State Dependency and the Limits of Middle Class Support for Democracy

Abstract

Scholars have long viewed the middle class as an agent of democratization. This article provides the first rigorous cross-national analysis of middle class regime preferences, systematically investigating the importance of an authoritarian state’s economic relationship with the middle class. Using detailed survey data on individual employment histories from 27 post-communist countries, I show that, under autocracy, state-sector careers diminish support for democracy, especially among middle class professionals. The results are robust to changes in the measurement of both the middle class and democracy support. I also show that neither selection nor response bias, redistributive preferences, communist socialization, or transition experiences can explain the results. The findings imply that a state-supported middle class may, in fact, delay democratization.

Keywords

democratization East European politics public opinion non-democratic regimes Russia/former Soviet Union

Introduction

Scholars have long viewed the middle class as an agent of democratization (Acemoglu & Robinson, 2006; Ansell & Samuels, 2014; Boix, 2003; Fukuyama, 2014; Lipset, 1959; Moore, 1966). There are, however, no cross-national studies on the political preferences of the middle classes in contemporary autocracies using multi-country survey data. Despite the centrality of the middle class to prominent theories of democratization, quantitative cross-national research on the subject lies primarily outside the field of political science.

Economists have examined the size, savings rates, and consumption patterns of the middle class as well as the impact of the middle class on economic growth globally. However, studies of the middle class in development economics have remained remarkably devoid of politics. Ravallion (2010) finds that 1.2 billion people in the developing world joined the global middle class from 1990 to 2005, using an income-based definition. He attributes cross-national variation in the expansion of the middle class to initial differences in poverty and rates of economic growth. Yet his view of this process is entirely apolitical and ignores the fact that regimes in China and Russia, where growth of the middle classes has been greatest, have explicitly invested in middle class formation (Chen, 2013; Remington, 2011). Banerjee and Duflo (2008), meanwhile, find that steady, salaried employment rather than an entrepreneurial spirit distinguishes the developing world’s middle classes from its poor. This too points to politics, since the main source of stable employment in developing nondemocracies is often the public sector.

Among existing cross-national studies of the middle class, only Easterly (2001) deals explicitly with politics. Easterly finds that a larger share of income for the middle class is associated with higher growth, more democracy and greater political stability. However, by treating the middle class as an aggregate, Easterly cannot establish whether it is middle class demand for democracy that explains this relationship. In short, the ability to make inferences about micro-level preferences and mechanisms based on these macro-level relationships is limited. Instead, using appropriate data, we should look for evidence that the implied relationships exist also at the individual level. Micro-level evidence of middle class demand for democracy would both increase our confidence that the apparent cross-national relationship between middle class size and democracy is not spurious and help to clarify otherwise vague causal mechanisms.

There are currently no individual-level analyses of these questions across multiple nondemocracies, despite new research opportunities and data that have emerged over the past 30 years. Instead, empirical research on the middle class and democratization has generally taken the form of single-country or comparative historical studies. Jones (1998) writes that as beneficiaries of state-planned development, East Asia’s new middle classes have played an ambiguous role in processes of democratic transformation (see also, Brown & Jones, 1995; Yang, 2007). In China, Chen (2013) finds that the new middle class, and especially employees of the state apparatus, are more supportive of the current regime and less supportive of democratization, while Tsai (2005) argues that the Communist Party’s clientelistic relationship with the private sector has limited its emergence as a coherent political force. Scholars of post-soviet politics have also expressed skepticism about the democratic credentials of the middle class and drawn attention to its relationship with the regime (Gontmakher & Ross, 2015; Maleva & Ovcharova, 2008).

One contribution of this literature has been to highlight the critical intervening role of the state, often overlooked in arguments about the impact of economic modernization on democratization (Bellin, 2002; Greene, 2010, p. 808; McMann, 2006). Since the collapse of communism, a number of scholars have emphasized the importance of new economic activity in the private sector to sustaining democratic coalitions in the region (Frye, 2003, 2006; Jackson et al., 2005; Mach & Jackson, 2006; Przeworski, 1991).¹ Still others point to the post-soviet middle classes’ continuing reliance on state support (Remington, 2011; Shevtsova, 2012) and have shown that their participation in pro-democratic protest is limited by structural dependence on the state (Rosenfeld, 2017). This article builds on these insights and links them to broader debates about democratization and authoritarian resilience. Empirically, it provides the first rigorous cross-national evidence on middle class regime preferences. I ask: Across the countries of eastern Europe and the former Soviet Union, is the middle class more likely to hold liberal democratic attitudes? In the autocracies, does dependence on the state diminish middle class support for democracy?

To test the argument, I use detailed information about labor market participation and political attitudes for a large cross-national sample of post-communist citizens in 29 countries of Central and Eastern Europe and the former Soviet Union. This approach ensures greater generalizability across different institutional contexts, including resource and non-resource states. Broadening the sample to include both democracies and nondemocracies helps to clarify why prior research, which has overwhelmingly focused on democratic settings, has failed to identify the sectoral patterns of middle class preferences over democracy that this article highlights. By contrast, a sample that includes multiple regime types reveals that autocratic states gain a constituency that is more hesitant about democracy through the provision of public sector jobs to middle class professionals.

My results confirm that the state middle class is less democratic than the private-sector middle class in the post-communist autocracies. By contrast, in the post-communist democracies, where I do not expect state employment to weaken support for democracy, indeed, no such relationship is evident. These results are robust to different definitions of the middle class and democracy support. I find that as the share of an individual’s career spent in the state sector since the collapse of communism increases, support for democracy decreases. The analysis makes clear that this finding is unlikely to be an artifact of age or Soviet socialization. I also consider several other alternative mechanisms by which state employment might matter for the middle classes’ democratic preferences (or appear to), but find weak support for each of them. These include: blaming democracy for the economic dislocation of the 1990s, different preferences over redistribution or sensitivity to inequality, communist education and experience. The analysis also addresses social desirability bias as well as bias resulting from selection. Across a variety of tests, I find that employment experiences affect the political values of state workers and that observed differences between state and non-state workers are unlikely to be due solely to sorting into career paths based on prior characteristics.

The upshot of this article is that control over public employment benefits autocrats and can supply them with a middle class that is economically self-interested in the maintenance of an autocratic system. Its findings suggest that demand for democracy will remain limited where middle class groups fail to gain economic independence from an authoritarian state.

The Argument

Throughout, I use the term middle class to describe individuals who are distinguished by their resources, occupational status, human and social capital. The middle class defined in this article most closely resembles the middle class of educated professionals in modernization theory (Lipset, 1959) and its values-based variants (Welzel & Inglehart, 2008) and shares with Collier and Collier’s (2002) “middle sector” an emphasis on members of a broad range of occupational groups between the working class and economic elite. At the same time, it is broader than Moore’s (1966) capital-owning bourgeoisie and empirically has above-median income like the middle class of elite competition theories (Ansell & Samuels, 2008). I do not use the term in a normative sense to refer to an innovative or modernizing class or to refer to a class defined by its mindset and values. I also do not use the term to describe a unified class actor. In fact, I show clear evidence that the state middle class holds political preferences which differ from those of the private-sector middle class, despite similar incomes, educations, and occupations.

Canonical theories expect the middle class to advocate for institutionalized political responsiveness. Modernization theory, in particular, predicts that a growing middle class of educated white-collar professionals will be the carrier of democratic values and will drive democratization forward. In contrast, this article argues that failure of the middle class to gain economic autonomy from the state actually stymies support for democracy. More specifically, public sector positions provide incentives for the middle class to prefer autocracy under certain labor market conditions.

