What Factors are Responsible for Decision Making in a Threshold Environmental Goods Experiment: a Consideration of Hypothetical Bias in Stated Preference Yohei Mitani†and Koichi Kuriyama‡ November, 2005 (revised on January 30, 2006 )
Abstract Many previous studies of hypothetical bias show that hypothetical payment in stated preference overestimates real payment with economic commitments (Harrison and Rutstrom, 2005). List and Gallet (2001) and Murphy et al. (2005) conduct meta-analysis using the data of previous studies, and they indicate that, when public goods are being valued, the bias increases when compared to private goods. In previous studies on public goods, it is assumed that real payment is unbiased. However, there is no study that analyzes whether real payment for provision of public goods is unbiased in the field of hypothetical bias. This paper analyzes the main factors of decision making in a threshold environmental goods experiment which controls the hypothetical nature of stated preference methods, in both payment and provision. We find that real payment decision making is explained not only by the subject’s willingness to pay but also the subject’s degree of cooperation (or tendency to free-riding) and members’ decision. Our results indicate that real payment for provision of public goods is possibly biased. Key words: hypothetical bias, real payment, threshold public goods experiments, stated preference, endangered species JEL classifications: C91, H41, Q51, Q57
† JSPS Research Fellow and Graduate School of Economics, Waseda University. Email:
[email protected] URL: http://homepage3.nifty.com/ymitani/ ‡ School of Political Science and Economics, Waseda University
1
1
Introduction
Since Law for the Promotion of Nature took effect in 2003, nature restoration projects, for example a vegetation restoration, have become a great issue in Japan. In practical, valuing the nature restoration project has become a policy challenge. The stated preference methods (SP), including contingent valuation method (CV) and choice experiments (CE), have attracted much attention and have been widely used in the field of environmental economics because SP can estimate the economic value of environmental goods, such as the protection of habitat for endangered species. However, SP is based on respondents’ stated willingness to pay (WTP) in a questionnaire and the hypothetical settings in SP include two hypothetical natures in both the payment for and provision of the goods. Therefore, many economists have worried about the validity and reliability of SP (Hausman, 1993). Many previous studies about environmental economics actually show that hypothetical payment in SP overestimates real payment with economic commitments in an economic experiment (Harrison and Rutstrom, 2005). This divergence between stated hypothetical payment and actual real payment is often referred to as hypothetical bias. To attempt to identify the factors of hypothetical bias, List and Gallet (2001) and Murphy et al. (2005) conduct a meta-analysis using the data of previous studies. Their results indicate that when public goods are being valued, the bias increases when compared with private goods. Moreover, in previous studies on public goods, it is assumed that real payment with economic commitments is unbiased, that is equal to true willingness to pay (i.e. compensating surplus or equivalent surplus). However, there is no study which analyzes whether real payment for provision of public goods is unbiased. This paper analyzes the main factors of decision making in a threshold environmental goods experiment which controls the hypothetical nature of the stated preference methods, in both payment and provision. We find that real payment for provision of public goods is also influenced by something like warm glow and strategic bias, so it is possibly biased. The rest of this paper is organized as follows. The next section introduces the previous studies of hypothetical bias and elaborates our motivations more fully. In section 3, we discuss our experimental procedures and design in detail. In section 4, we explain our hypotheses and our estimation model. In section 5, we report on the main results of our analysis with a hypotheses testing. Section 6 concludes the paper with suggestions for future research.
2
2
Previous Studies
As we pointed out in the previous section, there exist two hypothetical natures in a questionnaire of the stated preference methods. First, SP is hypothetical in the payment for the goods, that is, respondents actually do not pay their stated WTP but are only asked in the questionnaire. Second, SP is hypothetical in the provision of the goods, that is, the goods is not actually provided but it is only assumed the goods is provided. Therefore, there are strong doubts about the reliability of the stated hypothetical payment in a questionnaire because of not controlling economic incentive (Friedman and Sunder, 1994). In contrast, CV researchers have attempted to remedy this problem by carefully designing the survey. Hoehn and Randall (1987) showed that well-established dichotomous choice elicitation format could be incentive compatible under certain conditions. However, it is assumed that respondents believe that there are no hypothetical natures in both payment and provision. As the hypothetical natures are not controlled, the problem that there is no incentive for respondents to state their true value remains the key issue for the reliability of SP (Cummings and Harrison, 1994). Since Bohm’s (1972) classic experimental laboratory study, many experiments have been conducted, which compared hypothetical payment in a SP questionnaire and real payment with economic commitment in an economic experiment to examine this reliability of SP (Shogren, 2003). Harrison and Rutstrom’s (2005) recent survey concludes that hypothetical payment overestimates real payment in 34 of 39 studies. List and Gallet (2001) conduct a meta-analysis using 29 previous studies and attempt to identify the factors of hypothetical bias. Moreover, Murphy et al. (2005) conduct an analysis based on 83 data in 28 studies omitting low reliable data in meta-analysis from 59 studies which include List and Gallet’s (2001), and other new case. The mean of hypothetical / real payment in 83 observations is 2.6, and the median is 1.35, so it is widely recognized that hypothetical payment overestimates real payment1 . Their results of a meta-analysis indicate that the magnitude of hypothetical bias is statistically 1) less for a choice-based elicitation mechanism; 2) more for the use of student subjects; 3) more for public as compared with private goods. However, their analysis also indicates that results are quite sensitive to model specification. Furthermore, their result that the magnitude of hypothetical bias is higher for public goods is robust throughout the sensitivity analysis (Murphy et al., 2003). Hence, one could 1
There are several exceptions such as Johannesson (1997). In addition, the median of hypothetical / real payment in Murphy et al. (2005) is 1.35, that is, it comes from a highly skewed distribution.
