Unemployment in the US: a tale of two surveys

The latest US non-farm payroll numbers increased by 275,000 for February (payroll survey) incorporating a two-month payroll revision of -167,000, most likely due to seasonal factors, leaving a net revision of just 108,000 jobs added. On the other hand, however, the unemployment rate ticked up to 3.9% as the household survey showed a loss of 184,000 jobs.

It is confusing to say the least. These two statistics, i.e. the unemployment rate and the payroll numbers, are determined by two very different surveys and sometimes give very different views of the labour market. February’s payroll number was highly anticipated given the previous month’s data had shown an elevated 353,000 payroll number and, arguably, expectations for revisions were high.

The Bureau of Labor Statistics (BLS) has two monthly surveys that measure employment levels and trends: the payroll survey (also known as the establishment survey) and the household survey. These two different statistical references can be somewhat confusing to say the least, so it is useful to outline the differences between the two surveys and hopefully add some clarity around how non-farm payroll data in the US is derived. Basically, the payrolls and wage numbers are derived from a survey of businesses and other employers, while the unemployment figures come from a separate, smaller, poll of households.

The payroll survey is designed to measure employment, hours, and earnings by industry in the non-farm sector. Conducted monthly, a representative sample of around 119,000 businesses and government agencies representing approximately 629,000 individual worksites provides the data for this survey.

The household survey, meanwhile, is designed to measure the labour force status of the civilian non-institutional population, with the unemployment rate being the best-known statistic produced from this survey. A much smaller representative sample of approximately 60,000 households provides the information for this survey. It includes the unincorporated self-employed, unpaid family workers in family businesses, agriculture and related workers, workers in private households, and workers on unpaid leave. It excludes workers on furlough as, even if they receive pay for the furlough period, they are considered unemployed.

While both are valuable for understanding employment levels and trends, they often present different pictures of the job market due to their unique methodologies and coverage.

Additionally, one survey that hasn’t garnered too much attention is the national federation of independent business (NFIB) survey, which also provides useful insights into the US labour market as it pertains to small business sentiment and is a very important barometer to watch. Produced on the second Tuesday of each month, February’s data marked three straight declines in the NFIB survey’s measure of hiring intentions, suggesting that higher interest rates and tighter credit conditions are forcing small firms – which employ about half the workforce – to tighten their belts. With labour being their biggest variable cost, owners’ plans to fill open positions continue to slow, with a seasonally adjusted 12% planning to create new jobs in the next three months, down two points from January and the lowest level since May 2020.

The rise in the most recent unemployment rate was meaningful and it’s worth noting that the household survey is based on a much smaller survey size than the payroll data. Nonetheless, when taken together with the recent NFIB survey, it is not unreasonable to anticipate a deteriorating labour market with labour cost pressures continuing to ease. Given the Fed’s data dependent policy, there may not be enough information to determine a rate cut is forthcoming in May, but a deteriorating labour market where wage pressures are easing, could be a contributing factor to the Fed’s first rate cut in June.