Ryan Avent has responded to my post on the European Welfare State bill at Ezra Klein’s blog. Ryan is a smart, practical guy who raises some interesting points worth discussing, so I’ll take them one at a time.
After taking on the Wall Street Journal editorial page, which he characterizes as a “hub of conservative snake-oil salespersons”, he moves on to my post by saying:
He [Manzi] criticizes the timing, despite the fact that 80% of the spending stimulus is delivered within 30 months, and 98% of the tax cut stimulus within 18 months.
True; but I don’t think very meaningful. As I said in the post, the advantage of tax reductions is that they can act quickly. The timing issue I raised was with spending, and 30 months is hardly a standard that I think makes sense for classifying a unit of spending as stimulus.
The relevant point that I made in the post was that if this is a “normal” recession, then most of the stimulus would be too late, and if we are headed for a ten-year Great Depression-style or ten year 1990s Japan “lost decade”-style meltdown, then this isn’t really stimulus, but a long-term reallocation of control of resources to the government.
According to the NBER, there have been a total of 17 recessions outside of the Great Depression in the past 100 years. The average length has been just about one year (12.4 months). The longest recession in this period was the recession of 1910-1912, which lasted 24 months. This is also just about two standard deviations from the mean for length. No recession since the great Depression has exceeded 16 months. I think it’s fair to use 24 months as the maximum expected length of a “normal” recession.
According to the NBER, the current recession started in December, 2007. If we start the clock for the bill at February 1, 2009 (which seems pretty aggressive), that means we are already 14 months into the current recession. So we have ten more months until this recession is at the maximum length of a “normal” recession. About 20% of the stimulus spending is projected to occur by that time, so about 80% would occur after the onset of recovery if we had a 24-month recession. If 30 months were to make sense as a standard, this would imply a 44-month recession, which is unheard of outside of the Great Depression.
Ryan goes on to say:
And what Manzi doesn’t seem to get is that social program funding makes good stimulus. … Food stamp assistance will go directly toward reducing inventory at places like grocery stores, which will directly contribute to continued production at places that supply grocery stores, which means saved jobs and a substantial multiplier.
As I said in my post, that is the theory. I linked to a number of economists on various sides of this question. It’s pretty easy to find very respectable economists who will argue precisely the position that Ryan espouses here, and provide lots of data and analysis to support their claims. It’s about equally easy to find equally respectable economists who disagree, and who can also provide lots of data and analysis to back up the contrary position. As I also said in the post, what I’ve yet to see (and am confident doesn’t exist) is an empirical demonstration of a causal model that can reliably predict the impact of any of these proposals.
Finally Ryan goes on to say:
And he wraps up by pointing at uncited studies which purportedly show no benefit to increased education spending. Well, I’ll see that assertion and raise him Claudia Goldin and Lawrence Katz, who credit postwar innovations in education policy, including increased public funding, for generating broad-based productivity and wage growth — trends which have recently reversed themselves; we’re now producing cohorts less skilled than their parents. And I’ll raise him economics Nobelist James Heckman, who has repeatedly shown the value of increased funding for early childhood education.
It’s a fair enough comment that I did something that I hate when other people do it, which is to wave my hands at “studies” without linking to anything. I’ll remedy that here, but first it’s important to note that I actually made a subtly, but importantly, weaker claim than Ryan asserts I did. He says that I pointed to studies “which purportedly show no benefit to increased education spending”. What I actually said was that “Study after study has shown that, at a minimum, there is no clearly demonstrable educational benefit from more aggregate spending on schools.” That is, not that there is proof that more spending will fail to improve achievement, but rather that there is a lack of proof that it will.
The classic study is Hanushek, which is a meta-analysis of more than 400 published, peer-reviewed analyses of spending increases vs. outcome changes that show what is essentially a distribution of measured effects scattered around zero. To be fair, Hanushek had an epic methodology debate with Greenwald et. al., who claimed to find a small positive effect. Brookings did a book in which a number of scholars weighed in, including Hanushek and Greenwald, and I think that it would be hard to conclude upon reading it that there is good evidence of a relationship between aggregate spending and outcomes. I think it’s a fair statement of analytical opinion that there is no widely-accepted evidence that there is a material causal relationship between aggregate spending and student achievement. For what it’s worth, this is consistent with accepted opinion, to my knowledge, in other OECD countries.
Goldin and Katz’s book is excellent. I cite it approvingly and use data from it in an upcoming National Review article. Their reconstruction of the wage structure of early 20th-century manufacturing industries is worth the price of the book by itself. Further, I fully believe in their basic thesis (at least at a high level of abstraction), which I would state as roughly: workers in our economy are in a race between development of as-yet-non-commoditized cognitive capabilities on one hand, and wage reductions as capabilities are commoditized through technological advances on the other. But, in my view, they fail to make the case that additional years of schooling consistently create the relevant capabilities. In other words, I would say that we are in a race between capabilities and technology, rather than education and technology.
Goldin and Katz make the common-sense case that the interaction of supply, demand and institutions (what they call the SDI framework, see page 293) accounts for changes in wage structure. This rests on the central insights they develop in Chapter 3, where they say on page 101:
The most important conclusion to be drawn from this analysis is that the driving force behind the explosion of the college wage premium of the last 25 years was a sharp decrease in the growth rate of the relative supply of educated workers and not an increase in the growth rate of the relative demand for skill.
But what they fail to address (and I don’t think such an analysis is feasible) is to disentangle correlation from causality. To what extent does attendance at various academic institutions create skills versus act as a marker for capabilities that drive higher wages? Has there been a structural shift in this balance over the past several decades? Does post-1980 technology, on balance, interact with capabilities and educational attainment in structurally different ways than earlier technology shifts? And so on.
I yield to no man in my admiration for James Heckman’s work in program evaluation. Ironically, he won the Nobel Prize for the so-called Heckman Correction, which is a way of addressing, in certain circumstances, the kind of fiendish selection bias effects that can dominate program effects for many social programs–which is extremely closely related to the basic analytical problem that I am raising about Ryan’s implied interpretation of Goldin and Katz’s work.
I don’t deny, and in fact gratefully acknowledge, that schools can improve capabilities; and further, that schools are not free. Therefore, at some level, funding is required to get these benefits. There are surely specific early childhood interventions that create benefits. But this is a very long way from saying that $100 billion of spending as currently directed by the Congress on our heavily and poorly regulated education system should be expected to create material gains in student achievement.