Discussion Paper No. 69 Research Institute for Labour Economics and Labour Law

Working-Time Policies in Switzerland: An Analysis of Desired Working Time, Overtime, and Hours Constraints of Swiss Salaried Employees

Alfonso Sousa-Poza Department of Economics and Research Institute for Labour Economics and Labour Law at the University of St. Gallen, Switzerland

Fred Henneberger Department of Economics and Research Institute for Labour Economics and Labour Law at the University of St. Gallen, Switzerland, Department of Administrative Science, University of Konstanz, Germany

Abstract: In this paper, desired working time, overtime, and hours constraints of Swiss salaried employees are analysed with data from the 1998 Swiss Labour Force Survey and with data from the 1997 International Social Survey Programme. It is shown that (i) most Swiss workers do not face hours constraints, when we compare the desired working time with the contractual working time; (ii) the largest portion of constrained workers would like to work less and earn less (17% of all male and 22% of all female salaried employees); (iii) there is a deficit of part-time jobs for both men and women, especially for high-qualified individuals; (iv) men and women work on average 3.15 and 1.69 hours overtime per week, respectively; the most common form of compensation for this overtime is holidays, followed by no compensation and money; (v) a comparison between actual and desired working time reveals that a significant amount of workers are constrained in the sense that they would prefer shorter actual working hours; (vi) an international comparison of desired working time shows that workers in Switzerland are quite satisfied with their current workload/pay combination; Switzerland does, however, have the highest portion of workers wanting to work and earn less. Some policy implications of these findings are also discussed.

1 1. Introduction The debate over the length of the workweek has re-entered the public sphere in most industrialised countries. There are two main reasons for this renewed concern: first, the persistent unemployment in many of these countries has led to an increased interest in "worksharing" policies in order to distribute the available work more equitably. It is commonly argued that a reduction of the working time will increase the number of available jobs in an economy and that, thereby, unemployment can be reduced or prevented.1 A second reason is the willingness of families to divide paid and unpaid employment more equally among family members. By reducing the working time, men can devote more of their time to the running of the household and women can participate in the labour market. In this century, working time has declined substantially in all industrialised countries, and, currently, the most commonlyreported workweek in OECD countries is 40 hours (see OECD, 1998a, p. 153). Nevertheless, this decline has slowed significantly in recent decades in almost all OECD countries (the exceptions being Germany, Japan, and the Netherlands). In a few countries, this decline appears to have stopped, while, in others, there has even been an increase in hours (most notably the United States). A further characteristic of the current workweek is that there has been a growing diversity in the hours worked by employees, with the proportion of employees working 40 hours having fallen considerably. The debate over the length of the workweek has also entered the public sphere in recent years in Switzerland. The largest labour union umbrella organisation, the Swiss Federation of Trade Unions (Schweizerische Gewerkschaftsbund, SGB) has recently launched a popular referendum ("Volksinitiative") for a nation-wide introduction of an upper limit of 1872 work hours per year, which corresponds to a 36-hour workweek (see SGB, 1999, especially p. 27). Furthermore, this de facto reduction of the workweek would take place without significant changes in earnings.2 At this point in time, it is difficult to judge whether

1

The most ubiquitous examples being France and Italy, in which the normal workweek fell from 39 hours to 35 hours by legislation. See also OECD (1998a), p. 174. Whether new jobs have been created or not is yet unclear. See Freeman (1998), p. 29, OECD (1998a), p. 182, and, for Germany, Henneberger (2001). Recently, the media has also reported industrial action by French truckers in protest to a reduced workweek.

2

According to the referendum text, all employees who earn less than 1.5 times the average wage should not face a reduction in earnings (proposed article 24, paragraph 2 of the constitution). Due to the relative unequal distribution of income in Switzerland, this covers the largest portion of all Swiss employees.

2 the voters will accept the proposals in this referendum.3 The Swiss employers associations generally oppose such measures (see, for example, Schweizerischer Arbeitgeberverband (AGV), 1998; Clemmer, 1997). The Swiss Federal Railways (Schweizerische Bundesbahn, SBB) have also recently introduced a 39-hour workweek (as in the SGB referendum, based on a yearly upper limit of the working time), which meant a reduction of two hours per week for approximately 30,000 employees.4 The Swiss Post offers a further example of how working time reductions are trying to be introduced in Switzerland. In a pilot project in four cantons, the Swiss Post is employing unemployed individuals as part of a job-sharing scheme. For groups of three currently employed individuals, an unemployed person is given a job, and the workload is divided among the four. The working time is thereby reduced by approximately 25%.5 Finally, the Swiss telecommunications company (Swisscom) has recently, as part of a pilot project in three cantons, introduced a four-day workweek, which corresponds to a 36hour workweek.6 There are two main reasons for this resurgent interest in working time reductions in Switzerland (see also Henneberger, Graf and Sousa-Poza, 1999). First, in the 1990s, Switzerland faced its highest levels of unemployment since the 1930s, and, therefore, "job sharing" policies and working time reductions have become popular again. A second reason is that Switzerland has one of the highest weekly and annual working times in the industrialised world and also one of the highest standards of living. Thus, one could argue that the marginal value of leisure is relatively high. In Switzerland, the most commonly reported (contractual) workweek is currently 42 hours followed by 40 hours per week, and these figures cover over 70% of all employed individuals. Compared to its neighbouring countries, Switzerland therefore has a very long workweek.7

3

The SGB did, however, have problems in collecting the required signatures for a referendum (see Neue Zürcher Zeitung, 7th of January 1999). In fact, there even appears to be recent evidence that the SGB itself is giving this topic less importance (see Neue Zürcher Zeitung, 14th of January 2000).

4

See also Neue Zürcher Zeitung, 22nd of April 1999. It is also interesting to note that the Federal Government has no intention of reducing the working time for its employees (see Eidgenössisches Personalamt (EPA), 1999). The Federal Government does, however, offer individual working-time reductions (see Hablützel and Rebecchi, 1998).

5

See Graf, Henneberger and Schmid (2000), p. 45f.

6

See also St. Galler Tagblatt, 31st of January 2000.

7

In Italy, the most frequently reported workweek is 40 hours (covering 51% of the active labour force), in Austria, also 40 hours (covering 55% of the active labour force), in Germany, 38 hours (covering 32% of the active labour force), and, in France, 39 hours (covering 55% of the active labour force). See OECD (1998a), p. 157.

3 What has received surprisingly little attention in this debate on working-time reductions is the preferences that workers themselves have with regard to the length of the workweek. This is especially the case in Switzerland.8 In our opinion, however, understanding workers’ willingness to change their working time and the extent to which workers are constrained (i.e., for which desired working time deviates from actual working time) is a prerequisite in any discussion of working-time policies. In fact, since a large portion of Swiss voters are employed, knowing the extent to which these workers desire shorter working hours could give us a rough indication of how the SGB referendum will fare.9 The general aim of this paper is to analyse Swiss employees’ working-time (and thus, earnings) preferences.10 In section 2, we first take a brief look at the characteristics of the Swiss labour market. In this paper, we use data from two interesting data sets: the 1998 Swiss Labour Force Survey (SLFS) and the 1997 International Social Survey Programme (ISSP). These data sets are discussed in section 3. In section 4, we take a look at contractual and desired working time. More specifically, we analyse the extent to which desired working time corresponds to the contractual working time. In section 5, we extend our analysis by comparing desired with actual working time where actual working time is defined as contractual working time plus usual overtime. In this section, we also analyse the determinants of overtime and the determinants of the form of compensation for overtime. In section 6, a cross-national comparison of hours constraints is undertaken with the ISSP data set. Section 7 concludes with a few policy implications of our findings.

