The puzzling disconnect between production and employment

One of the puzzling economic developments during the first half of 2022 (2022H1) was the sharp disconnect between output and employment growth. Data released so far suggest that the economy shrank during 2022H1. Yet, employment growth was exceedingly robust during 2022H1.

Why are companies adding so many workers if aggregate production is falling? Is the disconnect mostly the result of measurement errors or is something more structural affecting the U.S. economy? How does the output-employment divergence affect the Federal Reserve’s ongoing battle against high inflation?

On the measurement errors front, issues associated with employment data from the two surveys used to capture key labor market developments as well as the ongoing divergence between two widely used measures of aggregate output are worth noting. On the structural front, the sharp decline in U.S. labor productivity during 2022H1 and the challenges this poses for the central bank’s inflation fight are matters of grave concern.

Evaluating the state of the job market requires a basic understanding of the two key surveys – establishment survey and household survey – utilized by the Bureau of Labor Statistics (BLS) to offer a crucial overview of overall U.S. labor market conditions.

The establishment survey, formally referred to as the Current Employment Statistics (CES) Survey, generates total non-farm payroll data by undertaking a monthly survey of “about 131,000 businesses and government agencies, representing approximately 670,000 individual worksites.”

The household survey, formally referred to as the Current Population Survey (CPS), forms the basis for determining the official unemployment rate and other key labor market statistics. The “CPS is administered by the Census Bureau using a probability selected sample of about 60,000 occupied households.”

Crucially, the establishment survey counts jobs rather than people, while the household survey counts people. This implies that a person who holds more than one job can show up multiple times in the payroll data but will be counted as just one employed person in the household survey. Furthermore, unlike the household survey data, the establishment survey data is subject to multiple revisions.

In the context of explaining the output-employment divergence puzzle, the information from the two surveys initially appeared to suggest significant divergence in the pace of hiring during 2022H1. But recent analysis indicates that the gap between the two may not be quite as large as originally assumed and that, in fact, overall job market was strong during 2022H1. Job vacancy rates also suggest that the labor market was robust in 2022H1.

On the output measurement front, there is a discrepancy between real GDP (gross domestic product) and real GDI (gross domestic income). Data released so far indicates a sizable divergence between the two measures of aggregate output. It is likely that final revisions will indicate that the output growth picture was less bleak. In recent years, GDI has often yielded a better real-time estimate of actual output, and follow-on revisions have typically resulted in GDP data converging toward the GDI estimates.

So, part of the explanation for the disconnect between output and employment may be the overestimation of job growth and the underestimation of output growth during 2022H1. However, expected revisions to data will not be large enough to fully resolve the discrepancy between the level of production and the pace of hiring observed so far this year.

A more pernicious development involves the staggering decline in US labor productivity. Labor productivity is typically measured as output per worker or as output per hour of work. As such, falling output amid soaring employment implies a precipitous decline in labor productivity. Initial estimates (subject to revision) indicate that the decline in U.S. labor productivity during 2022H1 was the worst since 1947.

It is possible that the recent decline in labor productivity reflects sectoral composition shifts to some extent. Early on in the pandemic, severe job losses were narrowly concentrated in low productivity service sector jobs (restaurants, travel, leisure, retail, etc.), and, consequently, there was a sharp uptick in average labor productivity as jobs were retained (or added to) in highly productive sectors that were able to take advantage of remote work and online-based transactions.

As spending on services soared in 2022H1, we saw a sharp uptick in job growth in traditionally low productivity sectors, which very likely dragged down overall average labor productivity. On the margins, “quiet quitting” and low worker morale may also be affecting productivity.

Another potential factor driving the decline in average labor productivity is the dispersion of critical workforce away from major knowledge centers such as New York City and San Francisco. Recent research from the Dallas Fed notes that the positive network externalities associated with agglomeration and clustering of workers and firms has historically boosted labor productivity.

Now, with the rise of remote work and decentralization of economic activity, economists at the Dallas Fed suggest that “Spontaneous workplace interactions are significantly reduced. ‘Water cooler talk’ doesn’t occur, reducing the intensity of knowledge exchange and slowing the building of new relationships and professional networks.”

Even after taking measurement errors into account, it is likely that U.S. labor productivity is downshifting towards (or maybe even falling below) the subpar trend growth rate observed between 2005 and 2018. The problem for the Federal Reserve is that low labor productivity amid surging unit labor cost (labor costs per unit of output) significantly complicates the battle against inflation.

Vivekanand Jayakumar is an associate professor of economics at the University of Tampa.