HELSINGIN YLIOPISTO - UNIVERSITY OF HELSINKI METSÄEKONOMIAN LAITOS - DEPARTMENT OF FOREST ECONOMICS

TUTKIMUSRAPORTTEJA - REPORTS

21

SENSITIVITY TO SCOPE IN CONTINGENT VALUATION OF FOREST CUTTING PRACTICES

EIJA POUTA

HELSINKI 2002

TUTKIMUSRAPORTTEJA - REPORTS 21 SENSITIVITY TO SCOPE IN CONTINGENT VALUATION OF

FOREST CUTTING PRACTICES

EIJA POUTA

THIS REPORT WILL BE EDITED BASED ON FEEDBACK RECEIVED FROM SEVERAL SOURCES. QUOTATIONS FROM THIS REPORT SHOULD BE TAKEN FROM THE MOST RECENT VERSION. http://honeybee.helsinki.fi/MMEKN/FrameWWW/research/research.html TÄTÄ RAPORTTIA TYÖSTETÄÄN SAADUN PALAUTTEEN POHJALTA. VIITATTAESSA KIRJALLISESTI RAPORTTIIN TULEE KÄYTTÄÄ VIIMEISINTÄ SAATAVILLA OLEVAA VERSIOTA.

Helsingin yliopisto, metsäekonomian laitos University of Helsinki, Department of Forest Economics PL 27 FIN-00014 HELSINGIN YLIOPISTO, FINLAND Telephone: (+358) 09 1911 Telefax: (+358) 09 191 57984 ISSN 1236-6218 ISBN 952-10-0523-8 PDF-versio 11.4.2002 Helsinki 2002

Eija Pouta 2002. Sensitivity to scope in contingent valuation of forest cutting practices.

This study applies contingent valuation (CV) in measuring the environmental benefits of a forest regeneration cutting policy that is designated to encourage cutting practices in Finland to take environmental concerns into account. This study examines the benefits of program, which conclusively regulates landowners to follow environmentally oriented cutting practices either in limited or extensive scope. The dichotomous choice between status quo and environmentally oriented cutting is found to be insensitive to the scope of the environmental alternative, as the scope variable was insignificant in the logit model. Even though, the truncated means of willingness to pay did not differ between the levels of scope, the overall means were statistically different. Based on the theory of planned behavior the path models of choice for each scope level indicated that the beliefs about negative side effects of extensive program, e.g. beliefs about unemployment effects, were one reason for insensitivity. Tutkimus mittaa metsien uudistushakkuita ympäristöä huomioon ottavaan suuntaan kehittävän politiikkaohjelman tuottamia hyötyjä. Markkinattomien ympäristöhyötyjen arvottamiseen käytetään ehdollisen arvottamisen menetelmää (contingent valuation). 1100 satunnaiselle vastaajalle suunnatulla postikyselyllä mitataan kahden erilaajuisen ohjelman ympäristöhyötyjä. Suppea ohjelma säätelee uudistushakkuualoille jätettävää säästöpuumäärää. Laaja ohjelma säätelee lisäksi useita muita uudistushakkuisiin liittyviä ympäristöön vaikuttavia attribuutteja. Tutkimus tarkastelee vastaajien dikotomisen valinnan nykytilan ja ympäristölähtöisen hakkuukäytännön välillä sekä maksuhalukkuuden herkkyyttä ohjelman laajuudelle. Vaikka ohjelman laajuus ei dikotomista valintaa selittävässä logitmallissa osoittaudu merkitseväksi eivätkä katkaistuun maksuhalukkuuden jakaumaan perustuvat maksuhalukkuudet poikkea ohjelmien välillä, todetaan katkaisemattoman jakaumaan pohjautuva keskimääräinen maksuhalukkuus kuitenkin laajemman ohjelman arvioineessa ryhmässä suuremmaksi kuin suppean ohjelman ryhmässä. Dikotomista valintaa selitetään sosiaalipsykologian suunnitellun käyttäytymisen teoriaan pohjautuvalla polkumallilla. Malli näyttää, että laajempaa ohjelmaa arvottaneet ovat päätöksessään painottaneet uskomusta ohjelman työttömyyttä ja hallintoa lisäävästä vaikutuksesta. Key words: forest attributes, regeneration cuttings, path model

Authors: Eija Pouta Research Scientist, Email: [email protected], Department of Forest Economics, P.O. Box 24, FIN-00014 University of Helsinki, Finland Agknowledgements: The Academy of Finland is acknowledged for financial support.

3

Introduction

The recent discussion of the validity of contingent valuation (CV) has, to a great degree, focused on the issue of sensitivity of valuation to the scope, or scale, of an environmental good (Carson & Mitchell 1995, Carson 1997, Frederick & Fischhoff 1998, Hoevenagel 1996). The lack of such sensitivity is also called embedding. Economic theory suggests that if an individual is willing to pay something to obtain a certain environmental good, she should be willing to pay more to obtain more of that good. The issue of embedding has arisen especially in cases involving large absolute changes (Kahneman & Knetsch 1992, Desvouges et. al. 1993). Since then, many studied have shown sensitivity to scope (Carson & Mitchell 1993, Smith & Osborne 1996, Carson 1997, Smith et al. 1997), or insensitivity (Diamond et al. 1993, Schade and Payne 1994, Svedsäter 2000, and some have even shown both insensitivity and sensitivity (Loomis et al. 1993, Giraud et al. 1999). Varying the quantity of the public good has been a typical technique used to test the effects of scope in cases involving endangered species or in connection with other wildlife conservation issues (Desvouges, 1993 Giraud et al. 1999). The effect of scope in terms of the size of the forest area to be conserved has also been tested (Loomis et al., 1993, Li et al., 2001). Although forest-related valuation studies of scenic quality or quality of forest for recreational use have been conducted, they have not taken scope into account by including varying numbers of quality attributes in valuation. In these studies, forest quality has been defined, for example, in terms of the number of trees remaining after an insect invasion (Walsh et al., 1989; Walsh et al. 1990, in terms of silvicultural management and tree species composition (Mattsson & Li, 1994), as well as in terms of the scenic properties of camping sites, based on tree age classes, densities, stories and species (Daniel et al. 1989). This study tests the effect of scope by varying the number of quality attributes in a program designed to improve forest cutting practices by requiring that environmental consideration be taken into account. In Finland, forests and forestry issues have high public relevance, and the public's interest in forests and forestry is high. Inasmuch as all forests are open to public access for recreation purposes, forest management practices influence most Finnish citizens' everyday living environment. Over 90% of Finns enjoy outdoor activities, mainly in forested environments (Pouta & Sievänen 2001). Consequently, in the 1960's and 1970's commercial

