Experimental Design
Design overview
Participants will be recruited from a panel of Mexican voting-age adults. About 5,600 adults will be invited to enter our online system at a specific date and time. Those who show up (about half are expected to show up) will be asked to fill a baseline survey, on the basis of which they will be assigned to one of two respondent pools pairs by an algorithm: the paired pool, and the leftover pool.
Paired pool
All pairs will consist of one morenista (i.e., pro-incumbent-party) and one non-morenista. Partisanship will be determined by asking respondents who they would vote for if a presidential election were held today (we will also collect standard measures of affective polarization and partisan preferences, including feeling thermometers party ID). Pairs will be created according to the following blocking procedure. First, we will divide the respondents into two sets (morenistas and non-morenistas) using the aforementioned question. Within each of these two sets, we will create groups of 4 similar persons, on the basis of Mahalanobis distance computed on the basis of various other questions in the baseline (including partisan feeling thermometers usually employed to measure affective polarization, as well as a self-reported measure of socioeconomic status). To create pairs, we then draw one group of 4 similar persons at random from each of the morenista and non-morenista sets. We randomly pair these up into four pairs of 1 morenista and 1 non-morenista. These four pairs constitute a block. Which each block, pairs will be randomly assigned to three treatment branches (about 350 pairs each) or a control condition (about 700 individuals). We expect block random assignment to substantially increase statistical power and help ensure balance on blocking covariates across treatment conditions.
Leftover pool
Leftovers are those individuals who could not be paired up. The reason leftovers are likely to exist is because the fraction of morenistas and non-morenistas in the sample is not something we can control. Therefore, the maximum number of pairs is the number of individuals in the smallest of the two sets. Our blocking procedure may further imply that one incomplete block will exist, in which case all individuals in that block will be reassigned to the leftover pool.
Procedure for the paired pool
After completing the baseline survey, pairs will be directed in two different directions according to their treatment assignment status. Those who get paired up and assigned to one of three treatment branches (but not those pairs assigned to the control condition), will be directed to chatrooms, where they will be instructed to interact in two activities: (a) discuss and rank two life priorities in terms of their importance for Mexicans in general, and (b) answer a set of trivia questions about current events, where those pairs who answer at least two questions correctly will be entered in a raffle. We expect these two activities will allow them to meet and discuss important things in life. We designed the questions so as to create a high degree of agreement in the ranking about what matters in life among the multiple choice answers. This activity will make commonalties in important values salient. On the other hand, we conjecture that the trivia questions will allow them to see themselves more as a collaborating team. We will evaluate the effect of these activities directly with survey questions.
Those in the paired pool who get assigned to the control condition will be asked to complete the same two tasks (ranking of life priorities and answering trivia questions), but they will do so individually.
Procedure for the leftover pool
These participants will be assigned to a secondary set of treatments, consisting of exposure to a video designed to elicit pride in being Mexican or to a placebo (a video about dolphins). The idea is that emphasizing a common national identity might decrease the salience of particularistic differences between subgroups of Mexicans, and therefore reduce affective polarization, increase identification with the nation and other people from the same country, and increase trust in others---especially fellow country people. We may or may not use this analysis as part of a separate paper, but we have not made a firm decision on this point yet.
Post-treatment procedure for both pools
All participants will be asked to fill out a short endline survey individually. Finally, a short follow-up survey will be fielded between two and six weeks later (staggering will create exogenous variation in time between treatment and follow-up, making it possible to estimate a treatment persistence curve).
Treatment branches and research design for the paired pool
Pairs in the paired pool will be randomly assigned to one of four conditions: N (control), E (equal), U (unequal), or ES (equal + socioeconomic status). Those in the control group will experience no interpersonal contact. All others will.
All individuals in the paired pool are told the following: (a) their ranking of life priorities for Mexicans will be used in a brochure to teach university students about Mexicans, and (b) those who respond to two or more trivia questions correctly will be entered into a raffle for 500 panelist points (an attractive amount---the full 25 minute activity including baseline, chat, and endline nets panelists 100 points). Answers are filled out individually in all cases, but pairs (i.e., those assigned to E, U, or ES) are encouraged to deliberate and cooperate with their partners via the chat.
For pair members in the E condition, we emphasize that the answers to activities (a) and (b) of both pair members will count equally for purposes of the brochure and the raffle. For pair members in the U condition, we randomly design one of the pair members as the Leader and the other as the Follower, and notify both pair members of these designations. We further explain that only the Leader’s answers will count for both the brochure and the raffle. In this case, the follower is encouraged to chat in order to persuade the leader, and the leader is encouraged to listen to the follower. Pair members in the SE condition are treated like those in the E condition, with the difference that they are initially provided information about the self-selected SES (socioeconomic status) of their chat partner. Importantly, this information is in pictorial form, as one of the first few questions in the baseline asks respondents to pick among 5 sets of pictures (where each set contains a photograph of a house façade, a bedroom, and a kitchen) which one most resembles their own living conditions. Pairs assigned to the control condition (N) do not chat, but instead complete tasks (a) and (b) individually. They are also entered into the raffle if they get at least two trivia questions right.
Treatment condition Status in the chat interaction Revealed information
N (700 individuals) No contact NA
U (350 pairs) Unequal Partisanship
E (350 pairs) Equal Partisanship
ES (350 pairs) Equal Partisanship+SES
Treatment branches and research design for the leftover pool
Individuals in the leftover pool are not paired up. Instead, they are randomly assigned to watch one of two videos: a video about things mexicans share or video about dolphins, as mentioned previously, between the baseline and the endline. The video is followed by a handful of attention/learning questions and respondents are then exposed to the endline.
Hypothesis
H1: Collaborative contact under equality in the interaction (E > N) increases
• Preference for democracy
• Pro-social, pro-democracy behavior as in declared willingness to be a poll worker
• Willingness to participate in future meetings, and willingness to participate in mix-partisanship meetings
• Willingness to donate to the aforementioned NGO
• Willingess to donate in dictator games to the fellow party sympathizers, vis a vis people who sympathize with a different party.
• Generalized trust and trust in fellow country people
H2: Collaborative contact under equality in the interaction increases the variables in H1 more than under inequality in the interaction (E > U)
H3: Collaborative contact has asymmetrical effects on the variables in H1 under equality in the interaction, when real-world socioeconomic inequality in SES is revealed to the parties:
H3.1: For the high SES individual, ES > E
H3.2: For the low SES individual, ES < E
Membership in overarching group and polarization
H4: Exposure to the video highlighting pride in common belonging (vis-à-vis the placebo video) increases the variables in H1. It also increases the sense of common identity or belonging.
These are the key hypothesis that we wish to preregister. We will additionally test a battery of secondary hypotheses.