Comparing Presidents – Responding to the Economist
Its a beautiful day, but despite that, I have to work. I should be doing productive stuff right now, but by way of this comment at Betsy’s Page, a right wing blog, I became aware of this piece at the Economists’ Free Exchange.
It seems that my look at Dems versus Rep Presidents has come under attack. I was going to write up a well reasoned reply, but, to quote John Candy in Canadian Bacon: “There’s a time to think, and a time to act. And this, gentlemen, is no time to think.”
So here goes…
This sort of analysis is, of course, quite beloved of Democrats with a smattering of statistics; not so beloved of the Republicans who spit soda all over their new flatscreens when they see it.
Funny. I started looking at data in response to stuff I saw on the Cato Institute, put out by the Club for Growth, or in articles by Thomas Sowell. These always seemed to me to be very carefully constructed so as to show the desired outcomes.
The problem with most of these sorts of analyses are plentiful. They have too few sample points…
Which is what degrees of freedom are for… Maybe there’s something wrong with the textbooks on my shelf, but the t distribution tables in the back of those textbooks have as few as 1 (one) degree of freedom. When the degrees of freedom are low, the t-statistics has to be really high in order in to reject H0. Or something like that – what do I know? I only have a smattering of statistics.
and worse, they generally treat the points as if they were independent events, rather than (as they are) dependent on the points immediately preceding
Let’s see. I did this several ways, with several data sets. Maybe you can argue that with quarterly data (in which case the few observation issue goes by the wayside, even for someone that is unacquainted with the t statistic). But I also did the study by administration (hence the argument about few points, I guess). If one wants to argue some sort of path dependence between Ike’s performance, JFK/LBJ’s, and Nixon, let’s hear it.
I have no doubt that the first year of JFK/LBJ’s term was heavily dependent on Ike’s administration – after all, the budget was Ike’s. But a few years later? I’d be more inclined to accept an argument from someone on the right that the “Great Society” had no effect on the economy if I hadn’t heard so many times that this massive expansion of the role of government in the economy was bad for the economy. One can have one’s cake or one can eat it.
you are more likely to get very rapid growth if you have had very poor growth somewhat recently
Ah yes. But I also put up figures leaving out periods of recessions, and also leaving out recessions plus two years. I note… Republican administrations seem to have more recessions and to spend more time in recession (shouldn’t this phase Republicans?), so this in effect means cherry-picking Republican performance. Despite that, Republicans still didn’t do as well as Dems.
And does this also apply to other series that I looked at, like the growth rate in real health care costs or youth crime? Why?
Also… is this a comment on lags? I note that I had an entire post on lags and how they relate to this issue. For the lag argument to work, it has to vary in length. I’ve heard folks on the right attribute the growth rate in the Clinton administration to Reagan, but never the growth in the GHW administration that immediately follows Reagan’s. And I have yet to find a right winger willing to attribute Reagan’s growth to Nixon/Ford, or worse, Carter.
The next point in the piece was a graph… showing percent change real gdp per capita quarterly. The author decided it helped her back up her point of fast growth following slow growth… I have a bit of a headache right, so maybe I’m missing something here and I’m not seeing what she’s getting at. Whatever it is, it doesn’t apply to the administration as a whole analysis. And it’s also sure a lot of points…
Since Ray Fair has argued that these variables are also likely to determine election results, this makes it a little tricky to attribute any sort of causation. The results are also pretty sensitive to choice of start year; for some reason, none of them ever starts with FDR.
Why not start with FDR? Regular readers know I consistently try to start with Ike’s election (assuming the data goes back to that point). Too far back, and the world changes too much. The Republican and Democratic personnae that we know today are not hugely different from the ones that existed immediately after WW2. I could likewise ask why, if the Republicans are the Party of Lincoln, they aren’t the most liberal party out there when it comes to, say, rights for Black people?
But OK. Fine. Let’s look at FDR because it seems that the author of this piece feels his inclusion will change things. I wandered over to Bureau of Economic Analysis’ National Income and Product Accounts Table 7.1 and pulled data on real GDP per capita. In 1932, the last year before he took office, real GDP per capita was $5,056. In 1945, when he died, it was $12,766. This translates to a growth rate of 6.84% per year. (Note… this is how I’ve been doing the analysis so far… last year before taking office to last year in office.) If we sum the annual growth rates of every Republican President starting with Nixon and running through 2006, it comes to 6.46%. I’m not sure this is a comparison the author had in mind.
