Research

Three of my paper projects rely on data from a custom survey of 1300 Muscovites collected 2018, carried out in the aftermath of the announcement of the Moscow Housing Renovation Program — a massive state-funded urban renewal program. Together with Ekaterina Borisova, Israel Marques II, and Regina Smyth explore the effects of this top-down, autocratic housing policy on social norms, policy preferences, and accountability.

In the project, I was responsible for the development of the sampling strategy, and partly for survey design and subsequent data analysis. We wanted to compare the residents of buildings that were included in the program with residents of the excluded building. At the same time, we were aware that the government’s decision to include buildings in the program was nonrandom and was shaped partly by the expected level of resistance to the program. So the challenge was to produce a matched pair sample based on factors not related to pre-existing variables of interest — social capital, norms, or redistribution preferences. To solve this problem, I used information on voting in previous elections at the finest level it was available. The intuition was that the Moscow city government had no finer-level information on the preferences of building residents; hence one could compile the list of election precincts that had both included and non-included buildings and survey the residents of those buildings.

In the first project (Borisova, Smyth and Zakharov, 2023), we study the effects of the top-down authoritarian policy on the formation of pro-social norms. Much of the existing work on the changes in pro-social norms and behavior in contemporary societies relies on path-dependent and historically-based arguments, or on the effects of democratic institutions, and cannot account for the periods of rapid change or the short-term accumulation of norms, especially under non-democratic regime types. Our work contributes to this literature by demonstrating that social norms of cooperation can arise quickly and as a response to government policy. We argue that the demands for social interaction embedded in the structure of the policy prompted residents of included buildings to interact to understand the program and organize for a house-level vote to secure benefits. Comparing a matched sample of residents living in buildings included and excluded from the program, we find that these interactions led to changes in pro-social norms. Notably, we also find spillover effects of the housing renovation program on the activity related to pension reform and voting in local elections. Strikingly, the effect of policy is larger in smaller buildings, in accordance with the prediction that collective action is easier to achieve within smaller groups. Our findings show the effect of authoritarian policy on societal attitudes and behavior and contribute to theories of short- to medium-term accumulation of societal capacity to resolve collective action problems in response to autocratic policy initiatives.

In Marques and Zakharov (2024), we use data from the survey to investigate whether, in autocratic settings, inclusion in specific government programs leads to spillovers in support for future social policy programs and redistribution generally. We find that residents of buildings included in the program have more pro-redistribution attitudes, with the effect being mediated by more positive attitudes toward the government. We believe that such programs can shape subsequent attitudes if they are carried out in a way that builds institutional trust by engaging citizens in the policy design process and educating them on its implementation. Our results shed light on how programs can be used strategically to promote a redistributive agenda and also suggest a pathway for the co-persistence of redistribution preferences and redistributive state policies.

In Borisova et al. (2023), we look at whether and how citizens assign responsibility for policies imposed by their authoritarian governments. Earlier work suggests that the assignment of blame is shaped by regime strategies that rely on framing to deflect blame toward lower-level officials or institutional actors. Our results show that citizens' capacities to assign responsibility vary across individuals and across specific policies in response to state strategies to consult with those most affected by specific policy initiatives.

One of my primary research goals is to examine the strategic behavior of political actors under authoritarian regimes. In a recent working paper with Ora John Reuter, Vladimir Shuklin, and Denis Volkov (Zakharov et al., 2023), we use an unexpected event during survey design to study how the 2023 coup attempt in Russia affected attitudes toward the regime and the coup’s leadership. Our findings show that authoritarian regimes can leverage control of the information space to demonize coup leaders and avoid major public opinion fallout swiftly. In the theoretical paper (Zakharov, 2016), I focus on the issue of the loyalty-competence tradeoff in authoritarian regimes — why do autocrats prefer to hire incompetent subordinates. The argument I make in my paper is that less competent subordinates will not invest in loyalty as they do not expect to be hired by a dictator’s potential successor. This was one of the first works modeling autocratic promotion decisions. 

In a working paper (Zakharov, 2018), I investigate resource curse, modeling a dictator’s decision to allocate state resources between economic development and security, and predict an economic resource curse in the absence of institutions that limit the dictator’s ability to prolong his term in office. With Ruben Enikolopov, Vasily Korovkin, Maria Petrova, and Konstantin Sonin, we use a field experiment to study electoral fraud, which is a widespread phenomenon, especially in electoral autocracies. Our work is one of the first to provide quantitative evidence of the extent of electoral fraud, despite the common knowledge that fraud exists and sometimes is widespread. To estimate the extent of electoral fraud in a 2011 election in Moscow, Russia, we exploit the random assignment of independent observers to polling stations. We find that the mere presence of independent observers at the polling stations decreases reported vote shares of the incumbent party, clearly demonstrating the presence of electoral fraud (Enikolopov et al., 2013). Moreover, our results suggest that the extent of the fraud was sufficient to have had a substantial impact on the outcome of the elections; they also confirm that the presence of observers is an important factor in ensuring the integrity of the procedure. In a working paper with Konstantin Sonin and Austin Wright (Sonin, Wright, and Zakharov, 2023), we model the interplay between civilian informing and insurgent extortion/retribution in a civil war. Drawing on newly declassified military records and a novel instrumental variables approach, we find robust, direct evidence that civilians respond to insurgent predation by providing intelligence to security forces in Afghanistan. 

