Testing commitment cost theory in choice experiments.
Consumer willingness to pay (WTP) for private as well as public goods is an important indicator of consumer response to different choice contexts. On the basis of the Hicksian welfare theory, the WTP can be interpreted as the compensating variation (CV) assuming that individuals' choice decisions regarding the value of a good are made in certainty and under static conditions (Mitchell and Carson 1989; Smith 2000; Zhao and Kling 2004). However, in real purchasing situations, individuals might be uncertain about the utility they can derive from a good or service.
Uncertainty in decision making is a crucial aspect in various economic settings (e.g., financial investment and environmental policy) in which agents generally make choice decisions without knowing their effects on future rewards (Arrow and Fisher 1974; Dixit 1992; Dixit and Pindyck 1994; Fisher 2000). In addition, individuals' uncertainty about the product may be a key factor in the adoption of new or novel products (Castano et al. 2008; Hoeffler 2003). Moreover, consumers' uncertainty about the value of the good has often been associated with the degree of availability of product information. In this regard, several studies have documented that, depending on the kind of information (positive or negative), individuals' WTP for a good or service can increase or decrease when the information is provided, especially in cases when individuals are not familiar with the good in question (Bower, Saadat, and Whitten 2003; Corrigan et al. 2009; Depositario et al. 2009; Hoehn and Randall 2002; Lusk et al. 2004; Marette et al. 2008; Meenakshi et al. 2012; Nayga, Aiew, and Nichols 2005; Protiere et al. 2004; Tkac 1998).
However, in real purchase or choice situations, consumers may not be able to acquire information at the moment of purchase. As such, when uncertainty exists regarding the features of a good, they either take a risk and purchase the product immediately or delay the purchase until they obtain more knowledge about the product's quality. Furthermore, individuals might have the opportunity to consider the purchase and return the product later if they are uncertain, among others, about whether its use is beneficial. Hence, in contrast with assumptions of neoclassical theory, in real choice settings, choices are typically made in a relatively dynamic context, in which individuals have the option to delay the transaction when future information can be gathered and to return the product in case they are uncomfortable with their purchase (Corrigan 2005; Corrigan, Kling, and Zhao 2008 ; Kling, List, and Zhao 2013 ; Lusk 2003; Zhao and Kling 2004).
Individuals' decision making in dynamic settings has been significantly investigated in environmental economics and finance (Arrow and Fisher 1974; Dixit 1992; Dixit and Pindyck 1994). However, consumers' choice behavior in a dynamic context is still a scarcely explored issue in the literature. Zhao and Kling (2001, 2004) re-examined, for the first time, the quasi-option value (QOV) concept to explain consumers' WTP formation, (1) demonstrating that theoretical divergence exists between static and dynamic welfare measures. The authors assume that in real choice situations, consumers' WTP not only depends on the intrinsic value of the good but also on a variety of factors such as the level of uncertainty about the good, the timing of decision making, and the degree of reversibility of a transaction (Zhao and Kling 2001, 2004). Hence, committing to a decision at the moment of the transaction may represent a cost for the individual. This cost has been termed by Zhao and Kling (2001, 2004) as the "commitment cost" (CC), that is, the "cost of forgoing the opportunity to learn more about the value of a good if a purchase is made today" (Lusk and Shogren 2007, 43). Theoretically, CC represents the difference between consumers' WTP and the static Hicksian CV when (1) individuals have uncertainty about the value of a good, (2) there is the possibility of delaying a purchase and gathering future information, and (3) the degree of irreversibility of a decision varies (Lusk 2003; Zhao and Kling 2004). Zhao and Kling (2004) stated that if individuals' uncertainty about the value of a good decreases, the CC related to the choice of making the purchase today decreases while individuals' WTP increases. On the other hand, in cases when consumers need to consider the possibility of gathering additional information in the future, their WTP today decreases and CC increases. Finally, in cases when the reversibility of the purchase is easier, the CC for buying today decreases and individuals' WTP increases.
The aim of this research is to test potential CC formation in individuals' product evaluations in choice experiments. We performed a field choice experiment focused on the creation of different decision-making settings using different treatments which are aimed at addressing the three main aspects of CC theory: change in degree of uncertainty, effect of delayed information, and change in the degree of irreversibility of the purchase.
This topic is important since, despite the intuitive appeal of CC theory, only a few studies have examined this theory empirically and they have done it using different methodological approaches. Moreover, the majority of these studies tested only some aspects of CC theory (see Table 1 for a list of studies on testing CC theory).
For instance, Corrigan, Kling, and Zhao (2008) investigated individuals' choice behavior only in the case of the CC related to the option of delaying the purchase to gather information in the future. The authors performed a hypothetical referendum-format survey in Iowa to estimate residents' valuation of improved water quality of Clear Lake. Their results show that respondents were less inclined to vote yes and therefore less inclined to pay a price premium for the actualization of the referendum in case they were offered the possibility of delaying the vote and acquiring new information from studying the lake. The authors concluded that when the knowledge of the good in consideration is low, making a forced decision leads to the formation of a CC. Hence, this study omitted relevant aspects of CC theory, such as the effect of a change in the degree of individual uncertainty and decision reversibility. In addition, given their use of a nonmarket good, the authors relied on the use of a contingent valuation method (CVM) and not on a nonhypothetical preference-elicitation mechanism. Corrigan (2005) used an experimental auction (EA) approach to estimate individuals' WTP formation in dynamic settings by investigating both option values related to the delay and the degree of reversibility of the transaction; however, he did not examine the degree of individuals' uncertainty about the value of the good. His results indicate that participants' WTP for a coffee mug was higher for subjects who perceived that reversing the transaction (selling the good outside of the experiment) was more difficult than delaying the transaction (buying the good outside of the experiment). Respondents' attitudes toward delaying and reversing the transaction were not experimentally elicited; rather, they used information from a self-reported survey to test or control for these factors in their empirical analysis. Similarly, Kling, List, and Zhao (2013) assessed the difficulty in delaying or reversing the transaction using a field experiment. In order to create a more realistic dynamic context, they gave groups of respondents the opportunity to purchase or sell the product (i.e., sports cards) the week following the experiment if they did not purchase or sell them during the experiment. (2) The results from their experiment confirmed a disparity between WTP and willingness to accept (WTA) under dynamic purchasing conditions. Although their study explored the effect of potential future information and reversibility of the transaction, it did not investigate the effect of a change in the degree of uncertainty about the value of the good on respondents' valuations. Only the study by Lusk (2003) investigated the formation of CCs by testing in a laboratory experiment all three aspects of the theory, namely the degree of individuals' uncertainty regarding the value of the good, potential future information, and purchase reversibility. Using a lottery ticket and a mug auction, three treatments were utilized in his study. These treatments differed depending on (1) the degree of information regarding the value of the goods in question, (2) the degree of potential future learning, and (3) the degree of reversibility of the transaction. The findings from Lusk's study only partially confirmed CC theory. Specifically, no significant difference in terms of WTP was found by the author in the case of less or more uncertainty and reversibility.
This study builds on the pioneering work of Lusk (2003) by assessing the effects of provision of information (present and future) and reversibility of purchase on WTP values from a nonhypothetical or real choice experiment (RCE). While the study of Lusk (2003) used a laboratory experiment with a subject pool of students, we conducted, for the first time in the literature, a field experiment on real shoppers in a supermarket. In addition, we chose to use an RCE approach because RCEs more closely represent individuals' purchasing behavior in a supermarket setting than EAs do. In contrast to RCEs, EA mechanisms can be characterized by peer pressure and the need to make decisions without knowing the final price (Akaichi, Nayga, and Gil 2013; Gracia, Loureiro, and Nayga 2011 ; Grebitus, Lusk, and Nayga 2013).