4

forestry was criticized for large-scale clear-cutting, which had a negative effect on the landscape, and for intensive land surface management on clear-cut areas, which hindered forest recreation. In Finland, forest regeneration consists either of clear-cutting and planting or of natural regeneration with seed trees left for some years on an otherwise clear-cut area. In the 1980's and 1990's, criticism of regeneration cuttings was motivated by biodiversity concerns (Hellström 2001). The absence of decaying wood in timber production forests was suggested to endanger species that favored old-growth forests. In addition, siltation due to regeneration cuttings played a role in these discussions. Private, non-industrial forests account for about 75% of the forest land in Finland, and they are distributed among a large number of forest owners, for about every sixth Finn is a forest owner. To ensure the quality of private forests for recreation and nature conservation, public agencies have issued instructions to guide forest owners in their present cutting practices (PP). However, changes in these practices outside biotopes under special protection have been based on the voluntary actions of private forest owners. For forest policymaking purposes it is interesting to determine to what extent Finnish citizens value programs of varying scope, which, by law or by mandatory regulation, force landowners to follow a particular environmentally-oriented forest regeneration cutting practice (EOP). The influence of considerations of scope on willingness to pay for an environmentally oriented harvesting practice is investigated here by measuring the support for a limited program versus an extensive program using a split sample survey design. A limited program determines only the number of trees that must be left on the cutting area to provide habitats for species living in the decaying wood. An extensive program, on the other hand, includes a wide range of other attributes, in addition to the number of living trees left in a cutting area. In addition to testing the sensitivity of WTP with respect to the scope of the environmental change, the respondents' decision-making processes — especially differences in them, depending on the scope of the good — are also studied. The study focuses on the question of whether the number of regulated attributes influences the processes an individual uses to make valuation choices. Choice processes are analyzed on the basis of the theory of planned behavior (TPB) (Ajzen & Madden 1986, Ajzen 1991). TPB and its relative, the theory of reasoned action, have been used in CV studies to explain and predict contingent valuation results (Barro et al. 1996, Kerr and Cullen 1995, Pouta & Rekola 2001). TPB can help us understand valuation behavior, not only in general, but also specifically in connection with

5

insensitivity to scope. Using a structural equation modeling technique, we compare path models based on TPB in two groups whose survey questionnaires presented different levels of scope of the environmental good. By analyzing the differences in the beliefs of these two sample splits regarding the effects of the program, it is possible to determine whether the respondents realized the differences between the policy packets or, conversely, whether they valued the benefits of the two packets equally. Analyzing the whole structure of attitudes and beliefs that explain the choice behavior produced information on which beliefs, attitudes and other variables are important in explaining choices between different scope levels. The next section reviews the discussion of the issue of scope in CV. Then we present the theoretical basis for analyzing the issue of scope in this study. Thereafter, the sample, measurements and estimation are explained. In the results section, the effects of scope are tested, and models are presented to explain them. Finally, conclusions are drawn concerning the reasons for insensitivity to scope. Insensitivity to scope The issue of scope gained prominence in the discussion of the validity of CV in connection with a study published in 1992 in which Kahneman and Knetsch reported evidence of insensitivity to scope and suggested that CV measures willingness to pay for the moral satisfaction of contributing to a public good, not for the economic value of that good. Several concepts, including perfect embedding, regular embedding, part-whole bias, and nesting have been used in the discussion of the validity of CV with respect to the matter of scope. "Perfect embedding" means that the value of the specific good is the same as the value of the more comprehensive good (Hoevenagel 1996), whereas in "regular embedding," a good receives a lower value in connection with a more comprehensive good than when valued on its own. Carson and Mitchell (1995) clarified the concepts connected to the issue of insensitivity to scope by referring to "quantitative nesting" and "categorical nesting." In quantitative nesting two goods are measured with a common scale and the first is more extensive than the second, e.g. project A might protect 300 trees per hectare from damage, while project B undertakes to protect 200 trees per hectare. In this study, we focus on categorical nesting, which occurs when goods are distinguished by changes in more than one attribute in a multivariate utility function, e.g. project A might seek to improve biodiversity,

6

scenic conditions and recreation opportunities, while project B might seek to improve only one of these attributes. There are two competing explanations for insensitivity to scope (Hovenagel, 1996). Critics of the CV method argue that because of the hypothetical nature of the method, respondents will always express similar WTP's across related goals. They argue that the reason for the insensitivity is that the respondents will perceive moral satisfaction, or a "warm glow" from their altruistic contribution, and they will contribute similarly, regardless of the scope of the good. They also argue that respondents have a kind of mental account book for "good causes," and irrespective of the description of the environmental good, they will contribute from the balance of that account in the form of WTP. The other line of explanations has attributed insensitivity of scope to poor survey design and/or problems in administering the survey (Carson 1997). According to this line of thought, a vague description of the good can cause any of several biases. In symbolic bias, the respondent reacts to an amenity's general symbolic meaning instead of to the level, or scope, of provision. A poorly designed good may be perceived as symbolic of a larger good.1 A metric bias means that the researcher might be defining the good in different units than those used by the respondent. In probability of provision bias, respondents might be skeptical that the good will actually be provided. Another reason for insensitivity has been seen in the possibility of different interpretations of joint production: the researcher believes that one good encompasses another, but the respondent finds the two goods to be indistinguishable. For example, the researcher attempts to value the health effects of clean air, but the respondent also takes the effects of clean air on visibility into account. Carson (1997) concluded that studies reflecting insensitivity to scope tend to suffer from small sample sizes or poor survey design. He also suggested that beliefs that the good will be provided vary among sub-samples involving different scopes. As one reason for this he mentions ways of administering the survey that do not encourage respondents to pay close attention to the questions being asked. As examples, he mentions telephone or shopping mall intercepts. However, the respondent's familiarity with the good and the high personal relevance of the good are assumed to improve the sensitivity of the respondent to the scope of the good (Carson 1997). Sensitivity of scope has been tested internally within subjects and externally between subjects. In internal tests the same respondent is asked about willingness to pay for different levels of the good in question. In external tests, several respondents are asked about a

7

single level of the same good. External tests between subjects using split sample survey designs have been considered a reliable way to test the insensitivity hypothesis (Arrow et al. 1993), and this method is also used here.

Theoretical framework The economic model of benefit measurement considers the utility of a respondent before and after the implementation of the EOP. It includes the scope of the program as an attribute of the utility function. The utility level can be described by an indirect random utility function (Hanemann 1984): Vij = V (wi , x a , ci ) + e ia ,

(1)

where Vij is the utility level of individual i wi is the income of individual i xa is the scope of the program A ci is a vector of variables describing the respondent, and eia is a stochastic component. The welfare measure CS (compensating surplus) can be defined as an amount of money which can be subtracted from (added to) an individual’s income after environmental change without his/her utility level from the level which existed before the EOP:

(

)

(

)

V w i , x0 , ci + e i 0 = V w i - CS , x a , ci + e ia ,

(2)

where x0 = status quo of forest state, and CS= compensating surplus (willingness to pay). In the referendum question, an individual faces an offer to pay a given sum of money (Bid) to gain better quality of forest environment. The probability of accepting the proposed program instead of the status quo can be written as follows:

8

Pr( project) = Pr[V (wi - Bid, xa , c) + e ia > V (wi , x0 , c) + e i 0 ] ,

(3)

where V(.) is the observable component of utility. If the cumulative distribution of error term e is logistic, the logit model can be used for the estimation. The probability of choosing the status quo can be written as follows: Fh (Dv) = Pr( No) =

1 1 + e Dv ,

(4)

where D v = v1 - v0 is the change in welfare and F h ( D v ) is the cumulative distribution function of standard logistic variate h = e 0 - e 1 . Let us assume that the survey respondent is presented with forest program B, which covers attributes of program A along with other attributes that increase forest quality with scope xb. If the bid is the same, according to economic theory the probability that program B will be chosen is higher than the probability of choosing program A. Thus, the hypothesis we are testing in this study can be stated as follows: Pr[V ( wi - Bid , xb , c) + e ib > V ( wi , x0 , c) + e i 0 ]

> Pr [V ( wi - Bid , x a , c) + e ia > V ( wi , x0 , c) + e i 0 ].

(5)

In addition to testing hypothesis (5) we apply a theory and methods from social psychology to study the decision process of individual facing programs of different scope. The theory of planned behavior clarifies the belief and attitude structure behind dichotomous choice. TPB is a series of hypotheses linking behavioral intention (in this case choice) to a weighted combination of attitudes, subjective norms and perceived behavioral control. According to TPB, an attitude is a function of salient beliefs about the attitude object, i.e. behavior in question. Each salient belief is linked to an attribute of the attitude object. The attitude is determined as a function of the strength of these beliefs (bi) and the evaluations (ei) associated with the attributes related to the behavior. In a CV context, predicting choice behavior (behavioral intentions) may involve any one of three types of attitudes: 1) an attitude toward the public good at the status quo level (PP) and in an improved state (EOP), 2) an attitude toward a policy dealing with the public good, and 3) an attitude toward paying for the public good (Figure 1). The attitude toward

9

paying is a counterpart of the perceived behavioral control in the TPB (Ajzen & Peterson 1988, Pouta & Rekola 2001). Subjective norms are not included in this study because of their weak explanatory power in a similar setting studied by Pouta & Rekola (2001). The products of beliefs and evaluations are called expectancy-value components (be). They explain attitudes toward the public good before and after implementation of the EOP, as well as attitudes toward a policy dealing with the public good. It is important also that bid, an explanatory variable from the economic model of dichotomous choice, is added to the model. In this study we compare the structure of path models based on TPB between respondents who were given a questionnaire containing the limited environmental change and respondents who were given a questionnaire containing the extensive environmental change.

be 1 (PP) be 2 (PP) be 3 (PP)

attitude (PP) choice

be 4 (PP) be 1 (EOP) be 2 (EOP)

attitude (EOP)

be 3 (EOP) be 4 (EOP)

attitude (policy)

be 1 (policy) be 2 (policy) be 3 (policy)

attitude (paying)

bid

be 4 (policy)

Figure.1. Starting point model which follows TPB. (PP= present cutting practice, EOP = environmentally oriented cutting practice, be= expectancy-value components as products of beliefs and evaluations)

10

Methods Sample The data for this study were collected in year 1998. The sample used for the mailed questionnaire consisted of 1100 Finns from age 18 to 70 and was drawn at random from the census of Finland. After the first mailing, reminder postcards and, after that, reminder questionnaires were sent to those respondents who did not respond to the earlier contacts. This produced a response rate of 49%, i.e., 541 at least partially completed forms. Gender and age distributions of respondents were similar to general population distributions. Measurement The questionnaire began with items that measured first participation in forest related activities. After these warm up questions the questionnaire elicited information about purpose, procedure, magnitude and effects of forest regeneration cuttings in Finland. The attributes of forest regeneration cuttings were also clarified with a drawn illustration of a cutting area. After that the environmentally oriented cutting practice (EOP) was introduced to the respondent. The sample was randomly divided into two sub-samples of equal size. One subsample was asked to respond a limited EOP program, and the other was asked to respond to a more extensive program. The limited program included only one environmental attribute: the amount of trees left on the cutting area. In addition to the amount of trees left on the cutting area, which was 35 trees for both levels of scope, the more extensive program included five environmentally meaningful attributes (Table 1). The attributes were selected based on the public discussion concerning forest management in Finland and on the opinions of a pool of experts representing forest and environment administration and environmental organizations.

