Labor Market Information > Wyoming Labor Force Trends > September 2011 > CES and Projections
New Monthly Feature in Trends: Comparison of Current Employment Statistics (CES) Estimates to Short-Term Employment Projections
Beginning this month, Wyoming Labor Force Trends features a new table comparing the official Bureau of Labor Statistics (BLS) Current Employment Statistics (CES) estimates to short-term employment projections (see Table). For several years Research & Planning (R&P) has produced short-term employment projections based on Quarterly Census of Employment & Wages (QCEW) data (Bullard, August 2009). QCEW data represent a census of employment from employers' quarterly unemployment insurance (UI) tax filings. R&P's short-term employment forecasts have been used internally to analyze the CES estimates. Now, in an effort to help readers better evaluate the monthly BLS-produced numbers, R&P will publish this table every month.
The official CES numbers are estimates based on a sample of approximately 1,300 Wyoming employers. The BLS took over production of state and area nonfarm employment estimates earlier this year; previously, R&P performed this function.
As noted in the table, the projections were based on QCEW data through March 2011. Thus, they represent a forecast of only four months (April through July). In contrast, the CES estimates start with QCEW data in September 2010 and use sample-based estimates of employment change each month to produce an estimate of current employment. The BLS analysts producing the CES estimates ignore QCEW data for fourth quarter 2010 and first quarter 2011 and rely exclusively on current sample data.
The Table shows that the official CES estimates were 4,870 jobs (1.6%) higher than the short-term projections produced in August. Across the industry sectors, the largest percentage difference was in the construction sector, where the official CES estimate was 2,675 (10.9%) jobs higher than the short-term projection. Other large differences were seen in the professional & business services sector (1,220 jobs, or 6.2%), natural resources & mining (1,151 jobs, or 4.1%), and leisure & hospitality (1,043 jobs, or 2.6%).
Based on R&P's projections, it appeared that the CES estimates were too low in the retail trade sector (-973 jobs, or -3.3%), other services (-395 jobs, or -3.4%), and educational & health services (-363 jobs, or -1.4%).
Which is more accurate, the estimate based on current data from a sample of employers, or a projection based on a census of employers in an earlier period? There is no simple answer to this question. If economic conditions are stable from quarter to quarter, it is likely the short-term projection will provide a good estimate of current employment. The accuracy of sample-based estimates depends critically on the size and quality of the sample. Particularly at the industry sector level, the CES sample may be too small to produce accurate estimates. Thus, the table shows that 10 of the sectors have larger percent differences than the estimate of total nonfarm employment (1.6%). Furthermore, previous research identified data quality issues with the CES sample (Bullard, March 2009).
Based on three years of data (2007-2009), the average absolute percent error of short-term projections of Wyoming total nonfarm employment for four months out was 1.2%, a level similar to the annual CES benchmark revision.
This new feature will give those interested in the state of Wyoming's economy another tool to use in evaluating the official employment estimates.
Bullard, D. (2009, March). Methodological note: a caution regarding employer reports to the Current Employment Statistics program. Wyoming Labor Force Trends, 46(3). Retrieved October 26, 2011, from http://doe.state.wy.us/LMI/0309/a2.htm
Bullard, D. (2009, August). Can Wyoming employment be modeled accurately? Wyoming Labor Force Trends, 46(8). Retrieved October 26, 2011, from http://doe.state.wy.us/LMI/0809/a1.htm