,
15 tweets,
4 min read
Read on Twitter
NEW from me: What if everyone had voted in 2016? We have quantified the left-leaning nature of non-voters and project that Hillary Clinton would be president today under such a scenario. But the findings run deeper than this headline. THREAD 1/14 economist.com/graphic-detail…
2. We first thought up this piece as a simple addition to our annual "World If" magazine supplement, but it soon ballooned into something larger. It seemed that nobody had suggested a satisfactory answer to the question of what would happen in US politics if everyone turned out.
3. We hypothesized that the US would drift more to the left under universal voting, but by how much? Though non-voters are more diverse & thus left-leaning than voters, there are plenty of working-class whites who don't turn out either. Our math says they make up ~50% of adults.
4. There is also the challenge of figuring out how much national shifts would have impacted key states, which ultimately decide the electoral college and thus the presidency. No typical method could provide an answer. We had to think outside the box, but we did find a solution.
5. I used a statistical method called multi-level regression and post-stratification (Mr P) to predict voting patterns among different demographic and geographic groups under varying hypothetical turnout scenarios. (I discuss the method in a Medium post to be released soon.)
6. In a nutshell, we trained models of voting behavior on survey data from the Cooperative Congressional Election Study, a 2016 poll of 64,600 adults. We then predicted voting habits with granular demographic data from the Census Bureau's American Community Survey.
7. With MRP we can determine how many additional votes candidates would have won in 2016 if 100% of certain groups cast ballots. See here how Trump makes gains when all working-class whites are added to the electorate. We did this for every group (interactive coming shortly). 
8. We can go further and break down these extra voters at the state level. Adding them up gives us a projection of Clinton and Trump's vote share in every state under a hypothetical mandatory/universal voting scenario. From there we assess electoral college victory!
9. We can go further, though, and quantify the exact changes in the probability that Hillary Clinton would have won certain states when the electorate shifts from actual 2016 voters to all adult citizens. Here's a small clipping of the print graphic. She goes from 288 to 337 EVs.
10. Note that there are bigger changes in more diverse states. Swing states like Florida and North Carolina would become Democratic-leaning. Texas, too, shifts far to the left—further even than PA, WI and OH and into likely Dem territory. That's b/c TX non-voters are very Dem.
11. In order to stay competitive, the political parties would have to adjust quickly to this new, more liberal electorate. Republicans would bear much of this burden, probably being forced to shift many of their policy positions to the left. They'd have several paths to victory.
12. Notably, even under 100% turnout, Republicans retain a structural advantage in the electoral college. If they double-down on working-class whites—a likely scenario—they can feasibly go back to a scenario where they win they win more electoral votes but lose the popular vote.
13. So not only have we solved an impressively difficult mathematical/statistical problem, but we have also further quantified how biased America's antiquated electoral system becomes when partisanship and geography become intertwined—a favorite topic of the newspaper.
14/14. This exercise taught us a lot about American voting behaviour, a lot about its electoral systems and even more about how we can use exciting new statistical methods. We hope you share our enthusiasm. Read here: economist.com/graphic-detail…
