Can Openers for Beginners

We haven’t had a Beginners post in a long time, but David Kestenbaum’s Planet Money post about traffic court got that part of my brain going again.

Kestenbaum’s story is that he went to traffic court and the judge was a friendly populist, but not an economist:

“The judge went on to say that this was the “people’s court” and explained that if she gave probation on a ticket, no points would appear and the insurance companies wouldn’t find out. ‘This court is not in the business of enriching the insurance companies,’ she said.”

Aha! Kestenbaum, who after a year on Planet Money is an economist, even if his Ph.D. is in physics, points out that whether or not people get points on their licenses doesn’t affect insurance company profits. They need to charge bad drivers more because their loss payments for those drivers will be higher; if they can’t find out who the bad drivers are, they will just raise premiums on everyone.

You can take this line of reasoning further.

  • The first-order effect of hiding information from insurers is that those drivers will enjoy lower premiums than they otherwise would have paid.
  • The second-order effect is that insurers will raise premiums on all drivers to make up the difference.
  • The third-order effect is that fewer people will buy insurance. The reason is that the value of insurance is different for different people; if you are a good driver, insurance is worth less to you than if you are a bad driver. (Let’s leave aside the very real problem that most people think they are good drivers, even though they aren’t.) If the price curve for insurance flattens out (same price for all levels of riskiness), then the price for good drivers will exceed the value of insurance to them, and they will elect to “go bare” (that’s the term used in the commercial insurance world, honest). This is bad because if fewer people are buying insurance, then the economy is producing less than the socially optimal amount of insurance.
  • The fourth-order effect is adverse selection. If the good drivers drop out of the pool, the insurance companies know this (they will find out after a year when the accident rates are higher than they would be if all drivers were in the pool), and they will raise rates for everyone left. Now the middling drivers are being overcharged for insurance (the bad drivers are still being undercharged), so then they will drop out, and after another year the insurers will raise rates again. Theoretically this could go on until only one person is being insured, though in practice it obviously doesn’t.

So by this line of reasoning it’s a very, very good thing that bad drivers get points on their licenses and insurers use that information to set premiums.

There are a couple of other angles you can take on this scenario. For example, if only one traffic court judge is behaving this way, then her customers are getting a break, and what you have is a transfer of wealth from every other insured driver in the country to the people in her courtroom.

There. Now you know how to sound like an economist at a cocktail party.

But like much of this kind of economistic speculation, which you will find all over the Internet, including here (hey, at least I’m aware of my shortcomings), is that it rests on a host of implicit assumptions. One of the big assumptions is perfect competition. Perfect competition is what makes things “balance out” all the time. Insurers can’t simply fix premiums arbitrarily high above everyone’s willingness to pay for insurance, because then each individual insurer would have a marginal incentive to lower premiums until all of the profits were competed away. Perfect competition lets you say that more or less nothing will affect an industry’s profits.

But there is no such thing as perfect competition. If you consider a world where some companies happen to be better than others, then things get more interesting. You could argue that giving insurers information about driving violations (points) helps the better insurers, because they will be able to use that data more effectively in their pricing models. You could also argue the reverse: state-assessed points are an easy way for all insurers to get data, and hence they level the playing field; if they couldn’t get that data, then the best insurers would find other ways to predict the likelihood of accidents (credit reports, for example, although that is highly controversial and I believe illegal in some states), which would give them more of a competitive advantage. Without assumptions, things get messier.

Another common assumption is no regulation. In all states, however, drivers are required to buy liability coverage (though not collision and comprehensive), just like in most states companies are required to buy workers’ compensation insurance. When there is a legislated captive market, the third-order effect above doesn’t apply, and the adverse selection problem becomes more complicated (it doesn’t go away as long as people can choose different attributes of their policies, like limits and deductibles).

Then of course there is the assumption of rational actors. Because everyone thinks he or she is a good driver, people underestimate the value of insurance. Or at least, they would underestimate it if they actually thought about it. But they don’t actually try to calculate the value of insurance to them and make a buying decision on that basis – they just buy it from whoever offers them the lowest price for some package of coverages they decide they need based on certain rules of thumb, like “whatever my parents told me to buy the first time I had a car.” Without rational actors, most of this type of cocktail party economics breaks down – but where’s the fun in that?

So now you should be familiar with a large proportion of the arguments made by economists. Extra credit to those who can explain the title.

By James Kwak

30 responses to “Can Openers for Beginners

  1. “Assume a can opener” ?

    Of course, the obvious solution is to forbid insurance companies from charging different rates to different drivers (regardless of driving record), then pass a law requiring everyone to buy identical insurance, then provide a “public option” to ensure some competition for the private insurers.

    Problem solved!

