German Parties’ Positions Mostly Match the Preferences of Their Voters. The Exception Is the SPD.

The GLES data are a gift that keeps on giving. Today, we are looking at preference profiles of voters of different parties. To recap, here is what the overall German electorate looks like:

Distribution of voters’ preferences with positions of political parties

Voters’ preferences are distributed on two dimensions, a left-right dimension that describes economic positions (from 0 = “left” = “more taxes, more social transfers” to 10 = “right” = “less taxes, less social transfers”), and a societal or “new politics” or “identity” dimension that stretches from 0 = “left” = “Green, Alternative, Libertarian” to 10 = “right” = “Traditional, Authoritarian, Nationalist”.

We can look at these preferences separately by parties. In the GLES survey, respondents were asked which party they voted for in the 2017 federal elections. We can use this information to distinguish between the preference profiles of voters of different parties. Note that the absolute numbers of self-proclaimed voters are different across parties. Warmer colors denote a higher density of voters in a location only relative to the numbers of voters of that party,  not relative to all voters.

Distribution of voters’ ideal points, by party

The figures reveal important differences between the ideal points of voters of different parties:

  • The AfD is the party with the most extreme distribution of voters, with voters concentrated mainly at the right-wing identity politics edge.
  • In contrast, voters of all three “old” parties, i.e. the established parties of old West Germany before ca. 1983 – CDU/CSU, SPD, FDP – are concentrated in the center of both distributions.
  • The Left party’s voters are the most dispersed, with preferences reaching into the right-wing extremes on the societal dimension.
  • The Green party’s self-proclaimed electorate is the most clearly concentrated “post-materialist”, even though it comprises a solidly center-left mass on both dimensions.

If we compare these distributions to the positions of parties, we find that the SPD is the party that is farthest away from the mass of the distribution of its voters. This is corroborated by statistical  analysis: The SPD is the only party for which the distance between a voter’s ideal point and the party’s position is not a significant predictor of vote choice. In other words: Those who voted for the SPD did not vote for the party because it was close to their preferred positions, but despite the fact that it was far away from their ideal points.

More specifically, the political positions of the SPD are too “left” on both dimension for the liking of their voters. Those voters who hold positions similar to those of the SPD vote for either the Greens or the Left party. Given the overall distribution of voters’ preferences, any further moves to the “left” are likely to result in a further decrease of  the SPD’s political support.

Spatially speaking, the SPD performed better than expected in the 2017 German Federal Elections, while the CDU did worse. And the AfD is the least-liked party of them all

There is a long political science tradition of modeling voting decisions as based on which party a voter is “closest” to in a policy space. This space can, in principle, be of any dimensionality; for both theoretical and empirical work, the assumption of a two-dimensional policy space is frequently used.

Voting in such a framework is simple – voters calculate the distances between their own ideal point in the policy space and the positions of all parties competing for votes. They then vote for the party that is closest to their ideal point. There is no consideration of e.g. how likable a candidate is or whether a voter feels particularly attached to a specific party. All that matters is distance. In addition, voters’ preferences (their ideal points) are fixed and exogenously given and there is no abstention.

We can use this model to gauge which parties over- or underperformed in the 2017 German Federal elections, given their policy stances and given the stated preferences of voters. For voter’s ideal points, we turn to the GLES post election survey data and use the questions on  preferences over taxation and immigration to measure the ideal point of a voter on an economic and a societal dimension, respectively. For parties’ positions on the same two dimensions we use data from the Chapel Hill expert survey. Here is what this looks like:

Political parties are positioned around the center of the distributions on the two dimensions. Warmer colors denote that more voters with those preferences can be found in these regions of the policy space, while the dark blue color that can be found towards the extremes of the distributions indicate that there are few voters whose ideal points are located there.

