ADP Does A Better Job Of Estimating Employment In Real Time than BLS
Written by Steven Hansen
Quite honestly, I have been downplaying ADP's monthly employment report because the methodology changed in October 2012 - and I had the feeling this report was now trying to mimic the Bureau of Labor Statistics (BLS) data instead of being an independent source of labor data. I felt the old ADP methodology provided independent employment data, and was concerned that the "new" ADP report simply was trying to "guess" what the BLS employment numbers would be.
We now have an entire year of non-farm private employment data to compare the BLS to ADP - and the results are rather surprising. The table below compares the employment growth announced at the beginning of each month (real time estimates) to the current estimate of employment growth.
|Variance From BLS Current Estimate for 2013|
|BLS Original Estimate for 2013 (Seasonally Adjusted) - headline data||18.4% too low|
|BLS Original Estimate for 2013 (Non-Seasonally Adjusted) - basis for Econintersect analysis of BLS data||17.2% too low|
|ADP Original Estimate for 2013||11.8% too low|
The BLS uses a methodology of announcing job growth using incomplete data, and then for months following the initial release continues to revise the data. If this approach was not bad enough - then there is the annual BLS benchmarking revision.
Each year, the CES survey realigns its sample-based estimates to incorporate universe counts of employment—a process known as benchmarking. Comprehensive counts of employment, or benchmarks, are derived primarily from unemployment insurance (UI) tax reports that nearly all employers are required to file with State Workforce Agencies.
So, at the end of a years worth of data - the BLS benchmarks it - and this year the change was fairly large. Now we know that ADP's use of its internal employment records was more than 1/3 more accurate in real time in 2013 to whatever methodology the BLS uses to "estimate" in real time.
If the USA was a company, and the BLS was my control group feeding back the number of people employed - how could this company make decisions in real time? The government can spy on our internet habits - but has no good idea of how many people are employed. A 18% error rate in real time growth is unacceptable.
This is 2014 - and there is no excuse to not having accurate data in real time.
Other Economic News this Week:
The Econintersect Economic Index for May 2014 is showing marginal growth acceleration - but the pattern over a half a year remains in a fairly tight range. The major soft data point remains personal income which is not part of our Economic Index. My position remains that the economy remains too strong to recess, and too weak to grow.
The ECRI WLI growth index value has been weakly in positive territory for many months - but now in a noticeable improvement trend. The index is indicating the economy six month from today will be slightly better than it is today.
Current ECRI WLI Growth Index
Initial unemployment claims went from 329,000 (reported last week) to 344,000 this week. Historically, claims exceeding 400,000 per week usually occur when employment gains are less than the workforce growth, resulting in an increasing unemployment rate. The real gauge – the 4 week moving average – degraded from 330,000 (reported last week as 316,750) to 320,000. Because of the noise (week-to-week movements from abnormal events AND the backward revisions to previous weeks releases), the 4-week average remains the reliable gauge.
Weekly Initial Unemployment Claims - 4 Week Average - Seasonally Adjusted - 2011 (red line), 2012 (green line), 2013 (blue line), 2014 (orange line)
Bankruptcies this Week: Energy Future Holdings (EFH)