When economists notice that nobody sells insurance for X, they have a standard explanation: adverse selection. Why can’t you buy flood insurance? Supposedly, because the highest-risk people will be first in line to buy it, which raises premiums, which encourages low-risk people not to buy, which raises rates further, and so on. The market “unravels.”
This all sounds good on a homework problem, but it doesn’t fit the facts very well. Do homeowners around the country really know more about their flooding risks than the actuarial profession? It is particularly hard to stand behind the adverse selection story when insurers could condition rates on the best meteorological data, the age and elevation of the home, distance from water, and other simple statistics.
It is also worth mentioning that a number of recent empirical studies find that selection is more likely to be advantageous than adverse. Low-risk people often buy more insurance than high-risk people! Check out Cawley and Philipson on life insurance and Chiappori and Salanie on auto insurance.
What’s the real story? Here are a few ideas.
- Adverse selection runs the other way. If you got an offer for flood insurance in the mail, maybe you’d figure “They only offer it because it isn’t worth it.” I don’t find this too plausible, as long as the issuer is already reputable.
- The insurance just isn’t worth it. It costs a lot to fix a scratch on your car, but most of us can live with a scratch. Maybe flood damage is similar.
- Consumers underestimate these risks, so they mistakenly conclude insurance isn’t worth it. The problem here is that the uninsured risks (like flood) seem to get more news coverage than the insured risks (like auto accidents). So if anything you’d expect the bias to go the other way.
- People figure that the government will bail them out if there is serious damage, so there is no need to get their own insurance.
I don’t know which, if any, of these alternative theories are right, but I suspect it’s mostly a blend of #2 and #4. What I am pretty sure of, though, is that “adverse selection” is more a convenient slogan than a convincing explanation for missing insurance markets.
READER COMMENTS
Jim Glass
Sep 24 2005 at 10:36pm
In the case of long-term care health insurance — or lack thereof, which causes people to lose their homes to nursing home and similar costs — and which is seriously underused, data show the primary answer clearly is a combination of #3 & #4.
People greatly underestimate the risk they face (perhaps due to psychological reasons: “I’m not going to be suddenly disabled for years to come”), and most people to the extent that they do think about it mistakenly assume Medicare covers it.
There are also complications in determining the value of this insurance, as its value can depend on one’s financial situation many years later, which one can’t know for sure in advance. That’s #2. But that can be handled by anyone who takes the insurance seriously enough to get that far — which most don’t, because #3 & #4 keep them from giving it much thought at all.
Bill Stepp
Sep 25 2005 at 7:02am
You are right that adverse selection (a k a antiselection) does not provide a theory of insurance pricing. Economists might think it does, but that would be news to actuaries and underwriters, who apply the law of large numbers to statistics of specific perils (mortality for life insurance, morbidity for disability, accidents for car, etc.) to determine how to price their products. Adverse selection is factored in by requiring insurance applicants to meet underwriting procedures and standards, which screens out unacceptable risks and prices acceptable risks correctly.
Cawley and Philipson find that people buy more policies of smaller size than predicted by the “canonical theory,” but there is a reason for this. They buy insurance over their life cycle to meet specific needs that arise along the way. For example, Mr. Jones gets married so he buys his first life policy. Then Junior is born so he buys a second policy. Junior’s sister later makes her appearance so Mr. Jones buys more. He later leaves his firm to start a new business with a partner, so they purchase insurance to fund a buy-sell agreement. As the business grows they buy more. Years later he and his wife buy insurance to pay the estate taxes on their estate. (Which is why the life insurance lobby is against repealing the estate tax.)
If economists spent more time in the empirical world and less time in the ivory tower, they might figure this out and avoid the errors that crop up in their “canonical theories.” They could also read an insurance text book, such as Harold Skipper’s classic work on life insurance, or the short piece by the life insurance agent David Hilton Jr., “How Insurance Works.”
Jon
Sep 26 2005 at 7:11am
Auto insurance — the highest risk drivers tend to be young and live in low income parts of the city. They also have less to lose and are less able to afford auto insurance. College and grad students often purchase the minimum–they have no assets to go after. Because the qualities that make one a safer driver tend to give advantages in the job market. Low income=less to lose in a liability case = less available to pay insurance costs.
Sol
Sep 26 2005 at 9:22am
I absolutely use the “They only offer it because it isn’t worth it” metric when Best Buy offers me their laptop computer “insurance”. It’s just common sense — they must be expecting to make money on the deal, else they wouldn’t offer it — so the expected return on the deal is clearly less than the cost of it. (I suppose there might be some quibbling on them being able to replace parts at cost rather than at your price, but I’m pretty sure the overall point is sound.)
