth income quartile 117

Appendix B: Summaries

Graham & Felton (2006)

Latin America (Latinobarometro), 18 countries

Graham & Pettinato (2001)

Latin America (Latinobarmetro) & Russia (RLMS) & US (GSS)

Greenfield & Marks (2004)

US (MIDUS), 1995

Haller & Hadler (2006)

International (WVS), 34 countries 19951997

Hayo (2004)

Eastern Europe (1991)

Headey & Wooden (2004)

Australia (HILDA)

increase and are both significant. An index of ownership of 11 goods is positively related to overall happiness. Average wealth of the country and average wealth of cities the size of where the individual resides has a negative effect on overall happiness (once country dummies are included). Including relative wealth (own wealth minus average wealth at the city size or country level), which is positive and significant, makes average wealth no longer significant, suggesting a relative wealth effect. Average economic ladder question (1-10) has a linear relationship with country average wealth and strong relationship with city’s wealth, suggesting people compare themselves internationally and between cities. There is also a strong relationship between city’s average wealth and individuals ELQ, suggesting a local comparison. For Latin America, the log of wealth increases life satisfaction, as does having a higher perception of ones relative income position and perceptions of past mobility and prospects for future mobility. For Russia, the log of income increases overall happiness. If the percentage change in the log of income over the last 4 years is entered instead, this is also significant. As with Latin America perceptions of past mobility and prospects of upward mobility, and perception of one’s relative financial position are all positively related to overall happiness. For the US, income category and economic satisfaction are positively related to overall happiness, as is perceptions of past mobility. Income is not significantly related to negative affect, or purpose in life, but is significantly related to positive affect. Controlling for GNP, GDP growth, financial satisfaction and subjective class, income is not significantly related to life satisfaction or overall happiness. However, financial satisfaction, and subjective class increase both. Having no chance of escaping poverty reduces life satisfaction and overall happiness, but increases in poverty over the last 10 years is not significant. Controlling for individual income, GNP and GDP growth are positively related to life satisfaction and overall happiness. Higher income quartiles report greater life satisfaction. The marginal effect of being very satisfied (on a 1-3 scale) is 0.05 for the 2nd income quartile, 0.10 for 3rd and 0.2 for fourth. Correlations with life satisfaction and wealth (0.15) are higher than for life satisfaction and income (0.11). 118

Appendix B: Summaries

Headey & Wooden (2004)

Australia (HILDA) & Britain (BHPS) & Hungary (Tarki) & Germany (GSOEP)

Helliwell (2003)

International (WVS)

Helliwell (2004)

International (WVS) & Europe (EVS) & US (Benchmark Survey) Europe (Eurobarometer)

Hudson (2006) Inglehart & Klingemann (2000)

International (World Values Survey)

Johnson & Krueger US (MIDUS) (2006)

Kenny (2005)

International (WDH)

Lee et al (2001)

US (NSFH), 1987/88

Similarly between mental health and wealth (0.16) and mental health and income (0.10). Moving up from the 25th percentile of the income distribution to the 75th percentile increases life satisfaction by 2 percentiles. Equivalised household income (net) significantly increases life satisfaction but has not significant effect on mental health. The log of net worth has a larger effect on both life satisfaction and mental health than does income. Equivalised income has a positive effect on life satisfaction in all four countries. However, stanardized coefficients on the log of net worth are larger in Australia, Germany and Britain, than coefficients on equivalised income. Equivalised consumption is available for Britain and Hungary, but is only significant for the latter. In Hungary consumption is a stronger predictor of life satisfaction than wealth or income. Income has a non-linear relationship with life satisfaction. Income decile has a positive effect and income decile squared a negative effect. Similarly, log of national income has a positive effect and log of national income squared a negative effect. Income decile is more important for non-OECD countries. In OECD countries life satisfaction is greater in all deciles than the first 3, but there is no pattern beyond that. Controlling for a range of social capital variables, per capita median income has no significant effect on life satisfaction or happiness in the WVS, or life satisfaction in the EVS, or happiness in the US Benchmark Survey. Controlling for trust variables, the ratio of household income to average national income has a positive effect on life satisfaction. GDP/capita is positively related to a combined life satisfaction and overall happiness score. The coefficient and significance level reduces once an index for democracy is included in the model. Household income per capita and household assets per capita increase life satisfaction (index of three life satisfaction questions). However, these become nonsignificant once perceived financial situation and control over life are included in the model. The coefficient on GDP/capita is positive and significant, but once controlling for the year it is no longer significant. Neither income nor assets were related to CES-D.

119

Appendix B: Summaries Lelkes (2005)

Hungary (Tarki) & Europe (ESS)

Louis & Zhao (2002)

US (GSS) 19891994

Macdonald & Douthitt (1992)

US (Wisconsin)

Magdol (2002)

US (NSFH), 19871988, 1992-1993

Marks and Fleming Australia (1999)

Marikainen et al (2003)

UK (Whitehall II study)

Martin & Westerhof (2003)

US (MIDUS), 1995

For Hungary: the log of equivalised household income increases life satisfaction. When log of personal income was used instead this had a smaller coefficient, it also led to the positive coefficient on being a student to increase, and the variable ‘having no friends’ to become significant. For Europe: Income increases life satisfaction. The probability of reporting life satisfaction as 8-10 or more is 0.050, 0.074, 0.122 and 0.168 for the 2nd, 3rd, 4th and 5th income quintiles compared to the lowest quintile, respectively for life satisfaction. For overall happiness the marginal effects are slightly smaller: 0.041, 0.066, 0.103, and 0.127. Family income at 16 does not have a significant affect on SWB (measured by a combination of overall happiness and life satisfaction). Being currently dissatisfied with one’s finances, finances having deteriorated or remained the same reduces SWB. An index of satisfaction with different life domains is positively related to the log of household income, but negatively related to the log of expected income, which is inconsistent with a life cycle income hypothesis. Annualized net worth was also not significantly related to life satisfaction. A model to test Duesenberry’s relative income hypothesis found that household income had a positive effect of the index of domain satisfactions, and the difference between average expenditure and household expenditure had a negative effect. This suggests that, controlling for income, if households spend above average this has a negative effect on SWB. A third specification finds the difference between income and the level of income reported to be ‘terrible’ has a positive effect on the index of domain satisfactions. Income significantly reduces CES-D for women only. This becomes non-significant once other variables are included. Standardised income is positively related to an index of domain satisfactions, controlling for occupation status. There was no significant interaction between age and the income effect. One standard deviation change in income increased SWB index (0-100) by about 1.5 units. Income and wealth has a positive effect on psychological well-being, however, the effect reduces after adjusting for baseline health. Controlling for subjective financial assessment, assets are positively related and debts negatively related to life satisfaction. Controlling for assets and debts, 120

Appendix B: Summaries

McBride (2001)

O’Connell (2004) Pichler (2006)

Phelps (2001) Van Praag & Bernard (2005) Van Praag et al (2003) Rehdanz & Maddison (2003)

Ritchey et al (2001) Schyns (2001) Senik (2004)

financial satisfaction and perceived control over finances increase life satisfaction. US (GSS), 1994 Both the log of income and the average income of (n=324) those in a similar age group (+/- 5 years) were not significant. However, having a standard of living which is much worse than one’s parents at the same age has a negative effect on overall happiness. Additional analysis suggests the relative income effect is smaller at low income levels where absolute income is more important. Rerunning the model on the high income group (n=228, income >=$20000) finds that increases in log income reduce the probability of being in the highest SWB group, and cohort income has a stronger marginal effect. For a low income sample, the cohort income has less effect; “at low income levels, the relative-income effects appear to be smaller and income becomes more important” (p. 276). Controlling for inequality, GDP/capita significantly Europe increases life satisfaction in 1995, but not 1996, 1997 (Eurobarometer), or 1998. 15 countries Europe (ESS), 2003 Compared to earning €750-2000 per month, earning €400-750 reduces SWB (combined overall happiness and life satisfaction) for those under 30 and 30 and over. Earning under €400 also reduces SWB, but particularly so for the 30 and over age group. Earning over €2000/month increases SWB for the 30 and over group, but not the under 30 group. In a separate model, perceived financial pressure lowers SWB, particularly for the old. US (Americans’ Family income is positively related to overall mental health) happiness. Netherlands Log of household income is positively related to assessing life more positively. Germany (GSOEP) Financial satisfaction is strongly related to overall happiness. GDP/capita increases overall happiness, GDP squared International has a negative sign but is not significant, GDP growth (WDH), 67 has a positive sign but is not significant, and GDP countries shortfall (the extent to which past income has been above the level reached in the survey year), is also negative and not significant. Controlling for the extent of activities one engages in, US (Americans’ income is not significantly related to life satisfaction, Changing Lives), depression or overall happiness. 1989 Russia (RUSSET) Income has a small positive effect on life satisfaction, although they have a reciprocal relationship. Russia (RLMS) The log of household income, household expenditure and individual income all increase life satisfaction (positive and significant when entered separately). 121

