Nonresponse Imputations and Related Measurement Issues in the CPI for Shelter

Shelter is the largest component of US consumer price index (CPI) inflation; therefore, the accuracy of shelter inflation is critical for the accuracy of overall CPI inflation. Nonresponse in the BLS Housing Survey, which underpins the measurement of CPI shelter inflation, has increased since 2000 and now represents roughly 40 percent of total observations. Missing rent data are currently imputed using a class-mean approach based on rent tier, potentially resulting in biased imputations, as we find that nonresponse is correlated with factors beyond rent tier. We study alternative simple imputation methods based on variables correlated with both nonresponse and rent growth, including structure type and tenure length. A simple model demonstrates that alternative methods could yield sharply different index biases. However, in practice, we find that these alternative methods yield similar shelter inflation indexes, suggesting that any index bias may be modest.