Does the Highway Patrol Keep Us Safe?
Most government programs have some sort of constituency that will fight to the bitter end to keep the money flowing. However, it is probably safe to say that few voters will take to the streets to defend their right to be arrested, fined, and temporarily blinded by a flashlight.
So in these rough budgetary times, maybe the highway patrol would be a good place to start chopping. Virginia and Michigan have already done so, laying off substantial numbers of state troopers, and Illinois recently came very close. Might this be a win-win, saving us both money and aggravation?
On the other hand, the highway patrol exists for a reason: theoretically, by deterring dangerous driving, it’s supposed to keep us safer on the roadways. Does it?
Greg DeAngelo and Benjamin Hansen have written an interesting working paper on this subject that takes advantage of some nice data. They also cite Freakonomics’ own Levitt multiple times, so you know they must be smart.
As Levitt and Dubner discussed in Freakonomics, it’s tough to determine the effect hiring police has on crime rates, due to what’s called the “simultaneity” of causation. We might suppose police deter crime. But ironically, we often see lots of police in high-crime areas, not because police cause crime, but because crime causes police—that is to say, high-crime jurisdictions are likely to hire lots of police officers. So if we want to study the true effect police have on crime rates, we need to somehow factor in the reverse causation.
Levitt has dealt with this problem using an “instrumental variable” approach. This involves finding a variable that is related to the number of police officers but not the number of crimes. He used the election cycle, which is related to police staffing levels (grandstanding politicians tend to hire more police in election years). Presumably, the point in the election cycle has very little to do directly with how much crime is going on, except through its influence on police staffing levels. Using this approach, Levitt and others have found what we would expect: the crime rate goes down in election years, when more police are on the beat.
There is, however, another way to get at this problem—use of an experiment. Ideally, what we’d like to do is take two very similar cities, order one of them to hire more police while forbidding the other from doing so, and then see what happens. While we’re at it we might also want to train their mayors to salivate when we ring a bell, just for fun. But unfortunately we don’t get to do these experiments because urban planners don’t yet rule the world.
But once in a while we get a second-best opportunity, when a city (or state) voluntarily makes a substantial change in the size of its police force, giving us a chance to see what then happened. This is called a “natural experiment,” and one took place in Oregon in 2003.
In that year, the Oregon Legislature mandated an approximately 35 percent cutback in state highway patrolmen for budget reasons. What effect did this have on accident rates?
DeAngelo and Hansen report that in the three years after the layoffs, statewide traffic fatalities rose by 19 percent, incapacitating injuries rose by 14 percent, and visible injuries rose by 12 percent compared to the three year period before. This is especially striking given that, nationwide, the fatality rate per vehicle mile traveled actually fell by 3.7 percent during the period under study.
Other characteristics that might have to do with accident rates—miles driven, precipitation, snowfall, and the number of young drivers on the roads—stayed about the same before and after the layoffs.
Summary statistics of this type can be misleading, so the authors used some fancy econometrics, the details of which I won’t bore you with. The upshot is that, controlling for other factors, for each one percent drop in the number of state troopers, road deaths on highways outside of cities (where state troopers do most of their work) rose about 0.32 percent, incapacitating injuries rose about 0.23 percent, and visible injuries rose about 0.26 percent.
Enforcement appears to have the greatest effect in dry weather conditions, outside of cities and in the summer—that is, precisely when we’d expect speeding to be the most prevalent, and thus where the presence of troopers should have the biggest impact.
Another important effect: after Oregon laid off the troopers, doughnut and coffee consumption in the state dropped 2.3 percent. (Just kidding).
The authors also looked at state trooper levels in Washington state and Idaho from 1979-2005, and—controlling for other factors like mandatory seat belt laws, speed limits and young and old drivers—they found very similar effects. Again, the biggest impact was outside of city limits and in dry weather.
Counterfactuals are always a tough business, but according to the data and the authors’ calculations, if Oregon state trooper levels had stayed at their 1979 levels (in terms of officers per mile driven on the highways), about 3,800 road fatalities might have been prevented between then and now. That’s about one and a quarter 9/11s.
Granted, hiring more troopers would have been expensive—in Oregon, putting one in the field costs about $100,000 per year. And Oregon would have needed 1,159 patrolmen in 2005 to match 1979 per-mile-driven levels, as opposed to the 250 that were actually working in 2005.
But the investment would have been worth it: the authors find more patrolmen would have cost about $320,000 per life saved, a real bargain given that economists generally value human life much higher than this. And those savings don’t even count other kinds of injuries and property damage that might have been averted.
This suggests that cutting state troopers is penny-wise but pound-foolish. I know cash is tight these days, but if we can find tens (hundreds?) of billions for a program that is penny-foolish and pound-foolish—high-speed rail—we surely should be able to find a way to keep our roads decently patrolled. Using the $100,000 figure, the $45 billion (optimistically) we’re planning to spend building just the California portion of HSR is enough to support the entire force of about 7,300 California highway patrolmen for over 60 years (and plenty more, assuming a decent rate of return on capital, since the savings for HSR would come now and the troopers would be paid over many years).
In my opinion, it’d be better to keep our state police budgets intact—even if it means we all have to spend plenty of time practicing our stories about being late for flights and having medical emergencies—than take the chance that many more people will die on our roads.