Saturday, October 23, 2010

Some Statistical Analysis of Parties in Power

Due to my endorsement of Republican Jon Runyan in my home House district (NJ-3), I was recently quoted on the site, a site dedicated to promoting the notion of divided government, a split in control between Congress and the White House. The site's author is supporting Republicans in the mid-terms, since we know the Democrats will control the White House for the next 2 years, but anticipates, presuming the GOP victory projected on this site and elsewhere, that he will be supporting President Obama's re-election campaign in 2012 in order to maintain divided government.

There is certainly some anecdotal evidence to support the notion that divided government works. The time period from 1995-2000 perhaps best illustrates this trend, as that period was typified by strong economic growth and a balanced federal budget while Congress was mostly controlled by the GOP and President Clinton was in office. Many would also site the period from 1983-1989, when Democrats controlled Congress and Ronald Reagan was in office, which saw a similar strong economic uptrend (although without the balanced budget, in that case.)

All of this got me to there a statistical correlation between divided government and economic growth. So, I decided to run the numbers. I chose the last 60 years as my time window and looked at GDP growth figures from the Bureau of Economic Analysis.

Before sharing the results, let me provide a lot of caveats and potential criticisms of this methodology:
(1) GDP Growth is a Trailing Indicator of Policy
It could be argued that economic growth in the present is not as driven by the economic activities of the present as it is by the policies of the past. For instance, it would be hard to blame President Obama for the onset of the "great recession" (you could argue whether his policies have helped or hurt the recovery, but that is another discussion), but certainly his Presidency has born the brunt of the economic contraction associated with this recession. So the potential lag between power and policy and between policy and economic result will no doubt make this analysis imperfect.

(2) All Republicans and Democrats are Not Equal
John F. Kennedy cut top bracket taxes. Richard Nixon instituted price controls. Bill Clinton reduced welfare benefits. George W. Bush vastly expanded Medicare benefits and spending. I could go on and on. The point is that being a Republican or Democrat is not entirely indicative of economic policy or philosophy, so any statistical analysis that treats Republicans and Democrats as a group ignores the vast difference in policies of Republicans and Democrats over time.

You could also certainly argue that the parties have evolved over time, with Republicans having a strong deficit reduction focus prior to Ronald Reagan and more of a low tax / supply side philosophy from Reagan on. Likewise, Democrats have evolved to favor a more progressive tax system from approximately Jimmy Carter on, whereas support for such a system was less consistent before then.

This poses some problem with looking at the effect of one party versus another, but has less of an effect on the analysis of whether divided or unified government works better.

(3) GDP Can Be a Mirage
You could certainly argue that growth in the 2000s, built on the back of leverage and the housing bubble was unsustainable, so giving "credit" for that growth in a statistical analysis might seem incorrect. Similarly, the stock market bubble of the late 90s artificially inflated growth.

So you could argue whether GDP is the ideal metric to measure this sort of thing on, but I struggled to come up with a better metric.

(4) The Time Window is Arbitrary
I chose the last 60 years because it is after World War 2, which was a unique historical event that might severely distort the analysis and because it approximately represents the modern political era and parties. But this choice is arbitrary and a different time window might yield different results.

(5) Correlation Does Not Indicate Causation
Murders and ice cream consumption can statistically correlated in every major American city. Does that mean that ice cream causes murder? Of course not. Murders peak in the summer, when more people are outside coming in contact with each other. Ice Cream consumption peaks in the summer because of the heat. The heat causes both effects, which causes them to be correlated.

What does this have to do with this analysis? It means that just because one party controlling a particular body of government correlates to stronger or weaker growth does not mean that one party controlling a particular body of government CAUSES stronger or weaker growth. So take all of this with a grain of salt.

(6) Past Results Do Not Project Future Performance
Things change. Even if there is an ironclad correlation, it may or may not continue to be true in the future.

Having caveated all that, let me share with you the results of my statistical analysis.
First, let's look at the overall average economic growth based on control of individual institutions:
Average GDP Growth by House Control:
Republican -- 3.3%
Democrat -- 3.1%

Average GDP Growth by Senate Control:
Republican -- 3.2%
Democrat -- 3.1%

Average GDP Growth by Presidential Control:
Republican -- 2.6%
Democrat -- 4.0%

So, you can see...Democratic President's tend to do a LOT better than Republicans in terms of economic growth, and Republican control of each individual body of Congress tends to do modestly better than Democratic control of those bodies.

In terms of unified government (defined here as one party having control of both Houses and the Presidency) versus divided government (defined as all other combinations), we see:
Average Growth - Divided -- 2.8%
Average Growth - Unified -- 3.7%

But since we've already established that Democratic Presidents have higher average growth than Republicans, so we need to tease out that effect. If we break these scenarios apart by party control of the Presidency, we get:
Democratic President -- Unified -- 4.1%
Democratic President -- Divided -- 3.8%
Republican President -- Unified -- 2.7%
Republican President -- Divided -- 2.6%

So, interestingly, if anything, divided government slightly depresses economic growth rates versus unified government, regardless of which party is in office.

So why does the initial analysis show that Republican control of the House and Senate produces slightly higher growth if Democratic Presidents produce higher growth and unified government is better? Simply because we've HAD divided government more often than not with Democratic Presidents than we've had unified GOP governments, which means that Republican Houses and Senates have been associated far more often with Democratic Presidents than with Republican ones. So, basically, the GOP House and Senate gets a halo from the statistical effect known as co-linearity.

* Statistically, Democratic Presidents have a much stronger record on economic growth than Republican Presidents in the past 60 years. The difference is statistically significant, but explains only 12% of the year-to-year variation in economic growth.
* Unified governments slightly outperform divided governments, but the difference is not statistically significant.

Bottom line: On GDP growth, the evidence runs against divided government being a benefit over the past 60 years, but the statistics aren't strong enough to form a definitive conclusion. Macro events appear to have a much larger impact than what party is in control of what body of government. This is probably not a compelling enough case to convince anyone (myself included) to switch his or her vote.

Disagree with my conclusions? Have some different statistics worth looking at? Let me know.

1 comment:

mw said...

Thanks for the link. Since you mentioned my two favorite words, you are back for a return appearance in the current compilation.

An interesting analysis, but I don't put much stock in any statistical analysis attempting to link macro economic performance or stock market indexes to parties in power (for many of the very reasons that you list as caveats).

I am more interested in analysis looking for divided/united gov correlation in things that are directly under federal government policy purview, like federal spending, deficits and taxes. I think that is a more meaningful exercise, even if subject to some of the same caveats.

As you might expect I've accumulated links to a few such studies during my four years of blogging on the topic. Since you asked:

Richard Vedder (older)
Niskanen and Van Doren