  2. OK, Nemo got it. I knew it was too easy for this crowd.

  3. Please get me a one-handed can opener.

    Of course, getting sick is more of a volitional act than crashing a car, so it would make sense to ban price discrimination in car insurance and preserve it in health insurance. Plus, it’s easier (and has more of an effect on our lives) to shop for healthcare after a heart attack than it is to shop for auto repair after an accident, reinforcing out first conclusion.

  4. The presumption here is tickets have something to do with being a bad driver. I think relying on actual costs is probably much more accurate.

  5. Of course the courts don’t want to enrich the insurance companies, just the state. Based on all the economic and mathmatical knowledge possessed by Dr. Kwak, Dr. Johnson and many of this boards contributors – notably Stats Guy – I wonder what percentage of traffic tickets are actually given out based on a desire for public safety and what percentage are given out to generate revenue. My uneducated guess is the ratio is somewhere around 20/80.

  6. I was sorta thinking along Lord’s line. Consider that on any given day almost all drivers could be given multiple speeding tickets. In going to work I travel 70-75mph on the highway when it is clearly 65mph. Even in the city streets getting to the highway I try to go the 45mph limit but sometimes I am going 50mph before I notice. So I don’t know that getting caught for speeding amounts to the obvious increased risk. Of course traveling 100mph is different. But is the bar too low for the criteria of speeding tickets = more prone to crash.

  7. Here’s a paper which would imply that there isn’t such a strong correlation between moving violations and accidents.

    http://deepblue.lib.umich.edu/bitstream/2027.42/33862/1/0000123.pdf

  8. Based on Richard Posner as the example, I figure judges don’t know diddly-squat about economics anyway.

  9. What about the people who know they are good drivers, but are not sure about everyone else? Without the mandate, those with high value cars, high value lives, or high future value on ready access to personal transportation would insure anyway, but would end up paying a premium because of the high likelihood of having to pay out due to a collision caused by an uninsured motorist.

    I guess economics may be too complicated for cocktail parties after all.

  10. “the judge was a friendly populist, but not an economist:”

    Well, who is both, these days? ;)

  11. “Well, who is both, these days? ;)”

    You could just see this as tacit acknowledgment that there’s no correlation between the costs insurance companies bear for a particular driver and the number of tickets that driver receives.

  12. Perhaps insurance companies are aware of evidence that the information we get from tickets is swamped by the fact that traffic stops are correlated with something besides bad driving. Take Lamberth (1998), writing in the WaPo “To summarize: African Americans made up 13.5 percent of the turnpike’s population and 15 percent of the speeders. But they represented 35 percent of those pulled over. In stark numbers, blacks were 4.85 times as likely to be stopped as were others.”

    Somehow I doubt blacks were 4.85 times more likely to smash their cars.

    Depending on the district, it might really be very populist to keep insurers from getting ahold of traffic stop information, as the cops are already doing what they can to muddy the data in ways that transfer income the other way.

  13. You’re assuming two things — one that tickets correlate highly with accidents, and the second that insurance companies raise rates in proportion to this correlation.

    It is perfectly possible that insurance companies raise rates on drivers with points in excess of the risk they pose — such drivers may be a more captive and profitable segment for insurance companies than drivers with a good record … in which case, both the judge and you have a valid point each.

  14. We’re missing the classic moral hazard here… that is, without punishing bad _behavior_, we have less incentive for people to try to behave better. Given our recent experience with moral hazard in financial services, this deserves some mentiond too…

    This is fundamentally different from the selection problem (some people are just better/worse drivers, given the same level of effort).

    I suppose my issue with the judge are twofold – are personal decisions and effort really impactful in determining road hazards? Well, texting-while-driving increases risk of accidents by 23 times (in some studies)… http://txtresponsibly.org/ I suspect many other behaviors are impactful as well, so I would think yes.

    The question is how many people being ticketed are just unlucky (in which case they are being punished with higher rates unfairly) vs. worse drivers (in which case they are being charged more for insurance to cover inherently higher risks) vs. really deserve it (aka, putting other peoples’ lives in jeopardy due to avoidable bad behavior).

    So an efficient system will try to discern this, and one way it does this is by not penalizing for first/minor offenses that are possibly unavoidable.

    For people that are just dangerous drivers, no matter how hard they try, the system needs to balance their needs/rights vs. the dangers to other people. Should an 80 year old with failing vision be allowed to drive? (http://www.sfgate.com/cgi-bin/article.cgi?f=/c/a/2003/07/17/MN249526.DTL) Not an easy issue.

    But what about driving 80 in a 45 mph zone? Well, if you’ve ever had a loved one hurt (or worse) in a car accident, you may be less sympathetic than the populist judge.

    The classic economic problem is that any attempt to strengthen incentives increases wrongful punishment (harsh punishment to people who really were trying to drive safely) and increases the insurance premium spread between people who are inherently bad drivers vs. inherently good drivers through no fault of their own (e.g. genetics or age or disability etc., which people have no control over).

