Instrumental or procedural democrats? The evolution of procedural preferences after democratization

ABSTRACT This paper addresses instrumentalist attitudes to democracy – attitudes according to which democracy is not valued for itself, but accepted only as a means to specific policy goals. Pippa Norris has argued that in the process of democratic consolidation, such instrumentalist conceptions of democracy are replaced with proceduralist ones, leading to an enlightened understanding of democracy. We use the unique case of German reunification to show that this process takes at least a generation to complete. Based on data from a novel battery of items fielded via the German GESIS panel, we show an East–West divide in democratic instrumentalism, which, however, is smaller among younger generations. While our findings do confirm Norris’ thesis that growing democratic experience leads to a shift from instrumental to procedural understandings of democracy, we also show that instrumental democrats still make up a sizeable portion of the citizenry that might withdraw support if dissatisfied.


Introduction
The stability and resilience of democracy depend on broad 'diffuse' or 'system support' for democracy as a form of government (Easton 1979;Fuchs 1999). Support for democracy, as frequently measured in numerous surveys around the world, seems to have at its core the comparison between aspirations derived from a normative ideal on the one hand, and experiences with real existing institutions and practices on the other hand. Where satisfaction with democratic practice diverges from aspirations, the resulting gap produces a 'democratic deficit' (Norris 2011) that potentially destabilizes the existing institutional order.
If citizens value democracy only as a means to specific ends, i.e. for instrumental reasons, support for democracy as a form of government is likely to be less stable, as it is contingent upon the regime achieving these ends. Dissatisfaction with the government's performance is in this case more likely to result in doubts about the capacities of the regime as such. By contrast, where democratic procedures are valued for themselves, as an expression and institutionalization of individual and collective autonomy, support for them will be less contingent upon specific decisions and policies. Citizens holding such a proceduralist conception of democracy will continue to support the regime even if they end up on the losing side of elections and policy decisions.
This distinction between procedural and instrumental conceptions of democracy has been proposed by Norris (Norris 2011, Chapter 8), who argued that in the process of democratic consolidation, instrumental (as well as authoritarian) conceptions of democracy are replaced with procedural ones, ultimately leading to an enlightened understanding of democracy (Norris 2011, Chapter 8). The German case constitutes a unique and interesting natural experiment that allows us an opportunity to test Norris' theory. Before reunification, around 20 per cent of Germans lived in an authoritarian socialist state, the German Democratic Republic (GDR), whereas the West German majority had, at the time, over 40 years of democratic experience in the Federal Republic of Germany (FRG). Socialization theory tells us that political attitudes are shaped early in life and unlikely to be entirely transformed by later experiences (Jennings 1996), meaning that a considerable share of East German citizens today may still be expected to possess different understandings of democracy from their West German contemporaries.
In what follows, we will first define instrumental as compared to procedural conceptions of democracy, drawing on arguments from both political theory and political sociology, and in particular on Pippa Norris (1999Norris ( , 2011 work. While we appreciate Norris' argument about the progressive development of conceptions of democracy, we point out some problems with her operationalization of conceptions of democracy. We argue that Norris' operationalization conflates preferences over specific policies with preferences over democratic procedures, and suggest an alternative measurement of instrumental and procedural conceptions of democracy.
We use a new item that we fielded in the GESIS panel in August-October 2015 and which in our eyes allows us to identify differences within comparatively consolidated Western democracies more adequately. Our results reveal significant differences between East and West Germans in older birth cohorts, which become smaller and insignificant in younger cohorts. The findings thus support Norris' hypothesis about the democratic consolidation process. Importantly, we show that her theoretical argument holds up under tests using a more direct measure of procedural preferences. We thereby update and expand upon the extant literature by showing that, even a quarter century after reunification, citizens, who have been socialized under an authoritarian communist regime, still prioritize policy outcomes over procedure to a greater extent than their compatriots, who have been socialized under democracy.
Citizens' conceptions of democracy: proceduralism and instrumentalism Support for democratic government and the acceptance of democratic decisions require a broad procedural consensus. The approval of democratic decision-making procedures as such allows those who have not found a majority for their preferred policy to accept and comply with the laws passed by the democratic system. In secular, pluralistic societies, where a policy consensus is out of reach, the procedural consensus is what in most matters allows constructive and peaceful dealing with dissent.
