Statistics / Geography

Nowhere is safer — the Northeast just fades later

One regional signal survives scrutiny, and it is about timing, not fragility. Everywhere else, no state or region's record can be separated from the national average once you count the uncertainty.

The Northeast pattern

Northeast rescues don't fail fast — they fail eventually. At two years their survival matches the rest of the country (66.7% vs 68.4%); at five years it collapses:

Northeast at 5 yrs — 18 bars27.8% CI 10.1–48.9

Everywhere else at 5 yrs — 255 bars53.1% CI 46.2–59.5

Their median dated lifespan is 2.77 years — the longest of any region. They don't die young; they die eventually. The 18 bars: NY 5 · PA 4 · MA 3 · MD 2 · NJ 2 · CT 1 · RI 1. See them in the survival explorer →

Association — not proof of causeConsistent across eras and methods, but n = 18 — and only 5 Northeast bars were still being followed at five years.

Regions

Still-open rate with its interval (navy tick = the national 46.1%), the Kaplan–Meier five-year survival, and the median dated lifespan including pre-air closures:

Pacific — 25 of 59 open42.4% CI 30.6–55.1

KM 5-yr 54.8% [40.6–66.9] · median dated lifespan 2.73 yrs

Mountain — 28 of 67 open41.8% CI 30.7–53.7

KM 5-yr 49.9% [36.7–61.7] · median dated lifespan 1.23 yrs

Southeast — 37 of 65 open56.9% CI 44.8–68.2

KM 5-yr 52.7% [36.8–66.3] · median dated lifespan 0.88 yrs

Midwest — 23 of 48 open47.9% CI 34.5–61.7

KM 5-yr 46.5% [28.7–62.5] · median dated lifespan 0.94 yrs

South Central — 13 of 25 open52.0% CI 33.5–70.0

KM 5-yr — not shown (too few at risk) · median dated lifespan 0.70 yrs

Northeast — 3 of 18 open16.7% CI 5.8–39.2

KM 5-yr 27.8% [10.1–48.9] · median dated lifespan 2.77 yrs

South Central's five-year KM figure is not shown: only 11 of its bars were still being followed at five years — too thin to headline. 2 bars fall outside the six regions and appear in the national figures only.

States

Every state's interval includes the national 46.1% — there are no state champions, only a nominal ordering. Search or sort:


California	55	24	43.6% [31.4–56.7]
Florida	39	24	61.5% [45.9–75.1]
Nevada	27	10	37.0% [21.5–55.8]
Texas	20	10	50.0% [29.9–70.1]
Arizona	16	8	50.0% [28.0–72.0]
Colorado	13	6	46.2% [23.2–70.9]
Illinois	12	5	41.7% [19.3–68.0]
Missouri	10	3	30.0% [10.8–60.3]
Tennessee	10	4	40.0% [16.8–68.7]
Georgia	8	4	50.0% · small n
Michigan	8	3	37.5% · small n
Wisconsin	7	5	71.4% · small n
North Carolina	6	4	66.7% · small n
Ohio	6	4	66.7% · small n
Louisiana	5	3	60.0% · small n
New York	5	2	40.0% · small n
Idaho	4	3	75.0% · small n
Pennsylvania	4	0	0.0% · small n
Utah	4	1	25.0% · small n
Massachusetts	3	0	0.0% · small n
New Mexico	3	0	0.0% · small n
Maryland	2	0	0.0% · small n
Nebraska	2	1	50.0% · small n
New Jersey	2	0	0.0% · small n
Oregon	2	1	50.0% · small n
Puerto Rico	2	2	100.0% · small n
Washington	2	0	0.0% · small n
Connecticut	1	1	100.0% · small n
Iowa	1	0	0.0% · small n
Kansas	1	1	100.0% · small n
Kentucky	1	1	100.0% · small n
Minnesota	1	1	100.0% · small n
Mississippi	1	0	0.0% · small n
Rhode Island	1	0	0.0% · small n

States under 10 bars show counts only, with the rate greyed — at that size the interval would span most of the scale. States also differ in era mix, which raw rates don't adjust for.

The neighbourhood doesn't predict it

It isn't the neighbourhood. Richer, poorer, older, denser, more educated — the census tracts of bars that survived are statistically indistinguishable from those of bars that closed. 279 of the 284 bars are matched to a census tract and analysed here; income data covers 274 of them (ACS 2024 5-year estimates; per-measure counts in the table):

Tract measure	Open bars: median	Closed bars: median	Mann–Whitney
Median household income	$75,114 (n=125)	$78,237 (n=149)	p = 0.216
Tract population	3,965 (n=127)	3,996 (n=151)	p = 0.942
Median age	38.3 (n=127)	38.5 (n=151)	p = 0.807
Share with a bachelor's degree	30% (n=127)	35% (n=151)	p = 0.060
Share in poverty	13% (n=127)	12% (n=151)	p = 0.455
Population density (per sq mile)	3,715 (n=127)	4,918 (n=151)	p = 0.008

Density is the only sub-0.05 result in six tests (p = 0.008) and does not survive a multiple-comparison correction — treat it as exploratory, entangled with the Northeast/urban pattern above.

Association — not proof of causeTract medians describe the area around the bar, not its customers or its rent roll; a null here rules out large neighbourhood effects, not neighbourhood effects entirely.

Raw regional and state rates count every bar equally regardless of how long it has had to fail; the survival explorer separates timing properly. Back to the overview →