Tuesday, 11 April 2017

Right-leaning politicians are better looking, but we know less about right-leaning scholars

Two recently published research papers caught my attention. Both compared the relative attractiveness of people on the right and left of the political spectrum. One of the research papers was good, and one was decidedly less so.

Let's start with the good paper, which was written by Niclas Berggren and Henrik Jordahl (both of the Research Institute of Industrial Economics in Sweden) and Panu Poutvaara (LMU Munich), and published in the Journal of Public Economics (ungated earlier version here). In the paper, they compare the relative attractiveness of politicians on the right and on the left. Why would anyone care about this? The authors explain:
If one side of the political spectrum has a beauty advantage, it can expect greater electoral success and to have political decisions tilted in its favor. We put forward the hypothesis that politicians on the right look better, and that voters on the right value beauty more in a low-information setting. This is based on the observation that beautiful people earn more... and that people with higher expected lifetime income are relatively more opposed to redistribution...
This is a very nice study. The authors first demonstrate that politicians on the right are indeed more attractive than politicians on the left, using data from Australia, the European Union, Finland, and the United States. Then they:
...study beauty premia in municipal and parliamentary elections. The former can be regarded as low-information and the latter as high-information elections, where voters know little and reasonably much, respectively, about candidates. We show that in municipal elections, a beauty increase of one standard deviation attracts about 20% more votes for the average non-incumbent candidate on the right and about 8% more votes for the average non-incumbent candidate on the left. In the parliamentary election, the corresponding figure is about 14% for non-incumbent candidates on the left and right alike.
They argue with a nice theoretical model that the reason for these differences is based on two things: (1) attractiveness is itself valuable, and voters are more likely to vote for attractive candidates; and (2) attractiveness signals that politicians have views that are further to the right. So, this explains why the attractiveness premium is greater for politicians on the right in low-information settings (where both effects work in the same direction) compared to politicians on the left (where the effects work in opposite directions, since left-preferring voters are more likely to see an attractive left candidate as being to the right of their views).

Finally, the authors confirm their results with an experiment:
Experimental election results confirm the observational findings from real elections. When matching candidates of similar age, the same gender and the opposite ideology in a random manner and asking respondents whom they would vote for solely on the basis of facial photographs (i.e., with low information), we find that candidates on the right win more often because they look better on average. Candidates on the right get a higher vote share, both from voters on the right and voters on the left, but with larger success among the former.
One of the cool things the study does is that it uses a different survey group to assess the attractiveness of the politicians from those who provided an assessment of whether the politicians were from the left or right, and different from those participating in the experiment. This ensures there is no cross-contamination across the study (they also got Europeans to rate the American politicians, and vice versa). The conclusions, that politicians on the right are better looking, that attractiveness is a cue for voters as to a candidate's conservatism (in a low-information election), and that attractiveness confers an extra benefit for a right-leaning politician in a low-information setting, are all fairly robust.

On to the second, not-so-good paper, by Jan-Erik Lönnqvist (University of Helsinki), and published in the journal Personality and Individual Differences (sorry I don't see an ungated version anywhere). This follow-up study (Lönnqvist cites an earlier version of the Berggren et al. paper):
...sought to investigate whether the attractiveness advantage of the political Right is specific to politicians. To investigate this, the attractiveness of Right-leaning scholars (referring to people who professionally engage in mental labor, such as academics or writers) was compared to that of Left-leaning scholars.
The data collection was ok, although both the subjective assessments of the attractiveness of the scholars and of their political orientation were both provided by the same respondents (being five research assistants, compared with the hundreds in the Berggren et al. study). However, that isn't the main problem with the study, which is this:
We first regressed the ideological tone of the magazine on ratings of attractiveness and perceived political orientation.
So, in all of Lönnqvist's regression models, actual political orientation is the dependent variable (the variable he is trying to explain), and subjective perception of political orientation is one of the explanatory variables. This creates two big problems that I can immediately see. First is related to interpretation. I'll try to keep this from getting too pointy-headed, but if you have the actual variable on the LHS of your econometric model, and a subjective measure of the same variable on the RHS, you think you have this model:

[Actual Political Orientation] = f{[Perceived Political Orientation], other stuff}

What that equation says is that actual political orientation is a function of perceived political orientation and some other stuff (which includes attractiveness). But actually what that equation really is, is a rearranged version of this:

[Actual Political Orientation - Perceived Political Orientation] = f{other stuff}

So, really this is a model of deviations between actual and perceived political orientation (being a function of other stuff, including attractiveness). Keep that in mind when you read the results:
...perceived political orientation accurately predicted actual magazine-related political orientation. However, physical attractiveness did not.
Then Lönnqvist splits the attractiveness variable into attractiveness and grooming ("the extent to which the target person appeared to have prepared physically for the photograph") and finds:
Perceived political orientation still predicted magazine-related political orientation, but now also physical attractiveness and target grooming predicted magazine-related political orientation.
In other words, there was a negative correlation between attractiveness and right political ideology, and a positive correlation between grooming and right political ideology.  Lönnqvist argues this means that left-leaning scholars are more attractive, while right-leaning scholars are better groomed.

However, go back to the equations above. What these results really mean is that the difference between actual and perceived political orientation is more negative for attractive scholars. This may be because attractive scholars are more left-leaning, or because the research assistants who did the rating thought that the attractive scholars were more to the right than they actually were! Notice that this result could easily just be in line with the Berggren et al. results above - people believe that better-looking people have a more right-leaning political orientation.

Similarly, the results demonstrate that the difference between actual and perceived political orientation is more positive for well-groomed scholars. This may be because well-groomed scholars are more right-leaning, or because the research assistants who did the rating though that the well-groomed scholars were more to the left than they actually were.

So, the results of the second paper tell us very little about the attractiveness of scholars and their political ideology. At the very least, we'd need to know about the results, excluding the subjective rating of political orientation from the models.

[HT: Weird Science on the NZ Herald, but I had seen the Berggren et al. paper earlier but I forget the source]

No comments:

Post a Comment