Democratization is fraught with uncertainty, not unlike other forms of regime change. Fear of uncertainty is heightened when the public sector is extensive, and has crowded out private-sector alternatives (Behar & Mok, 2015). Indeed, across the post-communist region state control in sectors like education, health, energy, transport, oil and gas trends monopsonistic. Consequently, some occupations requiring specific training and credentials are located almost exclusively in the public sector. Second, and relatedly, many public-sector professions are highly specialized and require skills, networks, and know-how whose value is highly specific. Whereas individualized, project-based work and flexible forms of employment are more abundant in the private sector, public-sector organizations and enterprises tend to be larger with idiosyncratic organizational procedures, products, tasks, and technologies, which “bind work experience to a specific context” (Marx, 2011, p. 8). Public administration is a case in point, but other occupations in the public sector are similar. Third, under autocracy, government jobs typically provide opportunities to earn informal rents. Change in political control of the state threatens these benefits.

Further, successful democratic transitions often involve simultaneous political and economic reform. In many contexts including the one examined here, democratization raises the specter of state retrenchment. Not only in the post-communist countries but around the world, the downsizing of public payrolls has frequently accompanied or soon followed democratization. Fears of state retrenchment, particularly in less competitive state enterprises, may shape the state middle class’s attitude toward democratization. Yet even where neoliberal reforms are not expected to accompany the triumph of democratic forces, middle class groups dependent on the state may be reticent about democracy for several reasons.

While all regime change threatens status quo rules about whose interests are represented in policy, state-dependent groups can also expect democratization, under an optimistic scenario, to broaden representation and accountability, while increasing transparency and reducing opportunities to earn informal rents. If democracy provides a more meritocratic basis for hiring, promotion, and retention in the government sector, state employees whose primary asset has been loyalty or who are less competitive may see their prospects diminished. Under a less optimistic scenario, structural conditions of corruption and a weakly professionalized public sector are unlikely to produce a democracy capable of delivering on programmatic policies (Kitschelt et al., 1999). Hence, the state middle class in a country with a corrupt public sector and state apparatus that is only weakly guided by law, should have little expectation that democracy will bring strong public goods provision. Instead weak parties are more likely to form on clientelistic bases, packing the public sector with their own loyalists—and plausibly disadvantaging those associated with the preceding regime. Again, for those who retain their jobs, rising attention to the rule of law under democracy promises to staunch the flow of informal rents.

For all of these reasons, the cost of supporting democratization can be higher for the middle class in state-dependent sectors than for those whose economic opportunity and life chances are independent of the state. Not all of these considerations apply equally to every type of public employment. But together I argue that they generate a consistent set of incentives which makes the state middle class more diffident about democratization than the private-sector middle class in the post-communist region. Meanwhile, if the argument I have laid out is correct, the negative effect of state employment on democracy support should be present in autocracies but not in democracies. Dependence on a democratic regime should not make public sector employees less democratic. If anything, we would expect the opposite. All the same forces—long-run dependence on the state and fear of being fired—plus an interest in maintaining a nonpolitical basis for promotion and retention² suggest that employees under a democratic regime should be at least as likely to prefer democracy.

Empirically, regime preferences, support for the status quo, and support for a particular incumbent are often intertwined. A middle class that is satisfied with the status quo and loyal to the regime will naturally be more reticent about democracy. However, these attitudes need not and, indeed, do not always go together. Even where the state middle class expresses grievances with the status quo (as in Russia during the 2011–2012 protests) or is generally unsupportive of the regime (as in Ukraine following the Orange Revolution), for the reasons outlined above it may not prefer democracy. This study focuses on attitudes toward democracy and only secondarily on support for the status quo or any particular autocrat.

Finally, the basic contours of this argument imply that democracy requires democrats. Recent research in the region finds that sector of employment and social class, as well as expressed support for democracy, predict the politically consequential protest behavior that increasingly influences regime trajectories (Beissinger, 2013; Rosenfeld, 2017; Smyth et al., 2013). The state-employed post-communist middle classes are not merely more fearful of expressing support for democracy. They are also less likely to take to the streets during episodes of anti-regime contention. In this light, studying individual preferences is important for understanding the sources of authoritarian resilience. I take up the question of how this argument is likely to travel beyond the post-communist region in the conclusion.

Empirical Strategy, Concepts, and Measures

The European Bank for Reconstruction and Development’s Life in Transition Survey (LiTS) covers 27 countries in Central and Eastern Europe and the former Soviet Union and approximately 27,000 individuals.³ This study focuses on Wave I, which was conducted in 2006 and includes a detailed employment history for all respondents over the entire post-communist period.⁴ For each year since 1989, respondents are asked to recall their occupation, industry, and the ownership of their employer—whether state, private, or foreign.

In addition, the survey includes common demographic questions as well as measures of household consumption. It also asks about economic satisfaction, income inequality, perceived social mobility over the post-communist period, and subjective position in the country’s income distribution. These allow me to compare the effect of state employment on individual regime preferences to a variety of competing explanations.

The 2006 LiTS is unique insofar as it combines detailed data on an individual’s employment history with measures of political preferences. The fact that the survey asks identical questions across a large number of democracies (N = 18) and nondemocracies (N = 9) which share important developmental legacies of communism provides additional analytical leverage (Beissinger & Kotkin, 2014; Pop-Eleches, 2007; Pop-Eleches & Tucker, 2017).⁵ Because these countries share a common history of universal state employment, we can be more confident when we compare across the democratic and nondemocratic cases that inheritances from state employment institutions of the communist era do not confound the results. A sample that includes both democracies and nondemocracies also helps to clarify why prior research has missed the contingent relationship between an expanding middle class and the emergence of democratic values that this article demonstrates.

Democracy Support

The analysis in this article seeks to explain variation in patterns of democracy support. To measure democracy support, I use eight survey items. The first asks: “With which of the following statements do you agree most: (1) Democracy is preferable to any other form of political system; (2) For people like me, it does not matter whether a government is democratic or authoritarian; or (3) Under some circumstances, an authoritarian government may be preferable to a democratic one.” This item captures both the salience of regime type and the preference for democracy. It probes whether a respondent views the type of government that they live under as relevant and whether they prefer democracy, or see authoritarianism as sometimes preferable.

The remaining items ask to what extent respondents think various democratic institutions are important for their country: (1) free and fair elections; (2) freedom of speech; (3) an independent press; (4) courts that defend individual rights against abuse by the state; (5) equality before the law; (6) minority rights; and (7) a strong political opposition. Responses to these items are coded on a 5-point Likert scale from strongly disagree to strongly agree.

The Cronbach’s α for standardized scores of these eight items is .86, indicating strong internal consistency (Nunnally, 1978). In Supplemental Appendix A, I report the results of an exploratory factor analysis, which supports the decision to treat these indicators as constitutive of a single underlying concept. While imperfect like all survey questions, together these items have several advantages. In particular, the second set of items provides a concrete and multifaceted assessment of respondents’ attitudes toward democracy, which does not rely on the meaning respondents assign to the word “democracy.” Meanwhile, the first item allows me to distinguish respondents who offer interviewers their views on various democratic institutions, but feel that for people like themselves the system of government is largely irrelevant.

Since democracy support requires both the belief that regime type matters and a preference for democratic institutions, my primary dependent variable is a product of respondents’ scores on each of these two attributes. I also disaggregate these items and use several alternative measures described below (Munck & Verkuilen, 2002, p. 24). Based on the first question, I create an indicator for the belief that democracy is preferable. I code as zero those who believe authoritarianism is sometimes necessary and those who believe regime type does not matter for people like themselves.

