How do we come to our political views?

"The German people"? The unequal responsiveness of the Bundestag


In a democracy, when politics systematically follows the political preferences of certain social groups while disregarding those of others, the principle of political equality is damaged. The new responsivity research examines whether political decisions are in line with the will of the citizens and, if so, whose opinions are implemented. In the USA, there is a clearly selective policy responsiveness to the detriment of the poor. We are investigating for the first time whether there are similar patterns in political responsiveness in Germany as in the USA. To this end, we evaluate 252 factual questions asked in the DeutschlandTrend surveys for the period from 1998 to 2013. The questions relate to policy changes and reforms discussed at the time of the survey from a wide range of politically relevant topics. The analysis of these data shows a clear correlation between the political decisions made and the attitudes of people with higher incomes, but no or even a negative correlation for those with low incomes. In Germany, there is a similar imbalance in political representation as in the USA, although the two countries differ greatly in their institutional prerequisites. Against this background, we finally discuss possible explanatory mechanisms for unequal responsiveness.


Democratic systems face a steady tension between formal political equality and actual inequalities of their citizens. If disadvantaged groups cannot hope that their interests will have an equal chance to prevail against those of more privileged groups in society, a key principle of democracy is violated. New studies on political responsiveness analyze whether or not decisions confirm to the citizens ’will and if so, whose preferences prevails. A number of studies focusing on the United States have found a pattern of selective responsiveness, in which the interests of the poor and median income groups are mostly ignored. In this article, we ask whether similar patterns are discernible in Germany. We analyze an original dataset of 252 detailed survey questions posed between 1998 and 2013. The questions deal with specific political decisions debated at the time and cover a broad range of politically relevant topics. Our results show a notable association between political decisions and the opinions of the rich, but none or even a negative association for the poor. Representational inequality in Germany thus resembles the findings for the US case, despite its very different institutional setting. Against this background, we conclude by discussing potential mechanisms of unequal responsiveness.


Liberal democracies find themselves in a constant field of tension between the claim to political equality and the existence of economic inequality.Footnote 1 The legally guaranteed opportunities to participate in an egalitarian manner based on citizenship status are intended to prevent otherwise existing inequalities from translating into unequal political influence. However, a number of research papers in recent years have shown that there is a negative correlation between social inequality and the degree to which political equality is achieved (Solt 2008). This relates to political participation, the representation of different social groups in parliament as well as the responsiveness of political decisions to the preferences of the socially disadvantaged. However, when the poor, precariously employed or formally poorly educated do not have the same chance that their concerns are heard and implemented in the political process, the promise that political equality will be implemented despite all actual inequalities is violated. The question of the leveling effect of democracy becomes more relevant as social inequality increases.

In this essay we concentrate on one aspect of this debate and ask whether there is a similar imbalance in political responsiveness in Germany as it has been attested for the USA. As recent research shows, there is a clear imbalance in political decisions at the expense of the poor and middle income groups. In particular, when rich and poor have different preferences, politics follows those with high incomes (Bartels 2008; Gilens 2012; Gilens and Page 2014). Many American authors see the reason for this imbalance in the predominantly private party and election campaign financing in the USA, which encourages political decision-makers to become dependent on large donors (Gilens 2015a; Lessig 2014).

In many European countries, on the other hand, the institutional conditions are different. At the same time, research on unequal responsiveness in Europe is still in its infancy. Although initial findings point to an imbalance in political decisions (Giger et al. 2012; Peters and Ensink 2015), existing studies do not allow a direct comparison with the American results due to their different empirical approach. In this article we therefore examine the relationship between the attitudes of different population groups and the political decisions of the Bundestag for Germany. Germany is well suited as a comparison case to the USA, because parties and election campaigns are financed by the state and through membership fees to a much greater extent. In our empirical study, we evaluate a new database that we created, which contains information for 252 factual questions about how strong the approval of a policy change was in different income groups and what political decision was subsequently made. We are using the complete ones for the first time GermanyTrend-Data since 1998 not previously available for research.

