Fred Barnes discounts the role of the economy in the outcome of the midterms:
Yes, the economy is always a factor in elections. But a wretched economy doesn't automatically doom Washington's ruling party to disaster in a midterm election. Since World War II, the average midterm loss by the president's party is 24 House and four Senate seats. In 1982, despite a deep recession and joblessness above 10%, Republicans lost only 26 House seats and none in the Senate. The difference between 1982 and today is that President Reagan's policies—cutting spending and taxes, firing striking air-traffic controllers—were popular.
Another pretty important difference is that Republicans were only defending 192 seats in 1982. The higher your seat total, the deeper you're stretched into hostile territory, and the easier it is to lose seats.
Put differently, after the 1982 election, Republicans controlled just 166 House seats. That would be quite a disaster for Democrats, wouldn't it?
Obviously, nobody claims that the economy controls political outcomes entirely. You can have your Watergate scandal or your 9/11, and smaller events, that play a huge role. But the relationship between presidential popularity and economic conditions is so strong it's amazing we have to debate this: