The myth of turnout

As part of our election forecasting project of the 2015 Croatian general election, during the two months of pre-election build up we were required to analyze the four previous electoral cycles and find some underlying trends that could be helpful in our forecasting model, but also to bust some deeply embedded domestic political myths. One of such myths was the impact of turnout on electoral results.

There is an almost tautological view amongst domestic political analysts, the general public, the media, and the politicians themselves, that higher turnout benefits the left political parties, i.e. the center-left coalition led by the SDP (Social-Democrats), while lower turnout implies a much better electoral result for the right parties, primarily the HDZ (Conservatives), which allegedly has a more disciplined voter base. This myth was most likely garnered after an impressive electoral victory of the SDP-led coalition in the 2000 elections, when turnout was significantly above average (2.9 million compared with the usual average of 2.4 million; in fact election turnout in Croatia was only once more so high – in the 1991 independence referendum, when it was slightly above 3 million). This led many pundits to believe that higher turnout was directly responsible (highly correlated) to the left coalition’s high margin of victory. However, if we look at the data (Figure 1), it points to no such relationship. The coalition has won in 2000 due to a variety of other factors, turnout was not one of them.

Figure 1. Turnout and vote share for HDZ (blue dots) and the SDP+HSLS coalition (red dots) for the 2000 elections

Ever since the 2000 elections, turnout has been considerably lower, but the electoral results differed. We depict a two-variable relationship between turnout (in %) and votes received (in %) across the municipality level for each given electoral year. We therefore show static representations of the given relationship. The problem with observing turnout dynamically in Croatia is lack of data. There were 5 general elections to date (under the same PR-system rules), which means we only have 5 data points for which we can dynamically observe the relationship between turnout and electoral success. This is not nearly enough to make a viable conclusion, so we are faced with observing a static picture rather than a dynamic one.

In the cases of 2007 and 2011 elections there is no apparent trend noticeable (we take these two years since they had a very similar total turnout, but different electoral winners). The same conclusion is reached when looking at all other elections (Presidential, EU Parliament, etc.). Turnout does not benefit neither of the two parties. There is not even a correlation present, let alone anything else.

Figure 2. The difference between turnouts and the difference between total vote share for HDZ (blue) and SDP (red) in 2007 and 2011 elections.

One exception to that rule were the latest Presidential elections in 2015. In this case we could have noticed a marginal benefit of higher turnout towards a single candidate. The reason could be that these were direct elections between only two candidates (second round run-off), in which the entire country was a single electoral district (which is not usually the case). The challenger and current President Kolinda Grabar Kitarovic from HDZ beat the incumbent Ivo Josipovic from SDP by a very small margin of roughly 30 thousand votes, after the incumbent Josipovic had a similarly small lead after the first round (amid one of the lowest turnouts ever in the first round of Presidential elections – only 1.7 million). The prevalent opinion among domestic pundits was that all Josipovic, as the center-left incumbent, needs to do is to increase turnout and he could be certain of victory. Turnout was higher. An extra 600 thousand people turned out to vote in the second round (total turnout a bit over 2.3 million). However higher turnout did not help the center-left incumbent, but the center-right challenger. We look at the data on the level of voting stations (a total of 6500 of them), and a slight trend is noticeable in favor of Grabar Kitarovic.

The graph shows the difference in % of votes for Josipovic and Grabar Kitarovic (y-axis) with respect to total turnout (x-axis) for each voting station. Even though this represents only a slight trend in favor of the center-right candidate, it is interesting to notice that areas of overwhelmingly high turnout did not favor the left, but the right instead.


Figure 3. The graph shows the difference in % of votes for Josipovic and Grabar Kitarovic on presidential elections with respect to total turnout for each voting station.

Finally we also observe this relationship on an electoral district level. We again look at the data on voting station level, but within each of the 10 electoral districts in Croatia. Unlike the national level where we found no apparent trend, within a few electoral districts we did notice some trends that could testify of a positive relationship between turnout and electoral result. In most districts, across most electoral years, there is no apparent trend, but in districts that can be considered as strongholds for each of the two major parties, higher turnout does help the dominant party. In particular this means that SDP will fare better in the 8th and 3rd electoral district (west and north), and HDZ in the 5th and 9th district (east and south), which represent their electoral strongholds. The thing is, once all of these districts are taken together on a national level, the individual district trends cancel each other out.

We use the following heat maps on a voting station level for each district to illustrate our point.

Heat maps illustrate the two-variable relationship where denser clustering around some point is painted with warmer colors. The denser clustering around the middle points to no relationship between turnout and votes, while a longitudinal clustering points to an existence of a trend where higher turnout increases of decreases the vote share of a given party. Also note that some of the clustering pointing out to a trend in one electoral cycle does not have to hold for the next one.

In conclusion, we can say that turnout does not in general affect electoral results, and that it has a minor influence only in districts which are party strongholds. In such districts the dominant party which mobilizes its support can achieve a significant electoral advantage over its opponent, which however does not necessarily have to translate into a national electoral victory.

Share This