8

A number of studies exist that analyse desired working time in other countries (see, for example, Kahn and Lang, 1992 for the United States; Kahn and Lang, 1996 for Canada; Bell and Freeman, 1995 for the United States and Germany; Bundesmann-Jansen et al. 2000, Bauer et al. 1994 and 1996, Klauder, 1998 and Dathe, 1998 for Germany; Sousa-Poza and Henneberger, 2000, 2001 for several countries).

9

Although one must take into consideration that our subsequent analysis is based on reported desired working-time data, which assumes that changes in working time are associated with corresponding changes in earnings. As pointed out above, the SBG referendum, however, does not, in general, foresee such a change.

10

In this paper, we do not analyse the reasons for the existence of hours constraints. This is in itself an important topic since the existence of hours constraints is often not easy to explain, and, in traditional economics, they are often assumed not to exist. The interested reader is referred to Lang and Kahn (2000).

4 2. The Swiss Labour Market: An Overview Switzerland has recently experienced its most severe economic recession since the 1930s. This was most notably felt by the high level of unemployment, which reached its maximum in 1997 with an unemployment rate of approximately 5.2%. As was mentioned above, one potential reason for this increased interest in working-time policies could have been caused by this surge in unemployment (the unemployment rate in 1990 was below 1%). In table 1, a few summary statistics for the Swiss labour market are presented (see also Birchmeier, 2001). Table 1: some descriptive statistics of the Swiss labour market males

females

total

participation ratea

87.2

71.8

79.7

full-time employmentb (in %)

92.3

53.5

part-time employment (in %)

7.7

46.5

average contractual weekly working time (hours)c

40.3

29.5

35.1

median contractual weekly working time (hours)c

42.0

34.0

41.0

unemployment rated

2.4

3.3

2.7

71,500

54,600

66,000

median income (SFr. per year; gross)e a

1999; as a percentage of individuals between the age of 15 and 64 defined as at least 30 hours working time per week (1999) c based on the 1998 SLFS (own calculations); only employees who work at least one hour per week and are at least 15 years old. d 1999 e 1998 Sources: 1998 SLFS, Bundesamt für Statistik (1999a), OECD (2000), Staatssekretariat für Wirtschaft (2001) b

The Swiss labour market also has, compared to other industrialised countries, a few unique characteristics: •

It has the highest participation rate in Europe. In 1999, for example, about 80% of individuals between the age of 15 and 64 were employed. In the European Union (EU), the corresponding figure was approximately 63% (see OECD, 2000).

•

The female participation rate (defined as a percentage of all females between the age of 15 and 64) is also remarkably high with a rate in 1999 of about 72%. The corresponding EU average was only 53% (see OECD, 2000).

5 •

As was mentioned above, Switzerland has one of the longest workweeks in Europe and also the highest annual working time, i.e., if one includes vacations and public holidays (see IW, 1999, p. 153).

•

Another interesting characteristic is that the Swiss labour force has a very high portion of part-time jobs. In Europe, only Holland has a higher portion (see Schaufelberger, 1997, OECD, 2000).

•

Switzerland also has very few labour disputes and, in an international setting, very few lost days due to strikes (see IW, 1999, p. 154).

•

According to the OECD, Switzerland has the third highest disposable GDP per capita among 29 industrialised countries (see OECD, 1998b).

3. Data In order to analyse employees’ working-time preferences in Switzerland, we use data from the Swiss Labour Force Survey (SLFS). The SLFS is a nation-wide, representative survey conducted annually by the Swiss Federal Statistical Office (Bundesamt für Statistik, BFS). During telephone interviews lasting approximately 20 minutes, individuals (including foreigners) are questioned on a number of labour-market related topics. The first SLFS survey was conducted in 1991, and the sample size is approximately 16,000 individuals (see BFS, 1996). One of the advantages of the SLFS, in the context of this study, is that it has very precise information on desired working time. The wording of the relevant question, posed to all currently employed individuals11, is as follows: "How many hours per week would you like to work assuming that your wage is adjusted proportionally?" One drawback of this data, however, is that not all workers were posed this question. More specifically, the following workers were not set this precise question: (i) employees who work full-time and would not prefer to work part-time; (ii) certain part-time workers who would like to work full-time. In the former case, it was assumed that these full-time workers are not constrained, i.e., their actual working time corresponds to their desired working time, and, in the latter case, these part-time employees were assumed to want to work 42 hours per week. Naturally, assuming that these full-time workers are not constrained need not always apply, since they may, for

11

An analysis of hours constraints faced by non-employed individuals would also merit investigation. This would especially be interesting for women since it is often claimed that (frequently well-qualified) women remain non-employed due to the fact that there is a deficiency of part-time jobs in the economy.

6 example, be working several hours overtime and therefore their current full-time status may not correspond to their desired working time.12 We restrict our analysis to salaried employees. The sample size is equal to 7,822 individuals, of which 4,095 are men and 3,727 are women. In section 6, we make an international comparison using data from the 1997 International Social Survey Programme (ISSP). The ISSP is a continuing annual programme of cross-national collaboration, which started with the first survey in 1985. The data for the ISSP are collected by independent institutions in several countries. The topics change from year to year by agreement with a view to replication approximately every five years. The ISSP’s official data archive is the Zentralarchiv at the University of Cologne, Germany. The topic of the 1997 survey is "Work Orientations", and covers issues on general attitudes toward work and leisure, work organisation, work contents, collective interests, and second jobs.13 The wording of the question related to desired working time (which we use in this paper) is as follows: "Think of the number of hours you work and the money you earn in your main job, including regular overtime. If you only had one of these three choices, which of the following would you prefer? (i) Work longer hours and earn more money; (ii) Work the same number of hours and earn the same money; (iii) Work fewer hours and earn less money". Thus, compared to the SLFS data on desired working time, the ISSP data is less precise.14 We analyse a sample of full and part-time workers in the following 14 industrialised countries: Germany, Denmark, France, Great Britain, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, and the United States.

12

In the first five waves of the SLFS (i.e., until 1995), all workers were questioned on their desired working time. Thus, those surveys have better data on desired working time than the corresponding data in the 1998 survey. Since the aim of this study, however, is to provide up-to-date information about desired working time, the 1998 survey is analysed, despite this shortcoming. The 1995 SLFS data have been analysed in Ramirez (1998) and in Sousa-Poza and Ziegler (2001). It must be noted, however, that the use of the more detailed information (i.e., data until 1995) on desired working time does not fundamentally change our main conclusions in this paper.

13

Official documentation on the 1997 ISSP is available from the Zentralarchiv at the University of Cologne, Germany.

14

We are not aware of many studies that have such precise information on desired working time as is available in the SLFS. Most studies apply data that are based on a discrete-type question as in the ISSP data set. See, for example, Altonji and Paxson (1988), Ilmakunnas and Pudney (1990), Kahn and Lang (1992), Stewart and Swaffield (1997). Exceptions (i.e., studies that also use precise data on desired working time) include Kahn and Lang (1995) for Canada, Bundesmann-Jansen (2000), Bauer et al. (1996, 1994) for Germany, and Euwals and van Soest (1999) for Holland.

