In November, did 20,037 additional people find employment in North Carolina or did jobs at nonfarm businesses fall by 6,500? Actually, both were correct, and this simple example illustrates the problems in measuring the health of the states labor market.
Each month the federal government issues two jobs reports for the nation and for individual states. Both are not 100 percent counts but are based on statistically extrapolating the totals from a sample.
The two reports are the household employment survey and the establishment job survey. The household survey contacts a sample of households and gathers information about their work status. Unemployment rates are derived from this survey.
The establishment survey gathers information about jobs from existing nonfarm businesses. Data about hours worked, wages and the types of companies employing workers are in this survey.
The two surveys are different, give different information and have different advantages and disadvantages. The household survey indicates only if a person is employed. So if an individual has more than one job, it counts as only one employed person. In contrast, the establishment survey would recognize the multiple jobs.
The establishment survey counts jobs where the jobs are located, even if the holder of the job lives outside the state. The household survey does the opposite attaching employment only to households living in North Carolina.
The household survey is better at capturing employment of individuals in newly created businesses, in self-employed businesses and at farms. The establishment survey is run from a list of existing businesses.
A major difference between the two surveys is their sample size. The household survey uses only 60,000 households nationwide, which is 0.05 percent of all households. The establishment survey uses a sample of 400,000 nonfarm businesses covering one-third of all workers. In statistical terms, this means more confidence in the establishment survey numbers.
Hence, the two job surveys are very different in their construction and measurement. Does this mean they will automatically give different results? The standard answer is that over long periods of time, the two surveys give very similar results, but over short periods they can diverge significantly.
In November (latest data available), the household survey showed 4,314,502 persons employed in North Carolina, while the establishment survey listed 4,084,100 jobs (both numbers are seasonally adjusted). This is a difference of near 5.5 percent. Perhaps more important, both surveys show a recovery in the labor market in the state since it hit bottom during the recession. Employment in the household survey is up 5.8 percent from its bottom in late 2009, and total nonfarm jobs in the establishment survey has increased 6.4 percent since its low point in early 2010.
But comparing the two surveys over a shorter period of time shows much less conformity. Looking at the change in the counts for the first 11 months of 2013, the household survey indicates a 5,699-person loss in employment in the state, but the establishment survey gives a 51,800 gain in non-arm jobs. This is obviously a big gap.
Perhaps even more telling, in five of the 11 months the two surveys were not consistent in showing employment and jobs moving in the same direction. That is, in almost half of the months, one survey showed employment or jobs rising while the other survey suggested employment or jobs declining.
One other related issue is the measurement of unemployment. The most frequently quoted rate is criticized because it excludes as unemployed those individuals who want a job but have simply stopped looking (so-called discouraged workers). What is generally not recognized is that the government does publish an alternative jobless rate including such individuals as unemployed, as well as an even broader rate adding as unemployed those workers who have taken part-time jobs only because they cant find full-time work.
In the third quarter of 2013, these three rates for North Carolina were 8.6 percent, 9.9 percent and 14.9 percent respectively. However, each of them was down about 2 percentage points from their highs three years earlier.
I have three recommendations for people and policymakers tracking the job market. First, recognize there are two separate labor market surveys measuring different aspects of the market and giving different information. Dont mix them. Second, the two surveys are most consistent over multiple years. Third, be wary of reading too much into any short-term such as monthly result. Sometimes we expect too much from statistics.
Michael L. Walden is a Reynolds Distinguished Professor at N.C. State University.