Our analysis shows that in Germany, too, political decisions are more in line with the preferences of higher than with those of lower income groups. The greater the differences of opinion, the more pronounced this imbalance is. These results raise doubts as to whether party donations alone explain differences in responsiveness, although we can only discuss competing explanatory approaches in this essay, but cannot conclusively examine them. The article is structured as follows: In the next section we go into the previous findings of responsiveness research, then we present the data basis and the methodical approach. In the following section we present our results, first considering the differences of opinion between social groups and then addressing the question of whose political preferences have a better chance of being implemented. The article closes with a conclusion and an outlook on future research perspectives.

State of research

In order to theoretically conceptualize the relationship between population and politics (or government) and its significance for political representation, the term “responsiveness” was developed in the literature on democratic theory. A government is responsive if it takes into account the concerns and interests of the population and reacts to them.Footnote 2 As a result, the will of the population and political action should be in harmony with one another. If, on the other hand, political decisions contradict the majority opinion of the population, they require special justification (Pitkin 1967, p. 232). In a representative system, the rulers must have room for maneuver that also allows unpopular decisions, but permanent or systematic ignorance of the will of the population undermines the legitimacy of politics (Pitkin 1967, pp. 232–233). Democratic responsiveness also requires that not only the interests of a few, but the interests of all those represented can influence political decisions.

Empirical responsiveness research has a long tradition in political science (Miller and Stokes 1963; Achen 1978). Until a few years ago, however, it dealt exclusively with the question of whether there is a general connection between public opinion on the one hand and political decisions on the other. Specifically, empirical studies examined whether the political decisions of the governments (or the voting behavior of individual MPs) correspond to the majority opinion of the population (for an overview see Manza and Cook (2002); Shapiro (2011)). Methodologically, public opinion in these works is recorded through surveys on specific political issues. Overall, many of these studies find a high degree of general political responsiveness, albeit to varying degrees in different policy areas (Page and Shapiro 1983; Wlezien 2004). Brettschneider (1995) examines 94 cases between 1949 and 1990 in which there was a change of opinion in the population as to whether the Bundestag reacted to this. In 55 to 60% of cases of change of opinion there is a parliamentary action by the Bundestag, which is why, from the author's point of view, the responsiveness of the German Bundestag is "greater than is often claimed" (Brettschneider 1995, p. 145).

Against the background of increasing income inequality and unequal political participation, responsivity research has turned in recent years to the question of whose Opinions are considered politically. In these more recent studies, US authors in particular examined the extent to which economic inequality translates into political inequality. They show that the socially disadvantaged have little influence on the political process and therefore warn of the consequences of increasing social inequality for democracy (Bartels 2008; Flavin 2012; Gilens 2005, 2012; Gilens and Page 2014). The new responsivity research is initiated by the analyzes of Martin Gilens (2005, 2012). In them he evaluates almost 1,800 factual questions from representative population surveys that were carried out between 1981 and 2002 in the USA and each asked for approval or rejection of a currently discussed policy change. Gilens compares the opinions expressed in the surveys with the political decisions made up to four years after the question asked and comes to a clear conclusion: If the interests differ between the income groups, the policy follows the high-income, while the Concerns of the "median voter" and the low-income are not taken into account. Bartels (2008, Chapter 9) comes to similar conclusions: the voting behavior of US Senators depends heavily on the interests of the high-income groups, whereas the opinion of the lower income groups has no influence. In a further study, Gilens and Page (2014) not only analyze the attitudes of different income groups, but also collect the political positions of important organized interest groups on the respective policy changes and compare them with the decisions made. Their results show that the demands of business-related interest groups and high-income individuals are reflected much more frequently in political decisions than those of employee-related organizations and middle and lower income groups.

