« Twitter roundup | Main | Midterm Postmortem at Boston Review »

November 10, 2010


Interesting analysis. Here are a couple of questions:

1. I understand research shows that economic conditions play a very big role in election results. Is there research showing which specific economic measure has the best predictive value? GDP growth? Unemployment rate? Job growth? "Misery Index" (unemployment rate + inflation rate)?

Also, what period is most significant? The latest quarter? The latest year? Some weighted average of recent quarters?

The point I'm driving at is that Brendan may have chosen GDP Growth in the 2nd Quarter of the Election Year because GDP Growth in Q2 1948 was so high.

Suppose Brendan had looked at GDP Growth in the 3rd Quarter of an Election Year. GDP Growth was much lower in 3Q 1948 than in 2Q. I would guess that Truman substantially overperformed relative to a plot that was based on GDP Growth in 3rd Quarter of an Election Year. Why prefer the 2nd Quarter to the 3rd Quarter? One might think that results closer to an election ought to be more significant than results farther away from it.

In statistical principle, a researcher needs to choose the method first, independent of the data. If the method is chosen to suit the data in order to produce a desired result, then the analysis is flawed.

2. Don't know whether there's enough data to study this one, but...

Do economic results have the same effect on elections even if nobody knows what they are? In this case, we presumably now know that economic growth in 2Q 1948 was considerably higher than was believed at the time. So, which has a bigger impact: the real figure or what was reported to voters at the time?

Income growth that comes in time to be felt before the election but not too far in advance - see e.g. Douglas Hibbs's model here: http://douglas-hibbs.com/Election2008/2008Election-MainPage.htm  Abramowitz is presumably using Q2 for the same reason.

I would also argue that Q2 data is more likely to be publicized and better understood i.e. "felt", by the public than Q3 data, particularly in the past. That assumption might be more questionable for recent years as the speed of information has seemingly increased.

I also was curious about GDP growth used as the measure and not some other measure. And does it account for "revisions" to reported growth (these measures seem to be frequently revised in the next quarter) or use the final figures?

In any case, it can dangerous to use just one measure like this to say anything with particular confidence.

I wonder how good the economy was when Lincoln was campaigning for reelection.

Knowing Brendan, he's got a chart for that.

The comments to this entry are closed.