Another new dimension in this study is our use of an unusual food product in the area of interest. Consumers' uncertainty is a relevant topic in the literature related to food consumption. Examining the dynamic choice context for an unusual food product is an important issue because new products generally embed a source of uncertainty in consumers' choices that may affect CC formation. We used apple sauce as the product of interest because while it is largely consumed in North American and North European countries, it is a food product that has been introduced in the Italian market as a healthy food product only recently; therefore, it is not considered a traditional food of the area of interest, that is, Italy. However, in food choice settings, the novelty of a food product is not the only feature that can produce uncertainty in consumer decision making. For instance, consumers' uncertainty about quality features of food products has been mostly associated with credence attributes such as safety, origin, and sustainability (Aprile, Caputo, and Nayga 2012; Caswell and Mojduszka 1996; Costa-Font, Gil, and Traill 2008; Grunert 2005; Grunert, Hieke, and Wills 2014; Grunert et al. 2001; Van Wezemael et al. 2010; Vermeir and Verbeke 2006). This is because credence attributes represent those features of the product that individuals cannot personally evaluate before or after consumption. Hence, consumers' valuations of the credence attributes depend on their level of trust in the product claims and the sources of these claims. Therefore, we used a food product with credence attributes (i.e., organic and locally produced) so that we could test CC theory with a higher degree of uncertainty concerning individuals' valuation.
To summarize, we conducted for the first time in the literature a field experiment that was aimed at identifying and creating a set of characteristics necessary for CC formation by (1) inducing a higher degree of uncertainty in choice making with the use of an unusual food product characterized by credence attributes and (2) by using different treatments to address the three main predictions of CC theory. The originality of this study also involves the testing, for the first time in the literature, of CC formation in a choice experiment context. This topic is important because choice experiments are one of the most widespread preference elicitation mechanisms used by applied economists. Choice experiments normally ask respondents to make irreversible choices in the moment, and they are based on the assumption that the decision maker has access to and makes use of all relevant information concerning the good of interest when making decisions. However, real-world choices are usually made in a dynamic context in which individuals have the option to delay or reverse a transaction owing to, among others, uncertainty about the product. While the choice experiment approach has been implemented by several studies to investigate information effects on consumers' valuation for various products, to our knowledge, no study has tested the effects of option values on obtaining future information and of reversing purchase transactions on consumers' choice behavior in a RCE context. RCEs are becoming increasingly popular in studies eliciting consumers' preferences and WTP for various goods because they provide more realism by allowing the conduct of choice experiments in a nonhypothetical setting. Hence, an evaluation of the potential effects of CC concepts in an RCE context may be helpful in the design of future choice experiment studies.
The remainder of this article is structured as follows. First, we describe the experimental design and the econometric models implemented in our study. Then, we present the results we obtained from the performance of our RCE. Finally, we discuss the study results, propose some conclusions, and provide suggestions for future research in the area of interest.
II. EXPERIMENTAL DESIGN
The data used in this study are drawn from a field RCE involving 248 consumers in a hypermarket located in Bologna, a city in the Emilia Romagna region of Italy. Food shoppers were randomly intercepted and recruited at the entrance of the retail store. They were informed about the opportunity to participate in a survey on consumers' valuations for apple sauce. When approaching the participants, interviewers asked them a set of screening questions: whether they were the main household food shoppers, whether they were at least 18 years old, and whether they were available to taste different types of apple sauce. If the responses to all of these questions were affirmative, the interviewer started the RCE. In the case of negative responses, the interviewer continued to randomly select other customers and ask the screening questions until eligible participants were found. Each participant was incentivized with a 5 [euro] purchase coupon.
Three attributes (price, production method, and area of production) were used to describe the different types of apple sauce. Four price levels were specified to approximately reflect the actual market prices for the apple sauce products (0.95 [euro], 1.45 [euro], 1.95 [euro], 2.45 [euro]). (3) The two-level production attribute was specified as either organic or conventional. Lastly, for the area of production attribute, we used two levels: locally produced and not locally produced. All apple sauces used in the study were produced in Italy, but the ones produced outside the borders of the Emilia Romagna region were defined as nonlocally produced and those from Emilia Romagna were considered locally produced.
Following Scarpa, Campbell, and Hutchinson (2007), the allocation of attributes and attribute levels for product alternatives was designed using a sequential Bayesian approach to minimize the Db error. Three different phases were performed. In the first phase, the choice set design follows Street, Burgess, and Louviere (2005). Accordingly, the selected attributes and their levels were first used to determine an orthogonal fractional factorial design, reducing the original 16 (4 x 22) combinations to just 8. Then, the generators described by Street and Burgess (2007) were used to obtain a practical set of eight pairs with a D-efficiency of 96.6%. This design was used for the pilot survey (second phase). In the last phase, we used the data from the pilot survey to estimate a multinomial logit (MNL) model whose coefficient estimates were then used as Bayesian priors to generate an efficient design (Choice Metrics 2011). This procedure has the advantage that the number of respondents necessary to reveal significant coefficients is reduced to a minimum. The resulting choice set is described in Table 2. (4)
Before answering the RCE questions, the participants were asked to taste the four types of apple sauce products (local/organic, nonlocal/ organic, local/nonorganic, nonlocal/nonorganic) in order to equalize the level of experience with the product in question among the respondents. We chose to adopt a blind tasting approach so that the organoleptic characteristics of the different kinds of apple sauce would not affect respondents' preferences for the production origin and production method attributes. After completing the blind tasting, participants also had the possibility of visually examining the apple sauce products (two cups of 100 g of apple sauce). Information regarding the RCE mechanism was then provided in detail to all participants. Specifically, they were first informed that they would face eight different choice tasks, each describing three choice options: two different apple sauce products and a "no-buy" option. Next, they were informed that after completing the eight choice tasks, one of these choice tasks would be randomly selected as the binding choice task. In case participants picked one of the two product alternatives, they would have to purchase the product he/she chose in the binding choice task. If they chose the "no-purchase" option, then they would not purchase any product and would not pay anything. Finally, the participants were clearly told that an actual payment would have to occur right after the choice experiment if they chose one of the two product options in the binding choice task and that every choice task would have the same probability to be picked as the binding choice task. Once the participants completed the RCE, they were asked to fill out a questionnaire concerning sociodemographic information.