7 Data on desired working time do have some problematic characteristics. The main problem with such data is that they are based on hypothetical questions. As with most such subjective data, one could question its reliability and validity. It should, however, be noted that there is ample evidence that such data is not purely "noise", i.e., it correlates with observable actions such as job mobility and labour supply (see, for example, Ham, 1982; Altonji and Paxson, 1986; Euwals, 1999). In section 5, we do discuss some problematic aspects of the SLFS data on desired working time.

4. Desired vs. Contractual Working Time in 1998 Table 2 depicts the portion of salaried employees in 1998 that would like to change their contractual working time, and also the extent of the desired change. As can be seen, however, the large majority of workers do not face hours constraints (81.4% and 64.3% of males and females, respectively). Of those workers that are constrained, the majority would like to work less and earn less. This is especially the case in the male sample. Table 2: portion of workers wanting to change their contractual working time in 1998 and to what extent > 10 hrs less

6-10 hrs less

1-5 hrs less

not

1-5 hrs more

6-10 hrs more > 10 hrs more

no. obs.

per week

per week

per week

constrained

per week

per week

per week

males

3.9%

9.0%

3.6%

81.4%

0.7%

0.6%

0.7%

4095

females

5.0%

9.3%

7.8%

64.3%

6.0%

3.6%

3.9%

3727

total

4.4%

9.2%

5.6%

73.3%

3.2%

2.1%

2.2%

7822

The relationship between desired and contractual working time is shown in figures 1 to 3. In figures 1 and 2, distributions of desired and contractual working time are plotted for females and males, respectively. These figures clearly show that women are more constrained than men, and the most predominant difference is the "excess demand" of jobs ranging from 15 to 38 hours per week.15 This is most notably the case for women working between 20 and 38 hours per week. In figure 2, one notes that men’s desired working time mirrors their contractual working time remarkably well. There is, however, a notable difference between

15

It is interesting to note that, despite the fact that women are less satisfied with thier working-time/wage combination, they often have higher job-satisfaction levels than men. This is also the case for Switzerland (see Sousa-Poza and Sousa-Poza, 2000).

8 desired and contractual working time in the male sample at around 30 to 38 hours per week. More specifically, the desire to work 30 to 38 hours per week is more common than the number of contracts offered in this range. Figure 1: distribution of contractual and desired working time for women

percent

20

10

contractual working 0

desired working time 1

6

11

16

21

26

31

36

41

46

51

56

contractual/desired working time (hrs./week)

The fact that, in both the male and female samples, there is an apparent deficiency of parttime jobs does speak in favour of the often-proposed introduction of flexible part-time work with a work rate varying between 70% and 90% of a full-time job (and with an appropriate adjustment of the hourly wage rate). In a survey conducted by the State Secretariat for Economic Affairs (Staatssekretariat für Wirtschaft, seco), about 57% of all Swiss firms have a good opinion of such flexible part-time contracts (see Blum and Zaugg, 1998, p. 180). Although there still appears to be a deficiency of part-time jobs, there has, in the past years, been a gradual increase in the number of offered part-time jobs. In 1999, approximately 41,000 new part-time jobs were created, approximately 60% of which are occupied by women (see BFS, 1999b, p. 40 and BFS, 1998, p. 38).16 As was mentioned above, Switzerland has, in an international comparison, one of the largest portions of part-time jobs (see Schaufelberger, 1997).

16

The fact that about 40% of the new part-time jobs were filled by men is an interesting observation. Between 1997 and 1998 the corresponding figure was only 20%.

9 Figure 2: distribution of contractual and desired working time for men 40

percent

30

20

10 contractual working 0

desired working time 1

6

11

16

21

26

31

36

41

46

51

56

contractual/desired working time (hrs./week)

In figure 3, the difference between desired and contractual working time is plotted. In the female sample, we observe that the desire to work more than is foreseen in the contract is most pronounced at weekly working hours below 26 hours, whereas the desire to work less is most favoured by women who work between 36 and 46 hours per week. It also appears to be the case that women who work less than about 10 hours per week are the most likely to want to work more. The same applies to women who have an actual working time between 17 and 25 hours per week. In the male sample, the difference between the contractual and desired working time more or less fluctuates around zero. Only males working more than 38 hours per week seem to prefer being employed a few hours less.

10 Figure 3: differences between contractual and desired working time for men and women

desired-contractual working time (hrs./week)

8 6 4 2 0 -2 -4 -6

males

-8

females 1

4

7

10

13

16

19

22

25

28

31

34

37

40

43

46

49

contractual working time (hrs./week)

In order to analyse the determinants of hours constraints in a multivariate way, we estimate regressions for males and females. In table 3, the results of an ordered probit model are presented, for which the dependent variable can take on seven possible values corresponding to the categorisation in table 2, i.e., ’0’ for workers wanting to work more than 10 hours less per week, up to ’6’ for workers wanting to work more than 10 hours more per week. Thus, this variable captures the magnitude and direction of hours constraints among employees. Several independent variables are included in the regressions: six demographic variables, five working-time characteristics, three other work-related characteristics, and twelve dummy variables characterising different sectors.17

17

Since, in Switzerland (as in many other industrialised countries), working conditions are negotiated on a sectoral level, the inclusion of the twelve dummy variables captures sector-specific differences. For the sake of clarity, however, the estimated coefficients for these variables have been omitted in the subsequent tables.

11 Table 3: determinants of hours constraints - estimated coefficients and marginal effects of the ordered probit model males females MEb

coef. constant demographic variables marrieda childrena high educationa low educationa age×10-1 age2×10-3 working-time characteristics works on weekenda works at nighta works block timesa flexible working timesa works shiftsa other work-related characteristics tenure tenure2×10-3 is in managementa small firm (< 100 employees)a number of observations log likelihood pseudo-R2

3.135**

MEb

coef. 1.541**

0.042 0.067 -0.232** 0.262** -0.680** 0.769**

0.008 0.013 -0.047 0.052 -0.136 0.200

0.023 0.268** -0.153** 0.107* -0.027 0.156

0.002 0.023 -0.013 0.009 -0.002 0.122

0.081 0.014 -0.076 0.034 -0.038

0.016 0.028 -0.015 0.007 -0.008

0.013 0.081 -0.090 0.286** -0.097

0.001 0.007 -0.008 0.024 -0.008

0.009 -0.068 -0.049 -0.013

0.002 -0.123 -0.010 -0.003

-0.021** 0.255** -0.189** -0.064

-0.002 0.212 -0.016 -0.005

4037 -2784 0.026c

3670 -4428 0.034c

Note: The dependent variable can have seven possible values ranging from "work more than 10 hours less per week" to "work more than 10 hours more per week" (see also table 2). Furthermore, the regressions included twelve dummy variables for different sectors, which, for the sake of clarity, are not presented here. The reference industry is the banking and insurance sector. a dummy variables b ME = marginal effect calculated at a score equal to 3 c the pseudo-R2 measure is that of McFadden (1973) */** significant at the 5%/1% level, respectively

Besides presenting the estimated coefficients, we also show the marginal effects calculated at a score equal to ’3’.18 These marginal effects therefore show the effect that an explanatory variable has on the probability of being unconstrained. In the male sample, one notes that, in general, middle-aged men with a high education19 would most likely want to work less than they are contractually required to. The relationship between the dependent variable and age is

18

The marginal effects (calculated at the sample means) are defined as follows: where φ is the standard normal probability density function.

