Corruption, the misuse of public office for private gain, has well-documented negative consequences. These consequences include reduced growth, exacerbated inequality and poverty, and the under-provision of public goods (Olken and Pande 2012; Dimant and Tosato 2018). Alongside these economic and social consequences, a rich literature investigates the political response of citizens to corruption. For the most part, this literature argues that a well-informed citizenry with non-corrupt alternatives will hold corrupt incumbents accountable by voting them out of office or protesting (Manin et al. 1999; Gingerich 2009). The thrust of this argument is that citizens use electoral institutions to restrain corruption.

While compelling, this argument ignores the fact that individuals have another option when they believe corruption to be high: exit (Hirschman 1970). While holding leaders accountable at the ballot box is a clear voice-based response, people also emigrate to avoid corruption’s negative impacts. I argue that emigration is especially attractive to citizens who feel their concerns cannot be adequately addressed by voice-based mechanisms that are supposed to curb corruption. Highly educated citizens in particular, who are often most informed about corruption and hold valuable human capital, likely see exit as both attractive and viable in the face of persistent corruption.

The possibility that people emigrate in response to corruption would suggest that at least some citizens grow skeptical that they can restrain corruption using domestic institutions. I argue that those who are most sensitive to corruption’s negative impacts are likely to view emigration as more effective at addressing corruption than voting or protesting, both of which require costly collective action and have uncertain outcomes. The empirical finding that electorates often do not hold corrupt politicians accountable may be due, in part, to emigration as a behavior that falls outside the realm of domestic politics, and is attractive to people most concerned about fighting corruption’s negative impacts.

I draw on a wealth of micro-level data from the Gallup World Poll to investigate how perceptions of corruption drive potential emigration, an outcome that a consensus of recent scholarship shows is highly correlated with actual emigration (Docquier and Peri 2014; Tjaden et al. 2019; Clemens and Mendola 2020). I find that when citizens perceive corruption to be widespread, they are more likely to intend to emigrate, to have made concrete plans to leave within the next year, and to have made active, costly preparations for their move. This relationship persists even when accounting for other attitudes closely related to corruption, including approval of the nation’s leadership and perceptions of economic conditions.

Educational attainment strengthens this relationship: advanced degree holders are much more likely to have made preparations for emigration when they perceive corruption to be high. I also show that corruption-concerned people who plan to emigrate choose less corrupt destination countries than those who are unconcerned with corruption, consistent with the logic that corruption-sensitive people use emigration to address their concerns. My empirical setup allows me to analyze how potential emigrants view voice-based alternatives while planning for exit. I find that potential emigrants have less favorable views of the electoral process in their home countries and are significantly less likely to have recently voted. The results suggest that deciding to exit is associated with declining confidence in, and use of, institutions that promote the exercise of voice.

My argument and findings bring scholarship on the political consequences of corruption into conversation with the international migration literature. In so doing, I identify a potential feedback loop between corruption, exit, and accountability. When corruption-sensitive citizens give up on voting and protesting and instead emigrate, they could inadvertently undermine domestic collective action capabilities to reduce corruption (Sellars 2019). This feedback loop may contribute to an equilibrium in which reducing corruption proves increasingly difficult. On the other hand, emigration can have positive economic and governance effects (Adams and Page 2005; Escriba-Folch et al. 2015; Cordova and Hiskey 2015), emigrants may “remit” anti-corruption norms from their destination countries (Ivlevs and King 2017), and corruption-sensitive emigrants may remain politically active from abroad (Ahmadov and Sasse 2016). Disentangling these potential effects is a topic ripe for future work.

Corruption and Political Accountability

The political consequences of corruption is a vibrant area of research.Footnote 1 Given that corruption has harmful effects on the wellbeing of most ordinary citizens, and that people consistently express opposition to corruption, this research agenda often takes political accountability as a point of departure: citizens should punish politicians who they perceive to be engaged in corrupt activities (Peters and Welch 1980). In a democracy, this involves voting out corrupt incumbents and engaging in protest (Ferejohn 1986; Besley 2006; Gingerich 2009). The resulting empirical prediction is that we should observe citizens voting out politicians who are credibly accused of corruption.

The evidence supporting this baseline expectation is mixed. Some studies find that voters punish incumbents who have engaged in corruption (Krause and Méndez 2009; Winters and Weitz-Shaprio 2013; Klasnja et al. 2016). However, a separate body of research finds that punishment for corruption is mild or non-existent. Politicians involved in corruption scandals are re-elected across a range of contexts, suggesting that punishing corruption is not as certain as our theoretical priors might suggest (Chang et al. 2010; Vivyan et al. 2012; Fernandez-Vazquez et al. 2016).