Appendix B: Summaries

Shields & Wailoo (2002) Shields & Price (2005) Smith (2003)

Stutzer (2004)

Subramanian et al (2005) Theodossiou (1998) Thoits & Hewitt (2001) Welsch (2002) Winkelmann (2004) Winter et al (1999)

However, the household variables were stronger predictors. Controlling for the log of household income and individual income, the log of reference income (based on education, occupation, industry, work experience, gender, and region) increases life satisfaction. The author argues that comparison income is acting as an ‘information’ tool. However, this finding is predominantly for those < 40 and those with volatile incomes suggesting it may be due to measurement error. Britain (Ethnic Household income is not significantly related to an minority survey) index of unhappiness. England (HSE) There is an inverse U shaped relationship between household income and inverse-GHQ; those in the £20,800 to £31,200 band having the best mental health. Compared to being in the bottom income quartile, Former Soviet Union, USA, West being in the 2nd and 3rd has no impact on overall Germany (WVS) happiness in USA or West Germany, although being in the highest group increases overall happiness. In the former-USSR, being in the 2nd, 3rd and 4th income quartile increase overall happiness compared to being in the lowest income group. Switzerland (Leu The log of household income is positively related to data) life satisfaction. The log of the income perceived to be sufficient for one’s household is negatively related to life satisfaction. A 10% increase in income increases aspirations by 4.2%. The average income in the community and the proportion of rich people in the community increases aspirations. Furthermore, the interaction between contact with neighbours and the proportion of rich people also increases aspirations suggesting this is not due to cost of living differences. US There is a strong income gradient on the probability (BenchmarkSurvey) of being unhappy (however, only 5% of the sample are classed as unhappy). Britain (BHPS) Having a low paid job is not significantly related to feeling less happy. US (Americans’ Family income is positively related to overall Changing Lives) happiness, life satisfaction and negatively related to depression. Controlling for environmental and scientific variables International the log of GNP/capita is positively related to the log (WDH), 54 of overall happiness. countries Germany (GSOEP) The log of family income is positively related to life satisfaction. Poland The log of household income per capita is positively related to an index of domain satisfactions.

122

Appendix B: Summaries

Correlations between income and well-being, controlling for other variables and unobservable individual effects Di Tella et al (2006)

Dockery (undated)

Dorn et al (2005)

Ferrer-i-Carbonell (2005)

Germany (GSOEP)

The log of household income is positively related to life satisfaction, the coefficient is larger for men than women and larger for left wingers than right wingers. Income in previous periods (up to four years previously) is negatively, but not significantly, related to life satisfaction. Average income over the last five years is not significantly related to life satisfaction (except for right wingers). The authors are not able to reject the hypothesis that people “adapt totally to income within four years” (p. 23). Australia (LSAY, For the LSAY, controlling for a range of personality HILDA) variables an index of family wealth (presence of consumer items) was positively related to overall happiness at 16, 18 (only 10% significance) and 19. Considering only those at work, controlling for satisfaction with pay, the log of hourly wage is negatively related to life satisfaction for 1999 data. However, for 2000 data controlling for satisfaction with pay, those in the highest wage quintile are happier. Switzerland (SHP) Subsistence income (40% of the average income) at (Leu data 1992) & the Canton level does not have a significant effect in International (ISSP) SHP data, but reduces overall happiness in the Leu 1992 data. Relative income, or the difference between own and subsistence income, increases overall happiness. However, since they are not able to reject a null hypothesis that the coefficients on subsistence and relative income are equivalent, it does not disprove the absolute income hypothesis. Where relative income is positive, its square is always negative and significant suggesting decreasing marginal utility of income. In the ISSP data, subsistence income (40% of national average equivalent income), has no significant effect on overall happiness, relative income increases overall happiness, and relative income squared reduces overall happiness. Again the null of equivalence can not be rejected. Germany (GSOEP), Using an ordered probit model with individual 1992-1997 random effects and controlling for the individual mean of log income, log education, log of the number of children and log of the number of adults (Mundlak approach) the log household income has small positive effect on overall happiness, and the effect is stronger in East Germany than West Germany. The 123

Appendix B: Summaries coefficient on the log of reference income (based on region, age, and education category) is negative and very similar to that on family income. Hence an increase in family income with an accompanying increase in reference income does not change SWB. However, the finding was not robust for Eastern Europe which may be due to increased instability in Eastern Europe resulting in relative income giving more mixed messages. The difference between log of own income and log of reference income is positive and significant for whole sample and West Germany (at 10%), but not East Germany. This may be due to mixed messages from reference income (cf. Senik for Russia). Comparison effects are asymmetric for West Germany and the whole sample. Being above ones reference income has a positive but not significant effect on overall happiness, being below ones reference income has a negative and significant effect, suggesting that for West Germany and whole sample only upward comparisons matter.

Ferrer-i-Carbonell & Frijters (2004)

Germany (GSOEP)

Frijters et al (2004) Germany (GSOEP)

Gerlach & Stephan (1996)

Germany (GSOEP)

Graham et al (2004)

Russia (RLMS)

Luttmer (2004)

US (NSFH)

“increases in family income accompanied by identical increases in the income of the reference group do not lead to significant changes in wellbeing” (p1015). The log of household income increases life satisfaction but the coefficient is about one third lower where individual effects are controlled for (in an ordered logit, controlling for fixed effects reduces the coefficient from 0.47 to 0.19 (0.11 in the OLS fixed effects). The log of household income is positively related to life satisfaction for males and females. One unit increase in log income leads to around 0.5 increase in life satisfaction (on a 0-10 scale). The log of household income per capita increases life satisfaction for women and men under 30 and 30-49, but not over 49. Using a first difference model, log of equivalent income has a positive effect on life satisfaction. If unexplained life satisfaction from 1995 is included in a model explaining life satisfaction in 2000 then income in 1995 and 2000 are still both significant. The percentage change in equivalised income, controlling for initial level and unexplained life satisfaction, has a positive effect. The log of average (predicted) earnings within the local area reduces overall happiness, and the log of household income increases happiness. The coefficient on the latter increases if the value of the 124

Appendix B: Summaries respondents home and tenure status are not also controlled for. These findings are robust to a range of specifications, including controlling for individual fixed effects. The log of neighbourhood earnings are not significant for the depression index. Interaction effects show that the log of average earnings is strongest for married or divorced, those aged 30 to 60, but similar for men and women, children versus no children, renting and home ownership. Those who socialise with their neighbours less than once a month are less effected by log of neighbourhood income (the coefficient is about one third of the size, and only significant at 10%). Meier & Stutzer Germany (GSOEP) The log of hourly wage and log of household income (2006) are both positively related to life satisfaction. Oswald & Britain (BHPS) Household income per capita increases life Powdthavee (2005) satisfaction. Stutzer & Frey Germany (GSOEP) The log of household income increases life (2006) satisfaction. Weinzierl (2005) Germany (GSOEP), The log of income is positively related to life 1984+ satisfaction, the log of lagged income is not significant, and the log of peer income (based on age, gender and education) is negatively related. The coefficient on log income of 0.035 (SE 0.004) is virtually matched by coefficient on log peer group income of -0.031 (SE 0.013), with lagged income having a negative, but non-significant effect on wellbeing. Hence these data suggest that “steady growth is associated with no significant increases in well-being” (p. 13). Wildman & Jones Britain (BHPS) Log of household income has a non significant effect (2002) on the GHQ, however, perception of financial status is strongly related. For men, considering ones financial situation to be very difficult increases the GHQ Likert (0-36 scale) by 3.412 points compared to assessing it as ‘getting by’. For women there is some evidence that relative deprivation has a negative impact. Controlling for household income having an income lower than 50% of the average increases the GHQ. Germany (GSOEP) The log of household income increases life Winkelmann & satisfaction, but the effect is reduced once individual Winkelmann effects are controlled for. (1999)