These findings have stimulated an ongoing discussion on the relationship between social inequality, responsiveness and political representation. Since unequal responsiveness presupposes that different social groups have conflicting views on certain political issues, a central point of discussion is whether there are significant differences of opinion between income groups. Critics argue that the preferences of the population do not differ significantly between income groups in many policy areas, which is why there is little scope for unequal responsiveness and the findings of unequal responsiveness are exaggerated (Enns 2015; Soroka and Wlezien 2008; Ura and Ellis 2008). In the American case, however, it can be shown that differences of interest become more visible the more differentiated opinions on specific political programs, measures or laws are asked (Gilens 2009, p. 340) and the more important issues that the groups deal with are often particularly important are contrary to this (Gilens 2015b). In addition, people with very high incomes or wealth are often underrepresented in surveys, which tends to lead to an underestimation of existing differences in attitudes. To examine the preferences of the top income groups more closely, Page et al. (2013) in a pilot study 83 interviews with Americans, whose wealth averages 7.5 million dollars. They find that the respondents are, on the one hand, more politically active and maintain closer contact with important politicians, and on the other hand they have decidedly different opinions in many political areas than people with average incomes, especially in social and tax policy.

Both empirical studies and the subsequent scientific debates have so far focused primarily on the USA. Previous work on political responsiveness in Europe has in most cases been limited to measuring the congruence between public opinion on the one hand and the opinion of politicians or party positions on the other (Adams and Ezrow 2009; Bernauer et al. 2013; Giger et al. 2012; Petring 2015). Donnelly and Lefkofridi (2014), on the other hand, examine in a comparative study of Western European countries the extent to which public opinion on selected political issues agrees with policy results. For the period from 1980 to 2010, they compare, among other things, opinions on topics such as the amount of public spending in various policy areas or the right to abortion with the political situation in the respective state in which the survey took place. They conclude that policy outcomes reflect the opinions of the upper income brackets more than those of the lower income brackets. The same applies to the work of Peters and Ensink (2015), who examine the relationship between redistributive preferences in the population and the level of social spending (as a percentage of gross domestic product) for 25 European countries. Their results show that the level of social spending correlates more strongly with the preferences of upper than lower income groups, especially when voter turnout in a country is relatively low. Since these studies are politicsResults (e.g. the distribution of income or the amount of expenditure on pension policy at a time) and not specific ones made by the policy decisions investigate, it remains open whether this result really reflects unequal responsiveness. If, for example, the population speaks out in favor of less social inequality, but inequality increases, this can have various reasons. In addition to lower redistribution, economic developments also influence the degree of inequality and, if in doubt, it can even increase, although politics tries to prevent this.

Even outside the American context, initial studies provide indications that politics is selectively responsive, but a comprehensive investigation is still lacking. Against the background of other institutional requirements in many European countries - such as electoral systems with proportional representation and greater public funding of parties and election campaigns - such a study is particularly promising for identifying possible mechanisms of unequal responsiveness. A mechanism that is often cited for the USA is the great dependence of politics on private (election campaign) donations, which in the USA make up the majority of funding. While the election campaigns are becoming more and more expensive, the group of large donors who carry a considerable part of the funding is getting smaller and smaller (Gilens 2015a, p. 226). The possibility of direct political influence by the “super-rich” therefore seems to many American authors the most obvious explanation of unequal responsiveness. In Europe, where party funding is primarily public in most countries and campaign donations are much lower, this mechanism appears less plausible (Woll 2015, p. 8). A comparative study with the USA therefore helps to clarify whether and to what extent unequal responsiveness also exists in other western democracies and to what extent the thesis that this is due to the influence of large donors stands up to international comparison.

Data basis and methodical approach

Germany is well suited as a comparison case to the USA to examine the question of unequal responsiveness. As a parliamentary democracy with proportional representation, it differs in many ways from the American political system in terms of its institutional structure. In particular, the rules on private party financing are very strict in the Federal Republic of Germany. In 1959, Germany was one of the first European countries to introduce state funding for political parties (Woll 2015, p. 8), which is why the parties are largely financed through the reimbursement of public election campaign costs and membership fees.In contrast to the USA, private party and election donations play a much smaller role, which could lead to a different pattern of responsiveness on the part of parliamentarians. At the same time, suitable surveys are available that allow a procedure analogous to that of comparative studies.