For the second component, I code an indicator for those who “agree strongly” that the seven democratic institutions mentioned above are important. I also relax this measure to include those who “agree somewhat” with the importance of democratic institutions and show that this choice makes little difference. It is worth noting, however, that the more lenient measure of democracy support is less highly correlated with democratic outcomes. This suggests that accounting for preference intensity as I do here results in a measure with greater predictive validity.⁶

The analysis in this article focuses on the group I call “liberal democrats” who express strong, consistent support for liberal political institutions as well as a preference for democracy over other political systems.⁷ This group of respondents has three important characteristics. First, the intensity of their democracy support is greater than that of other respondents. Second, their responses indicate that they consider regime characteristics salient; that is, they matter to the respondent personally. Third, their answers demonstrate ideological consistency and coherence across dimensions.

In addition to my primary measure of democracy support, I also test several alternative conceptions. I examine separately the items on institutions and salience. I also investigate more and less aggregated measures of the preference for democratic institutions. Indeed, the pattern is broadly similar whether democracy support is measured using a single question on the importance of a political opposition or a standard index of the democratic institutions items, as I report in the section on robustness.

The Middle Class

Table 1 gives my operational definition of middle class. Together with state employment, this measure of the middle class is one of two key independent variables in my analysis. Again, this measure supposes that middle class status is attained as individuals gain social and professional status markers: higher education and non-manual, white-collar occupations. I also include university students. For those who are retired or temporarily not working, I code the respondent’s most recent occupation. My definition of the middle class thus includes a modestly-sized cohort who are not economically active but last possessed a middle class job. To be sure that this choice is justified, I test whether my results are sensitive to the inclusion of retirees. I find that they are not.

Table 1.

Operational Definition of Middle Class.

Middle class	=	Education	+	Occupation^†
		Incomplete higher education		Technicians and associate professionals
		University degree		Professionals
		Postgraduate degree		Managers

†

The occupations included correspond to the ILO’s ISCO-08 categories 1 to 3.

I define as non-middle class all respondents who have lower educational and/or occupational attainment. Though the population can be stratified into elites, the middle class, and the working/lower class, this study assumes that only the middle and working/lower classes are represented in survey samples. Elites comprise no more than a fraction of the population in developing countries and are least likely to be accessible to survey researchers. Though owners and high-level managers of enterprises in the commercial, financial, and industrial sectors are captured by my measure, the bourgeoisie, as typically defined (e.g., Collier & Collier, 2002; Moore, 1966), makes up only a small part of the middle class examined in this article.

Last, to be sure that my results do not hinge on defining the middle class in terms of education and occupation, I also code the middle class according to Ravallion’s (2010) income-based criterion.

State Employment

The primary state employment variable that I use in the analysis takes a value of one if an individual has been employed in the state sector since the collapse of communism and zero if not. Across the post-communist countries in my sample, the share of respondents who have been state employees at any time since 1991 is 42%. This figure is even higher, 59%, among the middle class, reflecting the important role that public employment has played for professionals. In the nondemocracies with less reformed economies, the figure is higher still: 67%. These figures capture the extent to which state employment has remained a way of life for a large segment of the post-Soviet middle classes since the collapse of communism.

Taking advantage of the pseudo-panel nature of my employment data, I also code three additional variables in order to probe why and how state employment affects political preferences. Since only current employees are vulnerable to being fired for expressing “politically incorrect” views, including a dummy for current state employment allows me to test whether the apparent effect of a state career could be due to biased survey responses. In 2006, at the time of the survey, the overall share of state employees among the employed was about a third in the democracies and nearly half in the nondemocracies. If reporting bias is a problem, we would expect that current employment would absorb much of the effect of the other state employment variables and that its coefficient would be negative.

In addition to the binary indicators for state employment, I also code continuous variables for years of employment and career share in the state sector during the post-communist period. Years of post-independence state employment vary from zero to 15, while post-independence state career share ranges between zero and one. Public sector career share captures well an individual’s demonstrated access to private sector alternatives. It also reflects the extent to which an individual’s professional networks and know-how are tied up with the regime. I interpret this variable, therefore, as long-term dependence on the state. I expect state career share to be inversely related to democracy support.

After controlling for state career share, also including years of state employment allows me to investigate differential attrition as a source of selection bias and socialization effects. An account based on differential attrition implies that nondemocrats last longer in state employment under a nondemocratic regime. In a regression with both state career share and years of state employment, we would expect years in the state sector to be negative and significant, if differential attrition is driving the results. Using years of state employment, I can also test for evidence of workplace-based socialization. A significant negative coefficient on years of state employment would be consistent with socialization, understood as a cumulative deepening of anti-democratic attitudes.

Of those in my dataset who are middle class and employed by the state, about half are budget sector workers. Some 42% are in education, while 11% are in public health.⁸ Another 12% are in public administration, including the military and security apparatuses. In total, approximately two-thirds of the state-employed post-communist middle class works in these three industries. The remainder works in various state-owned enterprises. Among those who have worked in the public sector since 1991, the average length of post-communist employment is 8.3 years. Across the full sample, approximately 10% have worked in the state sector continuously since the collapse of communism, from 1991 until 2006.⁹

In the analysis that follows, I use variation in the type and timing of state careers as well as their duration to gain leverage on the question of whether state employment itself or merely selection (self-sorting) accounts for the apparent effect of a public sector career on middle class support for democracy. In particular, I exploit the fact that medical professionals are plausibly less motivated by political considerations than those working directly for the state administration. To preview, I find consistently that the sorting of less democratic types into state employment accounts poorly for patterns in the data that are better explained by state employment subsequently shaping regime preferences. These results are thus similar to Gans-Morse et al. (2017), who conclude that corruption in Russia results from the transformation of bureaucrats’ behavior and attitudes after entering the civil service, rather than through a process of corrupt self-selection.

Controls

I also control for a variety of alternative explanations. First, I include age and age squared to capture generational differences in attitudes toward democracy and differences in communist and post-communist socialization. Since state employees tend to be older, controlling for age ensures that state employment is not confounded by generational effects. Second, I include gender with males coded as one. Like age, gender potentially confounds the effect of state employment, since the state sector workforce in many countries is disproportionately female. If women’s regime preferences differ systematically from those of men, failure to include gender as a control could bias the results.

Third, I use a logged measure of household expenditures that has been equivalized per adult as a proxy for income. In regions where wages are often paid informally and not reported to tax authorities, many scholars prefer this measure to current income, which is often underreported. I expect this proxy for income to be positively related to support for democracy in line with Ansell and Samuels (2014) and contrary to redistributive theories. If most of the effect of being middle class runs through occupation and education and the additional explanatory power of income is negligible, its effect should be substantively small.

Fourth, I include respondents’ subjective evaluation of their location in the country’s income distribution (subjective income decile). Fifth, I control for attitudes toward the free market, since studies of democracy support in the region have often found a close connection between economic and political attitudes. Sixth, I use a measure of trust in the president and two items on economic and overall life satisfaction to capture support for the autocrat and the status quo, respectively. Controlling separately for unemployment has no effect on the results, so I interpret the findings for non-state employees as indicative of preferences in the private sector. The full specifications include all controls as well as country fixed effects.¹⁰

In a final section on other alternative mechanisms, I also control for economic experiences since the collapse of communism, preferences over redistribution, and communist socialization.

Empirical Results

I begin by examining whether being middle class is associated with greater support for liberal democracy. I use figures to intuitively display predicted probabilities of democracy support, separately for the democratic and nondemocratic post-communist countries. Table A2 in the Supplemental Appendix reports the logistic regression coefficients from all models subsequently discussed. The first set of empirical results, for the democracies, is summarized in Figure 1.¹¹

Figure 1.

