The Roofer Shortage Index

Why a count of roofers needs a denominator

A “roofer shortage” can be stated as a bare assertion or as an absolute count: so many thousand unfilled positions, so many contractors reporting difficulty hiring.

A count with no denominator cannot tell you whether a place is short of roofers. Two hundred roofers is a glut in a town of six thousand houses and a crisis in a city of four hundred thousand.

This study puts a denominator on it. The work a roofer does is, at bottom, work on a roof, and the closest public measure of the number of roofs is the number of housing units.

So the index is deliberately crude and deliberately reproducible: roofers employed, divided by housing units, times 100,000. Two federal datasets, one division, no weighting, no composite score, nothing you cannot check.

The point of a single-variable index is that it can be argued with. A twenty-factor “best places for roofers” score cannot be falsified by anybody, which is why we did not build one.

This one can be: if you think housing units are the wrong denominator, the Method section below re-runs the whole thing on a second denominator and reports how much the ranking moves.

The index across 50 large metropolitan areas

The bottom of that chart holds some very large markets. San Antonio, Houston and Austin (three Texas metros in a state with no roofing license and a great deal of hail) sit at 31.9, 40.4 and 42.6 roofers per 100,000 housing units.

Greater New York, with more than eight million housing units, records 3,100 employed roofers: 38.1 per 100,000, the second-lowest of the fifty.

The top is Florida and the West Coast, and Florida has its own reasons: see inside the Florida roofing market for the statutes and 9,402 DBPR license records. North Port–Bradenton–Sarasota records 384.9, Orlando 221.3 and Tampa–St Petersburg 216.1; Portland–Vancouver–Hillsboro records 209.7 and Seattle–Tacoma–Bellevue 169.5.

HyreRoof analysis: the ordering is not a ranking of how well served homeowners are.

It is a measure of how much roofing labor is carried on somebody’s payroll in that market, which is a function of storm-driven replacement demand, of how large the average roofing firm is, and of how much of the work is done by people the survey never sees.

If roofers were scarce, would they cost more?

This is the test that separates a genuine labor shortage from a shortage in the colloquial sense.

A shortage, properly speaking, is excess demand at the going price, and the signature of excess demand is that the price rises.

If the states with the fewest roofers per house were genuinely short of roofers, roofers in those states should be expensive.

HyreRoof calculation: across the 50 states and the District of Columbia, the correlation between the index and the state annual median wage for roofers is r = −0.035 (Spearman ρ = 0.066). That is not a weak relationship; it is the absence of one.

The ten states with the lowest index pay a mean state median of $53,572. The ten with the highest pay $54,615. A difference of $1,043 a year, under two percent, between the thinnest and thickest roofing workforces in the country.

The extremes make the point better than the coefficient does. Illinois pays roofers a median of $77,900, the highest in the country, on a middling index of 96.7.

Florida has more than twice Illinois’s density of roofers and pays $47,590. South Carolina has the thinnest roofing workforce of any state relative to its housing stock (33.6) and pays $45,760.

HyreRoof analysis: the honest reading is that state roofing wages are set by things this index does not contain (cost of living, union density, the residential-versus-commercial mix, prevailing-wage rules on public work), and that whatever scarcity the index measures is not strong enough to show through them.

It does not prove no contractor anywhere struggles to hire. It does mean that “there are not enough roofers” cannot be supported by pointing at a low ratio of roofers to roofs, because the low-ratio states are not paying like it.

The openings are churn, not growth

The Bureau of Labor Statistics projects roofing employment rising from 166,900 in 2025 to 175,600 in 2035: an increase of 8,800 jobs (BLS’s own figure, computed before rounding), or 5 percent, described as faster than the average for all occupations. Over the same decade it projects about 12,000 openings for roofers each year.

HyreRoof calculation: 8,800 additional jobs over ten years is 880 a year. Subtract that from 12,000 and 11,120 of the 12,000 annual openings, 92.7 percent, exist because somebody left, not because the industry grew. BLS attributes those to workers transferring to different occupations or leaving the labor force altogether.

That reframes the whole question. The recruitment problem in roofing is not principally that demand is outrunning supply. It is that the occupation loses roughly one worker in fifteen every year and has to find another.

A trade with that much turnover will always feel short of people to the person doing the hiring, whatever the ratio of roofers to roofs looks like from the outside.

Why roofers leave is not a question this dataset answers, and we do not guess at it here. The risk of the work is measured separately in our roofing fatality atlas.

Method

Research question. How many employed roofers are there per unit of housing stock, how much does that vary geographically, and does the variation behave the way a labor shortage would?

Numerator. BLS Occupational Employment and Wage Statistics, May 2025 estimates, occupation SOC 47-2181 (Roofers), cross-industry, retrieved from the OEWS time-series files published by BLS on September 3, 2026.

Series identifiers are constructed as OEU + area type + seven-digit area + 000000 (all industries) + 472181 + two-digit data type, so any figure here can be pulled again from the same file. Employment is data type 01, annual median wage 13, location quotient 17. BLS rounds employment to the nearest ten.

Denominator. US Census Bureau, American Community Survey 2024 1-year estimates, table B25001 (total housing units), taken from the table-based summary file and joined on the CBSA code in the accompanying geography file.

The secondary denominator is table B25034 (year structure built): total housing units less the units built 2000 or later, which we label pre-2000 stock.

The calculation. Index = employment ÷ housing units × 100,000. Nationally: 135,490 ÷ 146,740,964 × 100,000 = 92.3, equivalently one employed roofer per 1,083 housing units. Over pre-2000 stock: 135,490 ÷ 106,509,700 × 100,000 = 127.2.

Inclusion. Every OEWS metropolitan area whose seven-digit area code resolves to a CBSA present in the ACS 2024 1-year file. That is 265 of the 393 metropolitan areas in the OEWS area list; the other 128 have no published roofer estimate for May 2025.

BLS nonmetropolitan balance-of-state areas are excluded because they are not CBSAs and have no ACS counterpart. The ranked table is then restricted to metros with at least 500,000 housing units, leaving 50, because OEWS employment rounded to the nearest ten moves the index of a small metro by several points.

Sensitivity on the self-employment assumption. The numerator omits self-employed roofers. Run at three values rather than one, the national index becomes 108.6 if self-employed roofers are 15 percent of the trade, 115.4 at 20 percent (the Handbook’s figure), and 123.1 at 25 percent.

The geographic ranking is unaffected by any of these, because a constant multiplier cannot reorder a list, which is precisely why the ranking is more defensible than the level.

Sensitivity on the denominator. Re-ranking the 50 large metros on pre-2000 housing units instead of all housing units gives a Spearman rank correlation of 0.94 with the published order; across the 52 state-level rows it is 0.97.

The choice of denominator moves individual metros (Austin and Raleigh in particular, both with young stock) but does not change the picture.

The wage test. Pearson correlation between the state index and the state annual median wage for SOC 47-2181, n = 51 (50 states and DC; Puerto Rico excluded because its wage level is far outside the state range). r = −0.035, Spearman ρ = 0.066.

Means of the ten lowest and ten highest states by index: $53,572 and $54,615.

What we did not do. We did not build a composite score, weight the factors, adjust wages for cost of living, model demand, or estimate an unmet-demand figure. Each of those would require an assumption we could not source, and an index nobody can check is not research.