How reliable is the jobs data? Economists and Wall Street still trust it

San Francisco Chronicle​

18 hours ago

WASHINGTON (AP) — The monthly jobs report is already closely-watched on Wall Street and in Washington but has taken on a new importance after President Donald Trump on Friday fired the official who oversees it.
Trump claimed that June's employment figures were 'RIGGED' to make him and other Republicans 'look bad.' Yet he provided no evidence and even the official Trump had appointed in his first term to oversee the report, William Beach, condemned the firing of Erika McEntarfer, the director of the Bureau of Labor Statistics appointed by former President Joe Biden. The firing followed Friday's jobs report that showed hiring was weak in July and had come to nearly a standstill in May and June, right after Trump rolled out sweeping tariffs.
Economists and Wall Street investors have long considered the job figures reliable, with share prices and bond yields often reacting sharply when they are released. Yet Friday's revisions were unusually large — the largest, outside of a recession, in five decades. And the surveys used to compile the report are facing challenges from declining response rates, particularly since COVID, as fewer companies complete the surveys.
Nonetheless, that hasn't led most economists to doubt them.
'The bottom line for me is, I wouldn't take the low collection rate as any evidence that the numbers are less reliable,' Omair Sharif, founder and chief economist at Inflation Insights, a consulting firm, said.
Many academics, statisticians and economists have warned for some time that declining budgets were straining the government's ability to gather economic data. There were several government commissions studying ways to improve things like survey response rates, but the Trump administration disbanded them earlier this year.
Heather Boushey, a top economic adviser in the Biden White House, noted that without Trump's firing of McEntarfer, there would be more focus on last week's data, which points to a slowing economy.
'We're having this conversation about made-up issues to distract us from what the data is showing," Boushey said. 'Revisions of this magnitude in a negative direction may indicate bad things to come for the labor market.'
Economists and Wall Street trust the data
Most economists say that the Bureau of Labor Statistics is a nonpolitical agency staffed by people obsessed with getting the numbers right. The only political appointee is the commissioner, who doesn't see the data until it's finalized, two days before it is issued to the public.
Erica Groshen, the BLS commissioner from 2013 to 2017, said she suggested different language in the report to "liven it up", but was shot down. She was told that if asked to describe a cup as half-empty or half-full, BLS says 'it is an eight ounce cup with four ounces of liquid.'
The revised jobs data that has attracted Trump's ire is actually more in line with other figures than before the revision. For example, payroll processor ADP uses data from its millions of clients to calculate its own jobs report, and it showed a sharp hiring slowdown in May and June that is closer to the revised BLS data.
These are the figures is Trump attacking
Trump has focused on the revisions to the May and June data, which on Friday were revised lower, with job gains in May reduced to 19,000 from 144,000, and for June to just 14,000 from 147,000. Every month's jobs data is revised in the following two months.
Trump also repeated a largely inaccurate attack from the campaign about an annual revision last August, which reduced total employment in the United States by 818,000, or about 0.5%. The government also revises employment figures every year.
Trump charged the annual revision was released before the 2024 presidential election to 'boost' Vice President Kamala Harris's "chances of Victory," yet it was two months before the election and widely reported at the time that the revision lowered hiring during the Biden-Harris administration and pointed to a weaker economy.
Here's why the government revises the data
The monthly revisions occur because many companies that respond to the government's surveys send their data in late, or correct the figures they've already submitted. The proportion of companies sending in their data later has risen in the past decade.
Every year, the BLS does an additional revision based on actual job counts that are derived from state unemployment insurance records. Those figures cover 95% of U.S. businesses and aren't derived from a survey but are not available in real time.
These are the factors that cause revisions
Figuring out how many new jobs have been added or lost each month is more complicated than it may sound. For example, if one person takes a second job, should you focus on the number of jobs, which has increased, or the number of employed people, which hasn't? (The government measures both: The unemployment rate is based on how many people either have or don't have jobs, while the number of jobs added or lost is counted separately).
Each month, the government surveys about 121,000 businesses and government agencies at over 630,000 locations — including multiple locations for the same business — covering about one-third of all workers.
Still, the government also has to make estimates: What if a company goes out of business? It likely won't fill out any forms showing the jobs lost. And what about new businesses? They can take a while to get on the government's radar.
The BLS seeks to capture these trends by estimating their impact on employment. Those estimates can be wrong, of course, until they are fixed by the annual revisions.
The revisions are often larger around turning points in the economy. For example, when the economy is growing, there may be more startups than the government expects, so revisions will be higher. If the economy is slowing or slipping into a recession, the revisions may be larger on the downside.
Here's why the May and June revisions may have been so large
Ernie Tedeschi, an economic adviser to the Biden administration, points to the current dynamics of the labor market: Both hiring and firing have sharply declined, and fewer Americans are quitting their jobs to take other work. As a result, most of the job gains or losses each month are probably occurring at new companies, or those going out of business.
And those are the ones the government uses models to estimate, which can make them more volatile.
Groshen also points out that since the pandemic there has been a surge of new start-up companies, after many Americans lost their jobs or sought more independence. Yet they may not have created as many jobs as startups did pre-COVID, which throws off the government's models.
Revisions seem to be getting bigger
The revisions to May and June's job totals, which reduced hiring by a total of 258,000, were the largest — outside recessions — since 1967, according to economists at Goldman Sachs.
Kevin Hassett, Trump's top economic adviser, went on NBC's 'Meet the Press' on Sunday and said, 'What we've seen over the last few years is massive revisions to the jobs numbers.'
Hassett blamed a sharp drop in response rates to the government's surveys during and after the pandemic: 'When COVID happened, because response rates went down a lot, then revision rates skyrocketed.'
Yet calculations by Tedeschi show that while revisions spiked after the pandemic, they have since declined and are much smaller than in the 1960s and 1970s.
Other concerns about the government's data
Many economists and statisticians have sounded the alarm about things like declining response rates for years. A decade ago, about 60% of companies surveyed by BLS responded. Now, only about 40% do.
The decline has been an international phenomenon, particularly since COVID. The United Kingdom has even suspended publication of an official unemployment rate because of falling responses.
And earlier this year the BLS said that it was cutting back on its collection of inflation data because of the Trump administration's hiring freeze, raising concerns about the robustness of price data just as economists are trying to gauge the impact of tariffs on inflation.
U.S. government statistical agencies have seen an inflation-adjusted 16% drop in funding since 2009, according to a July report from the American Statistical Association.
'We are at an inflection point,' the report said. 'To meet current and future challenges requires thoughtful, well-planned investment ... In contrast, what we have observed is uncoordinated and unplanned reductions with no visible plan for the future.