State x State

US News & World Report has out this week its rankings of the states. The Best States, it says, based on several categories. What categories? Health care, education, crime, infrastructure, opportunity, economy, and government. How those are measured, you have to dip into the methodology, but for the moment let's accept the measures as reasonable.

Here's what's weird unless you read the fine print. Not every category carried the same impact.

Health care and education were weighted most heavily. Then came the opportunity states offer their citizens, their crime & corrections and infrastructure. State economies followed closely in weighting, followed by measures of government administration.

And now, some partisan politics.

Four of the five overall top-ranked states went for Clinton. "Yeah," you say, but what about all those categories? How'd Trump do there, or Clinton?" The answer, Clinton dominates nearly every category in the top states, the exception being the one on the health of state government, where Trump captured four of the top five states. That's interesting in and of itself, suggesting Trump did best in states that are well run even if they have other, clearly significant problems.

Top Five Rankings By Category and Who Won Each State

	Health	Educ	Crime	Infrast	Opportunity	Economy	Govt
1	Clinton	Clinton	Clinton	Clinton	Clinton	Clinton	Trump
2	Clinton	Clinton	Clinton	Clinton	Clinton	Trump	Clinton
3	Clinton	Clinton	Clinton	Trump	Clinton	Clinton	Trump
4	Clinton	Clinton	Clinton	Clinton	Trump	Trump	Trump
5	Trump	Clinton	Clinton	Clinton	Clinton	Clinton	Trump

Okay, fascinating in a certain kind of way, but an argument can be made that Trump's upset victory may have come from the states at the bottom of the list. This is indeed the case. On health care, for example Trump won the 13th worst states in that category (Arkansas being #50). Trump won three of the five worst on education, three of the worst five in crime, four of the worst five in infrastructure, four of the worst five on opportunity, four of the worst five on economy, but only two of the last five in lousy state government.

In overall rank, Trump did well on the state's scoring the lowest. Four of the worst five went for Trump, and nine of the worst 10. The states that appeared to have made all the difference on Election Day in going for trump, none of them make the Top 10 and most are far down in the pack, like Pennsylvania (30th), Michgan (33rd), Ohio (35th), North Carolina (25th) and Wisconsin (16th).

So what's it all mean? Clinton did well in states doing well, at least comparatively speaking. The average overall rank for states voting for Clinton was 16.8, while the average overall ranking for states voting for Trump was 31.3. That alone says a great deal. Trump did well in, to put it bluntly, crappier states, or at least states in which people had fewer opportunities and more struggles.

On a whim, I ran a logistic regression analysis on the data. In other words, I wondered which of the various "best state" factors mattered most in predicting that state going for Trump or Clinton.
If you're a statistical nerd and want more info, email me, otherwise I'm going to gloss over the PhDweeb stuff and get to the results.

Keep in mind the rankings on the various categories are statistically controlling for one another, kind of a pit fight to see which ones truly matter. So which do? No surprise, Health Care wins (Wald = 7.4, p<.01), and the Health of Government also achieves statistical significance (Wald = 4.0, p<.05). The lower the rank in health care, the more likely a state was to vote for Trump. The better the government's healthy ranking, the more likely that state went for Trump. None of the other factors matter when put in competition with one another. Using these macro data, states went for Trump if the residents had lousy health opportunities but their government was reasonably well ran. If I were more of a political scientist, I could dig more, but for the moment we'll have to settle for an overlong blog post.

Data available upon request