From a political philosophy perspective, different types of justifications for democracy are possible (see, e.g. Coleman and Ferejohn 1986). An instrumentalist justification is based on the assumption that democracy is instrumental to the achievement of a higher order goal that is external to it. These goals will typically consist in some version of a common good or overall welfare, but could in principle also be of a moral or religious kind. What is central here is that the justification of democracy hinges upon the justification of this higher order goal and on democracy's instrumentality for achieving that goal. From such an instrumentalist perspective, the justification of a democratic procedure is thus logically independent of its definition, i.e. from the characteristics and institutionalization of the procedure (Coleman and Ferejohn 1986, 7). A proceduralist justification of democracy, by contrast, is dependent on the definition of the procedure: what justifies democracy is not merely a consent on it or its instrumentality to achieve specific goals, but a quality of the procedure itself: for example, that it ensures democratic equality and allows citizens to exercise individual and collective autonomy.
Moving from political philosophy to political sociology, and thus to the level of individual attitudes, we cannot assume ordinary citizens to possess complete and coherent theories and justifications of democracy. However, we do know that citizens have specific understandings or conceptions of democracy that are more or less consistent. From such conceptions of democracy, citizens derive support for specific democratic procedures or democracy more generally and as a regime (see Landwehr and Steiner 2017;Goldberg, Wyss, and Bächtiger 2018). Citizens' support for democracy is thus to a significant degree based on normative reasons and justifications.
Recently, several authors have interpreted the rise of both technocratic decisionmaking and populist movements as a crisis of democratic proceduralism, and more specifically, representative party democracy (Urbinati 2014;Caramani 2017). Hence, it is important to understand what citizens mean by 'democracy' and what their democratic aspirations are. As Inglehart (2003, 51) notes: 'Although lip service to democracy is almost universal today, it is not necessarily an accurate indicator of how deeply democracy has taken root in a country'. As a citizen, I might thus support democracy either because I expect it to deliver outcomes that are overall superior to those delivered by any alternative regime (democratic instrumentalism) or because I regard it as a normatively superior way to deal with disagreements and arrive at collective decisions (democratic proceduralism), or for a mix of those reasons. This distinction between an instrumental and a procedural justification or conception of democracy is normatively and empirically relevant: Is citizens' support for democracy contingent upon the political system producing particular outcomes, and thus merely instrumental, or is it dependent on qualities of the democratic procedures, and thus procedural?

Conceptions of democracy, transformation and democratic consolidation
As we move a step further from political sociology to comparative politics and democratization studies, this question gains salience. The stability of democracy as a regime depends upon citizens' continued support in times of crisis. Only where citizens value democratic procedures for intrinsic and not only instrumental reasons and are willing to defend them can democracy be viewed as truly consolidated. Pippa Norris has described the development of citizens' conceptions as a learning process. In the process of democratization, authoritarian and instrumental understandings of democracy lose support while procedural ones gain support (Norris 2011, Chapter 8): 'Democractic experience […] undermines instrumental understandings of democracy' (Norris 2011, 161). This learning process is completed when proceduralist ideas have completely replaced authoritarian and instrumentalist ones, resulting in an 'enlightened' conception of democracy. On the other hand, socialization theory tells us that political attitudes, which are shaped early in life, are unlikely to change completely through later experiences (Jennings 1996).
For the German case that we study here, this implies that 30 years after democratic transformation, a considerable share of East German citizens may still be expected to possess different understandings of democracy from their West German contemporaries.
Norris' theorizing and empirical findings leave us with quite different hypotheses for East and West Germany. West Germany can, almost 70 years after the foundation of the FRG, be regarded as a consolidated and economically highly developed democracy. A culture shift towards post-materialism is reflected politically in the rise of the Green party since the 1970s, a party that has always emphasized citizen participation in politics. For West Germany, we would thus expect predominantly 'enlightened' procedural conceptions of democracy. Where citizens are dissatisfied with democracy, Norris would expect dissatisfaction to be caused at least in part by 'critical citizens' excessive expectations for democratic participation. While they embrace the procedural understanding of democracy, these 'critical citizens' are disappointed by the opportunities for effective participation the system has to offer (see Norris 1999).
