Resolving the Payroll-vs.-Household Debate
By Barry Ritholtz
RealMoney.com Contributor
8/13/2004 3:23 PM EDT
A week after the disappointing July employment report, it is hard to imagine that some economists are still sticking to the canard that the household survey more accurately reflects job creation and economic expansion than the payroll survey.
Several economists -- with agendas ranging from "refusal to admit error" to pure partisan politics -- continue to claim that the divergence between the two surveys understates the true strength of the economy.
How significant is the difference between these two data points? Consider the recent July payroll data, which came in significantly below expectations, with only 32,000 jobs created vs. a consensus of 240,000. The household survey covering the same month showed 629,000 new jobs. This might make investors more bullish, if they did not fully understand how the household survey is different from the payrolls survey. About 92% -- 577,000 of the jobs in the household survey -- were part-time only. Further, these 629,000 jobs also included uncompensated family employees and agricultural jobs, which, by definition, are not included in the nonfarm payroll survey.
An objective review of the data thoroughly discredits the "household survey is stronger" rationale, which has become the last refuge of economic scoundrels.
For those of you less interested in drilling down through the economic data, consider these comments from Federal Reserve Chairman Alan Greenspan during his congressional testimony at the House of Representatives on Feb. 11: "I wish I could say the household survey were the more accurate. Everything we've looked at suggests that it's the payroll data which are the series which you have to follow."
If you choose to ignore the congressional testimony of the Federal Reserve chairman, than consider the source of these reports themselves: The Bureau of Labor Statistics has recently looked into the statistical divergences between the two data sources and showed why the so-called household/payroll survey discrepancy is in actuality a "non-issue."
Different Methodologies
First, let's review the methodological differences. It turns out they are quite significant. The household survey relies on asking people -- in person, at their homes -- about their personal employment situation. A total of 60,000 households are contacted this way each month, according to analysts I spoke to at the Bureau of Labor Statistics, Current Population Survey.
The establishment, or payroll, survey, on the other hand, takes data directly from 400,000 established employers, typically though a payroll or human resources person.
The two different data series are extrapolated out to produce statistics on the 129.9 million employed people in the U.S.
The smaller survey size -- BLS contacts only one-seventh as many households as businesses -- helps explain why the household survey is so much more volatile, when projected out to the 100 million-plus level. Analysts at the BLS advise not putting too much stock into any one data point. What matters is the overall trend of job creation. Hence the reliance on moving averages by most economists.
Once we get past the quantitative data, consider the qualitative side of surveying: What is the objectivity of the persons providing the data? The payroll survey is derived from 400,000 businesses. It is a summary of corporate payroll data. There is simply no reason, nor any upside, for a corporate human resource person to falsify this information.
It is quite easy to imagine, however, an individual "puffing up" their own situation: It's simply a matter of pride or ego. Perhaps this helps explain why, ever since the technology bubble burst, there has been such a large increase in the numbers of "self-employed, work-at-home contractors" in the household survey. Very often, the phrase "work-at-home contractor" is merely a polite euphemism for being unemployed.
Different Jobs Measured
Second, let's look at what they measure: The two surveys actually quantify different jobs. When the BLS modified the household survey "to make it more 'similar in concept and definition' to the payroll survey," this divergence essentially disappears.
How? The BLS subtracted from the household survey those jobs not represented in the payroll data, specifically all agricultural and related employment, self-employed, unpaid family and private household workers, and workers absent without pay from their jobs.
The use of the broader standard (including farm labor, unpaid family workers and part-time employees) is what created the divergence. This is seen in the green line in either of the charts below. Using data "similar in concept and definition" to the payroll survey, the BLS found, eliminated the phantom missing jobs.
The BLS notes: The household series presented here has been smoothed for population control revisions. The "adjusted" household series has been smoothed for population control revisions and adjusted to an employment concept more similar to the payroll survey. Shaded area indicates recession. Source: Bureau of Labor Statistics , March 5, 2004
The BLS notes: The household series presented here has been smoothed for population control revisions. The "adjusted" household series has been smoothed for population control revisions and adjusted to an employment concept more similar to the payroll survey. Shaded area indicates recession. Source: Bureau of Labor Statistics , March 5, 2004
Conclusion
One of the inherent challenges of the market is what I like to call the "folly of forecasting."
Since the future is unknowable, the best we can do is make imperfect predictions based upon what we know at present. I give other analysts and strategists leeway when their conjectures prove incorrect -- as long as their models are rigorous and their intentions sincere. Call it the rule of "cutting the other guy some slack."
I have, however, no tolerance for erroneous economic expectations based upon faulty data, hidden agendas or personal biases. We saw examples of this during the heyday of the bubble years, when analysts regularly skewed their research in order to serve their investment banker masters, much to the detriment of investors.
This most recent episode with the household/payroll surveys has been similarly disingenuous.
To quote John Irons, senior economic research and policy analyst at OMB Watch, a Washington, D.C., nonprofit: "The [fastest] way to lose all economic credibility is to claim that either, one, the household survey is a better measure of employment than the payroll survey, or two, that 'economists disagree' about which survey should be used."
I couldn't agree more.
My research assistant Julie Heckman, soon to be heading back to Cornell, provided invaluable help with this column.
thestreet.com |
