Why Deer-Crash Maps Lie

It is one of the most repeated shortcuts in deer analytics: to find the deer, map the car crashes. The logic feels airtight: more deer, more collisions. In Iowa's own records it barely holds. Across 99 counties the correlation between deer harvested and animal-vehicle crashes is 0.306 — weak. The reason is arithmetic: a crash requires a deer and a car, and the number of cars varies far more across counties than the number of deer. A raw crash count is a traffic-exposure index wearing a deer costume, and it systematically points hunters at the wrong ground.

The idea is intuitive and cheap. Deer-vehicle crashes are logged by police, tabulated by state DOTs, and free to download by county — a ready-made proxy for where the deer are, no hunter survey required. Insurance companies publish state rankings on it every autumn. The trouble is that a collision is not an observation of a deer; it is an observation of a deer meeting a vehicle. Change the number of vehicles and you change the map without changing a single deer. That is exactly what happens across a state, and it is why the proxy fails where it is asked to work hardest — telling a rural deer county from a suburban one.

We can put a number on the failure. Summing Iowa's registered deer harvest and its animal-vehicle crashes by county over 2015–2024 and correlating the two gives r = 0.306: the two share under ten percent of their variance. If crashes tracked deer, that figure would sit near one. It does not, and normalizing crashes by county land area makes it slightly worse, because area is not the missing denominator. Traffic is.

Set Iowa's premier deer county beside its capital county. Clayton, in the driftless northeast, is the state's harvest leader — bluff country, timber, and river-break agriculture that grows and holds deer. Polk County is Des Moines: a metro of highways, arterials, and commuters. Over the decade Clayton hunters registered 42,354 deer to Polk's 10,415, more than four to one. Yet Polk logged 2,405 animal-vehicle crashes to Clayton's 1,401. A crash map hands the crown to Des Moines. A harvest map, the measured one, hands it to Clayton. Only one of those is about deer.

A crash needs a deer and a car. Across counties the cars vary far more than the deer. A raw crash map is mostly a map of where people drive.

The distortion is not a Clayton-versus-Polk quirk; it runs through the whole ranking. Order Iowa's counties by raw animal-vehicle crashes and the top of the list is a tour of the state's traffic: Cedar Rapids, the southeast river corridor, Des Moines, Dubuque, the Quad Cities, Ames. Order the same counties by registered deer harvest and an entirely different Iowa appears: the driftless northeast and the southern-border timber. A hunter handed the crash map is pointed straight at the interstates and away from the ground that actually produces deer.

The repair is to restore the denominator the raw count throws away. A crash rate (collisions per unit of travel, per vehicle-mile) nets out how much driving a county does and leaves the part that actually varies with deer. Only a rate is rankable as-is. In our data the deer-crash series that qualify are the ones states publish already normalized to travel; Illinois carries a deer-specific rate spanning three decades, and it is the series we treat as a signal. Raw counts we keep as counts, flagged, and never rank.

One more distinction hides in the species label. Iowa and Ohio do not publish a deer-specific crash code at all — only 'animal-involved' collisions, a category roughly ten percent broader than deer alone. The Iowa figures above are that animal-involved series, correct for what they are and never silently pooled with a true deer-specific count from another state. When the underlying series measure different animals on different denominators, stacking them into one leaderboard manufactures a comparison the sources cannot support — which is the same mistake as the crash map itself, one level up.

The practical rule is short. A raw crash count answers 'where do deer and cars collide most,' which is mostly a question about cars. If the question is 'where are the deer,' a crash map is the wrong tool unless it has been turned into a per-exposure rate. Even then, check whether it counts deer or every animal on the road. The measured harvest, county by county, remains the cleaner answer where a registration state provides it.