19

[ ] = φ(β′x )× β , and ∂x

∂E y x

Degrees from the following institutions were considered to be "high education": university, technical college ("höhere Fachschule", "Technikon"), and high school ("Matura", "Diplommittelschule"). The following categories were considered to be "low education": no degree (i.e., still in compulsory schooling), only compulsory schooling, and lower apprenticeship schemes ("Anlehre", "Haushaltslehrjahr"). The

12 a parabolic one, with its minimum at around 44 years. Thus, as age increases to 44 years, the probability that individuals report wanting to work less than they are contractually required to increases; thereafter the probability declines again.20 Stated somewhat differently, men in this age group are, ceteris paribus, the most willing to reduce their working time. Men with a low education are generally more likely to report wanting to work more, and the marginal effect shows that these individuals are less constrained than workers with a medium education (reference group). Women with a high education, women with higher tenure21, and women in management positions22 have a tendency to prefer to work less. Women with children, women with a low education, and women with flexible working times prefer to work more than stipulated in their contract.23 A look at the marginal effects also reveals that these women have a lower probability of being constrained. It is, therefore, interesting to note that women with flexible working times are less constrained than women who do not have the freedom to regulate their working times. This is a result that one could expect. The level of education has the same effects as in the male sample; the magnitude of these effects is, however, much smaller in the female sample. The coefficients of the education variables thus imply that there is a deficiency of part-time jobs for high-qualified individuals, especially among males.24 It should finally be noted that the explanatory power of these models is quite small, implying that a large degree of variation remains unexplained. This is a standard result in these kinds of models.

reference category was primarily made up of apprenticeships ("Berufslehre") and similar qualifications ("Vollzeitberufsschule", "höhere Berufsausbildung"). 20

The marginal effects show, on the other hand, that the probability of not being constrained is at its minimum at around 34 years of age, i.e., men in this age group are most likely to be contrained.

21

The underlying question for the tenure variable is as follows: "How many years have you been working for your current employer?"

22

The management-position variable is defined rather broadly. The exact wording is: "Do you have a management position?". About 20% of the female sample has such a position.

23

The survey question associated with the variable "flexible working times" is worded as follows: "How are your working times regulated in your main job?" The respondents then had three options: (i) fixed working times; (ii) block working times ("gleitende Arbeitszeit"); (iii) fully flexible working times. Totally flexible working times imply that the employee can plan his or her own working day. Running working times imply that the worker has to be in his or her office at certain times of the day (e.g., between 8 and 10 in the morning). Fixed working times are the most common and imply that the worker cannot, in any way, choose his or her daily working time.

24

In a study on women's working-time preferences in Germany it is shown that there also exists a deficiency of part-time jobs for high-qualified individuals. See Beckmann and Kempf (1996).

13 We can therefore conclude this section by stating that: (i) the overwhelming majority of both male and female workers appears to be satisfied with their contractual working time; (ii) there seems to be a deficit of part-time jobs for women, especially for jobs with around 20 hours per week; (iii) there is a deficiency of part-time jobs for high-qualified individuals and part-time jobs in the range between 30 and 38 hours per week; (iv) women with very low contractual working times (especially less than 10 hours per week) would like to work more; (v) a few socio-economic and demographic variables have an influence on hours constraints, most notably the level of education: especially middle-aged men with a high education as well as highly qualified women prefer working less than they are contractually obliged to. The contrary seems to apply to individuals, most notably women, at lower working-time levels, which, in turn, are associated with lower income levels.

5. Desired vs. Actual Working Time in 1998: The Role of Overtime In the above discussion, we analysed the relationship between contractual and desired working time. Our daily experience tells us, however, that actual working time may deviate from contractual working time since most employees work non-trivial amounts of overtime. In table 4, overtime statistics for Swiss salaried employees in 1998 are presented.25 Table 4: overtime in 1998 - summary statistics (hours per week) mean no. obs.a

standard deviation

Total hours per week

males

3782

3.15

4.60

4’365’994

females

3416

1.69

3.27

1’802’365

total

7198

2.47

4.10

6’168’359

a

based on the whole sample, i.e., including employees that do no overtime

As can be seen, men work, on average, more overtime hours than women (3.15 compared to 1.69 hours per week).26 Taken over the whole population, men (women) work approximately

25

The wording of the relevant question is as follows: "Taken over the past 12 months, how many hours per week did you work in excess of the number of hours stipulated in your contract?" Thus, this variable measures the overtime usually worked in a week (and not necessarily the actual overtime worked in, say, the previous week).

26

A total of 56% of all salaried employees do overtime work on a regular basis (males 65% and females 47%). These results show that, on average, the demand of the SGB referendum to limit the number of overtime hours to 100 per year is, currently, clearly not being met. According to a recent representative

14 4.4 (1.8) million hours overtime per week. Thus, in total, over 300 million hours overtime are performed each year (based on 1998 data). It therefore comes as no surprise that overtime plays an important role in any discussion on working time policies. Overtime, however, can take on many different forms. In general, one can distinguish between "transitional" overtime, which is compensated with additional free time, and "permanent" overtime which is paid out or not compensated in any way. In table 5, we show the form of compensation for overtime. A total of 59.8% (57.9%) of all males (females) that usually do overtime get compensation in the form of additional holidays. Only 10.6% (16.8%) of males (females) receive payment for their overtime. Finally, 29.6% (25.3%) of males (females) received no compensation. It is interesting to note that forms of compensation differ between genders: males are more likely than females to receive no compensation or compensation in the form of free time, whereas females are more likely than males to receive payment for their overtime.27 These differences between genders are, needless to say, primarily a result of segregated labour markets. Table 5: forms of compensation for overtime in 1998 - summary statistics percentage no. obs.a none money holidays none

hours per week money

holidays

males

2454

29.6%

10.6%

59.8%

6.50

5.70

3.82

females

1610

25.3%

16.8%

57.9%

4.44

3.96

3.06

total

4064

27.9%

13.1%

59.0%

5.76

4.81

3.52

a

based on a sample of salaried employees that reported their form of compensation and usually do overtime

In table 5, we also depict the average number of hours overtime according to the different forms of compensation. We note that, in both the male and the female samples, employees that do not get compensation for their overtime also work more overtime than workers who get some form of compensation. Another interesting result is that, although women are more likely to receive paid compensation, they, on average, work less overtime than the corresponding group of men (3.96 as opposed to 5.70 hours per week). Figure 4 depicts the distribution of overtime according to gender. Compared to men, a larger portion of women do

study for Germany, a strikingly similar result was obtained, i.e., 56% of all employees are confronted with regular overtime (see Groß/Munz, 1999). 27

Below we show that primarily white collar workers do unpaid overtime. Thus, it comes as no surprise that females are more likely to receive payment for their overtime. Note, however, that, on average, women work less paid overtime than men. This can easily be deduced from the results in tables 4 and 5.

15 shorter overtime hours per week. Men, on the other hand, are more likely to spend longer hours per week on overtime. Figure 4: distribution of overtime for men and women 60

50

percent

40

30

20

10

males

0

females 0

5

10

15

20

30

42

overtime (hrs./week)

In table 6, we present the results of probit and tobit regressions. In the case of the probit model, a dichotomous dependent variable is used with a value equal to ’0’ if the respondent does not usually work overtime and equal to ’1’ otherwise. Thus, these regressions reflect the determinants of doing overtime. In the case of the tobit model, the dependent variable has a value equal to ’0’ if the respondent does no overtime and equal to the number of overtime hours usually done per week otherwise. These regressions show how the number of overtime hours is affected by the explanatory variables. We have used the same explanatory variables as in the regressions of the previous section.28 Furthermore, and for the sake of clarity, we only present the marginal effects (calculated at sample means) here.