Recognizing these findings, a new wave of research instead focuses on identifying the conditions under which voters hold politicians accountable using domestic institutions. De Vries and Solaz (2017) illustrate how a variety of contextual factors create stumbling blocks to electoral accountability. To punish corruption, citizens must acquire information about corrupt acts, attribute blame to the correct political actors, and weigh other issues and considerations before choosing their behavioral response. At each stage, a number of factors could intervene between corruption and behavioral responses associated with accountability.

The information environment, for instance, is central: providing citizens with higher-quality information on malfeasance often leads to more attuned perceptions and more effective accountability (Ferraz and Finan 2008; Chang et al. 2010; Larreguy et al. 2020).Footnote 2 In-group biases also condition whether citizens ascribe blame for corruption to certain political actors. Complex institutional structures make it difficult to connect corrupt acts to the politicians who are responsible (Solaz et al. 2019; Anduiza et al. 2013; Tavits 2007).

A number of other factors influence how citizens respond to corruption, even when they have high-quality information and accurately attribute blame. There may be no non-corrupt alternative around which to coordinate, economic conditions may reduce the salience of corruption, or citizens may receive partisan representational benefits by voting for corrupt politicians (Zechmeister and Zizumbo-Colunga 2013; Eggers 2014; Schleiter and Voznaya 2018). Many individuals may be willing to tolerate some corruption if the ruling government compensates them appropriately (Shehaj 2020). Ordinary people may see all politicians as uniformly incompetent at controlling corruption (Pavão 2018). In such a case, voters may overlook corruption or abstain from political participation (Davis et al. 2004).

Corruption and Emigration

What do corruption-sensitive people do when they face these stumbling blocks that constrain the electorate’s ability to punish corrupt politicians? While scholarship on the political consequences of corruption investigates voice-based options (e.g., voting and protesting), I argue that when corruption persists, some people see exit as an effective way to seek private recourse (Hirschman 1970, 1978). When citizens grow dissatisfied with high corruption, they can emigrate, withdrawing from the state. Emigration is an alternative political response that, at the individual level, is a means to bring about reduction in perceived corruption, a political response that does not fall within the boundaries of accountability.Footnote 3

Why might we expect people to consider emigration in the face of high corruption when we would typically expect them to take political action in a spirit of voice? Political action aimed at reducing corruption is uncertain and costly. Voting out corrupt incumbents or organizing anti-corruption protests requires overcoming significant collective action and coordination problems, and people know that such efforts are often unsuccessful. Importantly, many individuals likely experience episodes in which elections and protests fail to significantly curb corruption (Klasnja et al. 2021). While emigration is also costly, it does not require coordination or collective action, its benefits can be internalized to the emigre, and its outcome in terms of reducing the negative impacts of corruption is more certain.Footnote 4 For people most concerned about avoiding the costs of corruption, I contend that exit can become attractive when compared to more uncertain voice-based actions like voting or protesting.

Additionally, corruption is often deep-rooted and pervasive (Mishra 2016). Even in democratic settings, countries find themselves stuck in high-corruption equilibria, in which citizens perceive that they have no non-corrupt alternative around which to coalesce (Pavão 2018). Elections or protests are unlikely to succeed in addressing corruption, and faith in political institutions weakens as a result (Morris and Klesner 2010; Clausen et al. 2011). In these environments, we might expect exit to be especially appealing. Finally, emigration could play an important role when accountability mechanisms are poorly institutionalized. Electoral punishment is not an option in autocracies, for instance, and electoral institutions in weak democracies are similarly lacking. While protest may be possible, participants often face costs associated with repression, and protesting to change the status quo in autocracies and weak democracies is highly uncertain (Carey 2006; Kalandadze and Orenstein 2009). For people who want to reduce the costs of corruption in such environments, my argument suggests that emigration might emerge as the most attractive response.

The decision between exit and voice can also be characterized in terms of the relative gains of each action for potential emigrants. Both exit and voice require incurring some cost in exchange for a potential benefit: reducing experienced corruption. My theoretical framework characterizes migration as a high-cost but high-benefit action to redress corruption concerns. Emigration is costly in a range of ways – it requires using scarce resources to finance mobility and entry into a destination country, acquiring information about employment opportunities in an unfamiliar labor market, integrating into a new political, economic, and cultural context, and severing proximate social ties with friends and family. While these costs are high, the benefits of migrating are also high – because many potential migrant destinations offer much less corrupt environments, migrating significantly reduces experienced corruption and its associated negative consequences. Despite the costs, the relative gains of migrating could be particularly high for potential emigrants who prioritize reducing the negative impacts of corruption on their own welfare.