Elite purges are an instrument of power control in authoritarian regimes. The purges in the Soviet Red Army during the 1937-38 Great Terror were one of the most intensive purges on record. Within two years, almost two thirds of the 1863 Soviet officers holding general-grade military ranks in 1936 were arrested; almost a half were executed. To analyze the patterns of repression among the Soviet high command, we compile a biographical dataset from a large number of open and archival sources sources, aided by over 15 research assistants over the past 3 years. To our knowledge, this is the first large-N study of a military purge in an autocracy, looking at how repression was related to background and biographical aspects of individuals, as well as their structural role; whether particular groups were targeted; and how different theories measure up to data.

Some of our results are familiar to historians: the probability of being repressed was higher for certain ethnic minorities, higher ranks, and those with prior foreign contacts. We also demonstrate that Stalin targeted most competent officers: controlling for other characteristics, including the military rank and party history, probability of repression was much higher for younger cadres. Combined with the results that military promotions in 1941, the first year of World War II on the Soviet territory, were inversely related to age, this is the first systematic evidence that the Great Terror directly impacted the disastrous Red Army performance in the first years of German invasion.

For some 1400 individuals, we were able to obtain information on their service records, yielding a total of 25000 individual-position pairs. We find that commander-subordinate relationships during the early 1930s seemed to be the source of additional risk for the subordinates during the Great Terror. In general, our findings suggest that current and former contacts played a role in the selection of targets during the purge, but that role was not decisive; this suggests that the purge was largely preventive in nature and not a response to an imminent coup threat. 

Three of my ongoing projects are going to use data from the COVID-19 vaccine preference and opinion survey (or CANDOUR II study). The goal of the survey is to conduct global research related to health, economics, and politics in response to the COVID-19 pandemic. The second wave, for which data collection is now in progress, is a nationally representative survey of 26000 individuals in 19 countries, including 5500 individuals who were surveyed in the first wave of the survey in 2021.

In the first project, in collaboration with Raymond Duch, Peter Loewen, and Thomas Robinson, we look at how the COVID-19 pandemic has affected support for, or satisfaction with the performance, of incumbent governments, and how citizens evaluate the performance of governments in managing the public health and economic crises (Duch, Loewen, Robinson and Zakharov, 2024).

We are going to use a conjoint experiment embedded in the survey. The subjects are presented with eight separate profiles describing government performance during the COVID pandemic, for a total of over 200000 observations. The profiles will differ on the following characteristics, including GDP growth, employment growth, lockdown duration, and the number of deaths. The subjects will then be asked whether they are willing to reelect a leader given this policy outcome. Our preliminary findings indicate that all characteristics of the profile affect the decision to support the incumbent; hence, we find that the COVID-19 pandemic has had significant political ramifications in democratic regimes. Moreover, the effect of the features was markedly heterogeneous across voters, depending on COVID exposure, left-right orientation, age, and behavioral characteristics such as patience and risk aversion.

In the second project, in collaboration with Raymond Duch, Thomas Robinson, and Sonja Vogt we study the effects of COVID-19 on cooperative behavior in a public goods game, embedded within the second wave of CANDOUR II study (Duch, Zakharov, Robinson and Vogt, 2022). Preliminary findings indicate that being exposed to COVID affects cooperative behavior, but the exact effect depends on the severity of exposure. Free-riding behavior is always reduced, while conditional cooperation is negatively affected by traumatic events (death of a relative) but is positively affected by less traumatic events (knowing somebody who was sick with COVID).

In the third project, we look at the effect of COVID-II on economic preferences, such as risk preferences, reciprocity, altruism, time discounting, and preferences for redistribution (Duch, Zakharov, Robinson and Clarke, 2022). These preferences are central to a large number of economic decisions, such as investment, savings, and education choices, and to support for state redistributive policies that ultimately affect the level of inequality in the society.

The goal of our analysis is to investigate how these preferences were impacted by the COVID-19 pandemic, and to explain the causal mechanisms for the effect — to what extent the effect was due to personal traumatic experiences, lockdowns and other compliance measures, and loss of income. A significant advantage of the CANDOUR project over surveys localized to individual countries is that we will be able to see how the effect of COVID exposure on economic preferences and cooperation depended on the extent of state response to COVID, such the provision of relief during the pandemic.

