When Detroit Heats Up,
So Does Its Crime

01 — The headlineWarmer days are busier days for Detroit police

Across 3,441 days, the daily mean temperature tracks about 36% of the variance in daily crime counts (Pearson r = 0.60) — though most of that is the shared seasonal swing, not genuine day-to-day movement. Stripped of the calendar, the link is weaker but real: warmer-than-normal days still carry more crime than their season predicts (about 10% of the within-season variance, r = 0.32; section 04). The relationship is strong, positive, and broadly monotonic (though a quadratic term detects mild curvature, so the per-10°F figure is best read as an average slope across the observed range): each additional 10°F is associated with roughly 5% more reported crime.

02 — The seasonal engineCrime and temperature rise and fall together

Aggregated by month, the two series trace nearly the same arc: a winter trough in January–February and a summer peak in July–August. July, the hottest month, is also the single highest-crime month of the year.

This co-movement is the heart of the story — but it also raises a fair objection: maybe summer brings more crime for reasons that merely coincide with heat (school being out, longer daylight, more travel). Section 04 tackles that head-on.

03 — Not all crime is equalHeat inflames confrontation, not paperwork

Breaking the effect out by offense category reveals a sharp split. Violent and interpersonal crimes are the most temperature-sensitive: aggravated assault, weapons offenses, and homicide all climb steeply with the mercury. Property and outdoor-opportunity crimes respond moderately. And administrative or indoor offenses — fraud, drug cases, court-process violations — barely move at all.

The heatmap below makes the texture vivid. Read across a row: deep red means a category runs well above its yearly average on hot days, deep blue means well below. Aggravated assault and weapons offenses swing from ~0.69× their average in deep cold to ~1.29× in heat — a near-doubling. Fraud and drug cases stay flat all the way across.

04 — Is it really the heat?Yes — even after stripping out the seasons

To separate temperature from everything else that makes summer summer, we computed anomalies: we fit each series a smooth seasonal cycle (a low-order harmonic curve through the year) and subtract it, leaving only how much warmer-or-cooler and higher-or-lower-crime each specific day was than its time of year would predict. If a 70°F day in May behaves like a 70°F day in September, season — not heat — would be doing the work, and the anomaly correlation would vanish.

It doesn't. Warmer-than-normal days still carry more crime than normal (r = 0.32), and the effect is strongest for the high-volume violent categories flagged above (aggravated assault, assault, weapons, homicide). A few lower-volume offenses whose raw temperature slope looked steep — sexual assault, disorderly conduct, OUIL — lose significance once the season is stripped out, a reminder that the per-category “sensitivity” in section 03 still carries seasonal confounding. But for the violent core, an unseasonably hot day is a higher-crime day on its own.

05 — The clockHeat hits hardest after dark

Joining each violent incident to the ambient air temperature in that very hour, the dose-response is strikingly clean: the city averages 2.6 violent incidents in a sub-20°F hour, rising to 4.6 in a 75°F+ hour — a 76% increase, hour for hour.

But the heat doesn't act evenly around the clock. Comparing the warmest third of days with the coldest third, daytime violence (8 am–6 pm) runs about 21% higher on hot days — while the late-night window (8 pm–3 am) jumps 43% higher. Warm evenings keep people outdoors and in contact long after a cold night would have emptied the streets.

06 — The calendarHot weekends are the most violent days of all

Weekends already carry more violence than weekdays at every temperature, and heat lifts both. Stacking the two effects, a hot summer weekend averages 116 violent incidents a day — versus just 65 on a frigid weekday, a 79% swing from the calm extreme to the volatile one. Across the temperature range, weekends run about 10 extra violent incidents per day above weekdays.

For total crime, the temperature slope is nearly identical on weekends (+5.0% per 10°F) and weekdays (+4.7%) — heat raises the baseline rather than steepening it. The weekend's distinct signature is in violence specifically, where social activity and temperature compound.

07 — The mapHeat sensitivity is a downtown, riverfront story

Crime is not spread evenly across Detroit, and neither is its responsiveness to heat. The density map below traces the familiar geography — corridors along the major avenues, concentration through the greater downtown core, the river defining the southern edge.