In East Germany, however, a significant share of the population was politically socialized under an authoritarian regime, consequently losing out on the opportunity to acquire 'democratic knowledge' in Norris' sense. These differences in socialization are reflected in a different party system structure in the East: the post-socialist party (PDS, now Die Linke) has gained much higher vote shares in the East than in the West, whereas the Green party failed to gain ground in the East. Overall, the share of citizens with a party identification is still lower in the East than in the West (Ohr and Quandt 2012).
In sum, the differences in political socialization before reunification and the comparatively smaller experience with democratic institutions and procedures, along with the hopes for prosperity associated with reunification are more likely to reinforce instrumental conceptions of democracy in the East. Following Norris, we thus hypothesize that instrumental conceptions of democracy will be more prevalent in East than in West Germany. Furthermore, we expect instrumentalism to be particularly strong among those age groups that were exclusively or primarily politically socialized in the GDR (born between 1940 and 1975). That is, we hypothesize that the difference in instrumentalism between East and West Germans decreases in the age of our respondents.
While seeking to test Norris' hypotheses on the development of political knowledge and citizens' understandings and expectations of democracy, we depart from Norris where the measurement of democratic instrumentalism is concerned. As noted above, Norris uses data from the WVS 2005. The items show draws on ask respondents whether they regard specific properties as 'essential characteristics of democracy' and groups items into three sets. A first set of items assembles characteristics that are part of a minimal definition of democracy as polyarchy (Dahl 1971) plus liberal rights and the rule of law. A person who correctly identifies these characteristics as essential to democracy scores high in the procedural dimension.
In contrast, if a respondent classifies instrumental qualities of democracy (such as a prospering economy), the second set of items, or even authoritarian characteristics (such as 'the army takes over when the government is incompetent'), the third set of items, as essential to democracy this would indicate a less enlightened understanding. While the majority of participants in all countries score high on the procedural index, Norris finds that countries with less democratic experience have comparatively high averages on the instrumental and authoritarian scales as well (Norris 2011, 155). The socialist state of the German Democratic Republic (GDR), however, was neither a military nor a religious dictatorship, which is why the kind of authoritarianism identified with respective items in Norris' study does not seem relevant here.
We, however, take issue with the set of items Norris uses to measure an instrumental understanding of democracy. She regards agreement with the following items being essential to democracy as indicative of an instrumental understanding of democracy (see Norris 2011, 156): . The economy prospers. . Criminals are severely punished. . People receive state aid for unemployment. . Governments tax the rich and subsidize the poor.
In our eyes, this measurement seeks to capture understandings of democracy through policy preferences in a problematic way. While it may reveal instrumentalist ideas about democracy in underdeveloped transformation countries where public authority is frail and social security hardly existent, it cannot adequately capture differences within and between consolidated democracies. For example, a person with liberal attitudes on criminal law or with libertarian preferences for a minimal state could not hold an instrumental understanding of democracy according to Norris' measurement. Unless we conceive of understandings of democracy being intricately linked to these specific policy preferences, Norris' measurement does not capture democratic instrumentalism adequately.
Instead, we propose to disentangle procedural from substantial aspects in citizens' understanding of democracy by using an alternative measure that we believe to be a more direct measurement of democratic instrumentalism. With an item fielded in the GESIS panel survey (see below), we ask survey respondents whether, where collective decisions are concerned, they would prioritize 'a fair and democratic procedure' or a decision outcome that 'I think is right'. We thus operationalize a continuous scale ranging from a complete priority on the procedure itself (extreme proceduralism) to complete priority on decision outcomes (extreme instrumentalism) and allow respondents to indicate their preference in the trade-off between the two extremes.
We expect our analysis of understandings of democracy in East and West Germany to reveal insights into the process of democratic consolidation in a very special case: an economically highly developed democracy that with the German reunification in 1990 received 16 million new citizens politically socialized under an authoritarian regime.

Political attitudes after democratic transformation in Germany and beyond
We chose Germany as a unique case to study the development of political attitudes and preferences after political transition as a number of studies have done before (see for instance Dalton 1994;Roller 1994;Fuchs and Roller 2006;Alesina and Fuchs-Schündeln 2007;Neundorf 2009;Sack 2017). In particular, Alesina and Fuchs-Schündeln (2007) stress the relative homogeneity of Germany before its separation and point to the fact that its division into two separate countries has been exogenously determined with little regard to prior administrative or cultural borders. These authors also demonstrate that in terms of economic prosperity as well as electoral politics the parts of the country which were later to become the FRG and GDR did not differ significantly prior to World War II.
