Emotions and Information Processing

We examine how emotions shape the way people process and react to information about immigration using a 2x2 and a 2x3 design.

2026-05-18

2026-05-31

• Negative emotions index: mean of fear, anger, sadness, disgust, and contempt, each measured on a 1 to 7 scale after the video.
• Negative emotional valence index: net emotional valence constructed from negative emotions relative to joy and surprise.
• Each of the seven emotion items separately: anger, contempt, disgust, fear, joy, sadness, and surprise.
• Posterior beliefs for five immigration-related factual questions (Q1 to Q5).
• Absolute error of posterior beliefs relative to the true statistic revealed in the survey.
• Belief updating for each factual question, defined as posterior belief minus prior belief.
• Understanding of statistics, measured as the closeness of answers to the correct ones
• Interpretation of randomized research evidence on border-wall effectiveness, measured on a 0 to 10 scale.
• Posterior belief about how increased southern-border enforcement affects unauthorized crossings, measured on a 0 to 10 scale.
• Preferred level of government spending on border security, measured on a 0 to 10 scale.

Primary Outcomes
Outcome variables: Emotions
Participants are asked to report the intensity with which they experienced a set of specific emotions while watching the video. For each listed emotion —sadness, surprise, fear, joy, anger, disgust, and contempt—respondents provide a rating on a scale from 1 to 7, where
1 indicates very low intensity and 7 indicates very high intensity. We create two measures of emotions. The first is constructed as the average of all the negative emotions (fear, anger, sadness, disgust, contempt), with values between 1 and 7. We also also construct a
variable labeled “negative emotional valence” as the difference between negative emotions (fear, anger, sadness, disgust, contempt) and positive ones (joy and surprise). The variable is standardized to take values between -1 and 1, with higher values representing a more
negative emotional state, and 0 a perfect balance between positive and negative emotions. We will also examine how watching the immigration video influences each of these 7 emotions separately.

Outcome variables: Beliefs
For each of the five factual questions about statistics, we will use the following outcomes:
• Posterior beliefs and absolute distance between the posterior and the truth based on the revealed statistics
• Beliefs Updating: difference between posterior and prior beliefs (controlling for prior beliefs in the regression)

Outcome variables: understanding of statistics:
We include a battery of four multiple-choice comprehension questions. We ask respondents to base their answers only on the provided statistics.
Comprehension outcomes are coded such that higher values indicate better understanding of the statistical information. We will also look at whether the outcomes reflect a negative bias against immigrants in the interpretation of the statistics.

Outcome variables: Interpretation of Research Evidence on Border Wall
After reading the research summary, respondents are asked to evaluate what the evidence implies about the effectiveness of the border wall in reducing unauthorized crossings. Responses are recorded on a continuous scale from 0 to 10, ranging from “not at all effective”
to “extremely effective,” with the midpoint indicating that the wall is “somewhat effective.” This scale allows for a nuanced assessment of respondents’ interpretations, capturing variation in how strongly they perceive the evidence to support (or refute) the effectiveness of
the policy. Importantly, this question does not directly measure respondents’ policy preferences or prior beliefs, but rather their interpretation of the same piece of information. As such, it provides a way to assess how individuals process and make sense of evidence, including whether different respondents draw systematically different conclusions from identical research summaries.

Outcome variables: Posterior Views on Border Wall
To measure posterior views on the border wall, we design two questions to capture respondents’ beliefs about the effectiveness of border enforcement and their policy preferences regarding government spending on border security.

The first question asks respondents how increased enforcement and surveillance at the southern US border effects the number of unauthorized immigrants entering the country. Responses are recorded on a 0-10 scale ranging from “decreases it a lot” to “increases it a
lot,” with a midpoint indicating no real change. This question measures respondents’ beliefs about the impact of border enforcement on migration flows, capturing whether they perceive such policies as effective deterrents, ineffective, or potentially counterproductive.
The second question asks respondents whether government spending on border security should be reduced, maintained, or increased. The question is framed by noting that the US currently spends substantial resources on border enforcement. Responses are again recorded
on a 0-10 scale from “reduced a lot” to “increased a lot.” This question captures respondents’ policy preferences regarding the allocation of public resources to border enforcement and reflects their support for more or less intensive investment in immigration control.

