Presidents, Congress, and Economic Growth – The Return of Presimetrics
by Mike Kimel
Presidents, Congress, and Economic Growth – The Return of Presimetrics
If you’re a regular reader, you may know that a couple of years ago, Michael Kanell of the Atlanta Journal-Constitution and I had a book published called Presimetrics. Nobody much read the book, but we had a lot of fun writing it. In it we looked at how Presidents performed on a wide variety of issues – everything from curbing abortions to the murder rate to the national debt. We tried to be as objective as possible, always using data from whichever source collected it, and treating each variable the same way.
Thus, for instance, when looking at the murder rate – we collected data from the FBI. Then we measured the change in the murder rate from right before a President took office to right before a President left office. We found that on a number of issues, particularly economic ones, Presidents who did well followed similar policies to other Presidents who did well. Conversely, Presidents who did poorly on a given issue tended to either show it not much attention or follow the same approach to the issue as other Presidents who did poorly. That is to say, policies matter.
In this post I want to focus on economic growth, but I want to do something a bit different than what we did in the book, and a smidge more complicated than the graphical approach in Presimetrics.
In the book, we looked at the annualized change in real GDP per capita from before a President took office to right before he left office. This time, I want to look at the following variable: Annualized Growth rate over the next four years less Annualized Growth rate for the previous four years.
Thus, for, say, 1981, Reagan’s first year in office, that variable would have this formula:
annualized growth rate from 1981 to 1985 less annualized growth rate from 1977 to 1981
Thus, it would show the difference in the growth rates between Reagan’s first four year term, and the four year term of his predecessor, Jimmy Carter.
I went with four year periods because just about every President in the sample served at least one four year term (or most of at least one four year term), and four years is enough time to see if policies are working or not, on average.
The BEA’s data on real GDP goes back to 1929, so the variable could be created for the years from 1933 to 2007.
I ran a regression attempting to explain this change in the growth rate using the following dummy variables:
1. A Republican becomes President. This variable took values of 1 in 1953 (the year Ike took office), 1969 (the year Nixon took office), 1975 (Ford actually took office in August of 1974, but I rounded up for Presidents who took over in the second half of the year), 1981 (Reagan), 1989 (Bush 1) and 2001 (Bush 2). The variable took values of zero all other years.
2. A Democrat becomes President. This variable took values of 1 in 1933 (FDR’s first year), 1945 (Truman took office in April 1945), 1961 (JFK), 1964 (LBJ took office in November of 1963), 1977 (Carter) and 1993 (Clinton).
3. A Republican President begins his second term. This variable took values of 1 in 1957 (Ike), 1973 (Nixon), 1985 (Reagan), and 2005 (Bush 2).
4. A Democrat begins his second (or plus) term. This variable took values of 1 in 1937 (FDR), 1941 (FDR), 1949 (Truman), 1965 (LBJ), and 1997 (Clinton).
5. A sitting Republican Veep ascends to the Presidency. This one takes values of 1 in 1975 and 1989.
6. A sitting Democrat Veep ascends to the Presidency, which occurred in 1945 and 1964.
7. Republicans take over both houses of Congress from Democrats. That happened in 1947, 1953, and 1995.
8. Democrats take over both houses of Congress from Republicans, which we saw in 1933, 1949, 1955 and 2007.
Note that variables 7 and 8 only count instances where one party took control of both houses of congress from the other party and not from a mixed Congress. Some instances of mixed Congresses looked a bit like one party control so I decided to simply those out.
Obviously, the fit of this simple model won’t be that great as all the variables are dummy variables, and that’s the sort of thing that really doesn’t help your cause when it comes to patterns in the residuals, but hey, the goal here is not to explain the change in growth rates (I’ve done that before, and perhaps will follow up this post) but to look at which of these variables.
If, for example, a given party’s Presidents tend to enact policies that produce growth, we would expect to see the coefficient associated with that party’s presidents be positive and statistically significant.
Results of the regression are shown below.
