Politicians making a difference
I’ve just come across a post from a week ago by Dani Rodrik which previews a forthcoming book by Larry Bartels. The post includes this fascinating diagram:
What this diagram shows is the percentage growth in income under a Republican and a Democratic administration over the course of a four year Presidential term. It gives the lie to the common assertion that it doesn’t matter to the ordinary citizen who is in power. Apparently the economy doesn’t just ‘keep on truckin’: Democrats actually generate more income growth and also shift wealth to the poor. Republicans tend to shift wealth to the rich and still generate less extra wealth for the rich than Democrats do.
Obviously the economy may not be the major determinant of one’s voting decision. What Bartels’ plot shows is that Democrats’ assertions that they help the common man are more than hot air. The different policy agendas of the major US parties really do make a significant difference to both wealth generation and equality within the nation.
For a more detailed analysis, check out Larry Bartels’ discussion on Rodrik’s blog.
Update: Commentary from Paul Krugman, Tyler Cowen and Alex Tabarrok.
Yawn. Sample size = 11 (6 Republican, 5 Democrat), and I’m generously counting Kennedy and Ford. Plus we’ll just ignore the fact that policy has long lags and it’s ridiculous to attribute all economic impact over the presidential term to the incumbent. Let’s also ignore tax and benefit changes, which are the most plausible way policy impacts on income. And then there’s the little matter of the senate and the house (not to mention state legislatures), all of which have a little impact on policy too…
The idea seems even sillier if we try and nominate the specific policies of Truman, LBJ, Carter (LOL!) and Clinton that we think led to an income growth rate about 1.5%pa higher. Or if we ask why countries with predominantly left-wing governments haven’t been surging ahead of the US over the last fifty years.
Cowen’s explanatory mechanism is plausible but I just don’t think there’s anything here that needs to be explained. Anyway, with a more independent central bank regime one would assume that a president has even less power to influence the economy in the short-term than in the past.
PS You’re confusing income and wealth.
Did you read Bartels’ piece on Rodrik’s blog? He answers most of your criticisms there, I believe. Note also that he appears to be using each year of each presidency as a separate data point.
I’m not sure that US Democratic policy can be extrapolated to ‘any left leaning government’. Furthermore, trying to compare this growth rate with other countries ignores a huge number of country-specific effects that are not relevant to Bartels’ results. I just don’t think that’s evidence against it at all.
PS. I’m not confusing them: I’m using them interchangeably 😉
I did: Bartels’ makes the standard claim that his result is robust to alternative specifications; but it’s amazing how often these guys find ways to ignore obvious alternative specifications that make there results vanish.
Statistical significance doesn’t mean a whole lot when you don’t really much in the way of a causal mechanism. US president is significant for income in much the same way the superbowl winner is significant for stock prices.
The only mechanism Bartels cites is the Cowen one – more expansionary policy – and if that is the case the differences aren’t sustainable (assuming a standard macroeconomic model). The Taborrok version gets the same result with no meaningful policy differences at all.
You don’t think that if there were a few simple policy variables (noting the large dispersion in R and D policies over the sample period) that produced an massive difference in income growth other countries (eg Europe) wouldn’t have stumbled upon those simple policy variables? The counter-factual of sixty years of D presidents would imply a US with a median income around 80% higher than it currently is.
To be honest one of the reasons I’m so sure that this study is bollocks is that no single variable ever explains such a huge gap, it’s just preposterous. It’s like the difference between D and R presidents being equivalent to the difference between the US and Africa. I’d have an open mind to the idea that the two parties have a causal impact on income inequality (properly measured with benefits and taxes), but no one has identified actual fiscal policies with anything like this level of impact on income.
If i was forecasting the average difference in income growth rate between D’s and R’s for the next fifty years, I’d still go with 0%pa.
And since we’re being silly it is only fair to remind your readers that a Democrat was responsible for prolonging the great depression
1) You’re criticising his statistical analysis without having read the details of it. There is no basis for thinking that he made elementary mistakes so I’m not sure why you persist in accusing him of them. I’d like to think that professors at Princeton are unlikely to make such simple errors when writing a book on the subject.
