Demographics is not destiny – election edition

Next week Tuesday, Americans will go to the polls to determine who controls Congress for the next two years. I have no special insights into the midterm elections, so I am not going to divulge my views here.

But it does give me another opportunity to go back to my mantra that demographics is NOT destiny. In my view, demographic changes are important but so slow that many, MANY other factors will dominate them.

You see, in spring, I was rattled enough by an opinion post in the New York Times to bookmark it. In that post, Thomas Edsall claimed that while the Democratic Party has a good chance of winning the 2028 Presidential election, the demographic changes in the US are such that the Republicans will likely increase their inbuilt advantage in the electoral College from 2032 to 2040.

His main argument is that the 2030 census will remove votes in the Electoral College from states with declining populations like New York and California and give more votes to states with increasing populations like Texas and Florida. Because New York and California are reliably blue states and Texas and Florida reliably red, this could shift some 12 votes in favour of Republicans until the next census comes along in 2040.

Nice prospects if you are in the blue camp like me.

But of course, what this analysis implicitly assumes is that voters in the 2030s will vote along the same demographic lines as they do today. So, it assumes black and highly educated voters vote reliably Democratic, while older voters and working-class white people vote reliable Republican.

But you see from this description that this is a brave assumption. Old people tend to die (such is life) so who knows if the old people in the 2030s will still be voting reliably Republican or swing more to the Democrats. Also, working class white people were historically a key constituency for the Democrats. Who know if they will still shun the Democrats in ten years?

Meanwhile, Hispanics are currently a demographic that tends to vote more Democrat than Republican. But they are also highly religious and may well switch into the Republican camp over time if they are alienated by Democratic policies.

And so on.

Which is why I recommend a new paper by Richard Calvo and his collaborators. They have analysed how well election predictions based on demographic variables worked in the past. The charts below show what would happen if you used demographic trends at the last election to forecast the next election or use demographic trends to forecast three elections from now as Edsall did in his piece.

The dark black lines show the forecast based on demographic considerations and the light grey line the actual election result. As you can see, there isn’t any resemblance between the forecasts and the actual results. In other words, demographic forecasts of election trends are a complete waste of time.

Demographic forecasts vs election outcomes