Few data can make the market excited but also nervous like non-farm payrolls. As an important indicator of the health of the US economy, it plays an important role in the Federal Reserve and investors' assessment of future interest rate trends.

However, when data such as non-farm payrolls, unemployment rates and wages show that the labor market is still hot and there are a lot of job opportunities, which is inconsistent with the economic situation, high interest rates, reduced recruitment by small and medium-sized enterprises and weak growth momentum, it is very important to dig out the abnormal reasons behind the good employment data. For example, previous analysis has found that a large part of the new jobs in the United States are driven by the health and social care sectors.

A major contradiction in the US employment data is that the non-farm data obtained through institutional surveys shows that jobs are constantly increasing, including 272,000 new jobs in May, but the household survey data shows that employment is slowly declining, with 408,000 jobs lost last month. The two surveys have different samples and scopes, and the difference is expected. The problem is that they are not in the same direction and tell completely different stories.

Institutional and household survey results differ widely

So what are the main differences between the two surveys? The institutional survey sampled more than 650,000 employers, while the household survey sampled 60,000 households. Employees in multiple positions were counted multiple times in the institutional survey, but only once in the household survey, which also includes the agricultural sector. In addition, due to the different samples, the two surveys also rely on different models for aggregation. The institutional survey does this by estimating the value of enterprise registration and deregistration data, while the household survey is based on population forecasts. Another problem is that the response rate of the enterprise survey in particular has dropped significantly since the epidemic, which may also affect its accuracy to some extent.

Corporate survey response rate drops sharply

Essentially, there are four sources of discrepancy between the two survey results: the underlying sample, the data definition, and errors caused by scaling up the sample data based on population and firm count estimates. The first is difficult to quantify, but we can assess the other sources of discrepancy.

Rogier Quaedvlieg, senior economist at ABN Amro, delved into the "labor market conundrum" in the United States. He found that the two surveys had roughly the same definition of data before 2022, but then diverged. As for the third and fourth sources of difference, the household survey may not fully reflect the large influx of foreign workers into the United States in recent years, which may be more easily reflected in the employer survey of the non-farm payroll report. Secondly, after the epidemic, it has become more difficult to estimate the registration and deregistration rates of enterprises and the impact on employment, partly due to changes in the way of working and the rise of new business models. In terms of population, the U.S. Bureau of Labor Statistics assumes that the net birth rate is higher than before the epidemic.

In summary, ABN Amro finds that the gap between the two surveys is largely due to underestimation of immigration, overestimation of business additions and deletions, and data definitions. As a result, when the actual population and business registration data are released, the US employment data may also be significantly revised, which is not uncommon.

So what's the takeaway? The U.S. labor market may be weaker than the nonfarm payrolls suggest, but not as bad as household surveys suggest. The Fed and the market may want to look more closely at other economic data for clues about the true performance of the job market, including unemployment claims, hiring surveys and the unemployment rate.

In addition, there were earlier reports that the U.S. Bureau of Labor Statistics will need to cut its sample size due to budget constraints, and the next White House owner may have to consider taking more measures to support this most important statistic in the United States.

The article is forwarded from: Jinshi Data