The latest employment report released by the U.S. Department of Labor’s Bureau of Labor Statistics (BLS) shows that nonfarm payrolls in July increased by 73,000, far below expectations. June's figure was revised down to an increase of just 14,000, compared to the previously reported 147,000,a nearly 90% downward revision. After the report's release, Donald Trump posted on Truth Social, claiming that the July employment data from the BLS had been manipulated and ordered the dismissal of BLS Commissioner Erika McEntarfer. Some media outlets mocked Trump for “shooting the messenger” due to “unpleasant numbers”, but it must be said that the BLS’s increasingly frequent downward revisions in recent years have gradually eroded public confidence in key U.S. economic data, to the point that even Trump could not stay silent.

The BLS’s recent trend of frequent and significant downward data revisions is hard to ignore. In March 2023, the BLS revised nonfarm employment figures for the period from April 2022 to March 2023, revealing that job growth had been overestimated by 818,000, the largest annual downward revision since 2009. After the adjustment, the average monthly job growth dropped from 242,000 to 173,000, a decline of about 28%. From March 2023 to March 2024, the BLS revised its monthly nonfarm payroll data downward in 9 out of 12 months, with an average monthly downward revision of approximately 42,000 jobs. In August 2025, the BLS released the July 2025 employment report, which showed that job growth in May had been revised down from an initial estimate of 144,000 to just 19,000, a downward revision of 125,000. June’s figure was also revised down from 147,000 to 14,000, a cut of 133,000. Combined, the two months saw a total downward revision of 258,000 jobs, the largest non-pandemic-related adjustment since 1979. These steep revisions triggered a sharp market reaction, with the Dow Jones Industrial Average falling more than 600 points, as investor concerns over a potential recession and the reliability of economic data intensified. Michael Green, a portfolio manager at Simplify Asset Management, sees that there are clearly flaws in the U.S. government's modeling, and without reliable data, it would be impossible to formulate sound policy.

Due to factors such as data lag, sample bias, and the complexity of the modern economy, the measurement errors in the key economic indicators the Federal Reserve relies on for decision-making have been steadily increasing, to the point where they can no longer be ignored. These economic data have become increasingly disconnected from economic reality, potentially distorting both the Fed’s policy decisions and the market’s perception of the U.S. economy.

Take employment data, one of the core indicators of the economy, as an example. The BLS relies on two major survey systems to capture employment trends: the Current Population Survey (CPS) and the Current Employment Statistics (CES). These two surveys differ in methodology and scope, and each has its limitations, which may result in a distorted depiction of employment conditions.

The CES counts the total number of nonfarm 'jobs' based on payroll reports submitted by businesses. Its unit of measurement is the number of positions, not individuals, so it cannot distinguish when one person holds multiple jobs. If a worker has two jobs, they are counted as two separate positions in the CES, which can lead to an overestimation of employment growth. Amid rising living costs, the gig economy and part-time work have become increasingly common in the U.S. According to CPS data, around 8.431 million Americans held multiple jobs in 2024, accounting for about 5.2% of total employment. This makes double-counting more likely and may help explain why the BLS frequently revises nonfarm employment figures downward.

The CPS surveys approximately 60,000 U.S. households each month to calculate the official U-3 unemployment rate. Under the strict definition of U-3, the unemployed are those who have not worked in the past four weeks but have actively looked for work and are available to start immediately. This narrow definition presents several issues. First, the official unemployment rate from the CPS does not include “discouraged workers”, i.e., those who have stopped job hunting because they believe no jobs are available, or the underemployed, such as people working part-time involuntarily. Since the COVID-19 pandemic, the U.S. labor force participation rate (LFPR) has declined, meaning that some working-age individuals have simply exited the labor market altogether and are no longer counted as “unemployed”, which in turn lowers the U-3 rate. In 2023, the U-3 unemployment rate stood at just about 3.6%, while the broader U-6 rate, which includes discouraged workers and involuntary part-timers, was as high as around 6.7%. In other words, a significant portion of idle labor is not captured in the mainstream unemployment data, causing the official figures to underestimate the true weakness of the job market. Additionally, in recent years, the CPS survey response rate has declined to around 70%, down from the pre-pandemic level of 83%-87%. The drop in response rates has been particularly pronounced among low-income and less-educated groups, potentially leading to an overestimation of employment conditions in the official data.

Beyond employment, the CPI also faces significant challenges. The BLS calculates the CPI each month using approximately 90,000 price samples, covering more than 200 categories of goods and services. This process involves hundreds of data collectors operating across 75 urban areas. When price data is not directly available, the BLS typically estimates about 10% of the CPI data points. As of July 2025, this estimation ratio has risen to 30%. In other words, over 30% of CPI price data is not based on actual collection but is instead filled in through model-based estimation, which may increase the measurement error in CPI data.

The rising proportion of estimated CPI data is mainly influenced by several factors.

First, similar to employment data, the response rate for traditional BLS price surveys has dropped significantly. The BLS Housing Survey, which collects rent information and estimates Owner’s Equivalent Rent (OER) to calculate housing costs, had a response rate of 70% in 2014. By 2022, that had fallen to 52%. The Current Expenditure Survey (CE), which gathers household spending data to assign weights to observed price changes and calculate inflation, saw its response rate decline to 43% by 2023. The Medical Expenditure Panel Survey (MEPS), which collects data from households and healthcare providers on individuals and their medical services, had already seen its response rate fall to 46% even before the pandemic. The American Community Survey (ACS), which collects a broad range of social, economic, housing, and demographic data at the community level each year, had a response rate as high as 97% a decade ago, but this had dropped to 71% before the pandemic.