But I can hear the response now… it was the war. It was the war that saved FDR. The US entered the war on December 7, 1941. So let’s go with 1941. Real GDP per capita in 1941 was $9,079. A growth rate of… 6.03%. Maybe 1940 is better? 4.98% annual.
Anyway, I could go on, but this is really silly. Going back to FDR is not going to hurt my hypothesis.
This analysis gets around some of this first problem by using quarters, rather than years or (stupider still, and most common–presidential terms), and testing various lags, but the dependency problems remain.
Why? Why is it stupider still? If Reagan’s performance was everything we hear, why shouldn’t I look at his annualized performance? Isn’t that a relevant metric? Or is opinion the only thing that matters?
And why did Reagan’s third best in sample performance follow Carter’s fourth best? I thought good performance followed bad ones. It certainly works when you look at each of the non-Reagan Republican administrations.
no one ever really posits a good mechanism that would make it so.
And then she mentioned a few mechanisms I suggested, including a few she said I dismiss out of hand are…
Democratic policies aren’t any better for the economy than Republicans, actually, but people think they are, so it becomes a self-fulfilling prophecy
If this were true, wouldn’t it be a good reason to vote Democratic? What is leadership except making people believe enough that they act upon it? What Reagan sold was “morning in America” and by selling it, he made it true. On the other hand, GHW’s delivery of “a thousand points of light” or whatever never quite rang true.
It’s a spurious correlation
Well sure. But I looked at a bunch of series. These results don’t stand in isolation. And I ran the means test on a series that doesn’t include debt. Looking at Real GDP per capita less change in real debt per capita (after all – you can boost real GDP per capita simply by spending money, but eventually it has to be paid back so this is a more fair measure of performance), for example, the P-value of the the same test is 0.00764. Let me repeat… eight tenths of a percent. The probability of a Type 1 error in the tests that I mentioned in the blog post is either 1% or 5% when you’re using quarterly or administration long data. Sure, it may be spurious, but at this point I think those who feel it is have the onus on them.
There’s some third factor that produces both economic growth, and Democratic administrations
If there is such a thing… why does GW’s performance fall in with other Reps rather than Dems? He did lose the popular vote in 2000. If something delivers both economic growth and Democratic administrations, and the something delivered the votes for a Democratic administration, where is the growth? And what happened in 1984? A landslide for Reagan, but Democrat-quality growth. This hypothesis doesn’t fit known facts.
people might vote Democratic when they’re not worried about the economy, and Republican when they want someone to fix it, which would mean that Republicans would end up with a disproportionate number of awful quarters. I have no particular evidence that this is true, mind you, and there may be some disconfirming data on the wind. But I can’t rule it out.
1992… “It’s the economy, stupid.” 1976? In two of the elections in the sample that produced Democratic administrations, the economy was an issue, and not because things were going well. Reagan also was elected partly on the basis of the economy. Was GW? She may not be able to rule it out. I think the rest of us can.
The problem is this: which policies?
Which policies, as practiced by Bill Clinton, Jimmy Carter, Lyndon Johnson, John F. Kennedy, and Harry Truman, but not by the Georges Bush, Ronald Reagan, Richard Nixon, and Dwight Eisenhower, are good for the American economy? Economic policy has changed so much over the last fifty years that Nixon and Eisenhower have, economically, far more in common with their Democratic predecessors and successors than they do with Reagan or the Bushes.
OK. If we focus on the post-WW2 era, it would seem the policies to emulate are those of JFK/LBJ, wouldn’t it? Almost 3.5% annual growth in real GDP per capita! Clinton, the second best performer, would also do in a pinch. Or, if we go with the eras beginning with FDR, it would seem the policies to emulate are those of FDR. 6.84% growth in real GDP per capita? That’s more than 2.5 times Reagan’s growth rate. Give me some of that, please! And seconds too. But I’m not sure this is where the author of the piece really wants to take things – my guess is she’d prefer something akin to Reagan’s term.
And then the author bring sup the Superbowl effect… Well, a few problems. First, I’m pretty sure the probability produced by the Superbowl effect is not as high as in the tests I ran. Second, there is no mechanism for the Superbowl effect. But is someone that works for the Economist really trying to argue that the President of the United States has no effect on the growth rate of the US Economy?