Housing

America's 25 most affordable small towns

Where the median home costs under three years of household income — and what "cheap" actually means in a region that's losing people.

By Commerce Institute · · 880 words
Data drawn from: Census ACS 5-YearCensus Place definitions

In Aspen, Colorado, the median home costs about 35 years of household income. In Bruceton Mills, West Virginia, it costs 0.96. That's the spread that shapes everything else about where Americans can plausibly live, and it sits inside a single measure: years-to-own, the median home value divided by the median household income.

The U.S. average sits a little above four. Below three, you're in a place where the median household could in principle pay off the median home in a Bachelor's-degree's worth of working years. Below two, you're somewhere with deep structural housing slack: the homes that exist outnumber the households who can afford to leave, the population is often declining, and incomes haven't fallen as fast as values have.

Figure 1 · Years-to-own across 19,000 US Census places

Distribution of years-to-own ratios across all U.S. Census places with population 1,000–50,000. Bars to the left of the red dashed line are the 25 most affordable. The U.S. median sits around 4.

024681012Years of household income to buy the median homeU.S. median (3.7)top-25 cutoff (1.19)Aspen, CO: 35.2 (off chart)15,585 US Census places, 1k–50k pop. 122 places have ratios ≥12 (not shown).

Source: U.S. Census American Community Survey 5-Year (B25077, B19013), 2019–2023. See methodology.

The list

Across all U.S. Census places with population between 1,000 and 50,000, twenty-five of them post years-to-own ratios under 1.7. They cluster in five regions: the West Virginia and Pennsylvania Appalachian coal belt, the lower Mississippi River valley (Illinois, Missouri, Arkansas, Mississippi), the Oklahoma–north Texas oil-decline corridor, and a handful of small Ohio Rust Belt remnants. There is one notable absence: no Great Lakes cities, no New England, no Mountain West.

Figure 2 · The 25 most affordable

U.S. Census places, population 1,000–50,000, ranked ascending by years-to-own (median home value ÷ median household income).

RankPlaceStatePopulationYears-to-own
#1AaronsburgPA2,7750.77
#2GirardvillePA1,5460.79
#3SentinelOK1,1150.82
#4MullensWV2,3610.87
#5Bruceton MillsWV8,3900.96
#6SylvesterWV2,1431.01
#7BrentonWV2,6941.01
#8Washington ParkIL10,3631.05
#9AshlandKS1,0481.06
#10BuffaloOK1,5871.06
#11AtlasPA7,1741.09
#12NewmanIL1,1961.10
#13White HallIL2,4881.12
#14FairfaxOK1,5011.12
#15BushnellIL3,0451.13
#16Searles ValleyCA1,8221.13
#17BenhamKY1,2631.13
#18WittIL1,0181.13
#19DavyWV1,9291.14
#20KirkwoodIL1,0531.15
#21MeredosiaIL1,2641.16
#22RotanTX1,6621.17
#23Mahanoy CityPA4,1241.18
#24AstoriaIL1,6121.19
#25CairoIL1,7851.19

What "affordable" actually means here

The numbers at the top of this list are sometimes startling — sub-1.0 ratios mean the median home costs less than a single year of median household income. But that floor is almost always demand-driven, not supply-driven. Bruceton Mills isn't building cheap new houses; existing houses have lost value faster than incomes have. Helena-West Helena, Arkansas, has lost roughly a fifth of its population since 2010. Pine Bluff has lost about a quarter.

That doesn't make the affordability fake — for a buyer with stable income that doesn't depend on local labor markets (remote workers, retirees, multi-state families), it's real. It does mean that the years-to-own number alone won't tell you whether the place is going to feel healthy. Combining years-to-own with population trajectory or net migration from the IRS data fills in that picture; we'll publish that companion piece in a future installment.

Why the Midwest and Appalachia dominate

Three structural patterns produce the regions on this list. First, sustained population decline pulls home values down faster than it pulls incomes — incomes are anchored by what's left of the local economy and by federal transfer payments, neither of which fall in proportion to outmigration. Second, the housing stock in long-settled small towns isn't easily reduced: when households leave, the houses stay, increasing the supply-to-demand ratio. Third, school district and county-property-tax reputations create self-reinforcing devaluation cycles that the housing-affordability literature has been documenting for thirty years.

This is the same set of mechanisms that make these places cheap to enter but hard to sell out of. A buyer entering the bottom of the affordability list at 1.5x household income may also find, twenty years later, that the next buyer's pool has shrunk further.

What this list isn't

This is a places-of-residence affordability list, not a quality-of-life ranking, not a growth-potential ranking, and not an investment-return forecast. The list is just one cut of one ACS estimate against one BEA-adjacent income estimate, restricted to a particular population band. The methodology page shows the equation, the source variables, the known limitations, and the replication SQL. Every place named in this article has a per-place profile linked above; they're worth reading in pairs.

The bigger story, for the housing-affordability debate, isn't the bottom of this list. It's the top. The same warehouse that produces this list also produces an Aspen-CO years-to-own of 35.2 (a value that places it in the top 1% of all U.S. Census places, for housing cost burden) and a Provincetown-MA years-to-own of 26. A country whose affordability range runs from 0.96 to 35 is a country with at least two completely separate housing markets, and any policy that treats them as one will miss.

Notes & data sources

  1. Years-to-own = median home value ÷ median household income, both from Census ACS 5-Year 2019–2023. See methodology.
  2. Population figures are ACS 5-Year 2019–2023; population change figures compare ACS 2010 to ACS 2023, where both are available.
  3. Aspen, CO years-to-own figure (35.2) is the Pitkin County median-home-value vs median-household-income; for the village itself the figure is similar.
  4. We restrict to Census Places with population 1,000–50,000. Below 1,000 the ACS confidence intervals widen too much to compare; above 50,000 the dynamics are different.

Read more

Methodology

Places mentioned