Middle class more supportive of democracy regardless of sector in post-communist democracies. The left panel of this figure shows the predicted probability of being a “democrat” given that a respondent is middle class (squares) or non-middle class (circles). The second panel shows the predicted probability of being a “democrat” for middle class and non-middle class respondents by sector of employment. The vertical bars (hash marks) are 95% (90%) confidence intervals.

The results conform strongly to the predictions of modernization theory, reproducing at the individual level what the classical literature on the middle class leads us to expect: that the middle class is more supportive of democracy than the working class. Whether the middle class is employed in the state or private sector is unrelated to democracy support. Both groups are equally democratic. Tentatively transposing these relationships to the macro level, we would conclude with Easterly (2001) that a large middle class is good for democracy. The second set of results in Figure 2, however, implies clearly that unconditional statements about the democratic virtues of the middle class are misleading for autocracies.

Figure 2.

Middle class in post-communist nondemocracies more likely to support democracy, only if employed outside the public sector. The left panel of this figure shows the predicted probability of being a “democrat” given that a respondent is middle class (squares) or non-middle class (circles). The second panel shows the predicted probability of being a “democrat” for middle class and non-middle class respondents by sector of employment. The vertical bars (hash marks) are 95% (90%) confidence intervals.

The left panel of Figure 2 shows that although being middle class is associated with more liberal regime preferences, the average effect of being middle class is small. To quantify the size of the effect, I calculate the predicted probability of democracy support for the middle class (squares) versus others (circles). While the difference in the predicted probability of democracy support for the middle and non-middle class is significant at the .001 level, it is only about 4 percentage points (with a 95% confidence interval of [.020, .063]).

The right panel of Figure 2 tests the proposition that support for democracy will be attenuated among a middle class whose economic opportunities and life chances are tied closely to a nondemocratic state. It shows that, under authoritarianism, the association between the middle class and liberal regime preferences is virtually zero for the state-employed middle class (solid square), whereas it is sizable and significant for those members of the middle class who are not employed by the state (open square).¹² In fact, even at the .1 level, we cannot reject the null hypothesis that the actual difference in democracy support between the public sector middle class and non-middle class is zero. Non-middle class respondents and state workers have about the same (low) probability of being democrats, just above 10%.

At the same time, the association between being middle class and support for a liberal democratic regime is considerably stronger for those who are not employed by the state. Members of the middle class outside the public sector are about 8 percentage points more likely than the non-middle class to prefer liberal democracy. This difference is both statistically distinguishable from zero at the .001 level and substantively meaningful, as it represents more than a 70% increase in the probability of democracy support.¹³ These results imply that, under autocracy, middle class support for democracy is conditioned by the middle class’s relationship to the state.¹⁴

Robustness Checks

These findings hold up to a number of changes in the scope of the middle class and the definition of democracy support. In this section, I also demonstrate that the state middle class’s reticence about democracy is not simply reticence toward the free market nor is it synonymous with support for the status quo or a particular autocrat. I take up each of these tests in turn. First, I relaxed the educational criteria I used to define the middle class by including those with specialized post-secondary training in addition to those with university and post-graduate degrees (Supplemental Appendix Table A3). This change makes very little difference. Second, I probed potential differences in the effect of education on the democratic attitudes of the middle class, distinguishing between the middle class educated in the Soviet period and the middle class that gained higher education after the collapse of communism (See Supplemental Appendix C.1). Given the ideological and explicitly anti-democratic character of the Soviet educational system and changes to the content and quality of education in the post-Soviet period, it makes sense to further investigate whether communist education could be responsible for the antidemocratic attitudes of the middle class. The results imply that low support for democracy among the state middle class is not merely a function of Soviet education. Whether educated under communism or the product of post-Soviet higher education, the state middle class’s attitudes toward democracy are statistically indistinguishable from those of the working class, in keeping with the main results.

I next add a series of controls that further probe the robustness of my middle class measure. While the main results reflect the net effect of middle class status without including a measure of income, I now introduce a control for logged equivalized household expenditures per adult (Supplemental Appendix Table A3).¹⁵ The coefficient on expenditures is positive and significant. However, its substantive effect is small. Moving from the 25th to the 75th percentile in expenditures increases the likelihood of democracy support by only 2.5 percentage points.¹⁶ Further, the inclusion of expenditures has virtually no effect on the coefficient for middle class, defined according to educational and occupational criteria. These results imply that (1) the independent effect of current income, above and beyond the effect of middle class status that runs through human and social capital, is small; and (2) it is unlikely that these results can be explained by a wage gap between the public and private sectors. Moreover, if we substitute Ravallion’s (2010) income-based definition of the middle class for our own and interact it with state employment in a model with all of the same controls, the pattern is very similar (Supplemental Appendix Table A3). The results remain substantively the same with the inclusion of all controls and country fixed effects.¹⁷

Next, I tested the robustness of my dependent variable. To be sure that the findings are not contingent on my coding of democracy support, I first relax the definition to also include respondents who “agree somewhat” that liberal institutions are important for their country. Though the coefficient on state employment is close to zero, the interaction between state employment and middle class remains large, statistically significant, and negative (Supplemental Appendix Table A5).¹⁸ The results from this model, displayed in Supplemental Appendix Figure A6, are very similar to those reported above and imply that the state middle class is 25% less supportive of democracy than the private-sector middle class (p < .05).

I also examined several other alternative measures of democracy support (Supplemental Appendix Table A5). The results of linear models in which democracy support is measured alternately using a conventional averaged index of the seven democratic institutions items and the belief that a political opposition is essential imply that the state middle class is about 5% to 10% less supportive of democracy than the private-sector middle class, though these differences are imprecisely estimated. I further find that public sector employees are both more likely to prefer authoritarianism and to view the type of political regime governing their country as personally salient. Substantively, these results suggest that the state middle class is about 25% more likely to support authoritarianism than the private-sector middle class. The one departure in these findings is that state employees’ stronger preference for authoritarianism and conviction that regime type matters does not appear to be conditional on class.¹⁹ In each case, the results of these additional tests lean in the expected direction, though they are not all statistically significant at conventional levels. Comparing across them, it is interesting to note that the more demanding the definition of democracy support, the clearer are the differences between the state and private-sector middle classes. To summarize, then, the contrast between state and private sector-middle class largely conforms to the theorized pattern across all of the alternative specifications of the dependent variable, though the results should be interpreted with some caution.

Last, I show that my measure of democratic attitudes is not simply picking up support for a market economy, satisfaction with the status quo, or loyalty to a particular autocrat (Supplemental Appendix Table A6). First, I added a variable that measures the preference for free markets, as much of the literature on democratic attitudes after the collapse of communism has found a connection between attitudes toward political democracy and a market economy (e.g., Duch, 1993; Frye, 2003; Gibson, 1996; Miller et al., 1994).²⁰ The main results are unchanged. While attitudes toward the market are positively and significantly associated with support for democracy, all other coefficients remain very similar (compare Table A2 column 1 and Table A6 column 1 in the Supplemental Appendix). These results suggest that democracy and a market economy are intertwined in the minds of post-Soviet citizens but are not synonymous.²¹ Were views toward democracy and economic liberalism basically the same, the preference for free markets would account for more of the variation in the model.²² Second, I demonstrate that reticence about democracy is not merely a function of support for the incumbent autocrat or satisfaction with the status quo. The inclusion of these additional controls does not change the main results. In short, then, each of these tests again confirms that state employment has a substantively important independent effect on middle class support for democracy.

