« Todd Kendall wishes to know | Main | Andreessen on The Psychology of Human Misjudgment »
Pay-As-You-Drive Car Insurance
The new issue of Democracy: A Journal of Ideas (registration, but easy and free) is very interesting. Here is one proposal, from Jason Bordoff:
Drivers who are similar in all respects—age, gender, driving record—pay roughly the same premiums whether they drive 5,000 or 50,000 miles per year, even though the likelihood of a collision increases with each mile. This “all-you-can-drive” pricing scheme imposes significant costs on society: more traffic accidents, congestion, air pollution, greenhouse gas emissions, and dependence on oil.
...the effect of PAYD on miles traveled and gasoline consumption would be significant: a 6.5 percent reduction under conservative estimates, and others suggest the reduction could be as high as 10 percent. To put that in perspective, it would take an 81-cent-per-gallon increase in the gas tax to achieve a 6.5 percent reduction in miles driven.
Monitoring costs seem workable, at least in principle with computerized odometers, so why don't companies do this?
Posted by Tyler Cowen on March 28, 2008 at 04:03 AM in Economics | Permalink
Comments
Why not just jack up the gas tax? I feel like I've said this before.
Posted by: Paul Gowder at Mar 28, 2008 4:13:41 AM
Aren't insurers already doing this? I know that my Dad in Germany has a car insurance which covers damages only if he drives less than 15000 km/year although I am not sure how they check this.
Posted by: Chris at Mar 28, 2008 4:32:45 AM
Companies do, see www.hollard.co.za
A South African company has exactly this product.
Posted by: Rory Mackay at Mar 28, 2008 4:52:29 AM
I'm insured in the U.S. and also get a discount for driving less than 15,000 miles per year. As with the discounts for being a non-smoker or being female, they don't demand proof, although I imagine they would look more closely if I were to make a large claim.
Posted by: David Wright at Mar 28, 2008 5:10:34 AM
This is common in Europe. Annual driving distance, checked if you have an accident, in which case your insurance payout is limited if your odometer is more than it should be. I doubt if it has any significant impact on the length driven, unless you happen to be close to the limit in both time and distance.
Posted by: Espen at Mar 28, 2008 5:11:14 AM
UK company also does this, with GPS to monitor.
http://www.norwichunion.com/pay-as-you-drive/
Posted by: at Mar 28, 2008 5:15:45 AM
Andrew Tobias proposed this years ago. And he proposed collecting it at the pump with each gallon sold. This would have the effect of gas guzzlers paying higher insurance rates than small cars. I'm not sure that's so inequitable and it would be much easier than a parallel system to collect from odometer readings.
Posted by: Hal Miller at Mar 28, 2008 5:16:47 AM
Could very easily be structured around one's mobile phone rather than the car--quick text of current location (confirmed by GPS?), model of car, etc. would register the driver for the trip at hand and for any vehicle in which the driver wishes to make that trip.
Posted by: J.Lo at Mar 28, 2008 5:33:14 AM
It is not very common among drivers yet, but I can tell you insurance companies here in Italy do offer contracts in the form of PAYD schemes.
An example from England: http://www.norwichunion.com/pay-as-you-drive/index.htm
Posted by: Lamiadestra at Mar 28, 2008 5:51:23 AM
All-State asks how much driving you do on an average week and what the purpose of it is. I don't know if this ends up in the premium's or not, but I don't know why they would ask it if they didn't consider it in someway.
Posted by: Justin Ross at Mar 28, 2008 7:33:55 AM
State Farm asks how long my typical round trip commute to work is. I know they put that into their premium.
Posted by: mike at Mar 28, 2008 7:56:47 AM
I'm in the auto insurance in the U.S. It is done in the U.S. -
Progersive has two patents on the technology and is testing it now in
Wisconsin(?). Americans typically are adverse to having thier insurance
company track them.
Posted by: Dave at Mar 28, 2008 8:09:32 AM
Are we really sure that an 81-cent increase in the gas tax only decreases miles driven by 6.5%? I wonder if the studies estimating that mix cause and effect, since the price of gas is higher when people are driving more in the summer.