By contrast, my framework characterizes voice as a relatively low-cost but low-benefit action. Exercising voice through voting or protesting, while certainly entailing some cost, does not require incurring the large costs associated with emigration – it requires few financial resources, does not impose the same informational costs, and social ties are not severed. However, the benefits of voice are arguably low in terms of reducing experienced corruption. As discussed above, collective action problems, high-corruption equilibria, and poorly institutionalized accountability mechanisms render voice-based options relatively ineffective at tackling corruption. While exercising voice is clearly less costly than emigration, for corruption-concerned potential emigrants, it also clearly provides far lower benefits in terms of reducing corruption’s negative impacts. In other words, I argue that the comparative gains from emigration are likely to be attractive for the most corruption-concerned citizens.

Existing research (e.g., Carreras and Vera 2018) conceives of abstaining from politics as an alternative form of exit in reaction to corruption. I instead conceive of exit in geographic terms – physically leaving a country. Geographic exit entails not just disengaging from the political process, but escaping the political process entirely by emigrating. While corruption might also cause citizens to abstain from politics (i.e., “internal exit”), abstention does not meaningfully redress citizens’ concerns about corruption, as they remain in the corrupt environment. By contrast, exit via emigration to a less corrupt destination country can result in lower experienced corruption.

While all individuals may see exit as attractive, I argue that highly educated people are most predisposed to emigrate as a response to corruption for two primary reasons. First, the highly educated are most likely to be well-informed about the extent of corruption, and crucially, are most likely to disengage from political action when they perceive corruption to be high (Seligson 2002; Croke et al. 2016; Agerberg 2019). Given that more educated citizens are most informed about corruption and often most pessimistic about the effectiveness of accountability mechanisms, emigration becomes relatively more attractive. When one becomes convinced that domestic institutions have difficulties meaningfully restraining corruption, emigration grows even more tempting.

Second, the highly educated are the most internationally mobile demographic. We know from the migration literature that advanced economies face high-skill labor shortages (Zimmerman and Kahanec 2011). To compensate, they have expanded policies that attract highly educated foreigners, leading to a global competition for mobile talent (Czaika and Parsons 2017). As a result, advanced degree holders have relatively more lucrative exit opportunities. I argue that given those same people are most likely to perceive corruption to be high, then they are most likely to consider emigrating, both to avoid the costs of corruption and because the human capital they hold makes mobility more frictionless. In other words, we should expect the relationship between corruption and emigration to be especially strong among highly educated individuals because they are likely to be well-informed, pessimistic about domestic redress, and face lower mobility constraints. By contrast, lower-educated individuals are more likely to face barriers to their mobility as a result of restrictive immigration policies that target them. As a result, while lower-educated individuals may also want to migrate due to corruption, they are less capable of doing so.Footnote 5

While my argument focuses on the role of corruption, a crucial alternative explanation is that emigration decisions are driven by purely economic considerations. This perspective suggests that wage maximization and employment opportunities – rather than corruption – explain the decision to migrate internationally (Clark et al. 2007; Fitzgerald et al. 2014). Poor economic conditions at home and stronger opportunities abroad might explain any observed relationship between corruption and emigration, since economic performance is correlated with corruption. I do not deny that economic considerations shape migration decisions; opportunities for employment and higher wages are certainly salient, and I account for them in the empirical analyses that follow. Instead, I argue that in addition to wage maximization, corruption plays an independent role because of its negative impacts on individual and social welfare. Additionally, economic considerations and corruption are arguably not totally distinct explanations. High corruption constrains opportunity by weakening economic growth (Dimant and Tosato 2018). As a result, corruption is arguably an underlying driver of emigration via its negative effects on economic development.

My argument also presumes that there are low-corruption destinations to which corruption-concerned individuals can move – if potential destination countries were just as corrupt as origin countries, exit would not redress corruption concerns. Indeed, commonly used measures indicate modest upward trends in corruption in popular migrant destinations like the US, Germany, the UK, and other Western European countries (Coppedge et al. 2020). If this trend were to continue, it would constrain the effectiveness of exit, as corruption differentials would grow smaller between origin and destination countries. Despite this possibility, there still remain large, meaningful disparities in corruption between origin and destination countries that corruption-concerned individuals can leverage to redress experienced corruption.

In short, I argue that emigration is an attractive individual-level response to corruption when citizens perceive that voice-based alternatives like voting and protests are insufficient. Emigration can act as an effective non-accountability-based mechanism, especially when individuals perceive that domestic institutions are too weak to meaningfully curb corruption. I expect education to be a meaningful moderator: the highly educated in particular are likely to view emigration as attractive when they perceive corruption to be high, given that they are relatively most informed, often most pessimistic about the prospects of domestic institutions, and have viable exit options.


