Regional analysis · Collin and Denton counties, north of Dallas–Fort Worth

Built together, due together

Name: Housing construction-era cohorts and storm exposure across 100 US counties
Creator: Before Regret
License: https://creativecommons.org/licenses/by/4.0/
Keywords: housing age cohorts, roof age, construction era, severe weather, North Texas, county data

Half the houses in Collin and Denton counties — McKinney, Plano, Frisco, Denton, Lewisville and Flower Mound among them — went up inside the same twenty years, in one of the country's heaviest hail corridors. Roofs do not age one house at a time here. They age in a cohort, and the cohort comes due together.

50.9%of Collin County's 421,938 homes were built 2000s or 2010s

50.7%the same for Denton County, the 7th most concentrated stock of the 100

1,240hailstorms recorded across the four DFW counties, 2015–2024

3 of 100counties are both this concentrated and this hail-exposed

Look up any of the 100 counties

Choose a county

NOAA records reports of severe weather, not a census of it, so counts describe how often something damaging was written down. Insurance shares are what mortgaged households told the Census they pay, in bands — there is no average premium in this data and none should be quoted from it.

The screen

Two conditions, applied to all 100 counties, leave three standing.

Take the share of a county's housing built in its busiest twenty-year window — a crude measure of whether the place went up gradually or in a burst. Then take the number of hailstorms NOAA recorded there between 2015 and 2024. Ask for at least 45% in one window and at least 100 hailstorms, and almost nothing qualifies.

County	Built in one 20 years	Which years	Hailstorms	Housing units
Denton, TX	50.7%	2000s–2010s	355	364,120
Collin, TX	50.9%	2000s–2010s	283	421,938
Wake, NC	45.0%	2000s–2010s	122	481,999

All 100 counties screened. Two of the three survivors share a border.

Denton ranks 3rd of the hundred for recorded hail and Collin 5th. They are also the 6th and 7th most concentrated housing stocks in the study. Very few places in America are both.

Why concentration is not the finding on its own

Plenty of counties were built in a burst. Almost none of them get hailed on.

County	Built in one 20 years	Which years	Hailstorms
Kings, NY	55.6%	pre-1940–1940s	2
Fort bend, TX	55.4%	2000s–2010s	33
Clark, NV	54.3%	1990s–2000s	28
San francisco, CA	54.1%	pre-1940–1940s	10
Suffolk, MA	51.0%	pre-1940–1940s	6

The most concentrated housing stocks that see almost no hail. Brooklyn, San Francisco and Philadelphia are concentrated in pre-1940 construction; Fort Bend and Clark are modern, like Collin and Denton, but sit outside the hail corridor.

That is the comparison that matters. Fort Bend County is more concentrated than either North Texas county — 55.4% of its homes went up in the 2000s–2010s — and recorded 33 hailstorms in ten years against Denton's 355. Clark County, Nevada is built the same way and recorded 28. Synchronised construction is common. Synchronised construction under hail is not.

What that does to a roof

The component hail destroys is the one with a fixed service life.

Hail damages roofs before it damages anything else, and an asphalt shingle roof has a finite life regardless of what the sky does — commonly quoted at fifteen to thirty years depending on product and installation. Insurers price on that directly: roof age is one of the few things an underwriter asks about a specific house, and past a certain age carriers move from replacement cost to actual cash value, raise the wind and hail deductible, or decline the risk. The same underwriting logic is what makes a flat roof hard to insure, where the age caps are shorter still.

In a county where the housing arrived gradually, that is a rolling problem. Some roofs are new, some are old, and the replacement bill for the county is spread across every year. Allegheny County, Pennsylvania has housing from every decade and a largest single decade of 28.2%, which is the ordinary case.

Collin County, TX

pre-1940 0.6%

1940s 0.5%

1950s 0.9%

1960s 2.2%

1970s 6.9%

1980s 13.5%

1990s 20.4%

2000s 25.3%

2010s 25.6%

2020s 4.3%

Allegheny County, PA

pre-1940 28.2%

1940s 8.5%

1950s 17.9%

1960s 12.0%

1970s 10.6%

1980s 7.1%

1990s 5.6%

2000s 5.0%

2010s 4.7%

2020s 0.5%

Share of housing units by decade built. Bars scaled to 30%. Census ACS 5-year.

Collin and Denton did not arrive gradually. Over half of each county's housing went up in a single twenty-year window, which means the original roofs went on in that window too. They do not come due one house at a time; they come due together, in a place that records a hailstorm somewhere in the county roughly every ten days of the storm season.

The county with the hail but not the timing

Tarrant records more hail than either and does not have this problem in the same shape.

Tarrant County logged 416 hailstorms over the decade, 2nd of the 100 counties and more than Denton or Collin. But its housing is spread much more evenly — 32.7% in its busiest twenty years, against Collin's 50.9% — so Tarrant has the hail without the cohort. Its roofs were not all installed at once and do not expire at once.