Figure 1
I could comment, but from past experience, I know that if I do, I’ll be excoriated by people insisting I’m a partisan. So I highlighted the significant and almost significant variables, and I’m calling it a day. Please leave your thoughts about what this means and why in the comments section.
Well, I guess there is one comment I do have to make as I wipe a touch of egg off my face… the coefficient (and quasi-significance) of the Democrat Veep ascendance variable kinda throws into question at least part of what I wrote here.
And now, my usual closing statement: if anyone wants my spreadsheet, just drop me a line. I’m at: my first name (mike), my last name (kimel – one m only) at gmail.
The only two examples of sitting Dem VPs becoming President are, of course, both via the death of the President.
What happens if move the dates for those events to 1965 and 1949 (i.e., pretend their predecessor finished his term, since both VPs were subsequently re-elected)?
Looking at snapshots like this is kinda silly, really.
Carter inherited a much different situation in 1977 than Reagan did in 1981.
Policy was different too, but policy takes years to effect itself.
Gasoline shortages thanks to Iran supply disruptions, all the stuff Nixon was doing 1970-74 to the economy, etc.
People have their narratives about stagflation, who was at fault, inflations doves vs. hawks, but when it gets down to where the rubber meets the road, what does this graph:
http://research.stlouisfed.org/fred2/graph/?g=79t
really mean?
I think it has great explanatory power, but wrt what exactly, I’m not sure.
Ken H,
Truman had to contend with the demobilization recession at the end of WW2. That said, he didn’t handle it well. If you look at the series (Growth next four years less growth prev four years), it goes negative in 1942…. because four years later is 1946, when demobilization occurred. But as I always note, it was his responsibility to deal with it.
As to LBJ, this excercise has been interesting as LBJ doesn’t look as he does when you look at only at annualized growth in real GDP. Part of that… he had very rapid growth, which resulted in large part from the accelerating growth under JFK. But he didn’t leave behind the greatest economy. Growth slowed in his last two years in office.
Put another way… the unelected Veeps just weren’t that great at Presidenting over the economy.
Troy,
Yes, there are special cases. But as I keep pointing out, a model for which every observation is a special case is a lousy model. That’s the point of statistics. And the presumption, given the data, is that an elected Democrat President is going to produce an increase in the growth rate – Carter was the exception.
I think by this rule, even Obama is going to work out. After all… growth from 2005 to 2009 was just about zero, so what recovery Obama has presided over is still better than the downturn that occurred in GW’s last term.
Mike,
I would be really interested to see your figures for Presidents/real per capita growth extended to cover the Obama administration. Do you have those numbers anywhere?
Stan – drop me a line (email address at the bottom of the post) and I’ll send you my spreadsheet. The reason the analysis stops in 2007 is that the dependent variable in 2007 takes this form:
(Annualized growth between 2007 and 2011) less (Annualized growth between 2003 and 2007)
2011 is the last year for which the BEA has the data. (The link to the original BEA data set is in the post.) Essentially, we’re going to have to wait until 2013 data is available to do the full analysis of Obama’s current term in office.
gah, I hate Bush with the heat of a million suns but I don’t blame him for 2001-2003.
yes, 2004-2006 was an utter fraud but at least we did have positive PAYEMS prints:
http://research.stlouisfed.org/fred2/series/PAYEMS/
2008 was just the system readjusting to the reality that we were in 2001-2003, which in turn was the reality of all the mistakes of the 1990s coming home to roost, albeit temporarily.
Troy,
I don’t hate Bush, but I do blame him for 2001 to 2003. Or rather, I blame the combination of Greenspan and Bush for 2001. Greenspan tightened the money supply before the 2000 election when the economy was slowing slightly out of fear that an economy growing too rapidly would tip the scales toward Al Gore. (I’ve had a post looking at Greenspan’s behavior during each presidential election while he was Fed chair – it’s pretty clear he was actively trying to boost the economy when a rep was Pes and slow the economy when a dem was pres).
GW, on the other hand, campaigned on “We’re going to reverse every policy that seems to be working in this economy.” What exactly was the outcome expected to be?