I use theoretical and empirical tools to examine strategic behavior in politics and how it is shaped by political institutions and preferences. In a recently published paper coauthored with Kemal Akoz I examine the effect that the homogeneity of preferences within the opposition electorate has on voter turnout, given a single pro-incumbent and multiple pro-opposition groups (Akoz and Zakharov, 2023). If each opposition group is represented by a separate candidate, there is a free-rider effect: the opposition turnout is lower if different opposition candidates are more substitutable. If there is a single pro-opposition candidate, the effect is the opposite under the proportional representation, and under a winner-tale-all system it depends on the size of the opposition, weighted by the intensity of preferences toward the opposition candidate.

In several theoretical and empirical papers, I examined the effect of valence or parties or candidates, or features which are seen in a positive light by all voters (such as the experience or trustworthiness), on candidate policies. Empirical studies suggest that valence is a key input in voting decisions, sometimes being even more important than policies chosen by parties and candidates. In a paper coauthored with Alexander Shapoval and Shlomo Weber we modeled the effect of incumbent valence on an equilibrium with candidate entry (Shapoval, Weber and Zakharov, 2019), finding that high-valence incumbents tend to adopt centrist positions. In an earlier paper (Zakharov, 2009) I modeled the possibility that valence can be chosen endogenously and at a cost, and found that under these assumptions parties can choose strategically divergent policy positions. In other papers (e.g. Schofield and Zakharov, 2010) valence of parties was estimated directly using empirical methods.  

In a series of formal-theoretical works, I examined how electoral competition is affected by how party or candidate vote shares are translated into payoffs (for example, a political party may care not only about winning a majority, but also about its share of seats). Building on an earlier work (Zakharov, 2012), in a paper with Constantine Sorokin I found that, with a finite number of stochastic voters, Downsian policy convergence is only possible if the if the electoral competition game is constant-sum, such as when both parties or candidates are probability-of-win maximizers or vote share maximizers (Zakharov and Sorokin, 2014). In a subsequent work (Sorokin and Zakharov, 2018) we assumed that the electorate can be large, with an infinite number of voters, and show that the presence of aggregate uncertainty (such as changing economic conditions or political scandals) is a necessary condition for the utility functions of candidates to have an effect on political competition. This result is not trivial as it does not directly follow from the law of large numbers.

With Denis Ivanov and Alexander Libman (Ivanov, Libman and Zakharov, 2023), we use an online survey of 2107 Moscow residents and find evidence of a human capital externality in civic and political participation in an authoritarian setting. Russia is a particularly interesting case for studying the authoritarian middle class because it is a relatively rare example of a modern authoritarian state with a large and internationalized educated class; the majority of autocracies have relatively small educated strata. We find that educated and upper-middle-income individuals are more politically and civically active; at the same time, the presence of such individuals in the neighborhood creates an externality, creating more vibrant local politics and lowering the cost of civic and political participation for both middle-class and those outside of it. Our observational findings are supported by an embedded conjoint experiment.

Another kind of social norms that I am interested in are preferences for truth-telling that facilitate social interactions by reducing lying. In an experiment involving over 1000 individuals from the U.K., Russia, and Chile (Duch, Laroze and Zakharov, 2020) we find that both incidence and extent of lying do not depend on the extrinsic benefit of lying, which is not consistent with the moral cost of lying being a smooth, increasing, and concave function of the extent of lying. 

In a theoretical paper, I exploit the possibility that lying can be managed through a mechanism design approach, and find that less lying may occur in equilibrium if the reports by the agents are not perfectly monitored (Zakharov, 2022a). In Zakharov (2023), I again model lying behavior - this time, assuming that the image concerns of individuals can be heterogeneous, and find that making lying more socially costly can lead to an opposite effect of making lying more frequent due to a crowding-out effect: increasing the social cost for a subset of agents will make them choose high-value reports less frequently, making such reports less socially costly for the rest of the agents, and inviting them to lie more. 

In Zakharov, 2024, I investigate whether the preferences for redistribution and social policy are affected by the awareness of the taxes that one pays. I use a survey experiment in Russia (a country where a significant fraction of individual payroll taxes is paid by the employer and is in effect hidden from the taxpayer) and find evidence of fiscal illusion — redistribution preferences are indeed affected by whether the tax beliefs are biased, and by the direction of the bias.

In Zakharov and Chapkovski (2025) we analyze whether redistribution preferences are affected by interstate conflicts using a preregistered survey experiment conducted in Russia during the Russo-Ukrainian war. The order of questions in the survey was manipulated to remind some respondents of the war before measuring the outcome variables. We find that among individuals who favor the war, the war reminder increases preferences for redistribution. Our two-wave design allows us to investigate the channels for this effect; we find that it is partly due to increased trust in the government. We also observe an increase in prosocial preferences among individuals who support the war, but this effect is not associated with the increase in redistribution preferences. 

In my other projects, I leveraged experimental methods to look at such phenomena as social polarization and the resulting toward holders of opposite opinions in an autocratic context (Chapkovski and Zakharov, 2022), or the role that social status plays in social interactions (Zakharov and Bondarenko, 2021).