International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
Around and beyond the cheap talk script in Choice Experiments
Roberta Raffaelli, Sandra Notaro

Last modified: 11 May 2015


Finding truth telling mechanisms to reduce hypothetical and social desirability biases remains an issue in environmental valuation studies. The ‘cheap talk' script has been extensively used in stated preference studies since its introduction (Cummings and Taylor, 1999) but the evidence about its effectiveness in reducing hypothetical bias is still mixed. From a psychological and behavioral point of view this result is not surprising. As highlighted by Carson and Groves (2011), in the hard version of the cheap talk the researcher tells respondents that some other respondents lie when they answer survey and "it is  not  clear how a respondent should interpret such a statement that clashes with the usual social norm of truth telling that survey researchers try to advance in surveys."

Inspired by the experimental psychological literature, we identified two scripts to be tested as potential truth telling mechanism. The first one (honesty script - HS) was inspired by the natural experiment carried out by Pruckner and Sausgruber (2013) who find that the simply script "thanks for being honest" was able to increase the honesty of people purchasing newspaper on the street. The second one was inspired by the experimental results of Gneezy (2005) on the role of consequences on people deception. In our script we emphasized the consequences of respondent's choices on social welfare (emphasis on consequences script - EcS). Given the recent interest for different forms of solemn oath mainly employed in lab experiments (Jacquemet et al. 2013; Stevens, Tabatabaei, Lass 2013; Carlsson et al. 2013; Sue et. al 2012) we add a third treatment to test the effects of a solemn promise proposed to be signed just before the choice task (solemn promise - Sp). We designed a between-sample choice experiment with four treatments to test the effects of these two scripts and the solemn promise against a control treatment (CT).

Moreover, following the stream of studies demonstrating the potentialities of the indirect questioning (or inferred valuation method) in reducing social desirability  - not only for private goods (Lusk and Norwood 2009a) but also for public goods (Lusk and Norwood 2009b; Yadav, van Rensburg, Kelley, 2012) - each respondent was presented with both the direct and the indirect format of the elicitation question.

The empirical study concerns the management of sheep grazing in a high natural value farming area in Italy (Regional Park of Porto Conte, Sardinia). The considered attributes are the landscape value of sheep grazing, the effects of sheep management on biodiversity and a cultural-heritage attribute (possibility to directly experience life and traditions of shepherds). The field CE was carried out in August 2014 and 390 tourists were interviewed through a CAPI system. The respondents were presented with six cards in which they had to make their choice and the same six cards in which they were asked to indicate how other tourists will choose among the three alternatives. The order of presentation of direct and indirect questions was randomized.

First estimates (MNL and RPL models) suggest that the simple honesty script (HS) is unable to reduce the WTPs in our empirical context. On the contrary, the script emphasizing the consequences (EcS) is able non only to reduce the WTPs with respect to the control treatment but also to produce WTPs that are very similar to the WTPs obtained by asking respondents to sign a solemn promise (Sp).  Moreover, from the latter two treatments (EcS and Sp) the same clear change emerges in the relative importance of the attributes. While in the control treatment the differences among  WTPs for the attributes are not too large,  in the EcS and Sp treatments landscape emerges as the most important  and the cultural attribute loses its importance. This seems to reveal the ability of these treatments to lead respondents to pay more attention to the trade-offs among the attributes. Future elaborations will also include stated attribute non-attendance  and investigate the effect of treatments on the precision of choices, exploiting also the information about  the time elapsing between responses to each question. Finally, the comparison between WTPs based on direct and indirect questions shows that the inferred valuation produces smaller WTPs for the public attributes but not for the private one.

These preliminary results should be appealing for researchers and practitioners involved in nonmarket valuation of environmental goods. Our "emphasis on consequences" treatment is able to reduce hypothetical bias as well as the solemn promise, but from a practical point of view, it is much easier to implement, reducing the stress for the interviewer and interviewees. The indirect questioning, asking to share others' responses among the alternatives is instead quite demanding for the respondents who require to be strongly motivated in this task.  In our empirical investigation, estimated WTPs from indirect questions can be considered a lower bound for the WTPs for the public attributes but not for the private one.  More in general, our paper suggest that drawing on experimental psychology literature can be very fruitful for defining new scripts aiming at reducing biases in stated preference techniques when monetary incentives cannot be used. More precise estimates could be useful for more informed cost-benefit analysis and decision making in the interest of the whole society.

Conference registration is required in order to view papers.