On the other hand, the reason I can do this is because I can afford to replace the laptop if needed. On the other hand, potentially a health crisis could destory my finances if I have no health insurance. Ergo, I do buy health insurance.
It seems that flood insurance might fall in a grey zone between trivial and devastating, with the exact risk involved depending on your circumstances.
Hmmm… how do banks handle it if your house is flooded before your mortage is paid off? Is the place just a write-off then? Or do they require you to have flood insurance if floods are likely?
Dennis Gildea
Sep 26 2005 at 11:20am
One issue specific to flood insurance: underwriters want to feel confident that they can estimate the frequency of future floods, but much of the literature and publicity about global warming suggests to them that historical data on frequency of weather-related events may be unreliable. So they invest in areas where they feel more confident that they can set a price which will return a profit.
rvman
Sep 26 2005 at 2:31pm
The existence of federal flood insurance probably crowds out any private market. The essentially infinite assets of the guarantor (The US Government) makes that insurance attractive to mortgage providers who require it specifically in flood-prone secured loans. What is being sold with insurance is the ability to use a mass of existing assets and consolidated premiums to bring the insurer closer to risk-neutral than the customer.
Floods are difficult because losses tend to be geographically clustered, so an insurer can get whalloped even if he has a large customer base, if that base is all caught in the same flood. Property insurers will often quit selling in certain areas because they already have too many customers there, and are big-loss prone. (For example, several property insurers quit selling homeowners for a while a few years ago in Fort Worth, TX, because they had too much risk on the books in the case of big hail.) Bigger insurers can reach closer to risk-neutral than small ones, reinsurers (insurance for insurance companies) can get closer still, and government can get closer than all.
Government insurance is probably underpriced, anyway, because its premiums probably do not account for improbable high-cost events, which would be bailed out through tax dollars anyway. It is FEMA, not flood insurance premia, that the US government will lean on to bail out New Orleans.
R.J. Lehmann
Sep 27 2005 at 11:23am
The problem with flood insurance, historicall, is a combination of 2&4. From an insurer’s perspective, the peril is a troublesome one, as it tends to cause a large number of total loss claims, with the additional problem of those losses being geographically clustered. Thus, actuarial rates for flood coverage would tend to be extremely high, and offering the coverage as part of an all-perils homeowners policy would mean sacrificing a good deal of premium from potential policyholders who would otherwise represent perfectly attractive risks if one didn’t have to worry so much about exposure to excess aggregation.
Given the price controls imposed by state regulators on insurers, getting those actuarially appropriate rates approved wouldn’t be an easy task to begin with. But even if it were, the rates would tend to be so high that they just simply aren’t market-clearing, and many homeowners would rather eschew go barren than pay the premium. Hence, insurers long ago began excluding the peril.
This trend was exacerbated by Good Samaritan problems in the wake of disasters. Governments would respond to major floods by paying to help homeowners rebuild their properties, thus further crowding out private insurance. In fact, the flood insurance program was created to try to at least begin to manage the moral hazards of government response and force at-risk homeowners to bear SOME of the costs of rebuilding upfront.
Scott Peterson
Sep 30 2005 at 4:30pm
With respect to flood insurance, the length of time between major floods in most parts of the US is greater than fifty years. Given that the longest mortgages are thirty years, and the fact that most homeowners do not stay in the same house for fifty years or more(probably 99% move at least once in thirty years), the expected value to an individual homeowner of a flood insurance policy will be very low.
Because the frequency of major floods is so low the homeowner will likely underestimate the (already low)risk of losing their home to a flood. The flood protection preparations in the Gulf Coast before Katrina are a great example of this. Since the last major hurricane had been Camille in the mid-sixties, no one had much comprehension of what a major hurricane would be like and clearly the communities of the Gulf Coast grossly underestimated the risks to their property.
Scott Peterson
Sep 30 2005 at 4:43pm
With respect to Bill Stepp’s comments, insurers are in the business of providing protection against risks where the distribution of policyholders who file claims fits to a normal distribution or “bell curve”. Most insurance policies have a clause that excludes events such as “acts of God” where the frequency of occurrence cannot be shown to be normally distributed. “Missing markets” generally would fall into the category of risks with non-normal distributions.
Related to this, Dennis Gildea is on the right track with his comment about the uncertainty regarding weather events. Reliable meteorological data in the US go back only about one hundred years; thus, due to the small sample size of weather data we can not make statistically meaningful estimates of the true probability of hurricanes, floods, noreasters and so forth.
Comments are closed.