11

Table 1. Choice settings. Limited program Forest regeneration cutting alternatives

Alternative A. (present practice)

Alternative Bb)

Amount of trees to be left on the cutting area a)

15 trees per hectare

35 trees per hectare

Cutting potential

14% more than actual use

12% more than actual use

Increase in annual tax of your household 1999

no change

tax increase of FIM 100c)

Extensive program Forest regeneration cutting alternatives

Alternative A. (present practice)

Alternative Bb)

Amount of trees to be left on the cutting areaa)

15 trees per hectare

35 trees per hectare

Valuable biodiversity sites

existing sites are conserved

existing sites are conserved and sites are restored from timber production

Share of large (over 5 ha) cutting areas

4%

0%

Adequate scenery consideration

in 80% of cutting areas

in all cutting areas

Adequate water quality zones

in 80% of cutting areas

in all cutting areas

Ruts of land surface management

visible 10-20 years

disappear in two or three years

Cutting potential

14% more than actual use

12% more than actual use

Increase in annual tax of your household 1999

no change

tax increase of

FIM 100c) a) In addition to giving the level of attribute in each alternative, the effects of varying attribute on forest environment were described. However exact effects in quantity terms were mostly unknown. For example, information about the effects of the trees left on the cutting area was: · Trees left on the cutting increase slowly the amount of decaying wood in the forest. Decaying wood is necessary for old growth insect and plant species. · We have very little information about the effect of varying amounts of decaying wood on endangered species. · Trees left on the cutting area have an effect on the scenery. b) Environmentally oriented cutting practice (EOP) c) Bid levels were FIM 100, 700, 1300, 1900, 2500 in the limited and in the extensive program.

12

In addition to these attributes, the effects of the policy on cutting potential and expenses to respondents household were clarified in the choice table. From respondent to respondent the expenses varied from the lowest bid of FIM2 100 to the highest of FIM 2500 and the respondents were informed that these were the result of compensation paid to landowners. The same bids were used for both levels of scope. In the choice table the first alternative was the present cutting practice (alternative A) and the second one was called alternative B, but it constituted of EOP following either the limited or extensive program. After being presented with the alternatives, each respondent was asked to make a choice between the status quo practice and either the limited or the extensive program depending on which questionnaire they received. For half of the sample we implemented the measurement of the variables of TPB. In this sample the number of questionnaires containing the limited program was the same as of those containing the extensive program. The included attitude and belief items did not effect response rate, which was 50% in this sample split. For the other half of the sample attitudes and beliefs were not measured. To determine the items for belief measurement, an elicitation study was carried out by telephone interview. A systematic sample of 50 people selected from the telephone directory answered open-ended questions about the positive and negative outcomes of forest regeneration cuttings and their regulation in Finland. Belief statements in the mailed questionnaire were constructed for the beliefs determined to be most salient in the telephone interviews. They focused on the PP, the EOP, and the policy implementing the EOP. Belief and evaluation measurements concerned the effects of cuttings with PP or with EOP on forest scenery, future growth of the forest, forest fauna, flora, accessibility of forest, and economic profitability of timber production. The policy beliefs concerned the effects of EOP on administration, forest planning, timber supply to forest industry, conflicts between interest groups, decision power of forest owner, unemployment and the economic status of forest owners. To determine the belief strength (bi) of the 6 statements concerning the outcome of regeneration cutting alternatives, respondents were asked whether they agreed or disagreed on a 7-point scale (ranging from -3 to +3). A belief evaluation (ei) was obtained by asking respondents to state the importance of the outcome on a 7-point scale ranging from somewhat important to extremely important. The products of the corresponding belief and evaluation measures formed the expectancy value components used in the model.

13

In addition, we measured general attitude toward PP and EOP. Respondents expressed their general attitudes by completing the following statements: “Forest regeneration cuttings in their present state are…” using two scales (useful-useless, pleasant-unpleasant) and “Forest regeneration cuttings according to program B are…” using two scales (useful-useless, pleasant-unpleasant). Attitude toward policy implementation and attitude toward respondents’ participation in paying the expenses of the new cutting practice were measured with two semantic differential scales. In the case of each attitude both scales were used to form a sum variable. Estimation In the following analysis, we used all available usable observations to, first, test the scope sensitivity with chi-square test of dichotomous choice results. We also used the same observations to build a logit model of dichotomous choice and to test the significance of scope variable in that model. The difference in models for limited and extensive program was tested with Hausman test. Logit models were used to estimate truncated and overall willingness to pay for both programs. After that we, secondly, focused on that half of the observations in which attitudes and beliefs were measured. To study attitude and belief structure behind the choice we built a path model for the limited program group and the extensive program group. Structural equation modeling software, such as LISREL 8, allows estimation of systems of regression models, such as this kind of path model (Jöreskog & Sörbom, 1996; Bollen & Long 1993). The models, which follow theory (TPB) (Figure 1), was used as a starting point for estimation. Both path models were developed further using chi-square test statistic, root mean square error of approximation (RMSEA), goodness-of-fit index (GFI), normed fit index (NFI) and critical N.3 Because the endogenous variable, choice, is dichotomous, the estimation in LISREL is based on a polyghoric correlation matrix, which allows dichotomous variables. The path model of the limited program was compared to the path model of the extensive program by analyzing the difference in correlation matrixes of model variables between groups and differences in regression coefficients between groups using chi-square test statistics. In addition, the standardized total effects of each variable on choice were compared between groups.

14

Results Test for scope sensitivity Against assumptions, the share of supporters of the EOP did not vary with respect to the scope of the practice (Table 2). The split sample test did not show a statistically significant difference between limited program and the extensive program. In the extensive program with six environmental attributes the support for EOP was even lower (41%) than for the limited program with only one attribute (43%), although, the difference was not significant. This reveals insensitivity to the scope of the good.4 One reason for this insensitivity might be that some respondents did not consider the sum total of the five additional attributes of the program as a good. Table 2 reports the share of the respondents who considered the EOP as a good. If EOP had not resulted in additional costs to the respondents, 71% and 74% of them would have supported it in a limited form or an extensive form, respectively. Although, the difference is not significant, it nevertheless excludes that explanation that most respondents considered additional attributes of extensive program as a bad. Table 2. Choice between status quo and environmental cutting practice (EOP) according to the program scope.

Supporters of the EOP Supporters of the EOP with no cost

Chi-Square

Sig.

