More than Missing Data: Survey Response Rates Following the 2025 Government Shutdown
April 09, 2026
The 2025 federal government shutdown had an immediate and well-documented impact on labor market statistics. Most visibly, there were delays in the release of the jobs report, along with other statistics, and the October release of the household survey (the Current Population Survey, or CPS) was canceled altogether. Here at the Atlanta Fed, those delays prevented us from publishing our data tools, such as the Wage Growth Tracker and the Jobs Calculator, as the data are derived from the CPS. Less visible, and the focus of this analysis, is a shutdown's possible lingering effect on CPS response rates. Following the shutdown, newly enrolled households have been responding at persistently lower rates than pre-shutdown trends would have indicated. If the lower response continues, the confidence in monthly national employment statistics will diminish.
Impact of the fall 2025 government shutdown on labor market statistics
The CPS, which is conducted by the US Census Bureau for the US Bureau of Labor Statistics (BLS), relies on continuous monthly interviewing under a 4-8-4 rotating panel design. A household is interviewed for four months in a row, then is out of the sample for eight months, and then is interviewed again for another four consecutive months. The first round of interviews is denoted as month-in-sample (MIS) 1–MIS4. The second round is denoted as MIS5–MIS8. This framework allows for the calculation of year-over-year changes in behavior for individuals. The aforementioned Wage Growth Tracker is a prime example of this: it calculates the wage growth from MIS4 to MIS8 a year later.
When the government shutdown suspended BLS operations in fall 2025, the October survey was not completed, resulting in a loss of all CPS data for that month. In addition, the shutdown caused processing delays in the collected September and November data, leaving policymakers (and anyone else interested) without timely information to gauge the health of the labor market. Furthermore, any labor force measures that are calculated on a year-over-year basis will be missing in both October 2025 and October 2026, as the base year for the 2026 measure was not collected. The missing data also affected the calculation of sample weights, as detailed by the BLS, which affects the precision of national statistics, although this impact should diminish over the next few months. Finally, beyond the data loss and its impact on Census methodology, the willingness of households to respond to the survey appears to have dipped.
CPS response rates have been declining since 2013
Declining survey response rates are not new for the CPS, nor is this the first time that external events have had an impact on response rates (see figure 1). The CPS had a 90 percent response rate in 2013. However, the response rate fell to 82 percent by February 2020, an average decline of 1.14 percentage points per year. Had that trend continued, the response rate would currently be over 75 percent, as shown by our estimated linear time trend—the blue dotted line in figure 1. Instead, in the first six months of the COVID-19 pandemic, the response rate plummeted to 70.2 percent. Although there was a partial recovery to 79 percent by September 2020, the response rate began to decline again after that point, falling at a significantly higher rate of over 2 percentage points per year (yellow dotted line) in the period between the pandemic and the shutdown, falling below 70 percent in 2024.
CPS response rate after the government shutdown
Immediately following the recent government shutdown period, the CPS response rate fell sharply, declining by almost 5 percentage points from the September to November surveys. A potential explanation for the initial low response rate could be the reduced outreach before entry into the CPS. Typically, according to the BLS, the Census Bureau sends letters to households before their initial survey and before the first survey after the eight-month break, which are both primarily done in person. According to contacts at the Census Bureau, no letters were sent before the November survey as a result of the government shutdown, so the November in-person outreach was the first contact for both the October and November 2025 cohorts. The Census then returned to normal procedures for the December survey. In spite of these efforts, the response rate remained below trend for the first three months after the shutdown. However, by February, the response rate was no longer statistically different from the pre-shutdown trend, although it was 3 percentage points lower than in September. Taken at face value, this trajectory suggests that any potential “shutdown effect” on response rates had largely dissipated by then.
An examination of individual MIS response rates complicates that conclusion. In each of the four CPS months following the shutdown, above-trend response rates for households that entered the CPS before the shutdown offset below-trend response rates among households that entered the CPS after the government shutdown (that is, the post-shutdown households). The net effect is an overall figure that appears back on trend—but for compositional reasons rather than a broad-based recovery.
This differential in response rates between pre- and post-shutdown cohorts is displayed in figure 2, which compares the actual response rate to the predicted trend by MIS for the four CPS months following the shutdown. Panel A covers households whose entry into the CPS occurred after the shutdown, while panel B covers households that entered the CPS prior to the shutdown. Within each panel, blue dots represent actual response rates for each MIS by month of the survey, while the green dots indicate the predicted response rates (along with the upper and lower bounds of a 95 percent prediction interval) based on the September 2020 through September 2025 trend. Readers can interpret a blue dot falling below its corresponding green dot as that MIS group underperforming relative to trend, while a blue dot exceeding its green dot indicates an above-trend response rate.
Of the 13 observable post-shutdown cohort response rates shown in panel A, all but one are below trend. Although not all the trend gaps are statistically significant, the consistency of the negative deviation from trend across post-shutdown cohorts in panel A is notable. In contrast, pre-shutdown households display response rates that are generally above or at trend, as panel B shows, with less than one-third of the pre-shutdown cohorts' response rates falling below trend. The remaining responses were all above trend, although most were not significantly so.
Historically, there has not been much of a gap, if any, between the two rounds of interviews for the CPS. There was no statistical difference in survey response rates between the first round (MIS1–MIS4) and the second round of interviews (MIS5–MIS8) for either the pre-covid period of January 2012–February 2020 or the post-covid/pre-shutdown period from September 2020–September 2025. However, since the shutdown, the second rotation survey response rates for the pre-shutdown cohorts are, on average, 1.9 percentage points higher than the response rates for the post-shutdown cohorts in their first rotation. These significantly higher response rates among pre-shutdown households in their second rotation have propped up the overall response rate, masking the persistent underperformance among post-shutdown cohorts.
This response rate gap presents a potential concern moving forward. If this dynamic persists, the compositional offset will naturally unwind over time. The pre-shutdown households with higher response rates will gradually exit the sample by January 2027. If below-trend response rates among post-shutdown households persist, the overall trend gap could worsen progressively throughout that period. For example, the overall response rate in February 2026 was 0.61 percentage points above trend. Had the response rates of the pre-shutdown households simply remained at trend rather than overperforming, the trend gap would have been 0.36 percentage points below trend, roughly a 1 percentage point swing.
Implications of the low CPS response rate
If lower post-shutdown response rates among households that are new to the survey persist, the decline in the overall CPS responses rate will accelerate, diminishing confidence in month-over-month changes in the unemployment rate and other labor force statistics from the household survey. McIllece (2023) suggests that at the current rate of response, a two-month average unemployment rate is necessary to have the same precision as a one-month estimate when the response rate was over 90 percent. The lower response rates also increase concerns about non-response bias. Cai and Baker (2021) found respondents who indicated they were unemployed or not in the labor force were more likely to be missing in future observations, suggesting that the unemployment rate could be biased downward. These effects become amplified every time the response rate declines further. Only time will tell if the government shutdown in fall 2025 has fundamentally changed the willingness of newly enrolled households to participate in the CPS. As a result, anyone who relies on national statistics, such as the unemployment rate, for policy analysis must monitor these response rates and keep in mind the impact of their decline on the reliability of the statistics derived from them.