The basis of the following study is the database we created "Responsiveness and Public Opinion in Germany (ResPOG)". It is based inter alia. to surveys of the Germany trend from the period 1998 to 2013, which ask about the approval or rejection of a political proposal or reform.Footnote 3 The questions contained in the data set cover very different policy areas and range from assessing a statutory quota for women to deployments abroad by the Bundeswehr to questions about a minimum wage or cuts in social spending. They reflect which questions dominated the public debate in a given month and which reform initiatives existed at the time of the survey. A total of 252 factual questions can be evaluated. For each factual question, the approval rate of different social groups was determined in order to be able to systematically analyze differences in attitudes. For this purpose, the respondents were categorized according to region (East and West Germany), gender, age, education, occupation and income, and the proportion of those who supported the requested policy change was determined. In order to be able to measure the responsiveness of the policy, it was then coded for each question whether a corresponding policy change took place over a period of two or four years or whether it did not materialize. In the case of laws, the date of issue is decisive for the point in time at which a change was made; in the case of resolutions, the day on which the resolution was passed. The corresponding changes in the law were researched mainly with the help of the Documentation and Information System (DIP) of the German Bundestag. Newspaper research and information from the portal complemented the survey. We limit ourselves to an investigation period of a maximum of four years after the question has been asked because we assume that subsequent political action can hardly be seen as a responsive reaction.

In order to be able to make statements about how much more likely (or less likely) a change will be if, for example, it receives significantly more approval from the 90th percentile of the income scale than from the 10th percentile, we calculate a logistic regression of the policy change on the approval rates in the various Income groups.Footnote 4 So that our analyzes are comparable with those of American studies, we follow their research design. Since the income categories surveyed vary over time and in size, we are guided by Gilens' approach (2005, 2012) in order to be able to compare the income groups with one another and over time. Each respondent is assigned the mean value of the percentage of his income group. If, for example, the lowest income group comprises five percent of the respondents and the next higher 10%, the respondents in the first group are assigned the value 2.5, the second the value 10 (as a mean value between 5 and 15%, the lower and upper end of this income group). A logistic regression is then used to estimate the probability with which a respondent agreed to an alternative answer. The group mean calculated in this way and its square flow into the regression as explanatory variables, so that non-linear relationships can also be taken into account (Gilens 2012, p. 61). When interpreting the income differences, it should be noted that these are statistically determined probabilities.Footnote 5


Differences of opinion between social groups

A prerequisite for distorted political decisions to the detriment of certain social groups is that they differ substantially in what they expect or demand from politics. In a first step, we therefore determine how great the differences of opinion are on the questions we have examined. In the following, we will mainly look at differences in attitudes between income groups, but we always use other groups for comparison. For each question in the data set, the proportion of respondents in different subgroups who agreed to a policy change was determined. If one compares the approval values ​​with one another, a positive correlation is initially visible: If the requested policy change is assessed positively by one group, the comparison group also agrees to a greater extent. However, the level of agreement can differ significantly, and the extent to which the responses from two groups differ from one another depends on how far apart the incomes are. Attitudes are most similar among respondents in the bottom two income groups. The number of questions with larger differences of opinion increases gradually the further the incomes differ. Although there is a positive correlation between the proportions of responses in all cases, there are still numerous factual questions on which income groups disagree.

In order to get a systematic impression of how great the differences of opinion between income groups are, the absolute difference between the proportions of the answers of two groups is calculated for each question. Fig. 1 shows the mean, the standard deviations from the mean and the range of differences of opinion for different income groups. The attitudes of four income percentiles are compared with those of respondents whose household income is in the lowest percentile. As the income gap increases, so do the differences of opinion, which shows both higher mean values ​​and higher maximum values. While on average only a few percentage points separate the answers of the lower two income groups, this value increases to more than 16 percentage points when the lowest group is compared with the highest group. On some questions there is 50 percentage points between the two ends of the income distribution.

Looking at the response behavior of respondents with different incomes alone, it cannot be conclusively answered whether the deviations shown are large. This only becomes clear when compared with the differences between other social groups. In order to enable a direct comparison of the differences of opinion, Fig. 2 compares the average differences between income groups with those of other comparison groups. In addition to the income groups, the left sub-graph also contains the difference between East and West Germans, men and women, those under 30 and over 60, and respondents with a high or low level of education. This figure shows that respondents from different income percentiles differ more in their attitudes than other comparison groups.

