Forecasting the NH State House Election - Postscript

by: William Tucker

Sat Nov 06, 2010 at 17:47:25 PM EDT

My statistical model projecting Republicans would win 273 seats in the New Hampshire State House captured the impact of the GOP wave, though slightly underestimating the magnitude. Still, it was more accurate than most of the pundits who get paid for this sort of thing, including Fahey (210-220), DiStaso (226) and Landrigan (230). Only Pindell was closer with his prediction of 290 Republican seats and 27 tossups.

The model was based on the projected national congressional vote without regard to local candidates. In my mind, the results validate the approach and provide valuable insight to understanding electoral fundamentals. But this could lead one to then ask, do campaigns matter? And I say, you need only consider the 72,000 vote differential between John Lynch and Paul Hodes to conclude campaigns and candidates and issues matter -- a lot.

Geeky detail below the fold.

• Nate Silver estimates Republicans will receive about 51.8% of the total votes for the U.S. House compared to 45.1% for Democrats. That's a two-party difference of 6.9 points and within shouting distance of the model's projected 6.2 point margin. Averaging the generic ballots polls was a very good predictor of the final margin, certainly better than the Gallup generic ballot poll and its 15-point margin.

• Nine generic ballot polls were released after I published the forecast on Friday. Had I had the time to update the forecast based on those polls (which trended Republican) the projected two-party margin would have increased to 7.2 points, even closer to the actual result, and would have increased the projection to 281 Republican seats.

• The projection for the number of partisan seats was within one seat of the actual result in 100 of 103 House districts. The model predicted the correct partisan makeup in 62 House districts, overestimated Republican seats in 10 districts and underestimated Republican seats in 31.

• Nate Silver has a very interesting analysis showing New Hampshire had a higher enthusiasm gap between Republicans and Democrats than any other state in the country -- perhaps explaining why the model slightly underestimated Republican wins.

• This graph indicates midterm seat change for the President's party in the U.S. House tracks with State House losses, supporting the general approach used in this model.

Cross-posted to Miscellany Blue.