The Academic Debate on LoJack - an Actuary Weighs In

LoJack is an anti-theft device for vehicles. Sort of. It emits a radio signal so that a stolen car can be recovered. So it's not so much a theft deterrent as it is a stolen vehicle recovery device. That is, unless you prominently advertise that "This vehicle is protected by LoJack," in which case LoJack protects your car but potentially just causes car thieves to steal a different car.

There was an interesting academic debate on the effectiveness of LoJack on deterring theft. Steven Levitt and Ian Ayers argued that LoJack is a positive externality. It does not provide a specific deterrence against car theft. That is, the thief doesn't know you have LoJack, so having it doesn't effect whether your car gets stolen. But it does provide a general deterrence. That is, if 10% of vehicles in any given city have LoJack, that makes it a lot riskier to be a car thief. It becomes far more likely that your stolen car will lead the police to a "chop shop" and shut down a car theft operation. In this sense, there is potentially a market failure. As a car owner, you'd like, say, 10 or 20% of cars to have LoJack. This would provide a herd immunity of the sort that vaccination provides to populations against spreading viruses. At some level of saturation, LoJack will dramatically reduce car thefts. But everyone has the incentive to shirk. Why bother with the expense of a LoJack when it doesn't protect my car from theft? I could just install the Club (a device that locks the steering wheel). But, like I said above, this just diverts the car thief to another car. Having 10-20% of vehicles using the Club might not reduce total vehicle thefts at all. Likewise, installing LoJack and advertising the fact might not have any effect on total thefts. It just diverts thieves to cars that don't have LoJack. Ideally, you have a lot of cars that 1) have LoJack and 2) don't advertise that they have LoJack. That's a hard sell. You're paying for theft deterrence that goes almost completely to someone else. ("But your stolen car gets recovered!" you might object. Which is true, but your car might have been badly damaged or even "chopped" by the chop shop by the time the police recover it.)

John Lott argued that all this is nonsense in his book Freedomnomics. He starts by basically laying out the positive externality case in the paragraph above.

LoJack would seem to have an overall societal benefit - since criminals won't know which cars are protected, even cars without LoJack should benefit. But, as the argument goes, this creates a problem: you don't install the device on your car, but hope that other car owners will. That way, if auto thieves don't know which cars are protected, you benefit from the overall drop in car thefts stemming from the presence of LoJack on some cars, while only other car owners bear the cost of installing the device. So in the end, no one installs it, because everyone hopes that everyone else will do it.

 So far, so good. So what's the problem with the argument? Lott argues that the "externality" is actually internalized if a single company advertises that all its cars carry LoJack. For example, Porsches could install LoJack as a standard feature, and car thieves will from then on avoid stealing Porsches. He writes, "Rather than having too little of an incentive to install LoJacks, any single company would have too much incentive to do so." True enough, but I think this misses the point about positive externalities. Thieves will avoid Porches, but turn to other brands to steal.

(That's not to take anything away from John Lott. I admire him for making this often-overlooked argument: a single large company or community, acting privately and without the hand of government, can solve a so-called "externality problem". They can "internalize the externality", so to speak. I happen to think that the Porsche example is a bad illustration of this, because it preserves the problem of sending the thieves to victimize other car owners.)

Next he turns to an insurance discounting argument. This is my bread and butter. It's what I do for a living, so my ears perked up when I heard this argument. He points out that there is never a discount for LoJack except in states that mandate such a discount. He quotes an insurance agent stating that such a discount is silly. LoJack doesn't actually reduce the cost of insuring the car, because by the time it's recovered the car is likely to be damaged ("wrecked" in the agent's language). It's a fair point, but I want to make a more quantitative argument.

Say it costs about $1,000 to insure a single car for one year. That's roughly the average insurance premium in many states. About $200 of this premium will be your Bodily Injury coverage (if you are liable for injuring a person/persons), another $200 your Property Damage coverage (if you are liable for damaging property with your car), another $300 for first-party Collision coverage (covers damage to your auto if you get into an accident, and usually when you are the liable party), and maybe $100 of it is "Other Than Collision" aka "Comprehensive" coverage. (That's $800 so far, the rest is lower-dollar coverages: uninsured and under-insured motorist, medical payments, personal injury protection, and various coverage endorsements and features.) Thefts are covered by the "Other Than Collision" coverage, but this coverage covers other things, too. Animal strikes, fires, weather (flooding and hail in particular), and glass repair. Theft represents only fraction of the "Other Than Collision" coverage (say, 10 or 20%, and of course this depends on where you live), which in turn represents maybe 10% of your insurance premium. Let's say it's twenty percent of ten percent of your insurance premium, or 2% of your total premium. An anti-theft device that completely eliminated this risk would warrant a discount of 2%, or $20 out of a $1,000 annual premium. More realistically, an anti-theft device lowers but doesn't eliminate the risk. Say we come up with an incredibly effective anti-theft device that cuts the cost of theft in half. That's a 1% discount, or ten bucks on your thousand-dollar auto premium. If you make monthly payments on your insurance premium, you won't notice the less-than-a-dollar-a-month benefit of installing an anti-theft device. And your insurer might not waste the effort to price this discount. Feel free to redo this exercise with different numbers, but you are likely to get a small answer, whatever you do. The upshot: looking to insurers to estimate the social value of LoJack is barking up the wrong tree.

Lott mentions that car thefts cost $8.4 billion worth of social harm a year (in 2002 anyway), implying that there's a large treasure to be had if only someone could solve the problem of vehicle thefts. Okay, but there were 235 million automobiles in the US fleet in 2002. So the cost per car per year comes to about $36. One needs to ask, is this a problem worth solving? It may be large in aggregate, but so small to each individual that prevention is more costly than living with the problem. LoJack might be useful only in a few communities with out-of-control car theft problems, and the externality argument might be correct for those special cases. The rest of us might just live with the problem, small as it is.

Contra Lott and contra Levitt and Ayers, LoJack could be extremely effective at deterring theft but still not be worth implementing, either as a government policy or as a corporate policy. LoJack has a one-time cost of $700-$1,000 (by a quick Google search). Say that's $35 to $50 per year over a 20-year usable life of a vehicle (I'm being sloppy about depreciation and discounting here). It's a toss up, but probably not worth the cost except in very high-risk areas. If only 10% or 20% of vehicles need to install LoJack to achieve the "herd immunity", it looks a lot better.

(Why doesn't this $36 number match my $20 (2% of a $1,000 annual premium) in the prior paragraph? Possibly because more expensive cars are more likely to be stolen. And the way it appears in Freedomnomics, "$8.4 billion worth of cars were stolen in 2002", it makes it look like this is the total value of cars stolen. Presumably some of these were recovered, and some damaged vehicles were salvaged (repaired or sold for scrap) so the 8.4 billion is an upper-bound of the cost to society or to insurers.)