Schedule - Parallel Session 5 - Field Experiments 2

WMG IDL 1st Floor Syndicate Room - 14:00 - 15:30

What Determines Contributions to a Real-Life Public Good?

Ronen Bar-El


We examine the determinants of contributions to a real-life public good. We conduct five rounds of contribution experiments for the procurement of sustainable supplies to two colleges’ synagogues, a religious college and a secular college. The choice of the campus synagogue as the real-life public good has two advantages. First, it is located within the campus and may be voluntarily used by the students. Second, like many religious institutions around the world, synagogues raise funds from their users (unlike other types of public goods, such as roads and schools). Thus, it is likely that the participants of the experiment had a prior position as to what extent they should contribute (or not) to synagogues. Nevertheless, it is important to note that there is no religious obligation to donate to synagogues. In addition, being part of the campus, the campus synagogues are financed by the academic institution, so the students had not contributed to their campus synagogue prior to the experiment. The results show that the intensity of use at the campus synagogue (measured by the reported number of monthly visits at the campus synagogue) affects the total contributions as well as their path. Accordingly, we find that the contribution path of the religious subjects (who make frequent use of their campus synagogue) is upward sloping, while that of the non-religious subjects (who rarely use their campus synagogue) is lower and decreases toward the end of the experiment. The upward slope of the religious’ contribution path holds also under terms of lack of information about the others’ contributions, although it is lower that the path under terms of information about the accumulated sum of contributions after each round. The non-religious group of subjects consisted of individuals who rarely use their campus synagogue and individuals who reported that they do not use their campus synagogue. The latter’s contribution cannot be considered a contribution to a public good. We separately analyzed the contributions of the group of non-religious subjects who reported zero visits at their campus synagogue. The results show that the latter group’s average total contribution and the contribution path are lower than those of former. We examined whether the religious population is more inclined to cooperate and to contribute compared with the non-religious population, by conducting a standard linear public good experiment among religious students. The results of the experiment show that the average contributions decline over rounds, that is, free-riding increases over rounds, and the end-game effect is present. We also found that the contribution level in the standard VCM is not related to the number of visits at the campus synagogue. In addition, we find in all the experiments that religious females consistently contribute more than religious males. Overall, our findings may indicate that contributions are motivated by contextualization, by the effect of asking and by peer effect. Finally, we derive practical implications from our findings.

Ronen Bar-El

Fellow Researcher, The Open University of Israel

Focal Foods in Schools: Identifying Pluralistic Ignorance with Behavioural Games

Nicholas Bardsley; Nicholas Bardsley; Rachel McCloy; Simone Pfuderer; Aljaz Ule


The quality of schoolchildren’s food intake is a matter of concern even in affluent societies. Evidence suggests that children misperceive the dietary choices of their peers, believing that their intake of healthy foods is lower than it really is (Lally et al. 2011). But it is less clear what role social norms play in guiding their perceptions or choices. We sample 54 children aged 14-15 at a comprehensive secondary school in the North of England, using behavioural games to identify norms. We compare norms to individual preferences, elicited using a questionnaire, and beliefs about others’ preferences, elicited using a guessing game. A stated preference survey using Likert scales was administered under three conditions. In the “coordinate” condition, pupils have to state what the most common response will be in circumstances that all other pupils are attempting to identify the most common response. This constitutes a coordination game and has been used by researchers in numerous contexts to identify social norms. In the “answer” condition, pupils simply state their own preferences for various food items using rating scales. In the “guess” condition pupils have to guess what the most common response was in the “answer” condition. Two protocols were used. In the first (N=30) pupils responded under “answer” followed by “guess” conditions. The second (N=24) used “coordinate” followed by “answer”. Responses in the guess and coordinate conditions were incentivised using £10 music store vouchers for questionnaires with the most correct responses. We find strong evidence of misperceptions of preferences, with children systematically under-rating attractiveness to others of healthy items and over-rating that of unhealthy items (p<.01; signed rank tests). The bias is generally consistent with a perceived influence of social norms: norms are against healthier items and in favour of less healthy items, often contrasting with modal preferences (p<.05; signed rank tests). Norms coincide with beliefs about others’ preferences in direction (positive versus negative attitudes) but not intensity (p<.05; rank sum tests) A tendency to overestimate the extent to which others have internalised a norm would explain our results but also implies ‘pluralistic ignorance’, whereby individuals privately reject a norm but believe that others mostly accept it. Consistently with this interpretation, very few pupils admit to ‘fitting in’ when choosing food, but nearly 50% predict fitting in responses by others. The overall pattern of results suggests potential for shifting pupils’ norms towards healthier options, by making public the distributions of their stated preferences.

