The Employment Rorschach Test

Two roads diverged in a yellow wood,

And sorry I could not travel both

-The Road Not Taken, by Robert Frost

Last week we received the Bureau of Labor Statistics’ May jobs report. The jobs report, more formally known as the Employment Situation, is composed of two key components: the Establishment Survey and the Household Survey. The Establishment Survey covers roughly 120,000 businesses and government agencies and estimates the number of nonfarm employees, hours worked and earnings. The Household Survey polls around 60,000 households and tabulates individuals’ employment status, which is used to calculate the closely followed unemployment and labor force participation rates, among other statistics.

Bond markets initially reacted negatively to the better-than-expected headline number of 272,000 jobs added during the month and acceleration in wage growth reported in the Establishment Survey. The yield on the 10-year U.S. Treasury rose about 15 basis points on the day of the release as the market priced in fewer interest-rate cuts from the Federal Reserve in 2024. In contrast, the Household Survey indicated that not only did employment contract in May, but it declined over the last six months. As this week’s chart shows, the divergence between the Establishment and Household surveys (adjusted to match the construction of Establishment payrolls) has grown with the Establishment Survey indicating 4.1 million more jobs than the Household Survey. Which data series is better gauging the health of the economy?

The actual state of the U.S. employment market probably lies somewhere in the middle. Employment reported in the Establishment Survey tends to be less volatile than employment reported in the Household Survey. This discrepancy could be due to the Establishment Survey being dependent on its birth-death model, which has accounted for 93% of employment gains in 2024. This compares to 66% over the same period in 2023 and 41% in 2022, according to calculations from Morgan Stanley. The large contributions from this model have led to persistent downward revisions recently to the historical data as more accurate tabulations of employment data are captured in the benchmark data.

The Household Survey, on the other hand, might be undercounting the employment impulse from immigrants, including those entering the country illegally but authorized to work. The Household Survey is susceptible to volatility in the underlying components. For example, the under-25-years age cohort represents nearly all job losses in the last six months. The unemployment rate, which is part of the Household survey, has risen 0.6% from the cycle low to 4.0% and is now above its 36-month moving average, a strong historical predictor of recession.

Broader measures of the labor market, like the Kansas City Federal Reserve Bank’s Labor Market Conditions Indicators, which track 24 labor market variables including employment statistics from both the Establishment and Household surveys, have softened since peaking in 2022 but remain well above their long-run average. The ratio of job openings to total unemployed has normalized to the pre-pandemic peak, a sign that the tightness in the labor market is abating but not collapsing. Higher-frequency indicators, like weekly initial jobless claims, signal laid-off workers are finding new jobs quickly. The divergence in the labor market makes the Fed’s job more difficult, as it adjusts monetary policy based on a dual mandate of promoting maximum employment and price stability.