- The Washington Times - Wednesday, February 6, 2008

ANALYSIS/OPINION:

Why don’t schools of journalism require their students to take statistics 101? They should.

Saturday’s Page One stories around the country blared the news that U.S. jobs in January fell for the first time in more than four years. But is that true? Let’s take it piece by piece.

One thing students learn about in beginning statistics is sampling error. Data based on samples are subject to a measurable range of error around the sample estimate. When a plus or minus measured change in data is within that band of error, the change is not statistically significant and hence not different from zero — i.e., there is no change.

At the beginning of this month the Bureau of Labor Statistics (BLS) reported a minus 17,000 change in nonfarm payroll jobs between December and January. (The estimate was labeled preliminary and subject to revision.) Because the estimate was within the sampling error range that BLS uses for month-to-month changes in payroll jobs, the bureau was careful not to say jobs declined in January.

But the negative number was far too juicy for the mass media to resist, and, conjuring up lurid synonyms for decline, they sent a scare across the land.

That wasn’t the media’s only goof. The widespread claim that this was the first decline in payroll job numbers in more than four years, while it made good drama, was also false. Just five months earlier the BLS reported a negative 4,000 change in payroll jobs for August, a preliminary number subject to revision like the latest estimate. Here, too, the BLS didn’t label the August change a decline. It was simply another no-change month.

But according to the jazzed-up and truth-be-damned criteria used by the media, in which technical esoterica like sampling error have no place, it was a calamitous decline. Now, it doesn’t take a statistician to tell you that five months is shorter than four years. True, the preliminary August estimate was subsequently revised to a positive number, but that doesn’t get the media off the hook since the January estimate they labeled a decline was also preliminary.

The more relevant employment story went unreported. While the labor market has unquestionably weakened in recent months, the picture is not as bad as the business-reported payroll numbers would have us believe.

Total employment, as measured by the government’s household survey, has been painting a decidedly less dismal picture. That’s true for last month, the last three months, and the last six months. Since month-to-month changes in total employment are more volatile than changes in payrolls because of the former’s larger sampling error, comparing six month changes in the two employment series increases reliability and makes more sense.

Since last July, payrolls jobs have risen by 420,000, or an average of 70,000 a month, whereas overall employment (adjusted for the annual revision of population controls used in the household survey) has risen much more, by 769,000, or 128,000 a month. That’s the difference between a modestly and a moderately expanding labor market. It stands in contrast to the prior six months, from January to July 2007, when payroll jobs rose significantly faster than household-measured employment. Then, payrolls went up by 574,000, or 95,000 a month, while total employment rose hardly at all. But these days it’s what happened lately that counts most.

Studies have shown that averaging the two employment series probably brings us closer to the truth. On that basis, for the last six months employment has risen by nearly 600,000, or about 100,000 a month — not particularly good, but not bad either. Certainly it’s better than the off-the-wall Humpty-Dumpty stories told by a lot of journalists and politicians.

There is no convincing evidence employment is headed into negative territory. Indeed, many believe the prevailing pessimism is exaggerated and not supported by an objective assessment of a broad range of data.

As occurred for August, in another month we may find that the January jobs estimate revised upward and the reported negative change forgotten. Since 1980, there have been 16 occasions in which an initially reported negative job change was revised to a positive number a month or two later.

In the months and quarters beyond, the outlook is murkier than usual because of the ongoing correction in housing, lingering uncertainties in the financial sector, and the stage of the business cycle. I’m betting on no recession, weak but positive economic and job growth in the first half of this year, and a rebound in output and employment in the second half. But to forecast with confidence these days is a self-deception.

Alfred Tella is former Georgetown University research professor of economics.