"Bavarian voters are cool customers, while Lower Saxony residents have a sunny disposition.
When parties search for reasons for their poor performance in elections, the weather is often at the top of the list. Exception: the recent federal election. Meteorology was not used as an excuse. However, statistics truly show: Good weather is bad for democracy.
Warm and sunny? Then the eligible voter, preferably the Bavarian one, would rather go to the beer garden than to the ballot box; the Berliner enjoys themselves at the lake. Cold and rainy? Then nobody wants to ruin their day further by placing two crosses on dreary lists in gloomy school gymnasiums, enveloped in the stale smell of sweat. The spirit of the voter was willing, but the sitting endurance was weak… Why not question such claims critically for once? Data on this is freely available on the internet. Wahlrecht.de publishes voter turnout for all federal and state elections. Wikipedia documents the exact election dates, and Darksky.net provides the corresponding weather. And voilà: There is a significant, meaning statistically explainable beyond random fluctuations, correlation between weather and voter turnout. The warmer it is, the fewer citizens cast their votes.
Five federal states are highly temperature-sensitive
When separated by the 17 types of elections, 16 state parliament and federal elections, the influence of temperature is also significant in five federal states. This result shows much more than would be expected through alpha-error cumulation. "Not significant" does not mean that there is no correlation. It may simply mean that the data does not (yet) suffice to confirm it with sufficiently high certainty. That means, there is a significant correlation between temperature and voter turnout. Remarkably, the direction of influence varies between the types of elections. In Bavaria, Berlin, and Saarland, voters are more likely to cast their votes in cool temperatures, while the opposite is true in Hesse and Lower Saxony. (Causal interpretations, such as speculations about the available alternatives for Sunday activities, are left to the reader.)
The influence is much weaker in federal elections. Here, statistics can only establish the correlation but not causally explain it: correlation does not imply causation. Perhaps the elections are more important, and people are more willing to sacrifice an hour of their free time for them? Or maybe the nationwide weather varies regionally too much, so the effects overlap?
Statistical thinking means: Don't believe everything you see - at least not right away. An alternative explanation immediately suggests itself: The correlation may simply be due to the tendency towards decreasing voter turnout over the years, which coincidentally coincides with a gradual warming of the climate, or it may be explained by shifts in the months in which the elections were held. Fortunately, there are statistical tools that can account for and exclude such alternative hypotheses. Regression analysis shows: Temperature remains a significant factor even with control variables for year, month, and weather type.
Mathematically, each degree Celsius costs our democracy 0.46 percentage points of voter turnout. With weather changes that can easily result in differences of ten to 15 degrees Celsius within a few days in September, that's potentially more than five percentage points. Thus, temperature is more important than democracy fatigue, as the average voter turnout decreases by only 0.3 percentage points each year. This means: Good weather is bad for democracy.
Politicians often emphasize how important voter turnout is, but they are only interested in how much they benefit from higher or lower turnout. Since Peter Altmaier's advice to protest voters to stay at home instead of voting for the AfD, there should be no doubt about this anymore. How fitting that it is the Union that claims to suffer more from holidays and good weather. But is that true?
A first look at the raw data seems promising: Clear trends emerge for almost all parties in the Bundestag – except (ironically) for the CDU/CSU. However, if you again control for the interfering influences of year and month, not much remains of the power of the weather. Not a single significant result is left for parties with prospects for the Bundestag. However, some other parties show voters who are motivated by good weather to cast their votes. Particularly noteworthy is the ÖDP, whose voters might be expected to be in the great outdoors on weekends and in sunshine.
While the influence of temperature on the absolute results is not significant for any party, such an influence can be observed in the ratios of the results. We initially focus on the two major parties, the Union and the SPD, because their results are the most stable due to the high number of elections and because they also show an apparent drifting tendency. Compared to the Union, the SPD loses significantly, by 0.2 percentage points with every degree. With the possible fluctuations in September, this means that there can be easily two percentage points change in the result. This can make the difference between Black-Yellow and Grand Coalition.
As a consolation for the SPD, it remains: If only federal elections are considered, the result is significantly weaker and not significant. However, this may not be as comforting as the comrades would like. Firstly, there are far fewer federal elections. Thus, it is automatically harder to obtain significant results. Secondly, the FDP performs significantly better than the SPD; they gain 0.1 percentage points with every degree. Thus, statistically speaking, the unusually warm election Sunday is at least partly responsible for the rather cloudy future of the Grand Coalition.
(This article appeared in the print edition of "Tichys Einblick" in October 2017.)"Video: