Austrian General Elections 2019: Preference votes

elections stringr tidyr::pivot Austria

A look at the distribution of preference votes casted in the 2019 general elections.

true
11-20-2019

Context

On 29 September, Austria held its general elections to the national assembly (Nationalrat). By now a flurry of analyses and comments has been published and the caravan of political commentary has largely moved on, at least when it comes to the analysis of the results itself (the government formation train is finally also getting up to speed). One aspect, I personally have never looked into, and which seems to fall a bit under the radar, is the use of preference votes by Austria’s electorate. This post is essentially my first ‘exploratory’ go at preference votes. No specific question, no specific theoretical framework in mind, just poking. I’ll first present the results, then detail some of the steps implemented in R to obtain, analyze and visualize the relevant data. The entire code for the analysis is available on my github account. As always, if you spot any glaring error etc, don’t hesitate and let me know.

Preference vote (Vorzugsstimme) refer to voters’ option (but not obligation) to indicate their support/preference for specific candidates running for the party of their choice. Since Austrian candidates run on a nation wide federal district list (Bundeswahlkreis), one of nine state district lists (Landeswahlkreis), or/and one of 39 regional district lists (Regionalwahlkreis), voters can cast up to three preference votes (here a sample ballot slip). Eventually, a candidate’s number of preference votes can be consequential when it comes to distributing the seats which a party won to the individual candidates. Without preference votes, seats are distributed in accordance to candidates’ position on the electoral list. The lower a candidate is on a list, the lower the likelihood that she can actually get a seat. However, with a sufficient large number of preference votes, the pre-election order on the electoral list can change and initially lower ranked candidates can move up and eventually gain a seat. For details see here.

Results

by district list

Overall, there were 1,8 million preference votes cast. The largest part (more than 60 %) were cast for candidates running on regional district lists. Probably this could be read as voters been particularly keen to support ‘their’ local/regional candidates to make it into the new parliament. Ties between voters and regional candidates are likely to be different than those between voters and candidates running on the more ‘removed’ federal or state constituency list.

by constituency list and party

If we further disaggregate these numbers and differentiate between parties, we see that preference votes obtained on the regional level accounted for the majority of preference votes for all parties. However, there is a noticeable difference between the comparable low 55.9 % of the Freedom Party (FPÖ) and e.g. the 65.7 % of the Social Democrats (SPÖ) or the 64.1 % of the NEOS. Or put differently, with 34.1 % the share of preference votes obtained on the federal level, the Freedom Party stands out. As for the underlying reasons, I can only speculate. One reading could be that voters of the FPÖ have weaker ties to their candidates on the regional or state level and hence felt less inclined to cast a preference vote for them. Another one could be that the relation among FPÖ’s candidates on the federal list were particularly competitive and hence a disproportionate high number of voters were mobilized to cast their preference votes on the federal list. Considering the exit of former party leader HC Strache (Ibiza…) and the subsequent (not yet consolidated?) change in the leadership, this seems plausible (see below). It however does not explain why FPÖ voters did not cast preference votes for candidates on state or regional lists.

Top 5 candidates

If we are interested in candidates, which candidates were particularly successful in securing preference votes?

The graph above clearly demonstrates the predominate role of former chancellor Sebastian Kurz when it comes to preference votes. With more than 150,000 votes his result clearly outclasses those of any other candidate. With the notable exception of the FPÖ, the leading candidates/party leaders of all parties succeeded in securing most of the preference votes. In the case of the FPÖ, former Minister of Interior Herbert Kickl overtook his party’s leader, Norbert Hofer. What is also noticeable is that party leaders almost exclusively secured their preference votes from the federal constituency list, while their colleagues largely relied on votes from the regional constituency list.1

Looking at the rankings within parties, I have to admit that I was a bit surprised by the strong performance of a some candidates which I at least had never heard of. The result of T. Mete for the SPÖ, and J Hechenberger for the ÖVP were particularly surprising to me.

Regional dimension

by district magnitude

The plot below displays a party’s total number of preference votes as percentage of a party’s total vote in each regional electoral constituency. Let’s call this for the sake of easier reference parties’ preference votes - party votes ratio. Or to predend a serious color, P-PVPV ratio.2 This PVPV ratio are grouped by constituencies’ magnitude (number of mandates). The motivation behind is the assumption that the fewer mandates are available the higher will be also intra-party competition. With a smaller cake to distribute (fewer mandates), candidates will try particularly hard to win preference votes and move up the electoral list. Here the analysis is limited to the 39 regional electoral lists. To avoid the indivudal dots overlapping, some random vertical variation was introduced.

The plot seems to provide at least some backing to this proposition. The constituency with only 1 mandate to compete for (Osttirol) features the highest median of parties’ PVPV ratio. While the picture is not entirely clear cut, there seems to be indeed a decreasing median share as constituencies’ magnitude increases from to two to seven mandates. The constituencies with eight and nine mandates (Graz-Umgebung, Hausruckviertel, OberStmk), however, do deviate. Hover over the individual dots to get details.3

by state

If we are interested in whether there are any regional differences when it comes to parties’ PVPV ratios, grouping them by states can be of some help. As the boxplots below show, on average parties’ PVPV ratio was the largest in Burgenland followed by Vorarlberg. These two states are Austria’s two smallest in terms of population. On the other end, Upper Autria (Oberösterreich), Lower Austria (Niedrösterreich), Styria (Steiermark), and Vienna featured on average clearly lower PVPV ratios. These states are Austria’s largest. This difference would be in line with the proposition that candidates in smaller states are better in mobilizing personal support, i.e. preference votes. Closer candidate-voter relations due to smaller population size sounds rather probable to me. But this would certainly need further analysis to corroborate the claim.

Intra-party dynamics

As already outlined above, preference votes are first and foremost interesting from an intra-party perspective. With sufficient votes, candidates which embarked on the electoral campaign from a lower ranked position can move up the intra-party ladder and secure a seat.

So when do candidate’s move up the ballot list?4

Did this happen often?

Candidates’ preference vote party vote ratio

The graph below presents candidates’ preference votes - party votes ratio, or, to stay serious clumsy, C-PVPV. This ratio is candidates’ number of preference votes expressed as % of the total number of votes of her party (in the respective constituency/list). The vertical red lines indicate the stipulated thresholds which candidates have to pass in order to move up on the party’s electoral list. Hover over the individual points to get the data pertaining to each candidate. Note that the x-axis is log-transformed to ensure candidates with low C-PVPV ratios remain (somewhat) visible. Note also that only those candidates are included which actually obtained at least one preference vote.