Nicholas Bardsley

Associate Professor, University of Reading

The Misery of Spending Down the Nest Egg

Yu Gao; George Loewenstein; Peter Wakker; Xianghong Wang


Empirical research (Panis, 2004) on happiness and retirement found that retirees with higher annuitization level were more satisfied with their lives, and maintained their satisfaction throughout retirement; retirees without annuities became less satisfied over the years. However, even after controlling for income, wealth, health, gender, marital status, age, etc., it is still possible that people with and without annuities belong to completely different groups. Do people behave differently, and feel differently when spending down from flow vs. lump-sum? To identify the effect of flow payment, we conducted a field experiment where we assigns subjects randomly into these two payment schemes. The main experiment was conducted at Renmin University in Beijing, China in April, 2015. The experiment consists of two stages. The first stage was a pre-survey to recruit subjects and obtain relevant information. In the pre-survey, we measured respondents’ “spendthrift-tightwad (ST-TW)” scale which is a measure of individual differences in the pain of paying (Rick, Cryder, Lowenstein, 2008), and asked about their consumption behaviors (frequency of eating at campus, monthly allowance, funding resource and frequency, financial satisfaction). The second stage was the main part, which involves two different payment schemes and daily surveys that last for 10 days. In the Flow treatment, subjects received CNY20 (≈USD3.09) in their campus card each day for 10 days consecutively (in total CNY200), whereas in the Lump-sum treatment, subjects received CNY200 (≈USD30.9) in their campus card at the first day of the experiment. Spending behavior is recorded by students’ campus card spending history provided by the campus card management center. Feeling is measured by self-reported happiness and money anxiety in the daily survey. We obtained spending records for 30 consecutive days: 10 days before the experiment (from 28th March to 6th April), 10 days during the experiment (from 7th April to 16th April), and 10 days after the experiment (from 17th April to 26th April). The spending records consist of location, time (precision up to a second) and amount for each transaction. Coincide with our conjecture, we found that flow treatment makes people spend more, and feel less anxious about money. The treatment effect is interacted by people’s ST-TW scale: the treatment effect decreases if the person has a higher ST-TW score (a person with higher ST-TW score means that it is difficult for her to control spending).

Linking Experimental and Survey Data for a Uk Representative Sample: Risk and Time Preferences

Matteo M Galizzi; Glenn W Harrison; Raffaele Miniaci


We report evidence from the first “artefactual field experiment” that directly incorporates experimental measures for risk and time preferences for a representative sample of respondents within the Innovation Panel (IP) of the UK Household Longitudinal Survey (UKHLS, also known as Understanding Society), the world-largest multi-scope panel survey. In wave IP6 we randomly allocated a representative subsample of 708 IP respondents to an experimental module where discounting rates and a-temporal risk preferences were elicited using incentive-compatible methods. Households were independently randomly allocated to either a face-to-face or a web interview mode. Individual respondents had the opportunity to self-select into their preferred interview mode, and to opt out of the experimental risk and time preferences questions. Time preferences were measured with 72 questions asking to choose between a Smaller-Sooner or a Larger-Later monetary reward. Risk preferences were measured with 18 binary-lotteries questions with low and high monetary stakes, and an extra question using a multiple-lotteries task. One question was randomly selected and subjects paid real money at the due date based on their preferred choice in that question. Respondents also answered survey questions on risk and time attitudes. We structurally jointly estimate risk preferences under Expected Utility Theory and Rank Dependent Utility, and time preferences considering a broad class of discounting models such as exponential, hyperbolic, quasi-hyperbolic, and Weibull discounting. The models, and their finite mixtures, are estimated using Maximum Likelihood calculating individual-specific levels of daily ‘background consumption’ from linked household income data. The structural estimations model the self-selection into the interview mode and into the experimental module, and use sampling weights to adjust for differential attrition and non-response. We have three findings. First, there is high heterogeneity in the responses from our UK representative sample. Overall responses seemingly support risk aversion and non-constant discounting. The finite mixtures models suggest that about half the responses are in line with an implicit exponential discount rate of around 17 percent a year, while the remaining responses could be explained by non-constant, in particular Weibull, discounting. Second, because of the self-selection into the interview mode, web respondents are younger, more educated, and less likely to report poor health. Despite self-selection, there is no significant difference between the web and the face-to-face responses and the structurally estimated parameters of risk aversion and time discounting. Third, the different risk and time preferences measures are systematically cross-validated with a comprehensive range of linked survey, administrative, and biomarkers data in the UKHLS, providing mixed evidence on their external validity.

Matteo M Galizzi

Assistant Professor of Behavioural Science, ESRC Future Research Leader Fellow, London School of Economics