I use the Gallup World Poll (GWP), a globally representative public opinion poll of individuals, to test the implications of my argument. The GWP conducts a yearly survey of at least 1,000 randomly selected individuals from each country.Footnote 6 The result is a wealth of data that is representative of more than 99 percent of the global population. The survey asks consistent questions about corruption and potential emigration over a long time period. I use the GWP’s yearly survey from 2009 to 2015 because of data availability for key control variables.Footnote 7 Sample size varies by model, but respondents come from about 135 countries. I include respondents from all countries in the analysis.Footnote 8 While the GWP admittedly offers only a snapshot of an individual pondering emigration at a single point in time, the high quality and representativeness of the data allow for a rich associational analysis that speaks to my research question.

Potential Emigration

I use the GWP because it provides the most extensive worldwide data on potential emigration and is widely used to study migration behavior (e.g., Dustmann and Okatenko 2014; Otrachshenko and Popova 2014; Dao et al. 2018; Manchin and Orazbayev 2018; Clemens and Mendola 2020). Three nested survey questions about emigration are asked. The first asks about emigration intentions: “Ideally, if you had the opportunity, would you like to move permanently to another country, or would you prefer to continue living in this country?” Respondents can answer yes, no, don’t know, or refuse to answer.Footnote 9 Those who answer yes are asked about concrete plans to emigrate: “Are you planning to move permanently to another country in the next 12 months, or not?”Footnote 10 Those who answer that they both intend and plan to emigrate are asked about active preparations to emigrate: “Have you done any preparation for this move (for example, applied for residency or visa, purchased the ticket, etc.)?” I use all three questions as increasingly restrictive measures of potential emigration.

The GWP data on potential emigration is widely regarded as strongly predictive of actual emigration. A consensus of recent research in economics and migration studies empirically confirms that potential emigration measured in the GWP is strongly correlated with actual emigration rates and migration flows (Docquier and Peri 2014; Tjaden et al. 2019; Clemens and Mendola 2020). In particular, Tjaden et al. (2019) find that potential emigration as measured by emigration plans and preparations is strongly correlated with actual macro-level migration flows, at around 0.8. The fact that the proportion of people who have made any plans or preparations for emigration is closely correlated with actual emigration rates means that these variables can be safely used as proxies for actual emigration. Given micro-level time series data on realized emigration for a sufficiently large selection of countries over time is not available, the GWP represents by far the best available data source on individual-level emigration behavior worldwide.

An alternative approach might employ country-level data on actual migration rather than individual-level data on potential emigration. While there are advantages to using realized rather than potential emigration, I use these data for a few reasons. First, emigration is a micro-level decision that happens at the individual or household level. The GWP therefore allows me to study individual perceptions and emigration behaviors at the most theoretically relevant level of interest. Second, the Gallup data offer uncommon insight into how corruption perceptions (as well as other factors) shape the process by which individuals make the decision to emigrate while other behaviors (such as political action) are still viable alternatives to address their concerns. For this reason, I argue that potential emigration data are especially helpful for testing my argument. Combined with existing analysis demonstrating robust links between potential and realized migration, I regard the GWP as the relatively more appropriate data for the analyses that follow.

Fig. 1
figure 1

Percentage of sample who state they have made costly preparations to emigrate by country. Percentages calculated by pooling all survey waves from 2009-2015 for each country. Source: Gallup World Poll

Figure 1 maps the percentage of individuals for each country who state they have made costly preparations to emigrate (the most restrictive measure). These emigration preparations rates are calculated by pooling all survey waves from 2009-2015 for each country. The map illustrates that major corridors of emigration during the time period – Central America, Africa, and Eastern Europe – also have the highest percentage of individuals who state that they have made preparations to leave. The US and Western European countries see much lower emigration preparation rates by comparison.

Corruption Perceptions

The GWP also affords an individual-level measure of corruption perceptions. Respondents are asked the following question: “Is corruption widespread throughout the government in (country), or not?” Respondents can answer yes, no, or don’t know. I exclude all respondents who do not respond by saying yes or no, leaving a binary measure of corruption perceptions.Footnote 11 While a more gradated measure of corruption perceptions would be preferable to a binary response set, this measure still identifies differences in individual-level corruption perceptions and allows me to leverage the GWP’s rich potential emigration data.

Many studies that explore the consequences of corruption use individual or professional perceptions as a proxy for “actual” corruption.Footnote 12 Here, I argue that individual perceptions are of direct interest because the unit of interest is the individual. Ordinary people decide how they respond to corruption based on their own perceptions, and not necessarily on an objective measure of corruption’s existence, or on the perceptions of businessmen or country experts. If individuals have access to reliable information about corruption and appropriately incorporate their personal experiences, then perception and reality might be tightly linked. Nevertheless, even if perceptions are only loosely related to actual corruption, I argue that they can still have a significant effect on the individual-level decision to emigrate.


