Birth and Death in Non Farm Payrolls

The Non Farm Payrolls (NFP) is the most influential American economic statistic. In tracking the formation and destruction of employment in the United States the NFP provides a reliable view of the future course of the economy.

The Non Farm Payrolls (NFP) is the most influential American economic statistic. In tracking the formation and destruction of employment in the United States the NFP provides a reliable view of the future course of the economy. If the United States cannot generate jobs to replace the more than seven million lost since the beginning of the recession then no predictions based on returning consumer spending will be accurate. If the consumer economy does not revive, there will be no real economic recovery.

The employment data of the NFP is collected from a survey of about 150,000 businesses and government agencies employers called the ‘establishment survey’. The questions ask for a record of hiring and firings. However, not every solicited company answers the survey. Some firms have folded and some never bother to reply. The Bureau of Labor Statistics (BLS) estimates the change in employment at the firms that do not respond; they also predict the number of firms that have gone out of business and the number of new firms created. Their estimates are based on statistical models that draw on historical data.

Statisticians normally deal with the problem of non-respondents, at least for simple questions, by assuming their answers to be reflective of the respondents. If 60% of respondents preferred one election candidate, the surveyors posit that same percentage among those who did not reply.

The assumption that non answering firms will have hired and fired in similar percentages to the responding companies is similar to the BLS in the Non Farm survey. “The first component excludes employment losses from business deaths from sample-based estimation…to offset the missing employment gains from business births” *. “The second component … (is) designed to estimate the residual employment not accounted for by the imputation”, that is to compare the estimates to actual numbers from state unemployment insurance statistics. These actual job numbers are available at a several months delay, so the initial NFP release contains an educated guess of jobs lost and created by failed and new found firms. In the past the estimate had compared well to the actual numbers. “Earlier research indicated…that the net contribution (of jobs) is relatively stable”. When the unemployment insurance numbers were compared to the estimates from the BLS model, as is done once a year and included in the January statistics, no notable discrepancies were discovered. This is the benchmark or ‘birth/death adjustment’.

However, something happened to the accuracy of the model in 2008. The estimates of jobs created proved to be wildly over optimistic. Last year these adjustments subtracted several hundred thousand jobs. And in January this year 356,000 jobs were estimated to have been lost to failed companies and that is what produced the record loss of 741,000 jobs in that month.

The culprits for this underestimate of job losses are quite likely the recession and the financial crisis. Economic studies have noted that financial and banking downturns produce much slower and weaker recoveries than business cycle recessions. The credit crunch has been particularly hard on small business. It is likely that new company formation and hiring is much weaker in this recession than in prior recessions. The BLS job creation model was not accurate in 2008 and may have similar problems in 2009.

The BLS will make this birth/death revision to the jobs statistics in February with the January NFP report. According to the BLS 824,000 jobs are likely to be subtracted from those reported to have been created over the prior twelve months **. That is an average loss of 69,000 jobs per month that were not created, jobs that never existed.

But to the media, the stock market, job seekers and consumers those jobs are real and that pace of employment improvement is incorporated into their economic view. The actual performance numbers of the economy, consumer expenditures, home purchases, debt retirement and other measures are founded in part on estimates of the economic future. Those estimates may be partly based on erroneous job creation and will have to be revised when the jobs numbers are altered.

The picture of the economy portrayed by the jobs report over the past year has been one of gradual but steady improvement. From a loss of 741,000 jobs in January 2009, the numbers have gotten a little better each month, -303,000 in May and only -11,000 in November. It appears that the worst of the job losses are over and that the economy may even have added jobs in November. This is quite unlikely to be true. Or at least the pace of improvement has been far slower than given so far in the statistics.

United States total household and non-profit wealth rose by $2.67 trillion in the third quarter, primarily because of the tremendous gain in the equity averages since March, and a small rise in home values. As noted earlier, the equity averages are looking at the jobs improvement as one indicator, perhaps the most important one on which to posit a recovery in consumer spending. The feedback loop for equities runs through employment. If job creation has been considerably worse than portrayed then some of the underpinning of the equity markets, and the improvement in household wealth is taken away. If folks are once again feeling poorer, assuming the equities fall or stagnate, then quite possibly the projected consumer spending increases will be harmed or delayed.

In comparison, the ADP report reached its low in March 2009 with a loss of 736,000 jobs. In November it recorded a loss of 169,000, a 77% improvement. The NFP was at its low in January at -741,000. The -11,000 in November reading represents a 98.5% improvement over the January. At the same 77% improvement as the ADP the November NFP would have been about -170,000.

Since the last adjustment was applied to January of this year the economy has lost 3,337,000 jobs according to the current Non Farm Payrolls statistics. If we assume that the December and January numbers are flat or even slightly positive, as would seem possible given the trend, then the -824,000 preliminary estimate of the birth/death adjustment would mean that job losses during the year were close to 20% higher than estimated.

This is not just a statistical adjustment; it is a number large enough to affect employers and their hiring plans, equity, commodity and bond markets, consumers and job seekers alike. It is not a comfortable possibility.