III. EXPERIMENTAL TREATMENTS AND RESEARCH HYPOTHESES
We used four RCE treatments with a between-subjects design. Hence, each participant was randomly assigned (5) to only one of the RCE treatments. The four RCE treatments differed in terms of possibility of gaining information (present or future information) and in terms of degree of reversibility of the transaction. In the first treatment, named "control treatment" (CT), 80 respondents were introduced to the RCE without receiving information about the possibility of gaining information about the product or returning it. In the second treatment, named "treatment with immediate information" (IINT), a group of 56 individuals was provided with a brief description of the product and a brief explanation of organic certification and of "local food." In order to avoid providing information that could negatively or positively influence respondents' perceptions toward the two food claims, we decided to furnish neutral information (Aprile, Caputo, and Nayga 2012; Lusk et al. 2004). (6) In the third treatment (56 respondents), named "delayed information treatment" (DINT), respondents were informed right before the RCE that they would be provided information about organic and local food production (the same information given in the IINT treatment) after concluding their grocery shopping at the store exit. This allowed us to assess the effect of potential future information on consumers' WTP. Again, these participants were clearly told that an actual payment would have to occur right after the choice experiment if they chose one of the two product options in the binding choice task. Respondents were informed that an interviewer would be available to provide them this information after completing their purchases, if interested. They were provided with an ID number in order to be recognized by one of our interviewers. The respondents were also provided with the description of the interviewer, specifying the color of the shirt and that he/she was carrying a sign with university information. In addition, the interviewer was informed when a participant was selected to this treatment so that he/she could then readily identify the participant at the cash register area. The interviewer then approached the respondent after the cash register and asked whether he/she wanted to be given the information. Finally, the last treatment, called the "reversibility treatment" (RT), was designed to determine the effect of the possibility of reversing the transaction on respondents' WTP; that is, they could return the product if they purchased one. As such, before the RCE, the 56 participants in this treatment were informed that if they chose a product in the binding choice task, they could return the product near the store exit after concluding their grocery shopping. That is, they could return the product to our interviewer, who would then repay them the price they paid for the product if they decided to return the two cups of apple sauce. Also in this case, respondents were given an ID number, (7) provided with the description of the interviewer, and identified to the interviewer so that they could more easily be recognized and approached at the store exit. Table 3 shows a layout of the procedures followed in the RCE treatments.
With these RCE treatments, we were able to test a set of hypotheses aimed at testing CC theory in a choice context involving an unfamiliar food product and a set of credence attributes. In order to determine the effect of information on individuals' WTP, the estimates from the second and first treatment were compared. In regards to the first issue of CC theory, we tested the following hypothesis:
[H.sub.01] : ([WTP.sup.IINT] - [WTP.sup.CT]) [less than or equal to] 0
[H.sub.11] : ([WTP.sup.IINT] - [WTP.sup.CT]) > 0
A rejection of [H.sub.01] confirms that giving information reduces consumers' uncertainty regarding the value of the product. This would validate the assumption that when subjects are less uncertain about the value of a good, CCs decrease and WTP increases, as predicted by Zhao and Kling (2004).
Next, in order to answer our research question related to the effect of willingness to wait for future information, we tested the following hypothesis:
[H.sub.02] : ([WTPD.sup.INT] - [WTP.sup.CT]) [greater than or equal to] 0
[H.sub.12] : ([WTP.sup.DINT] - [WTP.sup.CT]) < 0
A failure to reject [H.sub.02] implies that when subjects expect to gather more information regarding the good, CCs increase and WTP decreases. The rejection of [H.sub.02] confirms Zhao and Kling's (2004) CC theory, which assumes that an individual's WTP today decreases if there is a possibility of gathering information in the future.
Finally, our third hypothesis is related to individuals' WTP formation in case of a change i n the degree of reversibility of the purchase. According to CC theory, individuals' WTP for a good should be higher when the individual may be able to reverse or return a purchase. Accordingly, the following hypothesis was tested:
[H.sub.03] : ([WTP.sup.RT] - [WTP.sup.CT]) [less than or equal to] 0
[H.sub.13] : ([WTP.sup.RT] - [WTP.sup.CT]) > 0
A rejection of [H.sub.03] confirms that when subjects expect an ease of reversing the transaction, CCs decrease and WTP increases, validating the prediction of Zhao and Kling's (2004) CC theory.
IV. ECONOMETRIC ANALYSIS
To test the research hypotheses concerning CC formation, we estimated the effect of the treatments on WTP estimates. The derivation of WTP measures across treatments first requires the selection of the econometric model to be used for data analysis. Different model specifications were explored, such as the MNL, the panel random parameter logit model (RPL), and the RPL with error component (RPL-EC). From this exploratory analysis, the RPL-EC model was selected. RPL-EC models are widely used in the analysis of discrete choice models in environmental economics and in food choice studies (Caputo, Nayga, and Scarpa 2013; Gracia, Loureiro, and Nayga 2011; Scarpa et al. 2013; Van Loo et al. 2014; Van Wezemael et al. 2014). The RPL-EC model was chosen because it accounts for unobserved taste heterogeneity and because our experimental design included a no-purchase option (status quo), which can cause systematic effects associated with both the status-quo and correlated random effects across utilities between product alternatives in the choice set design (Scarpa and Alberini 2005; Scarpa, Ferrini, and Willis 2007; Scarpa, Thiene, and Marangon 2007).
When estimating choice models with random coefficients, as a second step, researchers should determine the specification of the utility function. If the utility function is specified in preference space, the researchers should assume a distribution for the random coefficients and then derive the WTP for an attribute as the ratio of the attribute coefficient and an estimate of the marginal utility of money. In several choice studies that assessed differences in WTPs across choice data, the price coefficient is held constant across individuals, whereas the coefficients of the other attributes and attribute levels are treated as random variables that follow a normal distribution. This restriction is commonly used despite that it implies that the standard deviation of the unobserved utility (scale parameter) is the same across all observations. As pointed out by Train and Weeks (2005), it assures that the WTP distributions can be easily calculated from the distribution of the nonprice coefficient (normal distribution), because the two distributions take the same form. It also avoids identification problems that may occur in a model with all random coefficients (Ruud 1996; Train and Weeks 2005).
In a preference space approach, the utility function of individual n for selecting alternative j in choice situation t is a function of the price and nonprice attributes. Accordingly, in this application, the function is specified as follows:
(1) [U.sub.njl] = ASC + [[alpha].sub.1]. * PRICE + [[beta].sub.1], * [ORGANIC.sub.njt] + [[beta].sub.2] * [LOCAL.sub.njt] + [[eta].sub.njt] + [[epsilon].sub.njt]
where ASC is a dummy variable indicating the selection of the no-buy option; price (PRICE) is a continuous variable represented by the four experimental design price levels; nonprice attributes such as ORGANIC and LOCAL are dummy variables taking a value of 1 if the product carries the corresponding labels and 0 otherwise; [eta] is an error component distributed normally but with zero mean (which inflates the variance of utility for options other than the status quo, that is, the no-buy option); [[epsilon].sub.njt] , is an unobserved random term that is distributed following an extreme value type-I (Gumbel) distribution i.i.d. over alternatives and independent of [alpha] and [beta].
Using the estimated coefficients from Equation (1), we calculated the marginal WTPs across treatments as the ratio of the partial derivative of the utility function with respect to the attributes of interest and then divided by the partial derivative of the utility function with respect to the price variable. The WTPs and the standard errors of each attribute level were obtained using the Krinsky and Robb (1986) bootstrapping method, resulting in a distribution of 1,000 WTP values for each attribute. In particular, the 1,000 observations were drawn from a multivariate normal distribution parameterized using the estimated means and variances from the RPL-EC model estimated for each RCE treatment. The 1,000 generated WTP estimates were then used to perform the computational method suggested by Poe, Giraud, and Loomis (2005) to test our research hypotheses about CC formation.
Admittedly, deriving the WTP estimates from models specified in the preference space with the price coefficient treated as a fixed variable has some limitations. Hence, in addition to utility specifications in the preference space, the utility function can be expressed in the WTP space (Cameron and James 1987; Scarpa and Willis 2010; Train and Weeks 2005). In a WTP space approach, the utility is reparameterized and therefore the coefficients can be directly interpreted as marginal WTP effects (Scarpa and Willis 2010). In other words, the researchers make a prior assumption of the distribution of the WTP rather than the attribute coefficients. Several studies have reported the advantages of using WTP space instead of preference space (Scarpa and Willis 2010; Thiene and Scarpa 2009; Train and Weeks 2005). According to Scarpa and Willis (2010), for instance, the use of WTP space may be more practical for derivations of welfare estimates and when accounting for interpersonal scale variation. It also provides more reasonable distributions of WTP (Train and Weeks 2005) and it may produce more stable WTP estimates (Balcombe, Chalak, and Fraser 2009).