28

As in the above analysis, these regressions also included twelve dummy variables for different sectors. Although we do not present the results here, one interesting observation is that workers in the banking and insurance sector have a relatively high probability of working long overtime hours and also have a high probability of wanting to work less than contractually required to (relative to other industries). One possible explanation is that firms in these sectors often try to recruit very productive workers, and they do so by offering relatively high wages together with long required working times (especially overtime). This topic is analysed in Sousa-Poza and Ziegler (2001).

16 Taking a look at the probit results first we note that men with a high education, men that have block or flexible working times, and men in management positions are likely to do overtime. Men with a low education have a higher probability of not doing overtime. The ageovertime function for men is a parabolic one with its maximum at around 31 years. Thus, younger employees are primarily affected by overtime. The determinants of overtime in the female sample are very similar to those in the male sample with the exception that marriage and the presence of children reduce the probability of doing overtime. This result reflects the much-cited observation that women are primarily responsible for household tasks and childcare (see, for Switzerland, Sousa-Poza et al., 2001). Table 6: the determinants of overtime - marginal effects of the probit and tobit models probit males females males demographic variables marrieda 0.009 -0.059** 0.088 childrena -0.011 -0.116** -0.014 0.056** 0.025** 0.874** high educationa -0.225** -0.149** -1.547** low educationa 0.144** 0.193** 0.120** age×10-1 -0.227** -0.283** -1.160** age2×10-3 working-time characteristics -0.009 0.027 0.429* works on weekenda works at nighta -0.014 -0.023 0.309 0.179** 0.168** 0.964** works block timesa 0.086** 0.009 1.325** flexible working timesa 0.060 0.026 0.225 works shiftsa other work-related characteristics tenure 0.001 0.003 0.002 2 -3 0.044 -0.008 0.249 tenure ×10 is in managementa 0.220** 0.225** 1.852** 0.021 0.003 0.044 small firm (< 100 employees)a sigma number of observations log likelihood pseudo-R2

3775 -2150 0.122b

3412 -2102 0.109b

4.883** 3775 -8312 0.211c

tobit females -0.247** -0.605** 0.208* -0.676** 0.112** -1.152** 0.183 0.005 0.520** 0.222 -0.138 -0.001 0.363 1.061** -0.059 4.515** 3412 -5738 0.171c

Note: The dependent variable in the probit model has a value equal to ’0’ if the respondent does not usually work overtime and equal to ’1’ otherwise. In the case of the tobit model, the dependent variable has a value equal to ’0’ if the respondent does no overtime and equal to the number of overtime hours usually done per week otherwise.The regressions include a constant and twelve dummy variables for different sectors. Only the marginal effects are depicted here. a dummy variables b the pseudo-R2 measure is that of McFadden (1973) c the pseudo-R2 measure is that of McKelvey and Zavoina (1975) */** significant at the 5%/1% level, respectively

The results of the tobit model show how the independent variables affect the length of overtime work. We note that men with a high education work about 52 minutes more overtime per week than workers in the reference group, whereas men with a low education

17 work about one and a half hours less overtime per week. A further interesting result is that men with flexible working times work approximately one hour and 20 minutes more overtime per week than individuals that have fixed working times. As one would expect, the effect of having a management position is very strong: managers work nearly two hours more overtime per week than workers without a management position. The results are similar in the female sample with the exception of the marriage and child coefficients.

As was shown above, there are different forms of compensation for overtime work. In order to establish the determinants of the form of compensation for overtime, we estimate multinomial logit functions for both males and females. The dependent variable can have three possible values, depending on the form of compensation. The summarised results are presented in table 7. The marginal effects of these multinomial logits are presented in table A in the appendix. In order to keep things simple, we only select some of the most interesting variables for the exposition in table 7. Table 7: determinants of the form of compensation for overtime (results of the multinomial logit model)

males

none

holidays

money

- high education - works on weekend - does not have block times - flexible working times - does not work shifts - in management

- does not have a high education - does not work on weekend - does not work at night - works block times - does not have flexible working times - works shifts - not in management - does not have a high education - younger workersb - single - no children - does not work at night - works block times - works shifts - have high tenure - in management

- low education - not middle ageda - works on weekend - works at night - flexible working times - not in management

- low education - older workers - married - children - works at night - does not have block times - flexible working times - does not have high tenurec - not in management Note: Some stylised results which are based on the estimates of the multinomial logits presented in table A in the appendix. a is a parabolic function with its minimum at around 42 years of age b is a parabolic function with its maximum at around 26 years of age c is a parabolic function with its minimum at around 43 years of age females

- high education - elder workers - works at night - does not have block times - does not work shifts - in management

As can be seen in table 7, several factors influence the form of compensation. In fact, the picture presented here is that the determinants of the form of compensation are complex, and

18 it is difficult to establish some stylised facts. This is a result of the high heterogeneity within different groups of employees with regard to their form of overtime compensation. However, in general, we note that, in both the male and the female samples, white collar workers do overtime that is most probably not compensated in any form, i.e., individuals in a management position and with a high education.29 Young, single women who have no children are likely to receive compensation in the form of holidays, whereas older, married women with a low education and children tend to get overtime paid out. In addition, females not being in a management position and not having high tenure, but possessing flexible working times and working at night receive paid overtime. One could assume, that these women tend to be at the lower level of the income scale.30 Overtime compensation in the form of holidays is generally given to individuals without a high education (although not a low education, i.e., individuals in the reference group). An interesting gender difference is that women in management are (if at all) likely to get overtime compensated in the form of holidays. This does not apply to male managers. Note, however, that women without a high education (i.e., not necessarily women in management) also are more likely to receive compensation in form of holidays. Thus, a certain amount of heterogeneity in the group of individuals getting a certain form of compensation inevitably exists. A further interesting observation is that overtime is generally paid out to workers with primarily a low education and who do not have a management position. We also note that the working-time characteristics have an influence on the form of compensation, and that these effects do sometimes vary between genders. Males not working on weekends and both males and females who do not work at night, but have block times, or work shifts are most likely to receive compensation in the form of holidays. Males as well as females with flexible working times are most likely to get overtime paid out. Note also that men with flexible working times are also likely to get no compensation and that the corresponding marginal effect is much larger than in the case of paid compensation (see table A in the appendix). This result could arise from the fact that, on the one hand, flexibility may be associated with an individual’s ability to organise his own workday (which primarily would apply to white-collar workers31). On the other hand, flexible working time may be administered by the firm, and this could

29

A similar result was obtained by Bell and Hart (1999) for Great Britain.

30

As was shown in section 4, these are also women who prefer to work and earn more than they currently do.

31

According to two recent studies for Germany, the introduction of so-called working time accounts ("Arbeitszeitkonten") led to a reduction of paid overtime (see Groß/Munz, 1999, Bauer et al., 2000).

