Understanding Statistical Significance in Monthly Unemployment Data

Every month, the Bureau of Labor Statistics (BLS) releases the unemployment rate, one of the most closely watched economic indicators. Financial markets, policymakers and the media often focus intently on whether the rate moved up or down by even a tenth of a percentage point, and whether this change was in line with market expectations and expert forecasts. Markets can move and policy narratives can shift based on these small monthly changes.

However, an important statistical consideration receives very little attention: the role of sampling variation in monthly unemployment data. In this blog post, we highlight that individual monthly changes, even of 0.2 percentage points, are often too noisy to determine whether they reflect true fluctuations or sampling variation, while sustained movements over three months of at least 0.3 percentage points provide much stronger evidence of genuine labor market trends.

The Statistical Framework

The unemployment rate comes from the Current Population Survey (CPS). Each month, the BLS attempts to interview approximately 60,000 eligible households, though because of nonresponses, the actual number of interviews completed is lower. While a substantial sample, it is still only a small fraction of the U.S. population. Even if the true, population-wide unemployment rate stayed constant, we could expect month-to-month variation simply because the survey samples different households each month.

This monthly variation is mitigated by the CPS’ rotating panel design, in which households are interviewed for up to four consecutive months. Sampling error, however, is inherent to working with subsets rather than complete censuses. The BLS calculates standard errors to help assess the likelihood that observed changes reflect real economic shifts rather than sampling variation.

The FRED chart below shows the unemployment rate from January 2022 through January 2026.October 2025 data are unavailable because of the federal government shutdown, creating gaps in the monthly series and the change series that follow. Unemployment declined from a COVID-19 pandemic peak to historically low levels by April 2023. Following a period of stability through July 2023, unemployment began a clear upward shift that continued through 2024, with another notable increase in mid- to late 2025. These shifts in level represent genuine changes in labor market conditions, yet we also see meaningful month-to-month variation even during relatively stable periods. So, how can we distinguish true shifts from normal month-to-month volatility?

Understanding the Statistical Significance of Labor Data

The BLS publishes statistical summary tables along with each jobs report, identifying whether changes in labor force indicators are statistically significant at the 90% confidence level.These data are in the statistical significance summary table of the jobs report’s supplemental files. The next figure plots month-to-month changes in the unemployment rate alongside the minimum change required for statistical significance. When changes are outside the shaded area, we can be 90% confident that they reflect true changes in the labor market rather than sampling variation.

Statistical Significance of Monthly Changes in the Unemployment Rate

SOURCE: Bureau of Labor Statistics.

As the figure demonstrates, the vast majority of monthly changes from 2022 through 2025 remain within the shaded area, even during periods when unemployment was rising or falling (recall the first figure). Even a 0.2 percentage point change often does not reach statistical significance, meaning it is hard to tell whether monthly movements reflect sampling variation or true labor market changes.

Careful observers may notice that the significance bands in the above figure appear narrower at the beginning of the period than at the end, suggesting that measurement precision declined somewhat over time. We will return to this issue after examining how changes over multiple months provide clearer signals.

Sustained Trends Provide Clearer Signals

While individual monthly changes often fall within sampling variation, cumulative changes over multiple months provide more-reliable signals. When the unemployment rate moves in the same direction over three consecutive months by at least 0.3 percentage points and thus exceeds the standard error bands, we can be more confident we are observing genuine economic shifts.

The third figure plots three-month changes alongside the minimum change needed for statistical significance.The absence of October 2025 data because of the federal government shutdown means the three-month change for January 2026 cannot be calculated. This time frame reflects the reality that identifying genuine trends requires observing sustained movements over multiple months that are hard to infer from monthly changes, given the sampling error.

Statistical Significance of 3-Month Changes in the Unemployment Rate

SOURCES: Bureau of Labor Statistics and authors’ calculations.

The data reveal four distinct episodes in which three-month changes fell outside the shaded area, indicating statistically significant changes. Early 2022 shows a significant decline as the labor market continued its recovery from pandemic disruptions. Following a period of relative stability through mid-2023, three separate episodes of significant increases occurred: October 2023, June-July 2024 and September 2025. Each of these instances was interrupted by periods in which the unemployment rate fluctuated without sustained directional movement. In each case, these significant three-month changes corresponded to clear level shifts observable in the first figure, precisely the pattern to look for when distinguishing genuine labor market developments from sampling variation.

The Role of Declining Response Rates

Returning to our previous observation, response rates for the CPS have declined steadily over time, raising questions about whether this affects measurement precision.The BLS has a dedicated webpage on the CPS response rate. To isolate the impact of response rates, we can compare months with nearly identical unemployment rates and labor force participation rates but different response rates. Matching the first two rates is important for isolating the effect of response rates.The unemployment rate itself affects measurement variance; higher unemployment rates have higher variance, all else equal. The labor force participation rate determines how many people in the sample are actually in the labor force, which is the denominator used to calculate the unemployment rate. A higher participation rate means a larger effective sample size for measuring unemployment, and a larger sample size reduces the standard error of the estimate.

Consider February 2022 and June 2025, which had very similar rates (unemployment of 3.9% and 4.1%, respectively, and a labor force participation rate of 62.3% for both months). When the response rate declined 6.3 percentage points, from 73.7% in February 2022 to 67.4% in June 2025, the significance threshold rose from 0.18 percentage points to 0.21 percentage points. This reveals that declining response rates do modestly reduce measurement precision. However, the effect is relatively contained, and the CPS remains a reliable survey instrument despite these declines.

The Importance of Recognizing Sampling Variation

Understanding these statistical properties has important implications for interpreting economic developments. The unemployment rate remains an essential indicator of labor market health, and the BLS statistical summary tables help determine which changes can be confidently attributed to genuine economic shifts rather than sampling variation.

When evaluating monthly unemployment data, it is important to recognize that single-month changes of 0.1 percentage points and even of 0.2 percentage points often fall within normal sampling variation and are thus difficult to interpret, while three-month cumulative changes provide stronger evidence of genuine trends. Sustained movements over three months of at least 0.3 percentage points exceed the significance bands and thus provide much stronger evidence of true labor market shifts. While declining response rates to the CPS have modestly reduced measurement precision, the effect remains manageable and does not undermine this fundamental approach.

Importantly, the BLS has consistently provided comprehensive documentation of measurement uncertainty and published statistical significance tests with every release. Making full use of these resources provides a more complete assessment of labor market conditions than focusing on (changes in) the headline numbers alone.

Notes

1. October 2025 data are unavailable because of the federal government shutdown, creating gaps in the monthly series and the change series that follow.
2. These data are in the statistical significance summary table of the jobs report’s supplemental files.
3. The absence of October 2025 data because of the federal government shutdown means the three-month change for January 2026 cannot be calculated.
4. The BLS has a dedicated webpage on the CPS response rate.
5. The unemployment rate itself affects measurement variance; higher unemployment rates have higher variance, all else equal. The labor force participation rate determines how many people in the sample are actually in the labor force, which is the denominator used to calculate the unemployment rate. A higher participation rate means a larger effective sample size for measuring unemployment, and a larger sample size reduces the standard error of the estimate.