125

Appendix B: Summaries

B.2. Personal characteristics – who we are, our genetic makeup Table B.2.1: Age Evidence on Age Study Correlations Cummins, Eckersley, Kai Lo, Okerstrom, Davern & Woerner (2004)

Measure/Data Source (Country)

Summary

PWI (Australia)

Life satisfaction and happiness increases with age and depression decreases with age. Having a partner between ages 26 to 55 crucial to well-being. Wellbeing of the youngest group and the oldest group is much less influenced by income than the middle-age groups (26-55). Only in 18-25 age group is there no gender difference, it is greatest in 36-35 group and subsequently decreases with age.

Huta, Park, Peterson OTH (Internet) & Seligman (2006) Keyes, Schmotkin, & Ryff (2002).

Age significantly negative correlated with the pursuit of pleasure.

MIDUS (US)

Life satisfaction significantly positively correlated with age.

Controlling for other variables but not unobserved individual effects, individual level data In the US age is negatively related to happiness Alesina, Di Tella & GSS (US) (5%), but age squared (the quadratic) is positively MacCulloch (2004) Eurobarometer related to happiness (5%). This pattern is the same (Europe) for life satisfaction in Europe. The quadratic suggests a U-shaped function with both happiness and life satisfaction lower in middle age than early or late adulthood. Baker, Cahalin, ACL (US) Older people in an elderly cohort ( ≥ 60yrs) were Gers & Burr (2005) happier than younger people (5%) but there were no differences in life satisfaction or depressive symptoms. Age negatively related to happiness (US) & life Blanchflower and GSS (US) satisfaction (Britain) but significant quadratics again Oswald (1997;2004) Eurobarometer suggest a U-Shaped age curve. (Europe) ISSP (International) Age negatively related to happiness but significant Blanchflower and quadratic again suggests a U-Shaped age curve. Oswald (2005) Brown (2000)

NSFH (US)

Brown, Taylor, Wheatley Price

BHPS (UK)

Age negatively related to depressive symptoms (0.1%). Using age dummies, they found those with the most positive GHQ-12 scores to be the age group 55-

126

Appendix B: Summaries (2005)

64yrs.

Burchardt (2004)

BHPS (UK)

Age negatively related to satisfaction with income but the significant quadratic suggested a U-Shaped age curve. Life satisfaction highest amongst 20-29yr olds, although only in Portugal, Greece, France, Finland & Austria. No quadratic checked.

Christoph & Noll (2003)

Eurobarometer (Europe)

Clark (2003)

BHPS (UK)

U-shaped age curve (minimising at 36 yrs) this time using reversed GHQ-12 scores.

Clark and Lelkes (2005)

ESS (Europe) & BHPS (UK)

Age negatively related to life satisfaction but significant quadratic again suggests a U-Shaped age curve.

Clark and Oswald (1994)

BHPS (UK)

Loss in well-being (GHQ scores) due to unemployment is more acute for people in mid adulthood (30-49yrs) than young or old adulthood.

Dockery (unpublished)

HILDA (Australia)

Age U shaped minimum at 35-44, although for women it is more constant until 45

Easterlin (2003)

GSS (US)

A range of age related interactions for happiness, e.g. in relation to marital status income etc. Seems its is hard to fully disentangle age from these other changes despite regression models including them since many people follow similar life courses.

Easterlin (2006)

GSS (US)

Slight rise from age 18-51 and then a decline. Possible bias due to earlier death of unhappy people.

Fayey and Smyth (2004)

EVS (Europe)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve.

Ferrer-i-Carbonell (2005)

GSOEP (Germany)

Age negatively related to happiness but significant quadratic suggests a U-Shaped age curve with minimum around 44yrs.

Ferrer-i-Carbonell and Gowdy (2005)

BHPS (UK)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve.

Ferrer-i-Carbonell & Frijters (2004)

GSOEP (Germany)

Age quadratic significant assuming cardinality but not ordinality of life satisfaction measure.

Gerdtham and Johannesson (2001)

SLLS (Sweden)

"Daily life is a source of personal satisfaction" highest amongst those aged 64yrs or above. 127

Appendix B: Summaries

Graham and Felton (2006)

Latinobarometer (Latin America)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve with minimum at 51.5, later than most European data.

Graham and Pettinato (2001)

Latinobarometer (Latin America) GSS (US) RLMS (Russia)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve with minimum at 49yrs (Latin America), not noted for US or Russia.

Graham, Graham Carol, Eggers & Sandip (2004)

RLMS (Russia)

Age is U-shaped with a minimum of 47.

Greenfield & Marks (2004)

MIDUS (US)

Amongst an older sample (65-74yrs), older people reported greater negative affect but not less positive affect or more purpose in life.

Haller & Hadler (2006)

WVS (International)

Using age dummies the age group with the highest levels of happiness and life satisfaction were those younger than 30 yrs.

Hayo & Seifert (2003)

(NDB) Eastern Europe

Age has a U shaped effect on life satisfaction with minimum at 37 yrs.

Hayo (2003)

(NDB) Eastern Europe

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve.

Headey & Wooden (2004)

HILDA (Australia)

Age negatively related to life satisfaction and mental health (5 items from SF36 scores) but significant quadratics suggest a U-Shaped age curve for both.

Headey & Wooden (2004)

BHPS (UK) GSOEP (Germany) Tarki (Hungary) SEP (Netherlands)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve in all countries.

Helliwell (2003)

WVS (International)

Using age dummies all ages had lower life satisfaction than 18-24 year olds except those aged >65 years who had higher.

Helliwell & Putnam (2004)

WVS (International)

Replicated findings reported by Helliwell (2003).

Hudson (2006)

Eurobarometer (Europe)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve.

Koropeckyj-Cox (2002)

NSFH (US)

Depressive symptoms (CES-D) were not related to age for people without children either men or 128

Appendix B: Summaries women. Lee & Bulanda (2005)

GSS (US)

Age is positively related to happiness amongst both men and women.

Lee, DeMaris, Bavin, & Sullivan (2001)

NSFH (US)

No effect of age on depressive symptoms (CES-D), but only included over 65yr olds.

Lelkes (2006)

ESS (Europe)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve.

Louis & Zhao (2002)

GSS (US)

Luttmer (2004)

NSFH (US)

No effect of age on composite of happiness & life satisfaction though unclear why as used similar years to other authors and used models with few variables. In Model 1 only own and mother's education significant, not exactly variables that should take away any age effect. Age linearly related to happiness, with 19-35 year olds having lowest happiness and over 60s the highest happiness.

Magdol (2002)

NSFH (US)

Fewer depressive symptoms (CES-D) among older people, men and women.

Peterson, Park, & Seligman (2005)

OTH (Internet)

Age negatively related to life satisfaction - no quadratic tested.

Shields & Price (2005)

HSE (UK)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve.

Smith (2003)

WVS (USSR; US; West Germany)

Age dummies used and suggest a U-Shaped happiness function for USSR and Germany, but no significant pattern for US. West Germans reach low point (30s) earlier than those in USSR (40s).

