Experimental design Directory UMM :Data Elmu:jurnal:J-a:Journal of Economic Behavior And Organization:Vol43.Issue1.Sept2000:

K.M. Brown, L.O. Taylor J. of Economic Behavior Org. 43 2000 127–139 129 markets — stated values are significantly higher in the hypothetical treatments. 3 Differ- ences in responses under hypothetical and real conditions are attributed to hypothetical bias, which refers to the overstating of true values for a public good when the payment decisions are not binding. 4 While this literature has focused on behavioral differences as related to ex- perimental treatments, no one has examined the potential for gender-related behavioral dif- ferences in these types of experiments, i.e., for gender differences as related to hypothetical bias. The potential for gender differences in hypothetical bias is suggested by the work of Gilligan 1982 in which she found that females think about and act on moral dilemmas in a more inclusive manner, taking relationships into consideration, whereas males are more concerned with obligations and rules. The experimental literature using laboratory public goods has interpreted her work as implying that females are less likely to free-ride than males Brown-Kruse and Hummels, 1993; Nowell and Tinkler, 1994 or that females are more likely to respond to context than males Cadsby and Maynes, 1998. We ex- tend this notion and propose that if females are more likely to respond to context than males, then females would be more likely to respond to the market context than males. Hence, females would be more likely to truthfully reveal their WTP in the hypotheti- cal treatment, where the true WTP is assumed to be represented by the responses in the real treatment. That is to say, females may be more likely to search their preferences in response to the interviewer’s request and consider the context of the valuation scenario more closely than males. 5 If this is the case, then we would expect females to exhibit less hypothetical bias than males in experiments using a non-laboratory public good. To test this hypothesis, this research elicits values for a non-laboratory public good from a sample of subjects in two distinct experimental treatments: a hypothetical treatment and a real treatment. Results from these two treatments are used to test for differences in values stated by females and males within a treatment; differences in values by treatment within a gender i.e., for hypothetical bias by gender; and differences in hypothetical bias across genders.

