Experimental Design Details
More precisely, each subject is randomized into one of the following three groups:
1) Perception of wealth. This is subdivided into two sub-arms. In the main arm (chosen with a probability of 2/3), the beliefs and information are expressed in a scale that is easy/intuitive to grasp, by expressing income per hour. In the secondary arm, we use the raw scale (yearly income, in billions). Within each of these arms, we randomize half of the subject to receive information.
2) Perception of tax rates. This arm is divided in three groups, assigned with equal probability. One group received information on the average tax rate paid by billionaires. Another group receives additional information on the loopholes used by billionaires. The last group receives no information.
3) Public Image. This is subdivided into two sub-arms, with equal probability: one about the spending habits (i.e., luxury lifestyle) of the billionaires, and another one about the reasons why they became rich. Within each of these arms, we randomize half of the subject to receive information.
There are two sub-treatments that deserve a special explanation:
- In the treatment arm on perceived wealth, we randomize whether we use a familiar vs. unfamiliar scale (e.g., Elon Musk increasing his wealth by $1,847,000 per hour versus $16.18 billion per year). The hypothesis is that individuals cannot wrap their heads around the concept of a billion, so they will react to the information more when it is presented in a scale that is easier to grasp.
- In the treatment arm on tax rates, we randomize whether individuals are only informed about the effective average tax rate paid by billionaires or whether they receive additional information on the loopholes that they use to reduce their tax burden. The hypothesis is that teaching individuals that billionaires pay a lower tax rate may backfire, by reducing the reference point and thus making it harder to demand higher taxes. On the other hand, our hypothesis is that teaching individuals about the loopholes will increase the desired tax rates, because it will increase the perceived unfairness of the current tax rates.
It is important to note that while the average treatment effects of each piece of information are useful, they are not the only object of interest. Due to belief updating, we expect effects to be largely heterogeneous -- for instance, the information on tax rates may cause some individuals to update their perceived tax rate upwards and may cause other individuals to update downward. Thus, we will measure the treatment heterogeneity by leveraging the data on prior beliefs. Moreover, to better grasp the magnitude of the effects, we will estimate the causal effects of beliefs using the same two-stages least square model from Cullen & Perez-Truglia (2022).
At the end of the survey, we collect some basic socio-demographic information about the participants, such as their gender, income group and political orientation. This data will allow us to: (i) Provide descriptive statistics about the sample; ii) control for background characteristics in regression analyses; (iii) re-weight the observations to make it more representative of the U.S. population; (iv) explore the heterogeneity of the effects, particularly between left-wing and right-wing respondents. We also elicit people's attitudes towards redistribution in general and their trust in the government.
References
Cullen, Z. and Perez-Truglia, R. (2022). How Much Does Your Boss Make? The Effects of Salary Comparisons. Journal of Political Economy, Vol. 130 (3), pp. 766–822.