Recent economic data has produced conflicting stories: total employment has fully recovered all losses caused by the pandemic, but the labor force participation rate has not fully recovered. Resolving the statistical conflict provides important insight into the state of the economy and also a lesson in interpreting economic data.
In February 2020, the number of jobs peaked at 152,504,000. It fell 14% in just two months and then slowly increased. In July 2022, it finally regained its peak, reaching 152,536,000. But what does that mean? The number of jobs comes from the Bureau of Labor Statistics’ Current Employment Statistics report. It counts jobs, not people working. Thus, the person who holds two part-time jobs is counted twice, once for each job.
The participation rate comes from the Current Population Survey, which counts people. We ask people if they work. A scholarship holder with two jobs is counted once. The owner of a small business is considered to be working even if she does not have a salaried job but is instead profiting from her business.
One last statistic before we get to the bigger picture: the data sources are wildly different. The number of jobs comes from all large companies and from a sample of small and medium-sized companies. It’s pretty accurate but runs into a problem when big changes happen in the economy. The sample of small and medium enterprises is extended to correspond to the entire universe of small and medium enterprises. So if the government sampled 5% of businesses, it would multiply its number by 20 to cover all businesses. But the government has not known, at least for many months, whether its sample covers 5% or 4% or 6% of small businesses. This is because new businesses are being created all the time and some businesses are closing every day. Next year, all businesses will be counted, but in the meantime, the Bureau of Labor Statistics is making an educated guess.
But the survey problem is even greater for other data. Government workers telephone a sample of households and ask questions such as “Did you work last week?” and if the answer is no, the follow-up question is asked: “Have you looked for work in the past four weeks?” Labor force participation means that the person was working, looking for work, or waiting for a job to start. This survey does not affect a person’s eligibility for unemployment insurance benefits, but some people are likely suspicious. They may respond that they looked when they did not look at all. This question highlights the risks of using people surveys.
During my first years of work as a professional economist, I learned that a researcher is never so happy to dig into the source of the data.
For the comparison of recovery in employment and recovery in labor market participation, don’t think of two very thin and precise lines; instead, think of two blurry, blurry lines. We can’t really be sure right now if they are above or below their February 2020 level, but we do know for sure that they have improved over the past two years. We can say with great confidence that both metrics are within range of their previous peak, especially if we’re vague about what “within range” means.
The biggest lesson from these stats is that business leaders shouldn’t focus on lines as hard as “exceed previous peak” or “two quarters of negative growth” or “20% decline in stock prices” . These are expressions of journalists of little importance to business leaders. They imply false precision of the underlying data, and they imply greater significance than the data warrants. An economy that has recovered 100.1% of its past decline is not much better off than one that has recovered 99.9% of its decline.
An even more important lesson for business leaders would be to ignore most headlines and focus on the handful of economic factors most relevant to the business, such as with an economic dashboard. Looking at a consistent set of data known to be closely related to the business will help business leaders focus on what’s important and avoid the rabbit holes of economic statistics.