2. Experimental design

The experimental design used in this research consists of an in-person survey that was ad- ministered to 488 student and adult subjects in 25 experimental sessions between September 3 See Cummings and Taylor 1999 or Smith and Mansfield 1998, for examples in which differences are not found between responses in hypothetical and real treatments. 4 In this type of research, it could be the case that responses in the hypothetical treatments are biased upward, while responses in the real treatments are biased downward due to free-riding. It has been difficult to disentangle these effects in experiments using non-laboratory public goods. The validation literature typically assumes that differences in behavior between hypothetical and real treatments are due to hypothetical bias, not free-riding, following the results of Cummings et al. 1995 as well as Taylor 1998. Cummings et al. find evidence of hypothetical bias with a private good where there is no incentive to free ride. Taylor uses an incentive compatible revelation mechanism for a public good and still finds evidence of hypothetical bias. 5 In other words, since a contingent valuation survey involves a hypothetical transaction, the subjects’ responses are disciplined only by their efforts to cooperate and respond to questions by placing themselves, as closely as possible, in the situation being posed. 130 K.M. Brown, L.O. Taylor J. of Economic Behavior Org. 43 2000 127–139 1997 and September 1998. 6 The same interviewer conducted all surveys, which were si- multaneously administered to the subjects in each particular session. Student subjects were recruited through on-campus organizations at Georgia State University, such as fraternities and sororities and academic clubs. Adult subjects were recruited through church groups, computer clubs, and other community organizations in the Atlanta, GA metropolitan area. To recruit subjects, the interviewer contacted social and business groups and requested per- mission to conduct the survey at the beginning of a regularly scheduled group meeting. Participating groups were randomly assigned to either a hypothetical or a real treatment. In exchange for time to conduct the survey during the group’s meeting, the Environmental Policy Program at Georgia State University paid 5 per completed survey to each partic- ipating group’s general fund. The monetary payment served as an incentive for groups to participate, but since the money was not paid directly to individuals the payment should not have affected their responses. The survey involves contributions to the Nature Conservancy, a national nonprofit orga- nization that developed and now sponsors the Adopt an Acre program. The Adopt an Acre program works by allowing individuals to contribute money to directly purchase, and place under protection, sensitive rainforest land. Each year the Nature Conservancy focuses their fund-raising efforts on a particular rainforest. At the conclusion of a fund-raising year a new Adopt an Acre program is initiated with another rainforest, however the previous projects remain active in that conservation and protection efforts continue through the funds raised in a given year. For this research, subjects contribute to the Nature Conservancy’s rainforest project in Costa Rica, which was the focus of the Adopt an Acre program in 1994 and 1995. While the Costa Rican project was closed to the general public for fund-raising during the time the surveys were conducted, the Nature Conservancy agreed to allow participating sub- jects a one-time opportunity to contribute to the Costa Rican project as part of this research. 7 We employ two experimental treatments using this good: a hypothetical treatment where no payments are expected from the subject as a result of their responses and a real treatment where actual payments are expected if the subject reveals a positive value. Following a de- scription of the rainforest in Costa Rica and its destruction issues, the payment mechanism is implemented as follows. In the real treatment, each survey packet includes a payment form and a stamped envelope, pre-addressed to the Nature Conservancy. Subjects write their maximum willingness to pay on their survey questionnaire and on the payment form, both of which include a subject identification number. Everyone is told to take the payment form and stamped envelope home with them, regardless of whether or not they state a positive value. The interviewer instructs those who state a positive value to place a check for the correct amount in the envelope with the payment form and mail it directly to the Nature 6 Student and adult categories are not necessarily mutually exclusive students are adults, while subjects from the adult subject pool may be students. However, we use this classification to refer to the method in which the subject was recruited. 7 A closed project was explicitly chosen in order to minimize the potential for individuals to free-ride off the contributions of the general public. It should be noted, however, that this experimental design does not permit a distinction between the sources of behavioral differences as specifically related to hypothetical bias versus free-riding. The one-time contribution offer is an effort to reduce the free-riding effects, however it must be recognized that free-riding could still exist. Thus, attributing all of the differences in hypothetical and real behavior to hypothetical bias could be overstating the bias that actually exists. K.M. Brown, L.O. Taylor J. of Economic Behavior Org. 43 2000 127–139 131 Conservancy; those who state a zero value are told to throw away the payment form and en- velope in the privacy of their own home. Once the Nature Conservancy receives a check, they verify that the amount on the check matches the amount written on the payment form and then mails the payment form back to Georgia State University so that it can be matched by iden- tification number to the subject’s questionnaire subjects are also told this information. 8 The hypothetical treatment consists of an identical description of the payment mecha- nism with the exception that subjunctive language is used and subjects are not provided with payment forms or envelopes. In addition, subjects are reminded several times in the hypo- thetical treatment that they are participating in a hypothetical survey and are not actually being given the opportunity to send money to the Nature Conservancy. This experimental design allows us to test several hypotheses related to gender. First, we test for gender differences in stated values in the hypothetical treatment and then repeat this test for values elicited in the real treatment. These comparisons allow us to examine the preferences of females versus males for the public good offered in these experiments. While we have no expectations as to whether females or males might report higher values for this good, ceteris paribus, we do expect the results to be consistent across treatments if there are no behavioral effects due to the valuation mechanism. In other words, if males state higher WTP than females in the hypothetical treatment, then we might also expect males to state higher WTP than females in the real treatment. Second, and related to the previous tests, we conduct out-of-sample tests to examine whether or not hypothetical bias exists for females and males and whether or not it is differentiated according to gender. Following Gilligan, and the interpretation of her work by Cadsby and Maynes, we might expect hypothetical bias to be smaller for females than for males if females are more likely to respond to the market context.

3. Empirical results