Picture of Christopher A. Hopkins, CFA

Christopher A. Hopkins, CFA

Jobs report highlights data collection woes

The latest report from the Bureau of Labor Statistics presented a picture of a normalizing employment market that bodes well for continued progress against inflation and showed that the economy is gradually slowing but still holding up well.

But below the surface, the much-anticipated monthly jobs tally presented some starkly conflicting data, a distinction that was not widely highlighted among media reports of surprisingly strong headline jobs number in May. For example, how could the US economy blow away expectations for new jobs while at the same time the unemployment rate rises? Curious.

The conundrum illustrates a growing problem for government data nerds as well as policymakers who rely on that critical data. Collecting reliable and timely employment information by conducting surveys is becoming increasingly difficult as response rates plummet. Just ask any political pollster.

The jobs report is a compilation of two monthly surveys conducted by the BLS. The first is a sampling of 119,000 businesses and government entities, called the Current Employment Statistics Survey but commonly called the “Establishment Survey”. The Bureau contacts business establishments each month to inquire about the number of non-farm jobs, hourly wages, and hours worked. From this sample, a first estimate of jobs created or lost in the economy is interpolated and successively refined over the following 2 months as more responses are received.  The BLS reported that 272,000 new jobs were created in May, far more than the consensus forecast and the headline in most media reports.

The other half of the employment report, the Current Population Survey, is broadly known as the “Household Survey”. Here the BLS contacts around 60,000 households and asks how many residents are employed or looking for a job, and whether they are working full- or part-time. The Household survey has a higher margin of error but picks up data on contract workers and the self-employed that do not show up in the business survey. Contrasting with the Establishment gain of 272,000, the Household survey estimated that the US lost 408,000 jobs in May and 800,000 since November, with the unemployment rate rising from 3.7% to 4.0%. Since December 2022, the cumulative gap between the Establishment and Household job counts has grown to over 2 million. What gives?

An important contributor to the differential in the two surveys is immigration. Most economists now believe that the household survey markedly undercounts immigrants who are employed and contributing to the economy, based upon new research from the Congressional Budget Office. This is important because economic expansion requires a growing labor force, but native-born Americans haven’t made enough babies since Y2K, so a larger number of foreign-born workers is necessary to grow the economy. Note that this is not a comment on border enforcement or immigration policy more broadly, but a simple mathematical observation. The additional workers entering the US have expanded the labor force and contributed to GDP growth and help explain why the tight labor market has not proven to be even more inflationary.

Sophisticated research agencies like the BLS can adjust estimates based on a better understanding of immigration patterns. But what if the people being surveyed just won’t answer? It is becoming increasingly challenging to gather quality data from the Household survey as fewer people agree to participate, creating a dilemma for both consumers and policymakers.

A decade ago, 90% of people contacted in the Household survey responded to the inquiry. Today the response rate has fallen to around 70% and dips into the 60s during some months, presenting a huge obstacle to collecting essential data to inform timely policy decisions. Various explanations have been proposed, including the decline in land lines, fewer people answering calls, privacy concerns, and an increase in anti-government sentiment.

The decline in response to critical government data surveys has real-world consequences. Obviously, the smaller the sample size, the less reliable the data. Lower response rates also suggest the possibility of self-selection if the folks who refuse to answer share some traits in common, introducing bias into the data. Private pollsters grapple with this factor as a partial explanation for notoriously erroneous political polling, prompting many firms to abandon phone surveys altogether.

The Bureau is actively engaged in modernization efforts to address this behavioral shift, including video-based interviews, a web-based collection system to compliment phone and in-person contacts by 2027, and new operational software to increase efficiency and mitigate language barriers. The BLS also publishes research to inform consumers of their data of survey limitations.

The divergent jobs surveys do not signal an impending recession, but they do suggest a turning point. A dominant theme of the post-COVID economy has been a perplexing labor shortage which is finally abating. BLS reported 1.2 job openings for every unemployed worker in May, down from 2.0 at the peak in 2022 and close to equilibrium. Historically, the Household job count has peaked near the top of each economic cycle while the Establishment survey tends to lag. That peak occurred last November, and anecdotal evidence also suggests jobs are getting harder to find. Remember “quiet quitting”? Forget it and get back to work.

Taken together, the surveys show a labor market is normalizing after an anomalous distortion as the economy begins to cool to the right temperature for sustainable growth and easing inflation. And if you receive a call from the BLS, please answer the phone. The next Fed meeting may depend on it.

Share this post