In the right partial graphic of Fig. 2, differences of opinion between occupational groups are compared with those of the income groups. In the occupational groups, unskilled and semi-skilled workers serve as a reference group. The differences of opinion between the occupational groups shown on the vertical axis and this reference group are shown. Three points deserve special attention. First, the trend confirms that the greater the social distance between two comparison groups, the greater the size of the differences of opinion. Second, the differences in attitudes between skilled workers and ordinary employees, on the one hand, and unskilled and semi-skilled workers, on the other, are greater than those between respondents in the 1st and 10th percentiles. On the other hand, the differences to the better-off with either a high income or from higher professional groups are similar. Thirdly, the comparison with the left part shows that the differences of opinion between socio-economic groups are greater than those between other social groups.

So far we have analyzed all the questions together, but it is possible that differences of opinion are particularly large in some policy areas and smaller in others. Where differences of opinion are more pronounced, the scope for selective responsiveness on the part of political decision-makers is particularly large. The left partial graphic in Fig. 3 shows for six policy areas how much the response behavior of respondents with different incomes deviates from one another. The graphic shows that there are no policy areas in which income groups have identical attitudes. In all six areas, opinions differ so widely that political decisions can one-sidedly reflect the preferences of individual groups.

In the right part of Fig. 3, the size of the differences of opinion is shown again for six policy fields. This time, however, not only income groups but also respondents from other categories are taken into account. Three observations catch the eye. First, the differences between occupational and income groups are greater in all policy areas than between East and West Germans or men and women. The differences of opinion among respondents with different educational qualifications are also lower, with the exception of migration policy. Second, the differences in social policy are smaller than in other policy fields, while this is also the policy field in which young and old give particularly strongly divergent answers. This could be a sign of a change in social values ​​that cannot be observed in other policy areas.

The evaluations carried out so far show that income influences political attitudes and the greater the social distance, the greater the difference between positions. Since the underlying categories do not always clearly distinguish between different groups and, in particular, the income of the respondents is only approximately determined, a cautious estimate of the differences of opinion can be assumed, which underestimates the actual differences rather than overestimating them. The next section examines whether political decisions reflect citizens' preferences.


In this section, we first examine the question of whether public opinion is generally reflected in the decisions of the Bundestag and, second, examine whether the opinions of financially better off groups are given greater consideration than those of less well off groups. In order to investigate the extent to which public opinion and political decisions coincide, a survey was carried out for each individual factual question as to whether the policy change presented in the question actually occurred or not. The database records whether the policy change occurred up to two years or up to four years after the survey. The following analysis only uses the information from the two-year period, as it was not yet possible to conclusively determine which political decision was made four years after the survey for the surveys from 2012 and 2013. For the remaining years, all analyzes were carried out with the variables for both periods in order to check whether the results were changed by the selection of the period. The results are little different, as almost 90% of policy changes, if implemented, were implemented within two years of the survey.

In a first step, a logistic regression is used to examine the relationship between the average agreement of all respondents to the proposed policy change and the occurrence of the policy change. The result is shown in Tab. 1 in the first column. Although the coefficient has a positive sign, the relationship is not statistically significant in the estimated model. This could either mean that politics in general is only slightly oriented towards the opinions of the population, or that it takes into account the interests of some groups but not those of others, which would then balance out across all respondents.

In order to get to the bottom of this question, the connection between the opinion of individual income percentiles and policy changes is examined in a second step. It is estimated what effect the approval of those in the middle of the income distribution (50th percentile), those in the bottom 10% and those in the top 90% of the income distribution has on the implementation of policy change.Footnote 6 The results of the respective logistic regressions are given in columns 2 to 4 of Table 1 and shown graphically in Fig. 4. They show a clear picture: the higher the income, the more agree political decisions with the opinions of the respondents. For example, if only 20% of the top income bracket are in favor of a policy change, then the probability that it will be implemented is 39%. However, if 80% of respondents from the upper income group agree to a policy change, their chances of implementation are almost 65%. This clearly positive and statistically significant relationship only applies to the top income group. The relationship between approval and implementation is also slightly positive for the middle income group, but only very weak. There is even a slightly negative correlation for the lowest group. For middle and low incomes, there is no significant correlation between the degree of support for a policy change on the one hand and the likelihood of its implementation on the other. Politicians are not responsive to these two groups.