19 primarily affect blue-collar workers. A similar argument may apply for men who work on weekends. On the one hand, such men could be white-collar workers (and thus tend to receive no compensation for overtime) who have a fair amount of autonomy with regard to the planing of their workweek or, on the other hand, these are primarily blue-collar workers that have to work on weekend (and thus tend to receive compensation for their overtime). The same reasoning applies to women who work at night. Thus, once again, the picture presented here is a complex one with a certain degree of heterogeniety in groups of individuals receiving a specific form of compensation. Perhaps the only stylised fact which can be made is that white collar workers are the most likely to receive no compensation for overtime and blue collar workers are more likely to get compensation in the form of money or holidays. As we pointed out in the comparison above between desired and contractual working time, most workers are satisfied with their contractual working time (see section 4). It therefore comes as no surprise that the number of constrained workers increases substantially when we analyse actual working time, actual working time being defined as contractual working time plus overtime. Figures 5 and 6 compare actual with desired working time. As can be seen in both the male and the female samples, the distribution of the actual working time curve has become flatter with a larger tail in the upper end, i.e., a large portion of the distribution has shifted toward the higher hours. Thus, these figures reveal that actual working time is, in general, too long. More specifically, the predominant desire to work approximately 40 to 42 hours per week is not being fulfilled, i.e., a substantial portion of employees work more than 42 hours per week. Taken at face value, these results would imply that hours constraints can be reduced by decreasing the number of overtime hours worked. This, in fact, is often demanded by unions (see SGB, 1999, p. 27).

20 Figure 5: distribution of actual and desired working time for women

percent

20

10

desired working time 0

actual working time 1

6

11

16

21

26

31

36

41

46

51

56

actual/desired working time (hrs./week)

Figure 6: distribution of actual and desired working time for men 40

percent

30

20

10 desired working time 0

actual working time 1

6

11

16

21

26

31

36

41

46

actual/desired working time

51

56

21 At this stage, however, a word of caution needs to be included. Based on the available data, it is not clear to what extent the reported desired working time refers to the contractual or the actual working time. Although the wording of the question would make us assume that the respondents mean desired actual and not desired contractual working time, this need not always be the case. It is conceivable that respondents did not explicitly take overtime into consideration when reporting their desired (actual) working time. Stated somewhat differently: if some respondents reported desired contractual working time instead of desired actual working time and if these two responses are different in the sense that the former explicitly excludes overtime, then one does face a problem.32 In such a case, figures 5 and 6 are potentially flawed since they would be comparing desired contractual with actual working time. Unfortunately, the only way to solve this problem would be by stating a more precise question in the survey. In any case, the analysis in this section does show that the nature of overtime plays a very important role in the design of working time policies. Furthermore, it does not change our conclusion from section 3, namely that there does not appear to be much leeway for a collective reduction of the workweek below 40 hours.33 Thus, we can conclude this section by stating that: (i) a fair amount of overtime is performed each week and that the most common form of compensation for overtime is holidays; (ii) a substantial amount of overtime work is not compensated in any form and only a small portion is actually paid out; (iii) women are more likely to receive payment for overtime than men; (iv) no compensation for overtime is most probable among white collar workers, whereas blue-collar workers have a higher probability of obtaining paid compensation; (v) if one compares actual working time with desired working time, then one notes that a large discrepancy between the most common desired weekly working time of between 40 and 42 hours and actual working time exists. Thus, overtime would appear to increase the number of constrained workers, in the sense that these workers would like to work less.

32

Desired actual and desired contractual working times are only equivalent if a respondent does not want to work overtime.

33

A further possible problem with a comparison of desired with actual working time is that, as pointed out above, a substantial portion of overtime is compensated by free time. It is difficult to judge the extent to which this additional free time is taken into consideration when reporting desired working time. Thus, the relationship between contractual working time, desired working time, overtime, and the form of compensation for overtime is quite complex and cannot be clearly interpreted with the available data.

22 6. An International Comparison of Desired Working Time As was pointed out in section 2, with regard to the weekly working time, Switzerland is somewhat of an exception since it has one of the longest workweeks in Western Europe and one of the highest in the industrialised world (see, OECD, 1998a). It is therefore interesting to see how hours constraints in Switzerland differ from those in other countries. In order to do this we analyse the latest Work Orientations data from the 1997 ISSP (see section 3 above). We analyse a sample of workers in 14 industrialised countries. In figure 7, the portion of workers wanting to work less and earn less, work the same and earn the same, and work more and earn more are depicted. A first interesting result is that the extent of hours constraints is relatively small. Furthermore, in most countries, the portion of workers wanting to work more and earn more is larger than the portion of workers wanting to work less and earn less. Only in three countries is this not the case: Denmark, Norway, and Switzerland. Of all the countries considered here, Switzerland has the largest portion of workers wanting to work less and earn less. This could be a result of the relatively high income in Switzerland and long working hours which could give rise to a higher marginal value of leisure.34 In figure 8, a ranking of the degree of hours constraints is shown. As can be seen, Switzerland is ranked fourth out of the 14 countries analysed here. Approximately 70% of all Swiss employees reported no hours constraints. Thus, in an international comparison, Swiss workers appear to be quite satisfied with their current workload/pay combination. This is an interesting result if one considers that Swiss employees work relatively long working hours. Furthermore, this outcome confirms the results obtained with data from the SLFS (see section 4).35

34

This in itself, however, does not explain why these employees should be constrained. Sousa-Poza and Ziegler (2001) provide a possible explanation as to why employees may work inefficient long hours in the sense that they work more than they would like to.

35

The results are, in fact, strikingly similar. Since both surveys use somewhat different questions, this similarity does provide some support for the reliability of the results presented above. Another similar result was obtained by a survey conducted in October 1998 by the GfS-Research Institute. In this survey, a total of 641 Swiss workers were asked whether they would like to work more, work less, or work the same amount (no reference was made to earnings). 65% said that they were satisfied and 23% reported overemployment. See GfS (1998).

23

Figure 7: hours constraints by country 100%

80%

60%

40%

20%

work less and earn less

work the same and earn the same

United States

Switzerland

Sweden

Spain

Portugal

Norway

New Zealand

Netherlands

Japan

Italy

Great Britain

Germany

France

Denmark

0%

work more and earn more

Source: Based on Sousa-Poza and Henneberger (2001)

Figure 8: ranking by country according to the percent of unconstrained workers 80% 70% 60% 50% 40% 30% 20% 10%

Source: Based on Sousa-Poza and Henneberger (2001)

Portugal

United States

Italy

Japan

Spain

New Zealand

France

Sweden

Germany

Netherlands

Switzerland

Great Britain

Norway

Denmark

0%

24 7. Conclusions Workers’ desired working time, workers’ hours constraints, overtime, and the form of compensation of this overtime are fundamental aspects in the development of working-time policies. Yet, these issues have received surprisingly little attention in the current workingtime debate in Switzerland. The aim of this paper is to present a number of empirical results on these topics. With data from the 1998 Swiss Labour Force Survey and with data from the 1997 International Social Survey Programme we show that: •

Most Swiss workers (81% and 64% of all salaried males and females, respectively) do not face hours constraints when we compare the desired working time with the contractual working time. The largest portion of constrained workers would like to work less and earn less (17% of males and 22% of females).

•

There does appear to be a deficit of part-time jobs for both men and women. This deficit is most pronounced for women who would like to work around 20 hours per week, and also for men and women with a high education (i.e., there is a deficit of high-qualification parttime jobs).

•

Women with very short contractual working hours (less than about 10 hours per week) generally would prefer working more and earning more.

•

Men and women with a high education are more likely to want to work less and earn less, whereas the opposite applies to workers with a low education. Women who have flexible working times are less likely to be constrained.

•

Men and women work on average 3.15 and 1.69 hours overtime per week, respectively. The most common form of compensation for this overtime is holidays, followed by no compensation and money. Women are much more likely to receive paid compensation than men.

•

Men and women with a high education and who have a management position are predominantly affected by overtime. Furthermore, it is primarily such workers who receive no compensation for their overtime.

