Experimental Design
Participants will be randomized into two groups and will be asked three sets of forecast questions. The first set of questions will differ across the two groups; the remaining two sets of questions will be the same across the two groups.
The first set of questions asks participants to forecast the extent to which subjects in a previous experiment agreed with a sensitive statement about illegal immigration, as a function of the subjects' political affiliations. In other words, participants in the current experiment will be asked to forecast the behavior of subjects in the previous experiment who self-identified as democrats and of subjects who self-identified as independents of republicans. Subjects in the previous experiment were randomized into one of two treatments: a private treatment, in which they received an assurance that their answers would remain anonymous, and a public treatment, in which they were given clues suggesting that their answers would be shared with other students at their university together with their names. The results of the previous experiment show that subjects assigned to the public treatment tend to skew their answers to sensitive questions in the direction that is perceived to be more socially desirable on campus. Depending on the group that participants in the current experiment are assigned to, they will be asked to first forecast the answers of subjects in the private treatment of the previous experiment and then those of subjects in the public treatment of the previous experiment, or vice-versa.
The point of this first set of questions is to study whether the forecasts of participants in the current experiment are systematically different depending on whether they are forecasting the behavior of subjects in the private vs. the public treatment of the previous experiment and depending on whether they are forecasting the behavior of subjects who self-identify as democrats vs. as independents/republicans. Furthermore, by randomizing the order in which subjects are asked the first set of forecast questions, I can study how the forecasts change as social image is made more salient. Specifically, when making their first forecast, participants in the current experiment are not aware that the previous experiment had two treatments, so social image is arguably not very salient; conversely, when making their second forecast, participants in the current experiment are aware that the previous experiment had two treatments, so social image is arguably more salient. I will also study meaningful information transmission as defined in my paper "Biased decoding and the foundations of communication" (https://drive.google.com/file/d/1hmwUv-WSeNAR3XCf637t6kBbsXxs--ph/view).
The second set of questions asks participants to forecast the average level of agreement of subjects in the private treatment of the previous experiment given knowledge of the distribution of answers of subjects in the public treatment of the previous experiment. Here I will study whether subjects understand the direction in which the private levels of agreement with the sensitive statements are skewed compared to the public levels of agreement and whether subjects forecast the private-public gap in average level of agreement to be larger for the sensitive than for the placebo statements. I will also compare the subjects' forecasts to the actual results from the previous experiment.
The third set of questions asks participants to forecast the dimensions along which the treatment effects in the previous experiment are heterogeneous. Specifically, they will be asked to forecast whether the treatment effects in the previous experiment were heterogeneous by gender, race, political affiliation, year in school (upper vs. lower), and major (humanities/social sciences vs. sciences/engineering). Here I will study whether participant correctly forecast the dimensions along which treatment effects in the previous experiment were heterogeneous.
I will also study whether the students' forecasts themselves are heterogeneous along the margins from the previous paragraph.