The findings so far show an imbalance in responsiveness in favor of upper income groups. However, there are many topics or questions on which the opinion of different income groups does not differ widely (see above). In these cases it cannot be determined whether or not politics is geared its decisions more towards the higher income brackets, because everyone wants the same thing.

But how does politics decide when income groups have conflicting interests? This question is examined in the next step, in which we only consider those questions in which the differences of opinion between the income groups compared in each case amount to more than ten percentage points.Footnote 7 On the one hand the lowest with the highest percentile and on the other hand the median with the highest percentile are compared. The results of the logistic regressions are shown in Table 2. First, it is noticeable that there are significantly more often large differences of opinion between the top and bottom income groups than is the case between the median and the top percentile. For this reason, the number of cases varies depending on which groups are compared with each other. Second, it becomes clear that the relationship found above is stronger if we only look at questions where opinions differ widely between the income groups. For questions on which the bottom and top income groups differ greatly in their opinions (columns 1 and 2 in Table 2), there is a clear connection: In the top income group, the approval of policy changes is strongly positively correlated with their implementation, negative in the lowest income group. This means that the likelihood of implementation actually decreases if more people from the lowest income group support a certain political decision. However, the relationship found is only statistically significant for the top income group. In any case, these results show that politicians systematically reflect the opinions of upper income groups in their decisions when there are major differences of opinion. This finding also applies when the top income group is compared with the middle income group. In columns 3 and 4 of Table 2 we see that there is a very similar relationship here.

The results described are shown graphically in Fig. 5, which again shows the estimated probability of a policy change depending on the proportion of supporters. In the left part of the figure, this probability is shown for the lowest and the highest income group, if they do not agree. The right part of the figure shows the results for the middle and top income groups. They are almost identical to the comparison between the lowest and highest groups, which makes the overall result of this section seem even more serious. All in all, we can now state that the politics of the Bundestag reacts more frequently to the views and concerns of the upper income bracket, whereas the opinions of the lower and middle income brackets are hardly taken into account or even ignored. This is particularly evident when we look at cases where the top income bracket advocates different policy choices than the other income brackets.

In order to systematically record how responsiveness is related to the size of the differences of opinion, in the third step we again estimate the same models, but insert an interaction between the opinion of the lowest and middle income groups on the one hand and the respective difference to the highest groups on the other. This procedure makes it possible to determine from what degree of differences of opinion there are significant differences in responsiveness. The results are contained in Tab. 3 and are shown in Fig.6 shown graphically.

In the left part of Fig. 6, the marginal effect of the lowest income group is shown depending on the differences of opinion with the highest income group. There is only a significantly positive effect if the settings differ only minimally from one another. However, if the attitudes between rich and poor differ by more than 15 percentage points, there is a negative effect for the bottom group, but a significantly positive effect for the top group. Lower income groups can only hope for a match between their own preferences and political decisions if those with high incomes want the same. If they differ, however, political decisions reflect the preferences of high-income groups. A very similar pattern can be seen when comparing the middle with the top income groups, which is why we do not use a separate graphic representation.

Finally, it could be assumed that different government coalitions are responsive to different degrees to the income groups we examined. In the years from 1998 to 2013 there were three coalitions, which were led by the SPD until 2005 and then by the CDU. In order not to reduce the number of cases too much, we only divide the total period into two periods, which include the Schröder and Merkel governments. We carry out the same analyzes as before for these two comparison periods. Even with this separate analysis, the pattern does not deviate from the previous results, although the estimate is somewhat less precise due to the small number of cases. But in both the SPD and CDU-led governments, low-income groups can only hope that their concerns will be implemented if the differences of opinion with the highest income group are small. If the preferences differ significantly, the governments follow the high-income citizens (Fig. 7).

discussion of the results

Our research has shown that income influences political opinions. In a large number of cases, respondents with low incomes would like different political decisions than their higher-paid fellow citizens; this applies in particular to foreign policy, but also to economic and social policy. In addition, we were able to show for the first time for Germany that political decisions are more likely to agree with attitudes in higher income groups, whereas for low-income groups either no systematic agreement can be found or even a negative correlation. What low-income citizens wanted in particularly large numbers had a particularly low probability of being implemented between 1998 and 2015. The pattern of systematically distorted decisions that has been proven for the USA also applies to Germany.