Stutzer (2004)

SHPS (Switzerland)

No effect of age or age squared on life satisfaction

Theodossiou (1998)

BHPS (UK)

Older people are less likely to happy (GHQ question) though its unclear why age squared was significant with a log-ratio of 1.00.

Weinzierl 2005

GSOEP (Germany)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve. A good example of where the coefficient is tiny, β < 0.001, though significant, p < 0.001.

Wiggins, Higgs, Hyde & Blane

CASP-19 (UK)

People over 70 had lower CASP-19 scores than those aged 64-70, suggesting that the U-shaped 129

Appendix B: Summaries (2004) Winkelmann (2004)

curve has a downturn again in very old age. GSOEP (Germany)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve.

Controlling for unobserved effects, individual data US (NSFH) Depressive symptoms at Wave 2 (CES-D) were Kim & McKenry greater among younger people (controlling for CES(2002) D scores at Wave 1). Lucas et al (2004)

GSOEP (Germany)

A significant relation with age and life satisfaction is reported during the "reaction phase" to unemployment but the marginal effect is 0.00, so the direction of the relationship is unclear.

Lucas (2005)

GSOEP (Germany)

Younger people have higher life satisfaction than older people prior to divorce, but age does not affect life satisfaction in the reaction or adaptation phases.

Meier & Stutzer (2006)

GSOEP (Germany)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve without fixed effects, but not significant with fixed effects.

Oswald & Powdthavee (2005)

BHPS (UK)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve with and without fixed effects.

Stutzer & Frey (2005)

GSOEP (Germany)

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve without fixed effects, but not significant with fixed effects.

Thoits & Hewitt (2001)

ACL (US)

Wildman & Jones (2002)

BHPS (UK)

Controlling for happiness at T1, age negatively related to happiness but a significant quadratic suggests a U-Shaped age curve. Controlling for life satisfaction at T1, there was no relationship with age at T2 but a significant quadratic again suggests a UShaped age curve. Using age dummies, there appears to be the familiar U- Shaped curve for GHQ scores for both men and women though significance is not reported.

Winkelmann & Winkelmann (1998)

GSOEP (Germany)

130

Age negatively related to life satisfaction but significant quadratic suggests a U-Shaped age curve without fixed effects, but not significant with fixed effects.

Appendix B: Summaries

Table B.2.2: Gender Evidence on Gender Study Correlations Cummins et al., (2004)

Measure/Data Source (Country)

Summary

PWI (Australia)

Women has higher 'Personal Well Being' by about 1.5% points, being higher on all domains except safety. Men living alone have lower well-being than those in other arrangements but women living alone close to their normal range.

Controlling for other variables but not unobserved individual effects, individual level data Women happier than men in US and Europe. Alesina, Di Tella & GSS (US); MacCulloch (2004) Eurobarometer (Europe) ACL (US) No difference M/F for happiness or life satisfaction, Baker, Cahalin, but women reported more depressive symptoms Gers & Burr (2005) (CES-D) Bardasi & Francesconi (2004)

BHPS (UK)

Seasonal/casual work is associated with higher life satisfaction amongst women, but more depressive symptoms amongst men. However, these effects are very small.

Biblarz & Gottainer (2000)

GSS (US)

Women happier than men.

Blanchflower & Oswald (2004)

UK (Eurobarometer) UK - Women report higher life satisfaction than GSS (US) men. Being unemployed or at home affects both men and women but is worse for men's life satisfaction. US - Women happier than men.

Blanchflower and Oswald (2005)

ISSP (International)

Men report higher happiness than women, but not for Australia.

Brown (2000)

NSFH (US)

Women show more depressive symptoms (CES-D) than men, especially amongst low educated women.

Brown, Taylor, Wheatley Price (2005)

BHPS (UK)

Men show higher well-being (inverse GHQ scores) than women.

Clark (2003)

BHPS (UK)

Men show higher well-being (inverse GHQ scores) than women.

Clark and Lelkes

ESS (Europe);

Men show lower life satisfaction

131

Appendix B: Summaries (2005)

BHPS (UK)

Clark &Oswald (1994)

BHPS (UK)

Men show higher well-being (inverse GHQ scores) than women.

Dehejia, DeLeire & Luttmer (2005)

US (NSFH)

There is no difference in happiness levels, or changes in happiness levels, between men and women.

Di Tella, MacCulloch & Oswald (2001)

Eurobarometer (Europe)

Men have lower life satisfaction than women (though significance not noted)

Di Tella, MacCulloch & Oswald (2003)

GSS (US), Eurobarometer (Europe)

Women have higher happiness and life satisfaction scores in both continents.

Di Tella, HaiskenDe New & MacCulloch (2006?)

GSOEP (Germany)

No difference in life satisfaction between men and women.

Easterlin (2006)

GSS (US)

Women report higher happiness than men.

Fayey and Smyth (2004)

Europe (EVS)

No difference between men and women for life satisfaction.

Ferrer-i-Carbonell (2005)

GSOEP (Germany)

Men lower life satisfaction than women.

Ferrer-i-Carbonell & Gowdy (2005)

BHPS (UK)

Men have lower life satisfaction than women but only significant once a number of psychological controls (e.g. self-esteem) are introduced to the analysis.

Flouri (2004)

NCDS (UK)

Women show poorer GHQ scores, higher Malaise but also higher life satisfaction.

Frey & Stutzer (2000)

SHPS (Switzerland)

No difference between men and women on life satisfaction scores.

Gerdtham & Johannesson (2001)

SLLS (Sweden)

Daily life is less satisfying for men than women.

Graham & Felton (2006); Graham & Pettinato (2001)

Latinobarometer (Latin America)

No difference in reported happiness among men and women.

Graham, Graham

RLMS (Russia)

Men happier than women.

132

Appendix B: Summaries Carol, Eggers & Sandip (2004) Greenfield & Marks (2004)

MIDUS (US)

Women report more negative affect, less positive affect and less purpose in life. Note women lower on all indices of well-being here (positive and negative)

Haller & Hadler (2006)

WVS (International)

Women have higher happiness and life satisfaction than men.

Hayo (2004)

Eastern Europe (NDB)

No difference in life satisfaction between men and women.

Headey & Wooden (2004)

HILDA (Australia) BHPS (UK); GSOEP (Germany); Hungary (Tarki).

Australia - Women report higher life satisfaction but lower SF36 scores. In European countries - no significant difference between men and women

Helliwell & Putnam (2004)

WVS; EVS; ESC (International)

Males less happy in US, Canada and using the WVS data. However, no gender differences for life satisfaction.

Hudson (2006)

Eurobarometer (Europe)

Women report higher life satisfaction than men.

Lee, DeMaris, Bavin, & Sullivan (2001)

US (NSFH)

Women report higher levels of depressive symptoms than men (CES-D).

Lelkes 2006

ESS (Europe)

Men have lower life satisfaction than women.

Louis & Zhao (2002)

GSS (US)

No difference in happiness as a function of gender.

Meier & Stutzer (2006)

GSOEP (Germany)

Women report higher life satisfaction than men.

Oswald & Powdthavee (2005)

BHPS (UK)

No gender effect on life satisfaction amongst those who can't work due to health problems.

Peterson, Park, & Seligman (2005)

OTH; SWLS (Internet)

Women report higher SWLS scores than males.

Shields & Price (2005)

HSE (UK)

Men report higher well-being using inverse GHQ scores.

Smith (2003)

WVS (USSR, US & Germany)

Men report lower happiness than women in the USSR but there were no differences in the US or Germany.

133

Appendix B: Summaries Stutzer (2004)

SHPS (Switzerland)

No difference in life satisfaction as a function of gender.

Stutzer & Frey (2005)

GSOEP (Germany)

No difference in life satisfaction as a function of gender amongst those who commute and are employed/self-employed.

Stutzer & Frey (2006)

GSOEP (Germany)

Amongst people who got married in the period (1984-2000), females tended to report higher life satisfaction than men.

Subramanian et al (2005)

US (SCBS)

Females tend to show lower levels of happiness than males.