Coloring each neighborhood by its temperature sensitivity reveals a pattern with a clear lead: entertainment and riverfront districts dominate the top of the list. Among the 88 of 91 neighborhoods whose slope is statistically distinguishable from zero (Benjamini-Hochberg FDR < 0.05), the three most sensitive are Rivertown (+14% per 10°F), Greektown (+13%), and Grixdale Farms (+10%) — the first two riverfront/nightlife districts where warm weather draws crowds to bars, festivals, and the riverwalk, the third a reminder that the effect is not exclusively a downtown one. Downtown proper itself runs hot (+8%), while quieter residential neighborhoods hover near the citywide +4–5%. The ranking is of noisy per-neighborhood point estimates, so read the broad pattern, not the exact order.

At the precinct level the effect is broad but graded — every one of Detroit's precincts shows a positive point estimate (from +3.8% to +6.4% per 10°F), all individually significant after FDR correction.

08 — The skyRain quiets violence; snow blunts theft

Temperature is not the only thing the weather does. Across 1,668 wet days and 1,773 dry ones, precipitation leaves its own mark — but because rain and snow are tangled up with temperature (it only snows when it's cold), every figure here holds temperature constant via regression, isolating the precipitation effect itself.

The clearest signal is on violent crime: a wet day sees about -5% fewer violent incidents than a dry day at the same temperature. Property crime barely moves (-1.4%), and administrative offenses — fraud, warrants, court process — are statistically untouched (-0.0%, n.s.). Rain keeps would-be antagonists indoors and off the streets; it does little to stop a fraud report from being filed.

And the suppression is not a fluke of warm rainy days — it holds inside every temperature band, from freezing to sweltering, each wet column sitting a few percent below its dry neighbor.

Rain and snow do different jobs. Separating precipitation by type — again at equal temperature — uncovers a clean split. The effects are marginal rates per inch (most days see only a fraction of an inch, so these scale a full inch out to the wetter tail): an inch of rain is associated with about 10 fewer violent incidents on the day but leaves theft essentially alone. An inch of snow does the opposite: it cuts property crime by roughly 6 incidents — buried cars, shuttered storefronts and empty sidewalks shrink the opportunity for larceny and break-ins — while denting violence only modestly.

For reference, here is the raw picture by sky condition. Snow days look dramatically calmer — but most of that gap is the cold they ride in on, which is why the temperature-controlled figures above tell the more honest story.

09 — The wind & the stormRough weather keeps the peace

Wind tells the same story as rain, and tells it independently. Holding both temperature and precipitation fixed, each extra 10 mph of peak wind shaves about 2.5 violent incidents off the day (p < 0.001) — yet leaves property crime essentially untouched (+0.2/day, n.s.). A blustery day is an uncomfortable day to be loitering on a corner.

Bundling the rough days together — the 480 "storm days" with heavy rain (>0.5 in) or strong wind (top-decile, ≥20 mph) — violent crime runs about 7% below a calm day of the same temperature.

Step back and a single mechanism organizes the whole report. Every weather condition that makes the outdoors less hospitable — rain, wind, storms — pushes violence down by a similar ~5–7%. Only heat, which makes the outdoors more inviting, pushes it up. Violence in Detroit is, in large part, a function of how many people are outside and in contact with one another.

10 — The long hot spellA heat wave is no worse than its hottest day

If heat fuels aggression, do tempers compound over a prolonged hot spell? We flagged Detroit's heat waves — runs of 3+ consecutive days topping 85°F (the 35 such streaks cover 80 days) — and asked whether being deep into one adds crime beyond what the day's own temperature predicts.

It does not. Comparing day 3+ of a heat wave to an isolated hot day of the same temperature, the difference is a statistically insignificant +1.1 violent incidents (p = 0.64) — and likewise flat for total crime. Crime rises with the thermometer each day and resets with it; the heat does not "bank."

11 — Every category, rankedOne row per offense: heat, rain and snow

The complete picture for all 23 offense categories with enough volume to model. Per +10°F is the temperature sensitivity; Rain and Snow give the change per inch of each (as % of the category's average), both holding temperature constant. Green = more crime, red = less; a small dot (·) flags results that are not statistically significant.

Reading the precipitation columns: most street-facing offenses turn red in the wet (weapons -30%/in rain, disorderly conduct -31%), while burglary stands out in green — break-ins actually rise with rain (+11%/in), the cover of bad weather apparently working in the burglar's favor.