Headline: Regional Employment Is Weak, But You Wouldn't Know It

Source: Neil Bhutta, The Dismal Scientist, 7 August 2001

URL: http://dismal.com/thoughts/article.asp?aid=1323

[Key graphs accompany the article, worth viewing at the above URL.]

Regional economists may be perplexed when trying to establish where exactly the national slowdown's roots lie. While employment is falling nationally, state figures, when aggregated, portray an economy that is substantially more buoyant. But this picture is almost certainly wrong. Given the bleak manufacturing picture, difficulties in several service industries, and common estimation errors in measuring employment at the state level, there is a good chance that state level data for this year will be revised significantly downward.

Employment statistics are computed from a survey administered by the Bureau of Labor Statistics. State employment bureaus use their state's sample to estimate monthly state employment figures, while the national BLS office uses the entire sample to formulate nationwide numbers. Because states prepare their own estimates individually and without forced reconciliation by the BLS, the numbers very rarely match, though in theory they should. But the current gap between estimates of national employment and aggregated estimates of state-by-state employment has reached an uncharacteristically large 1 million.

To put this difference in context, adding 1 million jobs to the current U.S. total would add more than 75 basis points to job growth and likely lessen recession fears. The gap has been widening quickly since the middle of 2000, as the state aggregate has grown more than 1.1% since last year, while the national series shows an increase of only 0.3%.

According to sources at the BLS, sampling errors and difficulties at the state level are more acute relative to the national level. And state errors usually point in the same direction, so that when summing state employment figures, the small errors accumulate to yield a noticeable difference against the U.S. More importantly, when the U.S. slows precipitously, state data often experience difficulties reflecting the trend. A similar situation occurred during the 1991 recession, when revisions ended up axing more than 1 million jobs from the preliminary estimate, with several states suffering harsh and telling adjustments.

The largest discrepancies stem from the service, manufacturing and, ironically, government payroll data. Interestingly, these also happen to be the industries slowing the most severely in the U.S. series, along with the far smaller wholesale trade. The difference in government employment is consistent over time, due likely to accounting differences, making this disparity less of a concern. However, the manufacturing and service industries evade simple explanation. A glut of anecdotal evidence seems to substantiate the slowdowns purported by the national series. Manufacturing has certainly seen inimical conditions, both in the tech segment and more traditional industries of steel and chemicals. And the service industry has seen a spate of layoffs at advertising firms, tech companies, literally hundreds of dot coms and temporary help agencies.

In fact, a large part of the slowdown at the U.S. level may be traced to business services and, more specifically, temporary help services. Temporary help is classified at the 4-digit SIC level (7363), and is compiled only at the U.S. level. State and local samples rarely support robust estimation at the 3-digit level, let alone for 4-digit industries. As such, states may not be properly reconciling the slowdown in this industry with higher-level industry estimates. Indeed, estimates of the 2-digit business services industry (73), which is gauged by the vast majority of state employment offices, shows the slowdown is far more pronounced at the U.S. level than for the state aggregate.

Of the five states currently purporting the most solid job growth, three - Florida, Colorado and California - are also among the top five states in terms of business services employment share. Further, of the top 15 growth states, more than half maintains a disproportionate share of business services employment.

***************** [TABLE]

Business Service Concentrated States Show Higher Employment Growth Employment % change year-ago, 3 mo. MA Share of Business Service Employment Business Service Concentration Rank 1 Nevada 4.8 6.2 26 2 Florida 3.2 12.2 1 3 Colorado 2.8 9.8 3 4 California 2.4 9.7 5 5 Texas 2.4 7.9 11 6 Wyoming 2.3 3.5 46 7 Alaska 1.9 3.2 48 8 Georgia 1.8 8.1 10 9 Utah 1.8 7.7 13 10 Virginia 1.7 9.7 4 11 New Mexico 1.7 5.5 33 12 Hawaii 1.7 5.6 32 13 Massachusettes 1.6 8.6 7 14 Idaho 1.5 5.2 41 15 Arizona 1.5 10.0 2 ************************************

Data for 1999 that have been benched to the ES202 universe data show California, which accounts for more than 13% of U.S. business services employment, in line with U.S. business services expansion. As the U.S. reversed course in 2000, California's employment continued to accelerate. Though it has slowed more recently, it maintains modest growth, in stark contrast to the contracting business services industry at the U.S. level. Colorado follows a similar pattern. And given the tech concentration in these areas, the still healthy employment numbers they are showing are highly suspect.

Florida's business services industry is supposedly a juggernaut, currently growing by more than 11%. While it has consistently outperformed the national business services series since 1996, the divergence between growth in Florida and the U.S. has risen to uncharacteristic proportions. What's more, a chief component of business services in Florida is temporary help, according to Economy.com estimates. Even more disconcerting, the state employment office does not report temporary help's 3-digit parent industry (736), indicating that this industry is not under close scrutiny.

Despite the implications of the above analysis, state data could dodge a downward revision in the future. The U.S. employment series could be revised upward, or it may be a combination of adjustments that reconciles the state and national series when 2000 data are revised. However, given historical precedent, the sampling problems states face and continued employment growth at the state level in beleaguered industries, it is much more likely that many states will face painful downward revisions. State officials who are currently bragging how well their local economies are doing vis-à-vis the national economy may soon have egg on their faces.

-- Andre Weltman (aweltman@state.pa.us), August 09, 2001