Theodossiou (1998)

BHPS (UK)

Men are less likely to report being happy (GHQ question) than women.

Winkelmann (2004)

GSOEP (Germany)

No effects of gender on life-satisfaction.

Controlling for unobserved effects, individual data NSFH (US) Women report higher levels of depressive symptoms Kim & McKenry than men (CES-D) even controlling for prior levels (2002) of CES-D. Thoits & Hewitt (2001)

ACL (US)

Controlling for happiness at T1, and life satisfaction at T1, gender doesn't seem to have any effect.

134

Appendix B: Summaries

Table B.2.3: Ethnicity Evidence on ethnicity

Study

Measure/Data Summary Source (Country) Controlling for other variables but not unobserved individual effects, individual level data Alesina, Di Tella GSS (US) Whites happier than African Americans. and MacCulloch (2004) Baker, Cahalin, Gers & Burr (2005)

ACL (US)

Amongst respondents over 60, there were no effects of ethnicity on happiness, life satisfaction or depressive symptoms. Contrast this with the findings of Thoits & Hewitt (2001) using the same data set and it suggests that the ethnicity differences reported by them may be principally among those less than 60 yrs old.

Biblarz & Gottainer (2000)

GSS (US)

Happiness lower among blacks than whites, but no difference between whites and 'other' ethnicities.

Brown (2000)

NSFH (US)

More depressive symptoms (CES-D) among blacks than whites although this factor becomes unsignificant when relationship factors are added, especially the instability of marriage/cohabiting.

Blanchflower & Oswald (2004),

GSS (US)

Whites happier than African Americans but not 'other race'.

Dehejia, DeLeire & Luttmer (2005)

NSFH (US)

Being black, or from another non-white ethnic group compared to being white does not effect overall happiness. However, for blacks but not whites there is an ‘insurance’ effect of religion in that the impact of income is lower for those blacks who are religious.

Easterlin (2006)

GSS (US)

Whites happier than African Americans.

Ferrer-i-Carbonell (2005)

GSOEP (Germany)

East Germans have lower life satisfaction than West Germans. (Not quite ethnicity but the closest heading we have)

Frey & Stutzer (2000); and Stutzer (2004)

SHPS (Switzerland)

"Foreigners" (i.e. non-Swiss) had lower life satisfaction than native Swiss. Importantly, this category of citizens lacks certain voting rights which the authors argue are important for wellbeing.

135

Appendix B: Summaries

Graham (2005)

Review

Blacks less happy in US, and minorities less happy in Latin America, but minorities happier in Russia.

Greenfield & Marks (2004)

MIDUS (US)

No differences, among the over 65s, in positive affect, negative affect, and purpose in life as a function of ethnicity.

Koropeckyj-Cox (2002)

NSFH (US)

Amongst childless people, White women showed more depressive symptoms (CES-D) than non-white women. There was no difference in symptoms among men as a function of ethnicity.

Lee & Bulanda (2005)

GSS (US)

African Americans less happy than Whites. No difference between Whites and "other" ethnicities.

Lee, DeMaris, Bavin, & Sullivan (2001)

NSFH (US)

Among widows/widowers, race does not affect number of depressive symptoms (CES-D).

Lelkes (2006)

ESS (Europe)

Being a member of an ethnic minority decreases life satisfaction.

Louis & Zhao (2002)

GSS (US)

No significant difference for happiness between blacks and whites until "adulthood variables" are added (e.g. marital status, health and job satisfaction). When these are added whites tend to show higher levels of happiness.

Luttmer (2004)

NSFH (US)

Amongst married/cohabiting people, the average reported happiness of the couple was higher for Hispanics than Whites, and other races were no different than Whites (Blacks, Asians or Other).

Magdol (2002)

NSFH (US)

Whites show lower depressive symptoms (CES-D) than blacks. This pattern holds for both men and women even when a range of other variables are added concerning social contact and support.

McKenry & McKelvey (2003)

NSFH (US)

No differences in happiness between black and white mothers following divorce/separation and both recover equally well over a five year period.

Meier& Stutzer (2006)

GSOEP (Germany)

East Germans have lower life satisfaction than West Germans.

Peterson, Park, & Seligman (2005)

OTH (Internet)

No difference in life satisfaction as a function of ethnicity in this on-line study.

Pichler (2006)

ESS (Europe)

Members of "discriminated against groups" reported 136

Appendix B: Summaries lower life satisfaction than people who did not report such group membership. Stutzer & Frey (2005)

GSOEP (Germany)

East Germans have lower life satisfaction than West Germans, but EU foreigners have higher life satisfaction than Germans in general. Other types of foreigners have lower life satisfaction though.

Subramanian et al (2005)

US (SBSC)

Whites are happier than Blacks or other ethnicities.

Theodossiou (1998)

BHPS (UK)

No difference in happiness for Whites compared to 'other 'ethnicities.

Controlling for unobserved effects, individual data ACL (US) Controlling for effects at T1, Whites were happier, Thoits & Hewitt had higher life satisfaction, showed fewer depressive (2001) symptoms and higher levels of personal mastery than non-whites.

137

Appendix B: Summaries

Table B.2.4: Personality Evidence on Personality Study

Measure/Data Source (Country)

Summary

Correlations Controlling for other variables but not unobserved individual effects, individual level data NCDS (UK) Psychological problems at age 16 were associated Flouri (2004) with worse GHQ scores and more Malaise but no differences in life satisfaction. Helliwell (2006)

WVS (International)

Extraversion is only significantly related to life satisfaction at 10% (after controlling for social trust, belief in god and other variables). A psychoticism index is significant only when interacted with the measure of social trust. Trust being particularly important for those higher in the psychoticism index.

Lee, DeMaris, Bavin, & Sullivan (2001)

NSFH (US)

There was no significant relationship between sociability and the number of depressive symptoms (CES-D).

Marks, Lambert & Choi (2002)

NSFH (US)

Prior levels of happiness, CES-D and mastery were significantly related to later levels, even after controlling for many other factors for both men and women, suggesting a degree of stability.

Ferrer-i-Carbonell and Gowdy (2005)

BHPS (UK)

Most of the GHQ sub-scales (e.g. enjoy challenges and self-worth) are related to life satisfaction. The question is whether these are personality variables.

Controlling for unobserved effects, individual data Kim & McKenry NSFH (US) Controlling for prior CES-D scores, there was still a (2002) strong negative relation between self-esteem and CES-D at time 2. People higher in self-esteem reported fewer depressive symptoms.

138

Appendix B: Summaries

Table B.2.5: Physical characteristics Evidence on Physical Characteristics Study Correlations Cummins, et al. (2004)

Measure/Data Source (Country)

Summary

PWI (Australia)

PWI positively related to the population frequency distribution for height. The people with the highest personal wellbeing lie within the most common height range (160-169 cm) – very small/tall have lower PWI but sample sizes small. However, more of very small have lower incomes. PWI drops after weight of 100kg and sharply at 120kg. BMI shows dropping at moderate obese and severe obese but not mild. Also related to income.

Controlling for other variables but not unobserved individual effects, individual level data Gerdtham & Johannesson (2001)

SLLS (Sweden)

No difference in satisfaction with daily life amongst people with body mass indexes either above or below 30.

