America’s data credibility crisis

Trust the data?

The monthly jobs report is supposed to be the market’s most trusted compass—dropping before the CPI, setting the tone for bonds, stocks, and rate bets, and often triggering the fiercest bouts of Friday volatility. For decades, traders took its integrity as a given. But that foundation is wobbling. Between payroll revisions, thinning survey responses, cracks in the CPI, and the dismantling of long-standing statistical guardrails, markets are beginning to question the reliability of the very data they price off.

The Employment Situation rests on two surveys. The Current Employment Statistics (CES), or establishment survey, canvasses some 631,000 businesses each month, producing granular snapshots of payrolls, hours worked, and earnings. It’s the workhorse that anchors the headline jobs figure. Running alongside it is the Current Population Survey (CPS), or household survey, which measures the labor force directly by demographic group and produces the iconic unemployment rate. One counts jobs; the other measures workers. When they diverge, markets bristle; when they align, the message roars.

Payrolls, however, are not set in stone. The first release leans on partial survey responses—about two-thirds of the sample—before creeping above 90% after three months. Revisions are built into the system, and annual benchmark updates against the Quarterly Census of Employment and Wages (QCEW) often redraw the prior year’s landscape. Traders understand that the “first print” is a draft, not a verdict.

But the recent run of downward revisions has tested that tolerance. May’s payrolls were revised down by 125k to +19k, June by 133k to +14k, and July printed at just +73k—marking the weakest three-month stretch since the early pandemic, and before that, the Great Recession. Statistically, a 133k revision isn’t beyond the pale—given a standard error of about 83k—but the persistence of downward moves tells a story of bias, not noise.

Two forces are at play. First, late survey responses. Businesses swamped by tariffs and other real-world headaches have been slower to file voluntary CES forms, leaving early estimates on thinner ground. As more responses trickle in, the picture skews weaker, forcing revisions.

Second, the birth/death model. Since the CES can’t capture brand-new firms in real time, the BLS uses a net business birth/death adjustment to account for job gains from new enterprises minus losses from closures. Lately, this model has been crediting the economy with more jobs than the QCEW trend supports. That means early prints lean optimistic—only to be corrected lower when harder census data catch up.

The credibility debate doesn’t stop at payroll. The CPI has its own cracks. A federal hiring freeze in January 2025 thinned the ranks of BLS price collectors. Attrition deepened the hole, and with fewer boots on the ground, the agency scaled back collection in some regions and stopped altogether in others. The gaps were filled with imputation—statistical guesswork in place of observed prices.

Normally, about 90% of imputations come from local “home cell” averages, 10% from broader “different cell” regions, and almost none from “carry-forward” (last month’s price). But since March, the share of “different cell” imputations surged—15% in March, 29% in April, peaking at 35% in June, and still elevated at 32% in July. The BLS insists the impact on inflation is minimal. Markets remain skeptical, especially after consensus undershot the core CPI print in four of the last five months, missing by as much as 0.2 percentage points in March and May.

Worries aren’t confined to trading desks. A Reuters survey of 100 policy experts—including Nobel laureates, ex-officials, and leading economists—showed 41% “very concerned” about U.S. data quality and another 48% at least “slightly concerned.” Nearly three-quarters said authorities weren’t treating the issue with urgency, and two-thirds said statistical agencies lack the resources to keep pace. With Social Security and other key income benchmarks indexed to CPI, these aren’t just market squabbles—they’re questions of household livelihoods.

The unease only grew when the Federal Economic Statistics Advisory Committee (FESAC) was dissolved in February. For years, it had provided expert oversight to the BEA, Census, and BLS on methodology and statistical integrity. Its removal stripped away a vital guardrail. Even Powell flagged the shift in congressional testimony, acknowledging “a very mild degradation” in survey coverage and warning that the direction of travel was concerning. Data response rates, both household and establishment, have been falling steadily since 2015.

The path back to credibility is clear, if not easy. Ending the hiring freeze and restoring budgets would be the fastest win—the BLS’ funding has stagnated in real terms for over two decades even as the economy it measures has expanded dramatically. Making CES responses mandatory, as some states already do, would improve coverage and reduce revision whiplash.

Technology can help too. The BLS already taps one confidential firm for massive apparel and household goods price feeds. But two-thirds of CPI’s ~100,000 monthly price quotes are still gathered by in-person visits. Expanding alternative data—corporate datasets, secondary sources, web scraping, and APIs—would lighten the load and modernize coverage. The strain of 2025 may have only accelerated those efforts.

Despite the noise, the BLS still produces the gold standard of market-moving macro data. But credibility is not self-sustaining. A world that trades trillions on these numbers cannot afford for them to drift into question. Restoring resources, modernizing methods, and tightening mandates isn’t bureaucracy—it’s the cost of keeping the economic compass pointing true. Markets can handle bad news; what they can’t handle is bad data.

Trader’s coda: The desk takeaway is simple—first prints now carry a wider revision premium. Markets are quicker to fade, positioning leans more tentative, and trust is migrating to alternative sources and private gauges. Until Washington shores up the scaffolding, every jobs Friday and CPI Wednesday is trading not just the data, but the data’s credibility.