That difference is the whole argument. Hail frequency alone does not create a synchronised replacement wave, and neither does concentrated construction alone. It takes both, and across the 100 counties examined here only 3 have both.

For newsrooms

Free to reuse, including the underlying county data.

This analysis and its data are published under a Creative Commons Attribution 4.0 licence, so a newsroom can republish the figures, redraw the charts, or run its own cut without asking. Both source datasets are US Government works and public domain in their own right: the Census Bureau's American Community Survey year-structure-built estimates, and the NOAA Storm Events Database. The decade-by-decade housing counts behind the concentration figures are published as their own file, so the central number in this study can be recomputed rather than taken on trust.

If you are looking for Dallas–Fort Worth hail statistics, Collin or Denton County housing data, or a count of hailstorms by county, this is the whole of what we hold. One caution before quoting it: decade of construction is not roof age, and this study makes no claim about how many roofs in either county are past their service life — the ACS does not track replacement.

Happy to pull the same cut for Tarrant, Dallas, Wake or any other county in the file, walk through the method, or check a figure before it prints — hello@beforeregret.com. If you find an error, say so and it will be corrected on the page with a note.

Method

How this was built.

Housing counts by decade of construction are American Community Survey 5-year estimates, year structure built, expressed as a share of total housing units. Concentration is the largest combined share of any two adjacent decade buckets — a twenty-year window that moves, not a fixed one. Hail counts are every county-tagged hail entry in NOAA's Storm Events Database for the ten calendar years 2015–2024. The comparison set is the 100 counties in 31 states with complete data.

Raw hail counts are used rather than hail per square mile, for the reason set out in the companion study: across these counties, event density correlates with population density at a Spearman coefficient of 0.681, because NOAA records reports and an unwitnessed storm is frequently never written down.

Limitations

What this cannot tell you.

This study deliberately does not use insurance prices, and the reason is worth stating because the obvious version of this story does. ACS table B25141 reports premiums in absolute dollar bands, so a county of expensive houses places more households above $3,000 whatever its weather does. Collin and Denton do sit high on that measure, and they are also affluent; Hidalgo County has similarly new housing, sits in Texas, and has only 6.0% of households in the top band. Separating the two would need median home values, which are not in this dataset, so no claim is made here about what anyone pays.

Decade of construction is not roof age. A roof installed in 2005 may have been replaced twice since, particularly after a hail claim, and the ACS does not track that. What the concentration measure establishes is when the housing — and therefore the first roof on it — arrived, which sets the clock rather than reading it.

The twenty-year window is a choice. A fifteen- or thirty-year window would rank counties differently, and the full table below is published so a different threshold can be applied to the same numbers. The 45% and 100 cut-offs were chosen to be round, not to produce three survivors.

Nothing here describes an individual property. Every figure is a county aggregate.

Appendix

All 100 counties by housing concentration.