2) The lack of a definite causal mechanism does cast doubt on the results. However, it isn’t a reason to dismiss robust statistical analysis, as Paul Krugman eloquently points out. You say yourself that it’s not unreasonable to believe that the President’s policies have an effect upon incomes. That nobody has yet identified something that would cause effects of this magnitude is certainly reason to question the results; however, I would be loathe to throw them out entirely on this basis.
3) If we don’t know the causal mechanism then it’s because the policy reasons for it aren’t simple. That makes it very difficult to point to the differences between US and European policy that have helped the US to grow faster. It may be that country-specific effects exacerbate the effectiveness of particular policies. Until we know of a plausible mechanism it’s hard to comment on this.
1) Well it’s unpublished, non-peer reviewed research, so I don’t see why I have to take it at face value. I’m certain you’ve seen the research that suggests a large proportion of published peer-reviewed economics papers contain non-replicable statistical results, so yes I find it quite believable that “professors at Princeton” make mistakes writing their polemics. His list of control variables didn’t sound particularly comprehensive, and I found his dismissal of the role of congress pretty vague.
2) Based on existing growth research, a reasonable prior for the difference between R and D policies would be zero. So faced with a statistical result with no underlying theory to support it, I think it is much more likely that the result is spurious than that there is some secret mix of Democrat policies that lead to high growth.
3) I think it stretches credulity to claim that despite the wide dispersion of policies embraced by R and D presidents, the D presidents share some elements of commonality that lead to high growth, and yet this element of commonality is so complex that we can’t identify it and other countries haven’t stumbled upon it.
I’m not sure that we really disagree about how much theory supports Bartels’ result. We really differ in the way we weight it when we update our beliefs. Let me explain the reasons why I think more highly of it than you do.
The Substance:
The result is surprising and doesn’t have strong theoretical support. However, since the methodology is not available it’s impossible to critique the statistical analysis.
The Authority:
The result is one graph from a book written by a professor at a top university. It is highly unlikely that I can think of any problems with his analysis that he hasn’t already considered himself. He considers the result worth publishing and he has a reputation to consider. Therefore I am willing to give it some credence.
The Supporting Cast:
Rodrik and Krugman, both academic stars, consider the result worth mentioning and don’t dismiss it as implausible. They both harbour doubts because of the lack of a mechanism, but don’t think that the flaw is fatal to its credibility. Since their opinions accord with mine, I’m more confident in giving the result some positive weight when I update my beliefs.
If you don’t find the opinions of a number of economic luminaries persuasive then I am baffled as to what would convince you of a result’s worth. Do you have to replicate the result yourself before you’ll believe it? Or does it simply have to accord with your priors?
Well I accept the underlying result that that D presidents have enjoyed an higher average GDP growth rate (which seems to be the driving factor). Since we can’t discuss methodology, I place this result in the “interesting correlation” category. I think you’d agree that the vast majority of “interesting correlations” one reads about are not causal.
As to what would make the causality story more convincing: place it within a macroeconomic model that controls for monetary policy, anticipated and unanticipated fiscal policy changes, technological shocks, provides some measure of presidential power relative to congress, and see how the presidential party result holds up. Identifying actual policy variables and testing for their impact would be nice. Testing those policy variables on other countries for confirmation too.
To be honest, I find your support puzzling (or perhaps I am mistakenly inferring your level of faith in this result). You believe it because a Princeton professor thinks it’s causal (despite the topic being quite clearly outside his field of expertise), and because two economic luminaries like the result (and for both of them you know full well that the result accords exactly with their somewhat extreme priors)? Are there no economic luminaries who’s views you strongly reject? If Mankiw and Phelps both said it was nonsense, would you downgrade your weighting to zero?
I suspect you accept my point that the magnitude of the result is implausible (admittedly, we don’t know what the maximum significant difference in means is). It concerns me that neither Rodrik or Krugman bothered to mention this. In fact, I’m certain that Krugman devoted a whole chapter in one of his pop-economics books (from before he went off the deep end) to laughing at people who try and attribute economic growth spurts to presidential policy.
I do assign a pretty low level of weight to all new economic research though. When you consider statistical fiddling, publication bias, and the fact that plenty of research is later contradicted by other research, I think that’s justified. And in this case I view the causality claim as implicitly contradicting lots of existing research, namely all those studies that struggle to find policy variables that (positively) affect long-term growth rates.