Hence, to assess the robustness of our results, we also specified the RPL-EC model in the WTP space in which the price coefficient is treated as a random variable. The basic specification in the WTP space can be expressed as follows:
(2) [U.sub.njt] = [[theta].sub.njt] (ASC - [PRICE.sub.njt] + [[omega].sub.1][ORGANIC.sub.njt] + [[omega].sub.2][LOCAL.sub.njt] + [[eta].sub.njt]) + [[epsilon].sub.njt].
Here, [theta] = [lambda]/[alpha], where [lambda] is the Gumbel scale parameter and a is the price coefficient. This term has a log-normal distribution, which ensures randomness of the price coefficient in a fashion correlated with scale. In addition, [eta] is the error component. As shown in de-Magistris, Gracia, and Nayga (2013), differences in WTPs between treatments involved in a certain hypothesis can be tested by conducting tests on pooled samples in which treatments are adequately identified using dummy variables. Accordingly, in our case, it can be specified as follows:
(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
The significance of the estimated 5 and their signs establish the effect of the treatment on the marginal WTP estimate of interest. A joint restriction can be tested using a likelihood ratio (LR) test, while a single restriction can be tested using a t test on the coefficient estimate. A total of three extended utility functions were specified, one for each research hypothesis to be tested.
Table 4 reports summary statistics of sociodemographic information across the RCE treatments (gender, age, education, and income). A Fisher's exact test was used to test whether our RCE treatments differ in terms of gender, age, education, and income. The results show that the hypothesis of equality of means between sociodemographic characteristics across treatments failed to be rejected at the 5% significance level. We can then affirm that participants were equally distributed across the treatments in terms of sociodemographic characteristics.
As a first approach in the analysis of treatment effect on respondents' preferences for the different types of apple sauce, we calculated the summary statistics of participants' choices for the "Buy" option (i.e., if respondents chose one of the product alternatives) and the "NoBuy" option in each treatment (Table 5). We then conducted a chi-square test in order to observe whether the percentages of "Buy" and "Nobuy" choices were equally distributed across the treatments.
The RT treatment, characterized by the reversibility of the transaction, has the highest frequency of choices for the "Buy" option, followed by the treatment when immediate information was given (IINT). The DIN treatment (with potential delayed information) has the highest frequency of choices for the "No-buy" option. Finally, results from the chi-square test show that the hypothesis of equality between respondents' choices is rejected. These initial results suggest that there are existing differences in respondents' preferences across treatments in terms of buying and no buying behavior.
Next, we tested the hypothesis of equality across RCE treatments using the LR test for three model specifications: MNL, RPL, and RPL-EC. Table 6 displays the likelihood values using the pooled data and the data for each RCE treatment.
The results show that the null hypothesis of equality across the likelihood values of pooled and segmented treatments is rejected (LR= 1428.80 for the MNL, LR = 28.96 for the RPL, LR = 28.90 for the RPL-EC). This suggests that the comparison of WTP estimates across the different treatments would be appropriate when separately estimating the models. Furthermore, when comparing model fit between MNL. RPL, and RPL-EC models, the RPL-EC model is found to be superior to the MNL and the RPL models as reflected by the increases in LL function across all RCE treatments. Hence, the RPL-EC model is used to estimate the data in all RCE treatments.
The coefficient estimates from the RPL-EC models were then used to calculate the marginal WTPs across the treatments. (8) We then tested our hypotheses about the CC formation using the combinatorial nonparametric Poe test (Poe, Giraud, and Loomis 2005), which was performed with the 1,000 WTP estimates derived by the application of the parametric bootstrapping method proposed by Krinsky and Robb (1986). Table 7 displays the marginal WTPs as well as the p value of the Poe test.
As shown from Table 7, our first hypothesis ([H.sub.01] : ([WTP.sup.IINT] - [WTP.sup.CT]) [less than or equal to] 0; [H.sub.11] = ([WTP.sup.IINT] - [WTP.sup.CT]) > 0) is rejected in the case of the organic claim, indicating that respondents' WTPs significantly increase when information about the meaning of the attributes is provided to them. In contrast to Lusk's (2003) finding that increasing the certainty about the value of a lottery ticket did not significantly increase the bids for lottery as predicted by CC theory, our result is consistent with the CC theory assumptions, albeit only in the case of the organic production attribute.
Looking at the results of our second hypothesis test, we state that the null hypothesis ([H.sub.02]: ([WTPD.sup.INT] - [WTP.sup.CT]) [greater than or equal to] 0; [H.sub.12]: ([WTPD.sup.INT] [WTP.sup.CT]) < 0) failed to be rejected. This suggests that the potential future information did not significantly affect respondents' WTP formation. This result is in contrast with Corrigan, Kling, and Zhao (2008), who reported that respondents were less willing to pay for an environmental policy when there was the possibility of acquiring delayed information. On the other hand, our finding is partially consistent with the results of Lusk (2003), who found that when participants expected to gather less information in the future, their bids did not vary significantly for coffee mugs while their bids significantly decreased for lottery tickets, confirming CC theory.
In addition, in order to test whether respondents' willingness to receive the delayed information might have affected CC formation, we included participants' acceptance of the delayed information at the exit of the store (a dummy variable equal to 1 if participant received the information, 0 otherwise) as an interaction term in the RPL-EC model. (9) We did not find any significant effect on respondents' valuations for organic and locally produced apple sauce.
In the case of our third hypothesis (H03: ([WTP.sup.RT] - [WTP.sup.CT]) [less than or equal to] 0; [H.sub.13]: ([WTP.sup.RT] - [WTP.sup.CT]) > 0), the null hypothesis is rejected for both attributes (local and organic), indicating that the WTPs for organic as well as local labels are higher when the purchase transaction is reversible. Consistent with Kling, List, and Zhao (2013), this result confirms CC theory.
Finally, as mentioned in the data analysis section, we also estimated the RPL-EC in the WTP space to test the robustness of our results. Table 8 reports the estimates of local and organic parameters and the corresponding p values of the t test for the dummy variables, indicating the treatment effects (dtreatment). As shown, the results are consistent with those obtained using the test by Poe, Giraud, and Loomis (2005).
VI. DISCUSSION AND CONCLUSION
Neoclassical theory is based on the assumption that individuals make choices under certainty and static conditions. However, in real purchasing situations, uncertainty and the potential of delaying or reversing a transaction can affect choice decisions. Hence, the measurement of WTP under uncertainty conditions differs from Hicksian CV because of the formation of the CCs (Zhao and Kling 2001, 2004). According to CC theory (Zhao and Kling 2001, 2004), CCs decrease and WTP increases when individuals are less uncertain about the value of a good and when it is possible to reverse a transaction. On the other hand, WTP today decreases and CCs increase when potential future information can be gathered. Despite the intuitive predictions of CC theory, only a few studies have tested WTP formation in dynamic settings using experimental approaches such as referendum-format CVMs and EAs. In addition, with the exception of Lusk (2003), past studies have tested only two of the three CC concepts previously discussed; that is, they did not test the formation of the CC with a change in the degree of respondents' uncertainty about the value of the good (Corrigan 2005; Corrigan, Kling, and Zhao 2008; Kling, List, and Zhao 2013).
