Intergenerational Heterogeneity in Real Measures: A Migration Story

• In Event: Measurement and Data Collection in Migration Research

Abstract

Standard price indices such as the Consumer Price Index (CPI) are among the most consequential statistics produced by the federal government. They serve as inflation benchmarks, deflate nominal income and consumption into real terms, and underpin policy decisions ranging from Social Security adjustments to poverty measurement. Yet the CPI is constructed at the item-area level — it tracks how prices change within geographic areas but does not follow households as they move across those areas. When households systematically relocate from low-cost to high-cost locations, this design choice introduces a bias that grows silently over time.

This paper asks: how much does internal geographic migration bias standard cost-of-living indices, and what are the consequences for measured real economic well-being?

We focus on housing, the expenditure category with the largest cost-of-living differences across locations. In 2019, rents in the cheapest primary sampling unit (PSU) — Florence-Muscle Shoals, AL — were $680 per month, while the most expensive — San Francisco-Oakland-Hayward, CA — reached $2,690 per month, a nearly four-fold difference. Over the same period, the U.S. population systematically shifted toward higher-cost PSUs.

Our data combine three sources covering 1986 to 2019: the Zillow Observed Rent Index (ZORI), which provides metro-level rental prices; the Federal Housing Finance Agency (FHFA) Housing Price Index, used to extend ZORI backward; and March Current Population Survey (CPS) data to construct annual population shares across 59 Bureau of Labor Statistics PSUs.

We build a framework that separates household cost-of-living from the standard CPI construction and derive a migration-adjusted cost-of-living index. Empirically, we compare three price indices: a direct chained Laspeyres using lagged PSU population weights, a two-step chained Laspeyres that mirrors the BLS procedure, and a household-based Jevons index constructed by simulating 1,000 households whose PSU assignments evolve with observed population shares over time. The household index captures the price-level change a household experiences when it migrates — something item-area indices by construction cannot do.

Our central finding is that the household-based Jevons index exceeds the standard two-stage Laspeyres index by 67 percentage points over the 1986–2019 period. The two item-area indices, by contrast, differ by only about 2 percentage points from each other. Applying a housing expenditure share of one-third, this gap implies that real income growth has been overstated by approximately 22 percentage points when area-based deflators are used in place of household-tracking deflators. Furthermore, our estimates constitute a lower bound: omitted low-cost PSUs, the urban sampling frame, and a conservative migration algorithm all attenuate the true gap.

These findings have direct implications for studies of intergenerational well-being. Because younger cohorts are more likely to have migrated to high-cost cities, conventional real measures make younger generations appear relatively richer compared to older cohorts than they actually are. Accurately measuring real economic progress requires aligning price deflators with the households whose welfare we seek to evaluate.

Author

• Michael Alexander Navarrete, Federal Reserve Bank of Atlanta
  Presenting Author