My argument suggests an interactive relationship between educational attainment and corruption perceptions. The GWP codes individuals based on their reported educational attainment. I construct an indicator for whether an individual has a tertiary education (completed four years of education beyond high school and/or received a college degree or more). I focus on tertiary-educated individuals because this level of human capital is most relevant for lucrative international migration opportunities. My expectation is that the relationship between corruption perceptions and potential emigration is stronger among higher-educated respondents.

Control Variables

I account for additional potential confounders that could influence both perceptions of corruption and potential emigration. Women are less likely to condone or engage in corruption, and also gain less economic return on emigrating than men, especially if they are married (Swamy et al. 2001; Melgar et al. 2010; Morrison and Lichter 1988); I control for gender. Since older people tend to perceive higher levels of corruption and are less likely to emigrate, I control for age (Cabelkova and Hanousek 2004; United Nations Population Division 2013).

Socioeconomic characteristics are also salient. Higher-income individuals more frequently pay bribes due to their deeper pockets, which colors their perception of corruption (Mocan 2008; Hunt and Laszlo 2012). Meanwhile, the relationship between income and emigration is well-documented: people move to take advantage of wage differentials, and only those with sufficient financial resources can emigrate in the first place (Dustmann and Okatenko 2014). I include the individual’s within-country household income quintile. Existing literature documents that unemployed people tend to perceive higher levels of corruption, and unemployment is also a predictor of considering employment opportunities abroad (Gatti et al. 2003; Melgar et al. 2010). I include a binary measure of employment.

Transnational social networks are paramount in shaping an individual’s decision to emigrate, and new research suggests that transnational networks can influence perceptions of corruption (Clark et al. 2007; Fitzgerald et al. 2014; Ivlevs and King 2017). I use responses to the following question: “Do you have relatives or friends who are living in another country whom you can count on to help you when you need them, or not?” to account for this. In line with other questions, respondents can answer yes, no, don’t know, or refuse to answer. I again exclude all those without a yes/no answer to leave a binary response set.

Finally, perceptions of corruption are likely tied to other perceptions of the government or economy that we might also expect to have an effect on potential emigration. Individuals who may already have more generally negative perceptions of their country may be both more likely to respond that their country is corrupt and intend to emigrate. Two primary related perceptions are a) approval of the current government in power and b) perceptions of current economic conditions. I take advantage of the GWP to account for these other perceptions. Using responses to additional questions, I construct indicator variables for whether or not a respondent a) approves of the government’s current leadership, and b) believes that the country’s economic conditions are worsening. While it is difficult to account for every possible alternative perception related to corruption, I attempt to ensure that the results remain robust to the most plausible potential confounders. Appendix Table 2 displays summary statistics for the global sample for all variables used in subsequent analyses.

Empirical Analysis

I estimate a series of logistic regression models with country\(\times\)year combination fixed effects.Footnote 13 I use this modeling strategy because it results in a within-country-year model; that is, the country-year dummy variables capture factors that are constant within a given country-year, allowing us to focus on variation between individuals within that country-year. This estimation strategy allows me to compare two individuals who live in the same country and are surveyed in the same year, but who have differing corruption perceptions. This estimation strategy is beneficial because it obviates the need to control for any potential confounding variables that are fixed for all individuals in the same country and in the same year, like economic development or growth, the occurrence of elections, political institutions, or any other country-year-specific conditions. I also consider this strategy superior to only country fixed effects, which would result in the comparison of respondents in the same country as many as six years apart and would allow for over-time within-country variation to bias the results. My results are robust to choosing other modeling strategies, such as a linear probability model and random effects logistic regression. I report robust standard errors clustered by country-year combination.

Corruption and Potential Emigration

Table 1 presents the baseline results. Column (1) is a bivariate model of corruption perceptions and emigration intentions. Column (2) adds socioeconomic and demographic controls, while Column (3) further adds alternative political and economic perceptions. Column (4) instead models emigration plans, while Column (5) is a model of active emigration preparations; both models include the full set of control variables.Footnote 14 Due to missing data on key control variables, as well as on the independent and dependent variables of interest, the number of observations declines across Models (1) through (5). Specifically, the inclusion of controls for income quintile, employment status, and transnational social networks account for much of the missingness between Model (1) and the other models. While losing significant amounts of data is not ideal, I regard accounting for personal economic circumstances and social networks as essential to accurately identifying the association between corruption perceptions and migration intentions, because they are key competing explanations for emigration decisions. Additionally, despite missingness, Appendix Table 1 demonstrates that the sample used remains large and broadly globally representative.