In this study, we revisit the three main predictions of CC theory by examining the effect of (1) a higher degree of information about the product in question, (2) potential delayed information, and (3) a change in the degree of reversibility of the purchase. Our study advances the literature in a number of ways. First, this is the first field experiment on CC theory where all three of its main predictions were distinctly tested on real shoppers in a supermarket. Second, we used an unusual food product in the area of interest, which allowed us to induce a higher degree of uncertainty in consumers' choices and therefore reproduce purchase situations where the option values related to the degree upon which present and future information and transaction reversibility are present. Hence, we considered the nature of product attributes as a source of uncertainty in consumers' choices and therefore as a source of CC formation. Finally, to the best of our knowledge, this study is the first to test CC theory in the context of RCEs, and therefore, in testing whether the construction of a dynamic decision context might be of relevance in choice experimental designs.
Our results show that WTP increases when consumers are provided with information regarding the meaning of the products of interest. This is consistent with CC theory prediction that making a choice in conditions of less uncertainty induces CC formation and therefore an increase in WTP for the good in question. However, our findings are consistent with CC theory just for the organic production attribute. At first glance, it might be possible to deduce that the cause of these diverging results might be the nature of the given information. However, we provided neutral information for both attributes by providing a simple description of the regulations concerning organic certification and local production in the Emilia-Romagna region. We chose to provide neutral information precisely in order to avoid potential induced preferences for one of the two attributes. As regards the difference between the organic and local claims, we verified that the former has a universally regulated certification characterized by a specific label. In contrast, the Italian food system still lacks regulation that governs the identification or labeling of local food products. Hence, it is possible that awareness of a controlled certification system will significantly affect individuals' decision making and induce a reduction in uncertainty for the quality of the food product in question. This finding has an important implication for the marketing of local food products. Local foods tend to have a relatively short supply chain, and consumers can seek information directly from farmers. The introduction of a "Local Food" label might play an important role in providing information and encouraging the commercialization of local food products. A caveat here is that we cannot exclude the possibility that one reason for why we did not obtain any significant increase in WTP for the locally produced apple sauce is that the information we provided might have not been convincing or might have not answered respondents' concerns regarding the origin of production. For example, it is possible that "produced in Italy" might have not been perceived as a source of less uncertainty in comparison to a label such as "regionally produced."
The second prediction of CC theory is that individuals' WTPs "today" decrease and CCs increase when potential future information can be gathered. To date, there has been little agreement on this aspect of CC theory. For instance, while Corrigan (2005), Corrigan, Kling, and Zhao (2008), and Kling, List, and Zhao (2013) found that CCs increase when potential future information can be gathered, Lusk (2003) did not find any effect on WTP for coffee mugs. We also did not observe significant reductions in the WTPs when the possibility of gaining delayed information was offered to respondents in both cases (i.e., organic certification and local production). The failure to reject our null hypothesis on this aspect of CC theory cannot be attributed to the methodological approach used to elicit consumer WTPs, because Lusk (2003) also obtained a similar outcome using an EA approach. A possible explanation can be related to the nature of the attributes used to describe our products. In particular, we used two credence attributes in our RCE design. Individuals cannot personally evaluate credence attributes before or after consumption; therefore, this may be a source of uncertainty to individuals making decisions. As such, we might have obtained different results if we used search or experience attributes. Hence, testing WTP formation in dynamic settings using search or experience attributes could be an interesting area for future research.
Moreover, although in our experimental design we aimed to avoid the formation of transaction costs, a factor that might have also affected respondents' valuations is the hassle or awkwardness of getting the future information by having to delay one's departure from the store to visit with the interviewer. However, we can only assume that the respondents who decided to receive the information did not feel "hassled" by delaying their exit from the store since they would have not done it otherwise. On the other hand, individuals might have not approached the interviewer after their grocery shopping because they might have been less interested in receiving the information. As such, in this case, the respondents might have given less importance to the option value of gathering more information in the future. We may then conclude that respondents' willingness to receive the information and, therefore, respondents' willingness to approach the interviewer did not affect individuals' evaluations. Nevertheless, we cannot completely eliminate the possibility that we could be simultaneously testing both the information and the way the information is being provided. It is possible that the same treatment but with web-based information, an email follow-up, or any number of other approaches might generate different results. Another issue worth mentioning is that in our experimental design, we did not provide the respondents an opportunity to repurchase the apple sauce after they have gained the information. It is possible that this could have significantly influenced consumers' choices given that provision of future information could reduce the current WTP only if consumers will be able to utilize the information and make corresponding purchase decisions in the future. Therefore, a test that requires not only that future information is provided but also that consumers will have chances to purchase the same goods after the information is provided should be tested in future studies. (10) These issues would be good areas for future research.
Notwithstanding these potential limitations of our study, our results strongly confirm that individuals' WTPs from RCE decrease and CCs increase in case of transaction irreversibility. Respondents' WTPs were significantly higher when they had the chance to reverse the transaction, although they decided to keep the purchased product in most cases (i.e., just one subject returned the apple sauce). This might suggest that the option value related to reversibility is a crucial aspect that should be remembered when designing RCEs. In the real market, retailers generally have policies concerning the reversibility of customers' purchases. Hence, in real purchasing situations, consumers are usually aware of the possibility that they can return the product they just purchased, suggesting that the irreversibility conditions that generally characterize RCEs might be a source of bias in individuals' WTP estimation. This finding could also be of relevance in the marketing of new food products since it suggests that an easier return policy might encourage consumers to pay more "now" for the product in question.
Overall, results from this study partially support the predictions of CC theory given that we could confirm two of the three main CC theory predictions. In particular, our results strongly confirm that transaction reversibility can significantly affect consumers' WTP formation, suggesting that this issue should be considered when designing RCEs.
ABBREVIATIONS CC: Commitment Cost CT: Control Treatment CV: Compensating Variation CVM: Contingent Valuation Method DINT: Delayed Information Treatment EA: Experimental Auction LR: Likelihood Ratio MNL: Multinomial Logit QOV: Quasi-Option Value RCE: Real Choice Experiment RPL: Random Parameter Logit RPL-EC: RPL with Error Component RT: Reversibility Treatment WTA: Willingness To Accept WTP: Willingness To Pay
Online Early publication July 19, 2016
EXAMPLE OF CHOICE TASK
A package consists of two cups of apple sauce, 100 g each. In the first alternative, the apple sauce is organic, has been produced in Italy, but outside the Emilia-Romagna region and costs I.45[euro]. In the second alternative, the apple sauce is not organic, has been produced in the Emilia Romagna region, and costs 0.95[euro]. Which one do you choose? You can also choose to buy none of these (translated from Italian language).
Apple Sauce Apple Sauce Alternative A Alternative B Organic Nonorganic Produced outside Produced in Emilia-Romagna Emilia-Romagna 1.45 [euro] 0.95 [euro]
Akaichi, F., R. M. Nayga, and J. M. Gil. "Are Results from Non-hypothetical Choice-based Conjoint Analyses and Non-hypothetical Recoded-ranking Conjoint Analyses Similar?" American Journal of Agricultural Economics, 95(4), 2013, 949-63.
Aprile, M. C., V. Caputo, and R. M. Nayga Jr. "Consumers' Valuation of Food Quality Labels: The Case of the European Geographic Indication and Organic Farming Labels." International Journal of Consumer Studies, 36(2), 2012, 158-65.
Arrow, K. J., and A. C. Fisher. "Environmental Preservation, Uncertainty, and Irreversibility." The Quarterly Journal of Economics, 88(2), 1974, 312-19.
