Natural Disasters Perception and Housing Insurance Choices

The treatment conditions we will implement are as follows:
1. Condition I: Control group.
2. Condition II: Information treatment.
Participants will receive objective and accurate information about the probability of natural disasters occurring in Italy.

2024-11-27

2024-12-22

1. Estimation of beliefs regarding natural disaster risks in Italy.
2. Estimation of beliefs regarding a specific cause of death in Italy (placebo).

The estimation of beliefs about natural disaster risks in Italy will allow us to assess how Italian households perceive these risks. Furthermore, both within-group analysis (prior vs. posterior beliefs) and between-group analysis (treatment vs. control) will enable a precise estimation of the impact of the information provision in correcting these beliefs.

As an additional control analysis, we will examine the effects of the information provision on another outcome variable. This will help verify that the intervention operates through a specific channel while ruling out alternative explanations, such as general pessimism induced by the information provided.

These questions will be monetarily incentivized.

Other relevant outcome variables will evaluate the effects of the information provision in the following dimensions:
1. Beliefs about individual risks of natural disasters.
2. Confidence in the estimation of natural disaster risks in Italy.
3. Information-seeking behavior induced by the treatment.
4. Propensity to purchase insurance or modify existing coverage.

The overarching purpose of these outcome measures is to evaluate how objective information about these risks influences individuals' beliefs and behaviors in this context.

Importantly, the different outcomes will be assess both in aggregate and following teterogenous treatment effects that will include socio-demographic characteristics and financial and insurance litereacy.

At the outset of the experiment, participants will be asked to estimate probabilities related to the risks associated with natural disasters in Italy. Following this, using a between-subjects design, some participants will receive objective and accurate information provided by experts regarding these risks. We will then evaluate how the provided information shapes participants' beliefs—primarily their risk estimations—and subsequent behaviors, including information-seeking actions and their propensity to purchase or modify existing home insurance plans. Finally, at the end of the experiment, participants will complete demographic questionnaires as well as questions to assess their financial and insurance literacy.

At the outset of the experiment, participants will be asked to estimate probabilities related to the risks associated with natural disasters in Italy. Specifically, they will assess the risks of natural disasters and the potential damage to Italian households, both at the national and individual levels.

In a between-subject design, half of the participants (the treatment group) will receive objective information about these risks as estimated by experts (e.g., ISPRA). For participants in the treatment group, they will then be asked to estimate the natural disaster risks again, using the same framework as above. Additionally, all participants will be asked to state how confident they are in the estimations they provide.

As a placebo, both the treatment and control groups will also be asked to estimate the risk of dying from a specific health condition. This placebo question will help more carefully assess the impact of the information treatment on the specific risks we are focusing on, as well as any potential spillover effects induced by other mechanisms.

Following this, participants will be asked a set of questions to assess whether they are interested in seeking further information related to natural disaster risks and to what extent they believe informing themselves about such risks is appropriate. Subsequently, they will be asked whether they intend to maintain or change their current home insurance plans.

Finally, at the end of the experiment, participants will complete a socio-demographic questionnaire, along with questions to assess their financial and insurance literacy. Additional questions will elicit participants' risk attitudes and other factors relevant to the specific research question (e.g., place of residence).

Randomization done in office by a computer.

Individual.

No

No clusters.

300 individuals, equally split between the treatment and control groups.

150 per treatment arm.

At the conventional level for type I error (alpha = 0.05), we have more than 0.8 power, with our sample, to detect a Cohen's effect size of d=0.33. This calculation is based on explorative survey estimates we obtained for our main outcome variable. We assumed a mean estimate of 192 in the control group, with a s.d. equal to 194, in line with explorative estimates we obtained in a previous exploratory survey. We assumed the same s.d. for the treatment group. Given the information treatment should move the estimate to 800, if fully incorporated, we believe that we have power to even detect a small effect, corresponding to less than perfect updating to the information we provide. A Cohen's effect size of d=0.33 would correspond to a mean estimate of 255 in the treatment group, with these assumptions.