First, we simulate 1,000 separate outcomes of the national vote. The pooled generic polls
conducted 121 to 180 days in advance of the 2010 election show a very close division of 49.1% Democratic and 50.9% Republican. But a near tie in the polls in mid-summer projects to a significant vote plurality for the Republicans in November, close to a 53%-47% split. This prediction is not due to any bias in the polls, but rather stems from the electorate's tendency in past midterm cycles to gravitate further toward the "out" party over the election year -- ultimately gaining about two extra points beyond what summer polls would otherwise show.
The national vote only tells us part of the story, and we still need to determine how it would translate into seats. For each of the 1000 simulated values of the national vote, we simulate the outcome in 435 congressional districts. Open seats and incumbent seats are treated separately. Open-seat outcomes are estimated based on the simulated national vote swing plus the 2008 presidential vote in that district. Outcomes with the incumbent on the ballot are estimated based on the simulated national swing plus the incumbent's vote margin in 2008 and whether the incumbent is running as a freshman. The weight that these variables are given in predicting the final outcome depends on their explanatory power in past elections. Full details are presented in our forthcoming PS paper.
To sum up, first we generated 1,000 simulations of the national vote. Then, we applied each of the 1,000 simulated national outcomes to each congressional district, noting the party of the "winner." For each of the 1,000 simulated outcomes of the national vote, we project the partisan division of the 435 congressional districts.
The figure below displays the range of simulated results. As can be seen from the predominance of red bars, the Republicans win the majority of seats in 79% of the trials. On average, the Republicans win 229 seats, 23 more than the Democrats and 11 more than the 218 needed for a majority. However, the simulations yield considerable variation, with a 95% confidence interval of 176 to 236 Republican seats.
http://www.huffingtonpost.com/joseph-bafumi/a-forecast-of-the-2010-ho_b_697051.html