Matt Yglesias asks "What is it about the economy?" that impacts election outcomes, and links "Enik Rising", an interesting blog by a political scientist, who finds that income growth matters but that unemployment does not matter for mid-term election outcomes.
Having done a powerpoint slide on this issue once, I can say that this issue is actually trickier than it looks. His mistakes are three: First, he looks at change in house seats as the variable he's trying to explain. So, it would make sense to include "how many house seats the president's party holds" as a control variable. Second problem is that the Democrats usually lose about 8-12 seats in midterm elections simply because the share of young people, women, and minorities all decline during mid-term elections. Of course, young people, women, and minorities all turned out at record rates in 2008, but will sit out the midterms. (Good news for Dems is that they will be back in 2012... There is a lot of habit persistence in voting behavior, but it's really specific to the type of election.) But I digress. Third issue is that there are just too few data points here, and US politics has changed too much since 1912 to gain much by expanding the series. One thing the author could do is include Presidential election years, and that would help. Also, that the Bush 02 year is a chief counterexample to the "unemployment doesn't matter" is quite telling. Obviously, 2002 was an election dominated by 9/11. The Republican gain/loss was also helped out by the fact that they only started off with a slight majority, and that Republicans traditionally do well in midterms. Control for those two things and the Republican performance in 2002 loses a bit of its luster... But there's no way, really to control for 9/11 since it was a one-time event. Which means someone needs to write a careful international paper. And when they do, my guess is that they'll find that a change in unemployment matters a lot.
After all this criticism, however, I agree with part of the bottom line, that GDP growth matters more than unemployment. Why? The big reason why the economy matters has to do, I'm convinced, not with the actions of laid-off workers but with how the media covers the economy, the president, and the federal deficit. The deficit gets covered like it's a huge scandal, and it always comes across as though the President and Congress have been reckless with the nation's finances, even though the truth is that running smaller deficits would have been much more reckless. Since GDP and the stock market rebound before unemployment, of course, this means that GDP growth is a better indicator, because media types care much more about their stock portfolios and bottom-line GDP growth than they do unemployment, which is simply a remote statistic to them.