3
argue that valuing public goods is main factor which increases the divergence between hypothetical and real payments. The primary purpose of SP is to elicit the economic value of non-market goods like environmental goods. However, there are few studies focusing on public goods whereas most studies focus on private goods, as Foster et al. (1997) surveyed 13 previous studies of hypothetical bias and concluded. It is necessary to observe real payment with economic commitment in order to examine the hypothetical bias. In most previous studies of hypothetical bias, the hypothetical nature in the payment of the public goods is controlled by asking for the payment actually. On the other hand, there are few studies that can control the hypothetical nature in the provision of public goods. Brown et al. (1996) valued the road removal goods that were expected to create a non-use value in a national park. In a real payment measurement, respondents actually pay their bid and their payments could be used for the purpose. However, even if a respondent pays actually, it is not necessarily the case that the road removal goods is provided. Therefore, there is still the hypothetical nature in the provision of the good. Moreover, they cannot exactly remove the possibility of free-riding that depresses real payment (Brown et al, 1996). Previous studies on environmental goods, including Seip and Strand (1992; membership in a environmentalist association) and Brookshire and Coursey (1987; densities of trees in a neighborhood park), cannot control both hypothetical provision and the possibility of free-riding. Although our interests are the provision of environmental goods, previous studies of hypothetical bias on public goods fail to elicit the real payment because of following reasons: 1) the difficulty of provision of public goods; 2) the difficulty in controlling free-riding incentive. Now, the question here is to analyze the factors of decision making of real payment in the provision of public goods. In the field of environmental valuation, the problem of free-riding has been analyzed as a strategic bias. A number of CV studies show that strategic bias can be controlled as low as possible by well-established survey design (Mitchell and Carson, 1989). However, their previous studies on strategic bias have been developed under hypothetical settings in not only the provision but also the payment for the goods, so the magnitude of the bias under real payment setting is not well understood. In contrast, in the field of experimental economics, free-riding behavior has been observed, but it is less than what most economic theory predicted (Ledyard, 1995). Given this phenomenon, it is highly possible that free-riding is a considerable problem in real payment setting which controls the hypothetical nature in payment. Furthermore, note that the problem of free-riding is related to that of both 4
provision condition and hypothetical provision, because whether a public goods is provided depends on others’ decision making in realistic situation. Most studies of hypothetical bias assume that real payment with economic commitments is unbiased, that is equal to true WTP (Murphy et al., 2005). However, real payment as well as hypothetical one cannot control the hypothetical provision and the possibility of free-riding in the provision of public goods. In this context, can we assume that real payment for public goods is unbiased? Alternatively, does real payment for public goods depend on the true WTP at all? In addition, is real payment influenced by free-riding tendency and others’ decision under the situation which removes the hypothetical nature in both payment and provision? It is necessary to uncover the factors of subject’s decision making of real payment in a perfect realistic situation that controls both the payment for and provision of public goods and allows for free-riding behavior, in order to explain the factor of hypothetical bias of public goods. This paper presents main factors of subject’s decision making in a threshold environmental goods experiment which controls the hypothetical nature of SP.
3
Experimental Design
In this section, we introduce the outline of our experiment. Following this, we explain the detail of each stage in our experiment. Finally, we summarize some key features of our experiment.