Balcombe, K., A. Chalak, and I. Fraser. "Model Selection for the Mixed Logit with Bayesian Estimation." Journal of Environmental Economics and Management, 57, 2009, 226-37.
Bower, J. A., M. A. Saadat, and C. Whitten. "Effect of Liking, Information and Consumer Characteristics on Purchase Intention and Willingness to Pay More for a Fat Spread with a Proven Health Benefit." Food Quality and Preference, 14(1), 2003, 65-74.
Cameron, T. A., and M. D. James. "Efficient Estimation Methods for 'Closed-Ended' Contingent Valuation Surveys." Review of Economics and Statistics, 69, 1987, 269-76.
Caputo, V., R. M. Nayga, and R. Scarpa. "Food Miles or Carbon Emissions? Exploring Labelling Preference for Food Transport Footprint with a Stated Choice Study." Australian Journal of Agricultural and Resource Economics, 57(4), 2013, 465-82.
Castano, R., M. Sujan. M. Kacker, and H. Sujan. "Managing Consumer Uncertainty in the Adoption of New Products: Temporal Distance and Mental Simulation." Journal of Marketing Research, 45(3), 2008, 320-36.
Caswell, J. A., and E. M. Mojduszka. "Using Informational Labeling to Influence the Market for Quality in Food Products." American Journal of Agricultural Economics, 78(5), 1996, 1248-53.
Choice Metrics. Ngene vl.0.1 User Manual and Reference Guide. Sydney, Australia: Choice Metrics Ltd, 2011.
Corrigan. J. R. "Is the Experimental Auction a Dynamic Market?" Environmental and Resource Economics, 31(1), 2005, 35-45.
Corrigan. J. R., C. L. Kling, and J. Zhao. "Willingness to Pay and the Cost of Commitment: An Empirical Specification and Test." Environmental and Resource Economics, 40(2), 2008, 285-98.
Corrigan. J. R., D. P. T. Depositario, R. M. Nayga Jr., X. Wu. and T. P. Laude. "Comparing Open-Ended Choice Experiments and Experimental Auctions: An Application to Golden Rice." American Journal of Agricultural Economics, 91(3), 2009, 37-853.
Costa-Font. M., J. M. Gil, and W. B. Traill. "Consumer Acceptance. Valuation of and Attitudes Towards Genetically Modified Food: Review and Implications for Food Policy." Food Policy, 33(2), 2008, 99-111.
de-Magistris, T., A. Gracia, and R. M. Nayga. "On the Use of Honesty Priming Tasks to Mitigate Hypothetical Bias in Choice Experiments." American Journal of Agricultural Economics, 95(5), 2013, 1136-54.
Depositario. D. P. T. R. M. Nayga, X. Wu. and T. P. Laude. "Effects of Information on Consumers' Willingness to Pay for Golden Rice." Asian Economic Journal, 23(4), 2009, 457-76.
Dixit, A. "Investment and Hysteresis." Journal of Economic Perspectives, 6(1), 1992, 107-32.
Dixit. A. K., and R. Pindyck. Investment Under Uncertainty. Princeton, NJ: Princeton University Press, 1994.
Fisher, A. C. "Investment Under Uncertainty and Option Value in Environmental Economics." Resource and Energy Economics, 22, 2000, 197-204.
Gracia, A., M. L. Loureiro, and R. M. Nayga. "Are Valuations from Nonhypothetical Choice Experiments Different from Those of Experimental Auctions?" American Journal of Agricultural Economics, 93(5), 2011, 1358-73.
Grebitus, C., J. L. Lusk, and R. M. Nayga. "Explaining Differences in Real and Hypothetical Experimental Auctions and Choice Experiments with Personality." Journal of Economic Psychology, 36, 2013, 11-26.
Grunert, K. G. "Food Quality and Safety: Consumer Perception and Demand." European Review of Agricultural Economics, 32(3), 2005, 369-91.
Grunert. K. G., L. La, J. B. Poulsen, O. Ueland, and A. Annika. "Consumer Perceptions of Food Products Involving Genetic Modification--Results from a Qualitative Study in Four Nordic Countries." Food Quality and Preference, 12. 2001, 527-42.
Grunert, K. G., S. Hieke, and J. Wills. "Sustainability Labels on Food Products: Consumer Motivation, Understanding and Use." Food Policy, 44, 2014, 177-89.
Hoeffler, S. "Measuring Preferences for Really New Products." Journal of Marketing Research, 40(4), 2003, 406-20.
Hoehn, J. P, and A. Randall. "The Effect of Resource Quality Information on Resource Injury Perceptions and Contingent Values." Research and Energy Economics, 24, 2002, 13-31.
Kling, C. L., J. A. List, and J. Zhao. "A Dynamic Explanation of the Willingness to Pay and Willingness to Accept Disparity." Economic Inquiry, 51(1), 2013, 909-21.
Krinsky, I., and A. L. Robb. "On Approximating the Statistical Properties of Elasticities." Review of Economics and Statistics, 68(4), 1986, 715-19.
Lusk, J. L. "An Experimental Test of Commitment Cost Theory." American Journal of Agricultural Economics, 85(5), 2003, 1316-22.
Lusk, J. L., and J. F. Shogren. Experimental Auctions: Methods and Applications in Economic and Research. Cambridge: Cambridge University Press, 2007.
Lusk, J. L., L. O. House, S. R. Jaeger, M. Moore, J. L. Morrow, and W. B. Traill. "Effect of Information about Benefits of Biotechnology on Consumer Acceptance of Genetically Modified Food: Evidence from Experimental Auctions in the United States, England, and France." European Review of Agricultural Economics, 31(2), 2004, 179-204.
Marette, S., J. Roosen, S. Blanchemanch, and P. Verger. "The Choice of Fish Species: An Experiment Measuring the Impact of Risk and Benefit Information." Journal of Agricultural and Resource Economics, 33(1), 2008, 1-18.
Meenakshi, J. V, A. Banerji, V. Manyong, K. Tomlins, N. Mittal, and P. Hamukwala. "Using a Discrete Choice Experiment to Elicit the Demand for a Nutritious Food: Willingness-To-Pay for Orange Maize in Rural Zambia." Journal of Health Economics, 31(1), 2012, 62-71.
Mitchell, R. C., and R. T. Carson. Using Surveys to Value Public Goods: The Contingent Valuation Method. Washington, DC: Resources for the Future, 1989.
Nayga, R. M., W. Aiew, and J. P. Nichols. "Information Effects on Consumers' Willingness to Purchase Irradiated Food Products." Review of Agricultural Economics, 27(1), 2005, 37-48.
Poe, G. L., K. L. Giraud, and J. B. Loomis. "Computational Methods for Measuring the Difference if Empirical Distributions." American Journal of Agricultural Economics, 87(2), 2005, 353-65.
Protiere, C., C. Donaldson, S. Luchini, J. P. Moatti, and P. Shackley. "The Impact of Information on Non-health Attributes on Willingness To Pay for Multiple Health Care Programmes." Social Science and Medicine, 58, 2004, 1257-69.
Ruud, P. "Approximation and Simulation of the Multinomial Probit Model: An Analysis of Covariance Matrix Estimation." Working Paper, Department of Economics, University of California, Berkeley, 1996, 1-17.
Scarpa, R., and A. Alberini. Applications of Simulation Methods in Environmental and Resource Economics. Dordrecht, The Netherlands: Springer, 2005, 247-73.