In the aftermath of the price shock of Hurricane Katrina, when the price of gas ever so briefly spiked above $4, the roads in Northern Virginia were Empty with a capital E. It seemed like discretionary driving disappeared. Driving on 66 with my Prius was never more pleasurable.
Posted by: Ted at Mar 28, 2008 8:20:49 AM
But the sheer number of miles driven is a poor measure. What matters at least as much is the time, location, and types of roads. Somebody who puts in lots of miles on uncrowded rural interstates during daylight hours in snow-free regions is a much better risk than somebody who puts in far fewer miles on city streets that may be icy during 4 months of the year.
Insurance pricing via GPS tracking anyone?
Posted by: Slocum at Mar 28, 2008 8:37:35 AM
I suspect the measuring problem is deeper than you think.
Driving 5k miles in the city is much more dangerous than 5k miles in the country. Suppose your client lives in the suburbs, how do you determine how much of his/her miles were in the city (with a bunch of cars) versus the suburbs (where there is less congestion)?
I suppose GPS could determine how much congestion you''re driving in. Nonetheless, a computerized odometer will not be sufficient.
Posted by: Scott Wentland at Mar 28, 2008 8:40:54 AM
Ted, do you live in the same Northern Virginia as I do?? There's traffic here at 4AM. I don't remember any magical time of no traffic and find it hard to imagine that could result from gas prices 30 or 40 cents higher than they are now.
Posted by: Cliff at Mar 28, 2008 8:41:39 AM
I'm surprised nobody has brought up privacy issues yet. Andrew Tobias' idea (collecting at the pump), pointed out earlier, though, is a reasonable solution though.
Posted by: Dave at Mar 28, 2008 8:47:23 AM
As mentioned, pay-as-you-drive insurance is available in the UK, though it's uncommon.
I would say there are a couple of problems with it. First, from a practical technological point of view, the tech needs to be fairly inexpensive, reliable and tamper proof. Second, people don't like to be monitored, which is what this system is doing.
These are barriers which can certainly be overcome though. For example, I think one major insurer in the UK is requiring policyholders to install a monitoring device for another purpose - to check whether they drive the vehicle at night. That sort of thing will soften people up to the concept. I wonder if cultural factors don't mean monitoring is more acceptable in continental europe and less so in the UK and US. Though Americans may be more willing to hand over data to companies, if it gets them cheaper driving. I think it's a strong bet that as the technology solidifies, people get used to it, and insurance companies demonstrate it's a good underwriting model, this sort of pricing will become common.
Posted by: Tim at Mar 28, 2008 8:47:51 AM
Note that the quoted paragraph leads with the condition that we are talking about a class defined by a number of characteristics. In other words, the calculation of the premium is a multivariate exercise, with different weights to different factors. Why should miles driven have a large weight (which is really what is being argued).
Posted by: Paul McMahon at Mar 28, 2008 8:49:51 AM
the model should also include population density. rural dwellers that drive many miles on virtually empty roads are also less likely to have accidents. we have many more of these people in the u.s. than europe.
Posted by: oops at Mar 28, 2008 8:52:16 AM
I fail to see how paying a 'flat rate' for insurance results in more traffic accidents et al. People are supposedly paying 'through the nose' for gasoline, and it seems to be having little effect on how much they drive. Why would pay-as-you-drive insurance change people's behaviour more than raising gas prices?
Most insurance companies in North America will ask you the distance of your commute to work and how many miles you intend to drive the vehicle over the year. Longer commutes and more annual miles may mean higher premiums.
In Ontario, Canada, you have to provide your insurance company name, policy number and the odometer reading when your renew your plates (which is either every year or two years). Supposedly, this information provided to insurance companies.
Posted by: Vincent Clement at Mar 28, 2008 8:55:28 AM
oops: The facts seem to state otherwise. In Alabama, between 2000 and 2000 there were 28.9 motor vehicle fatalities in rural areas versus 18.6 in urban areas. The lack of seat belt use in rural areas is a factor in the higher fatality rates. Also speeding is a major factor, especially in rural areas - seems those "virtually empty roads" are a licence to drive very fast.
http://www.arhaonline.org/rururbcomp.htm
http://www.nrharural.org/about/sub/different.html
http://www.ncbi.nlm.nih.gov/pubmed/10634275
http://www.google.com/search?q=vehicle+accidents+rural+versus+urban&ie=utf-8&oe=utf-8&aq=t&rls=org.mozilla:en-US:official
The second link in interesting. It mentions that despite one-third of motor vehicle accidents occurring in rural areas, two-third of motor fatalities occur in rural areas.
Posted by: Vincent Clement at Mar 28, 2008 9:15:51 AM
Nalebuff and Ayres devoted several pages of their book "Why Not?" to this problem -which they classified as "Can you see how the problems of too much driving can be solved? How can we get drivers to feel the pain of driving the extra miles."
After surveying an number of benefits and obstacles to the scheme, they conclude by saying "Few [insurance] companies want to embark on a course that alienates half their clientele. ... So, as long as the low mileage drivers are willing to keep subsidizing the high mileage ones, the incumbents are happy not to rock the boat."
interestingly, their solution to externality problem requires a solution to what Schelling called a mulit-person's prisoner's dilemma among insurance companies. Nobody wants to be first in the field, but everyone would be better of if more insurance companies offered the product.
Posted by: michael webster at Mar 28, 2008 9:34:16 AM
oy, that's silly. Tyler, please recall that insurers are already free to price discriminate across a very broad range of factors, there are a large number of competitors, low barriers to entry and low costs for consumers to switch carriers, a statistically enormous number of customers and the hazard rate is nontrivial.
So as repeatedly noted above, insurers do this to some extent already; further, it's combined with other criteria such as where the miles are driven, type of car, age of driver and the driver's full history including their credit scores. Hypothetically their rating would be better if each insured car provided GPS tracking, full telemetry and dashboard video feeds to the insurer, along with the identity and blood chemistry of the driver and passengers. But, really.
What's really surprising is that this is a proposal to have private auto insurers levy carbon taxes. Nice work if you can get it!
Posted by: misplaced trust co. at Mar 28, 2008 9:37:02 AM
GMAC is offering me a low rate for driving less. Since I have a GM car that has (1) a 12,000 mile/year lease and (b) OnStar technology, they know that I am otherwise incentivized (by the mileage penalties attached to the lease) to drive less than 12,000 miles/year, and they can easily monitor it.
I'll see what kind iof deal they offer me when my current policy is up.
Posted by: bartman at Mar 28, 2008 9:44:07 AM
Even better, I'd like to see this pay-as-you-drive car/gas/insurance/maintenance/parking model get more use.
Posted by: Van at Mar 28, 2008 9:45:15 AM
Drivers who are similar in all respects—age, gender, driving record—pay roughly the same premiums whether they drive 5,000 or 50,000 miles per year,
What does this even mean? If I have the same "driving record" as someone driving ten times more miles than me, then I am getting into crashes 10 times more often. People who drive more miles get into more crashes, and therefore will pay higher premiums.
Why concentrate on a related factor (miles driven) as a proxy for what can be measured exactly (crashes)?
Posted by: Rich B. at Mar 28, 2008 10:15:01 AM
1. The perfect should not be the enemy of the good: measuring miles driven, even if we don't know if it's rural or urban, and discounting by miles is better than not measuring at all and charging a flat rate.
2. Significant problems with a more rational transportation policy are caused by driving's relatively high fixed cost and relatively low variable cost. There are benefits beyond the insurer for mile-based insurance, in other words.
3. No, the 'discounts' you get in this country are not the same thing. I get a piddling tiny discount for barely driving at all - and it's a one-level discount; it provides zero incentive to drive less beyond that number, and it's done on the honor system anyways. The thing people are talking about here is paying x per mile; not having a 5% discount for a leisure vehicle.
Posted by: M1EK at Mar 28, 2008 10:25:50 AM
Rich B., auto insurance companies need a way to gauge the probability of future payout, just as health insurance companies do. A mid-30s female is less likely to require major medical treatments (e.g. cancer treatments, emergency care for heart attack, stroke, etc.) than a late-70s male.
Should medical insurance companies quote rates based only on past medical history, or based on the expected rate of future payout given all variables?
Posted by: meter at Mar 28, 2008 10:44:57 AM
Miles driven do not correlate strongly to damage done to others, which is the main cost of insurance. There's no claim made (in the part you've quoted) that risk remains flat relative to miles driven; a case could be made that the risk per mile should fall with increasing miles driven.