• Desired immigration level, measured on a scale from reducing immigration a lot to increasing immigration a lot.
• Preferred policy toward undocumented immigrants, from full pathway to citizenship to full deportation.
• Views on when immigrants should be eligible for public benefits such as Medicaid, food stamps, or welfare.
• Views on when legal immigrants should be eligible to apply for U.S. citizenship.
• Composite policy-attitudes index constructed using principal components analysis (PCA) across the immigration-policy items.

Outcome variables: Views on Immigration Policy
The survey elicit respondents’ views on immigration policy using four questions. Each question captures a distinct dimension of attitudes toward immigrants and immigration policy.

The first question asks respondents to indicate whether immigration levels should be reduced, maintained, or increased. The response options range from “reduced a lot” to “increased a lot.” This question captures respondents’ general stance on the desired scale of immigration inflows.

The second question asks respondents to choose which policy best reflects their views on undocumented (illegal) immigration. The options span a wide spectrum, from granting all undocumented immigrants a pathway to citizenship, to allowing conditional pathways (e.g., based on requirements or childhood arrival), to denying any pathway, and finally to supporting full deportation. This question captures preferences over legalization and enforcement, reflecting how inclusive or restrictive respondents are toward undocumented immigrants.

The third question examines views on immigrants’ access to government assistance programs such as Medicaid, food stamps, or welfare. Respondents are asked when immigrants should become eligible for such benefits on the same basis as citizens, with options ranging from
immediate access upon arrival to never receiving access. This question measures respondents support extending public resources to immigrants.

The fourth question focuses on political integration by asking when legal immigrants should be allowed to apply for US citizenship. Because citizenship confers voting rights, the timing of eligibility reflects respondents’ willingness to incorporate immigrants into the political
community. The response options range from relatively short waiting periods (e.g., two years) to very long delays or complete opposition to granting citizenship.Taken together, these questions provide a comprehensive picture of respondents’ attitudes toward immigration policy across four key domains: the desired level of immigration, legalization of undocumented immigrants, access to social benefits, and political enfranchisement.
We will examine the treatment effect on each of these 4 domains, as well on index constructed using a PCA approach.

This is an individually randomized online survey experiment. After a disclaimer and baseline demographics module, respondents report prior beliefs about five immigration-related facts and complete a general attention check. The first randomization is a 2 x 2 design crossing video type with information about undocumented immigration.

Not available

Computer-based individual randomization embedded in the online survey platform. Respondents are assigned in equal proportions across the relevant arms at each randomization stage. The first-stage randomization assigns video type and the message about undocumented-immigration trends; the order of the video and statistics screens is randomized independently; the second-stage randomization assigns one of three research summaries about border-wall effectiveness.

Individual respondent.

No

N/A. Individual-level randomization; no clusters are planned.

9,000-10,000 adults

We will recruit subjects via Cint, a multinational market research platform. We will recruit about 9k-10k subjects; we will assign the same number of respondents to each of the to each of the different treatment conditions: emotional video versus cartoon, decrease versus increase statistics, order of the video and statistics, and the three different research summaries

In the first part of the analysis, we will compare outcomes either between emotional versus cartoon video, or between the type of video interacted with the decrease/increase statistics (in a 2x2 design)-- or interacted with the order of the video and statistics. In the simple treated-control comparison, we will have about 4,500 subjects per condition and hence a 80% power to detect an effect size of 6% of a standard deviation at a 5% significance level. In the 2x2 design, we will have about 2,250 subjects per condition and hence a 80% power to detect an effect size of 8% of a standard deviation at a 5% significance level. In the second part of the analysis, we will compare outcomes between the type of video interacted with the research summary. In this 2x3 design, we will have about 1,500 subjects per conditions, and hence a 80% power to detect an effect size of 10% of a standard deviation at a 5% significance level.