In the past, the ADP report has been used to help forecast BLS report, but in 2008 the ADP report has grossly underestimated the job loss reported by the BLS – 1.7 million jobs to be exact. Well, ADP got a new face! Macroeconomic Advisers (models payroll estimation) advanced their econometric methodology for a new and improved ADP employment report.
In 2008 through November, the ADP report outpaced the BLS report by 1.7 million jobs, or 154,000 on average per month. There are fundamental differences between the ADP and BLS reports – size and coverage of the report (Boeing is not included in the ADP report), survey methods, definition of employess (actively paid or just listed on payroll), etc. – that prevented the ADP report from being a perfect substitute to the BLS report, but this differential is alarming.
As always, click on the chart to enlarge.
The correlation between the unrevised ADP payroll and the BLS payroll growth is 0.84 – a strong correlation. However, in 2008 the unrevised ADP payroll report has veered miles from the BLS report.Macroeconomic Advisers changed their econometric methodology for estimating the ADP payroll. Specifically, they now estimate vector autoregression model, adding the previous week's initial unemployment claims report as an exogenous innovation. You can view the presentation here by following the like titled "Webex Presentation of Revisions" for a discussion of the econometric model.
The model was estimated both in and out of sample, and Macroeconomic Advisers concludes that the revised ADP methodology will improve the ADP payroll report. In fact, the new methodology (using the in sample estimation) surpassed the BLS payroll by just 221,000 jobs in 2008, or 20,000 per month, and the correlation improved to 0.94.
The December 2008 ADP report will include the new methodologies.