3.1
Experimental Procedure
In order to identify an individual’s decision making of real payment, it is necessary not only to observe the real payment in an environmental goods experiment which controls hypothetical nature in both payment and provision but also collect variables which can explain the decision making there. In addition, the environmental goods to be valued need to be concrete in order to elicit both stated WTP and real payment for the goods. Then, we recreate the provision of a concrete environmental goods in our laboratory combining SP with public goods experiment. This study evaluates a restoration project of an endangered species as an environmental goods. Firstly, we need to measure the stated WTP for the restoration project in order to test whether real payment is explained by the stated WTP. We measure it by conducting the SP using a questionnaire. Secondly, we need to observe the subject’s cooperation level as one of individual attributes, and it is also considered as either subject’s free-riding tendency or cooperative 5
tendency of payment. In this connection, we observe the contribution level by conducting a standard public goods experiment which has been widely done in the field of experimental economics, because it can be considered as individual attribute that everyone generally has in spite of the case of environmental goods experiment. Thirdly, a provision point of environmental goods depends on other members’ decision making. Then we need to measure the total payment of their group in order to analyze whether subject’s decision making of real payment is influenced by others. Our experiment consists of three stages as table 1 summarizes. Stage 1 is a standard public goods experiment and subject’s contribution level is observed in this stage. Next stage 2 is an environmental valuation of concrete environmental goods and subject’s WTP is measured in this stage. Finally, stage 3 is the environmental goods experiment that controls the hypothetical nature in both payment and provision and subject’s real payment for the same goods is observed in this stage. Table 1: Experimental Procedures and Data Collection ⇒ Coop. •with economic commitments, game situation •continuous public goods Stage 2: Environmental Valuation ⇒ WTP •stated preference methods, questionnaire •conservation project of endangered species Stage 3: Envinronmental Goods Experiment ⇒ Real Payment •with economic commitments, game situation •threshold environmental goods
Stage 1: Public Goods Experiment
The experiment was conducted at the laboratory for political economy at Waseda University on January 29 in 2005. The participants consisted of 40 public individuals (21 males and 19 females, the average age is 39.4, and the residences include Tokyo, Chiba, and Kanagawa) recruited by a pooling agency. An experimental session took approximately 1.5 hours and each subject earned 5750 Japanese yen (about 50 U.S. dollars) on the average. Subjects were randomly appointed to a computer with privacy shields and communication was not allowed between subjects. The administrator provided oral instructions with front screen and answered any question. In stage 1 and 3, the Z-tree software (Zurich Toolbox For Readymade Economic Experiments) was used and the original Web survey software was used in stage 2. 6
3.2
Experimental Design
Each stage was designed as follows. Stage 1 We ran the standard continuous public goods experiments in stage 1. The purpose of this stage is to observe a subject’s cooperation tendency (or free-riding tendency). The game structure is as follows. n
2 πi = (E − Ci ) + Cj , 5
(1)
j=1
where πi is the subject i’s payoff, E is the initial token, Ci is the subject i’s contribution to the public good, and n is the number of a group. There are 5 subjects in a group and each subject has 10 tokens in the beginning of each session. Each subject faces a decision of splitting E between savings E − Ci and contribution Ci . Nash equilibrium is characterized by Ci = 0 for all subjects, that is, it is a dominant strategy to free ride on the others. In stage 1, this public goods experiment is repeated 5 times, and we define the subject’s average payment( 5t=1 Ct /5) as subject’s contribution level (Coop): 5 Cit . (2) Coopi ≡ t=1 5 Many previous studies of a public goods experiment have shown that many people are cooperative although free-riding is a dominant strategy (Ledyard, 1995). As the explanation of this phenomenon, Andreoni (1990) and other some studies point out the ”warm glow” or ”impure altruism” that people experience from the act of contributing. However, our study treats this cooperative tendency as given individual attribute without the explanation of the reason. Stage 2 We conducted an environmental valuation regarding a conservation project of endangered species in stage 2. The purpose of this stage is to measure the subject’s WTP for the environmental goods to be provided in stage 3, after explaining the goods preciously in detail, and stabilizing a subject’s preference for it. Subjects are informed of that the goods to be valued is environmental goods2 . The goods to be valued is the restoration project of Asaza 2
Subjects were informed of just only economic experiment when subjects were recruited. Therefore, there is no sampling bias that the person was only recruited who was interested in environmental issues.
