Statisticians are concerned that Section 605 in the House Fiscal Year 2026 Commerce, Justice, Science appropriations bill “would prohibit the Census Bureau from making more than two contact attempts for any survey. While this directive might appear to reduce costs, it would, in practice, undermine the nation’s statistical infrastructure, distort representation, and erode the precision of the data on which communities, businesses, and governments depend.”
The October 21, 2025 letter from the American Statistical Association to Congressional leaders goes deeper into the negative impact of this provision, on top of that identified previously by the Census Project and leading business representatives.
The ASA noted that, “Nonresponse is not random. Hard-to-reach households such as rural families, renters, and low-income households are disproportionately excluded when follow-up is limited. Experience has shown that collecting efforts must continue beyond two attempts to reach them. Further, research shows these respondents most closely resemble the nonrespondents, and their inclusion is critical to reducing bias. Excluding them systematically biases results toward more affluent and easier-to-count groups. We saw these biases manifest in 2020, when pandemic disruptions forced the ACS to scale back field operations. The result was a data set skewed toward higher-income households such that the bureau had to classify it as “experimental.””
The letter also noted the significant cost to the taxpayer of the restrictions in Section 605 of the House CJS FY26 bill. “The monetary costs of additional follow-up are real, but the benefits are greater. A Bureau of Labor Statistics study found that while extreme follow-up yields diminishing returns, early and mid-level efforts are critical to reducing bias. Other research confirms that the hardest-to-reach respondents are the most important for reducing systematic error. A National Academies panel on nonresponse emphasized that limiting contact attempts threatens data quality for key federal surveys and called for innovation in follow-up strategies, rather than restriction. Recent studies linking survey records with administrative data show that when response rates fall, income and poverty estimates are biased upward, with Hispanic and low-income households particularly undercounted.”
