Brief 86: Do political preferences affect policy learning and uptake? Evidence from a field experiment with informal entrepreneurs
EGAP researchers: Martín Valdivia, Cesar Zucco
Other researchers: Anna-Katharina Lenz, Rafael Goldszmidt
Key takeaways: Providing informal entrepreneurs with one-on-one information and assistance meaningfully increased formalization in Brazil’s Individual Microentrepreneur (MEI) program, but political attitudes toward the enacting government (Lula/PT) had only weak and inconsistent effects on take-up. Making the program’s political origins more salient did not widen the gap between supporters and detractors as expected. Instead, it led detractors to learn more about the program while supporters relied on the political cue in place of information, suggesting endorsements and information can act as substitutes rather than complements. If so, raising uptake in polarized settings may call for communication tailored to different political groups rather than a single message.
Geographical Region: Brazil (Rio de Janeiro)
Type of study: Pre-registered field experiment (randomized controlled trial)
Preparer: Danielle Bohonos
Executive Summary
This study examines whether political preferences shape citizens’ uptake of and learning about government programs, using Brazil’s Individual Microentrepreneur Program (MEI). MEI offers informal entrepreneurs subsidized access to social security and business-related incentives and was introduced under President Lula. The authors randomly assigned MEI-eligible informal entrepreneurs in Rio de Janeiro to a control group, a main treatment arm (one-on-one information and assistance), or an endorsement arm (the same assistance but explicitly linking the program to Lula). They find that the informational treatment increased formalization, that supporters of Lula/PT responded somewhat more strongly than detractors, but that explicitly stressing the program’s political origins did not widen this gap as predicted. Instead, exploratory analysis shows that political endorsement affected the two groups differently in how much they learned: detractors absorbed substantially more programmatic knowledge when the political connection was emphasized, while supporters did not, yet the two groups ultimately formalized at similar rates, implying that political cues substituted for information acquisition among supporters.
Policy Challenge
Low take-up of beneficial government programs is a persistent problem across policy domains, commonly attributed to unfavorable cost-benefit calculations, lack of information about eligibility, enrollment hassles, social stigma, or behavioral frictions such as inattention and procrastination. A separate literature has established that political preferences shape how citizens evaluate politicians, policies, and information. Despite these two well-developed literatures, the intersection between political preferences and program uptake had received little direct empirical attention prior to this study.
This gap matters for policy design because if political identity functions as an “uptake heuristic”, then the success of a program may depend as much on citizens’ attitudes toward its sponsors as on the program’s actual terms. As such, this study has direct implications for how governments communicate about public benefits, especially in an era of rising political polarization worldwide.
Context
MEI was enacted in July 2009 during Lula’s second term as part of a broader portfolio of “social inclusion” programs that became a signature of his administration. The program offers a straightforward path to formalization: a low, subsidized monthly fee providing access to social security benefits (retirement, disability, sick leave, maternity leave) along with business-related advantages such as a taxpayer registration number. Despite generous terms, informality remains widespread among Brazil’s low-income entrepreneurs. At the time of the study (2017–2019), Brazil had grown sharply more polarized into pro- and anti-PT camps following corruption allegations against PT leaders and Rousseff’s impeachment, even as Lula himself remained fairly popular. This context offered the authors a chance to test whether latent political polarization spills over onto policy uptake for a program that predates the polarization itself.
Research Design
The authors first conducted a visual assessment of streets in selected low-income communities to identify informal entrepreneurs, then administered a baseline survey to 4,350 individuals (2017), excluding those already formalized or past retirement age. The remaining eligible sample was randomly assigned, at the individual level within neighborhood blocks, to a control group (N=434), a main treatment arm (N=432), or a smaller political-endorsement treatment arm (N=278). Both treatment arms consisted of a one-on-one, roughly 40-minute consulting session at the entrepreneur’s place of work, including a short informational video, printed materials, and an offer of further assistance; the endorsement arm’s video additionally featured a picture of Lula with a voice-over noting that he had enacted the program, while the main arm’s video instead featured Brazil’s Sebrae small-business support agency.