Alternative Explanations: Selection

Having shown a robust association between state careers and regime preferences, the following sections endeavor within the natural limitations of observational data to establish that it is causal. A key challenge is to untangle the effect of a state career itself from the factors that lead individuals to sort themselves into different career paths. In econometric terms, correctly identifying the effect of career choices on support for democracy requires a strategy to tackle reverse causation and selection bias (i.e., the influence of an unmeasured variable on both career choice and the outcome of interest). I next describe these problems substantively in the context of my analysis and outline the strategy I use to address them.

First, if people with authoritarian values seek out state employment, the two variables will be related, but not for the reasons I cite. The case in which people who value freedom and liberal democracy gravitate to the private sector poses a similar problem. In both cases, democratic values exert a simultaneous effect on employment choices. This effect would bias our estimates.

Similarly, our estimates would be biased if some selection criteria affect both employment choices and attitudes toward democracy. For example, if being unsuccessful in the new market economy leads some to both dislike democracy and seek the security of public employment, again our estimates will be biased. This latent character type (being a post-Soviet “loser”) would be an omitted variable. Aversion to risk, which might lead to a preference for public employment and a reluctance to embrace democratization would be another example. Insofar as these characteristics are a function of age, gender, or are proxied by low subjective well-being (or low social mobility), this problem is minimized by the controls. Nevertheless, I consider this issue in greater depth below.

The usual caveats regarding cross-sectional survey data apply. Ideally, we would have a measure of pre-employment regime preferences and could investigate directly whether regime preferences or their correlates are themselves drivers of a career in the state sector. However, comparable panel data in which respondents are reinterviewed over a period of years simply do not exist for such a large number of countries. The pseudo-panel nature of the LiTS employment data allows me to address these issues more fully than is typically possible with cross-sectional data, making the EBRD survey a very good testing ground for my argument and its alternatives. Not only does the LiTS survey tell us whether a respondent works for the state in her current occupation, it tells us how her career has progressed since 1989, just before the collapse of communism. As such, I can identify those respondents who have spent more and less time in the state sector and those who have moved into and out of it. The survey also further divides state employees by type.

I exploit this variation both within the state sector and across time to test the plausibility of competing arguments and speak to the question of causal mechanisms. Subsequent sections show that neither selection nor other conceivable alternatives explain my results. Instead, the analysis implies that, where private sector alternatives are lacking, authoritarians benefit from the presence of a large public sector.

Using occupational variation to test for selection

I begin this analysis of mechanisms by examining differences across occupational groupings within the state sector. Some serve the state directly, working within the state administration at various levels. Others work in education and public health and are paid out of the state budget. In this section, I exploit the fact that some public employees are more likely than others to choose their careers on the basis of preexisting political values.

Specifically, I expect that civil servants, persons in the military, and those employed in various security apparatuses are more likely to be politically motivated in their choice of career than are medical professionals. The fact that very few medical professionals in the post-Soviet countries make a career exclusively in the private sector further supports my identifying assumption, since those who wish to practice medicine are virtually required to do so within the state system.²³ If the negative effect of state employment shown in Figure 2 holds among medical professionals, this would support the idea that state employment does, in fact, have an effect on regime preferences that is independent of selection.

Figure 3 shows the results of this analysis. The figure displays the predicted probabilities of support for liberal democracy across all four major occupational groupings within the state sector: civil service, military, and security apparatuses; education; health; and those who work for the state in some other capacity (the largest group being managers and professionals in state-owned enterprises). For comparison, the first bar gives the predicted probability of democracy support among the non-state middle class (open circle). The probability of democracy support among the non-state middle class is higher than the probability of democracy support among all occupational groups within the state sector. For education and other state sector occupations these differences are significant at the .01 level, while for those who serve the state and for health workers these difference are less precisely estimated and significant only at the .1 level.²⁴

Figure 3.

State employment has a negative effect on support for democracy across all occupational categories. This figure compares the predicted probability of support for liberal democracy across four occupational groupings within the state sector: civil servants, military, and security apparatuses; educators; health workers; and those who work for the state in some other capacity (the largest group being manufacturing in state-owned enterprises). The predicted probability of democracy support among the non-state middle class is significantly higher and shown for comparison (open circle). The vertical bars (hash marks) are 95% (90%) confidence intervals.

Most importantly, no group within the state sector is significantly more likely than any other to prefer a liberal democratic regime. For selection to explain these results, medical professionals and teachers would have to be equally motivated by political considerations as civil servants and police in choosing state careers. This is not to say that selection plays no role in the choice of careers; however, these results suggest that state employment has an independent effect, and that this effect, rather than selection alone, better accounts for empirical patterns in the data. In subsequent sections, I present additional evidence that neither selection nor reverse causation can convincingly explain my results.

Using career trajectories to test alternative mechanisms

I next test several different mechanisms by which state employment could lower the probability of democracy support, exploiting variation over time in respondents’ careers. Whereas typical cross-sectional data only provide evidence on current employment, here I am able to juxtapose different career trajectories in order to better understand the relative importance of selection effects, immediate material incentives, long-term dependence on the regime, and socialization which deepens over time. This analysis also helps to address the concern that social desirability bias could be driving the results.

To sort out the mechanisms that link state employment with political preferences, I regress democracy support on a dummy for current state employment (at the time of the survey in 2006), state career share, years in the state sector, the variable for middle class, and the controls. The results presented in Figures 4 and 5 correspond with column 4 of Supplemental Appendix Table A2. I first consider the effect of current state employment. More than half of those who had spent their entire careers in the state sector were still employed by the state at the time of the survey. As such, this variable should absorb much of the unique effect of being a public-sector “lifer,” if such an effect exists.

Figure 4.

As post-Soviet career share in state employment increases, support for democracy decreases. (a) The left panel of this figure plots the predicted probability of being a “democrat” given the share of ones career spent in the state sector since 1991, controlling for class, gender, age, and age squared. (b) The right panel uses simulation to quantify the size of the difference in the predicted probability of democracy support for a given career share (x-axis) versus spending one’s entire career in the state sector. The vertical bars are 95% confidence intervals.

Figure 5.

The effect of additional years of state employment is insignificant, all else equal. (a) The left panel of this figure plots the effect of additional years of state employment on the predicted probability of democracy support based on a logistic regression that also includes state career share, class, gender, age, and age squared. (b) The right panel shows that the effect of additional years of state employment is insignificant, a finding that is inconsistent with both differential attrition out of the state sector and cumulative, long-term socialization.

Including the variable for current state employment also allows me to test whether response bias can plausibly explain the effect of a state career. If fear of being fired for expressing liberal views is driving stated democratic preferences, then we would expect current employment to be of principal importance in these regressions.²⁵

In fact, the coefficient on current state employment is negative as expected, but does not always reach conventional levels of significance (Supplemental Appendix Table A2, column 4). This implies that the association between state employment and the preference for an illiberal regime is not principally an artifact of social desirability bias. If fear of being fired were of primary importance, we would expect current employment to have greater substantive and statistical significance. Instead, current state employment explains little of the remaining variation after controlling for long-run dependence on the state.²⁶

The results in Figure 4 suggest instead that long-term dependence on the state has the most significant effect on the political values of state workers in the post-communist nondemocracies. To proxy long-term dependence on the state, I use the share of an individual’s post-Soviet career spent in the public sector. This variable captures well an individual’s demonstrated access to private sector alternatives. While those with a state career share of less than 1 but greater than 0 have also been employed outside the state, those who have spent their entire career in the state sector have no demonstrated ability to access private sector alternatives. I expect those with larger career shares in the public sector, who likely possess the least access to private sector alternatives, to be the least democratic, all else equal.

Consistent with this hypothesis, Panel 4(a) shows that as the share of one’s career spent in the state sector increases, support for democracy decreases. Panel 4(b) quantifies the size of the difference in the predicted probability of democracy support for those who have spent between 0% and 90% of their career in the state sector versus spending one’s entire career in the state. Substantively, the probability of preferring a liberal democratic regime if you have never been employed by the state is about 7 percentage points greater than if you have only had state employment (far left point and vertical bar). As the right panel of the figure shows, these differences are all statistically distinguishable from zero at the .05 level.

This pattern of results further helps to rule out an alternative account based on selection. Specifically, the downward sloping line in Figure 4(a) is inconsistent with a selection story that emphasizes initial sorting by democratic preference type. If that were the case, we would not expect to see any additional change after the initial sorting took place. Supplemental Appendix D.1 offers further evidence against political self-selection by exploiting regime transitions in Ukraine and Georgia.

Yet we might still be concerned that those who are more democratic leave state employment sooner than those who are more authoritarian (i.e., that differential attrition could explain these results). For example, imagine two individuals who have both spent their entire careers in the state sector. Now imagine that one has spent 15 years employed in the state sector, while the other has spent only 1 year. An account based on differential attrition would predict that the person who has been employed longer by the state would be less likely to be a democrat than the person who has been employed for only 1 year. Evidence to the contrary is presented in Figure 5. In fact, Panel 5(b) shows that additional years of employment are not significantly related to the sorting of democrats out of the public sector.

Finally, the insignificant effect of additional years of state employment, evident in Figure 5 also casts doubt on alternative explanations focused on cumulative, long-term socialization or indoctrination through repeated exposure (e.g., Pop-Eleches & Tucker, 2017). Only socialization that occurs very quickly, does not deepen appreciably over time, and is relatively sticky would be consistent with this evidence.²⁷

In sum, the findings suggest that neither pure selection of less democratic types into state employment nor certain versions of longterm socialization likely account for these findings. That said, separating employment trajectories from factors like age, skills, and education is a difficult task. Although we cannot entirely rule out these alternatives given the nature of the data, the fact that several pieces of evidence all point to a similar conclusion lends credence to the argument that a large public sector limits support for democracy, especially among the middle class that has long been expected to champion democratization.

Other Alternative Mechanisms

Economic performance

Before concluding, I consider three substantive alternatives, but find little evidence to support them. The first alternative explanation is that state workers have less favorable views of democracy because they are less satisfied with their country’s economic performance since independence and are more likely to associate economic insecurity with the transition to democracy. Indeed state workers, along with women and pensioners, are often characterized as losers in the post-communist transition, while private sector workers, along with the young and well-educated, are often viewed as winners. Public sector wage arrears were widespread in the former Soviet Union during the 1990s. Official salaries remain low for many teachers, doctors, and state sector professionals to this day (see e.g., Richter, 2006; Sharunina, 2013). If the state middle class blames democracy for downward social mobility and loss of the social status that was associated with their occupations during the Soviet period, this, rather than state dependence, could explain their attitudes toward democracy.

The analyses reported in Supplemental Appendix D.2, however, suggest that negative economic experiences since the collapse of communism do not explain the state middle class’s lagging support for democracy. While post-communist citizens may have once blamed democracy for declining standards of living, today they increasingly fault their nondemocratic governments for the failure to improve conditions. Indeed, I find that negative assessments of economic performance and the perception of downward social mobility are actually associated with less support for the status quo—that is, with more support for democracy.²⁸ The estimated effect of public employment for the middle class, meanwhile, remains surprisingly stable with the addition of these economic controls, suggesting that negative economic experience cannot account for lower support for democracy among the state middle class.

As Tucker et al. (2002, p. 559) note, ascribing winner/loser status on the basis of demographic characteristics may not be consistent with individual perceptions. In fact, perceived social mobility is exactly equal for the state and non-state middle classes.²⁹

Redistributive preferences

A second set of alternative explanations emphasizes the redistributive consequences of democratization. Two redistributive accounts are plausible. First, political economy models of democratization imply that the farther an individual’s income is below the median, the greater is her demand for redistribution. These theories further expect redistribution to be higher under democracy. It thus follows that those with below-median incomes should be more likely to support it. Perhaps, the state middle class has incomes that are higher than those of the private-sector middle class, making them less favorable toward redistribution and democratization? In fact, both the state and private-sector middle classes in my sample have median incomes slightly above the overall median of the distribution.³⁰ Moreover, the median income of the state middle class is slightly below the median income of the private-sector middle class, making the state middle class’s comparatively lower support for democracy puzzling in light of redistributive theories.

Given the striking rise in inequality as the region democratized, another possibility is that post-Soviet citizens expect the triumph of democratic forces to be accompanied by less rather than more redistribution, contrary to the assumptions of canonical redistributive models (Karakoc, 2018). This argument would imply that state workers are less favorable toward democracy because they are more sensitive than others to inequality. The survey asks respondents whether they agree or disagree that the gap between the rich and poor in their country should be reduced. There is no relationship between views on income inequality and sector of employment (see Supplemental Appendix Table A11). Concern about inequality is shared widely across lines of class and employment sector. Empirically, middle class respondents are no less likely to favor redistribution than those below them. Employees in the state sector are no more likely to favor redistribution than those outside of it. In sum, it seems unlikely that preferences over redistribution and attitudes toward inequality could explain the pattern of democracy support found in this article.

Communist socialization

A third potential concern is that the apparent effect of state employment actually reflects Soviet-era socialization. In most surveys it is impossible to distinguish the impact of Soviet-era workforce socialization from that of the post-Soviet era. The LiTS survey, however, offers a unique opportunity to disentangle the effect of post-Soviet careers from those of the Soviet era. Recent scholarship suggests that communist socialization has enduring effects (e.g., Beissinger & Kotkin, 2014; Pop-Eleches, 2007; Pop-Eleches & Tucker, 2017). If post-Soviet state workers were more imprinted by the communist system than non-state workers, and those effects persist, state workers may be less favorable to democracy in the present period.

Following existing literature, exposure to communism can be thought of as a cumulative experience (e.g., Pop-Eleches & Tucker, 2017). Hence, one strategy for dealing with the potentially confounding effects of communist socialization is to control for age, as I have done throughout. Indeed, age is almost perfectly collinear in this sample with communist socialization, when the latter is measured as years under a communist regime. Across virtually all of the models previously reported, age is highly statistically significant. The predicted probability of democracy support rises somewhat until about age 40, then steadily declines.³¹ These results imply slightly more democratic attitudes among those who were young adults at the time of perestroika, less support among those who grew up under declining pluralism after the mid-1990s, and the least support among those who lived longest under communism. Further analyses in Supplemental Appendix D.3 show that the effect of state employment on the middle class also cannot be explained by coming of age politically under communism, living through Stalinism, experiencing perestroika, or other communist-era experiences like party membership or a family history of repression. In sum, these results imply that being a state employee in a post-Soviet autocracy has an effect on attitudes toward democracy that is not driven by differential exposure to communist socialization. Communist experience matters too. But state employment after the fall of communism matters independently.

Conclusion

This article raises questions about the longstanding assumption that a large middle class aids democratization. Instead, I argue that autocrats benefit from control over public employment and the economic incentives it creates in particular for the middle classes to remain reticent about democracy. Using multi-country survey data, I have shown that across the post-communist autocracies the state middle class is generally unsupportive of democracy and no more supportive of democracy than the working class. I also find that as the share of one’s post-Soviet career spent in the state sector increases and private sector alternatives decrease, support for democracy declines—a pattern I attribute to the importance of long-run dependence on the state.

