Map of life expectancy at birth from Global Education Project.

Tuesday, June 15, 2021

You never see a correlation like this . . .

 ... in social science. Well, almost never. From Seth Masket:


Image 

The correlation coefficient (Pearsons r) for this association is .867. Defining the correlation coefficient in mathematical terms is a bit complicated -- it's the sum of the products of the difference from the mean of x and y for each observation, divided by their standard deviations. If the variables were completely unrelated, it would be zero; if they were perfectly correlated, it would be 1. As you can see, it's very close to 1 -- all of the observations are close to the straight line, called the regression line. Where Biden got 30% of the vote, less than 50% of adults have gotten a shot. Where he got 70%, more than 80% have gotten a shot.

 

Just looking at this doesn't tell you why these variables are so closely related. It obviously isn't that voting for Biden causes you to get a vaccination. But something about the states where people voted for Biden, or the people who voted for him, or both, are associated with getting vaccinated. Getting vaccinated is good because then you don't get sick and you don't make your neighbors sick. But the extent to which this association is a function of choice vs. opportunity we don't exactly know.

 

 



1 comment:

Don Quixote said...

I had a short newspaper clipping on my fridge twenty years ago. It was about research into a study that showed a correlation between childhood obesity and children who watched, say, at least four hours a day of TV. But the blurb ended by noting that the researchers couldn’t determine if the children were obese because they watched so much TV — or if obese children were just more likely to watch a lot of TV!