7
(Nymphoides peltata) in lake Kasumigaura. Asaza is endangered species and keystone species from an ecological viewpoint (Mitani and Kuriyama, 2005). After giving enough information for subjects to form their preferences, the 14 times choice experiments are conducted3 . Subsequently, after a question session 14 times choice experiments are conducted again. Finally, a contingent valuation with payment card elicitation format is conducted and the subjects are asked their WTP for the restoration project that can avoid extinction of Asaza absolutely. Let us assume that q 0 is an extinction level (status quo), and q 1 is an avoidable level that the project is provided, then WTP is defined by the expenditure function (e) as follows. W T P ≡ e(p, q 0 , u0 ) − e(p, q 1, u0 ),
(3)
where u0 is the utility level of status quo, p is a price of composite goods. Surly, the subjects do not know that subjects’ WTP stated in this stage is used in the next stage and the same project is dealt in the next stage. Stage 3 In stage 3, we ran the threshold concrete environmental goods experiment. The purpose of this stage is to observe subject’s real payment for the restoration project in the experiment which controls the hypothetical nature in both payment and provision. The goods to be provided is the same restoration project in stage 2. That is, the subjects face the decision making, ”how much are you willing to pay for the restoration project that can avoid extinction absolutely?”. There are 5 subjects in a group. Whether the restoration project to avoidable level of Asaza in lake Kasumigaura is provided or not depends on the total payment of their group. If total payment, called Group, is more than provision point, the project is provided, and if not, the project is not provided and Asaza in lake Kasumigaura goes extinct absolutely. When the project is provided, the subject’s welfare is improved, and the monetary measure of that improvement, that is, subject’s true WTP (compensating surplus) is added to their payoff. The game structure is as follows. n (E − Ci ) + λW T Pi if j=1 Cj ≥ 0.5 × n × E , (4) πi = E otherwise where πi is subject i’s payoff, E is initial token, in this case E is 10 tokens, n is the number of members, W T Pi is subject i’s willingness to pay, and λ is 3 Choice experiments is a one of stated preference methods. Kuriyama (2005) for details.
8
Refer to Mitani and
scale that adjusts unit between WTP and token. As you see, if the project is not provided, subject’s payment is refunded (money-back guarantee). In this stage, this game is repeated 5 times. Then, we define each subject’s contribution Ci in each session as subject’s real payment for the restoration project as follows. (5) Rit ≡ RealP aymentit = Cit . Here, let us check how to control the hypothetical nature in both payment for and provision of the restoration project, whereas allowing the possibility of free-riding. First, the hypothetical nature in payment is controlled by reducing subject’s Ci from his/her final monetary payoff4. Second, the hypothetical nature in provision is controlled to add subject’s WTP to his/her final payoff. If the project is provided, each subject’s welfare is improved by the provision. We design our experiment to add the monetary measure of subject’s benefit (λW T P = λ[e(p, q 0, u0 ) − e(p, q 1 , u0 )]) to final payoff. When the project is not provided, the endangered species goes extinct and the benefit from the project is zero and at the same time, the payment is also zero and the initial tokens are their payoff. Third, a provision point depends on the total payment of their group, so the unreality of provision condition is controlled. For instance, even if subject i pays his/her all tokens, the extinction can not be avoided under the condition that other members pay not so much. On the other hand, as long as total payment of a group is more than provision point, subject i profits by reducing his/her payment as low as possible. Here, Nash equilibrium is characterized by Ci = 0 for all subjects, C = {Ci | nj=1 Cj = 0.5nE, λW T Pi > Ci }. Note that the benefit from the restoration project varies across individuals by using individual WTP.
3.3
Key Features of our Experiment
Key features of our experiment are as follows. First, in order to induce the economic incentive, we employed an experimental economics approach. Final monetary payoff varies ranging from 5000 Japanese yen (about 40 U.S. dollars) to 6500 Japanese yen (about 55 U.S. dollars) depending on each subject’s payoff in experiment. We can observe a real payment with economic commitment by applying this induced value theory (Smith, 1976).
4 Our experiment has economic commitment because ∂πi /∂Ci < 0 and actual payoff is influenced by π.
9
Second, we employed a threshold environmental goods experiment because every environmental policy and goods have some thresholds in the real world. Croson and Marks (1998) categorize the previous studies of public goods experiments from the viewpoint that provision and contributions are continuous or discrete. Other previous studies of threshold public goods experiments include Marks and Croson (1999: provision point mechanism), Croson and Marks (2000: step returns), Cadsby and Maynes (1999: provision mechanism), Rondeau et al. (1999: provision mechanism), and Rose et al. (2000: provision mechanism). Third, we consider the heterogeneity of both subject’s preference for environmental goods and subject’s behavioral principal. Most of previous studies on public goods experiments have assumed that both subject’s preference and behavioral principal are homogeneous. However, preferences for environmental goods often vary across individuals and groups, and behavioral principals ranging from free-riding to cooperation also often vary across individuals and groups (Mitani and Kuriyama, 2005; Burlando and Guala, 2005). Our study allows for the subject’s preference heterogeneity by applying his/her WTP (λW T Pi ) to his/her benefit, and captures subject’s behavioral heterogeneity by his/her cooperative tendency (Coopi ). As remarked above, our study can control both the hypothetical provision and the possibility of free-riding by combing a threshold environmental goods experiment with the stated preference method.
4
Hypotheses and Estimation Model
This section presents our hypotheses about determinants of real payment and our estimation model which is used to test the hypotheses.
