Methodology: A county slice in time

This project set out with the seemingly straightforward task of seeking answers to these questions: How often, and why, do counties experience major employment loss? What happens to those counties' economies when they do?

Such a project requires that research parameters be established—including the unit of study, the time period to be studied and the employment decline necessary for a county to register as "shocked" (and thus worthy of further investigation).

In this case, counties were chosen for study over cities or other economic jurisdictions for the simple fact that this is the smallest economic "unit" with the most comprehensive and comparable data. To create a subset of counties that experienced the most severe employment decline, the bar was set at a 3 percent loss of jobs, which represents the highest 4 percent of employment declines during the study period. A numeric loss (say, of 200 or more jobs) was ruled out in favor of a percentage loss because numeric loss would have eliminated a number of small counties in the district from consideration because they have fewer than 1,000 jobs to start with.

The time period of 1990 to 1993 was selected because it met three basic criteria:

• It offered a manageable number of "shocked" counties to analyze (in this case, 40 counties, which experienced 48 total annual employment shocks).
• It was far enough in the past to observe recovery patterns in both employment and income.
• It was recent enough for local officials to recollect the county's economy at the time of the employment decline, thus anecdotally providing insights and analysis on the cause of employment shocks where government data typically cannot, in part because state and federal tracking of mass layoffs and other employment events at the county level do not go back to this period.

Having established these parameters, various data were collected to investigate the existence of relationships and tendencies. Annual data on total employment from 1989 to 2002 were collected from the U.S. Department of Commerce's Bureau of Economic Analysis for each of the district's 303 counties. (The BEA uses figures originally published by the widely cited Bureau of Labor Statistics, but then does supplemental work to fill in some data gaps in BLS figures. As such, BEA figures are likely more thorough, though less timely.)

For all 303 counties in the Ninth District, the annual percentage change in employment was calculated for each county during four periods: 1989-90, 1990-91, 1991-92 and 1992-93. Next, any county that suffered an annual employment loss of at least 3 percent in any one or more of those periods was flagged for further study.

In all, 40 counties qualified as shocked, or about 13 percent of all district counties. Eight counties appeared more than once, meaning they suffered multiple years with at least a 3 percent employment decline. Miner County, S.D., had this unfortunate distinction. Slope County, N.D., was the only county to endure three shocks.

To measure a county's employment recovery, the annual average growth rate was computed from the year of employment decline (or shock) to a "smoothed" 2001 figure (which is the average of 2000 to 2002 employment levels to smooth out any unusual one-year movements).

Next, statistics on income, educational attainment, age and industry composition, poverty rates and proximity to a metropolitan area were collected for each county. The analysis examined whether counties that suffered employment shocks had different overall characteristics than nonshocked counties before employment drops occurred. These statistics were also used to explore whether the counties that recovered successfully had different initial characteristics than those counties that did not. Certainly, the list of variables chosen for this analysis is not exhaustive; others could be used, but these were believed to be both familiar and diverse enough to uncover trends that might be described to fedgazette readers.

A point of emphasis: Each county has the same importance in this analysis, whether it is Hennepin County, Minn. (1989 employment of almost 900,000) or Petroleum County, Mont. (1989 employment of 269). So while a handful of counties account for a large share of employment and overall economic activity in the Ninth District, this analysis does not weight or factor in a county's employment size, but rather views each county as a relatively equal and discrete unit.

Return to: County employment: Shocks and rebounds