#	County	Peak 20yr	Years	Hail
1	Kings, NY	56%	pre-1940–1940s	2
2	Fort bend, TX	55%	2000s–2010s	33
3	Clark, NV	54%	1990s–2000s	28
4	San francisco, CA	54%	pre-1940–1940s	10
5	Suffolk, MA	51%	pre-1940–1940s	6
6	Collin, TX	51%	2000s–2010s	283
7	Denton, TX	51%	2000s–2010s	355
8	Philadelphia, PA	51%	pre-1940–1940s	16
9	Gwinnett, GA	50%	1990s–2000s	20
10	New york, NY	49%	pre-1940–1940s	4
11	Nassau, NY	48%	1950s–1960s	10
12	Will, IL	47%	1990s–2000s	89
13	Hidalgo, TX	47%	1990s–2000s	50
14	Pinellas, FL	46%	1970s–1980s	5
15	Lee, FL	45%	1990s–2000s	14
16	Utah, UT	45%	2000s–2010s	10
17	Wake, NC	45%	2000s–2010s	122
18	Palm beach, FL	45%	1970s–1980s	48
19	Bronx, NY	44%	pre-1940–1940s	2
20	Queens, NY	44%	pre-1940–1940s	3
21	Broward, FL	44%	1970s–1980s	54
22	Cobb, GA	43%	1980s–1990s	22
23	Providence, RI	43%	pre-1940–1940s	10
24	Fairfax, VA	42%	1970s–1980s	51
25	Travis, TX	42%	2000s–2010s	147
26	Honolulu, HI	41%	1960s–1970s	1
27	Ventura, CA	41%	1960s–1970s	0
28	Mecklenburg, NC	40%	2000s–2010s	38
29	Orange, CA	40%	1960s–1970s	3
30	San diego, CA	40%	1970s–1980s	11
31	Dupage, IL	40%	1970s–1980s	58
32	Maricopa, AZ	40%	1990s–2000s	82
33	Riverside, CA	39%	1990s–2000s	7
34	Essex, MA	39%	pre-1940–1940s	20
35	Essex, NJ	39%	pre-1940–1940s	2
36	San bernardino, CA	38%	1970s–1980s	13
37	Santa clara, CA	38%	1960s–1970s	33
38	San mateo, CA	38%	1950s–1960s	9
39	Wayne, MI	38%	1940s–1950s	23
40	Orange, FL	38%	1990s–2000s	47
41	Milwaukee, WI	38%	pre-1940–1940s	70
42	Erie, NY	38%	pre-1940–1940s	31
43	Suffolk, NY	38%	1960s–1970s	17
44	Cuyahoga, OH	37%	pre-1940–1940s	39
45	Middlesex, MA	37%	pre-1940–1940s	47
46	St. louis, MO	37%	1950s–1960s	77
47	Fulton, GA	37%	1990s–2000s	30
48	Allegheny, PA	37%	pre-1940–1940s	115
49	Polk, FL	36%	1990s–2000s	28
50	Hudson, NJ	36%	pre-1940–1940s	0
51	Pima, AZ	36%	1970s–1980s	54
52	Snohomish, WA	36%	1990s–2000s	0
53	Bergen, NJ	35%	1950s–1960s	42
54	Westchester, NY	35%	pre-1940–1940s	16
55	Cook, IL	35%	pre-1940–1940s	171
56	Montgomery, MD	34%	1970s–1980s	58
57	Sacramento, CA	34%	1970s–1980s	2
58	Bexar, TX	34%	2000s–2010s	268
59	Hillsborough, FL	34%	2000s–2010s	13
60	Dallas, TX	34%	1970s–1980s	186
61	Bernalillo, NM	33%	1970s–1980s	63
62	Worcester, MA	33%	pre-1940–1940s	62
63	Lake, IL	33%	1980s–1990s	64
64	Contra costa, CA	33%	1970s–1980s	12
65	Macomb, MI	33%	1960s–1970s	26
66	Los angeles, CA	33%	1950s–1960s	9
67	Dekalb, GA	33%	1970s–1980s	14
68	Harris, TX	33%	2000s–2010s	149
69	Prince george's, MD	33%	1960s–1970s	16
70	Tarrant, TX	33%	1990s–2000s	416
71	Norfolk, MA	33%	pre-1940–1940s	15
72	Tulsa, OK	32%	1970s–1980s	215
73	El paso, CO	32%	1990s–2000s	717
74	Pierce, WA	32%	1990s–2000s	0
75	Baltimore, MD	32%	1950s–1960s	45
76	Kern, CA	32%	1970s–1980s	8
77	Miami-dade, FL	32%	1970s–1980s	52
78	El paso, TX	32%	2000s–2010s	57
79	Hamilton, OH	31%	pre-1940–1940s	79
80	San joaquin, CA	31%	1990s–2000s	2
81	Oakland, MI	31%	1960s–1970s	25
82	Monroe, NY	31%	pre-1940–1940s	15
83	Oklahoma, OK	31%	1960s–1970s	296
84	Fresno, CA	30%	1970s–1980s	2
85	Middlesex, NJ	30%	1950s–1960s	29
86	Duval, FL	30%	1990s–2000s	69
87	Jefferson, AL	30%	1960s–1970s	46
88	Shelby, TN	30%	1970s–1980s	51
89	Multnomah, OR	30%	pre-1940–1940s	0
90	Davidson, TN	29%	2000s–2010s	64
91	Salt lake, UT	29%	1970s–1980s	29
92	Jackson, MO	28%	1960s–1970s	127
93	Jefferson, KY	28%	1950s–1960s	60
94	Alameda, CA	28%	1960s–1970s	19
95	Montgomery, PA	28%	1950s–1960s	59
96	Franklin, OH	28%	1970s–1980s	79
97	Hennepin, MN	27%	1970s–1980s	132
98	King, WA	27%	2000s–2010s	0
99	Denver, CO	26%	2000s–2010s	142
100	Marion, IN	26%	1950s–1960s	67

Cite this Before Regret, “Built together, due together: synchronised housing construction and hail exposure in North Texas.” 6 September 2026.
https://www.beforeregret.com/research/north-texas-roof-age/
Data source: US Census Bureau ACS year-built estimates; NOAA Storm Events Database.
Analysis of Census ACS year-built estimates and NOAA Storm Events across 100 US counties. Full county dataset: CSV · JSON · housing by decade. Questions and corrections: hello@beforeregret.com

Embed the county lookup Free to use, no tracking, no attribution required beyond the credit line it carries.

<iframe src="https://www.beforeregret.com/research/north-texas-roof-age/embed/"
        width="100%" height="420" style="border:1px solid #e5e0d8"
        title="Storm frequency and insurance cost by county"
        loading="lazy"></iframe>

It posts its own height to the parent page as beforeRegretEmbedHeight if you want it to resize with the content.