Some parties, like the CDU/CSU and the SPD, are closer to where the bulk of voters are, while others like the FDP, the AfD, the Left party and the Greens, position themselves more to the edges. How should these positions translate into votes? If we assume voting based solely on minimum-distance considerations, the SPD should have gotten a mere 14.3 percent of the votes instead of the 20.5 percent it actually received. In contrast to this over-performance, the CDU/CSU should have received 39.1 percent of the votes instead of the 33 percent it ended up with. The table shows these numbers for all the parties currently in the German Bundestag

Party Expected Minimum Distance Vote Share Actual Vote Share Difference
CDU/CSU 39.5 33 -6.5
SPD 14.3 20.5 6.2
GREENS 9.8 8.9 -0.9
LEFT 7.9 9.2 1.3
FDP 9.9 10.7 0.8
AFD 18.5 12.6 -5.9

While the Greens, the Left party and the FDP performed about as predicted by a minimum-distance model, the AfD performed much worse than minimum-distance considerations would suggest.  One of the reasons for the discrepancy between predicted support and electoral performance might be the likability of parties. The GLES data contain information on how voters view parties. Respondents are asked to voice their opinion on parties by using an 11 point scale with a range from -5 to +5 where -5 stands for “I don’t think anything of this party at all” and +5 denotes “I think very highly of this party”. This question goes beyond mere distances in a policy space. It also covers emotional and longer-term attachment to political parties. 

Using this measure, we find that the AfD is the least liked party with an average value of 2.77 on the -5 to +5 scale. A full 60 percent of voters chose the worst possible judgment for the AfD. By comparison, the SPD has a mean favorability value of 7.4. However, we find an almost identical number for the CDU. Even considering the somewhat lower values for the CSU, the favorability rating cannot explain why the CDU/CSU underperformed their minimum-distance spatial expectations.

Maybe, then, it was all about the candidates? Comparing the favorability ratings of Chancellor Angela Merkel and then-candidate for the Chancellory Martin Schulz of the SPD, we find that the difference between Schulz and Merkel is about 1.25 in Merkel’s favor. In other words: If Merkel had been seen as unfavorably as Schulz, the CDU/CSU would have done even worse. On the other hand, a candidate Schulz with Merkel’s favorability rating would have outperformed the SPD’s spatial forecast by an even wider margin.

Looking ahead, these findings are good news for the CDU and not so good news for the SPD. With the SPD already outperforming their levels of support suggested by a spatial perspective, it is not clear how the party could increase their vote share, especially not if the SPD moved to a position further to the left. For the CDU, the party’s longtime strategy of steering a calm course could just be what is needed to bring voters back who should have voted for the party based on their preferences, but didn’t do so in the 2017 elections. In any event, neither party stands to gain from moving further to the extremes.

Did the CDU move to the left? Hardly, but AfD voters think it did

We are still pondering the implications of the 2017 federal elections in Germany that propelled the far-right AfD to become the third-strongest party in the federal parliament. As new data on party positions and voter perceptions are becoming available, we took a look at the purported leftward shift of the Christian Democrats, Angela Merkel’s CDU.

With the party leadership’s decision to open up the vote to allow for the introduction of marriage equality, including the possibility for gay and lesbian couples to adopt children; the decision to end nuclear power in Germany; and, most importantly, chancellor Merkel’s 2015 decision to open Germany’s borders for about 1.1 million refugees and migrants, the previously conservative party seemed to have veered far into the left field, particularly on a policy dimension that can be described as GAL/TAN –“Green, Alternative, Libertarian” vs. “Traditional, Authoritarian, Nationalist” (Hooghe, Marks, and Wilson, 2002). 

We can look at two kinds of position data to see if the claim of a leftward shift of the CDU is actually true. The first data source employs objective criteria in coding party manifestos to gauge parties’ positions. We used data from the Comparative Manifesto Project  to compare the CDU’s position over time. The GAL/TAN variable we constructed takes positive mentions of a “national way of life”, “traditional morality”, “law and order” and negative mentions of “multiculturalism” as indicators of more right-leaning positions, while negative mentions of a “national way of life”, “traditional morality”, “law and order” and positive mentions of “multiculturalism” are indicators of more left-leaning policy positions. Normalizing the resulting indicator to a range from 0 (left-most position) to 100 (right-most position) puts the CDU at 45.35 for the 2017 elections. In 2013, this number was 44.68 – indicating that there was hardly any change at all. Going back further in time, the CDU is at 42.05 in 2009 and at 43.17 in 2005. This is certainly not a pronounced leftward shift.