The results illustrate that perceiving widespread corruption is associated with an increased probability of potential emigration, whether one considers intentions, plans, or preparations. While other kinds of political and economic perceptions also meaningfully influence potential emigration, even when they are accounted for in Columns (3) through (5), corruption perceptions continue to have a positive and statistically significant effect.Footnote 15

Table 1 Corruption perceptions and potential emigration

Substantively, perceiving corruption to be widespread is associated with a 4.7 percentage point increase in the probability of intending to emigrate, a 0.4 percentage point increase in the probability of planning to emigrate, and 0.2 percentage point increase in the probability of having made costly preparations to emigrate, on average.Footnote 16 The decline in absolute effect size across increasingly restrictive measures of potential emigration in large part reflects the lower baseline probability that an individual has made plans or preparations to emigrate. In percentage terms, perceiving corruption to be widespread is associated with a 25 percent increase in the probability of intending to emigrate, a 15 percent increase in the probability of planning to emigrate, and a 16 percent increase in the probability of preparing to emigrate.

My argument suggests that believing corruption is widespread reduces confidence in institutions that can address corruption, increasing the attractiveness of emigration relative to political action. One implication is that including measures of confidence in political institutions should mitigate the estimated coefficient on corruption perceptions. I explore this possibility in Appendix Table 6, where I add an indicator of whether an individual is confident in the honesty of elections. As expected, including this indicator noticeably reduces, but does not eliminate, the relationship between corruption perceptions and emigration intentions. This might reflect that corruption perceptions are associated with emigration through pathways besides confidence in elections alone. For example, corruption may be associated with reduced confidence in other institutions (such as the judicial system or local governments) that might address corruption. Worries about corruption could also be associated with non-institutional concerns about the ability to achieve economic mobility.

The Role of Education

These results support the proposition that corruption perceptions are meaningfully associated with potential emigration. However, they do not test whether education moderates the relationship between corruption and potential exit. Table 2 shows the interactive relationship of corruption perceptions and education. These models suggest that the relationship between corruption perceptions and potential emigration depends heavily on educational attainment. Regardless of whether we model emigration intentions, preparations, or plans, the interaction between corruption perceptions and higher education is positive and statistically significant.Footnote 17

Table 2 Corruption Perceptions, Education, and Potential Emigration

Figure 2 displays the predicted interaction of corruption perceptions and education for emigration intentions, plans, and preparations respectively. For lower-educated individuals, perceiving widespread corruption is associated with a 4.3 percentage point increase in the probability of intending to emigrate, a 0.3 percentage point increase in the probability of planning to emigrate, and just a 0.1 percentage point increase in the probability of having made costly preparations to emigrate.

For higher-educated individuals, the estimated effect is much larger across the board: a 7 percentage point increase in the probability of intending to emigrate, a 0.9 percentage point increase in the probability of planning to emigrate, and a 0.5 percentage point increase in the probability of having made costly preparations to emigrate. In percentage terms, perceiving corruption to be widespread is associated with a 41 percent increase in the probability of having made emigration preparations for high-educated individuals (compared to just an 11 percent increase for low-educated individuals). Even when looking at the most restrictive measures of potential emigration, corruption perceptions prove particularly salient for people who hold advanced degrees.

High levels of corruption may be be associated with specific political leaders rather than the broader political environment. To account for this possibility, in Appendix Table 10 I include an interaction of a respondent’s education and whether or not they approve of the country’s political leadership. The interaction between corruption perceptions and education remains positive and statistically significant across all models; the interaction between leadership approval and education is negative and statistically significant for migration intentions, but does not reach conventional levels of significance for migration plans or preparations. This suggests that while corruption indeed may be linked to specific leaders, broader corruption perceptions, interacted with education, continue to play an independent role.

My argument suggests that the interactive effect of education and corruption is in part due to information acquisition abilities. Higher-educated individuals are more likely to acquire and are better able to evaluate information about corruption, leading them to consider exit options. I test this mechanism directly by analyzing whether higher-educated individuals are more likely to seek out information. I estimate two logistic regression models using questions from the GWP: whether or not an individual reads a newspaper weekly, and whether or not an individual has used media to “get news and information” in the past week.Footnote 18 I display these results in Appendix Table 11 and indeed find that holding a higher degree substantially increases the probability of either reading a newspaper or using media to acquire information, consistent wtih my argument.

Fig. 2
figure 2

Increase in predicted probability of potential emigration due to perceiving widespread corruption. Panel (1) refers to emigration intentions. Panel (2) refers to emigration plans. Panel (3) refers to emigration preparations. Error bars are 95 percent confidence intervals. Predicted probabilities based on Columns (3), (4), and (5) of Table 2, respectively

These results so far support the main argument of the paper: all people see emigration as an attractive response when they perceive corruption to be high. But individuals with more human capital, who are most likely to be informed about the extent of corruption and who are most internationally mobile, are especially likely to consider exit when they perceive high corruption. A recent wave of research demonstrates that in developing democracies and electoral autocracies, increasing education can lead to decreased domestic political engagement and disillusionment with politics (Croke et al. 2016; Agerberg 2019). My findings suggest that rather than just abstaining from political participation, highly educated people may instead see emigration as the political response that most efficiently addresses their own concerns in an environment of high corruption.