4.1
Hypotheses
Does real payment depend on true WTP? First hypothesis is to examine whether real payment for environmental goods depends on the true WTP, as previous studies of hypothetical bias assume. Hypothesis 1 : Subject’s true WTP increases his/her real payment. If this hypothesis 1 is rejected, it means that there is no relation between real payment and true WTP. Then, we will need to reconsider the findings of previous studies on hypothetical bias. 10
Previous studies of the public goods experiments have shown that there are various subjects ranging from free-riding to cooperative behavior (Ledyard, 1995; Burlando and Guala, 2005). Then, hypothesis 2 verifies how many influences subject’s cooperation tendency exerts on real payment for environmental goods. Hypothesis 2 : Subject’s contribution level increases his/her real payment. Even if hypothesis 1 is accepted, in the case that hypothesis 2 is also accepted, real payment can be a measure of true WTP, but real payment is biased by cooperative tendency. Is real payment influenced by others decision? In the real world, a social decision making on provision of environmental goods depends on aggregated members’ decision as well as a decision making of one respondent. Therefore, it is undeniable that respondent’s decision making may depend on others regardless of hypothetical scenarios, ”Please assume that the policy decision depends on your choice.” Then, we observe the subject’s decision making in the situation that the provision of an environmental goods depends on total payment of members and examine whether subject’s decision making is influenced by others decisions. In addition, hypothesis testing here is conducted by using a time series data. There are 5 decision makings in this experiment. We analyze whether subjects’ decision making in period t are influenced by total payment of their group in the previous period t − 1. In this analysis, we add the variable of subject’s profit at previous period (P rof it) in order to control a ”profit effect”. In the game structure showed in equation (4), subject has the incentive that reduces his/her payment as low as possible. Free-riding equilibrium leads us to the following hypothesis. Hypothesis 3 : Total payment of the group in previous period decreases their real payment. If hypothesis 3 is accepted, the result indicates that subject’s decision making of real payment is influenced by others member. Next, let us reconsider the relation between Nash equilibrium and decision making of real payment. When the total payment of the group is more than a threshold all subjects have the incentive that reduces their payment. On the other hand, if the total payment of the group is less than a threshold, Nash equilibrium 11
can not predict their incentive because there are many equilibria including free-riding equilibrium, Ci = 0 for all subjects, and threshold equilibria, C = {Ci | nj=1 Cj = 0.5nE, λW T Pi > Ci }. Then, these predictions of Nash equilibrium lead us to the following hypotheses. Hypothesis 4 : When the goods is provided in previous period, the previous total payment of the group decreases his/her real payment. Hypothesis 5 : When the goods is not provided in previous period, it is not always true that the previous total payment of the group decreases his/her real payment. If both hypothesis 4 and 5 are accepted, the theoretical prediction is supported.
4.2
Estimation Model
Censored Regression Model In this subsection, we introduce our estimation model that is used to test these hypotheses. Table 2 shows a list of variables used in this analysis. Because our dependent variable, real payment in stage 3 (Ri), runs from 0 though 10, we apply a censored regression model to estimations (Greene, 2003). Let us define that R∗i is a latent variable of subject i’s real payment Ri , and xi is an explanatory variable vector which is explained in table 2. Then, the censored estimation model is as follows. R∗i = β xi + εi , εi ∼ N (0, σ 2).
(6)
This equation shows a latent underlying regression. εi is an error term with normal distribution. Observed dependent variable is described as follows. ∗ ∗ Ri , if 0 ≤ Ri ≤ 10 . (7) Ri = 0, if R∗i < 0 ∗ 10, if Ri > 10 Then, we can use maximum likelihood method, and get maximum likelihood estimator (MLE). Empirical models We show our empirical specifications to test hypotheses. First, model 1 is the estimation model examining hypothesis 1 and 2. And model 2 is a full model of model 1 with subject i’s socio-economic attributes (Zi) including 12
Table 2: Descriptions of Variables Variable R WTP Coop P rof it Group δ Success δ F ail GroupS GroupF IN COM E SEX AGE
Descriptions Real Payment in stage 3 (dependent variable) WTP in stage 2 (×10−2 ) Contribution Level in stage 1 (mean of contribution) previous profit in stage 3 previous Total Payment of the group in stage 3 dummy that equals 1 if provided in previous time dummy that equals 1 if not provided in previous time = δ Success · Group = δ F ail · Group income (×10−7 ) gender (male= 1, female= 0) age
Mean 6.53 27.09 4.81 13.71 22.91 29.54 18.37 4.2 0.5 39.38
income, gender, and age. Model 1:R∗i = α + β1 ln(W T Pi ) + β2 Coopi + εi Model 2:R∗i = α + β1 ln(W T Pi ) + β2 Coopi + βz Zi + εi Here, hypothesis 1 predicts that β1 has a positive sign and hypothesis 2 predicts that β2 also has a positive sign. Next, the model 3 is the estimation model examining hypothesis 3. And model 4 is a full model of model 3 with subject i’s socio-economic attributes. Model 3:R∗it = α + β1 ln(W T Pi ) + β2 Coopi + β3 P rof itit + β4 Groupt + εit Model 4:R∗it = α+β1 ln(W T Pi )+β2 Coopi +β3 P rof itit +β4 Groupt +βz Zi +εit Here, hypothesis 3 predicts that β4 has a negative sign. Next, the model 5 is the estimation model examining hypothesis 4 and 5. And model 6 is a full model of model 5 with subject i’s socio-economic attributes. Model 5:R∗it = α + β1 ln(W T Pi ) + β2 Coopi + β3 P rof itit + β4 GroupSt + β5 GroupFt + εit Model 6:R∗it = α + β1 ln(W T Pi ) + β2 Coopi + β3 P rof itit + β4 GroupSt + β5 GroupFt + βz Zi + εit
13
Variance 10.02 564.77 6.07 12.17 47.87 10.72 22.45 2.48 0.25 109.38
Here, hypothesis 4 predicts that β4 has a negative sign. On the other hand, hypothesis 5 predicts that β5 is not significant. We estimate these 6 empirical models, and test our 5 hypotheses with estimated parameters.