But almost every insurance company I've been with has asked either my one-way commute to work, or my annual miles driven, and I'm sure they've used that information in setting my rates.
Posted by: Anthony at Mar 28, 2008 10:55:54 AM
I've often wondered about this. I drive very little (less than 1/2 of what the average person does), and yet my discount is very small. I'm pretty sure the reason is that insuance companies know that if they refuse to pay claims on the grounds that the insured drove too much and thereby violated the terms of their policy, they would raise holy hell and it just wouldn't be worth it.
Posted by: David J. Balan at Mar 28, 2008 11:32:47 AM
Progressive is experimenting with this. See:
http://newsroom.progressive.com/2007/January/Tripsense-mich-ore.aspx
Posted by: Kevin O'Reilly at Mar 28, 2008 11:33:17 AM
The final plan, of course, should involve pay-as-you-go car insurance that hooks up to GPS, and charges you based on where you drive and time of day, because driving at certain times in certain places carries risks.
In addition, insurance companies could contract with the government to create congestion pricing based on this system.
Posted by: Keith at Mar 28, 2008 11:35:55 AM
Insurers would love to have this information, but the cost has been too high thus far, both in terms of technical implementation, and consumer good will. Americans (and probably many others) despise being tracked, and insurers already face an enormous PR problem, just for being in the business. Just as no one seems willing to trust health insurers with their genetic information, no one is lining up to trust their auto insurer to use real-time GPS data in a fair and honest manner.
Additionally, every state has the right to regulate the business of insurance in that state, as it sees fit. The regulators are no more enthusiastic about this sort of data collection than the consumers, and without their approval, no such system can be put in place.
This is unfortunate, because miles driven is probably the current best technically feasible indicator of the base level of risk the insurer assumes. All the questions insurers ask about usage, miles driven, and other proxies for “amount/type of driving”, only exist to modify a premium that is based on car-months, not car-miles. If actual miles driven could be collected in a verifiable manner, the price could be matched far more closely to the risk.
Posted by: at Mar 28, 2008 12:13:28 PM
Maybe the difference in expected accidents (and more importantly, expected insurance payouts) between someone driving 50,000 miles and someone driving 5,000 miles isn't all that big. Traffic accidents, especially expensive accidents, are rare enough that I don't think us non-experts on the subject have any idea how the relationship works. I'd imagine that driving *patterns* make a lot more difference than total miles, though, and I'm fairly certain that driving records provide a much better prediction than either. In fact, total miles driven might already show up and be factored in as more traffic tickets.
Posted by: Anthony at Mar 28, 2008 12:18:07 PM
As somebody who did actuarial work to pay for life while in college, I can assure everyone here that insurance companies do use mileage to gauge premiums, and have many methods for finding out accurate miles/year which include your self-reported (and checkable in many cases) commute, service reports, etc.
Pay as you go insurance is a very very old idea that is expensive to implement and provides tiny advantages at best. It's also true that not every mile is equally dangerous, so people who choose to pay this way will pay through the nose -- otherwise, the risk is that only the people who drive the most expensive miles will use the service.
There are a million other reasons why this is a bad idea. I'm guessing this guy didn't actually run this idea by any actuaries.
Posted by: infopractical at Mar 28, 2008 12:34:01 PM
Anthony,
Insurers utilize many variables to measure a risk (including all of those you mentioned), with one chosen as the “exposure base” variable, and the rest used as “rating variables”. The insurer will charge a “base rate” for each unit of the exposure base, and this base rate will then be modified, depending on the values of the other rating variables.
In the canonical pricing model, the ideal base has a uniform relationship with the level of risk, and so the selected base variable is whichever best displays this relationship. The current most used exposure base is the car-month, which is largely recognized to be an inferior exposure base compared the car-mile.
Once the exposure base is chosen, the impact of the other rating variables on the amount of risk the insurer assumes is modeled, by one statistical technique or another. The modeling incorporates all of the standard caveats which pop up in multivariate statistical modeling.