N

Limited program

Extensive program

42.6%

41.3%

0.085

.771

503

70.6%

74.2%

0.738

.390

453

value

Logit models were used to analyze the nature of the scope variable further (Table 3). In the logit model for all observations the bid, some socioeconomic background variables, responses to some opinion statements, and the scope variable were tested to determine their ability to explain the probability that the EOP would be chosen. 5 The coefficient of the bid was negative and significant. Of the socioeconomic variables, higher education increased the probability of supporting the EOP. On the other hand, the coefficient of the income was not significant in the model. Those respondents who lived in southern Finland were more likely to support the EOP than respondents from other parts of the country. Those respondents who managed their own forest lands as a leisure activity were more likely to support the status quo than an EOP. Some of the coefficients of the opinion statements were very significant. If

15

respondents reported that they could not afford the EOP or that they were pleased with the current cutting practice, they were more likely to support the status quo. However, the main goal of the logit model was to analyze the coefficient of the scope variable. The logit model confirms the scope insensitivity shown in table 2, as the coefficient of the scope is not significant. Table 3. Logit model for dichotomous choice. All observations

Limited program

Extensive program

B

Sig.

B

Sig.

B

Sig.

4.168

.0000

.8219

.4519

.3093

.7816

-.4278E-3

.0073

-.42845E-3

.0606

-.4693E-3

.0452

Education

.2271

.0052

.0885

.4280

.3621

.0031

Income

-.0371

.5940

-.0083

.9333

-.0665

.5274

Living in south

.5107

.0618

.7876

.0458

.2138

.5931

Forest work as a hobby

-6513

.0354

-.0622

.8905

-1.0817

.0163

Statement: EOP more important than income

.6386

.0000

.6738

.0000

.6111

.0000

Statement: Income more important than EOP

-.1684

.0675

-.1147

.4023

-.1891

.1475

Statement: Can not afford

-.2586

.0003

-.2890

.0037

-.2280

.0315

Statement: Pleased with current practice

-.4466

.0000

-.5486

.0001

-.3968

.0025

Scope

.0697

.7935

Constant Bid

N

424

221

203

Model Chi-square

228.87

128.62

108.33

Restricted log likelihood

-290.45

-151.28

-139.16

Log likelihood

-176.01

-86.97

-84.99

Percent correct

79.72

82.35

80.30

Hausman test statistic between limited and extensive program

7.8401

Hausman test significance

.4069

To analyze possible differences between sample splits in greater detail, logit models for both scopes of EOP were built. Although, the Hausman test between both models did not reveal any structural differences in the data of the sample splits, there were differences in the statistical significance of some coefficients. Education increased the probability of choosing the EOP only if the EOP was extensive. This can be related to the ability to weigh larger amount of arguments of extensive program. Living in southern Finland, on the other hand, increased the likelihood of choosing of the EOP only if the program was limited.

16

Managing forest land as a leisure activity decreased the probability of choosing the EOP only in the case of extensive program. Table 4. Willingness-to-pay estimates (FIM 1 » USD .17 in October 1999, FIM 1 » 0.17 € since the beginning of

2002).

All

Limited program

Extensive program

1782

1742

1680

452

639

548

[1739, 1825]

[1658, 1826]

[1605, 1755]

313

241

388

467

590

526

[269, 357]

[163, 319]

[316, 460]

test statistic

prob.

T-test for truncated means

1.0749

0.2830

T-test for overall means

2.7119

0.0069

The truncated meana a

Standard deviation

a

95% confidence intervals b

The overall mean

b

Standard deviation

b

95% confidence intervals

a

The willingness-to-pay distribution is constrained to be non-negative The willingness-to-pay distribution is not truncated, and the support range from - ¥ to

b

+¥.

Table 4 reports the mean willingness to pay for implementation of an EOP. In the case of means that were truncated to be non-negative there were no significant differences between the two scopes. Mean of one program fitted inside the 95% confidence intervals of the other program. Instead, based on the estimates of overall means the respondents in fact were sensitive to scope. There, t-test showed that WTP estimates differed significantly. The mean WTP for the extensive program was significantly higher than the mean WTP for the limited program. From the estimated means of WTP and from coefficients of the bid in logit models we can see that likelihood of choosing the EOP was higher in the extensive program than in the limited program with low bid levels but lower if the level of the bid was high. From these two contrasting means the over all mean can be recommended for policy purposes. The overall mean does not truncate the WTP so that it becomes non-negative, and the data reflected some negative WTP’s in this study (Table 2).

Understanding choice behavior with respect to scope We used analysis of variance to test if the beliefs about the effects of the regeneration cutting in a limited or in an extensive program differed (Table 5). The results

17

show that the means of beliefs regarding each outcome of cutting did not differ significantly in any of the cases (first six statements). This suggests that respondents in the limited program group considered the environmental benefits of new cutting practice similarly as did the group of respondents who were given the extensive program. Table 5. Beliefs concerning the outcome of regeneration cutting and its implementation with regard to the scope of the program (scale 1, fully agree –7 totally disagree). Analysis of variance. Limited program

Extensive program mean

F

sig.

3.32

.000

.984

4.57

4.47

.057

.811

diminishes fauna diversity

3.44

3.39

.071

.790

diminishes flora diversity

3.33

3.33

.927

.337

hinders hiking in the forest is economically efficient

3.41

3.53

.310

.578

4.04

4.09

.022

.886

3.81

4.45

8.35

.004

increases management based on planning reduces timber supply to forest industry increases conflicts between interest groups narrows the decisionmaking power of forest owner increases unemployment

4.73

5.07

3.169

.076

3.91

4.06

.482

.488

4.55

4.64

.203

.653

4.57

5.09

6.611

.011

3.23

3.99

10.941

.001

has negative effect on the finance of forest owners

4.08

4.33

1.245

.266

Alternative forest regeneration cuttings… reduces scenic beauty

3.33

guarantees future growth

Implementing the program… increases administration

However, in the beliefs concerning the implementation of the policy there were significant differences between the sub-samples of varying scope. The extensive program was regarded more as a source of increased administration (p-value .004), but also as a source of forest management based on forestry plans (p-value .076). The respondents who faced the extensive program considered that the EOP would increase unemployment more than the respondents in the limited program group (p-value .001). This happened in spite of the fact