5
Estimation Results
Summary of Estimation Results The estimation results are shown in table 3. Numbers in parentheses are t-value. And two asterisks are significance level of 5 %, three asterisks are significance level of 1 %. Let us note that all models perform quite well in that almost all the explanatory variables included show the expected sign of coefficients with statistical significance. That most variables which are significant in one model are also significant in others indicates that our basic specification is robust. Before hypotheses testing, we overview the estimation results of the control variables. First, about subjects’ socio-economic attributes, IN COM E and AGE are not statistically significant in all models. SEX is significant at 1 % level and has a negative sign in model 4 and 6. That is, real payment in stage 3 is lower for male as comparing with female. Second, P rof it is significant at 1 % or 5 % level and has a positive sign. That is, this result indicates that the higher profit in previous period is the more real payment. We compare our models from the viewpoint of fitting measures including AIC(Akaike information criterion) and BIC(Bayesian information criterion)5 . Comparing with model 2, model 1 is supported in both fitting measures. Comparing with model 3 and 5, model 4 and 6 are supported in AIC. On the other hands, model 3 and 5 are supported in BIC. Therefore, these results indicate that socio-economic attributes make little difference with model fitting. Then, we test our hypotheses by using the estimation results of model 1, 3, and 5 without socio-economic attributes. Hypothesis Testing First of all, we examine hypothesis 1 and 2. From the estimation result of model 1, the coefficient of ln(W T P ) is significant at 1 % level and has a positive sign. This result shows that true WTP has positive effects on real payment and hypothesis 1 is supported. Therefore, the claim is rejected that there are no relation between real payment and WTP. This result supports 5
AIC and BIC are respectively defined as follows, AIC = −2 ln LL + 2k, BIC = −2 ln LL + k ln N , where k is the number of estimated parameters, and N is the number of samples.
14
Table 3: Estimation Results: Censored Regression Models Constant ln(W T P ) Coop
Model 1 1.531 (-1.178) 0.911*** (2.605) 0.548*** (3.176)
Model 2 1.125 (0.618) 0.813** (2.344) 0.532*** (2.946)
P rof it Group GroupS
Model 3 -1.245 (-0.649) 1.081*** (4.093) 1.153*** (6.412) 0.328** (2.558) -0.201*** (-3.053)
Model 4 -1.465 (-0.707) 0.894*** (3.416) 1.085*** (5.933) 0.269** (2.193) -0.187*** (-2.979)
GroupF IN COM E SEX AGE Sigma LL N of Obs AIC BIC
2.536*** (7.793) -84.91 40 177.819 184.575
0.076 (0.259) -1.282 (-1.609) 0.027 (0.589) 2.414*** (7.782) -83.29 40 180.58 192.402
3.673*** (13.259) -330.921 160 673.841 692.292
0.149 (0.682) -1.861*** (-3.093) 0.046 (1.354) 3.484*** (13.287) -324.375 160 666.749 694.426
Model 5 -3.629 (-1.392) 1.117*** (4.217) 1.215*** (6.534) 0.399*** (2.884)
Model 6 -3.582 (-1.365) 0.929*** (3.538) 1.153*** (6.061) 0.339** (2.537)
-0.183*** (-2.751) -0.127 (-1.512)
-0.172*** (-2.702) -0.119 (-1.492) 0.166 (0.761) -1.894*** (-3.157) 0.041 (1.186) 3.468*** (13.297) -323.488 160 666.976 697.728
3.655*** (13.268) -329.985 160 673.969 695.496
Numbers in parentheses are t-value. **Significance level of 5%. ***Significance level of 1%.
