In Practice, State-By-State Analysis Of Presidential Elections Doesn't Tell You Much

Formally speaking, the presidential election is a series of state-by-state elections. But Jonathan Bernstein chides up and coming pundits and Dylan Matthews for giving too much credence to the idea that we should be peering into the specific details of the employment situation in key swing states.

The case against worrying too much about the Electoral College is summed up by this chart from Andrew Gelman. He shows that despite claims from the Obama campaign to have redrawn the electoral map, what they actually did was do slightly better than John Kerry all across the country:

You'll note that there's absolutely no sign that Barack Obama crafted a message designed to appeal to voters in Rhode Island, didn't buy ads in the Providence media market, didn't do a broundbreaking GOTV targeting effort in Rhode Island, etc. But he did better in Rhode Island than John Kerry did, by approximately the same extent as he improved in the swing states which, again, was about the same as how he did in a deep red state like Idaho.

That said, what this chart primarily says to me is that state effects are so small as to be uninteresting except in the interesting elections! The fact that the pro-Obama swing was bigger in North Carolina and Virginia than in Ohio and Missouri wound up not mattering because the 2008 election wound up not being very close. But Ohio was still a bigger state than North Carolina and a more Democrat-friendly one. You can easily imagine an alternative, closer version of the 2008 election in which instead of Obama narrowly winning North Carolina he ends up narrowly losing Ohio. In that case the fact that Obama got a relatively strong swing in North Carolina and a relatively weak one in Ohio would have been very interesting and we'd probably have spent a lot of time focused on the fact. In other words, these idiosyncratic state effects are generally quite modest in their scale, but the winner-take-all nature of the Electoral College means that in a close election even a very small effect could be a big deal.