    Weakening the incentives creates the problems described above.

    So what are the ways out?

    – better traffic policing, but this is costly
    – give judges/officers discretion to use information about their perceptions of the driver’s intentions, but this can cause bias and inconsistency (e.g. racial profiling, etc.)
    – get better information, like devices that track general speeding via GPS, but this is an invasion of privacy
    – scare drivers silly by making them think the punishment will be harsh even if it’s not

    Maybe the best solution is legislating better accident prevention technology:

    http://www.sciencealert.com.au/news/20072111-16620-2.html

    Plus we have proximity radar, automatic breaking, and other technologies that are being cooked up.

    But James is absolutely right that economists focus way too much on simple models (without focusing on practical solutions).

    Remember all the talk of the importance of increasing co-pays and deductibles to maintain strong incentives? Well, if people REALLY behaved rationally, then the monetary penalties of higher insurance would not be needed as much because drivers already have a very high copay… that is, the risk of death and serious bodily injury. But people seem psychologically prone to discount daily risks, show overconfidence in their abilities, and fail to balance immediate needs (being late to an event) with less tangible needs (avoiding a low risk of being disabled for life).

    Thus, we have a paternalistic state. And when it comes to driving, thank goodness! Because, let’s face it – cars are death machines, people do stupid things all the time, and that’s a bad combination. Plus, if you were to write down the names of fifty random people you know, I guarantee there are 2 of them that you do not trust behind the wheel without supervision.

    What’s interesting, though, is that the economic rational actor debate immediately focuses the argument on a particular set of problems – moral hazard and adverse selection – which are probably far less important (and less impactful) than technology improvements and simple legislation (like harsh penalties for texting while driving).

  15. Good point: as one retired traffic cop wrote, insurance companies donate speed guns to the police, so they can generate more revenue. Where I commute, you could ticket every second driver for speeding. Does it mean they are all bad drivers??
    Insurance rate should only go up after an accident-at-fault.

  16. Nemo
    except for the public option that’s exactly what we have and still the insurance companies compete
    – maybe they do because foreigners come in and want a chunk of the market maybe they do because the salesmen have to make a living
    – maybe if we had more “choice” in what we cover it may become cheaper but also it would make life a lot more complicated (ask any woman about the ordeal of chosing a lipstick)
    … and last but not least it would make lawyers oh so happy …

  17. Speeding tickets represent what, 90% of traffic tickets? And 30% of accidents are based on speed (which includes inappropriate speed for conditions, such as going 40 on an icy road with a 55 MPH speed limit)

    The one time I was in an accident (yes, an anecdote, which proves little) it wasn’t speed, it was when the other driver ignored the stop sign.

    We used to set speed limits by the rule of 85%, assuming that drivers were usually safe, and setting a speed based on drivers’ behavior. We now set those speed limits based on revenue needs, increase ticketing in recessions, etc…

  18. My same thought exactly. Noting like assuming laws on moving violations as perfectly efficient in terms of signalling likelihood of accidents. Of course all moving violations are created for public safety as opposed to say, oh I don’t know, local and state revenue sources…

  19. We are assuming of course that when municipalities are under fiscal pressure and increase ticket quotas that this is somehow indicative of relative driving skill?

    At least the insurance companies have access to accident data, which is probably a better predictor, as others have stated.

    Given that insurance companies prefer poor quality data to no data – actuaries have to have something to do – I imagine there are a lot of poor decisions being made in the interest of risk management.

  20. Amusingly, the recent book “Do Economists Make Markets”, edited by MacKenzie et al. (2007), has a can opener for its cover as a reference to that joke. The book itself is a fascinating look at new work in the sociology of economics, examining the key role economists (and other knowlede-producers) play in making markets possible.

  21. I like your arguments, but I have always had a problem with the “rational actor” debate for reasons that you state – “being late to an event” vs. “avoiding a low risk of being disabled for life” or dead. I don’t believe we are rational.

    But car companies have inadvertently gotten themselves into a negative “moral hazard” situation because they are constantly making cars safer and thereby enabling us to go faster with less risk. If we really believed in “moral hazard” we’d all be in Pintos. A highway full of cars “unsafe at any speed” will surely decrease the speed limit.

    Submitted for you approval is the fact that highway fatalities over the years have not decreased by any significant amount. http://www-fars.nhtsa.dot.gov/Main/index.aspx If cars are indeed safer, why haven’t fatalities decreased? Because we’re driving faster and more recklessly. Cars, being safer, create a false sense of security which we respond to by taking on more risk (insert your own hedge fund analogy here).