15
previous studies partly. Next, Coop is statistically significant at 1 % level and has a positive sign. This result shows that contribution level has positive effects on real payment and hypothesis 2 is supported. Therefore, real payment is influenced by cooperative tendency as well as true WTP. This result indicates real payment is possibly biased. In addition, this means that interpreting Coop as ”warm glow”, a warm glow occurs not only in hypothetical setting but also in real payment. Second, we examine hypothesis 3. From the estimation result of model 3, the coefficient of Group is statistically significant at 1 % level and has a negative sign. This result shows that total payment of a group has negative effects on real payment and hypothesis 3 is supported. Therefore, there is tendency that subjects reduce their real payment depending on other’s payment. This result indicates that subjects’ decision making is influenced by others’ decision making if subject slightly considers that the provision of environmental goods is decided by social members’ decision making. Third, we examine hypothesis 4 and 5. These hypotheses analyze the hypothesis 3 in more depth from the viewpoint of Nash equilibrium. From the estimation result of model 5, the coefficient of GroupS is statistically significant at 1 % level and has a negative sign. On the other hand, the coefficient of GroupF is not significant. That is, when the goods is provided in the previous period, total payment of the group decrease his/her real payment. However, when the goods is not provided in the previous period, it does not necessarily result in the same effect above. Therefore, these results support the prediction of Nash equilibrium.
6
Concluding Remark
This paper analyzes the main factors of subject’s decision making in a threshold environmental goods experiment which not only controls the hypothetical nature in both payment and provision but also allows for the possibility of free-riding behavior. Our first finding is that the subject’s WTP increases his/her real payment. Therefore, the argument is rejected that there are no relation between real payment and WTP. Because this result means real payment can explain the true WTP, this result supports previous studies of hypothetical bias partly. Our second finding is that the subject’s contribution level increases his/her real payment. This result indicates real payment is possibly biased in the provision of environmental goods because real payment is influenced
16
by cooperative tendency as well as true WTP. If we interpret Coop as a warm glow of contributing to environmental goods, the warm glow occurs not only in hypothetical settings but also in real settings. Our third finding is that total payment of the group decreases their real payment. That is, there is the tendency that subjects reduce their real payment depending on others’ payment. This result shows that subjects’ decision making is influenced by others’ decision making if subject slightly considers that the provision of environmental goods is decided by social members’ decision making, even if researchers assume to remove the influences of others. This result indicates the existence of free-riding behavior as the Nash equilibrium predicts and it is the viewpoint which the previous studies lacked. Moreover, this result theoretically and experimentally supports the results of the meta-analysis conducted by List and Gallet (2001) and Murphy et al. (2005), that the magnitude of hypothetical bias is statistically more for public goods as compared to private ones. Previous studies of hypothetical bias have assumed that real payment was equal to the true WTP (i.e. compensating surplus or equivalent surplus), without controlling the hypothetical nature in provision and the possibility of free-riding. Our study recreates in a laboratory more realistic situation by adapting the framework of public goods experiments and analyzes the relation between real payment and true WTP. As remarked above, we find that real payment for provision of public goods is influenced by something like warm glow and strategic bias, so it is possibly biased.
