Does Ticket Splitting Empower White, Old, Well-educated, Rich Men?

Evidence from Germany

When German voters elect a new parliament on September 24th, citizens will have two votes: one for a candidate in a single member district and one for a party list. In 2013, the major parties (Social Democrats, SPD, and Christian Democrats, CDU/CSU) received only 56.8 percent of the party votes, but gained 98.3 percent of the 299 single member districts. Only in 15.4 percent of all districts, a candidate from the left party (die Linke), the green party (Bündnis 90/Die Grünen), or the liberal party (FDP) performed second best (own calculation based on information provided by Bundeswahlleiter 2013). These figures suggest that many voters decide to split their ticket: They give their candidate vote to one of the two major players and their party vote to a minor one. A first strategic motivation for this type of voting behavior follows from the aim to make both votes count. Voters favoring small parties do not want to waste their district vote for runners unlikely to win and support their favorite candidate from a major party instead. Coalition formations constitute a second strategic consideration leading to ticket splitting. Citizens supporting major parties might give their party vote to a minor party they would like to govern with their preferred party to ensure that it passes formal thresholds (Gschwend 2007). According to the CSES post-election survey, 16% of the German voters split their tickets in 2013.

This blog contributions aims to explore whether all citizens stand equal chances to split their ticket. There are reasons to expect that women, the poor, the less-well educated, young people, and citizens of immigrant origin have a lower motivation and competency to make such a strategic vote choice. First of all, these groups might be less likely to believe into their ability to make efficient decisions about ticket splitting. After all, this type of voting behavior requires a high level of information about preference distributions in their districts and coalition formation processes. Women, the young and the poor are less likely to trust their political decision-making capacity. In addition, citizens with these attributes tend to interact less with parties and other organizations that provide information about strategic decisions. Hence, I would expect a lower probability for ticket splitting for this set of groups. If such a relationship persists, it would have profound implications: White, old, well-educated, rich men – who are also more likely to be present in parliaments, would use their vote more efficiently so that ticket splitting would re-inforce their dominant position and inequality in representation.

Data from the 4th wave of the CSES in Germany, which was conducted after the 2013 Bundestag election, allows predicting the chance to split the ticket for citizens with different characteristics. The dependent variable takes the value ‘1’ if respondents supported different parties with their district and party vote and ‘0’ otherwise. As explanatory variables, the models include gender, being born abroad, the level of education, household income, and age as well as their interactions. Additional controls include typical variables predicting turnout (political knowledge, East/West, ideological positons, support for the statement that voting makes a difference) and ticket splitting (existence of a party identification, district winning party matches party vote, party vote for minor party). I estimated multi-level logistic regressions with districts nested in states and random effects for both levels.

Table 1 presents four models predicting the chance to split the ticket and shows considerable variation across the different groups. In contrast to my expectation, the young are more likely to split their tickets than the old. Every additional year of living decreases the chance to vote for two different parties about 1.5 percent. The models further reveal that the electoral behavior of women and the foreign born barely differ from men and the domestically born. Even though the coefficients of the binary variables for gender (women) and birth abroad indicate that these characteristics increase citizens’ likelihood to split their tickets, the effects lack statistical significance in Model 1, 3, and 4. I did neither find any indication that these variables interact with factors such as age or political knowledge. However, Model 2 clarifies that at the cross-section of gender and immigrant status, the chance to split tickets is considerably lower than for men born abroad and women born in the country. The double disadvantage emerging from gender and the immigration experience thus constitutes a barrier to ticket splitting.

Table 1: Odd ratios for the chance to split tickets depending on membership in traditionally excluded groups, political knowledge, and probability for strategic voting.

Annotations: All models are multi-level logistic regressions with districts nested in states and random effects for both levels. The figures show odd ratios and the level of statistical significance is indicated as follows: a p < 0.15, * p < 0.10, ** p < 0.05, *** p < 0.01. All models also control for living in the East of Germany (for all coefficients p<0.05), left-right placements (for all coefficients p<0.05), support for the statement that voting makes a difference (for all coefficients p<0.01), and party identification (for all coefficients p>0.15).

The level of education – as expected – increases the chance to cast the ballots for different parties. Well educated are more likely to make strategic choices. Higher incomes also unfold a positive impact, even though the coefficient lacks statistical significance. As Model 3 and 4 reveal, political knowledge moderates the relationship between education and income on the one hand and the likelihood to split tickets. Figure 1 visualizes their combined effects. Low educated and poor citizens stand the lowest chances to make use of this instrument if they have little information about politics at hand. Yet, if they show a medium-high or high level of political knowledge, they are as likely as rich and well-educated citizens to vote for two different parties.

Figure 1: Marginal effects of education and income on the predicted probability for vote splitting.

Annotations: Based on Model 3 and 4. Outer lines show 95% confidence intervals.

These findings lead me to conclude that ticket splitting does enhance the unequal representation of some politically excluded groups, but not all of them. In particular foreign born women are unlikely to vote for two different parties, but also the poor and low educated, in particular if their political knowledge is limited. By contrast, young, female domestically born, and male foreign born citizens optimize their impact on the selection of candidates through ticket splitting and can through this mean compensate – to some degree – for their lower level of presence in parliament. This implies that group-specific factors are at hand and explain the frequency with which different groups of citizens make use of this instrument – a topic certainly worth exploring in more detail.

References

Bundeswahlleiter (2013) 'Endgültiges Ergebnis der Bundestagswa'. Available at: https://www.bundeswahlleiter.de/dam/jcr/ce166f8c-6137-4ed3-baa7-95abc5c093c1/btw13_arbtab4.pdf [accessed 24 August 2017].

Gschwend, T. (2007) 'Ticket-splitting and strategic voting under mixed electoral rules: Evidence from Germany', European Journal of Political Research, 46 (1): 1-23.