Hence, the German case provides us with a quasi-experimental setting to study political attitudes following political transitions. Dalton (1994, 470) suggests that studying reunified Germany is 'somewhat like […] a psychologist studying siblings separated at youth and then reunited in later life'. The treatment, in experimental parlance, then is the institutionalization of entirely different political and economic system in the two countries. East and West Germans thus grew up under very different circumstance with potentially longrunning consequences for their political orientations.
For reunified Germany, a set of studies conducted in the 1990s found considerable differences in political values between East and West German citizens (Fuchs 1999;Rohrschneider 1999;Hofferbert and Klingemann 2001) although proclaimed support for democratic principles is similarly high in both parts of the country even shortly after reunification (Dalton 1994). Rohrschneider (1996) demonstrated that similar differences could also be found between West and East German legislators. Ten years later, support for democracy in East Germany has become much more similar to West Germany than to other countries of Central and Eastern Europe which showed considerably lower levels of support (Fuchs and Roller 2006).
However, a more recent study by Sack (2017) shows that even in 2012, East Germans tend to prefer a socialist model of democracy, whereas West Germans preferred a liberal one. Other studies found similar results (Alesina and Fuchs-Schündeln 2007) demonstrate a persistent gap in economic policy preferences between East and West Germany. All of these papers show that higher demand for social security in the East is not just explained by the fact that the East is still lagging behind the West economically, exhibiting lower growth rates and higher unemployment, but also by its communist past.
These findings for Germany mirror results found in the broader literature on democratization in Central and Eastern Europe (Mishler and Rose 2007;Neundorf 2010;Pop-Eleches and Tucker 2014). For instance, Mishler and Rose (2007) show that while support for the new regime grows over time across the population differences between generations persist. Similarly, Pop-Eleches and Tucker (2014) highlight the importance of early socialization under communism. What differentiates the German case is the opportunity to compare citizens socialized under different regimes within one country.
Prior work has shown that East Germans, and citizens of post-communist states more generally, are prone to incorporating demand for policy in their conceptions of democracy. Beyond providing an update to that literature with newer survey data, we are also able to offer a new substantive perspective. Specifically, we will test whether East Germans are also more likely than their West German peers to prioritize these instrumental considerations over procedural considerations. In the next section, we present the data and explain how we measure instrumentalism in unified Germany.

Data and operationalization of democratic instrumentalism
To measure citizens' conceptions of democracy, we use data from the German GESIS Panel. The GESIS Panel is an off-line recruited mixed-mode access panel representative of the adult, German-speaking population in Germany (Bosnjak et al. 2017;GESIS 2017). It was launched in 2014 and at that time included about 4900 panellists. Since then, waves of the survey have been conducted every 2 months. The panellists can participate either off-line (via paper questionnaire) or on-line. 62% use the latter option. The data we use come from wave 'ce' (conducted between mid-October and mid-December 2015)approximately 3300 people participated in this waveas well as the base questionnaire, which all panellists completed.
The central item in the study for the purposes of this study is an item measuring how respondents weigh the procedural against the substantive quality of a decision: we ask survey respondents whether, where collective decisions are concerned, they would prioritize 'a fair and democratic procedure' or a decision outcome that 'I think is right'. We thus operationalize a continuous scale ranging from a complete priority on the procedure itself (extreme proceduralism) to complete priority on decision outcomes (extreme instrumentalism) and allow respondents to indicate their preference in the trade-off between the two extremes. Answers to this item allow us to infer whether respondents have a more procedural or instrumental understanding of democracy. For the former, the procedure takes priority over the result (a response on the left-hand side of the scale), for the latter the result takes priority over the procedure (a response to the right-hand side of the scale). Is there a procedural consensus in Germany? And, how frequent is an instrumentalist understanding of democracy? When we look at the descriptive statistics we see an almost uniform distribution of answers ( Figure 1).