•

Young, single women with no children are likely to get compensation in the form of holidays, whereas elder, married women, with children have a higher probability of getting paid compensation for overtime.

•

A comparison between actual and desired working time, where actual working time is defined as contractual working time plus overtime, reveals that a significant amount of workers are constrained in the sense that they would prefer shorter actual working hours.

25 •

An international comparison of desired working time reveals that workers in Switzerland are quite satisfied with their current workload/pay combination (Switzerland ranks fourth out of the 14 industrialised countries analysed here). Switzerland does, however, have the highest portion of workers wanting to work (and earn) less out of all the 14 countries. In the light of the current working-time debate in Switzerland, most notably the

referendum launched by the SGB, which foresees the introduction of a de facto workweek of 36 hours and a mandatory upper limit of 100 hours overtime per year, our analysis could suggest the following: •

Our results show quite clearly that the 42 and 40-hour workweek is preferred by an overwhelming majority of employees. Thus, one could at least question the optimality of a nation-wide reduction of the workweek by, on average, four to six hours per week. Naturally, the extent to which earnings change plays an important role in such a conclusion. If a reduction of the workweek leaves earnings unaffected (something which is questionable), then this conclusion obviously needs to be put into perspective.

•

If working-time policies at least partially aim at reducing the number of constrained individuals, then an increase in part-time jobs and a reduction of overtime is recommendable. Our results show that there is a deficiency of part-time jobs for highqualified individuals and also part-time jobs in the range between 30 and 38 hours per week. The fact that Switzerland has the largest portion of workers wanting to work less and earn less out of the 14 industrialised countries analysed in this paper reinforces this conclusion.

•

The determinants of hours constraints, overtime, and the form of compensation of overtime are quite complex. This is primarily due to the high and, to a large extent, unobservable heterogeneity of working time preferences between and within different groups of employees. As a result, individual (as opposed to collective) working-time regulations would seem to be more appropriate. The fact that the goodness-of-fit of the above regressions is relatively low implies that a high degree of variability exists, i.e., hours constraints appear to be quite randomly distributed among the underlying population, and, thus, the variability of hours constraints cannot be well explained by the numerous regressors implemented here. Hours constraints therefore are largely affected by unobservable factors such as personality, motivation, etc. Collective working-time regulations, by nature, cannot adequately capture this heterogeneity. Therefore, individual agreements - perhaps on the basis of a shorter average workweek - seem better suited to reduce workers’ hours constraints and generally improve their well-being at work.

26 References Altonji, J.G./Paxson, C.H. (1986), "Job characteristics and Hours of Work", in: Ehrenberger, R. (ed.), "Research in Labor Economics", Vol. 8, Greenwich, Westview Press, 1-55. Bauer, F./Groß, H./Schilling, G. (1994), "Arbeitszeit '93. Arbeitszeiten, Arbeitszeitwünsche, Zeitbewirtschaftung und Arbeitszeitgestaltungschancen von abhängig Beschäftigten", Resultate einer aktuellen Repräsentativbefragung bei abhängig Beschäftigten in Westdeutschland durch das Institut zur Erforschung sozialer Chancen, Köln, im Auftrag des Ministeriums für Arbeit, Gesundheit und Soziales des Landes Nordrhein-Westfalen, Bruns, Minden. Bauer, F./Groß, H./Schilling, G. (1996), "Arbeitszeit '95. Arbeitszeitstrukturen, Arbeitszeitwünsche und Zeitverwendung der abhängig Beschäftigten in West- und Ostdeutschland", Resultate einer aktuellen Repräsentativbefragung bei abhängig Beschäftigten in West- und Ostdeutschland durch das Institut zur Erforschung sozialer Chancen, Köln, im Auftrag des Ministeriums für Arbeit, Gesundheit und Soziales des Landes Nordrhein-Westfalen, Neusser, Neuss. Beckmann, P./Kempf, B. (1996), "Arbeitszeit und Arbeitszeitwünsche von Frauen in Westund Ostdeutschland", Mitteilungen aus der Arbeitsmarkt- und Berufsforschung, 29:388408. Bell, D./Hart, R.A. (1999), "Unpaid Work", Economica, 66:271-290. Bell, L./Freeman, R. (1995), "Why do Americans and Germans Work Different Hours?", in: Buttler, F./Franz, W./Schettkat, R./Soskice.D. (eds.), "Institutional Frameworks and Labor Market Performance: Comparative Views on the U.S. and German Economies", Routledge, New York, 101-131. Birchmeier, U. (2001), "Ausgewählte Arbeitsmarkt- und Bildungsindikatoren im Lichte der Statistik. Ein Vergleich der Schweiz mit der EU, den USA und Japan", Die Volkswirtschaft, 74(1):54-59. Blum, A./Zaugg, R.J. (1998), "Beschäftigung durch innovative Arbeitszeitmodelle. Schlussbericht", Bundesamt für Wirtschaft und Arbeit (BWA), BWA Schriftenreihe, Beiträge zur Arbeitsmarktpolitik No. 12, Bern. Bundesamt für Statistik (BFS) (1996), "Die Schweizerische Arbeitskräfteerhebung (SAKE): Konzepte, methodische Grundlagen, praktische Ausführung", Bern. Bundesamt für Statistik (BFS) (1998), "Die Schweizerische Arbeitskräfteerhebung (SAKE): Kommentierte Ergebnisse und Tabellen", Bern. Bundesamt für Statistik (BFS) (1999a), "Statistisches Jahrbuch der Schweiz 2000", Bern. Bundesamt für Statistik (BFS) (1999b), "Die Schweizerische Arbeitskräfteerhebung (SAKE): Kommentierte Ergebnisse und Tabellen", Bern. Bundesmann-Jansen, J./Groß, H./Muntz, E. (2000), "Arbeitszeit '99. Arbeitszeitstrukturen, Arbeitszeitwünsche und Zeitverwendung der abhängig Beschäftigten in West- und Ostdeutschland", Resultate einer aktuellen Repräsentativbefragung bei abhängig Beschäftigten in West- und Ostdeutschland durch das Institut zur Erforschung sozialer Chancen, Köln, im Auftrag des Ministeriums für Arbeit, Gesundheit und Soziales des Landes Nordrhein-Westfalen, Neusser, Neuss. Clemmer, P. (1997), "Arbeitszeitreduktion – vermag sie Arbeitsplätze zu schaffen", in: Arbeitgeberverband der Schweizer Maschinenindustrie (ASM), "Jahresbericht 1997", Zürich. Dathe, D. (1998), "Wechselwirkungen zwischen Arbeitszeitpolitik und Arbeitsangebotsverhalten – Eine Untersuchung zur Bedeutung von Arbeitszeitpräferenzen für eine Politik der Arbeitsumverteilung", WZB, discussion paper FS I 98-201, Berlin.