Outcomes included a binary “intent to formalize” measure, actual formalization, and formalization combined with two months of tax compliance. The authors estimate intention-to-treat effects using OLS regressions with block fixed effects and pre-specified controls, and test for heterogeneous treatment effects by interacting treatment assignment with the political support moderators, consistent with their pre-analysis plan (PAP).
Results
Consistent with Hypothesis 1, the informational and assistance intervention increased formalization: point estimates across both treatment arms show a roughly 5–9 percentage point increase in intent to formalize and about 5 percentage point increases in actual formalization and tax compliance, with the main formalization outcomes statistically significant. Support for Hypothesis 2, that Lula/PT supporters would respond more strongly than detractors, was weak and inconsistent: differences between supporters and detractors were clearest for the “softer” intent-to-formalize outcome, present but modest for actual formalization, and absent for the more demanding formalize-and-comply outcome. Hypothesis 3 was not supported: explicitly stressing the program’s political origins did not widen the supporter–detractor gap in formalization, and if anything, the gap was smaller in the endorsement arm than in the main arm for the two administrative outcomes.
Exploratory analysis of knowledge outcomes helps explain this null result. The endorsement manipulation produced a clear divergence in learning: detractors of Lula who received the endorsement treatment showed knowledge gains of roughly 19–20 percentage points, compared to only 0–6 percentage points (not statistically significant) among supporters. Because knowledge was strongly associated with formalization overall, this pattern suggests supporters formalized despite acquiring less information while detractors’ skepticism prompted them to scrutinize the material more closely, and they formalized in part because they learned more. This heterogeneity was not replicated using non-political trust moderators and does not appear to be explained by pre-treatment socioeconomic differences or by selective prior formalization among supporters, supporting the interpretation that the pattern reflects a political mechanism rather than pre-existing differences between participants.over, the effects are broadly effective across partisan groups, meaning that even Democrats were swayed by Republican politicians’ claims of misinformation, and vice versa. The sizes of the observed differences in support are also substantively meaningful. In one of the studies, for example, around 44% of participants who did not hear a rebuttal from the politician said they would not support them; among those who read about their claims of misinformation, that figure is just 32 to 34%.
In a follow-up test, the authors examine how claims of fake news compare against other possible strategies politicians might adopt. They find that alleging misinformation is only marginally more effective for maintaining voter support than simply denying that the scandal ever happened. But casting doubt on the truthfulness of the story does offer a significant advantage over directly apologizing. This result reveals a perverse incentive for politicians caught in a scandal: owning up to one’s mistake can help bolster the public’s trust in the current media environment to accurately report on politicians, but misleading voters offers the more promising way to maintain political support.
Interestingly, in two of the authors’ experiments, respondents were randomly assigned to learn about the scandal through either text-based or video evidence. Politicians’ claims of misinformation were generally only effective against textual reports, but not against videos (with one exception in their fourth experiment). This finding offers some reassurance that the increased prevalence of “deepfake” videos in recent years has not undermined voters’ trust in authentic video content, although the authors caution that the studies were conducted before the popularization of generative AI and more accessible and compelling deepfake-creation technologies.
Across the various studies, crying fake news does not appear to consistently reduce whether people actually believe in the scandal, although this depends slightly on how beliefs are measured. The experimental treatments also do not appear to reduce trust in the media, which might be explained by low baseline levels of trust among the American public. (In the authors’ sample, just 23 to 33% of participants who were exposed to the scandal but saw no misinformation claims indicated that they trust the media).
Lessons
For policymakers, the central takeaway is that providing convenient, one-on-one information and assistance can meaningfully raise take-up of well-designed benefit programs, even among historically hard-to-reach informal populations. The political findings call for caution. Linking a program to a prominent (but polarizing) political figure did not reliably increase uptake among supporters or reduce uptake among detractors. Instead, it appears to have changed how different groups processed the information.
Specifically, political framing appears to let sympathetic citizens to rely more on prior trust than on engaging closely with program details, suggesting that their decisions may rest on a weaker understanding of the program. In contrast, the same framing may prompt more skeptical citizens to scrutinize the information more carefully and ultimately acquire greater program knowledge, potentially helping overcome political resistance to a program if paired with a credible, non-partisan information channel.
More broadly, the findings suggest that political endorsements and policy information may function as substitutes rather than complements, underscoring the importance of pairing political communication with clear and accessible program information.