In a previous comment I discussed some thoughts of Paul Krugman and Matthew Kahn on why US Republicans are so hawkishly opposing any environmental regulation. I add some thoughts of my own to this HERE but still felt somewhat unsatisfied. And decided to take all this to the data. You wouldn’t expect my surprise when I found the following.
Basically, Paul Krugman and Matthew Kahn suggest that the Republicans are anti-green because they are the rich 1% who can, firstly, protect themselves against environmental impacts, and secondly, because they believe they lose more since the higher taxes needed for environmental policies will be mostly levied upon them. I then added that the rich 1% don’t really make up 47.2% of the US society, and wondered HERE where the rest of the 46.2% votes come from. I suggested they may be insecure poor who are fooled by the rhetoric of the Republicans that they lose their jobs if environmental regulation is increased. However, these things should show up in the data, right?
So I took the data from the World Value Survey, where every couple of years a representative sample of individuals from each country is asked about personal opinions and characteristics. I took this data and ran a regression, trying to estimate the underlying reason for voting for Republicans. The output of the regression for the Wave 5 (2010-2014) is shown below.
In order to interpret the coefficients, one should look at the following, explained at the variable sex: Simply said, the Odds Ratio states whether the odds of voting Republican for a female (female=1, male=0) respondent are higher if the Odds Ratio is above 1. Thus, we see that females may be inclined to vote for the Democrats, while males tend to vote Republican. A second information is given by the 95% Confidence Interval. If the lower end of the confidence interval is below one (for sex it is .43), while the upper end is above one (for sex at .79), then the results are not statistically significant, so we really cannot say much. In this case, the results for sex are statistically significantly different from each other. Let’s thus look at what the regression tells us.
I present the results from Wave 5 (2010-2014), those from Wave 4 (2005-2009) tend to be sufficiently similar. I add also a remark at the beginning if the result is weak or strong (in terms of statistical significance)
- (weak) The richer a respondent the more likely (s)he is going to vote Republican. Thus, the idea of Paul Krugman and Matthew Kahn is correct in that there are hints that the rich tend to vote for Republicans, but the results are not statistically significant.
- (strong) Females tend to vote less for Republicans.
- (strong) the more highly educated respondents vote Republican. Since there is a correlation between income and education, either of the two variables may capture part of the effect of the other.
- (strong) The more important environmental issues are for voters, the less likely they are going to vote for Republicans! Thus, there is a real partisan issue going on here. If you are anti-green you tend to be anti-democrat. (variable green comes from a principal component analysis of variables B001, B002, B003; the higher is green the less one cares about the environment).
- (strong) Housewifes, retired and self-employed tend to vote Republican.
So here is a (mostly) clear result: Those that are highly educated, anti-green, male, most likely rich and housewifes tend to be more likely to vote Republican. So yes, Paul Krugman and Matthew Kahn were sort of right, if we believe that the highly educated also tend to be richer, and thus the education variable captures some of the variation of being rich. But this still leaves the important question: How can this small, rich minory get such a strong base of supporters, especially if they tend to propose policies that are in their own interest? Why then do the 46.2%, those that are poorer follow the lead of the self-interested, anti-green, highly-educated/rich and vote Republican? As we can see from the regression above, being unemployed or poor does not present itself as a statistically significant factor here. So what could?