Contrary to what one might expect, the medium category, with which the wish for the compatibility of procedural and substantive quality of decisions could be expressed, is not the most frequently chosen one. However, this result can partly be attributed to the wording of the question which specifically refers to a possible trade-off between procedure and substance. However, we also do not observe a concentration of answers at the ends of the scale, which such a question could have caused. We deem it likely that the distribution arises from interviewees being torn between the desire for compatibility of procedure and substance and the recognition of the goal conflict stressed by the question. Further conclusions escape a purely descriptive assessment. Nevertheless, the determinants of attitudes on the procedural and substantial quality of decisions can be examined in further analyses.

Comparing instrumentalism in East and West Germany
In this section, we compare mean scores of instrumentalism across East and West as well as birth cohorts while controlling for other potentially confounding variables. We first provide a visual description of instrumentalism across birth cohorts in East and West Germany and then document still existing socio-economic and demographic discrepancies between East and West Germany as they are reflected in our data.
During almost 50 years of separation, East and West Germany, as well as more generally West and Eastern Europe, underwent dramatically different economic development. Because of this, the younger democracies of Central and Eastern Europe still lag behind Western Europe regarding their prosperity, as evidenced by GDP per capita or purchasing power, to this date. East Germany, while much more prosperous than neighbouring countries to the east, is, even more than a quarter century after reunification, still lagging behind West Germany in many regards. A study published 25 years after reunification by the Berlin Institut (2015) reveals a continuing divide between East and West Germany, particularly in terms of the economic situation. East Germany is still structurally disadvantaged: for instance, East Germans have only 80% of West Germans' average disposable income (Berlin Institut 2015, 35). Also, the share of people in risk of poverty is much higher in East than in West Germany; as is unemployment.
Hence, we are faced with a notorious question in the social sciences: whether differences at the systemic level, which we document in Figure 2 and Table 1, are due to systemic influences or merely to the composition of the population within the systems (Przeworski and Teune 1970). In our case of differences in understandings of democracy between East and West Germany, compositional effects concern an unequal distribution of individual-level characteristics that are related to our dependent variable. Systemic effects, by contrast, describe the influence a system has on the behaviour or attitudes of individuals within the system. In our case, this is the political socialization of citizens under a communist-authoritarian (in East Germany) and a market economy-democratic system (in West Germany and after 1990 in unified Germany).
We use a matching approach to compare citizens who resemble each other in all relevant characteristics but onebeing from East or West Germanyto ascertain that the differences in democratic conceptions between East and West Germany are indeed cultural and not merely results of current structural differences between East and West Germany. We consider the following socio-economic and demographic control variables: gender, education, employment status and income. These variables, in terms of which the composition of the two groups (East and West Germans) differs, are, at the same time, plausibly linked to instrumentalism.  [1985][1986][1987][1988][1989][1990][1991][1992][1993][1994][1995] is the first one to be completely politically socialized in the unified Germany. The size of the points indicates the share of respondents belonging to the respective cohort. Matching is preferable to OLS regression because it is a non-parametric technique that allows us to also account for complex interactions between the control variables which would require multiple interactive terms in a linear regression model. Using EM, we create a subsample of respondents in which the composition of these characteristics is exactly equal in East and West Germany. Our aim is to compare citizens of the same birth cohort from East and West Germany who are as similar as possible in all other relevant regards. For instance, we compare two or more women from East and West Germany whose educational level is equivalent to upper secondary, who are employed and live in a household with a monthly income of between 3200 and 4000 euros.
We begin by visually inspecting the differences in instrumentalism between East and West Germany across different birth cohorts. Figure 2 shows the mean values of instrumentalism per birth cohort, differentiating between East and West Germany. The vertical bars depict 95% confidence intervals around the group means. Two 'trends' are apparent in this visualization. First, instrumentalism tends to be lower on average among younger cohorts in both East and West Germany. Second, the difference between East and West Germans within birth cohorts also tends to be lower among the younger birth cohorts.
Within older cohorts, we see marked differences between East and West, while younger East Germans who have spent less (born 1975-1985) to no time (born 1985 or later) under the old system are indistinguishable from their West German peers. Unfortunately, since we present a cross-sectional analysisour item was only fielded in this single wave of this specific surveywe cannot say anything about the development of instrumentalism within individuals over time. We address this issue further in the discussion of our results.
These initial findings support Norris' argument that conceptions of democracy develop progressively in the process of democratic consolidation. Moreover, they do so while avoiding Norris' above-mentioned measurement problem of conflating procedural preferences with policy preferences. However, another worry, common to both prior studies (see for instance Alesina and Fuchs-Schündeln 2007;Neundorf 2009) and our present study, is that differences in understandings of democracy are not just cultural but also rooted in socio-economic and demographic structural factors. Figure 3 shows that younger cohorts are under-represented and older cohorts over-represented in East Germany. As older citizens tend to be more instrumentalist, this induces a bias in a simple comparison of instrumentalism in East and West Germany. One reason why East Germans on average, that is across cohorts, have more instrumentalist understandings of democracy than West Germans is that the average age is higher in East than in West Germany. By calculating mean values of instrumentalism within birth cohorts, as we do in Figure 2, we control for this bias (see Table 1). Consequently, the average of the differences between East and West Germans' instrumentalism within age groups is different from the average difference in instrumentalism between East and West Germany. The former is 0.35 points ('Average' in Table 1) while the latter is 0.44 ('Overall' in Table 1).
The same logic used in these within-age-group comparisons can be extended to further potential explanatory variables through the use of matching. Before we do so, we document the differences in our control variables between East and West Germany and explain their coding. Figure 3 compares levels of education, income and unemployment between East and Germany documenting lower incomes and higher unemployment in East Germany. Figure 3 also shows how these variables interact with gender, for instance, strong differences in unemployment between East and West German womensomething which the non-parametric matching allows us to easily incorporate.
We code education according to UNESCO's International Standard Classification of Education . For this, we employ two questions from the base questionnaire completed by all panellists, which ask respondents about their educational degrees and vocational training. 1 Household income is an ordinal variable defining multiple income brackets into which respondents can sort themselves. Unemployment is a dummy variable indicating whether a respondent is unemployed or of any other employment status (e.g. full time, part time, student, pensioner, etc.) (Figure 4).
Because all our control variables are categorical, they result in a modest number of unique combinations of variable values. Hence, we can employ an Exact Matching (EM) procedure, which we implement using the R package MatchIt to create a subsample of similar East and West German citizens (Ho et al. 2007). Our matched subsample consists of 1740 observations: 414 East German respondents were matched with 1326 West German respondents who had the exact same values on the control variables. In the matched subsample East and West German respondents do not differ from each other on average with regard to our control variables gender, education, household income and employment status. 2 Since we compare respondents who are very similar to each other except for their place of living, we are now more confident that our estimates represent the causal effect of having been socialized in East Germany on the one hand or West Germany and unified Germany on the other hand. 3 One caveat to this procedure is the reduction in sample size and corresponding loss of statistical power due to the matching procedure. In particular, 52 East German respondents had no exact match among West German respondents in our sample and hence were excluded from the analysis. 469 West German respondents were dropped from the dataset because they did not resemble closely enough any of our East German respondents. However, the reduction of the sample is also a feature because these 'outliers' removed by matching have the potential to bias our estimates of group differences. As an alternative and robustness check, we also use Optimal Matching which retains all 466 respondents from East Germany. Here, we match each East German respondent with the one West German respondent who is most similar to them, but not necessarily exactly the same, while still excluding some West Germans that do not closely resemble any East German respondent. We present the results of this analysis, which are substantively similar to the results obtained through EM, in the appendix. 4 Here, we only present the results obtained by using EM which is the more conservative approach. Table 2, just as Table 1, provides a comparison of mean values of instrumentalism in East and West Germany, within birth cohorts and across the complete sample. Positive differences indicate higher values of instrumentalism in the East while negative values indicate lower values of instrumentalism in the West. We still see a similar pattern in these comparisons in that the oldest birth cohort exhibits the most substantial difference in instrumentalism, whereas younger cohorts do not differ at all. Overall and despite a  [1943,1955) 0.56* 99 350 [1955,1965) 0.28 125 324 [1965,1975) 0.59* 99 340 [1975,1985) 0.47°56 209 [1985,1995) −0.25 35 103 Overall 0.4* 414 1326 *p < 0.05, **p < 0.01, ***p < 0.001.
