January 22, 2004
Bad Barro! Bad Barro! No Bone for You!
Oh God! Another piece of misinformation introduced into the stream of popular economic discourse!
In his most recent Business Week column, Robert Barro says a number of things that... well, to put it politely, are not as true as he appears to think they are. But the most untrue of all are his claims that:
...over the past year... household [employment] grew by two million.... In the past year, the increase in the labor force by 1.7 million represented 1.2% growth--close to the annual growth rate of 1.4% seen since 1980. So there is no validity to the argument that the labor force behaved in an unusual way in 2003.
Let's look at the numbers, most easily found in the CEA-JEC's monthly publication Economic Indicators:
Status of the Labor Force (Household Survey)
|Date ||Adult |
|2002:Dec ||218.74 ||145.15 ||136.44 |
|2003:Jan3 ||219.90 ||145.84 ||137.54 |
|2003:Feb3 ||220.11 ||145.86 ||137.41 |
|2003:Mar ||220.32 ||145.79 ||137.35 |
|2003:Apr ||220.54 ||146.47 ||137.69 |
|2003:May ||220.77 ||146.48 ||137.49 |
|2003:Jun ||221.01 ||147.10 ||137.74 |
|2003:Jul ||221.25 ||146.54 ||137.48 |
|2003:Aug ||221.51 ||146.53 ||137.63 |
|2003:Sep ||221.78 ||146.54 ||137.57 |
|2003:Oct ||222.04 ||146.89 ||138.09 |
|2003:Nov ||222.28 ||147.19 ||138.53 |
|2003:Dec ||222.54 ||146.87 ||138.48 |
|3Not strictly comparable with earlier data... an additional population adjustment in January 2003, and other changes. Data for February 2003 are not directly comparable with earlier data due to a weighting difference. |
Robert Barro compares the December 2003 labor force of 146.87 million with the December 2002 labor force of 145.15 million and calculates a twelve-month growth rate of 1.2%. He subtracts the December 2002 employment number of 136.44 million from the December 2003 employment number of 138.48 million, and calculates a twelve-month change in household survey employment estimates of +2 million.
And he doesn't read footnote 3.
What does footnote 3 mean? Take a look at the first column of the table above--the adult non-institutional population column. Note that the Bureau of Labor Statistics's estimate of this population grows by a steady amount of some 200,000 plus a year each and every month, except for January 2003. In January 2003 the estimated adult non-institutional population jumped by 1.16 million. Why? Because the Bureau of Labor Statistics talked to the Bureau of the Census, and as a result of its conversations adjusted its estimate of the adult population. It did NOT go back and change its estimates of the adult population in 2002--even though it now thinks that those estimated populations are too low by about the 1 million or so by which it raised its estimated population in January 2003.
The BLS estimates the civilian labor force and civilian employment by taking the estimates of the labor force-to-population and the employment-to-population ratios that come out of its Household Survey, and multiplying them by its estimate of the civilian non-institutional adult population. The labor force, as you can see, jumps by a large 700,000 between December 2002 and January 2003--but nobody (except Bob Barro, that is) thinks that a seasonally-adjusted 700,000 people joined the labor force in January 2003: this large jump is an artifact of the population adjustment. Civilian employment, as you can see, jumps by a large 1.1 million between December 2003 and January 2003--but nobody (except Bob Barro, that is) thinks that there were a seasonally-adjusted 1.1 million more people at work in January 2003 than December 2002: this large jump is an artifact of the population adjustment.
That's what footnote 3 is trying to tell Bob Barro: It's saying,
DON'T COMPARE DECEMBER 2002 LABOR FORCE ESTIMATES TO SUBSEQUENT ESTIMATES AND THINK YOU HAVE AN ESTIMATE OF THE RATE OF GROWTH OF THE LABOR FORCE!
DON'T SUBTRACT DECEMBER 2002 EMPLOYMENT ESTIMATES FROM SUBSEQUENT ESTIMATES AND THINK YOU HAVE AN ESTIMATE OF THE CHANGE IN HOUSEHOLD SURVEY-CONCEPT EMPLOYMENT!
What are the right numbers from the Household Survey? Elise Gould of the Economic Policy Institute estimates that consistent population-estimate household survey-concept employment has risen by one million over the past year--only half as fast as Barro claims (but still a lot better than the minus 100,000 or so of payroll survey-concept employment). And the corresponding estimate of labor-force growth over the past year is 0.7%--not "close to" but only half of the average rate of growth in the labor force, and a remarkably low number to see in a year in which the official unemployment rate is nearly constant.
Contrary to what Barro says, there is enormous validity to the belief that the labor force and employment behaved in unusual ways in 2003.
Posted by DeLong at January 22, 2004 01:00 PM
Walking the halls with these folks, I am repeatedly impressed that red will be claimed blue and yellow claimed green as long as it justifies the policies of George Bush and Republican leaders of Congress. But, red is not blue.
As Barro has apparently said, if the data doesn't support your theory the data is probably wrong.
Barro's corollary is apparently - if the data don't support a republican administration, then the data can be massaged.
Prof Barro had a rather bad hacking cough when I saw him last. I kept thinking `hack, hack, get that man some help'.