18

that both groups were informed that implementing EOP will not have an effect on employment. Also forest owners’ control over their forest property was considered to be restricted more in the group of extensive program (p-value .001). Two path models were estimated in order to permit us to focus to the choice processes and to determine which beliefs affect the choice in each scope group. We initially tested path models, which followed the TPB as accurately as possible. However these models, not reported here, did not significantly describe the observed choice making, as their chisquare values were high and their p-values were approximately zero.6 In order to improve the statistical performance, the variable “attitude toward policy” and “attitude toward EOP” were combined, because the correlation between these two was very high and significant. Also some expectancy-value components (products of beliefs and evaluations) that were not significant were removed from the model. In order to improve the models further we used modification indexes of LISREL and allowed expectancy value components that were targeted on PP to explain the attitude toward EOP and policy. Also vice versa, those expectancy value components that were targeted on EOP or policy were also allowed to explain attitude toward PP if the coefficient was significant. Having taken these steps, we selected the path models that fitted best (Figure 2 and Appendix 1), based on chi-square test statistic, RMSEA, GFI, NFI and critical N, for both groups, from among the several estimated alternatives. Both models included three regression equations; first explaining choice, second explaining attitude toward PP and third explaining attitude toward EOP and policy. The path structure of both models emerged when attitudes, which were endogenous in the second and the third equation, were used in the first equation to explain choice.

19

Limited program

scenery be (PP) 0.42

growth be (PP)

0.14

0.19

attitude (PP)

economic be (PP)

-0.48

-0.16

0.20

scenery be (EOP) -0.20

0.35 hiking be (EOP) planning be

attitude (EOP & policy)

0.34

0.38

0.21

timber supply be -0.14

choice

0.14 0.33 0.36 -0.07

-0.14 attitude (paying)

conflicts be

bid

Extensive program scenery be (PP)

-0.30

0.33 growth be (PP) economic be (PP)

0.22

0.18 0.44

attitude (PP)

-0.29 -0.21

0.33

scenery be (EOP) economic be (EOP) planning be

0.16 0.30

conflicts be impoyment be

-0.13

choice

-0.21

0.23 attitude (EOP & policy) 0.30

0.17 timber supply be

-0.20

0.41 0.17 0.20 -0.16

attitude (paying)

0.26 bid

Figure 2. Path models for explaining choice between the present and the environmentally oriented cutting practice (EOP) as made by the limited program group and the extensive program group.

According to chi-square test both models were adequate descriptions of the relationships among variables (Appendix 1). The overall fit of models was good, because the chi-square is associated with a probability level > 0.05. Also other test statistics showed that models’ fit were acceptable. In both models RMSEA was clearly under the critical level of

20

0.08. The goodness-of-fit index (GFI) and normed fit index (NFI) both exceeded the 0.90 criterion in both models. In addition, critical N was higher than the actual number of observations in both models. We used the final models as a basis to test the hypothesis that coefficients in the path models for both groups were similar. First we estimated a model for the extensive program group using the same variables and relationships as in the previous model of the limited program. The test statistic for the hypothesis is the difference between the chi-squares for these two models with a degree of freedom equal to the difference between the degrees of freedom (Table 6). Second, we estimated model with the same variables and relationships as in the previous model of the extensive program group for the limited program and obtained the test statistic in a similar manner. Both tests indicated that the hypothesis that the coefficients are equal should be rejected (p-value 0.000).

Table 6. Test statistics for hypothesis that model coefficients are equal between groups. Limited program

Coefficients in the model of limited program set invariant

16.57

Extensive program Chi-square df 144.17

Group comparison

16

55

39

Coefficients in the model of extensive program set invariant

117.88

15.13

102.72

51

15

39

127.61

Differences in coefficients in the models for both levels of scope indicated different structures in the relationships between choice, attitudes and beliefs. To test this further we compared polyghoric correlation matrixes between both groups. The chi-square statistic for this test produced a value of 174.09 with 84 degrees of freedom, indicating a significant difference between groups in correlation structure (p-value 0.000). To analyze which beliefs actually had an effect on choice, we compared the standardized total effects of each explanatory variable on choice (Table 7). Standardized total effects consist of both direct effect and indirect effects via attitudes. Comparing the standardized total effects and their significance in the models for the limited good and for the extensive good reveals differences in variables which explain choice and in their relative

21

importance. The coefficient of the attitude toward present cutting practice was higher in the limited program group than in the extensive program group. The choice made by those receiving the extensive good was driven more by the bid, which was significant in that group only. In addition, the total effects of expectancy value components and their significance also vary between the two groups. The group receiving the limited good placed more emphasis on the economics of timber production in its present state. In that group only the expectancy value component attached to possibilities in hiking in the forest became significant. The group with more extensive program placed greater emphasis on the scenery and economic expectancy value components of the EOP. Of those expectancy value components, which were connected to policy effects, components linked to planning and timber supply were given more weight by those receiving the limited program, while in the group receiving the extensive good, greater emphasis was placed on the component linked to employment. Table 7. Standardized total effects on choice in path models.

attitude toward PP attitude toward EOP and policy bid attitude paying scenery be (PP) growth be (PP) economic be (PP) scenery be (EOP) roam be (EOP) economic be (EOP) planning be timber supply be conflicts be employment be

Limited Extensive program program Standardized total effects (t-value) -0.49 -0.20 (-9.77) (-3.15) 0.33 0.42 (6.69) (5.71) -0.07 -0.16 (-1.73) (-2.95) 0.47 0.33 (9.03) (5.08) -0.27 -0.38 (-6.06) (-6.17) 0.05 0.19 (0.93) (3.08) -0.20 -0.09 (-5.47) (-2.51) 0.19 0.44 (4.57) (6.49) 0.14 (3.48) 0.12 (3.13) 0.11 -0.04 (4.17) (-0.62) 0.21 0.07 (4.67) (2.40) -0.07 0.12 (-1.37) (1.97) 0.11 (3.25)