Scarpa, R., and K. Willis. "Willingness-to-Pay for Renewable Energy: Primary and Discretionary Choice of British Households' for Micro-Generation Technologies." Energy Economics, 32(1), 2010, 129-36.
Scarpa, R., D. Campbell, and W. G. Hutchinson. "Benefit Estimates for Landscape Improvements: Sequential Bayesian Design and Respondents' Rationality in a Choice Experiment." Land Economics, 83(4), 2007, 617-34.
Scarpa, R., S. Ferrini, and K. Willis. "Performance of Error Component Models for Status-Quo Effects in Choice Experiments. Rationality in a Choice Experiment." Land Economics, 83(4), 2007, 617-34.
Scarpa, R., M. Thiene, and F. Marangon. "The Value of Collective Reputation for Environmentally-Friendly Production Methods: The Case of Val Di Gresta." Journal of Agricultural and Food Industrial Organization, 5(1), 2007, 1-26.
Scarpa, R., R. Zanoli, V. Bruschi, and S. Naspetti. "Inferred and Stated Attribute Non-Attendance in Food Choice Experiments." American Journal of Agricultural Economics, 95(1), 2013, 165-80.
Smith, V. K. "JEEM and Non-Market Valuation: 1974-1998." Journal of Environmental Economics and Management, 39(3), 2000, 351-74.
Street, D., and L. Burgess. The Construction of Optimal Stated Choice Experiments. Hoboken, NJ: John Wiley & Sons, 2007.
Street, D. J., L. Burgess, and J. J. Louviere. "Quick and Easy Choice Sets: Constructing Optimal and Nearly Optimal Stated Choice Experiments." International Journal of Research in Marketing, 22(4), 2005, 459-70.
Thiene, M., and R. Scarpa. "Deriving and Testing Efficient Estimates of WTP Distributions in Destination Choice Models." Environmental and Resource Economics, 44(3), 2009, 379-95.
Tkac, J. "Award-Winning Undergraduate Paper: The Effects of Information on Willingness-To-Pay Values of Endangered Species." American Journal of Agricultural Economics, 80(5), 1998, 1214-20.
Train, K., and M. Weeks. "Discrete Choice Models in Preference Space and Willing-to-Pay Space," in Applications of Simulation Methods in Environmental and Resource Economics, Chapter 1, edited by R. Scarpa and A. Alberini. Dordrecht, The Netherlands: Springer, 2005, 1-16.
Van Loo, E. J., R. M. Nayga, Jr, H. S. Seo, and W. Verbeke. 2014. "Visual Attribute Non-Attendance in a Food Choice Experiment: Results from an Eye-Tracking Study." Paper presented at the Agricultural & Applied Economics Association's 2014 AAEA Annual Meeting, Minneapolis, MN, July 27-29, 2014, 1-17.
Van Wezemael, L., W. Verbeke, J. O. Kiigler, M. D. de Barcellos, and K. G. Grunert. "European Consumers and Beef Safety: Perceptions, Expectations and Uncertainty Reduction Strategies." Food Control, 21(6), 2010, 835-44.
Van Wezemael, L., V. Caputo, R. M. Nayga, G. Chryssochoidis, and W. Verbeke. "European Consumer Preferences for Beef with Nutrition and Health Claims: A Multi-Country Investigation Using Discrete Choice Experiments." Food Policy, 44, 2014, 167-76.
Vermeir, I., and W. Verbeke. "Sustainable Food Consumption: Exploring the Consumer 'Attitude-Behavioral Intention' Gap." Journal of Agricultural and Environmental Ethics, 19(2), 2006, 169-94.
Zhao, J., and C. L. Kling. "A New Explanation for the WTPAVTA Disparity." Economics Letters, 73(3), 2001, 293-300.
--. "Willingness to Pay, Compensating Variation, and the Cost of Commitment." Economic Inquiry, 42(3), 2004, 503-17.
CLAUDIA BAZZANI, VINCENZINA CAPUTO, RODOLFO M. NAYGA JR. and MAURIZIO CANAVARI *
* This work was partly supported by the Tyson Chair Endowment at the University of Arkansas and by the National Research Foundation of Korea (NRF-2014S1A3A2044459). The authors want to thank the company "Natura Nuova" for donating part of the material used in the experiment and "Coop Adriatica" for granting permission to perform the survey in their retail outlet. The authors also thank Jayson Lusk for comments and suggestions on an earlier version of the manuscript. Finally, the authors thank Salar Jahedi for his suggestions on the implementation of the experimental procedures. All the responsibility for the content of this paper stays with the authors.
Bazzani: Research Postdoctoral Associate, Department of Agricultural Economics and Agribusiness, University of Arkansas, Fayetteville, AR 72701. Phone 1-479-575-3240, Fax 1-479-575-5306, E-mail firstname.lastname@example.org
Caputo: Assistant Professor, Department of Agricultural, Food, and Resource Economics, Michigan State University, East Lansing, MI 48824. Phoned -517-884-8656, Fax 1-517-884-8656, E-mail email@example.com
Nayga: Professor and Tyson Chair in Food Policy Economics, Department of Agricultural Economics and Agribusiness, University of Arkansas, Fayetteville. AR 72701; Adjunct Professor, Korea University, Norwegian Institute of Bioeconomy Research, and NBER. Phone 1-479-575-2299. Fax 1-479-575-5306, E-mail firstname.lastname@example.org
Canavari: Associate Professor. Department of Agricultural Sciences, Alma Mater Studiorum, University of Bologna, 50-40127, Bologna, Italy. Phone 39-0512096108, Fax 39-0512096105, E-mail email@example.com
(1.) Under the assumption of risk neutrality, the financial benefit in postponing an irreversible and uncertain investment is defined as a quasi-option value (QOV) (Dixit 1992; Dixit and Pindyck 1994).
(2.) This experiment set was limited to the analysis of WTA formation in dynamic settings.
(3.) The actual market price for two cups of apple sauce (100 g each) generally varies between 1.10 [euro] and 2 [euro] depending on the brand, production method (organic or conventional), and type of store. However, in the pretest phase, respondents were asked to indicate their reference price for two cups of apple sauce (100 g each). The range of the suggested prices was much wider than the actual market prices. Therefore, in this experiment, it was decided to use a slightly larger price range than the one defined by the market in order to avoid that respondents may consider the differences in price irrelevant.
(4.) An example of a choice task presented to respondents is reported in the Appendix.
(5.) Participants were randomly assigned to the different treatments in order to also avoid interviewing time effects. Results from the Kruskal-Wallis test show that there is no significant difference in terms of interviewing time across the treatments (results are available from the authors upon request), indicating that interviewing time did not affect WTP variation across the treatments.
(6.) Regarding organic production, we introduced the definition of organic certification according to Council Regulation (EC) No. 834/2007 of 28 June 2007. Because a universal definition of "local" in Italy does riot yet exist, we used existing regional legislative decrees and proposed regulations related to the "local food" issue. Regarding the product, we provided the following information: how to consume the product (as a snack or dessert), how to store it, and that it contained only apples.
(7.) The duration of individuals' grocery shopping was calculated in order to determine whether this factor affected respondents' willingness to return the product. However, only one participant returned the apple sauce, 25 minutes after completing the survey.
(8.) We elicited consumers' familiarity with apple sauce (using a 7-point semantic scale) and consumers' frequency of buying apple sauce in order to elicit respondents' uncertainty degree for the characteristics of the product in question. Results from the Kruskal-Wallis test show that no significant difference exists across the treatments both in terms of respondents' familiarity and frequency in purchasing apple sauce (results are available from the authors upon request), suggesting that these aspects do not significantly influence the difference in consumers' evaluations across the treatments. Hence, we did not include this information in the model.
