« Disappointing News on Income, Poverty, Health Insurance, and the Earnings of College Graduates | Main | Hoisted from Comments »

Tuesday, August 29, 2006

"Unarmed Men in a Battle of Wits"

Brad DeLong sets the record straight as he reviews claims in a new book on the use of economic statistics. His conclusion? Two thumbs down for Econospinning:

"Mix and Match", by Brad DeLong: Gene Epstein of Barrons writes, asking:

I wish Brad DeLong would ... open my book [Econospinning] ... to the 12 pages that make up the second chapter... The chapter recounts a garden-variety case of econospinning by New York Times columnist Paul Krugman. My version of the story is that Krugman not only confused one set of employment data with another to make a point about the job outlook in a May 2004 column. A year later, when Krugman was given the chance to correct the error by then-Public Editor Daniel Okrent, he denied he had made it. DeLong also weighed in, ostensiby to defend Krugman, but only succeeded in compounding the confusion.

OK. There is a certain horrifying fascination in watching the right wing's minions and useful idiots in the press attempt to attack Paul Krugman on matters of economic substance. The Mickey Kauses, the Andrew Sullivans, the Donald Luskins, the Danny Okrents--all seem unarmed men in a battle of wits, or perhaps an air assault by a circular firing squad of flying attack monkeys.

Our story so far:

Danny Okrent wrote, conclusively demonstrating to even his closest friends that he was grossly unfit to be New York Times ombudsman:

13 Things I Meant to Write About but Never Did - New York Times: 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.... [S]ome 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. I didn't give Krugman... the chance to respond.... I decided to impersonate an opinion columnist...

Paul Krugman responded:

Paul Krugman I: In Daniel Okrent's parting shot as public editor of The New York Times, he levied a harsh charge against me: he said that I have "a disturbing habit of shaping, slicing and selectively citing numbers in a fashion that pleases his acolytes but leaves him open to substantive assaults." He offered no examples...

And in reply Danny Okrent did come up with an example--but his example was wrong:

Paul Krugman II: When I asked Daniel Okrent for the specifics behind his final attack, he offered two examples of what he claimed was improper use of numbers. This was the first time I heard from him, or anyone else, about either alleged problem. Let me start with the example that, I think, sheds most light on what is going on: Mr. Okrent's claim that I engaged in "blending, without explanation, numbers from the household survey and the establishment survey -- apples and oranges -- apparently in order to make a more vivid political point about Bush (5/25/04)."

He's referring to two different surveys conducted by the Bureau of Labor Statistics, which provide alternative estimates of employment. Some people play games by mixing and matching numbers from the two surveys, and Mr. Okrent has apparently spent the past year firmly believing (without having checked with me) that I did the same thing, to score political points. But I didn't. All the numbers in my 5/25/04 column came from the establishment survey....

In correspondence with Mr. Okrent, I pointed out that his specific attacks -- especially the blatantly wrong characterization of my 5/25/04 column -- were unfair. I asked him to do what he would have expected me to do, and admit that he had been in error. He refused. Let me repeat that Mr. Okrent never raised these issues as public editor. He now says that he didn't because he "experienced your best-defense-is-a-good-offense approach, and found it futile to deal with it"...

Then Danny Okrent committed virtual self-immolation:

For a man who makes his living offering strong opinions, Paul Krugman seems peculiarly reluctant to grant the same privilege to others.... Because only a fool or a supply-sider would eagerly engage in a debate on economics with Prof. Krugman, I'll try to eschew argument and stick to facts.... The mixing of household and establishment numbers in his 5/25/04 column: Missing from the BLS chart he cites is any number that even resembles the 140,000 new jobs each month needed to keep up with the growing population a statistic he cites in the column, and upon which he seems to have based some of his computations. To my knowledge, that number only appeared in the household survey.

None of Okrent's sources--whoever they were--egging him on and telling him that Krugman "mixed-and-matched" numbers had, you see, told Okrent that the 140,000-a-month number came neither from the household survey nor from the establishment survey, but was the result of multiplying current payroll employment by a projected non-elderly adult population growth rate from yet a third source--the census.

