Non-farm payrolls are coming! Will one of these two employment data points be wrong?

The Federal Reserve always pays extreme attention to the US labor market. Therefore, the release day of non-farm payroll data is often one of the most important days on the bond trader's calendar.

Of course, it's not just bond holders who care about this issue. Stock prices also react, although we are not willing to delve into the question of "when bad data is beneficial or detrimental to the stock market."

In addition to non-farm employment numbers, there is a series of other official employment data released at the same time. These data come from two sources — establishment surveys and household surveys.

Establishment surveys estimate hourly wages, hours worked, income, and employment numbers by sampling a range of businesses and government agencies. Household surveys estimate data such as the unemployment rate by sampling a range of households. Both surveys are from the US Bureau of Labor Statistics. Both surveys are conducted in the week before the 12th of each month. They together form a puzzle of the labor market that analysts and policymakers can interpret.

However, in the past few years, these two surveys have provided drastically different information on employment growth. According to the household survey, total employment showed a mediocre performance after a strong rebound from the impact of the COVID-19 pandemic last fall. Meanwhile, according to the establishment survey, employment growth has remained strong.

The household survey showed a decrease of about 700,000 jobs since its peak in November last year, while the establishment survey showed an increase of about 1.6 million jobs, which is not a small margin of error.

In short, the Federal Reserve's next steps will partly be based on which version of the data they choose to believe.

Barclays economists Jonathan Miller, Mark Janoni, Pooja Sriram, and Colin Johnson firmly support the establishment survey.

Analysts, clients, and the media quickly complained that the establishment survey is misleading. This view has its merits: the establishment survey's methods for estimating business births and deaths are opaque, and it undergoes regular downward revisions compared to more accurate employment measures. This suggests that the strong employment data supporting the Fed's decision not to cut rates should be downplayed, and a faster and more substantial rate cut should be a priority.

However, Miller and others believe that those opposing the establishment survey are mistaken. They argue that the household survey is the unreliable narrator of the labor market. As they put it:

While it is reasonable to speculate that the establishment survey somewhat exaggerates employment growth, our analysis unequivocally shows that since 2022, the household survey has significantly underestimated job growth. While theoretically this mismeasurement may not affect the calculation of the unemployment rate, it is quite likely that the degree of increase in the unemployment rate is lower than the reported level.

After a detailed analysis of over 5,000 words, they make their argument. In short, despite the shortcomings of the establishment survey, it corrects quarterly employment and wage surveys (QCEW) every January.

In contrast, the household survey has never been corrected. By cross-validating the household survey, QCEW, and data released by ADP, Barclays believes that the household survey significantly underestimates new job positions In addition, they believe that the small sample size of household surveys, declining response rates, and the method of readjusting proportions make it statistically unreliable, while incorrect population estimates mechanically limit the growth of household employment.

They concluded: Household surveys have exaggerated job growth and the cooling of the entire labor market, which is hard to avoid.

They did not categorically state that the rise in the U.S. unemployment rate is meaningless. It is worth noting that other analysts believe that due to demographic issues affecting both the numerator and denominator of household surveys, the unemployment rate will still be a relatively accurate reading