In the previous post in my mini-series on insurance fundamentals I provided a numerical example that has a 1 in 100 probability of a loss of $50. I explained how the fair rate to insure this would be 50 cents, because that’s the amount at which the expected financial loss equals the premium payment. I then walked through how even at a significantly higher premium some amount of reinsurance was still required to keep this scheme solvent (so far nobody has taken me up on the homework assignment in that post!).
But that is not the only challenge in coming up with an insurance scheme that actually works. Another one carries the somewhat odd name of “Moral Hazard.” Let’s consider the source of the 1 in 100 probability of a $50 loss. In many real world insurance examples both the probability and the size of the loss are at least partially influence by human behavior. For instance, the likelihood and severity of a car accident depends on on your driving speed. If you drive recklessly fast you have a higher likelihood of a serious accident.
Now that raises an interesting dilemma: if you take out car insurance you may wind up driving faster than you would if you had to pay for the full damages yourself. That change in behavior after obtaining insurance is called “Moral Hazard” — the term apparently dates back to the 17th century and may have carried an actual moral connotation, meaning that it was considered immoral to change one’s behavior that way. Today we simply tend to recognize that this type of behavior change occurs as a result of insurance and more generally when actors do not carry the full risk of their actions. It is important to note that this is entirely different and separate from fraud, where someone claims exaggerates the size of a loss in their insurance claim (or files a claim without a loss having occurred in the first place).
The problem with moral hazard of course is that it needs to be priced into the insurance. If after taking out insurance the risk goes to 1.2 in 100 and a $60 loss, then the fair rate would have to be 1.2 / 100 * $60 = 72 cents (up from 50 cents). Also with the increased frequency the rate for the reinsurance would have to rise because now there is a higher probability of multiple losses occurring simultaneously.
One commonly observed feature of insurance plans that is aimed at reducing moral hazard is the deductible (it has other reasons for existence as well that I will get to in later posts on fraud and the cost of claims processing). The deductible is the amount that you have to pay yourself – sometimes also called “out of pocket” for that reason – before the insurance company pays anything. The idea is that since you are risk averse to begin with, after all otherwise you wouldn’t buy insurance, the deductible provides an incentive to keep moral hazard small.
We can look at moral hazard as a problem of information asymmetry. If the insurance carrier could perfectly observe your behavior after you take out insurance then it could address moral hazard in other ways. For instance, car insurance could charge a higher rate only to drivers who go too fast. In the extreme the insurance company could control the behavior of the car and not let it go above a certain speed. For instance in a possible future of only self driving cars the problem of moral hazard could be eliminated entirely.
In the absence of such information the insurance rate is designed to account for the average degree of moral hazard. This means that there will be a wealth transfer from the drivers who drive safer than the assumed degree to those who drive more recklessly. Some people mistakenly think that such a transfer is a necessary feature of all insurance. It is simply the result of insufficient information.
Moral hazard doesn’t just arise in the obvious places such as car insurance. Any time risk is shifted away from an actor whose decisions impact that risk there will be a distortion. For instance, mortgage originators often sold off the mortgages shortly after originating them. This leads to moral hazard in the diligence process. Again that’s different from outright fraud such as changing numbers on a mortgage application. I am simply talking about the effort that goes into verifying the information that has been provided by the applicant.
The next post will be about another problem of incomplete or asymmetric information in insurance know as adverse selection.