22

Conclusions Our study reports the results of a split sample scope test that compared the valuation results of two levels of environmentally oriented cutting practices. Two program levels were tested by introducing two different sets of regulated attributes of cutting areas, viz. a limited program and an extensive program. Increasing the number of attributes of the good had no effect on the probability of choosing the environmentally oriented forest regeneration cutting practice. In that sense the respondents were insensitive to the scope of the good. The scope variable was not significant in the logit model of dichotomous choice. However, the overall mean of the extensive program was significantly higher than the mean of the limited program. Analysis of variance showed that respondents’ beliefs with regard to biodiversity conservation, forest scenery, outdoor recreation environment or economics of timber production were similar regardless of the scope. Many attributes of the extensive program, including share of large cutting areas, adequate scenery considerations, and regulation of ruts of land surface management, could for good reason, be assumed to have an effect on perceptions of scenic beauty and recreation opportunities. However, the beliefs of the respondents who received the extensive program concerning these matters were similar to those who received the limited program. This reveals the difficulty of describing forest attributes in a meaningful way, which can be one reason for insensitivity. In the previous literature the respondents’ familiarity with the good and the high personal relevance of the good was assumed to improve sensitivity to the scope of the good (Carson 1997). We can conclude that despite the high relevance of forests to this Finnish sample, describing forest attributes in meaningful way proved to be difficult. This difficulty is especially apparent in policy level studies that are not linked to any specific location or site (Loomis et al. 1993). In addition, the lack of existing ecological information of the effects of proposed policies may be another reason for insensitivity. However, beliefs related to implementing the policy differed between the two groups. The respondents who valued the more extensive program believed that it would increase administration and unemployment and would limit the forest owners’ decisionmaking power more than did the group valuing the limited program. It is evident that respondents considered programs as packages that included not only the good but also the

23

policy through which the good is offered and the effects of that policy (Fishoff and Furby, 1988). Although, the sensitivity to scope was weak in the dichotomous choice itself, in the path models, based on TPB, choice differed depending on the scope. It seems that the scope affected the way respondents constructed their preferences. The scope of the good altered the weight placed on each separate component of the choice. Bid and scenic and economic beliefs were emphasized in the responses of those who received the extensive program. Recreation interests were given more weight by those who received the limited program. Differences in choice models also indicated that the choice is not only influenced by those beliefs that were connected to the good but also by those beliefs that were connected to the regulative policy. Even though both groups were informed that there would be no increase in unemployment, respondents who received more extensive environmental good nevertheless assumed larger unemployment effects. It seems that scope insensitivity is related to the joint production type of situation described by Carson (1997); the respondents related other benefits or negative effects to the good than those which the researcher intended. This study did not attempt to measure the moral satisfaction or warm glow of respondents related to the two program levels. However the warm glow literature describes that the decision structure and the main arguments of the decision are similar regardless of the scope of the good. In the case of warm glow, moral satisfaction is the main argument of the decision, and beliefs about the benefits of the program do not make any difference in the decision. However, in this case we found significant differences in the structure of the choosing process. In that sense our results did not support the hypothesis of “warm glow”. Instead, the reason for insensitivity to scope would, in this case, be the negative side effects that respondents connected to the extensive good. However, the integration of adequate measures of moral satisfaction into the path model of choice might broaden our understanding of the reasons for insensitivity to scope. In addition to reporting the results of the scope test, the study reported the actual willingness to pay amounts for two programs. The overall means of both programs that were designed to produce environmentally oriented cutting practices were positive. These results provide information for cost benefit analyses to evaluate cutting alternatives for policy-making purposes.

24

References

Ajzen, I. and Madden, T.J. 1986. Prediction of goal-directed behavior: attitudes, intentions and perceived behavioral control. Journal of Experimental Social Psychology. 22:453-474. Ajzen, I and Peterson, G. 1988. Contingent value measurement: the price of everything and the value of nothing. P. 65-76 In Amenity Resource Valuation: Integrating Economics with Other Disciplines, Peterson, G.T., Driver, B.L. and Gregory, R. 1988. (eds.). State College, PA:Venture Publishing, Inc. Ajzen, I. 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes. 50:179-211. Arrow, K. Solow, R. Leamer, E., Portney, P., Randner, R. and Schuman, H. 1993. Natural resource damage assessment under the oil pollution act of 1990. National Oceanic and Atmospheric Administration panel. Federal Register. 58(10):4601-4614. Barro, S.C., Manfredo., M.J. Brown, T.C. and Peterson, G.L. 1996. Examination of the predictive validity of CVM using attitude-behavior framework. Society and Natural Resources. 9:111-124. Bollen, K.A. and Long, J.S.(eds) 1993. Testing Structural Equation models. Sage Publication. 320 p. Browne, M.A. and Cudeck, R. 1993. Alternative way of assessing model fit. P.136-162. In Testing structural equation models. Bollen, K.A. and Long, J.S. (eds.). Sage Publication. Carson, R.T. 1997. Contingent valuation surveys and tests of insensitivity to scope. P. 127163. In Determining the Value of Non-marketed Goods: Economics, Psychological, and Policy Relevant Aspects of Contingent Valuation Methods. Kopp, R.J., Pommerehne, W.W. and Schwarz, N. (eds.). Kluwer Academic Publishers. MA, USA Carson, R. and Mitchell R. 1995. Sequencing and nesting in contingent valuation surveys. Journal of Environmental Economics and Management. 28:155-173. Carson, R. T. and Mitchell, R.C. 1993. The Issue of the Scope in Contingent Valuation Studies. American Journal of Agricultural Economics. 75:1263-1267. Daniel, T., Brown, T. King, D., Righards, M. and Stewart, W. 1989. Perceived scenic beauty and contingent valuation of forest campgrounds. Forest Science. 35:76-90. Desvouges, W.H., Johnson, F.R., Dinford, R.W., Boyle, K.J. Hudson, S.P. and Wilson, K.N. 1993. Measuring natural resource damages with contingent valuation: test of validity and reliability. P. 91-159.In Contingent Valuation: A Critical Assessment. Hausman, J.A. (eds.). Amsterdam, North Holland Publishers. Diamond, P.A., Hausman, J.A. Leonard, G.K. and Denning, M.A. 1993. Does contingent valuation measure preferences. P. 41-85. In Contingent Valuation: A Critical Assessment. Hausman, J.A. (eds.). Amsterdam, North Holland Publishers.