At the beginning, some possible explanatory models were already mentioned, to which we want to come back here. For the USA, Gilens (2015a) makes the argument that politics there is increasingly dependent on private donors, while at the same time the proportion of small and medium-sized donations in election campaign funding is declining and this is increasingly being borne by a few super-rich. It is therefore not surprising that their influence on politics is growing (Gilens 2015a, p. 226). However, due to the predominantly state election campaign funding, this argument is unlikely to be relevant to the German case.

The situation is different with other mechanisms discussed in the literature, such as unequal political participation. Political engagement has become more unequal in almost all western democracies over the past three decades, as people with low incomes vote far less often than those with high incomes (Schäfer 2015; Solt 2008). If it is assumed that the interests of those who actively participate politically are given greater consideration in the political process, this can result in unequal responsiveness and lead to a downward spiral in which social and political inequality mutually reinforce one another. Socially disadvantaged groups notice that their concerns are not being heard and are therefore turning away from politics, which are then even more oriented towards the interests of the better-off. This argument could apply equally to the USA and to European countries. The same applies to the argument that (business-related) organized interest groups would exert a strong influence on political decisions and change politics in the interests of capital and the better-off (Gilens and Page 2014; Hacker and Pierson 2010). Another mechanism discussed in the literature is the overrepresentation of the upper classes in the legislative bodies (Carnes 2013; Mansbridge 2015), which has increased almost everywhere in recent decades (Best 2007). The high proportion of academics in the Bundestag is particularly noticeable, but the occupational structure also differs significantly from that of the population as a whole.Footnote 8 As Carnes (2013) and Carnes and Lupu (2015) show, MPs from different social classes represent different positions in economic policy. The more parliaments become socially more homogeneous, the greater the likelihood that certain positions will no longer be represented or only insufficiently represented. However, corresponding research results are not yet available for the German case. In order to be able to make more precise statements about the extent to which the social origin of the members of parliament influences the decisions of the parliament, further research is necessary that deals with how members of the parliament or parliamentary groups reach their decisions and what role the individual socio-economic background and external influences play in this.


  1. 1.

    An earlier version of this essay was presented at the “Democracy Revisited” conference from October 28th to 30th, 2016 in the Tutzing Academy. We thank the participants for numerous helpful comments. At the University of Osnabrück, Rabea Mette worked very competently on the creation of the ResPOG database, for which we would like to thank her warmly. Finally, we would like to thank two anonymous reviewers from the Zeitschrift für Politikwissenschaft, who gave us very constructive advice to help us improve our contribution.

  2. 2.

    Responsivity differs from the concept of “congruence” in that responsiveness not only requires the political positions of the parliamentarians to agree with the majority of the population, but rather a connection between the will of the population and the political one decisions (Lax and Philipps 2012).

  3. 3.

    We thank the Federal Ministry of Labor and Social Affairs, which supported this research project financially, and the employees of infratest dimapwho have prepared the previously inaccessible datasets for analysis.

  4. 4.

    With the data sets available to us, it is not possible to calculate the net equivalent income, since the DeutschlandTrend only determines the income of the entire household without asking about the number of people living in the household at the same time. However, it cannot be assumed that this will result in a systematic distortion of the results. Using current data from the general population survey of the social sciences (ALLBUS 2014), it can be shown that the weighted income is highly correlated with the household income that does not take the number of people into account (r = 0.94; p = 0.000, N = 3061). It follows that the information available to us also reliably distinguishes low incomes from medium or high incomes. As in other surveys, however, very high incomes are not recorded.

  5. 5.

    The comparison with the income categories originally collected in the DeutschlandTrend shows that the effects do not change significantly if we use the existing income values ​​instead of the ones we calculated.

  6. 6.

    In the following we only use the 10th and 90th percentiles instead of the 1st and 99th percentiles. We are following Gilens' (2009