139

Appendix B: Summaries

3. Socially developed characteristics – our human and physical capital Table B.3.1: Education Study

Data (Country)

Summary

Correlation between education and SWB, not controlling for other variables Keyes et al (2002)

US (MIDUS)

SWB and education correlated 0.07 (p=3), 1.72 for 154

Appendix B: Summaries

Weinzierl (2005) Wildman & Jones (2002)

Germany (GSOEP) Britain (BHPS)

Winkelmann (2004) Winkelmann & Winkelmann (1998)

Germany (GSOEP) Germany (GSOEP)

women and 2.05 for men. Moving from unemployment to employment reduces the odds of a high GHQ score by 0.87 for men and 0.79 for women. Being employed has a positive effect on life satisfaction. Being unemployed increases GHQ Likert (0 to 36 scale) by 1.979 points, the effect is reduced to 0.989 once satisfaction with finances, and financial expectations are controlled for. Being unemployed reduces life satisfaction, controlling for the family specific random effect. Employees with past unemployment experience have a lower satisfaction than the average employee, hence might be intrinsically dissatisfied, or their jobs are insecure. Individuals who later become unemployed have an average satisfaction level of 7.3 in 1984, compared with an average of 7.6 for all employed. However, although unemployed are slightly less satisfied before becoming unemployed this effect is small compared with the drop in satisfaction while actually unemployed (5.6). Those unemployed in t-1 and t have similar levels of satisfaction, suggesting individuals do not get used to their situation. Being unemployed reduces life satisfaction in a pooled ordered logit model. Controlling for fixed effects makes little difference to the point estimates. The detrimental effect of unemployment is the same regardless of the reason for termination (involuntary or voluntary) suggesting unemployment causes dissatisfaction rather than the dissatisfied choose to leave. Duration is not significant, suggesting individuals do not get used to their situation. Unemployment as a worse effect on the under 30 age group, under 50 and least on the 50s and over.

155

Appendix B: Summaries

B.4. How we spend our time Table B.4.1: Hours worked Evidence on hours worked Study Correlations Schoon et al (2005)

Data (Country)

Summary

NCDS (UK)

Men who work part time have slightly lower life satisfaction than those working full time, especially divorced men.

Controlling for other variables but not unobserved individual effects, individual level data ACL (US) Amongst the over 60s, happiness is positively Baker, Cahalin, associated with the amount of time engaged in Gers & Burr (2005) 'productive' activities (e.g. paid/unpaid work). Bardasi & Francesconi (2004)

BHPS (UK)

Life satisfaction and GHQ scores are not affected by whether the numbers of hours worked is 1-15 pw or 16-29 pw, for either men or women. Working part time, compared to full time, had no significant effect on overall happiness.

Blanchflower & Oswald (2004)

GSS (US)

Blanchflower & Oswald (2005) Luttmer (2000)

ISSP (International)

Meier & Stutzer (2006)

GSOEP (Germany)

NSFH (US)

Working part time, compared to full time, had no significant effect on overall happiness. The log of average usual workings hours for husband and wife (controlling for being employed), does not have a significant effect on average overall happiness. However, when only the main respondent is included, log of usual working hours reduces overall happiness (coefficient of -0.111), when predicted income is used on the averaged variables, again log of usual working hours reduces overall happiness (coefficient of -0.137). Amount of hours worked was positively associated with life satisfaction (even in the fixed effects model). However, the square of the hours worked was negative suggesting an inverse U-shaped curve such that LS goes down after a certain number of hours are worked but the peak number of hours is not reported. The amount of time spent volunteering is also positively related to LS (although less so in the fixed effects model).

Controlling for unobserved effects, individual data Weinzierl (2005) GSOEP (Germany) Number of hours worked per week positively associated with life satisfaction.

156

Appendix B: Summaries

157

Appendix B: Summaries

Table B.4.2: Commuting Evidence on Commuting Study Correlations Kahneman, et al. (2004)

Data (Country)

Summary

DRM (US)

Commuting associated with lower levels of positive affect and higher levels of negative affect compared to all other activities, though still a positive ratio.

Controlling for other variables but not unobserved individual effects, individual level data More time commuting associated with lower Helliwell & Putnam WVS, ESS, SCBS (International) happiness in the benchmark study. (2004) Stutzer & Frey (2005)

GSOEP (Germany)

158

Commuting time negatively associated with life satisfaction although the commuting time squared was positively associated suggesting a U-shaped curve with the worst effects occurring for moderate, rather than very long, commutes. Using a fixed effects model, only the straight commuting time variable was significant. Moreover, they test to see if other households members benefit from respondent commuting, but find negative coefficient on partners commuting time.

Appendix B: Summaries

Table B.4.3: Housework Evidence on Housework Study Correlations Kahneman, et al. (2004)

Data (Country)

Summary

DRM (US)

Housework associated with more positive than negative affect and lies in the mid-range of activities in terms of ratio.

Controlling for other variables but not unobserved individual effects, individual level data NSFH (US) While there was no relationship with number of Lee, DeMaris, hours spent on housework and depressive symptoms, Bavin, & Sullivan amongst older people (>65yrs) the more they (2001) disliked housework the more depressive symptoms they exhibited. In other words, its not how much they did but how much they disliked doing it that mattered. Magdol (2002)

NSFH (US)

More housework was associated with more depressive symptoms among women.

159

Appendix B: Summaries

Table B.4.4: Caring for others Evidence on caring for others Study Correlations Kahneman, et al. (2004)

Data (Country)

Summary

DRM (US)

Caring for one's children associated with more positive than negative affect but the ratio was amongst the lowest across a range of activities and was less positive than spending time with family and friends.

Controlling for other variables but not unobserved individual effects, individual level data Hirst (2003; 2005) BHPS (UK) The number of hours of voluntary care provided positively associated with psychological distress (GHQ scores) for both men and women. Transitions out of health care also associated with poorer GHQ scores for women who previously provided high levels of care. Controlling for unobserved effects, individual data NSFH (US) The transition to care-giving for close kin (child, Marks, Lambert & parent spouse) is associated with more depressive Choi (2002) symptoms and less global happiness and feelings of personal mastery especially, for women and men. The effects are less strong for not immediate and non-kin. There are also little positive benefits on Ryff's categories either. People are not reporting growth and a purpose in life from caring for kin although there is some sign that caring for nonrelatives does have a beneficial effect for women but not men.

160

Appendix B: Summaries

Table B.4.5: Community involvement and volunteering Evidence on community involvement and volunteering Study Data (Country) Summary Controlling for other variables but not unobserved individual effects, individual level data Meier & Stutzer GSOEP (Germany) The amount of time spent volunteering is positively (2006) related to LS (although less so in the fixed effects model). Controlling for unobserved effects, individual data ACL (US) Respondents who were happier/satisfied/high selfThoits & Hewitt esteem/good health/low depression in time period 1 (2001) (1986) worked significantly more volunteer hours at time 2 (1989). However, volunteering in time period 1 and current levels of social integration explain much of the effect of well-being on volunteer work.

161

Appendix B: Summaries

Table B.4.6: Sleep Evidence on Sleep Study Correlations Kahneman, et al. (2004)

Data (Country)

Summary

DRM (US)

Poorer sleep associated with lower positive and more negative affect. For instance the difference between those who had "very bad" vs "very good" usual sleep quality was around 1.4 on the scale for enjoyment at both work and home.

Controlling for other variables but not unobserved individual effects, individual level data Ferrer-i-Carbonell BHPS (UK) Loss of sleep is associated with lower life and Gowdy (2005) satisfaction.

162

Appendix B: Summaries

Table B.4.7: Exercise Evidence on Exercise Study Data (Country) Summary Correlations Controlling for other variables but not unobserved individual effects, individual level data ACL (US) Amongst the over 60s, happiness is positively Baker, Cahalin, associated and depressive symptoms (CES-D) are Gers & Burr (2005) significantly negatively associated with the amount of time engaged in physical activities. Dockerty (unpublished)

HILDA (Australia)

Women who exercise 7 days a week are more satisfied – other amounts of exercise were nonsignificant.

Ferrer-i-Carbonell and Gowdy (2005)

BHPS (UK)

Working in the garden is associated with higher life satisfaction.

Ritchey, Ritchey & Dietz (2001)

ACL (US)

Among older people, life satisfaction and happiness were higher and depressive symptoms lower (CESD) for gardening.

163

Appendix B: Summaries

Table B.4.8: Religious practice Evidence on Religious practice Study Correlations Cohen (2002)

Kahneman, et al. (2004)

Data (Country)

Summary

GSS (US)

Amount of time spent in religious activities is associated with greater happiness for both Catholics and Protestants.