Corruption Perceptions and Destination Choice

One clear empirical implication of my argument is that individuals who perceive high corruption and emigrate in response should choose to move to countries with less corruption. In other words, people who are disillusioned with corruption should not just immigrate, but immigrate to low-corruption countries.

Fortunately, the GWP allows me to directly test this implication. When an individual responds that they plan to emigrate within the next year, they are also asked to which country they plan to emigrate. Not surprisingly, the most popular planned destination countries include the United States and Western European countries, but many also choose Gulf countries and an assortment of other developing countries. I match each respondent’s planned destination country to its score on V-Dem’s Political Corruption Index (Coppedge et al. 2020). This index is a summary measure of corruption’s pervasiveness in a given country and ranges from 0 to 1; higher values indicate higher corruption. I estimate a series of OLS models of corruption in a respondent’s planned destination country as a function of that respondent’s corruption perceptions. These models all include country\(\times\)year fixed effects, and the sample is limited only to respondents who plan on emigrating within the next year.

Of course, corruption-concerned individuals may choose specific destinations not because of their low corruption, but also because of their relatively higher wages. My argument implies that in addition to economic considerations, corruption-concerned emigrants specifically choose lower-corruption destination countries (as opposed to those less concerned with corruption). Yet because corruption and wage differentials are likely highly correlated, any empirical analysis must include a measure of relative wage differentials to account for this alternative explanation. In all models, I control for logged difference in real GDP per capita between the selected destination country and origin country. My expectation is that corruption perceptions will be associated with lower levels of corruption in planned destinations.

I present these results in Table 3. Column (1) shows the relationship between corruption perceptions and destination-country corruption only controlling for wage differentials, while Columns (2) and (3) add controls for income and education, respectively. The results illustrate that on average, emigration-planning individuals who perceive widespread corruption choose less corrupt destination countries than emigration-planning individuals living in the same country in the same year who do not believe corruption is widespread. This relationship holds even when controlling for household income and respondent education. These results provide further support for my core argument, validating the proposition that corruption-concerned individuals not only move, but systematically move toward relatively less corrupt destinations.Footnote 19

Table 3 Corruption perceptions and destination choice

Do Potential Emigrants Stop Using Voice?

Hirschman (1970, 1978) presents a seminal political economy theory in which exit and voice are competing responses to dissatisfaction with the political status quo. The fact that I observe potential emigration allows me to evaluate whether individuals preparing to exit preemptively disengage from voice-based alternatives. The empirical implication of my argument would be that people who decide to emigrate in response to corruption are likely to see voice as ultimately a weak or insufficient way of addressing their concerns.

Political participation measures in the GWP are unfortunately limited. Respondents in a wide sample of countries are asked whether or not they have confidence in the honesty of elections in their home country. Additionally, in some years (2009 and 2010), respondents in a selection of Asian and European countries are asked if they voted in the most recent elections. I use each of these variables to estimate the relationship between emigration intentions and political engagement.Footnote 20 Because data on voting is only available for a smaller set of countries and time periods, the analysis that follows is more speculative in nature.

In Table 4, I estimate a series of logistic regression models with country\(\times\)year fixed effects using these two variables. I include a range of control variables commonly associated with political behavior: education, age, gender, and income quintile. Column (1) analyzes the association between emigration intentions and confidence in elections, while Column (2) adds an interaction term with education. Column (3) analyzes the relationship between emigration intentions and voting in the most recent election, while Column (4) again adds an interaction term with education.

Table 4 Potential emigration and voice

The results indicate that potential emigrants are much more likely to lack confidence in the honesty of elections in their home country. On average, individuals who intend to emigrate are about 12 percentage points less likely to state they they are confident in elections (a decrease from 0.53 to 0.41). Yet this relationship is heavily dependent on education, as Column (2) illustrates. For highly educated individuals, intending to emigrate is associated with a 16 percentage point decrease in the probability of having confidence in the honestly of elections (compared to an 11 percentage point decline for lower-educated people).Footnote 21

Individuals who express an intention to emigrate are, on average, also less likely to report that they voted in the most recent elections. People who state that they intend to emigrate are about 5 percentage points less likely to have voted, on average (a decrease from 0.82 to 0.77). While the interaction term is statistically significant only at \(p<.1\), the negative relationship between emigration intentions and voting is larger for the higher-educated. For highly educated individuals, intending to emigrate is associated with a 6.4 percentage point decline in the probability of having voted (compared to a 4.5 percentage point decline for lower-educated individuals). Taken together, these results lend preliminary support to the idea that potential emigrants are more likely to have lost faith in their home country’s electoral process, and less likely to have exercised their right to vote. In other words, the decision to exit is associated with a fall in the confidence in, and use of, voice-based alternatives.