This result is corroborated by data from a second source, the Chapel Hill Expert Survey (CHES). The CHES data are compiled by asking political scientists what they think the positions of political parties in their country of expertise are.  According to the CHES data, the CDU changed its position on the GAL/TAN dimension only moderately, from 6.0 in 2014 to 5.8 in 2017 on a 0 to 10 scale where 0 stands for the left-most position and 10 for the right-most position.

Here is the interesting part: This notion of a party that remained essentially stable in their policy position on a GAL/TAN dimension is not shared by voters of the AfD. Using representative survey data from the German Longitudinal Election Study (GLES), we see that AfD voters see a pronounced shift to the left in the GAL/TAN position of the CDU.

While, in 2013, non-AfD voters placed the CDU at 6.79 and voters of the AfD saw the CDU at 6.48 on the GAL/TAN dimension (the GLES data use a 1 to 11 scale, with 1 denoting left, 11 denoting rightward positions), this perception changed massively in 2017. AfD voters placed the CDU at 4.98, while non-AfD voters placed the party at 6.1. Put differently, the perceptions of voters who didn’t vote for the AfD are broadly in line with objective and country expert data on the political position of the CDU, while AfD voters massively overestimate the leftward shift of the CDU.

This perception of a a pronounced leftward shift of the CDU on a new politics dimension by AfD voters also renders the CDU and the SPD more alike. In 2013, the perceived difference between the parties on the GAL/TAN dimension was 1.74 among non-AfD voters and 1.99 among AfD voters. In 2017, these differences changed to 0.87 and 0.58, respectively. The difference between AfD voters and non-AfD voters in the perception of the CDU in comparison to the SPD thus increased between 2013 and 2017. This is evidence of a successful mobilization that portrayed the CDU as being “too far” away from AfD voters’ ideal points and the AfD taking up the slack. The table below shows a summary of these results, together with data on the perception of party positions on a general left-right dimension and and economic dimension.

Obviously, perceptions matter in politics. One interpretation of these findings is that AfD voters were looking for a reason to rationalize their support for a party that is at odds with much of the post-war political consensus in Germany. Casting the CDU as having swung massively to the left may just have done that trick.

On the map: Turnout and party support in the 2017 German federal elections

Following up from yesterday’s post, we can also look at a map of Germany to see where the electoral losses for the  center-right CDU/CSU were most severe. The party incurred their most severe losses in parts of Saxony and Bavaria, former party strongholds, but also in the Southwestern state of Baden-Württemberg.

Change in electoral support (party vote) for the CDU/CSU in the 2017 German federal elections

By contrast, the far-right party AfD increased their vote share most markedly in East Germany and parts of Bavaria.

Change in electoral support (party vote) for the AfD in the 2017 German federal elections

Finally, we can look at the change in turnout. Turnout increases were largest in Bavaria and parts of East Germany. As we have seen,  increased mobilization benefited the AfD and hurt the CDU/CSU.

Change in voter turnout in the 2017 German federal elections

Interpretation:  The AfD was able to asymmetrically mobilize voters that supported their anti-immigrant and anti-establishment positions. In September 2017, less than two weeks before the election, more than 80 percent of AfD supporters said that the “AfD is the only party through which I can express my protest against current policies”. The same share of AfD supporters agreed that the “AfD does not solve any problems, but at least they tell it like it is.” (Source: ARD-DeutschlandTREND). This, together with the turnout analysis, suggests that the AfD mobilized previously alienated voters by taking positions that were not represented by other parties in the political arena.

How voter mobilization benefited the far-right German AfD

The September 24, 2017 German federal elections saw the far right, anti-immigrant and euro-skeptical party “Alternative for Germany” (“Alternative für Deutschland” – AfD) surging to third place behind chancellor Merkel’s center-left CDU/CSU and the Social Democrats. With the AfD’s success the focus of most observers, an interesting phenomenon has been largely overlooked, namely the fact that the AfD disproportionately benefited from increased turnout. We analyzed some of the available data at the level of electoral districts and found that the AfD was particularly strong in those districts that saw the largest increase in voter turnout.

