A Better Way to Forecast BLS Employment Data?
Every month, we forecast changes to levels of US Nonfarm Employment as reported by the Bureau of Labor Statistics (BLS) in its Employment Situation Summary. The BLS figures typically come out on the first Friday of the month, reporting on the previous month's employment counts, and according to the following schedule.
Our goal is to release our estimate two weeks prior to the release of the BLS figures. We can prepare this estimate in advance because we consolidate and analyze Hiring Demand data very rapidly following the close of the BLS survey period.
There are other sources of such forecasts, one of which is the well known "Consensus Estimate". The Consensus Estimate is usually the mean or median estimate of dozens of professional forecasts.
To keep track of the relative accuracy of our forecast and that of the Consensus Estimate, we provide a running calculation of their forecast standard errors:
Table: Monthly Change in Seasonally Adjusted Nonfarm Payroll Employment (thousands) and Historical Forecast Standard Errors.
|Source||Previous 4 Months||Since Jan. 2007|
|Jan 2010||Feb 2010||Mar 2010||Apr 2010||Avg. Error||Avg. Error, %|
(Difference with BLS)
(Difference with BLS)
The BLS issues three successive "estimates" of each month's employment levels. The third release is the "final" one, and is presumably closer to the truth than the first two preliminary releases. This is why we can compute a "BLS forecast error" – the first preliminary BLS data release can be viewed as a forecast of its own third and final estimate.
Our forecast is unique in that it uses "Hiring Demand" data (white paper), which is the number of new online job ads appearing on employment-specific websites. This data is available for download through our Hiring Demand Dashboard. Professional forecasters can include job ad data as a co-variate in their forecast models. Data series are also customizable to focus on individual companies, groups of companies, industries, or geographic regions.
Prior to each month's release, we publish our estimates alongside the other estimates.
We also keep track of the most recent month's estimates and the BLS's successive monthly revisions.