DRM (US)

All activities were associated with more positive than negative affect for this adult female sample. Leisure activities are associated with more positive and less negative emotions, working and commuting are associated with lower levels of positive affect and higher levels of negative affect. In the middle were chores (e.g. housework) and leisure activities (e.g. exercising). However, no inferential statistics are presented for this data so it is unclear how important these differences are. Poorer sleep associated with lower positive and more negative affect.

Controlling for other variables but not unobserved individual effects, individual level data ACL (US) Happiness, life satisfaction and CES-D scores were Baker, Cahalin, not associated with amount of religious activities. Gers & Burr (2005) Clark and Lelkes (2005)

ESS (Europe)

Religious attendance at least once a month is associated with higher life satisfaction.

Dehejia, DeLeire & Luttmer (2005)

NSFH (US)

Attendance at religious services is positively related to overall happiness, similarly a change in attendance frequency is positively related to changes in happiness levels. There are no significant differences between the different denominations. Being religious (measured by religious attendance) reduces the coefficient on income, significantly so for blacks. Hence religious blacks are ‘insured’ against the happiness effect of income changes.

Ferris (2000)

GSS (US)

Small effect of frequency of attendance at religious ceremonies on happiness.

Hayo (2004)

NDB (Eastern Europe)

Life satisfaction higher amongst people attend church at least weekly (compared to never). Less frequent attendance not associated with increased life satisfaction compared to never. 164

Appendix B: Summaries

Helliwell (2003)

WVS (International)

Church attendance once or more times a week associated with higher life satisfaction.

Helliwell & Putnam (2004)

WVS, ESS, SCBS (International)

Frequent attendance at religious services associated with higher happiness in WVS and Benchmark but not with higher life satisfaction in WVS or ESS.

Lee, DeMaris, Bavin, & Sullivan (2001)

NSFH (US)

Attending church is associated with fewer depressive symptoms (CES-D).

165

Appendix B: Summaries

B.5. Attitudes and beliefs towards self/others/life Table B.5.1: Attitudes towards our circumstances Evidence on attitudes towards our circumstances Study Data (Country) Summary Controlling for other variables but not unobserved individual effects, individual level data BHPS (UK) GHQ scores lower if one's financial situation is Brown, Taylor, perceived as being worse than last year and will be Wheatley Price even worse next year. (2005) Dockerty (unpublished)

HILDA (Australia)

Self assessed financial well-being show very strong effect of being very comfortable or prosperous relative to being poor or very poor. Perception of job security (type of contract was not significant), if 1050% chance of loosing job (20% of workers), negative effect for all groups (1%), and if over 50% (6% of workers), significantly negative for women.

Easterlin (2001)

GSS (US)

People typically think they were less happy in the past and will be happier in the future.

Frijters (2000)

RLMS (Russia)

Domain satisfaction at Time 2 not especially predictive of global life satisfaction once Time 1 global satisfaction is controlled for.

Gerlach & Stephan (1996)

GSOEP (Germany)

Satisfaction with health positively associated with life satisfaction for women and men under 50. However, the two are negatively associated for men over 50.

Graham & Pettinato (2001)

Latinobarometer (Latin America), RLMS (Russia), GSS (US)

Latinobarometer: Life satisfaction positively associated with perceptions of past and future social mobility and positive attitudes towards pro-market democracy and satisfaction with county's democratic processes. RLMS: People happier if they did not fear job loss, were pro market and pro democracy. GSS: Economic dissatisfaction and negative perceptions of past social mobility are both associated with lower levels of life satisfaction.

Hayo & Seifert (2003)

NDB (Eastern Europe)

Subjective economic well-being is highly correlated with life satisfaction.

Johnson & Krueger

MIDUS (US)

Perceived financial situation and control over one's

166

Appendix B: Summaries life fully mediated the effects of actual financial resources (income and assets) on life satisfaction.

(2006) Keyes (2000)

MIDUS (US)

Perceived declines in work and family relations over the last ten years are associated with lower amounts of positive and higher amounts of negative affect. However, counter to everyday thinking, improvements in functionings at work and with family are not associated with greater levels are also associated with greater levels of negative affect. This is explained by arguing that any change is unsettling and thus provides negative emotions.

Kim & McKenry (2002)

NSFH (US)

Perceived role strain (competing pressures of household, job, parenting, spouse) is associated with more depressive symptoms (poorer CES-D scores).

Louis & Zhao (2002)

GSS (US)

Poorer financial perceptions (now and in future) associated with lower life satisfaction.

Magdol (2002)

NSFH (US)

Martin & Westerhof (2003)

MIDUS (US)

Pichler (2006)

ESS (Europe)

Feeling that one's career had been sacrificed to aid partner's job mobility was associated with more depressive symptoms among women. Perceptions of finances, health and family support all positively (and relatively strongly) associated with life satisfaction. However, while finances and family support predominantly mediated the objective circumstances, objective health remained important in its own right even after accounting for perceptions. This suggest that the effects of health are not always realised. Perceived financial pressure associated with lower life satisfaction/happiness.

van Praag, Frijters & Ferrer-iCarbonell (2003)

GSOEP (Germany)

Global life satisfaction, strongly predicted by domain satisfaction including job, house, leisure, environment, financial and health all adding additional explanatory power. The main three determinants are health, job and financial satisfaction.

Wiggins, Higgs, Hyde & Blane (2004)

CASP-19 (UK)

Among over 65s, CASP scores lower if the area they lived in was thought to be "deprived", if they thought they didn’t have enough pension provisions, if recent events had been negative and the density and quality of social networks was poor.

Wildman & Jones (2002)

BHPS (UK)

Poorer GHQ scores are associated with perceptions of financial situations as currently poor, worse than last year and getting worse in the future. 167

Appendix B: Summaries

Table B.5.2: Trust Evidence on trust Study Data (Country) Summary Controlling for other variables but not unobserved individual effects, individual level data Helliwell (2003) WVS (International) Social trust (most people can be trusted) and beliefs about the wrongness of cheating on taxes both positively associated with life satisfaction.

Helliwell (2006)

WVS (International)

Higher social trust associated with higher life satisfaction and lower probability of suicide.

Helliwell & Putnam (2004)

WVS; ESC, SCBS (International)

Social trust and trust in police are positively related to life satisfaction (WVS, ESC) and happiness (WVS, Benchmark). Trust in neighbours positively associated with life satisfaction (ESC) and happiness (Benchmark) - [not measured in WVS].

Hudson (2006)

Eurobarometer (Europe)

Higher life satisfaction associated with greater trust in government, police, legal system, the European Central Bank, and business. No effect on life satisfaction of trust in EU, radio, unions, voluntary and non-governmental organisations and United Nations. Life satisfaction associated with higher relative trust (difference of individual from country mean), for the government, ECB, legal system, and the UN.

168

Appendix B: Summaries

Table B.5.3: Political persuasion Evidence on political persuasion Study Data (Country) Summary Controlling for other variables but not unobserved individual effects, individual level data Unemployment negatively affects happiness for US Alesina, Di Tella GSS (US); right wingers more than US left wingers. The life and MacCulloch Eurobarometer satisfaction of European left winger's is more (2004) (Europe) negatively affected by income inequality than European right wingers. Ferrer-i-Carbonell and Gowdy (2005)

BHPS (UK)

People have higher life satisfaction if they care about animal extinction but lower LS if they worry about the ozone layer.

Graham & Pettinato (2001)

Latinobarometer (Latin America), RLMS (Russia),

Latinobarometer: Life satisfaction positively associated with positive attitudes towards promarket democracy and satisfaction with county's democratic processes. RLMS: People happier if they were pro market and pro democracy.

Pichler (2006)

ESS (Europe)

Collapsed life satisfaction/happiness scores lower when attitudes towards politicians are negative, higher when attitudes are positive.