The following table shows results from a regression analysis that models vote share as a function of a number of covariates, including the change in electoral participation (the turnout variable). Focussing on model 3 where the AfD’s vote share is the dependent variable, we find that the AfD massively benefitted from an increase in turnout. Increasing the change in turnout (i.e. mobilization) from its mean by one standard deviation leads to a one percentage point increase in AfD votes.

Determinants of Electoral Success, German Federal Elections 2017
(1) (2) (3)
Change in Turnout from 2013 0.214* -0.872*** 0.510***
(0.116) (0.119) (0.103)
Population Density -0.000*** -0.000*** -0.000***
(0.000) (0.000) (0.000)
Disposable income per capita 0.000* -0.000*** 0.000
(0.000) (0.000) (0.000)
Higher Education (percentage Abitur) -0.002*** 0.001*** -0.002***
(0.000) (0.000) (0.000)
Unemployment -0.006*** 0.011*** 0.003**
(0.001) (0.001) (0.001)
Percentage migration background -0.002*** -0.001*** 0.001***
(0.000) (0.000) (0.000)
East -0.073*** -0.134*** 0.112***
(0.008) (0.008) (0.007)
Constant 0.427*** 0.286*** 0.107***
(0.032) (0.033) (0.028)
Observations 299 299 299
R2 0.718 0.724 0.729
Adjusted R2 0.711 0.717 0.723
Residual Std. Error (df = 291) 0.032 0.033 0.028
F Statistic (df = 7; 291) 105.830*** 108.835*** 111.919***
Notes: ***Significant at the 1 percent level.
**Significant at the 5 percent level.
*Significant at the 10 percent level.

This is substantively important, given that the AfD fell short of clearing the threshold for parliamentary representation by just 0.3 percent in the 2013 federal elections. On the other hand, the SPD massively lost from an increase in turnout, showing that the party was not able to speak to its voters.

Of the other variables, we observe a number of significant effects. Most importantly the AfD is stronger in the east. On average, going from a Western district to one in the East increases AfD vote share by more than 11 percent.

Secondly, the AfD is stronger in less populated areas, i.e. outside the big cities. The AfD is also stronger where voters are less educated and where unemployment is higher.

Finally, the AfD is stronger in districts where the share of the population with a “migration background” is higher. It is important to note that the Federal Returning Office defines “migration background” as follows: “Migration background means foreign nationals plus all those Germans who came to Germany after 1955 plus all those Germans with at least one parent who came to Germany after 1955” – “Als Personen mit Migrationshintergrund werden alle zugewanderten und nicht zugewanderten Ausländer sowie alle nach 1955 auf das heutige Gebiet der Bundesrepublik Deutschland zugewanderten Deutschen und alle Deutschen mit zumindest einem nach 1955 auf das heutige Gebiet der Bundesrepublik Deutschland zugewanderten Elternteil definiert.” (Federal Returning Office). Immigrants from Russia who came to Germany in the early 1990s and who disproportionately support the AfD could explain this effect. (see, for example, here.)

How important was each of these variables in relation to each other? To see this, we can turn to z-score standardized coefficients (beta coefficients) and their graphic representation on the following figure:

The strongest predictor in relative terms is the East dummy, then the education variable, followed by population density, the change in turnout from 2013, the percentage of people with migration background, and finally the unemployment variable. Note that the variable that captures disposable income is not significant.

We can conclude by pointing out that the AfD success cannot be explained by economic grievances. The variable that captures income is not significant and the unemployment variable is a poor predictor once other factors have been taken into account. What matters most is the division between Eastern und Western districts. In the East, many voters seem to be alienated from established parties and possibly from the system as a whole. The AfD with their anti-system rhetoric was able to mobilize these voters. The geographic and educational divide of the electorate is reminiscent of the situation in other countries, for example in the US.