Diffusion Profusion: Shares of Industries Contributing to Job Gains and Losses
Payrolls grew by a strong 272,000 last month, well above expectations, and the unemployment rate ticked up slightly to 4%. Since President Biden took office, payroll employment is up by an historically large 15.6 million and May marks the 30th consecutive month with the unemployment rate at or below 4.0%. While the overall labor force participation rate ticked down in May, the prime-age (25-54 year-olds) rate ticked up, with working-age women posting their highest participation and employment rates on record, based on data going all the way back to 1948. Wage growth was up 4.1% on a yearly basis last month, and while we don’t yet have May inflation data, that pace will likely beat price growth, last seen rising at 3.4% in April. For more details on today’s report, see CEA’s X thread. [link]
Of the many labor market statistics that get bandied about on employment-report days, one you don’t hear a lot about is the diffusion index. Yet the diffusion index (DI) shows up every month on the ever-important “Summary table B” with headline findings from the Establishment Survey. In today’s report, the index was 63%, meaning 63% of industries added jobs on net in May.
What makes this indicator worthy of its vaunted position in the employment report? One answer is that because the diffusion index reflects the health of job gains across 250 private-sector industries, it’s a useful summary of the scope of employment activity across the many industries that comprise the U.S. labor market. A DI of 50% means that half of those industries are adding employment while half are shedding jobs. Over a series that starts in 1991, the average DI is 58%. Since President Biden has been in office, job growth has been strong and wide-spread across industries, with the DI averaging 65%.
For this month’s job-day post, CEA took our own deep dive into some homemade DIs (see our data appendix for a description of how we’re crunching these numbers). First, however, we use the BLS data to examine how widespread job gains have been across private industries in recent years. Following the post-pandemic bounce-back, the DI was elevated around 70% and higher as the historically strong jobs expansion took hold. Since then, as with many other normalizing variables at this stage of the expansion, it has drifted back down to the longer-term average level of around 60%.
The next figure uses our own DI analysis, separating out 24 industries in the private education and health care sectors from the rest of the private industries, as they’ve been notable contributors to job gains in recent months, including 86,000 in May.[1] Here we see that while the education and health care DI has been elevated relative to the rest, that’s been the case for decades, and even as the other series (private industries excluding education and health) have drifted down, both DIs are in healthy ranges around their long-term averages and consistent with strong job growth. It would thus be mistaken to conclude that just a few sectors are expanding while most others are not.
We next contrast DIs among the top 25% cyclically-sensitive industries, the bottom 25%, and the middle 50% (see appendix for how we break out these categories) based on their historical relationships with the change in the aggregate unemployment rate. Note that in the prior two downturns covered in the figure, as we’d expect, the less-cyclically sensitive DI doesn’t fall as much as the others. In the pandemic downturn, however, the shock hit job growth hard in all industries.
The figure shows that less-cyclical industries are posting high DIs, with around 80% adding net jobs in recent months. The middle-sensitivity group has come down from its peak and is around its average since 2000. The most cyclically-sensitive series has come down the most, which is consistent with the cooling, though still very strong, job market.
In sum, these DIs show robust job growth across the majority of private-sector industries. Over 60% of private industries contributed to May’s above-expectations payroll number, and while education and health have elevated DIs, the other industries are consistently chipping in. Cyclical DIs are showing some cooling effects, as we’d expect at this stage of the expansion, but they too are around their 2019 average.
All of this is consistent with a steady, stable economic expansion, where the strong job market continues to provide robust opportunities for working Americans. As inflation has slowed, wage gains have outpaced price gains, boosting workers’ buying power. Of course, as President Biden has stressed, too many families are still struggling with high prices, and we will continue to press forward with our cost-cutting agenda. But it is good to be able to do so with the strong tailwind of this labor market at our backs.
Appendix
The diffusion index represents the percentage of distinct industries for which (seasonally-adjusted) payroll employment is growing. Across distinct private sector industries, indexed by
, CEA calculates the diffusion index as:
To construct its set of industries, CEA follows the procedure outlined in the CES technical documentation and matches their list of NAICS industry group data available in the payroll survey. In general, the data for three-digit NAICS subsectors are not disaggregated to the four-digit level, and so in these cases the three-digit subsector data are used.
High, middle, and low-sensitivity industries are classified based on the rankings of estimated in regressions of industry-level 6-month employment change on the 6-month change in the aggregate unemployment rate:
We estimate these regressions for each industry from 2000 to 2019. A higher (absolute) value of this coefficient implies a greater sensitivity of industry employment to overall labor market changes. Industries are ranked based on the absolute value of their estimated coefficients. The top 25% of industries in this ranking are classified as high-sensitivity, the bottom 25% of industries are calculated as low-sensitivity, and all others are classified as middle-sensitivity. The diffusion indexes for each group are then calculate in the same ways for each subset of industries.
[1] Note that while the BLS figure uses a one-month span of industry job changes, we use a 6-month span to smooth out the series; also, due to data availability constraints on sub-sectors, these data only go through April.
