State-Level Cost-of-Living Decomposition — Where Does the Premium Come From?

Findings

The 5 most expensive states average 109.5 on the all-items index, while the 5 cheapest average 87.5. The headline gap is roughly 21.9 index points.

Decomposing that gap:

• Rent-component gap: 138.2 − 61.2 = 77.0 points
• Goods-component gap: 107.7 − 94.2 = 13.5 points
• Rent-to-goods ratio: 5.7×

The rent-driven gap dominates the goods-driven gap by a wide margin at the state level — the same pattern we observe at the metro level. This is not a coincidence: state-level RPP averages are themselves area-weighted aggregates of metro and non-metro RPPs, so the metro-level rent-supply story propagates upward through the geographic hierarchy.

Top-5 most expensive states (live from database)

Top-5 cheapest states (live from database)

Interpretation

The decomposition above is a useful corrective to the common framing that "everything costs more" in expensive states. Tradable goods do not vary much across U.S. geography — the same brand of cereal at the same retailer carries similar prices in Mississippi and California. The cost gap shows up in housing first, services second, and goods last. For households making relocation decisions, this means: (a) if you own your home outright, much of the headline cost-of-living gap evaporates; (b) if you primarily consume tradable goods (large family, lots of online shopping), again much of the gap evaporates; (c) if you are a renter who relies heavily on local services (childcare, dental care, dining out), you bear the full cost-of-living gap.

Policy and Research Implications

The state-level decomposition has implications across several policy domains. Federal cost-of-living adjustments for programs like Social Security currently use a single national CPI series, ignoring geographic variation entirely. A retiree in Mississippi receives the same nominal benefit as one in Hawaii, despite Hawaii's overall RPP running roughly twice Mississippi's. Various proposals to introduce regional COLAs exist; the BEA RPP series provides a defensible methodology if Congress ever chose to act.

Federal poverty thresholds face the same uniformity issue. The official poverty measure applies a single national threshold; the Census Bureau's experimental Supplemental Poverty Measure (SPM) does adjust for housing-cost differences using HUD Fair Market Rent data. The decomposition above suggests that adjusting only for housing captures most of the geographic variation in the cost of living — goods and (to a lesser extent) services do not vary as sharply.

Academic and policy research on real-income comparisons across states or metros increasingly uses BEA RPP to translate nominal incomes into a common purchasing-power basis. The decomposition above explains why this matters: failing to adjust nominal incomes by RPP would systematically overstate real income in expensive states (because their nominal-dollar incomes appear higher) and understate it in cheap states.

Limitations

• State-level RPP averages mask substantial intra-state variation. California's rural Central Valley metros sit close to the national average; San Francisco and San Jose carry the bulk of the state premium. The same is true for New York (Manhattan vs. upstate), Massachusetts (Boston vs. western counties), and other large-state aggregates.
• BEA's index uses national-average household weights. Atypical households see different effective price levels. Renters bear the full rent gap; owner-occupants who hold their home outright are largely insulated. Healthcare-intensive households (chronic conditions, end-of-life care) bear a larger services share than the national average.
• The "top 5" tier is sensitive to small data revisions; a state on the boundary (rank 5 vs. 6) may swap between releases. The bottom-5 tier is more stable because the cheap-state cluster (Mississippi, Arkansas, West Virginia, Alabama, Oklahoma) has been remarkably persistent across BEA releases over the past decade.
• Index values are annual averages; intra-year volatility (especially in housing) is not captured. Hurricane- or wildfire-impacted markets may show transient distortions that smooth out within 12-24 months but create misleading single-year readings.
• State-level data inherently mixes urban and rural geographies. A state with a small urban core dominating its population (Hawaii, Nevada, Massachusetts) will see RPP track urban-area patterns; a state with substantial rural and small-metro population (Wyoming, North Dakota, Mississippi) will reflect a more rural cost basket.
• BEA RPP excludes state and local income tax burdens, which are a meaningful additional cost dimension when comparing across state lines. Tax-aware purchasing-power calculations require layering Tax Foundation State Business Tax Climate or Tax Policy Center state tax data on top of the RPP comparison.

Replication and Methodology

For the full data-pipeline methodology, see the PlainCost /methodology page. The two SQL queries underlying this article run at request time against PlainCost's BEA-mirror database.

Underlying SQL:

SELECT * FROM states ORDER BY rpp_all DESC LIMIT 5;
SELECT * FROM states ORDER BY rpp_all ASC LIMIT 5;

To replicate from BEA's source: download the SARPP table from the BEA Regional Data API, filter to the most recent year, take the top and bottom 5 by All-Items RPP, average the component columns within each tier.