Wednesday, September 03, 2008

Numbers Are Like Fire: Don't Play With Them

I question the potency of the argument put forth by Obama's chief economist (Laura D'Angelo Tyson) on the Colbert Report last night that the economy performs better under a Democratic president than a Republican one. While glancing at the data suggests such a correlation might be there (my analysis program is currently on the fritz so an eyeballing had to do), too many complications render the explanatory power weak.

Variance Within Presidents A party maintains control of the presidency for a minimum span of four years (even if the sitting president resigns of leaves office, his VP maintains the party "dynasty") though several do so for eight, sometimes more. But GDP growth fluctuates a great deal within such a span of time, sometimes going negative. Since our dependent variable doesn't change nearly as much as growth rates do, we lose a lot of our explanatory power.

Policy Variance The Democratic and Republican parties of today aren't constant. This limits our scale of analysis to a much smaller time frame (I went back to the 1930s) and even then there were changes. Also the politics that defined most the difference between the parties were Cold War issues, not economic ones.

Measurement Issues I used real GDP growth to analyze the relationship; Tyson no doubt used similar data. But this data is flawed because it treats government spending as growth, even when that spending is simply a transfer payment (such as Social Security). I'm currently looking for non-government numbers.

Lag Issues Colbert joked about this but there really are lags in policies, sometimes reaching ten years. Since the presidency oscillated between parties for a while, lagging the variables could result in the opposite conclusion.

Causation Issues Perhaps people are more willing to vote for a Democratic president when times are good. Since more government spending has been a staple of Democrats since FDR (Republicans play a similar game, of course, but to a lesser extent), the public might simply be buying a normal good. Wealthier people are more willing to buy feel-good government programs.

The Theory This is likely the biggest problem with the story. What's the explanation for why a Democratic presidency would fair better for the economy than a Republican one? The one I hear the most is that they help the "little guy", but it's not clear how paying someone because they are poor will incentivize them to be wealthier. Moral hazard issues abound. Their preference to cutting taxes for the non-rich helps, but it's usually paired with a tax increase for the wealthy. Unless you believe the richest tend to only buy and invest in each other, this story doesn't get you far. Income mobility remains quite high in the US, repeating the incentive issues that crop up in "helping the poor."

2 comments:

Ryan said...

David,
I had thought the claim behind the PBC idea was more or less monetary policy. Suppose there's a real tradeoff between unemployment and inflation. Then if one party cares (relatively speaking) more about inflation, we'd expect to see busts during those periods said party is in power, all else equal.

I don't like the "public spending is a normal good" story, unless the person advocating the theory swore to it publicly about 100 years ago -- that's a "heads my theory is right, tails all other theories are wrong" case if I've ever seen one. And in-term variance is fine: it just means your R-squared is low. My concern is with the sample size. Even going back to the 1930s, N=13 unless you're willing to count different years w/ the same president as independent trials. So how are we really going to get statistical significance?

Anonymous said...

Ugh. There's that "cum hoc ergo propter hoc" thing again. Good points from both of you.

Did this woman actually present anything to substantiate her claim?