A Hint at How to Solve the Puzzle of Fondness for Authoritarianism
Do some people prefer authoritarian regimes, or do they just underestimate their social and economic costs? Democratic backsliding and popular enthusiasm for governments with authoritarian tendencies aren’t going away. Why is this, and what could be done about it? In an era of geopolitical shifts and the growing influence of authoritarianism, understanding research in the field of regime change and democracy is crucial, not just for academic researchers but for a variety of stakeholders.
The Unjournal recently commissioned an evaluation of a paper by Acemoglu et al. (2025), "Misperceptions and Demand for Democracy under Authoritarianism." This paper considers the importance of providing accurate information about the role and implications of democracy and media freedom in voters’ preference for authoritarianism, and draws on evidence from an online and a field experiment in Izmir, Türkiye, to measure the effect of such campaigns on beliefs regarding the quality of institutions and media freedom, and whether the experiments affected voting behavior. Both of our evaluators rated the work highly, highlighting the importance of the research on grant-making decisions to empower democracy. The research is significant because it suggests that providing information to citizens about government interference in media freedom and democracy led to changes in voting patterns. The evaluators were generally positive about the research, but found a small number of limitations that opened avenues for future re-analysis and further research on the subject, such as a comparison of pre- and post-treatment values, rather than only the latter (for the online part of the field experiment).
Read the discussion by one of our field specialists with a background in history and geopolitics.
Daron Acemoglu and co-authors’ working paper Misperceptions of Democracy under Authoritarianism claims to show that providing information on the erosion of media freedom and democracy under authoritarian-leaning governments may change voter beliefs about the government. In turn, this change of beliefs is said to increase the vote share of opposition parties. This paper provides insights into “one of the most important puzzles of the last few decades”: the enduring popularity of authoritarian government, according to one of the two expert evaluators that The Unjournal hired to consider these claims. On reading these, you might notice that the evaluations differ greatly, as you’ll see below.
The authors provided residents of randomly treated areas in the city of Izmir, Türkiye, with factual statements about the quality of democracy and media freedom, such as ”...as media independence in a country worsens, corruption increases... media independence has worsened and corruption has increased in Türkiye in the last 30 years”. Factual information such as this may be less accessible owing to the government’s “crackdown on the opposition and independent media”. The information was provided through an online experiment consisting of a survey containing information promoted through Facebook adverts, and a field experiment facilitated by door-to-door canvassing and discussions with residents. The authors found substantial differences between treatment and control groups. For example, they found that treated respondents’ opinions on the extent of decline in institutional quality were more negative by 6.8% of a standard deviation compared to the control group. Furthermore, voting patterns changed, with the online treatment and field treatment accounting for a respective 3.7% and 0.8% increase in the probability of respondents voting for the opposition.
Both evaluators were generally positive about this work. Both believed there would be a strong chance of publication in a highly regarded subject journal. However, both saw some important limitations to the research, some of which could be remedied by a modification of the authors’ data analysis, econometrics, and characterization of their results; other issues would require future experiments and data collection. For example, the first evaluator, Korhan Kocak, suggests that the authors should have compared “the difference between post- and pre-treatment values instead of only the post-treatment values” for the online experiment — this would be an easy robustness check.
The second evaluator noted the limitation caused by possible spillover effects. It’s difficult to regulate the distribution of information on a large scale, as was attempted in this research, without some of the information being revealed to those who are not targeted by the study. A significant spillover effect could potentially bias the estimated effect size towards zero, thus underestimating the true effect size, as even the "comparison group" would have been somewhat influenced by the treatment.
The second evaluator also opined that the results are likely to depend greatly on contextual factors such as “salience of recent scandals or disasters, and audience receptivity to nonpartisan messengers” and are therefore difficult to generalise across other countries that may have authoritarian or hybrid governments but which have not suffered recent scandals or disasters, such as in the aftermath of Turkey’s 2023 earthquake. As The Unjournal’s evaluation manager pointed out, “people are naturally suspicious of freely-provided information as it is hard to be truly non-partisan when information may hurt or help one political side”, and the political groups and their causes likely to benefit in different elections around the world will vary because of the particular context of the election and the politicians involved. Furthermore, the same evaluator highlighted that a population’s dislike of a regime doesn’t necessarily mean they desire democracy, as was the case in numerous 20th-century revolutions from China to Iran, and thus correcting misperceptions does not necessarily build democratic support.
However, the assessors considered the article to be of significance with substantial relevance beyond academia, despite its limitations; for example, Evaluator 2 considers the results useful for donors and for policy decisions in pro-democracy NGOs and election-monitoring organizations, which would make more progress towards their goals by the use of non-partisan information dissemination than by direct normative appeals to democracy.
All of this sets the table for promising work. If and when the authors make the code and de-identified administrative data scripts publicly available (the sooner the better), other researchers would likely be keen to do robustness replication work that would give us more understanding and more confidence in the results. E.g., one evaluator suggested excluding areas adjacent to the treatment neighborhoods from estimations of the treatment’s effectiveness. More ambitiously, future researchers could extend these RCTs to other countries, adapting to local contexts, to directly test their generalizability. If the authors’ findings are indeed robust and generalizable, these experiments could also have major direct social benefits.
