Friday, July 30, 2010

Weak growth means the House is in play

It's the end of July and second quarter 2010 economic figures are now available from the Bureau of Economic Analysis.  And they're not good.  Anyway, these are the numbers I've been waiting for in order to calculate a forecast for the 2010 midterm elections.

Now, before I do that, a note about forecasts.  I do not claim to have any special powers to foresee the future.  I am not doing this to make money.  I think forecast models are helpful when issued before an election for two reasons.  First, they allow us political scientists to test what we know.  It's relatively easy to fit curves after the fact, but it can be very useful to say that this is what we think drives elections, and here's a test based on what's going to happen in the next few months.  Second, they give us a baseline for interpreting the election when it happens.  Republicans will no doubt crow about any electoral gains they make in the fall, but it's helpful to know whether they did better or worse than the models predicted.

Anyway, onto the forecasts.  In my previous posts on this topic (like here), my economic variable has been the growth in per capita real disposable income between the third quarter of the year prior to the election and the second quarter of the election year.  And growth for the current cycle has been pretty anemic -- just 0.88%, while the average prior to midterm elections since 1950 has been 1.69%.  The other figure I include in my calculations is the president's approval rating as of Labor Day, which I'm guessing will be 45%.  That's right about where he's been lately, and his numbers don't bounce around all that much.

Based on these figures, I calculate that Republicans will gain 40 seats in the House.  And guess what?  That's just one more seat than they need to control the chamber.  These figures also predict that the Democrats will lose control of seven state legislatures.

Now, it turns out that I can get a slightly better fit (an R-squared of .64 as opposed to .59) if I instead use the economic growth from the second quarter of the year prior to the election to the second quarter of the election year.  By that measure, Republicans stand to pick up 50 seats, and the Democrats could lose control of 10 state legislatures.

So what does this all mean?  Obviously, these numbers are aren't set in stone, and there are substantial error terms associated with these forecasts.  This is also a pretty simple model: I'm not using data on candidate quality or regional variations, which can be important.  I've used the two variables (economic growth and presidential popularity) that seem the most predictive of midterm election results, but of course these numbers could change drastically between now and November.  The economy could rally, or stall.  (A prediction based on second quarter economic numbers in 2008 would have totally missed the economic collapse that started in September of that year.)  Obama could capture Osama Bin Laden and beat him to death with a shovel on "The View," which I'm guessing would have some short-term impact on his approval ratings.  But barring any major surprises, this is where we are -- a very close contest with a better-than-even chance of a Speaker Boehner being sworn in next January.

Note: My data are available here.


Steve said...

How are you coming to your state legislative figures/losses? Your data has seats won or lost by President's party, but state legislative data seems to be excluded. Are you looking at chambers won or lost over midterm elections, or do you include controls for marginal chambers?

Seth Masket said...

The state legislature numbers come from NCSL. It's the number of total state legislatures gained (lost) by the president's party. I've gone ahead and inserted the numbers in the spreadsheet.