  22. I’m not taking the bait. It’s not statistically significant. Let’s look at the numbers.

    This one judge could have been in a favorable mood. Next time she takes the bench, she might not be so generous. Judges are human, too.

    How many judges are in this county? Are there other “hanging” judges who will hand out points like candy and negate any effects from this generous judge?

    How many traffic judges in the county? How many counties in the state? How many insured in the state? How many insurers in the state? I live in a suburban county in Pennsylvania and the above are so numerous that one generous judge in one county for one day would make zero difference.

    I feel that Adam is baiting us by suggesting that one judge letting bad drivers go free costs us all in higher premiums. Should we get the pitchforks and torches out for this judge? I think not. We would never be able to track the significance of this isolated incident to dollars per wallet. He’s also raising a single example into a story which might be good reporting, but horrible statistics.

    I think I know why there are lots of companies running lots of ads to sell us car insurance – there’s gold in them thar wallets!

  23. Howdy

    I would say that cars are safer now than in the past and the evidence comes from an increase in driving miles while fatalities remain essentially constant. We drive more — not recklessly more — just more.

  24. Very true. Also, population has grown even as fatalities have declined slightly…

    Thus, per capita death rate has dropped. How much of this is technology (vs other factors like more aggressive anti-drinking laws starting in the 80s) is the subject of study, I am sure. Likewise, some “technology” has partly canceled out new safety technology (like larger vehicles which have dramatically increased rollover deaths, and are more threatening to other vehicles as well).

    Yet while these factors have clear and substantial impacts, it’s hard to find any evidence that increase the distribution of speeding tickets and minor traffic violation tickets significantly suppresses fatality rates.

    The evidence (after a cursory search) seems mixed…

    New York, we seem to see an effect:

    http://www.streetsblog.org/2009/07/14/ta-report-reckless-driving-casualties-rising-as-nypd-enforcement-lags/

    Ottawa, we see a _reverse_ effect:

    http://www.thenewspaper.com/news/19/1989.asp

    But in the New York study, I have some face validity doubts. The study argues that we see a 12% decrease in fines for failing to yield at the same time as a 26% increase in related fatalities… But then they note that (on average) a driver could go 1,589 years while DAILY failing to yield before (on average) getting a ticket. So, um, before the 12% decline in ticketing rates they could go 1,806 years between failing-to-yield tickets.

    I kind of have to wonder if drivers even NOTICED the change in failing-to-yield ticketing rates.

    I wonder if anyone has seen a well designed cross-state or cross-jurisdiction study of impact of ticketing rates on fatalities… ?

  25. (a previous post evaporated, so if it shows up and this is a near duplicate, sorry)

    Yes. Also, even though fatality totals have declined only slightly, population has increased, so per-capita fatality rates have declined more. This, in spite of bigger/more dangerous cars (with rollover risks, and risks to “counterparties” in an accident).

    Here are a couple cases from google:

    This Ottawa article claims there’s a _reverse_ effect.

    http://www.thenewspaper.com/news/19/1989.asp

    This NYC article claims a big effect:

    http://www.streetsblog.org/2009/07/14/ta-report-reckless-driving-casualties-rising-as-nypd-enforcement-lags/

    I’m skeptical of the conclusions. For example, they observe a 12% decline in failure-to-yield tickets and a simultaneous 26% increase in failure-to-yield fatalities. BUT, they then notice that at the new rates, people could go 1589 years between failure-to-yield tickets (on average) while failing to yield every day. Um, this would mean that before the reduction in ticketing rates, they could go ONLY 1398 years on average between failing-to-yield tickets even if they violated the law every day.

    So, um… I have to wonder if drivers even noticed the change in ticketing rates.

    But, overall, I think the majority of academic studies do observe an effect in the expected direction. I don’t recall seeing a study comparing the ROI for traffic ticketing efforts vs. infrastructure vs. education vs. new rules against risky behaviors vs. safety technology… (anyone know of one?)

    But for all those pro-regulation folks (me included), it’s worth noting that when Massachusetts recently jettisoned its state-imposed point system, rates dropped substantially for most people. The Mass system for many years was a classic case of regulatory capture that benefited industry at the expense of consumers.

  26. Sorry, Mass jettisoned its state imposed _rate_ system, not the point system.

  27. If you charge everybody in proportion to what their future losses will be, you are no longer selling insurance.

  28. I learned nothing about can openers.

  29. Many states have similar programs for drivers with clean records. Why should normally good drivers be penalized twice, once with a ticket and then by the insurance company?

  30. Escape of ticket-driven points is easy enough – obtain a license from another state. While technically illegal in some states, it’s never enforced, as there is no revenue associated with the license in your pocket. Try driving with an out-of-state license on your vehicle in a state with a high sticker-tax, and that’s a different matter.

    Why did the Dept for Public Safety become the Dept for Revenue Collection, anyway?