Site Meter

Tuesday, May 31, 2005

Andrew Samwick comes close to proving a negative (too bad he was trying to prove a positive).

There has been some chatter in the blogosphere about Daniel Okrent's parting comments about Paul Krugman during their time at the New York Times. The offending sentences are:

Op-Ed columnist Paul Krugman has the disturbing habit of shaping, slicing and selectively citing numbers in a fashion that pleases his acolytes but leaves him open to substantive assaults. ... No one deserves the personal vituperation that regularly comes Dowd's way, and some of Krugman's enemies are every bit as ideological (and consequently unfair) as he is. ... But that doesn't mean that their boss, publisher Arthur O. Sulzberger Jr., shouldn't hold his columnists to higher standards.

[snip]

. The discussion that followed Okrent's piece might lead one to believe that there are no examples. I'll remind my readers of one that occupied our time back in October. It started with the post, "Paul Krugman, Meet Irony." The key quote (with the offending statement highlighted) from Krugman's op-ed, "Checking the Facts in Advance" is:

Mr. Bush will boast about the decline in the unemployment rate from its June 2003 peak. But the employed fraction of the population didn't rise at all; unemployment declined only because some of those without jobs stopped actively looking for work, and therefore dropped out of the unemployment statistics. The labor force participation rate - the fraction of the population either working or actively looking for work - has fallen sharply under Mr. Bush; if it had stayed at its January 2001 level, the official unemployment rate would be 7.4 percent.


As I noted in my original post and the considerable discussion that followed (here, here, here, here, and here), there are two channels that allow the unemployment rate and the labor force participation rate to fall while leaving the employment-population ratio unchanged. The first is that people who want to work give up looking for work. (This takes the same person out of both unemployment and the labor force, with no one entering or leaving employment.) The second is that people who have jobs decide they don't want them anymore (perhaps to take care of their kids or go back to school), and they get replaced by someone who was previously looking for work. (This takes one person out of employment and the labor force and another person out of unemployment and into employment. Same net effect.) The two channels have opposite implications for whether we think the statistics are bad news for the economy.


Samwick is a smart guy. This should be about the best shot at Krugman. Samwick's "example" does not support Okrent's claim that "Op-Ed columnist Paul Krugman has the disturbing habit of shaping, slicing and selectively citing numbers in a fashion that pleases his acolytes but leaves him open to substantive assaults." The number cited in the op-ed criticized by Samwick is payroll employment, unshaped, unsliced, and selected because it is very very important.

Samwick disagrees with Krugman's interpretation of the number. Actually Samwick only hints that he might disagree, since he only claims that another interpretation is logically possible, but does not argue that it is reasonable (it is not as explained by Brad DeLong explains).

I can imagine no stronger evidence that Okrent is wrong and that there is, indeed, no example (thus Samwick surpasses Patrick Sullivan in a comment to a post below). A smart economist sets out to find an example for Okrent, explicitly responding to the view that there seems to be no such example. He comes up with something else entirely. A negative can't be proven, but rarely in human history has anyone come so close.

Update: Thanks for the link Prof Samwick. However, this post doesn't deserve it at all. As Samwick gently notes back at his blog I was totally wrong. Krugman's article was based entirely on the household survey and not at all on the payroll survey. He discussed the employment to working age population ratio (best measured with the household survey) not employment (best measured with the payroll surve). The reason is that population working age or otherwise is measured once every 10 years. In between it is necessary to guess about illegal immigration. The anomaly in employment numbers (household survey looks better than payroll survey) might be explained by reduced illegal immigration after 9/11 and general tightening up of border security.

However, I think that Samwick is unfair to Krugman. In the quoted passage, Krugman does not use the word "discouraged" or any synonym or paraphrase. Thus that op-ed was neutral on the very interesting current Samwick DeLong debate (which seems to be going rather well for Samwick as Brad has turned to sociology. Brad really honestly loves sociology and believes in it, but talking about sociology in a debates between economists tends to be a sign of being in a tight spot). Like Brad, I find Samwick's recent post pretty convincing, totally aside from the fact that it contains a link here and he is nice about how dumb the post above is.

Still I think that Samwick combines two questions. One is who is leaving the labor force (unemployed people or workers who retire) the other is why people are leaving (discouragement or some other reason). Krugman is a bit sloppy in suggesting that the people who leave were all unemployed. He does not claim that they left, because they were discouraged. Krugman just notes the fact that the decline in unemployment is, in an accounting sense, fully explained by the decline in labor force participation. In other op-eds when referring to other time intervals Krugman did discuss discouragement. However, it is a more reasonable hypothesis for those periods (roughly 3/2001 through 6/2003). Also I really should find and re-read things before commenting on them.

Thus Samwick's valid point is just that Krugman doesn't mention the fact that gross flows are much larger than net flows. This is a very common omission in discussions of economics in newspapers (maybe I should say "almost universal" not "very common" but I am trying to be relatively careful). It seems to me that Krugman was very careful not to speculate about people's motives, that he just mentioned a fact which tends to undermine a claim based on the unemployment rate. I think that Samwick doesn's disagree with Krugman's interpretation, because I see no evidence that Krugman presented an interpretation.

I think more strongly than ever that Samwick's critique has nothing to do with Okrent's. However, I recognise that Samwick is not proving a negative, because it is clear he is interested in the topic because it is important (rather more important than Paul Krugman) and not because he was looking for misuse of statistics by Krugman. I think the record on that debate is coming close to proving the negative that none such can be found. Clearly many people have been looking hard, and no one has come up with any such thing.

No comments: