ADP Employment Reports, Truthiness and CDOs
I've taken issue with the accuracy and relevance of the monthly ADP Employment Reports, which use a large sample of private payroll data to estimate what's happening in the economy. (Here, for instance.) ADP has hired an economics firm to do a lot of fancy estimating, and lately the thing gets all sorts of attention, not just because it comes out a day before the government payroll numbers, but because CNBC gets everyone into a lather and talks about it as a market-moving data point. It's a great public relations opportunity for ADP.
My beef has been with its misleading techniques in estimating employment changes at small, medium and large firms, but today I am happy to add that a world-renowned economist, Paul Kasriel of Northern Trust, has thrown up his hands at the entire undertaking:
The ADP National Employment Report estimate for May is an increase in private nonfarm payrolls of 55,000. Before revision, the April ADP estimated an increase in private nonfarm payrolls of 32,000, subsequently revised to an increase of 65,000. (Why does ADP revise its data? Does not the company know with certainty how many payrolls it processed in a month?) The first guess by the BLS for April private nonfarm payrolls was 231,000. The chart below shows the difference between revised BLS and ADP estimates changes in private nonfarm payrolls in the 12 months ended March. Over these 12 months, the average absolute difference between revised BLS private nonfarm payrolls and revised ADP private nonfarm payrolls has been about 83,000 - not insignificant from an economic "truthiness" perspective.Here's the chart he's talking about. True, it's toughest to make forecasts at turning points in the business cycle, but these forecasts have been way off.
Click on the graphic for a larger image
Kasriel adds:
So, the ADP estimates of employment, even after revisions, do not provide much guidance in forecasting BLS estimates of employment. Just as CDO2s do not add much to the benefit of society, neither do ADP estimates of employment...
...
I would find the ADP underlying unadulterated data more important than the adulterated BLS nonfarm payroll data.I've tried talking to the ADP economists, without much impact. Maybe Paul Kasriel can get the message across.