Nobody Asks the Renter Before the Panel Gets Upgraded

The cross-tab nobody had run

HyreElectrical’s hazard-era panel study counted 67,068,006 US housing units — occupied and vacant together — built between 1950 and 1989, the years Federal Pacific Stab-Lok and Zinsco panels were on the market, and a narrower 34.7 million built in the 1960s and 1970s aluminium-wiring window. That study used Census table B25034, which records year built for every housing unit but carries no field for who lives there. It could size the hazard. It could not say whose problem it is. The construction-era norms behind that window are in housing age and electrical systems.

The split-incentive barrier is a well-described structural problem in energy research. The U.S. Department of Energy’s Better Buildings program has run peer exchanges specifically on landlord-tenant conflict over efficiency and electrification upgrades, and DOE’s own electrical-panel research program states plainly that “upgrading panels is not always feasible — especially for low-income families.” Add a lease to that constraint and the household facing it may not even be the one with the financial exposure. What that constraint looks like in households already behind on the bill is home energy insecurity. Evidence on the barrier is genuinely mixed: a 2010 evaluation of two Southern California Edison multifamily efficiency programs found that property owners frequently installed efficiency upgrades in tenant units anyway, for reasons unrelated to the split-incentive theory — tenant satisfaction, property value, or the belief that a lower utility bill leaves a tenant more room to pay rent. That evidence concerns discretionary efficiency retrofits a landlord can choose to skip. A panel that an inspector has flagged or an insurer has declined over is a different kind of decision, and nobody had counted how many renters actually sit inside it. That count follows.

The national split, by tenure

Of 127,482,865 occupied US housing units in the 2023 five-year American Community Survey, 82,892,037 are owner-occupied and 44,590,828 are renter-occupied — a 65.0/35.0 split. Cross-tabbed against year built (table B25036), 38,323,134 owner-occupied homes (±91,651 at 90 per cent confidence) and 21,710,304 renter-occupied homes (±53,624) fall in the 1950–1989 window, for a combined 60,033,438 (±106,186).

HYRE calculation. That combined figure is 47.1 per cent of all occupied housing — visibly consistent with the hazard-era study’s 47.1 per cent share of the total housing stock (occupied and vacant together), which is a useful internal check: vacancy is not concentrated in one construction era enough to move the headline number. Within the era-built stock, the tenure split is 63.8 per cent owner-occupied to 36.2 per cent renter-occupied — a full percentage point above renters’ 35.0 per cent share of the occupied stock overall. Framed the other way: 48.7 per cent of every renter-occupied home in America sits in the hazard-era window, against 46.2 per cent of every owner-occupied home. Neither gap is large. Both point the same direction.

Where owned and rented housing age differently

In 42 of the 51 jurisdictions measured, the rental stock has a higher hazard-era share than the owned stock — renters are more likely than owners to live in a home from this window, not less. The gap is largest in the District of Columbia (38.0 per cent of renters against 24.9 per cent of owners, a 13.2-point gap) and Nevada (39.5 against 30.5, a 9.0-point gap), and it runs to seven or eight points in Alabama, Mississippi, New Mexico and Delaware.

HYRE analysis. Nine states run the other way, with OWNED housing older than rented housing: New York (49.2 per cent of owners against 39.7 of renters), Vermont, Connecticut, North Dakota, Massachusetts and four others. The pattern in most of these states is a housing market where newer apartment construction has outpaced newer single-family construction — the rental stock skews toward recent multifamily building even as the owner-occupied stock, concentrated in older single-family homes, stays put. Neither pattern is universal, which is the honest finding: the split-incentive population this page counts is not evenly distributed, and a household in New York reasoning from the national 48.7 per cent figure would be using the wrong number for their state.

The largest rental markets, owner share against renter share

Restricting the comparison to the fifteen states with the largest rental populations makes the pattern easier to read at the scale that matters most: these are the states carrying most of the country’s renter-occupied hazard-era stock in absolute terms. Twelve of the fifteen — California, Florida, Michigan, Ohio, New Jersey, North Carolina, Georgia, Illinois, Pennsylvania, Washington, Texas and Nevada — show a higher hazard-era share among renters than owners. Connecticut, New York and the District of Columbia show the reverse, each for the reason given above: an owner-occupied stock that is, on average, older still.

The income context: who has less leverage to force the issue

ACS table B25119 puts 2023 median household income at $97,352 (±143) for owner-occupied households and $51,393 (±84) for renter-occupied households nationally — the renter median is 47.2 per cent below the owner median.

HYRE analysis. Income does not appear anywhere in the tenure-by-year-built table, and this page draws no causal line between a household’s income and whether its panel gets replaced. What the income gap does establish is which side of the split incentive has less room to force the question through a costlier channel — moving, hiring a tenant-side attorney, or covering the work itself and seeking reimbursement. A renter already earning roughly half the area’s owner-occupied median is the same renter this page finds has no legal standing to order the electrical work in the first place. The two facts are separately sourced and they compound.

Method

Primary source. US Census Bureau, American Community Survey 2023 5-year estimates, table B25036 (Tenure by Year Structure Built), for the United States and each of the 50 states and the District of Columbia. Retrieved 6 September 2026 from the public table-based Summary File at www2.census.gov after an unauthenticated request to api.census.gov returned a missing-key redirect — the same substitution the hazard-era study records, changing no figure.

Derivation. The 1950–1989 sum for each tenure is table lines 007–010 (owner: built 1980–1989, 1970–1979, 1960–1969, 1950–1959) and lines 018–021 (the same four decades for renter-occupied). Margins of error are the square root of the sum of the squared component margins — the Census Bureau’s published method for derived sums — at 90 per cent confidence. Median household income by tenure is table B25119, same vintage, same retrieval route.

RECS derivation. EIA’s 2020 public-use microdata file (recs2020_public_v7.csv, 18,496 responding households) was aggregated using its published final analysis weight, NWEIGHT. Renter-occupied combines KOWNRENT codes 2 (rent) and 3 (occupy without payment of rent), matching the Census Bureau’s renter-occupied definition. Built 1950–1989 combines YEARMADERANGE codes 2 through 5. Standard errors use EIA’s 60 Jackknife replicate weights (NWEIGHT1–NWEIGHT60) with the variance formula EIA states in its own 2020 RECS Methodology Report: Var = (59/60) × the sum of squared deviations of each replicate estimate from the full-sample estimate. The codebook and methodology report were downloaded directly from eia.gov on the retrieval date.

What this page reuses, and what it does not. The 1950–1989 construction-era window and its significance for Federal Pacific, Zinsco and aluminium-wiring installation dates are established in the hazard-era study and are not re-argued here; see that page for the CPSC and carrier-underwriting evidence. Every housing-unit count on this page is newly derived from B25036, a table that study does not use, because B25034 has no tenure field.

Reproducibility. Every input is a public file: the B25036 and B25119 Summary File extracts, and EIA’s RECS 2020 public microdata and codebook. Anyone with those files can rebuild every figure on this page.