Historically republican voters have been more likely to turn out in midterm elections, so getting (say) 50-50% in polls of registered voters in polls means the Republicans win the election. Also, the out-party outperforms the generic polls, and this year they are the out party.
Here is the historic relationship between poll results and vote outcomes in congressional elections from famous Columbia political scientist Andrew Gelman. It can be used to predict the results in the 2010 election.
Excluding Rasmussen, the republican advantage on the generic ballot is 50.6% to 49.4% according to Real Clear Politics. This means “Generic vote Dems minus 50%” is -0.6%. Let us err on the side of caution and say it is exactly 50-50.
We are about 200 days from the election. The incumbent party is Democrat, the blue line applies.
I have plotted the current data on the relevant graph. If this election follows the average for 7 historic elections, Republicans will win the national vote share by a decent margin.
Is that enough to capture the house? Again, following historical patterns, it should be.
Yet the prediction markets only asses the probability of a Republican takeover at 41-44%.
Either the markets anticipate that (perhaps because the country has changed, or due to 2010-specific reasons) the Republicans under-perform relative to their poll figures and relative to historical patterns.
Or the markets are miss pricing the probability of a shift in power, in which case we should buy at the current prices.