The results are robust to changes in the coding of middle class and alternative definitions of democracy support. I also assess a number of competing explanations, showing that neither age nor generation, views of the free market, communist era experience, or place in the income distribution can explain the effect of state dependence. Moreover, within the natural limitations of cross-national survey data, a variety of tests imply that the effect of state employment is not primarily due to processes that affect the types of individuals found in the state sector (i.e., to selection), but to the factors I identify. In the absence of true panel data, the ability to analyze variation in post-communist careers and career trajectories increases our confidence that state employment influences political preferences and not the other way round. Finally, in the democratic post-communist countries, where we would not expect state employment to weaken support for democracy, indeed, no such relationship was evident. This gives us greater confidence that the results reflect the effect of post-Soviet careers under post-independence authoritarian regimes—rather than some other shared legacy of universal state employment across the former communist bloc.

The fact that state employment is a non-trivial share of professional employment in middle-income autocracies around the world suggests that this article’s argument may well have broader reach outside the post-communist region. Better understanding employment relations between the middle class and the state and their implications for illiberal regimes beyond the former Soviet Union is a worthy avenue for future research. Where the share of middle class employment in the state sector is large relative to the private sector, as it is here (i.e., greater than half of formal employment on average across the post-Soviet nondemocracies), this framework is likely to be relevant.

In the ex-communist countries, democratization has uniformly been associated with neoliberal reforms and a shrinking state. Such reforms have often accompanied democratic transitions in other regions as well, though certainly at other times and places the neoliberal project has been more closely associated with authoritarian rule. Still even in Latin America, young democracies enacted painful market reforms beginning in the 1980s, dismissing many government employees and privatizing public enterprises (Weyland, 2004). Looking beyond the region, we would expect the political salience of a state-dependent middle class to be greater in countries where democratization is likely to be accompanied by neoliberal reforms than in places where these processes are decoupled.³²

Given the role that education plays here alongside occupation in defining the middle class, might the ideological and explicitly antidemocratic character of communist-era education make this study’s findings specific to post-communism? The fact that the state middle class which attended higher education after independence looks very much like the state middle class educated under communism suggests that these findings are probably not unique to the nature of the Soviet educational system. That the results also hold for younger respondents with less exposure to communist education can also be interpreted as implying the argument’s portability across different educational contexts. Of course, some caution is warranted. Educational systems in the former Soviet countries did not change overnight following independence and legacies of communist education persist. Whether the argument travels beyond the region and across different educational systems merits further investigation. Still, given that it holds both for those educated under a system that was highly ideological and at a time when many institutions of higher education across the region were working to instill democratic values does suggest greater generalizability.

Though the views of ordinary people alone do not determine the outcome of democratization, understanding the relationship between economic structures and attitudes toward democracy is of both theoretical and practical importance. Investigating whether the preferences this study documents inform voting, protest, and other political behaviors that shape democratic trajectories is a fruitful area for future inquiry as is their effect on democratic outcomes—or in the opposite case, suggested here, on authoritarian stability (see e.g., Dahlum et al., 2019; Rosenfeld, 2017). Relatedly, these findings cannot in the end tell us whether a more democratic middle class would generate democracy, as modernization theory implies, or whether the anti-democratic tendencies of the state middle class would fade were democracy to emerge for exogenous reasons. These are open questions for future study, though the post-communist experience of the 1990s may suggest that exogenous democratization is not easily consolidated when key social groups are unsupportive.

In conclusion, these findings suggests that there is no necessary connection between high levels of human and social capital and the preference for a liberal regime. I have shown that constituencies within the middle class whose livelihoods and status depend on public employment are less supportive of democracy than others with similar educational and occupational profiles. This evidence illustrates clearly how a large public sector contributes to authoritarian survival. It also speaks directly, at the micro-level, to one of political science’s abiding puzzles: development without democratization. Economic opportunities and employment security in the public sector may in fact delay democratization by helping to sustain a middle class that does not support democracy.

Supplemental Material

online-appendix-final – Supplemental material for State Dependency and the Limits of Middle Class Support for Democracy

Supplemental material, online-appendix-final for State Dependency and the Limits of Middle Class Support for Democracy by Bryn Rosenfeld in Comparative Political Studies

Supplemental Material

Replication_Files – Supplemental material for State Dependency and the Limits of Middle Class Support for Democracy

Supplemental material, Replication_Files for State Dependency and the Limits of Middle Class Support for Democracy by Bryn Rosenfeld in Comparative Political Studies

Footnotes

Acknowledgements

I wish to thank Chris Achen, Mark Beissinger, Ray Duch, Stephan Haggard, Henry Hale, Cynthia Kaplan, Grigo Pop-Eleches and seminar participants at Columbia, NYU, George Washington, UCLA, Princeton, Nuffield College/University of Oxford, Vanderbilt, WashU, Cornell, and the University of Southern California for excellent feedback on earlier drafts. I am grateful also to the journal’s editors and three anonymous reviews for very helpful comments. A previous version was presented at the 2017 annual meeting of the American Political Science Association.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Bryn Rosenfeld

Supplemental Material

Supplemental material for this article is available online at the CPS website

Notes

Author Biography

Bryn Rosenfeld is an assistant professor in the Department of Government at Cornell University. Her book, The Autocratic Middle Class, will be published by Princeton University Press. Her work has also appeared in such journals as The American Political Science Review, the American Journal of Political Science, and Sociological Research & Methods.

References

Acemoglu

Robinson

J. A.

(2006). Economic origins of dictatorship and democracy. Cambridge University Press.

Ansell

Samuels

(2008). Inequality and democratization. Paper presented at the 2008 meeting of the Midwest Political Science Association.

Ansell

B. W.

Samuels

D. J.

(2014). Inequality and democratization: An elite-competition approach. Cambridge University Press.

Banerjee

A. V.

Duflo

(2008). What is middle class about the middle classes around the world? Journal of Economic Perspectives, 22(2), 3–28.

Behar

Mok

(2015). Does public employment reduce unemployment? Topics in Middle Eastern and African Economies, 17(2), 111–121.

Beissinger

(2013). The semblance of democratic revolution: Coalitions in Ukraine’s Orange Revolution. American Political Science Review, 107(3), 574–592.

Beissinger

Kotkin

(Eds.). (2014). Historical legacies of communism in Russia and Eastern Europe. Cambridge University Press.

Bellin

E. R.

(2002). Stalled democracy: Capital, labor, and the paradox of state-sponsored development. Cornell University Press.

Boix

(2003). Democracy and redistribution. Cambridge University Press.

10.

Brown

Jones

D. M.

(1995). Democratization and the myth of the liberalizing middle classes. In Bell

D. A.

Brown

Jayasuriya

Jones

D. M.

(Eds.), Towards illiberal democracy in Pacific Asia (pp. 78–106). Macmillan.

11.

Chen

(2013). A middle class without democracy: Economic growth and the prospects for democratization in China. Oxford University Press.

12.

Collier

R. B.

Collier

(2002). Shaping the political arena: Critical junctures, the labor movement, and regime dynamics in Latin America. University of Notre Dame Press.

13.

Dahlum

Knutsen

C. H.

Wig

(2019). Who revolts? Empirically revisiting the social origins of democracy. The Journal of Politics, 81(4), 1494–1499.

14.

