Nonresponse Imputations and Related Measurement Issues in the CPI for Shelter

Shelter is the largest component of U.S. CPI, so the accuracy of shelter inflation is critical for the accuracy of overall inflation. Nonresponse in the BLS Housing Survey, which underpins CPI shelter measurement, has increased over time and now represents roughly 40 percent of observations. Missing rents are currently imputed using a class-mean approach based on rent tier, likely leading to upward-biased imputations. We study alternative simple imputation methods built upon variables correlated with both nonresponse and rent growth, including structure type and tenure length, and explore regression-based approaches in the appendix. While a simple model demonstrates that different methods could yield sharply different index biases, we find that in practice, these alternative methods yield similar shelter inflation indexes, suggesting that any index bias may be modest.