17
References [1] Andreoni, J.(1990). ”Impure Altruism and Donations to Public Goods: A Theory of Warm-glow Giving” Economic Journal, 100 , 464-477 [2] Bohm, P.(1972). ”Estimating Demand for Public Goods: An Experiment” European Economic Review, 3 , 111-130 [3] Brookshire, D.S. and Coursey, D.L.(1987). ”Measuring the Value of a Public Good: An Empirical Comparison of Elicitation Procedures ” American Economic Review, 77(4) , 554-566 [4] Brown, T.C.; Champ, P.A.; Bishop, R.C. and McCollum, D.W.(1996). ”Which response format reveals the truth about donations to a public good? ” Land Economics, 72(2) , 152-166 [5] Burlando, R.M. and Guala, F.(2005). ”Heterogeneous Agents in Public Goods Experiments ” Experimental Economics, 8 , 35-54 [6] Cadsby, C.B. and Maynes, E.(1999). ”Voluntary Provision of Threshold Public Goods with Continuous Contributions: Experimental Evidence” Journal of Public Economics , 71 , 53-73 [7] Croson, R. and Marks, M.(2000). ”Step Returns in Threshold Public Goods: A Meta- and Experimental Analysis ” Experimental Economics,2 , 239-259 [8] Croson, R. and Marks, M.(1998). ”Identifiability of Individual Contributions in a Threshold Public Goods Experiment ” Journal of Mathematical Psychology , 42 , 167-190 [9] Cummings, R.G. and Harrison, G.W.(1994). ”Was the Ohio Court Well Informed in its Assessment of the Accurach of the Contingent Valuation Method? ” Natural Resources Journal , 34 , 1-36 [10] Foster, V.; Bateman, I.J. and Harley, D.(1997). ”Real and Hypothetical Willingness to Pay for Environmental Preservation: A NonExperimental Comparison ” Journal of Agricultural Economics, 48 , 123-138 [11] Friedman, D. and Sunder, S.(1994). Experimental Methods: A Primer for Economists , Cambridge University Press [12] Greene, W.(2003). Econometric Analysis: Fifth Edition , Prentice Hall 18
[13] Harrison, G.W. and Rutstrom, E.E.(2005) ”Experimental Evidence on the Existence of Hypothetical Bias in Value Elicitation Methods ” in Plott, C.R. and Smith, V.L. (ed.) Handbook of Experimental Economics Results, North-Holland [14] Harrison, G.W.; Beekman, R.L.; Brown, L.B.; Clements, L.A.; McFaniel, T.M.; Odom, S.L. and Williams, M.B.(1999). ”Environmental Damage Assessment with Hypothetical Surveys: The Calibration Approach” in Boman, M.; Brannlund, R. and Kristrom, B. (ed.) Topics in Environmental Economics Kluwer Academic Publishers [15] Hausman, J.A. (ed.) (1993). Contingent Valuation: A Critical Assessment , Elsevier Science Publishers [16] Hoehn, J.P. and Randall, A.(1987). ”A Satisfactory Benefit Cost Indicator from Contingent Valuation ” Journal of Environmental Economics and Management , 14 , 226-247 [17] Johannesson, M.; Liljas, B. and Oconor, R.M.(1997). ”Hypothetical versus Real Willingness to Pay: Some Experimental Results ” Applied Economics Letters, 4 , 149-151 [18] Ledyard, J.O.(1995) ”Public Goods: A Survey of Experimental Research ” in Kagel, J.H. and Roth, A.E. (ed.) The Handbook of Experimental Economics, Princeton University Press , 111-194 [19] List, J.A. and Gallet, C.A.(2001). ”What Experimental Protocol Influence Disparities Between Actual and Hypothetical Stated Values?: Evidence from a Meta-Analysis ” Environmental and Resource Economics , 20 , 241-254 [20] Marks, M.B. and Croson, R.T.A.(1999). ”The Effect of Incomplete Information in a Threshold Public Goods Experiment ” Public Choice, 99 , 103-118 [21] Mitani, Y. and Kuriyama, K.(2005) ”Measuring the Value of the Vegetation Restoration: Capturing Preference Heterogeneity”, European Association of Environmental and Resource Economics (EAERE) 14th Annual Conference, Bremen, Germany [22] Mitchell, R. and Carson, R.(1989) Using Surveys to Value Public Goods: The Contingent Valuation Method , Resources for the Future
19
[23] Murphy, J.J.; Allen, P.G.; Stevens, T.H. and Weatherhead, D.(2005) ”A Meta-Analysis of Hypothetical Bias in Stated Preference Valuation ” Environmental and Resource Economics, 30 , 313-325 [24] Murphy, J.J.; Allen, P.G.; Stevens, T.H. and Weatherhead, D.(2003) ”Revisiting the Data and Estimation in List and Gallet (2000) ” [25] Neill, H.R.; Cummings, R.G.; Ganderton, P.T.; Harrison, G.W. and McGuckin, T.(1994). ”Hyphothetical Surveys and Real Economic Commitments ” Land Economics, 70(2) , 145-154 [26] Rondeau, D.; Schulze, W.D. and Poe, G.L.(1999). ”Voluntary Revelation of the Demand for Public Goods Using a Provision Point Mechanism ” Journal of Public Economics, 72 , 455-470 [27] Rose, S.K.; Clark, J.; Poe, G.L.; Rondeau, D. and Schulze, W.D.(2002). ”The Private Provision of Public Goods: Tests of a Provision Point Mechanism for Funding Green Power Programs ” Resource and Energy Economics, 24 , 131-155 [28] Seip, K. and Strand, J.(1992) ”Willingness to Pay for Environmental Goods in Norway: A Contingent Valuation Study with Real Payment ” Environmental and Resource Economics , 2(1) , 91-106 [29] Shogren, J.F.(2003) ”Experimental Methods and Valuation”, in Maler, K. and Vincent, J.R. (ed.) Handbook of Environmental Economics North-Holland Elsevier [30] Smith, V.L.(1976). ”Experimental Economics: Induced Value Theory ” American Economic Review, 66(2) , 274-279
20