significant reduction in our sample size, we still see a substantial and statistically significant difference between East and West German respondents. Differences are still as substantial as before and by and large correspond to the pattern established in Figure 2. Among the oldest cohort, East German respondents are substantially and significantly more instrumentalist than respondents from West Germany, while there is no difference among the younger cohorts. The birth cohort [1955,1965) is an aberration in that the difference between respondents in East and West Germany in this cohort is substantively smaller than in the neighbouring birth cohorts and statistically insignificant. East Germans in that birth cohort were 23-34 years of age when the wall came down in 1989. 30-50-year olds and under-age citizens saw the highest numbers of migration from East to West Germany. Even over two decades later net migration between East and Germany is still negative in the age groups below 50 years of age (Statistisches Bundesamt 2015). This implies that particularly young families were moving from East to West Germany at the time of reunification which would explain the pattern we see in Table 2. For the youngest cohort, those born in 1985 or later, East German respondents are even less instrumentalist than their West German peers. Although substantial, this difference is not statistically significant due to the small sample size.
Some caution is in order when interpreting these results. We use respondents' current place of living to distinguish between East and West Germans as no better measure is available in the survey. More specifically, there is no question in the panel asking respondents whether they lived or grew up in East Germany before 1990. Of course, using current place of living is not without problems because by now, 5 million people (roughly 6% of the German population) have moved from East to West Germany or vice versa (Statistisches Bundesamt 2015). What does this imply for our results? If East Germans, who on average hold more instrumentalist conceptions of democracy, move to West Germany, this will increase average instrumentalism in the West. Vice versa, if West Germans who on average have less instrumentalist conceptions of democracy, move to East Germany, this will decrease average instrumentalism in the East.
Accordingly, we regard our measurement of the differences between East and West Germany as a conservative one. This means that we are confident that differences between older cohorts of East and West Germans do indeed exist and are probably even more substantial than our estimates suggest. Every single year between 1957 and 2016 more people have moved from East to West Germany than the other way around implying that the likely attenuation bias in our estimates mostly accrue to East Germans having moved to West Germany. This is, as we have argued above, consistent with the patterns between cohorts we observed in Table 2. However, we are less confident about the convergence in attitudes among younger respondents. Particularly, many young adults from West Germany have moved to East German university cities for their studies. Consequently, we cannot rule out with certainty the possibility that young East Germans still hold more instrumentalist conceptions of democracy than their peers from West Germany.
These caveats notwithstanding, we do demonstrate persisting differences in citizens' conceptions of democracy in East and West Germany. We provide an update to the existing literature by showing that generational patterns rooted in political socialization under different systems continue even a quarter century after democratic transition. With our data, we are able to cover the first post-transition generation, which was exclusively socialized politically under democracysomething which earlier studies were not able to do. In our analysis, we saw patterns in line with a 'trend' of convergence between East and West driven by generational replacement, which will potentially end in full convergence: we observe no significant difference among our youngest cohorts.

Conclusion
Our analysis has revealed significant differences between East and West Germans in terms of what Norris labelled 'instrumental conceptions of democracy'. East Germans in older birth cohorts tend to attach a higher weight to the substance of decisions than to their origin in fair and democratic procedures. These differences persist even if we control for composition effects, i.e. for differences between the East and the West German population reflected in our sample. The unique case of Germany after reunification allows us to rule out country-level effects and to isolate the effect of political socialization. Our analysis confirms Norris' hypothesis about 'democratic enlightenment' in the process of democratic consolidation. Importantly, we show that her theoretical argument holds up under tests using a more direct measure of instrumental and procedural conceptions of democracy.
However, our results also show a persistent socialization effect: older birth cohorts living in East Germany are still significantly higher on democratic instrumentalism than their peers in the West and young people from both East and West Germany. Our findings mirror and update those of other researchers who have worked on the long-term impact of socialization using the case of Germany or post-communist transformation. We show that East and West Germany, even more than a quarter century after reunification, still differ in terms of instrumental versus procedural differencessimilar to other researchers who have documented differences in, among others, understandings of left and right (Neundorf 2009), economic policy preferences (Alesina and Fuchs-Schündeln 2007) or party identification (Ohr and Quandt 2012).
Our and other studies on the German case also find generational differences in political orientations which are also found in other post-transformation societies in Central and Eastern Europe. For instance, Mishler and Rose (2007) found persistent differences in the support for the new and old regime between different generations in Russia although all generation grow more supportive of the new regime over time. In short, our findings are broadly representative of patterns of generational differences and post-transformation learning over time found in the extant literature.
