Intervention (Hidden)
Recruitment occurs in three waves. To test for statistical power, we ran a pilot wave(n=100) in May 2019. Then, in July 2019 we will run the second wave (n=1,000). In this wave, we randomize treatment and estimate effects. Finally, conditional on funding, we will run a third wave in September 2019 intended to detect smaller behavioral changes that individuals may report. In our main estimates, we will pool all three waves and include a wave fixed effect in order to control for any wave-specific differences. We recruit from two different populations. First, we use Amazon’s Mechanical Turk population (MTurk) in order to crowd-source responses from an active and online community(waves 1 and 2). This also gives us a representative sample from across the United States during the Summer of 2019. Second, for wave 3 we plan to recruit from an alternative population to extend the external validity of the results.
Upon starting the survey, participants are asked a variety of demographic questions. This includes gathering information on gender, age, political orientation, and political preferences.Following this, participants are sequentially asked two things for two different types of government spending. First, for a given $100, how much they would prefer to have allocated between the two categories. This provides a measure of an individual’s preferred spending allocation,Pi. Second, how much they believe is actually being allocated to each. This gives a measure of an individual’s expectations, Ei, capturing what how think the government is actually distributing between the two categories. The difference between Pi and Ei provides an initial measure of polarization. In practice, the two categories that a participants chooses between are “Welfare and Government Assistance Programs” and “Military, Defense, and Homeland Security”. These categories are chosen due to the fact that they draw polar views despite being funded similarly (Oldendick and Hendren, 2018).After this, we randomly assign individuals into either a treatment or control group. The control group is shown their difference between Pi and Ei and is then simply required to finish answering outcome questions. The treatment group is similarly shown their difference between Pi and Ei but is also assigned an information intervention described below.
Our experimental design induces either an increase or a decrease in an individual’s degree of polarization. The directional change is dependent on the individual’s initial beliefs and expectations regarding the allocation of government spending. We randomize an information intervention that reveals to treated participants the real allocation, R, between the two categories. By doing so, depending on an individual’s initial Pi and Ei, participants are treated with either a reduction in polarization, an increase in polarization, or no change in polarization.