Unlike much of the nation, New York City has seen a robust rebound in employment since the recession. In early 2012, employment here reached 3.86 million, the largest number of jobs ever recorded. Yet the city’s unemployment rate has risen in recent months and is now 10 percent—its peak during the recession—and well above the 5 percent rate seen before the downturn.
Two estimates of New York City employment are reported each month—the count of the number of jobs (based on a survey of business establishments) and a count of the number of people employed (based on a household survey).
Between August 2008 and December 2010 the establishment survey showed that New York City lost 130,000 jobs, or about 4 percent of total city employment, and the household survey showed comparable declines.
As of June 2012, however, the establishment survey showed that city employment had rebounded by almost 200,000, reaching an all time high, while the household survey showed no rebound at all. In fact, the unemployment rate, which is calculated from the household survey, has recently crept up—from 9.1 percent in December to 10.0 percent in June.
[Anecdotes and arrows in dark red by Mish]
We look to three possible explanations for why these two employment measures have diverged so sharply in New York City. First, and most obvious, is commuters. Jobs in the city held by people who commute from the rest of the metro area are counted in the establishment survey, but not in the household survey. If most of the new jobs were going to commuters and few to city residents, that would help explain some of this divergence.
A second possible explanation for the rise of the gap might reflect the treatment of self-employed workers in the two surveys. As the recovery took hold, it could have been the case that large numbers of workers shifted from self-employment (counted in the household survey, but not the establishment survey) to a job in a business (counted in both surveys). This shift would not be reflected in the household survey, because the worker was already counted as employed, but it would show up as a rise in the establishment survey. While current self-employment data are not available for New York City, nationwide data indicate that there has indeed been a shift away from self-employment since 2009. If a similar pattern occurred in New York City—even to a considerably greater degree—it would only explain a fraction of the divergence between the two employment measures.
Finally, multiple jobholders could be a factor. A resident of the city who holds two jobs would be counted as one employed person in the household survey, but counted twice in the establishment survey. An increase in multiple jobholding in the city during this recovery would give rise to a gap. As with the self-employed, we do not have direct evidence of multiple jobholding among New York City residents, nor is there evidence of a rise in multiple jobholding in this recovery at the national level. So unless the city deviates significantly from the nation, this explanation holds little water.
Major Turning Points
One of the reasons the BLS uses the household survey at all is that at major turning points the household survey tends to lead. At the onset of recoveries, new business startups lead in recoveries. At the onset of declines, small businesses going out of business would tend to lead declines.
In this case, for New York City, there was no recovery at all as measured by the household survey.
In an attempt to get the establishment survey and the household survey in sync, the BLS uses a Birth-Death Model hoping to compensate for the lag.
Even still, there are problems as noted by the BLS.
There is an unavoidable lag between an establishment opening for business and its appearing on the sample frame and being available for sampling. Because new firm births generate a portion of employment growth each month, non-sampling methods must be used to estimate this growth.
The net birth/death model component figures are unique to each month and exhibit a seasonal pattern that can result in negative adjustments in some months. These models do not attempt to correct for any other potential error sources in the CES estimates such as sampling error or design limitations.
Note that the net birth/death figures are not seasonally adjusted, and are applied to the not seasonally adjusted monthly employment estimates to derive the final CES employment estimates.
In attempting to adjust the establishment (payroll) survey via the birth-death model, the BLS admits it may have introduced even more errors. Moreover, the BLS admits "lags". Those lags are especially critical at turning points.
Some believe this is purposeful manipulation. I don't. I think it is modeling error.
Regardless, if the US (in this case New York City) is at a turning point, the household survey is more apt to pick that up.