After an avalanche of data last week, there are more signs of a slowing economy. Real GDP rose modestly in the first half, and inflation started to drift up. The labor market was the last piece to fall in line, and it did with a bang on Friday. Job growth slowed sharply starting in May (including large downward revisions in May and June), and the unemployment rate increased in July. It’s a complicated mix of supply and demand shocks, but an unsurprising outcome given the significant policy changes this year, including higher tariffs, less immigration, and downsizing the government. Changes in labor supply, as well as tension between the employment and inflation mandates, create challenges for the Fed.
Job growth has slowed (since the spring).
The main takeaway from the employment report is that job growth has slowed considerably, and this slowdown began earlier than previously estimated. Before last Friday, job growth appeared to have held up well in May and June (the blue line), but with more data, those job gains were revised down sharply, and the increase in payrolls in July was modest (green line). The three-month average had been 150,000, and is now 35,000. That was a significant shift in the labor market data, but it aligns with other signs of economic slowdown.
The primary task is to interpret this slower path for job growth. More on that below. However, the large revisions in May and June have raised several questions, and President Trump used them as a pretext to fire the Commissioner of the BLS.
This is a policy problem, not a measurement problem.
The problem is not the Bureau of Labor Statistics. Large, unpredictable shifts in economic policy are placing unusual strains on our measurement apparatus because they are causing large, unpredictable changes in the behavior of consumers and businesses. These changes are difficult to measure in real time. The GDP statistics this year have struggled to isolate massive swings in imported goods around the start of tariffs from its measure of domestic production. The initial estimates of payrolls didn’t capture the slowdown in employment, but that’s more a reflection of how sharp the jobs slowdown is, rather than a limitation of the surveys. Increasing staffing and budgets at statistical agencies would be a wise investment, according to the revisions.
The revisions were large, but not mysterious.
The two-month downward revision for May and June, at -258,000, was the largest since the pandemic. Although unusually large, these revisions followed the standard procedures: the BLS publishes its first estimate for a month on the first Friday of the next month. It revises that month’s estimate in the next two reports. There are two sources for these revisions: 1) data collections from businesses and government agencies that did not meet the deadline for the first estimate, and 2) updated seasonal factors based on the new data. Before 2003, seasonal factors remained unchanged at the monthly revisions. (There are later revisions at the annual benchmarks.)
Separating the latest revisions into the two sources, the June downward revision is primarily due to the collection of new data, while the May downward revision is primarily due to new seasonal factors. Note, the seasonal factors changed due to the additional collections for May and June, as well as the first collection for July.
The June revision is the primary driver of the revisions. (It also influenced the revised seasonals in May.) The downward revision to payroll growth (as a percent change) in June is notable, but it is not historic. Even larger revisions from the first to the second estimate were common before 1990. On average, the revisions have gotten smaller over time. Larger revisions now tend to occur in times of transition in the labor market, such as recessions or early recoveries.
... continue reading