169

Appendix B: Summaries

Table B.5.4: Religious beliefs Evidence on Religious Beliefs Study

Measure/Data Summary Source (Country) Controlling for other variables but not unobserved individual effects, individual level data Baker, Cahalin, ACL (US) Being religious had no effect on life satisfaction, Gers & Burr (2005) happiness or depressive symptoms. Clark & Lelkes (2005)

BHPS (UK) & ESS (Europe)

Being catholic or protestant associated with higher life satisfaction in Europe. ESS: Interactions of religion and unemployment suggests that found being Catholic and to a lesser extent Protestant, reduces the negative effect of unemployment by about half. BHPS: Religious belief seems to be an 'insurance' against unemployment (coefficient on the interaction term between church and unemployment is significantly positive). Churchgoing divorcees report lower levels of life satisfaction, significantly for women in the conditional, fixed effects logit model.

Cohen (2002)

GSS (US)

No differences in happiness across religion (Jews, Catholics & Protestants). For Jews, the low Ns made interpreting the non-significant findings problematic. For Catholics, happiness was higher as a function of stronger religious beliefs, greater spirituality, more coping though God and amount of time spent in public practice. For Protestants, higher spirituality, coping, public practice and this time congregational support were all associated with greater happiness. Religious beliefs were uncorrelated for protestants.

Dockery (unpublished)

HILDA (Australia)

Importance of religion positively related to life satisfaction.

Ferris (2002)

GSS (US)

Not much difference in happiness as a function of specific religion (Jews, Catholics & Protestants) but all seemed to be associated with higher happiness than no religion. Doctrinal differences have a stronger impact on happiness than denominational preference, with happiness being especially high among evangelists and fundamentalists.

Flouri (2004)

BHPS (UK)

Being religious at age 42 is associated with better GHQ scores but no significant effects on life satisfaction (although the relationship is positive).

170

Appendix B: Summaries Haller & Hadler (2006)

WVS (International)

Church attendance seems important for both happiness and life satisfaction but actual denomination seems less important. In particular there are strong degrees of variance within each religion especially in relation to life satisfaction, although there do seem to be some advantages for protestants, though these tend to dominate in the richest countries.

Kahneman et al. (2004)

DRM (US)

Praying/worship associated with high levels of positive affect and low levels of negative affect. Since praying is essentially a solitary activity this suggests that the effects are not due to social interaction.

Smith (2003)

GSS (US)

Being religious is associated with more happiness than being atheist in the US but there was no effect in the former USSR and both were associated with greater happiness in Germany. Its not clear what this last finding means.

Hayo (2004)

NDB (East Europe)

Compared to catholics, life satisfaction not affected by religious denomination (protestant, orthodox, other, atheist).

Helliwell (2003)

WVS (International)

Life satisfaction higher if "god is important in person's life".

Helliwell (2006)

WVS (International)

Helliwell & Putnam (2004)

WVS, ESS, SCBS (International)

Belief in god is associated with higher life satisfaction and a lower probability of suicide. Importance of god/religion associated with higher life satisfaction and happiness in the US, but no effects on life satisfaction in Europe.

Lelkes (2006)

ESS (Europe)

Being religious is associated with higher life satisfaction and a significant marginal effect.

Rehdanz & Maddison (2003)

World database of happiness (International)

No effects on happiness of religion with respect to the proportion of the population who are buddist, hindu, muslim, christian, jew or orthodox, compared to atheist.

171

Appendix B: Summaries

B.6. Relationships Table B.6.1: Marriage/intimate relations Evidence on Marriage & Intimate Relations Study

Measure/Data Source (Country)

Summary

Correlations Controlling for other variables but not unobserved individual effects, individual level data GSS: Being married was associated with higher Alesina, Di Tella GSS (US); happiness than being divorced, separated or and MacCulloch Eurobarometer widowed. (2004) (Europe) Eurobarometer: Being married was associated with higher life satisfaction than being divorced, separated or widowed. Biblarz & Gottainer (2000)

GSS (US)

Looking at the well-being of adults whose childhood was affected by divorce and death of a parent. Coming from a divorced single-parent family is associated with lower happiness than either both parent or single parent families due to widowhood. However, subsequent remarriage of a parent seems harder on those whose natural parent died than was divorced.

Blanchflower & Oswald (2004)

Eurobarometer (UK data); GSS (US)

Eurobarometer: Being married was associated with higher life satisfaction than being divorced, separated or widowed. Living as if married lower than being married but higher than other states. GSS: Being married once associated with greater happiness than being divorced, separated or widowed and also married twice or more. Happiness was also lower if parents had been divorced by age 16. There is no difference in reported happiness between those who did not have sex at all in the last year and people who had sex less than once or twice a month. However, happiness was higher amongst those who had sex weekly or more often. Happiness was also higher if it was with the same partner.

Blanchflower and Oswald (2005)

ISSP (International)

Being married was associated with greater happiness than being divorced, separated, widowed or single.

Brown (2000)

NSFH (US)

Depressive symptoms (CES-D) are higher among people cohabitating than married. Prior experiences of marriage or cohabiting are also associated with more depressive symptoms of cohabitors/marrieds

172

Appendix B: Summaries than if the person was previously single. However the main pattern appears to be moderated by relationship stability - unstable relationships are more associated with cohabitating and it seems to be this instability that is the cause of depressive symptoms since if stability is maintained it doesn't matter. Longitudinal analysis further suggested that the differences weren't due to selection effects but the nature of cohabitating relationships. Clark (2003)

BHPS (UK)

Compared to being single, being separated and widowed is associated with worse GHQ scores but being married and divorced does not make a difference. The same is found for life satisfaction except for the fact that life satisfaction is higher for being married.

Clark and Lelkes (2005)

ESS (Europe)

Being married was associated with higher life satisfaction than being divorced, separated, widowed or never being married.

Clark and Oswald (1994)

BHPS (UK)

Compared to being single, being separated and divorced is associated with worse GHQ scores but being married and widowed does not make a difference. Note the large error term for widowed reflecting large individual differences.

Clark and Oswald (2002)

BHPS (UK)

Calculate (according to GHQ scores) that a shift from being married to: separated is equivalent to a loss of £8000 (per year at 1992 prices), to divorced is equivalent to a loss of £1000, to widowed is equivalent to a loss of £7000. The equivalent losses tend to be even higher if one looks at just the happiness question in the GHQ.

Dehejia, DeLeire & Luttmer (2005)

NSFH (US)

Compared to being married, being separated, divorced, widowed or never married is associated with lower happiness. Transitions into these states also lowers happiness.

Di Tella MacCulloch & Oswald

Eurobarometer (Europe inc. UK), GSS (US)

Compared to being single (?) being married is associated with higher happiness in the US and higher life satisfaction and happiness in Europe and the UK. The reverse if found for separation and widowhood. However, divorce is found to have a significantly negative effect on life satisfaction in the UK and happiness in Europe but not for happiness in the US or life satisfaction in Europe.

Dockery

HILDA (Australia)

Both men and women had higher life satisfaction if

173

Appendix B: Summaries married.

(unpublished) Fayey and Smyth (2004)

EVS (Europe)

Being married associated with higher life satisfaction than being divorced, widowed or single.

Ferrer-i-Carbonell (2005)

GSOEP (Germany)

Living together associated with higher happiness (presumably than not living together).

Ferrer-i-Carbonell & Frijters (2004)

GSOEP (Germany)

Living with partner associated with greater happiness in OLS model but not Ordered logit.

Ferrer-i-Carbonell and Gowdy (2005)

BHPS (UK)

Being married or living together associated with higher life satisfaction.

Flouri (2004)

NCDS (UK)

Having a partner is associated with better GHQ scores, higher life satisfaction and less malaise.

Frijters, HaiskenDeNew & Sheilds (2004)

GSOEP (Germany)

Compared to being not married, there was no difference in life satisfaction amongst men or women who were married, separated, divorced or widowed.

Frey & Stutzer (2000)

SHPS (Switzerland)

Life satisfaction lower for both single men and women.

Gerdtham & Johannesson (2001)

SLLS (Sweden)

Being single and living alone (as opposed to not single) associated with lower daily life satisfaction.

Gerlach & Stephan (1996)

GSOEP (Germany)

Fixed effects model with period effects finds marriage significantly positive for men and women

Suggest Documents