This made Paul Krugman just the weeniest, teeniest bit irritated:

[3] Paul Krugman III: Okrent is lying to cover his mistake when he accused me of blending data from the household and establishment surveys. He now claims that he was only referring to my estimate of how many payroll jobs the economy needs to add per month [to keep labor market conditions from deteriorating], which for some reason he thinks is based on the household survey. But that's not what he said to me: he claimed that the basic numbers I gave on job growth were mix-and-match. In fact, in our correspondence, when I said that it was all payroll data, he declared that "your insistence that you relied only on one set of numbers is very puzzling. I don't see how the math works any other way; maybe you could further enlighten me." In other words, [Okrent] screwed up completely...

That's the back story. Now for the new episode, in which Gene Epstein of Barrons arrives at the scene to do what Danny Okrent won't--to defend Okrent's claim of Krugman's "mixing and matching"--and thus to join Okrent in virtual self-immolation:

Epstein: Just re-read the second-to-last paragraph in [Krugman's May 25, 2004] column.

And employment is chasing a moving target: it must rise by about 140,000 a month just to keep up with a growing population. In April, the economy added 288,000 jobs. If you do the math, you discover that President Bush needs about four years of job growth at last month's rate to reach what his own economists consider full employment."

Isn't "140,000" a number? It comes from the Household Survey--more formally known, as I needn't tell you, as the Current Population Survey--which ties in with the reference to the "growing population." As I explain in my book, that 140,000 is clearly based on the plausible, ball-park expectation that the Civilian ( i.e., employment-eligible) Population will be increasing at an average of about 210,000 a month, and that the labor force participation rate will run about two-thirds. Since two-thirds of 210,000 is 140,000, the plausible assumption is that the labor force will grow by 140,000 a month. And what the first sentence of that paragraph is saying is that "employment... must rise by about 140,000 a month to keep up with a growing population" because if it doesn't, there will be unemployment....

The only way to get Krugman's "four years" is to mix the data together--"apples and oranges"--exactly as Daniel Okrent had originally said. As I point out in the book, that does not even make "good nonsense," but you do get Krugman's result!

Is Epstein correct? No. As I emailed Epstein:

No [the 140,000 a month number] doesn't [come from the Household Survey]:

The CPS is a survey. Each month, it surveys 18,000 clusters of about four housing units each in 754 sample areas. The CPS produces estimates of things like employment-to-population ratios, unemployment rates, unemployment durations, labor status transition properties. It doesn't produce any information about trend labor force growth. It doesn't produce any information about how fast payroll employment has to expand to keep the labor market at about the same level of tightness. It can't--it is a survey of households, not a count of how fast the number of households and thus the number of potential workers is growing.

IIRC, Paul calculated the 140,000 number by taking Census--not CPS--estimates of the rate of adult population growth and multiplying that growth rate by the current level of payroll employment.

I don't see how the CPS could have entered into it.

You see, the "plausible, ball park expectation that... [adult, non-elderly, civilian] population will be increasing at an average of about 210,000 a month comes from census projections, not from the household survey.

And it gets worse.

You see, the Bush Administration in its 2004 Economic Report of the President did the same calculation as Krugman, using the same census and establishment survey sources. You can see it on p. 94:

The Labor Market: Nonfarm payroll employment fell an average of 50,000 workers per month in the first seven months of 2003, before increasing 35,000 in August, 99,000 in September, and an average of 48,000 per month in the fourth quarter.... In the fourth quarter, the unemployment rate averaged 5.9 percent.... Because the labor force is constantly expanding, employment must be growing moderately just to keep the unemployment rate steady. For example, if the labor force is growing at the same rate as the population (about 1 percent per year), employment would have to rise 110,000 a month just to keep the unemployment rate stable, and larger job gains would be necessary (and are expected) to induce a downward trend in the unemployment rate...

The only difference between the Bush Administration and Krugman is that the Bush Administration assumed 1% per year as the rate of adult non-elderly population growth, and hence concluded that nonfarm payroll employment had to grow at 1/12 of a percent per month--110,000, that is--in order to keep labor market conditions stable. But growth is more like 1.25% per year. With payroll employment of 130 million, that's Krugman's number of about 140K per month.


As I said, a certain horrifying fascination.

    Posted by on Tuesday, August 29, 2006 at 09:38 PM in Economics, Press | Permalink  TrackBack (0)  Comments (5)


    TrackBack URL for this entry:

    Listed below are links to weblogs that reference "Unarmed Men in a Battle of Wits":


    Feed You can follow this conversation by subscribing to the comment feed for this post.