The role of higher-order risk preferences in belief updating and preventive action: the case of environmental risk

The field experiment consists of an incentivised activity in which participants will be asked to perform a series of tasks and answer a series of questions in order to determine the dependent, independent and control variables to be used in the analysis. The experiment will be conducted on the Pagai Islands (North and South Pagai), located off the western coast of Sumatra, Indonesia. These islands were among the areas most severely affected by the 2010 Mentawai earthquake and tsunami. Data will be collected on site with the help of local enumerators.

The fieldwork is planned to last 3 weeks in April 2026. Researchers will visit local communities and recruit individuals by randomly selecting households in the region to visit. Participants will be invited to take part in the experiment on the spot, with no prior communication. Respondents can decide whether to participate voluntarily and can reject the invitation at any time without penalty. To enable them to make an informed decision, potential participants will receive sufficient information about the study's objectives, topics and the types of data to be collected, as well as the purposes of the analyses, when they are invited to participate. Field enumerators will give respondents who decide to participate a printed copy of the informed consent, including contact information and the data protection regulation.

A total of 15 local communities will be visited, with around 20 interviews conducted in each community. The questionnaire will be administered using tablets. Survey participants were required to be at least 18 years old and to have a basic understanding of numerical concepts. The age threshold was set in reference to legal adulthood markers in Indonesia, where the minimum voting age is 17 and the legal age for marriage is 19, thereby ensuring that respondents possess a minimum level of legal and decision-making capacity.

2026-04-01

2026-04-17

The primary outcomes are the measures of beliefs about environmental risks and the uptake of preventive measures.

We elicit both prior and posterior environmental risk perceptions in order to accurately capture belief updating following information provision. In both cases, we elicit the mean and variance of risk perception to better characterise its distribution. Regarding the uptake of preventive measures, we assess two actions: the intention to donate to the cause and the willingness to voice concerns to local leaders. Both actions are extensively and intensively measured (i.e. uptake or not and the level of effort applied to each action). To minimise experimenter demand effects, respondents reply to these activities anonymously and discreetly.

As a secondary outcome, we elicit an alternative measure of beliefs about environmental risks. In this case, we use an instrument comprising three different belief elicitation questions.

This alternative belief measure will be constructed using the obviously related instrumental variable (ORIV) approach. This method combines different questions and formulations to provide a more efficient estimator, as outlined by Gillen, Snowberg and Yariv (2019). To ensure that the instrument is less prone to measurement errors, we ask three types of question concerning the perceived returns of the assets: qualitative, quantitative and probabilistic (Haaland et al., 2023).

In order to answer our research questions and test our hypotheses, we are planning to conduct a field experiment. This will enable us to elicit risk preferences, perceived environmental risk measures and willingness to take preventive action, as well as to apply information provision treatments. The experiment will comprise six sections, each containing tasks designed to elicit specific individual attitudes or preferences. Three of these tasks are incentive compatible, such that the answers given can influence the monetary compensation for participating in the experiment.

Not available

Randomisation will be performed at the individual level at the start of the questionnaire. This will be done using a computer-based randomisation process calculated by the experiment programming tool.

Individuals.

No

15 grids of 1 km² each.

300 individuals.

At least 20 respondents per grid and at least 100 respondents will be assigned to one of the three study arms: control and two treatments.

Assuming a significance level of 5%, statistical power of 80%, and a conservative ICC of 0.5, the minimum detectable effect size is approximately 0.16 standard deviations.