Prof. Barro is the Robert C. Waggoner Professor of Economics at Harvard????? Maybe I should apply for a job there.
Only if you promise to read the footnotes in BLS tables you use...
Aside from this glaring error -- which I'm really glad Brad explained, since when I read the Barro column I was baffled (although apparently not baffled enough to investigate matters myself) by his numbers -- that column is very interesting because of the way it ends. Although he can't really bring himself to say it explicitly, Barro in not so many words suggests the country would be better off if it elected a Democratic president in 2004. Of course, he believes this for reasons that not many social democrats would be happy about, but it's still testimony to the remarkable fiscal fecklessness of the Bush administration that someone with Robert Barro's politics is not sure that people should vote for the current president.
I suggest a forecasting/predictive-inference/data massage registry for economists -macro and micro. Whenever you publish a calculation, you post raw data, what you did to it before it saw any model, the model, initial and intermediate estimates and the final results. Publicly available, read only access after each thing is posted.
James, i have been waiting (and waiting) (and waiting) for signs of honest conservatism, which would, be definition, mean at least a rejection of bush as we know him. Other than the communists at cato, i haven't found any, but possibly, per your comment, barro is heading that way....
I must be an idiot. I did the calculations three different times: 146.87(Dec)/145.86 (Feb)=1.00692; this is for 10 months.
Ln(1.00692)/10 multiplied by 12 = 0.000828...
EXP(0.000828...) = 0.83% for all 12 months.
Okay, perhaps I screwed up: let's try (1.00692)^1.2 (12 months = 1.2*10 months).
Nuts: 0.83% for 12 months. Not a huge deal, but if the historic average is 1.4% or so (based on household surveys), it's *something*.
Sorry, just read Elise Gould's article (link in original post). She also makes the same mistake:
>>In contrast, the household survey indicates that employment has risen by 2.0 million since the recovery began and by 600,000 since the start of the recession.<<
I cut her a lot of slack because she plots the numbers over a longer period of time (so what else can she do?) and she explains a variety of other methodological shortcomings in the Household Survey. Barro, we'll agree, uses this adjustment to make 50% of his argument.
I can't understood this goof as this point has been made clearly. And Barro is a very bright economist who normally neither makes these kind of errors or shills for Karl Rove. What happened?
Barro goes on to say that the Reagan deficits curbed spending growth and he supports the Bush tax cut on the hope the same may happen now. If one looks at nondefense purchases and transfers relative to GDP, there is no decline in this ratio of any significance. The only restraint to spending was the peace dividend. Does Barro think the 1981 tax cut caused that? Does Barro even realize how deep the Federal spending cuts would have to be to pay for the Bush tax cuts?
Be sure Luskin knows about Barro's boo-boo; he'll certainly want to do a post about Barro's "lie."
Brad, your point is very well taken--one really should check the data before pontificating in Business Week. However your case is even stronger than you thought.
If you focus on the last six months of 2003, the labor force (SA) actually declined. This was only the second time in over forty years that this has happened. In contrast labor-force growth during the first half (even after correcting for the Barro blooper) was fairly strong. I find this pattern quite unbelievable; I'm sure that reality is much smoother. Natura non facit saltum, as Marshall used to say. I presume that these fits and starts in the published data are due to the substantial sampling error that is present in the relatively small (60K households) sample that underlies the Household Survey.
Thus Barro erred twice. He was wrong that the annual growth last year (Dec/Dec) was typical (Brad's point), and he was wrong about the distribution of that growth during the year (my point).
One last thing--your web site is terrific. I don't know how you find the time to do this, but I and lots of others really appreciate it!
Here is an interesting thought on employment and the labor market:
"Joe Carson, head of global economic research at Alliance Capital Management, applies new thinking to an old problem. Maybe companies have been slow to hire because they hired too many people in the late 1990s, Carson says. ``From 1998 to 2000, almost 8 million new payroll jobs were created, even though this period coincided with a protracted slide in corporate profits,'' Carson says. ``Normally, there's a very strong correlation between profits and job creation.''
In the second half of 2003, estimated GDP growth of 6 percent to 7 percent coincided with the creation of 221,000 net new jobs. Carson suspects hiring has been delayed, not disabled, as companies got themselves in better shape.
If Carson's theory is right -- that the over-investment in labor was as large as that in capital equipment -- and the corporate labor purges are winding down, job growth should accelerate to 150,000 to 200,000 a month in the near future."
150-200K jobs/month is too low to reverse unemployment numbers. Where were those dollars for workforce retraining when the economy first stalled in 2001. Why is it only now in this SOTU that worker training becomes a thought and even now it is underfunded??
Re James MacLean Jan 22, 2004 4:19pm: Actually, I don’t make the same mistake as Barro -- my CPS employment numbers are population adjusted, Barro’s are not. We both cite a 2 million employment gain from the household survey, but we’re comparing two different time periods. My employment numbers are since the recovery began, that’s November 2001. Barro’s numbers are from December 2002.
Thanks for pointing that out. My apologies.
That is precisely the whoop ass I was hoping for. Thanks.