Voters' Demand for Price Caps

We study citizens' support for price caps in the laboratory. Participants in our lab experiments will be exposed to alternative market regimes and information regarding their effects, such that we can understand citizens' support for price caps. We will examine several factors that could drive such support. First, we will test whether limited exposure to price caps increases the support for this policy. This may be the case if citizens fail to anticipate the negative effects of price caps on market outcomes. Second, because, in practice, citizens delegate the choice of policies to politicians, we will test whether citizens' lack of agency increases their support for price caps. Third, we will test whether providing information about the market consequences of price caps reduces citizens’ support. Finally, we shall also study the role of citizens' political attitudes and their attitudes toward market regulation.

2022-11-16

2023-02-28

The primary outcomes of the experiment are the regime choices and beliefs about market outcomes.

Subjects are randomly assigned into one of three treatment arms: Free Market First-Experience (FM-E), Free Market First-Information (FM-I), and Price Cap First-Experience (PC-E). To facilitate logistics, all subjects participating in a given experimental session are assigned to the same treatment arm. In the FM-E and FM-I treatments, both the practice trials and the first block of the Market Task feature a free market. By contrast, in the PC-E treatment, both the practice trials and the first block of the Market Task feature a price cap of 7 ECU.

Once the first block finishes, subjects in the FM-E treatment are informed that a price cap of 7 ECU will be imposed in the second block, whereas subjects in the PC-E treatment are informed that the initially imposed price cap of 7 ECU will be removed in the second block. Subjects in the FM-I treatment do not trade during the second block.

In all three treatments, before the second block begins, subjects choose the regime to be implemented in the third block, namely a free market or a market with a 7,000 price cap. Specifically, one buyer in each group is randomly assigned to choose the regime for the third block in his or her group. We refer to this person as the decisive chooser. Each of the three remaining buyers in the group also indicates which regime he or she would like to implement in the third block, but this choice is hypothetical. We refer to these three people as the hypothetical choosers. Importantly, each subject learns whether he or she is a decisive or hypothetical chooser before choosing the regime for the third block. At this point, subjects do not learn the choices of others.

Before the third block begins, subjects choose the regime to be implemented in the fourth block. In each group, the buyer who was previously assigned to be decisive remains decisive, while the others remain as hypothetical choosers. Buyers are reminded of their role before making a choice. At this point, subjects do not learn the choices of others.

Throughout the session, subjects report a series of beliefs. The details are provided in the Pre-Analysis Plan.

Randomization was done by a computer based on an index of political beliefs.

Individual.

Yes

175 groups of 4 participants.

700 observations.

700 participants.

Using a within-subjects design, we assume a small effect size of 0.2 standard deviations in the primary outcome following Cohen's suggestion. We also assume 80% power and a one-sided test for differences in paired means with a 5% significance level. To account for clustering, we assume a cluster size of 4 and an intra-cluster correlation of 0.5. We adjust for 3 comparisons using the Bonferroni approach. Under these assumptions, the minimum sample size is 240 subjects without accounting for clustering. Assuming a design effect of 2.5, the estimated sample size is 600 subjects. If we use a McNemar test instead for discrete outcomes, the estimated sample size is similar. In both cases, we assume an autocorrelation of 0.5. Performing exact power calculations for non-parametric tests is not feasible because we do not know the shape of the underlying distribution. Therefore, some scholars use simulations to approximate power, making some distributional assumptions. At this stage, we follow a rule of thumb proposed by Lehmann (2006), who suggests adding 15% more subjects to the sample size estimated under a parametric approach. Hence, the estimated sample size is 690 subjects. This sample size is similar to the one used by Dal Bo et al. (2018), the closest antecedent in the empirical literature. In their experiment, they used 768 subjects. Therefore, our power calculations suggest that a sample of close to 700 subjects will be enough for this study. A final piece of our power analysis is the sample requirement for studying the role of pivotality. Some of the proposed tests in this regard require the use of interactions. Power calculations for interactions are typically very demanding. Therefore, we consider this analysis as exploratory.