25

Fischhoff, B. and Furby, L. 1988. Measuring values: A conceptual framework for interpreting transactions with special reference to contingent valuation of visibility. Journal of Risk and Uncertainty. 1:147-184. Frederick, S. and Fischoff, B. 1998. Scope (in)sensitivity in elicited valuations. Risk Decision and Policy 3:109-123. Giraud, K. L., Loomis, J. B. and Johnson R. L. 1999. Internal and external scope in willingness-to-pay estimates for threatened and endangered wildlife. Journal of Environmental Management. 56: 221-229 Hanemann, W. M. 1984. Welfare evaluations in contingent valuation experiments with discrete responses. American Journal of Agricultural Economics. 66: 332-41. Hellström, E. 2001. Conflict cultures- qualitative comparative analysis of environmental conflicts in forestry. Silva Fennica, Monographs 2. 109 p. Hovenagel, R. 1996. The validity of the contingent valuation method: perfect and regular embedding. Environmental and Resource Economics. 7:57-78. Jöreskog, K.G. and Sörbom, D. 1996. LISREL 8 User’s Reference Guide. Scientific Software international, Inc. Chicago, IL. Kahneman, D. and Knetsch, J. 1992. Valuing public goods: The purchase of moral satisfaction. Journal of Environmental Economics and Management. 22:57-70. Kahneman, D., Ritov, I. and Schade, D. 1999. Economic preferences or attitude expressions? An analysis of dollar responses to public issues. Journal of Risk and Uncertainty. 19:203-235. Kerr, G.N. and Cullen R. 1995. Public preferences and efficient allocation of a possum-control budget. Journal of Environmental Management. 43:1-5. Li, C-Z., Kuuluvainen, J. Pouta, E., Rekola, M. Tahvonen, O. 2001 Using choice experiments to value Natura 2000 nature conservation program in Finland. Helsingin yliopisto University of Helsinki, Metsäekonomian laitos - Department of Forest Economics, Tutkimusraportteja – Reports 18. Loomis, J. Lockwood, M. and DeLacy, T. 1993. Some empirical evidence on embedding effects in contingent valuation of forest protection. Journal of Environmental Economics and Management. 24:45-55. Mattsson, L. and Li C-Z. 1994. How do different forest management practices affect the nontimber value of forest?- An economic analysis. Journal of Environmental Management. 41:79-88. Pouta, E. 2001. Testing the measurement of attitudes and beliefs in contingent valuation as a preference construction tool. Proceedings of IAREP 2001 Conference: Environment and Wellbeing. pp. 245-248. Pouta, E. and Rekola, M., 2001. The theory of planned behavior in predicting willingness to pay for abatement of forest regeneration. Society and Natural Resources. 14:93-106.

26

Pouta, E. and Sievänen, T. 2001. Luonnon virkistyskäytön kysyntätutkimuksen tulokset Kuinka suomalaiset ulkoilevat? Results of the demand study. In Luonnon virkistyskäyttö 2000. Summary: Outdoor recreation 2000. Sievänen, T. (eds.). Metsäntutkimuslaitoksen tiedonantoja 802:32-76, 195-196. Schkade, D.A. and Payne, J.W. 1994. How people respond to contingent valuation questions: A verbal protocol analysis of willingness to pay for an environmental regulation. Journal of Environmental Economics and Management 26:88-109. Smith, V.K. and Osborne, L.L. 1996. Do contingent valuation estimates pass a “scope” test? A Meta-analysis. Journal of Environmental Economics and Management. 31:287-301. Smith, V. K.; Zhang, X. and Palmquist, R. B. 1997. Marine debris, beach quality, and nonmarket values. Environmental and Resource Economics. October 10:223-47 Svedsäter, H. 2000. Contingent valuation of global environmental resources: test of perfect and regular embedding. Journal of Economic Psychology. 21:605-623. Walsh, R.G., Ward, F. A. and Olienyk, J.P: 1989. Recreational demand for trees in National Forests. Journal of Environmental Management 28:255-268. Walsh, R.D., Bjonback, R.D., Aiken, R.A., Rosenthal, D.H. 1990. Estimating the public benefits of protecting forest quality. Journal of Environmental Management 30:175-189.

27

Appendix 1. Path models for explaining choice between present and environmental cutting practice in the limited and extensive program groups.

Limited program choice

attitude toward PP attitude toward EOP and policy bid attitude paying

Partial models attitude attitude toward toward EOP PP and policy

-0.48 (-9.77) 0.33 (6.69) -0.07 (-1.73) 0.34 (6.46)

scenery be (PP) growth be (PP)

0.14 (3.04)

economic be (PP) scenery be (EOP) roam be (EOP)

0.17 (3.10) 0.14 (3.48)

0.42 (6.41) 0.19 (3.00) 0.41 (6.60) -0.16 (-2.54)

Extensive program choice

b (t-value) -0.20 (-3.15) 0.41 (5.71) -0.16 (-2.95) 0.38 0.20 (5.67) (3.20) -0.20 -0.30 (-2.98) (-4.95) 0.22 (3.63) 0.35 (4.99)

timber supply be conflicts be

0.34 (5.33) 0.21 (4.67) -0.14 (-3.12)

-0.14 (-2.30)

0.30 (4.72) 0.33 (4.57) 0.18 (2.05) 0.44 (4.16)

0.33 (5.28)

economic be (EOP) planning be

Partial models attitude attitude toward toward PP EOP and policy

-0.21 (-3.08)

0.23 (3.59) -0.29 (-3.00) -0.21 (-2.90)

0.17 (3.14)

employment be R2 N Model chi-square p-value RMSEA GFI NFI Critical N

0.80

0.60 130 16.57 0.41 0.017 0.98 0.97 244

be= expectancy value component PP= present cutting practice EOP= environmentally oriented cutting practice

0.55

0.62

0.39 143 15.13 0.44 0.0081 0.99 0.98 283

0.16 (2.43) 0.30 (4.17) 0.17 (2.64) -0.13 (-2.01) 0.26 (3.96) 0.55

28

End notes

1

The same bias is also called “judgement by prototype” by (Kahneman, Ritov and Schkade 1999).

2

FIM 1 » USD .17 (October 1999), FIM 1 » 0.17 € (since the beginning of 2002)

3

The root mean square error of approximation (RMSEA) takes into account the error of approximation in the population as well as the model degrees of freedom. Browne and Cudeck 1993 have proposed that 0.05.