Discussion and Conclusion

Existing theoretical and empirical research on the political consequences of corruption broadly suggests that people use electoral institutions to punish corrupt politicians, provided that conditions on the ground are conducive to accountability. I instead argue that when citizens grow increasingly concerned about persistent and pervasive corruption, they are more likely to become disillusioned with voice-based redress and instead turn to emigration as an exit option. Emigration can prove more attractive to people who are most sensitive to corruption’s negative impacts than voting or protesting, both of which require collective action and often fail to effectively curb corruption. Highly educated people in particular, who are most informed about corruption and most pessimistic about the efficacy of voice, are most likely to use emigration as a private pathway to escape corruption’s negative impacts.

I use a wealth of globally representative, micro-level data on potential emigration and perceptions of corruption to validate my argument. When citizens perceive corruption to be widespread, they are more likely to intend to emigrate, to have made concrete plans to leave within the next year, and to have made active, costly preparations for their move. While corruption-concerned people see exit as an attractive response to corruption regardless of education, higher-educated people are especially likely to have begun the emigration process. Consistent with my argument, corruption-concerned people who plan to emigrate choose less corrupt destination countries than those who are unconcerned with corruption. Potential emigrants also sour on voice-based alternatives: they are less likely to be confident in their country’s electoral process and less likely to have recently voted.

Considering emigration as a response to concerns about corruption generates new explanations for mixed empirical findings to date. A number of studies find that voters often fail to hold corrupt politicians accountable. Meanwhile, I argue and find that the most well-informed individuals who have clear exit opportunities are more likely to emigrate. This dynamic suggests that people who are most attuned to corruption’s negative impacts leave rather than voting or protesting against corruption. In other words, if the most corruption-sensitive people emigrate, they leave behind an electorate that, all else equal, is less concerned with addressing corruption. This possibility has not been explored, as studies to date do not consider this exit option. I certainly do not argue that variables like information provision and competing political interests do not matter. Rather, the exit option further complicates the relationship between corruption and behavioral response because it falls outside of the domestic political realm as traditionally understood.

My argument and findings also suggest new directions to explore emigration’s long-term effects on political accountability. From one perspective, emigration (especially of the highly educated) could make mobilization of the masses on issues like corruption more difficult. Sellars (2019) argues that emigration can undermine domestic collective action capabilities to bring about change, because it raises the opportunity costs of political action and lowers confidence in the eventual success of mass political movements. In addition, given that more educated citizens are often leaders in political movements, their exit may be especially damaging to mobilization capacity (Morris and Staggenborg 2004). Sellars’ argument, combined with my findings, suggests a perverse potential feedback loop: corruption concerns push people to emigrate, which then undermines domestic collective action capabilities, allowing problems like corruption to persist with less challenge.

On the other hand, corruption-induced citizen exit could have positive economic and governance effects. Emigration generates financial flows in the form of remittances, which are larger and more stable than many other external sources of capital (World Bank 2016). A large literature suggests that remittances have positive economic effects, reducing poverty and allowing households to insure against negative shocks (e.g., Adams and Page 2005). Remittances can promote democratization under certain conditions and foment protests against autocratic regimes (Escriba-Folch et al. 2015, 2018). And crucially, a growing literature indicates that remittances decrease corruption by increasing electoral accountability and creating incentives for reform (Tyburski 2012, 2014).Footnote 22

Besides their financial remittances, those who emigrate as a result of corruption may have more direct political impacts on their countries of origin via social remittances. Ivlevs and King (2017) argue that migrants remit anti-corruption norms back to their countries of origin. They find, in the Balkan context, that having families members abroad reduces the likelihood of being asked for bribes by public officials and makes bribe-taking behavior less socially acceptable. Corruption-induced emigration may therefore foster accountability over time if it results in the transmission of anti-corruption norms.

Finally, while not the main focus of this article, an emerging literature argues that exit and voice are not mutually exclusive. Many emigrants continue to participate in the politics of their home countries after leaving (Guarnizo et al. 2003). A number of demographic, socioeconomic, and destination-country factors shape how emigrants remain political actors of relevance.Footnote 23 However, no existing research explores whether emigrants continue to act from abroad on specific political problems, like corruption, that pushed them to leave (see Chaudhary 2018). Two opposing lines of reasoning seem plausible: emigrants who left because of corruption disengage from politics due to their disenchantment with their home-country government (see, for instance, Guarnizo et al. 2003, on the case of Colombia), or those individuals still view corruption as a salient political problem worth addressing despite their exit, and engage politically across borders. This is an area ripe for future work.