However, prior work on the German case, as well as post-communist states more generally, has often looked at instrumental or procedural understandings of democracy separately. That literature has established that citizens of post-communist are prone to incorporating demands for policy in their conceptions of democracy. We, more specifically, have shown that East Germans socialized under a communist authoritarian system are also more likely than West Germans to prioritize those policy concerns over democratic procedures.
Yet, in the absence of long-term panel data, we have to acknowledge that we are unable to assess the relative importance of generational replacement and democratic learning or 'enlightenment' (Norris 2011) within individuals for closing the gap in conceptions of democracy between East and West Germany. Alesina and Fuchs-Schündeln (2007), who use two waves of the German Socio-Economic Panel (SOEP) to study economic policy preferences, also demonstrate a convergence in attitudes and preferences between East and West Germany. They estimate generational replacement to account for about one-third and change of preferences within individuals for the remaining twothirds of the convergence.
Our results stand in interesting contrast, though not in a contradictory way, to earlier work by Dalton, Bürklin, and Drummond (2001) who find that support for direct democracy is higher in East than in West Germany. This finding could be explained by citizens who value obtaining the right decision over a fair procedure having a stronger tendency to believe that a majority of the population thinks as they doa phenomenon which is called 'false consensus' in psychological research (Ross and Greene 1977;Marks and Miller 1987;Wojcieszak and Price 2009). Relatedly, populist attitudes might also be more prevalent among instrumental democrats. Put differently, why would one want elaborate decision-making procedures aimed at balancing competing interests if one believes that there is a unified 'will of the people'?
Interestingly, Dalton et al. find that from 1991 to 1998 support for direct democracy decreased in East Germany while it increased slightly in the West. On the one hand, this could be explained by the net movement of about a million citizens from East to West Germany in the years immediately following the fall of the wall. On the other hand, it is also possible that as citizens in the East experienced their first democratic elections, they became aware of the multiplicity of interests that representative democracy is designed to moderate and, thus, in the process of 'enlightenment' became more sceptical of simpler institutions such as direct democracy. Unfortunately, we are unable to test these conjectures with our data. However, our results provide fertile ground for further theorizing and empirical analysis of how understandings of democracy develop and influence political attitudes and behaviour. As our measure is not wedded to a specific contexthowever, policy preferences might beit can also be deployed in surveys in other (post-)transformation countries. Given that political attitudes in East Germany seem to have quicker than in other post-transformation societies (Fuchs and Roller 2006), we should expect to see even higher levels of instrumentalism in, for instance, Central and Eastern Europe.
While our focus is on the convergence between East and West Germany we note that instrumentalism, as measured with our one-item strategy, is far from being a purely East German phenomenon. Overall, 40 per cent of respondents place themselves on the right side of the scale. This, however, does not necessarily mean that German democracy is in danger. Only those who position themselves at the extreme right-hand side of the scale seem to completely deny the importance of fair and democratic procedures, while those who chose any of the options left of it at least implicitly acknowledge a goal conflict between procedures and substance. Some degree of instrumentalism may in fact be entirely compatible with democratic attitudes and behaviour. It is not a problem if people expect democracy to deliver better economic outcomes in general or to secure peace and freedom. Only if they make their support for it dependent upon the delivery of specific policies does an instrumental understanding of democracy become problematic.
Moreover, the stability of the procedural consensus for democracy eventually depends on at least two factors: citizens' expectations of democracy and their experience with it. Only if democracy fails to meet the expectations of instrumental democrats are these likely to question existing democratic institutions. While instrumentalism may be problematic in itself, it only becomes an imminent threat in those instrumental democrats who feel that representative democracy systematically fails to fulfil their expectations.
What factors then lead proceduralists and instrumentalists to become dissatisfied with the way democracy works? Do these factors differ for proceduralists and instrumentalists? Are instrumentalists more prone to become dissatisfied with democracy? We consider these questions to be important topics for future research. Notes 1. Because respondents were not asked about doctoral degrees, we are unable to code category 6 'Upper Tertiary' of the ISCED-97 coding scheme. Respondents with a doctoral degree are thus contained in category 5 'Higher/Tertiary'. 2. For a comparison between East and West German respondents before and after matching, see Figure A.1 and Table A.1 in the Appendix. 3. For a comparison between East and West German respondents before and after matching, see Figure A.1 and

Disclosure statement
No potential conflict of interest was reported by the authors.