27 Eidgenössisches Personalamt (EPA) (1999), "Verbesserte Sozialpartnerschaft beim Bund", Pressemitteilung vom 21.6.1999, Bern. Euwals, R. (1999), "Hours constraints within and between jobs", Paper presented at the 11th Annual Conference of the European Association of Labour Economists (EALE) in Regensburg, Germany. Euwals, R./van Soest, A. (1999), "Desired and Actual Labour Supply of Unmarried Men and Women in the Netherlands", Labour Economics, 6:95-118. Freeman, R.B. (1998), "Wages, Employment and Unemployment: An Overview. Can the EU pass the jobs test?", in: European Commission: EC/DG V – OECD/DEELSA Seminar: Wages and Employment, Luxembourg, 21-31. GfS-Research Institute (GfS) (1998), "Der Wunsch nach individuell gestaltbaren Arbeitszeiten", Spezialstudie des GfS-Forschungsinstituts im Rahmen des "Sorgenbarometer" für das "Bulletin" der Credit-Suisse, Bern. Graf, S./Henneberger, F./Schmid, H. (2000), "Flexibilisierung der Arbeit. Erfahrungen aus erfolgreichen schweizerischen Unternehmen", Bern. Groß, H./Munz, E. (1999), "Erste Ergebnisse einer 1999 durchgeführten repräsentativen Beschäftigtenbefragung zu Arbeitszeitformen und -wünschen", Köln. Hablützel, P./Rebecchi, R. (1998), "Staat und Verwaltung im Wandel", Die Volkswirtschaft, 71(12):40-44. Ham, J.C. (1982), "Estimation of a Labor Supply Model with Censoring Due to Unemployment and Underemployment", Review of Economic Studies, 49:335-354. Henneberger, F. (2001), "Arbeitszeitpolitik als Beschäftigungspolitik? Zur Diskussion der Beschäftigungseffekte von Wochenarbeitszeitverkürzungen in Privatwirtschaft und Öffentlichem Dienst", in: Becker, U. (ed.), "Weniger Arbeit – Arbeit für alle?", Marburg, 1750. Henneberger, F./Graf, S./Sousa-Poza, A. (1999), "Arbeitszeitverkürzung, Arbeitszeitflexibilisierung und Arbeitszeitwünsche von Arbeitnehmern", VSAM-Revue, 80(9):18-22. Ilmakunnas, S./Pudney, S. (1990), "A Model of Female Labour Supply in the Presence of Hours Restrictions", Journal of Public Economics, 41:183-210. Institut der deutschen Wirtschaft Köln (IW) (Hrsg.) (1999), "Zahlen zur wirtschaftlichen Entwicklung der Bundesrepublik Deutschland. Ausgabe 1999", Köln. Kahn, S./Lang, K. (1992), "Constraints on the Choice of Work Hours: Agency vs. SpecificCapital", Journal of Human Resources, 27:661-678. Kahn, S./Lang, K. (1996), "Hours Contraints and the Wage/Hour Locus", Canadian Journal of Economics, 29:71-75. Klauder, W. (1998), "Arbeitszeitpolitik im Lichte gesamtwirtschaftlicher Zusammenhänge und des Strukturwandels der Arbeitswelt", Sozialer Fortschritt, 47:48-49. Lang, K./Kahn, S. (2000), "Hours Constraints: Theory, Evidence and Policy Implications", in: Houseman, S./Wong, G. (eds.), "Changes in Working Time in Canada and the United States", Kalamazoo (MI) (forthcoming). McFadden, D. (1973), "Conditional Logit Analysis of Qualitative Choice Behavior", in: Zarembka, P. (ed.), "Frontiers in Econometrics", New York, 105-142. McKelvey, R./Zavoina, W. (1975), "A Statistical Model for the Analysis of Ordinal Level Dependent Variables", Journal of Mathematical Sociology, 4:103-120. OECD (1998a), "Employment Outlook", OECD, Paris. OECD (1998b), "Main Economic Indicators", OECD, Paris. OECD (2000), "Main Economic Indicators", OECD, Paris. Ramirez, J.V. (1998), "Unemployment Rate and Working-Hour Constraints: Empirical Evidence from the Swiss Labour Force", International Journal of Manpower, 19:449-460.

28 Schaufelberger, C. (1997), "Wegkommen vom ’Standardmodell 8 Stunden, 5 Tage’", Schweizer Arbeitgeber, 92:240-244. Schweizerischer Arbeitgeberverband (AGV) (1998), "Kontraproduktive 36-WochenstundenInitiative", http://www.arbeitgeber.ch/arbeitsmarkt/980327.htm. Schweizerischer Gewerkschaftsbund (SGB) (ed.) (1999), "Kurz und knapp: Die drei Ziele der Initiative für eine kürzere Arbeitszeit", SGB-Dokument No. 62, Bern. Sousa-Poza, A./Henneberger, F. (2000), "Work Attitudes, Work Conditions, and Hours Constraints: An Explorative, Cross-National Analysis", Labour. Review of Labour Economics and Industrial Relations, 14(3):351-372. Sousa-Poza, A./Henneberger, F. (2001), "An Empirical Analysis of Working-Hours Constraints in 21 Countries", Research Paper, University of St. Gallen. Sousa-Poza, A./Schmid, H./Widmer, R. (2001), "The Allocation and Value of Time Assigned to Housework and Child-Care: An Analysis for Switzerland", Journal of Population Economics (forthcoming). Sousa-Poza, A./Sousa-Poza, A.A. (2000), "Taking Another Look at the Gender/Job-Satisfaction Paradox", Kyklos, 53:135-152. Sousa-Poza, A./Ziegler, A. (2001), "Asymmetric Information on Workers' Productivity as a Cause for Inefficient Long Working Hours", Research Paper, University of St. Gallen. Staatssekretariat für Wirtschaft (seco) (2001), "Die Lage auf dem Arbeitsmarkt Januar 2001", Pressedokumentation vom 19. Februar 2001, Bern. Stewart, M.B./Swaffield, J.K. (1997), "Constraints on the Desired Hours of Work of British Men", Economic Journal, 107:520-535.

29 Appendix Table A: determinants of the form of overtime compensation logit model males none holidays demographic variables marrieda 0.030 -0.031 0.016 -0.026 childrena high educationa 0.123** -0.128** low educationa -0.100* 0.059 0.006 0.005 age×10-1 2 -3 0.011 -0.140 age ×10 working-time characteristics works on weekenda 0.085** -0.151** works at nighta 0.042 -0.101** works block timesa -0.052* 0.064** 0.100** -0.147** flexible working timesa works shiftsa -0.156** 0.141** other work-related characteristics tenure -0.243 0.004 0.028 -0.047 tenure2×10-3 is in managementa 0.086** -0.040* small firm (< 100 employees)a -0.003 0.011 number of observations log likelihood pseudo-R2

3132 -2505 0.125b

- estimated marginal effects of the multinomial

money

none

females holidays

money

0.001 0.010 0.005 0.042* -0.111** 0.130**

0.018 0.031 0.169** -0.294 -0.009 0.152*

-0.121** -0.150** -0.074* -0.013 0.141* -0.272**

0.103** 0.119** -0.095** 0.043* -0.005 0.120*

0.066** 0.059** -0.012 0.046** 0.015

0.018 0.080* -0.105** -0.011 -0.288**

-0.021 -0.176** 0.225** -0.043 0.285**

0.003 0.096** -0.120** 0.054** 0.003

-0.002 0.020 -0.046** -0.007

-0.001 0.026 0.044** 0.031

0.008** -0.107 0.068** -0.045

-0.007** 0.081* -0.112** 0.015

2394 -1936 0.199b

Note: The dependent variable can have three possible values depending on the form of overtime compensation. The regressions include a constant and twelve dummy variables for different sectors. Only the marginal effects are presented here. The sample includes all employees that answered to the question dealing with the form of compensation for overtime. It should also be noted that over 20% of employees that do not usually work overtime also reported their overtime form of compensation. These individuals have also been included in these regressions (and therefore the sample size here does not correspond to the sample size in table 5). a dummy variables b the pseudo-R2 measure is that of McFadden (1973) */** significant at the 5%/1% level, respectively