Dalton

R. J.

Shin

D. C.

Jou

(2008). How people understand democracy. In Diamond

Plattner

(Eds.), How people view democracy (pp. 1–30). Johns Hopkins University Press.

15.

Duch

R. M.

(1993). Tolerating economic reform: Popular support for transition to a free market in the former Soviet Union. American Political Science Review, 87(3), 590–608.

16.

Easterly

(2001). The middle class consensus and economic development. Journal of Economic Growth, 6(4), 317–335.

17.

Finifter

A. W.

Mickiewicz

(1992). Redefining the political system of the USSR: Mass support for political change. American Political Science Review, 86(4), 857–874.

18.

Frye

(2003). Markets, democracy, and new private business in Russia. Post-Soviet Affairs, 19(1), 25–45.

19.

Frye

(2006). Ownership, voting, and job creation in Russia. European Journal of Political Economy, 22(2), 452–471.

20.

Fukuyama

(2014). Political order and political decay: From the industrial revolution to the globalization of democracy. Farrar, Straus, and Giroux.

21.

Gans-Morse

Kalgin

A. S.

Klimenko

A. V.

Yakovlev

A. A.

(2017). Motivations for public service in corrupt states: Evidence from post-Soviet Russia (Working Paper No. WP BRP 13/PSP/2017). National Research University Higher School of Economics. https://dx-doi-org.web.bisu.edu.cn/10.2139/ssrn.3072181

22.

Gibson

J. L.

(1996). Political and economic markets: Changes in the connections between attitudes toward political democracy and a market economy within the mass culture of Russia and Ukraine. Journal of Politics, 58(4), 954–984.

23.

Gontmakher

Ross

(2015). The middle class and democratisation in Russia. Europe-Asia Studies, 67(2), 269–284.

24.

Greene

K. F.

(2010). The political economy of authoritarian single-party dominance. Comparative Political Studies, 43(7), 807–834.

25.

Hanmer

M. J.

Kalkan

K. O.

(2013). Behind the curve: Clarifying the best approach to calculating predicted probabilities and marginal effects from limited dependent variable models. American Journal of Political Science, 57(1), 263–277.

26.

Hough

J. F.

(1994). The Russian election of 1993: Public attitudes toward economic reform and democratization. Post-Soviet Affairs, 10(1), 1–37.

27.

Jackson

J. E.

Klich

Poznanska

(2005). The political economy of Poland’s transition: New firms and reform governments. Cambridge University Press.

28.

Jones

D. M.

(1998). Democratization, civil society, and illiberal middle class culture in Pacific Asia. Comparative Politics, 30(2), 147–169.

29.

Karakoc

(2018). Inequality after the transition: Political parties, party systems, and social policy in southern and postcommunist Europe. Oxford University Press.

30.

Kitschelt

Mansfeldova

Markowski

Toka

(1999). Post-communist party systems: Competition, representation, and inter-party collaboration. Cambridge University Press.

31.

Lipset

S. M.

(1959). Some social requisites of democracy: Economic development and political legitimacy. American Political Science Review, 53(1), 69–105.

32.

Mach

B. W.

Jackson

J. E.

(2006). Employment change, attitude evolution and voting during Poland’s transition: Longitudinal evidence. European Journal of Political Economy, 22(2), 472–502.

33.

Maleva

Ovcharova

(2008). Rossijskie srednie klassy: nakanune i na pike ekonomicheskogo rosta. In Shastitko

A. E.

Shastitko

S. B.

Abdasheva

M. A.

Ovchiniko

T. M.

Maleva

T. M.

Ovcharova

L. N.

(Eds.), Rossijskie srednie klassy: nakanune i na pike ekonomicheskogo rosta (pp. 7–102). Ekon-Inform.

34.

Marx

(2011). The unequal incidence of non-standard employment across occupational groups (IZA Discussion Papers, No. 5521). Institute for the Study of Labor (IZA). http://nbn-resolving.de/urn:nbn:de:101:1-201104113948

35.

McIntosh

M. E.

MacIver

M. A.

Abele

D. G.

Smeltz

(1994). Publics meet market democracy in Central and East Europe, 1991–1993. Slavic Review, 53(2), 483–512.

36.

McMann

K. M.

(2006). Economic autonomy and democracy: Russia and Kyrgyzstan. Cambridge University Press.

37.

Miller

A. H.

Hesli

V. L.

Reisinger

W. M.

(1994). Reassessing mass support for political and economic change in the former USSR. American Political Science Review, 88(2), 399–411.

38.

Moore

(1966). Social origins of dictatorship and democracy: Lord and peasant in the making of the modern world. Beacon Press.

39.

Munck

G. L.

Verkuilen

(2002). Conceptualizing and measuring democracy: Evaluating alternative indices. Comparative Political Studies, 35(1), 5–34.

40.

Nunnally

J. C.

(1978). Psychometric theory (2nd ed.). McGraw-Hill.

41.

Pop-Eleches

(2007). Historical legacies and post-communist regime change. The Journal of Politics, 69(4), 908–926.

42.

Pop-Eleches

Tucker

J. A.

(2017). Communism’s shadow: Historical legacies and contemporary political attitudes. Princeton University Press.

43.

Przeworski

(1991). Democracy and the market: Political and economic reforms in Eastern Europe and Latin America. Cambridge University Press.

44.

Ravallion

(2010). The developing world’s bulging (but vulnerable) middle class. World Development, 38(4), 445–454.

45.

Rechel

Richardson

McKee

(Eds.). (2014). Trends in health systems in the former Soviet countries. World Health Organization.

46.

Remington

(2011). The Russian middle class as policy objective. Post-Soviet Affairs, 27(2), 97–120.

47.

Richter

(2006). Wage arrears and economic voting in Russia. American Political Science Review, 100(1), 133–145.

48.

Rosenfeld

(2017). Reevaluating the middle class protest paradigm: A case-control study of democratic protest coalitions in Russia. American Political Science Review, 111(4), 637–652.

49.

Rosenfeld

Imai

Shapiro

J. N.

(2016). An empirical validation study of popular survey methodologies for sensitive questions. American Journal of Political Science, 60(3), 783–802.

50.

Sharunina

A. V.

(2013). Iavliaetsia li rossijskij “budzhetnik” “neudachnikom”? Analiz mezhsektornykh razlichij v oplate truda. Ekonomicheskij Zhurnal, 17(1), 75–107.

51.

Shevtsova

(2012). Implosion, atrophy, or revolution? Journal of Democracy, 23(3), 19–32.

52.

Smyth

Sobolev

A. S.

Soboleva

I. V.

(2013). Patterns of discontent: Identifying the participant core in Russian post-election protest. http://www.ucrs.uu.se/digitalAssets/254/254668_3discontentconfsmythpatternsofdiscontent.pdf

53.

Tsai

K. S.

(2005). Capitalists without a class: Political diversity among private entrepreneurs in China. Comparative Political Studies, 38(9), 1130–1158.

54.

Tucker

J. A.

Pacek

A. C.

Berinsky

A. J.

(2002). Transitional winners and losers: Attitudes toward EU membership in post-communist countries. American Journal of Political Science, 46(3), 557–571.

55.

Welzel

Inglehart

(2008). The role of ordinary people in democratization. Journal of Democracy, 19(1), 126–140.

56.

Weyland

(2004). Neoliberalism and democracy in Latin America: A mixed record. Latin American Politics and Society, 46(1), 135–157.

57.

Yang

D. D.

(2007). Classing ethnicity: Class, ethnicity, and the mass politics of Taiwan’s democratic transition. World Politics, 59(4), 503–538.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

5.30 MB

0.00 MB

0.24 MB