The weight of Brexit: Areas with high obesity rates more likely to vote Leave
29/06/2016 5 Comments
I saw a presentation today that had an obesity map of the UK on one of its slides. Someone made a flippant remark about how much the map resembled to map of EU referendum results. And indeed, looking at the data confirmed that areas with high obesity levels were much more likely to vote Leave even when controlling for income, health, education, economic activity, and age. It would be naive to think that body-weight itself is a driver of voting preferences. However, it is a measurable characteristic that could help us understand the determinants of voting behaviour.
[It has crossed my mind to hide this post. Although its general reception has been positive (or neutral), I have also received messages how spiteful I was, that I’m part of the ‘elite’ who ridicules the less fortunate, etc (anyone who knows me personally knows how far my background falls from the ‘elite’). Eventually I decided to leave it on as evidence that nothing in my writing is spiteful. I have no reason to ridicule anyone and if you take the time to read my post you will find that I simply present a correlation between AREAS with high obesity rates and voter preference. Nothing in this post suggests that leave voters are obese, or that obesity is something to be ridiculed. Readers might project their own biases onto this post, but that doesn’t mean that the post itself is biased.]
I’m sure there will be numerous candidates who will offer us in-depth analyses on the outcome of the EU referendum. I have no intention of taking the fun away from them, all I do here is present a few patterns of a briskly compiled dataset. This is not full-fledged research, it’s simply a brief explanation of the data attached below.
The first thing I wanted to look at was whether there was correlation between obesity and the EU referendum vote. Luckily, data is available on the % of people voting on Leave/Remain in each English local area (England has 326 districts – I focused only on England because obesity data was most readily available for English districts), and Public Health England provides the obesity data.
Here are the top 10 districts with the highest and lowest percentages of Leave votes and the corresponding percentages of obese people (BMI 30kg/m2 or over). The percentage of obese people in each district varies between 13 and 35%. It does indeed appear that high proportions of Leave votes are associated with high proportions of obese adults.
Looking at all 326 districts the correlation between the % of obese adults in a district and the % of Leave votes is high, with a correlation coefficient of 0.8 (0 implying no correlation and 1 perfect correlation). Scatter plots below confirm this. On the left hand side you can see how the percentage of Leave voters decreases with the percentage of adults with healthy weight. The right hand side plots shows that the percentage of Leave voters increases with the percentage of obese adults.
OK, so the knee-jerk reaction will be: yes, but this is all driven by other factors. For example poorer areas tend to have more obese people and poorer areas were more likely to vote Leave and so on. So I quickly downloaded district level census data on some of the most obvious factors that might be affecting both the referendum vote and the % of obese adults. The factors I looked at were: income, the variation of income (my crude measure of inequality), the level of health of adults, the level of education of adults, and the economic activity of adults.
The table below shows that obesity is still a strong determinant of Leave votes. A 10 percent increase in the number of obese people leads to a 6 percent increase in the number of Leave vote. More adults with higher education and being economically active has a negative effect on the proportion of Leave votes.
So what’s going on here? My immediate thought was that obese people have poorer health and are probably more likely to worry about NHS access, which was one of the flagship Leave campaign arguments. But the results below reveal the opposite effect – areas with more healthy people have a larger % of Leave votes. Very peculiar.
|Variation in average income||0.00863***||-9.06|
|% Very good health||0.721***||-4.69|
|% Level 4 education||-1.196***||-20.89|
|% Economically active||-0.273**||-3.07|
|* p<0.05, ** p<0.01, *** p<0.001|
It is unlikely that being overweight itself affects people’s voting preferences. Instead this is probably driven by personality traits that characterise both Leave voters and obese adults. A Swiss study on obesity and personality traits reveals 5 important personality traits that affect eating disorders. Extrovert, open, conscientious, and agreeable people are less likely to have eating disorders, and neurotic people are more likely to have such disorders.
So can we jump to the conclusion that Leave voters are less likely to be extrovert, open, conscientious, or agreeable, and more likely to be neurotic? This work argues that personality traits are indeed important drivers of voters’ preference. This other post makes similar claims but at country level (credit to Chris Hanretty for the tip).
It would be naive to think that body-weight itself determines voting preferences. But body-weight and life-style choices are things that we can easily measure. If we know the policies that influence the prevalence of obesity in local areas, or the sort of personality traits that certain physical characteristics imply, it could help us understand the drivers, and psychology of voting behaviour much better. Certainly worth further exploring.
Data downloaded from the Office of National Statistics, Nomis (official labour market statistics), Public Health England, and the Electoral Committee. Data used for this blog available here.
Caveat 1: age was one of the main determinant of the referendum vote so it might appear to be an obvious omitted variable. However age doesn’t seem correlated with obesity so it’s unlikely to be biasing the obesity coefficient. Here I included a table where the other variables (except income) are further classified into various age groups. One thing that stands out is that health status, qualifications, and economic activity does not make any difference for the youngest and the oldest age groups. It appears preferences are set for these two age groups and demographic characteristics do not alter these preferences (i.e. the young tend to vote Remain and the elderly tend to vote Leave irrespective of their health status, qualifications, and economic activity).
Caveat 2: Of course one could control for a catalogue of other things from census data but as I highlighted before, this is not a full-fledged analysis, rather some food for thought. The same applies for the choice of model specification, etc.