Posted by: at Mar 28, 2008 12:41:49 PM
If the insurance companies succeed in getting routinely readable recording GPS units installed in cars, what will they do with information about that modest speeding and rolling through stop signs that almost everyone does from time to time? Will you pay a milage fee and a speeding fee and a rolling-through-stop-signs fee? My instinct tells me that this won't happen at first but some company will sneak such fees into the fine print and then there will be hell to pay. My instinct does NOT tell me what the eventual result will be.
Also, could the GPS unit cost the company lots of money in adverse fault determinations? Suppose two cars, one insured by a company that installs GPS units and one by a company that does not, get into an ambiguous-fault accident. Both cars were speeding modestly, as most people do on that particular expressway, but that fact can only be proven about the car that has the GPS unit installed, so the insurance companies that install GPS units could lose more fault contests than those who don't.
-dk
Posted by: dick King at Mar 28, 2008 1:22:55 PM
Are we using GPS or just the odometer? I wouldn't feel like I was being tracked if it was just the odometer; who knows where I went or when? If it's GPS, expect to read "wife files for divorce after GEICO's GPS shows husband was at brothel/crackhouse." Crack also being a good reason to raise car insurance rates.
Posted by: nick at Mar 28, 2008 1:41:25 PM
Wasn't it a few years ago that one of the car rental companies put data recorders (not GPS) on their cars, and started fining people who speeded? IIRC, there was a massive revolt, and the idea was dropped. Or am I wrong?
Posted by: Bartman at Mar 28, 2008 2:08:25 PM
"Miles driven do not correlate strongly to damage done to others"
This may be true in the aggregate - but once each drivers' base behavior is known (likelihood of reckless driving, let's say), then miles driven most definitely DO correlate strongly to damage done to others.
And, again, the number for "drives relatively little" is still very high - providing a big part of the high ratio between fixed cost and variable cost for driving which is a major component of why people who already own cars aren't willing to leave them at home and use alternative transport. Which, ironically, again works against the insurer - they aren't going to pay out if I take the bus, after all.
Posted by: M1EK at Mar 28, 2008 3:11:57 PM
"I would say there are a couple of problems with it. First, from a practical technological point of view, the tech needs to be fairly inexpensive, reliable and tamper proof. Second, people don't like to be monitored, which is what this system is doing."
GPS technology is already cheap & reliable and could be made tamper proof. And I believe the privacy concerns could be addressed as well. For example, rather than sending the GPS track to the insurance company, it would be possible to pull information about locations and per-mile insurance pricing down to the black box. The black box could then keep a running total of mileage charges without preserving or sending the GPS information to the insurance company. The insurance company wouldn't be able to give you a detailed account of why you'd incurred the charges you did, but nor would it be able (even under court order) to say where you'd been driving (that info would have been intentionally discarded).
Posted by: Slocum at Mar 28, 2008 4:37:25 PM
Yeah, uh, put me down as objecting for privacy considerations. I don't trust my government—regardless of the administration—to know where my car is at all times…and I don't trust my insurance company one iota more with that sort of information.
Posted by: Nathan Sharfi at Mar 28, 2008 5:29:53 PM
Instead of GPS the car could just track distance, acceleration, and speed. That way the insurance companies would still know if how much is on the high way (fast, constant speed), and what's in cities (lots of stop-and-go).
I think this is a great idea. I'm a male college student and therefore in the highest bracket for anyone without a crash or moving violation on their record. However I ride my bike whenever I can to school and work so the insurance company recognizing that in the form of a discount would be great.
Posted by: Nick M. at Mar 28, 2008 8:42:42 PM
"This may be true in the aggregate - but once each drivers' base behavior is known (likelihood of reckless driving, let's say), then miles driven most definitely DO correlate strongly to damage done to others."
OK, but how well can insurers determine a good "base behavior"? Without GPS-tracking their every movement (which is not the same as just checking the odometer reading for the mileage), I'd say not very well. At least not in the case of a driver with a clean driving record. In fact, the more miles a person drives while maintaining a clean driving record, the more likely they are to be a good driver.
That's the problem. I'd expect that most of the cost of an insurance policy, even for a driver with a clean driving record, goes to the cost of bad drivers. My property damage liability insurance of $145 for 6 months, on a male 32-year-old with a clean driving record, which pays out up to $10,000 in property damage with no deductible, implies that I have a 1 in 35 chance of totally destroying someone else's property each year. I don't think my real risk is anywhere near that high, whether I drive 1000 miles a year or 40,000 miles a year. It seems quite obvious to me that the insurance company has no clue what my true "base behavior" is. Rather, they are lumping me, in aggregate, with many other drivers with greatly varying "base behaviors". If that assumption is true, then it's easy to see that miles driven might not be strongly correlated to damage done to others.
Maybe the idea would be both more effective and less controversial if it was only applied to people with known, poor, driving records.
Posted by: at Mar 28, 2008 8:58:35 PM
Perhaps you are a better driver than I am, but $10,000 in property damage is nothing. The average price of a new car, in the US, is $30 grand.
Posted by: adam at Mar 29, 2008 12:23:32 AM
$10,000 is all that's required under the law, and the law is the only reason I have the insurance in the first place.
Posted by: at Mar 29, 2008 8:32:15 AM
"OK, but how well can insurers determine a good "base behavior"?"
The same way they do today, for good or for ill. Nobody's talked about any changes on that account; just providing some additional information which currently lacks, as in "today I'll ride the bus to work - so why on earth should I be paying as if I drove? The insurance company doesn't have to pay if my bus gets hit by a car".
Once again, don't make the perfect the enemy of the good.
Posted by: M1EK at Mar 29, 2008 9:52:13 AM
Pay at the pump for gas insurance has a couple of real benefits. 1) You no longer have un-insured drivers 2) Car sharing programs and van pools have significantly reduced costs.
These are both social goods. The bad news? Insurance companies are pretty much out of business if everyone pays their premium with their fill-up.
There was a trial of this in Texas a few years back. I never tracked the data beyond the fact that it wasn't going to be real for our car sharing non-profit. So... pointless except in a dreamable society where people make decisions for the long-term health of the society.
Go figure.
Posted by: Mud at Mar 29, 2008 11:39:32 AM
The real problem with pay-at-the-pump is that the rate is the same per mile for safe drivers and for those who only use their cars for bar-hopping at night.
-dk
Posted by: Dick King at Mar 30, 2008 11:40:22 AM
I do a form of Pay as You Drive. I have several cars and don't drive all of them very often. So I cancel the insurance and renew it the days I want to drive. With online insurance service this takes 5 minutes a day to save about 2/3 of my insurance bill. A miles based system would be easier, but doesn't seem to be available.
Posted by: Nathan Whitehead at Mar 30, 2008 2:18:58 PM
"The perfect should not be the enemy of the good: measuring miles driven, even if we don't know if it's rural or urban, and discounting by miles is better than not measuring at all and charging a flat rate."
But what if it's only *slightly* better, and the costs of implementing it outweighs the benefits? What are the benefits, anyway? To the insurance company, there isn't a direct benefit. They collect the same amount in premiums, it's just spread out differently. So it's only the consumer that'd get the benefit (and the insurance company would thus get the benefit of higher demand vis a vis its competitors).
I guess it makes sense for cars with extremely low mileage, which is probably a small fraction of the insured autos. For the rest of us, I really don't think it makes sense to go through the hassle of getting our odometers certified every year.
On the other hand, a GPS tracking device which could measure things other than just mileage, like time/location/speed, that'd probably make sense for any car whose primary driver pretty much always obeys traffic regulations. I'm not sure if that's a small fraction of insured autos or not. There's also the urban canyon problem. But that's significantly different from just paying per mile.
Posted by: at Apr 1, 2008 7:39:37 AM
Nice to meet you.
I had a look at blog.
Please link to this site.
http://www.geocities.jp/bom_2_08/
Posted by: bal at Apr 3, 2008 5:35:48 AM
Nice to meet you.
Your site was interesting and had a look at it.
Come to take my blog good.
Link if you like it.
Thank you.
⇒ http://analysis-of-the-insurance.blogspot.com/
Posted by: dual at Apr 4, 2008 3:56:53 AM
Why not just offer additional auto insurance policy option
selling insurance for distance driven(5000 or 10000 miles) instead for time (6 month or 1 year). All risks can be calculated by mile the same way as by time considering age, location and so on. People concerned about privacy
just have to buy old style insurance. I believe new product will be popular in the cities with public transportation, high traffic and expensive auto insurance.
Posted by: John at May 14, 2008 8:53:15 AM