Second, the rise of the online economy and regional differences have made price changes more complex, making it difficult for the traditional fixed consumption basket to fully capture real price movements. On one hand, online product prices fluctuate frequently, and traditional monthly price collection tends to lag behind. On the other hand, differences in consumption patterns and price trends across regions mean that national average weights often mask significant price disparities.

Finally, there are inherent limitations in how the CPI is constructed. The BLS typically determines the items and their weights in the CPI "basket" based on consumer expenditure survey data from the previous two years, updating the basket every two years, usually in January of even-numbered years. This means the CPI reflects consumption patterns that lag by 2–3 years. For example, the 2024 CPI weights are based on data from 2021–2022, which may fail to capture consumption shifts in 2023–2024 caused by post-pandemic economic recovery, e-commerce growth, or inflationary pressures. In 2025, CPI data may rely more heavily on historical data or proxy estimates for some goods due to budget cuts reducing sample collection, further exacerbating the lag in basket updates. Moreover, about one-third of the CPI, i.e., housing costs, is calculated using Owner’s Equivalent Rent (OER), which has a built-in time lag due to its six-month rolling sample design, making it unable to promptly reflect surges in rents from new leases. CPI data also excludes asset prices, thus overlooking the impact of asset bubbles on household wealth and consumption behavior, which can lead to an underestimation of actual inflation pressures.

These statistical inaccuracies have led to a serious disconnect between the data and the public’s "felt experience" of the economy. Although the annualized CPI increase reached 9.1% in 2022, most consumers felt that their cost of living had risen far more than that. Rent, utilities, and food expenses continued to climb, while official data showed inflation had "peaked and declined". On social media, discussions such as "untrustworthy statistics" and "prosperity on paper, hardship in reality" have become increasingly common, reflecting a rapid erosion of public trust in government economic statistics.

A similar situation has also emerged in the employment sector. Although the unemployment rate remains around 4%, the public is facing large-scale layoffs and stagnating real wages. Layoffs in the tech industry and among federal employees, combined with wage growth that fails to keep up with inflation, have made the notion of "full employment" seem hollow and ineffective.

The disconnect between data and the public's economic experience has led to a crisis of trust in core U.S. economic data. However, the more serious issue is that statistical errors not only affect public perception but can also cause a divergence between the data and actual economic conditions, potentially influencing the Federal Reserve's decision-making. After all, it is the Fed that decides the U.S. economic policies, and this is anchored to two key objectives: employment and inflation.

Looking back at 2021, the Fed insisted that the rise in inflation was only "temporary" when the CPI showed inflation to be relatively mild, which delayed interest rate hikes. However, the reality was that by that time, the economy had already harbored hotter inflationary factors, such as asset bubbles and soaring rents, which were not fully reflected in the CPI. The CPI "misreporting" ultimately led to a delayed response from the Fed, culminating in the uncontrollable inflation crisis of 2022. In early 2023, the CES data showed strong employment growth, and the Fed maintained its tightening stance based on this. However, subsequent data revealed that the actual number of new jobs was much lower than initially reported, and the policy adjustment was significantly delayed. By the end of 2024, after lowering the target range for the federal funds rate from 4.5%-4.75% to 4.25%-4.5%, the Fed took no further action. Some analysts believe that Fed chair Jerome Powell may repeat the same policy delay mistakes made in 2021.

The current market and the Fed are seriously divided on whether interest rates should be lowered. The market believes that the weak job market and economic slowdown require faster monetary easing, while the Fed is concerned about "sticky inflation" and has chosen to delay rate cuts in the face of seemingly strong employment data. At the July 30 press conference, Powell stated that the labor market remains strong, and the Fed is still in the early stages of understanding how Trump’s tariff reforms and other policy changes will impact inflation, employment, and economic growth. The market had originally expected the Fed to cut rates in July or September, but the FOMC meeting on July 30 kept rates unchanged, with Powell reaffirming the “wait for data” stance, reducing the probability of a rate cut in September from 80% to 40%. Even Trump has repeatedly criticized Powell for being too slow to lower rates, calling him “Too Late Powell” and even threatening to replace him.

The issue of distorted economic data in the U.S. has reached an unavoidable level. The core economic statistical system in the country is no longer adequate for monitoring the modern economy. Data distortion not only weakens public trust in the government, but also causes the Fed to veer off course in the fast-moving economic landscape. It is important to note that the Fed is a large institution with abundant experience, so data-related issues are actually correctable. The real problem lies in its stance, which heavily leans towards the Democratic Party and opposes Trump. In other words, the Fed is using the so-called "data" to counter Trump’s policies, acting as the leader and flag-bearer for Wall Street and left-wing media, which has led to the current situation. However, if the data does not accurately reflect the economic reality, the Fed’s so-called forward guidance will lose credibility. In the face of the modern economy’s volatility and complexity, overly relying on a statistical system designed in the last century undoubtedly comes at a high cost.

Final analysis conclusion:

The measurement error issue of core economic indicators in the U.S., such as the CPI and employment data, has been intensifying recently. In addition to flaws in data compilation, the Federal Reserve's alignment with the Democratic Party has led to the use of so-called "statistics" to counter Trump-era policies, exacerbating the deviation in core data. However, from the public's perspective, the "distortion" of data is continuously eroding the government's credibility, potentially triggering a crisis of trust. From a policy standpoint, inaccuracies in economic data could cause the Fed to veer off course while navigating the fast-changing economic situation, potentially having irreversible consequences for the U.S. economy.

______________

Chen Li is an Economic Research Fellow at ANBOUND, an independent think tank.