Thursday, February 7, 2013

IBO | Uses PUMS Data to Clarify NYC Job Trends

This is a good news story for people who care about trends in jobs and unemployment in the nation, state, and city.

The Independent Budget Office has used the American Community Survey to create a new household employment and unemployment series. The ACS is a new (since 2005) data collection program of the Census Bureau that uses a sample of 250,000 U.S. households to generate more frequent data on demographic and other trends between decennial Censuses. The IBO unveiled the new series in its February 2013 Fiscal Brief (http://www.ibo.nyc.ny.us/iboreports/febacsemployment2013.pdf).

This development is welcome because for decades the household employment numbers (based on a monthly sample of 70,000+ households nationwide) have been less consistent than the payroll jobs numbers, which are based on a much broader base of unemployment insurance filings monitored by the labor departments of every state and collected by the Bureau of Labor Statistics.

The divergence between the two numbers sets up a controversy as to which number to believe. On Long Island last year, an unsuccessful challenger to Rep. Tim Bishop (NY-1) sent out expensive mailings misusing the household survey to argue that if 100,000 more people were unemployed, it meant there were 100,000 fewer jobs. In fact, payrolls were rising during the period. The increase in unemployment meant that a handful of people were reporting that they had decided to look for work, and their response was inappropriately extrapolated and mis-characterized.

Two points to bear in mind, to avoid misusing the unemployment numbers, are: (1) They are based on a tiny sample when used for regional analysis as small as a Congressional District, and (2) They represent a sub-sample of people who are self-declared as looking for work.

In New York City, for example, if the sample of households is 1,500 and we are measuring the number of unemployed, and there are two people per household, an unemployment rate of 9 percent means that the 1,500 telephone respondents are reporting to the Census Bureau (acting for the BLS) that 270 people out of 3,000 have recently been looking for work. That's a tiny number of people on which to base a widely watched number. The unemployment rate can jump around as people in the responding households stop looking one month and start looking again the next month.

The fluctuations could feed on themselves, as a small change in the rate is reflected in headlines of New York City newspapers (or, on Long Island, Newsday) and create a false sense that the job market is getting better or worse. The fluctuations in an area could simply reflect changes in the weather. The BLS has tried to adjust the numbers to reduce the volatility, but that create a whole new set of problems, as real turning points are assumed to be noise.

So it's a real help that the IBO has identified another source of information as a benchmark for checking on the household survey, the Public Use Microdata Sample (PUMS) of 250,000 households that is used for the American Community Survey. This sample is rich in demographic data and being larger is more reliable, especially for use at the local level.  The data go down to the PUMA,  a local geographic area, which in New York City are the 50+ Community Districts overseen by Community Boards in each borough and are roughly equivalent in size to a Council Member district, although the political districts are not coterminous with electoral districts.

The IBO has already used the PUMS data to analyze traffic congestion, finding that 1.9 million people commute into the New York City congestion zone to work and of the commuters by car, more than one-fourth are from New Jersey and nearly 10 percent are from Bergen County alone (see http://www.ibo.nyc.ny.us/newsfax/insidethebudget154.pdf).

The bottom line of the new ACS-based series is that the recent extrapolated changes in employment runs between the payroll numbers and the household numbers. This suggests that the high BLS numbers for unemployed New Yorkers, despite strong growth in payrolls, are measuring something real. Several theories about what has been happening that have been widely promulgated - such as an increase in commuters - are shot down by the Microdata numbers. The persistently high unemployment rate in New York City is confirmed as real, and the IBO explanation is simply that as the economy is growing, more people look for work, thereby growing the labor force available for hire. Nothing new about that.

What is new is that we have another movie camera with which to follow these trends. Thank you, IBO.