(9.) Results are available from the authors upon request.
(10.) We thank anonymous reviewers for pointing these out to us.
TABLE 1 Summary of Previous Studies Testing the Commitment Cost Theory Paper Good in Question Method Corrigan, Kling, and Improved water Hypothetical Zhao (2008): quality of a local referendum format CV "Willingness to Pay lake survey and the Cost of Commitment: An Empirical Specification and Test" Corrigan (2005): "Is University coffee mug nth-price auction the Experimental Auction a Dynamic Market?" Kling, List, and Zhao Sport-cards nth-price auction (2013) "A Dynamic Explanation of the Willingness To Pay and Willingness To Accept Disparity Lusk (2003) "An Lottery ticket and Second-price auction Experimental Test of university coffee mug Commitment Cost Theory" Paper Experimental Design Corrigan, Kling, and Lab or field Field experiment Zhao (2008): experiment "Willingness to Pay Between subjects and the Cost of Between or within Commitment: An subject design Tested Delayed information Empirical aspects of CC theory Specification and in the treatments Test" Corrigan (2005): "Is Lab or field Lab experiment the Experimental experiment Between or Auction a Dynamic within subject design Within subjects Market?" Tested aspects of CC Delayed information theory in the treatments Irreversibility degree of the purchase Kling, List, and Zhao Lab or field Field experiment (2013) "A Dynamic experiment Between or Explanation of the within subject design Between subjects Willingness To Pay and Willingness To Tested aspects of CC Delayed information Accept Disparity theory in the treatments Irreversibility degree of the purchase Lusk (2003) "An Lab or field Lab experiment Experimental Test of experiment Between or Commitment Cost within subject design Between subjects Theory" Tested aspects of CC Uncertainty degree theory in the treatments Delayed information irreversibility degree of the purchase TABLE 2 Final Choice Set Used in the Study Apple Sauce A (a) Price Method of Origin of Production Production Choice task (b) 1 1.45 Organic Nonlocal 2 1.95 Nonorganic Local 3 0.95 Organic Nonlocal 4 1.45 Organic Local 5 0.95 Nonorganic Nonlocal 6 2.45 Organic Local 7 2.45 Nonorganic Local 8 1.95 Nonorganic Nonlocal Apple Sauce B Price Method of Origin of Production Production Choice task (b) 1 0.95 [euro] Nonorganic Local 2 1.45 [euro] Organic Nonlocal 3 1.95 [euro] Nonorganic Local 4 1.45 [euro] Nonorganic Nonlocal 5 1.95 [euro] Organic Local 6 0.95 [euro] Nonorganic Nonlocal 7 2.45 [euro] Organic Nonlocal 8 2.45 [euro] Organic Local (a) In each choice task, the respondent is also presented with a "none of these" option. (b) The order of choice tasks is randomized. TABLE 3 Layout of the RCE Treatments CT IINT DINT RT Blind tasting [check] [check] [check] [check] Visual examination [check] [check] [check] [check] Information RCE mechanism [check] [check] [check] [check] Neutral information [check] Information given about the [check] product and about organic and local production after grocery shopping Possibility of returning the [check] product RCE questions [check] [check] [check] [check] TABLE 4 Sociodemographic Characteristics of the Sample CT IINT INDT RT TOT Gender Female 55% 64% 64% 59% 38% Male 45% 36% 36% 41% 62% Fisher's exact test p value = .638 Age 18-34 20% 20% 27% 11% 19% 35-49 20% 27% 21% 25% 23% 50-64 32.5% 39% 34% 39% 36% >64 27.5% 14% 18% 25% 22% Fisher's exact test p value = .343 Education <High school 29% 16% 17% 23% 23% High school 31% 50% 43% 34% 38% Bachelor's degree 32.5% 23% 31% 37.5% 31% >Bachelor's degree 7.5% 11% 9% 5% 8% Fisher's exact test p value = .442 Income <15,000 [euro] 23% 22% 11% 14% 19% 15,000 [euro]-29,999 [euro] 42% 38% 41% 22% 37% 30,000 [euro]-44,999 [euro] 23% 24% 27% 47% 30% 45,000 [euro]-59,999 [euro] 5% 12% 14% 8% 9% >60,000 [euro] 7% 2% 7% 8% 6% Fisher's exact test p value = .126 TABLE 5 Percentage Choosing "Buy" and "No-Buy" Option in Each Treatment CT IINT INDT RT Buy option 64% 66.5% 61.4% 71% No-buy option 36% 33.5% 38.6% 29% Pearson's chi-square (3); 10.1132 p value: .018 TABLE 6 Hypothesis Test of Equality across Treatments Choices MNL All treatments 1984 -2596.33 CT 640 -632.08 INT 448 -412.08 1NDT 448 -409.74 RT 448 -428.03 [H.sub.0] = Test of equality 1428.80 *** across treatments (a) RPL RPL-EC All treatments -1759.66 -1701.30 CT -582.60 -562.55 INT -374.56 -365.38 1NDT -386.53 -374.72 RT -401.49 -384.20 [H.sub.0] = Test of equality 28.96 *** 28.90 *** across treatments (a) (a) Likelihood ratio test *** Significant at 1%. TABLE 7 Marginal WTP ([euro]/two cups 100 g each of apple sauce) across Treatments and Hypothesis Tests Hypothesis Tests (Poe Test) Organic Local [H.sub.01] : ([WTP.sup.IIN] - [WTP.sup.CT]) [less than or equal to] 0 WTP.sup.IIN] 1.06 0.70 [WTP.sup.CT] 0.80 0.53 p value 0.08 0.20 [H.sub.02] : ([WTP.sup.DIN] - [WTP.sup.CT]) [greater than or equal to] 0 WTP.sup.IND] 0.92 0.42 [WTP.sup.CT] 0.80 0.53 p value 0.76 0.28 [H.sub.03] : ([WTP.sup.RT] - [WTP.sup.CT]) [less than or equal to] 0 WTP.sup.RT] 1.13 0.96 [WTP.sup.CT] 0.80 0.53 p value 0.06 0.03 TABLE 8 Robustness Tests in WTP Space ([euro]/two cups 100 g each of apple sauce) Hypothesis Tests Coefficient Standard p Value Error [H.sub.01] : ([WTP.sup.IIN] - [WTP.sup.CT]) [less than or equal to] 0 Organic x dtreatment 0.69 ** 0.34 0.04 Local x dtreatment 0.19 0.33 0.56 [H.sub.02] : ([WTP.sup.DIN] - [WTP.sup.CT]) [greater than or equal to] 0 Organic x dtreatment 0.36 0.32 0.40 Local x dtreatment -0.27 0.30 0.23 [H.sub.03] : ([WTP.sup.RT] - [WTP.sup.CT]) [less than or equal to] 0 Organic x dtreatment 0.62 ** 0.31 0.05 Local x dtreatment 0.63 ** 0.31 0.04 ** Significant at 5%.
|Printer friendly Cite/link Email Feedback|
|Author:||Bazzani, Claudia; Caputo, Vincenzina; Nayga, Rodolfo M., Jr.; Canavari, Maurizio|
|Article Type:||Author abstract|
|Date:||Jan 1, 2017|
|Previous Article:||Trust, trustworthiness, and business success: lab and field findings from entrepreneurs.|
|Next Article:||Category reporting in charitable giving: an experimental analysis.|