WingRight

Alan Blinder has an article in the New York Times that discusses a book entitled Unequal Democracy: the Political Economy of the New Gilded Age by Larry M. Bartels, a political science professor from Princeton. We have a photograph of Professor Bartels above.

The first principle theme of Bartels’ book, as elucidated by Blinder is that “the United States economy has grown faster, on average, under Democratic presidents than under Republicans.” This is from 1947-2005.

The second theme is that income inequality, the liberals’ latest bugaboo, grows under Republican presidents and shrinks under Democratic presidents.

Blinder ties these two supposed findings together to demonstrate that a vote for Obama is a vote for greater growth and less inequality and who wouldn’t want that. Statistics prove this truth. He does deign to admit,

Such a large historical gap in economic performance between the two parties is rather surprising, because presidents have limited leverage over the nation’s economy. Most economists will tell you that Federal Reserve policy and oil prices, to name just two influences, are far more powerful than fiscal policy. Furthermore, as those mutual fund prospectuses constantly warn us, past results are no guarantee of future performance.

The first question that comes to mind is a classic - does correlation equal causation? There is the little issue of Congress that does things like pass income tax laws and spending bills, but Bartels couldn’t find any correlation between which party controlled Congress and GDP or inequality, so he concludes that it must be the President’s party that causes these two economic metrics to change.

The second concern is that we have a political scientist doing statistical analysis of economic data. I haven’t been able to determine whether this actually happened, but I’ll bet a buck that Bartels outsourced the statistical analysis because he couldn’t do it himself.

Jim Manzi does know something about statistical analysis and he dismantles the Bartels theory in an excellent blog post.

It seems that Professor Bartels used a one-year lag time to credit/debit a president for GDP growth and income inequality. In simple terms, Jimmy Carter lost the 1980 election but gets credited for what happened in 1981. Now you might say that some of Carter’s decisions would have effects that carried over for a year and others had effects that carried over for two years or not at all.

Professor Bartels will have none of that, however. It’s one year, no more, no less. I’m sure he would say that he had to pick a cutoff date and one year seemed like the best one to use.

Manzi demonstrates a possible reason why Bartels likes one year. If you set the lag at two years, there is no statistical correlation between the president’s party and GDP or inequality. If you set the lag at zero years, there is no statistical correlation between the president’s party and GDP/inequality.

This makes no sense at all. Bartels’ entire thesis depends an assuming that every one of Carter’s economic policies had an effect until exactly one year after he left office and none had effects that disappeared well before that time or continued well after one year.

Far be it from me to question Professor Bartels’ motivations, but if I were looking for a statistical correlation with a lot of data where there wasn’t any real correlation, I could probably build a spreadsheet that optimized for that correlation if I could vary the lag time. Who knows, if Bartels had cut off the effects of a president’s economic policies 347 days after he left office, there might even be a higher correlation between a president’s party and economic variables.

And none of this could possibly be tied to an election year. It’s all science.

Tags: Economy, Obama, Politics //