Why your Apple Watch and your cheap car insurance might be a threat to the welfare state?


A few years ago, car insurance companies started to offer so-called “Pay-How-You-Drive” policies that directly link a customer’s premium to their driving behavior. These insurance policies were rolled out once it was possible to observe and report individual driving behavior through GPS-enabled devices (including mobile phones). These trackers can record and transmit – in real time – information that has actuarial relevance for the insurer. Examples include kilometers driven, time of day, speeding (absolute, and relative to speed limits), acceleration, breaking events and their harshness, swerving, and so on. While seemingly inoffensive, this essay argues that similar technologies could, in the medium term, severely threaten the political foundations of the welfare state.

Pay-How-You-Drive (PHYD) policies that couple insurance rates to individual driving behavior are attractive for insurance companies because they allow them to use fine-grained classifications of their customers’ risks. This contrasts with traditional, static measures that are used for conventional risk classification – such as drivers’ age, their occupation, place of residency, or car model – which group customers in broad bins of risk classes. PHYD products, therefore, are actuarially more accurate, tying drivers’ probability of causing an accident more closely to the insurance premium they have to pay. That is why PHYD insurance appeals to safe drivers (“good risks”) – the customers insurance companies hope to sign up.

The ability to track behavior at the micro-level (“micro-tracking”) ameliorates the well-understood problem of asymmetric information that hampers insurance markets. When an insurance company cannot distinguish between good and bad risks, it has to charge premiums that are ‘too high’ for good risks. But this leads to ‘adverse selection’: at those relatively high premiums, insurance is only attractive to bad drivers, leaving the company with high-risk customers only. For insurance markets to survive, however, good and bad risks have to be pooled. (In car insurance markets, this is typically achieved by governments mandating compulsory car insurance for everybody.) Micro-tracking solves the asymmetric information problem – good and bad drivers can now be objectively distinguished – and it has fundamentally changed the car insurance market within a short time. This is all the more remarkable since it had remained largely unchanged for the previous hundred years, since its inception around the turn of the 20th century.

It does not take much imagination to predict that, in the near future, most car insurance policies will be of the Pay-How-You-Drive (PHYD) kind. Once in place, they attract safe drivers. This leaves traditional car insurance products with a worse risk pool, which requires an increase in premiums. This incentivizes even more safe drivers to select PHYD insurance, and so on – until everybody is covered by PHYD products. One result of micro-tracking, therefore, is that good risks pay less, and bad risks pay more. When it comes to car insurance, many observers would consider this a fair development. Overall, it may even lead to more careful driving!

But micro-tracking has the ability to fundamentally change ‘person insurance’ as well, and many observers would be concerned about the consequences of such a development. The most important examples of ‘person insurance’ are social policy programs – accident insurance, unemployment insurance, health insurance, long-term care insurance, old-age insurance – run by the government. Collectively, they are known as ‘social insurance programs,’ or ‘the welfare state.’ These programs are affected by asymmetric information problems as well. Governments can circumvent adverse selection by forcing (typically) every citizen to be part of the insurance program. Moreover, premiums are tied to incomes (which governments can observe), not risk profiles (which governments cannot observe). This is the basic structure of a typical social insurance program, then: all citizens are part of the risk pool, and contribution rates are based on income. These programs are solidaristic because good risks essentially subsidize bad risks. The lucky (often termed ‘socially strong’) support the unlucky (‘socially weak’): the healthy support the sick; the employed support the unemployed; the young support the old; and so on.

Once in place, social insurance programs tend to enjoy widespread support. This is especially true if risk profiles are very similar, or if citizens have little information about their individual risk profiles. In the absence of viable private insurance alternatives, even many of those who end up subsidizing others support social insurance programs because risk averse citizens are willing to pay for insurance. But the advent of ‘big data’ has the potential to provide viable private insurance markets in areas that are today dominated by government mandated social insurance. This could fundamentally undermine the broad public support welfare states enjoy.

The kind of data that may enable a health insurance company to differentiate good from bad risks include individual trackers (such as FitBit, AppleWatch, and the like), which can record – and can instantly transmit – physical activity, vital signs, sleeping patterns, and the like. They also include credit card records that allow insurers to assess dietary habits, among other things. But perhaps most importantly, they include genetic information that can increasingly be used to predict the likelihood – and, in some cases, certitude – of future sickness. To be sure, societies can, and do, regulate what information insurance companies can use for their risk classifications. And, indeed, many countries have decided to prohibit the use of genetic information for insurance purposes. But individuals may decide to voluntarily share their tracked data with an insurance company, as many already do, in order to qualify for lower premiums. More importantly, regulation is a political decision. Whether or not a majority of citizens will support regulations that prohibit insurers to use micro-tracking is an open question. The answer depends, in part, on the distribution of risk, which determines whether a majority of citizens stands to benefit from private insurance alternatives or not.

What is clear, however, is that the mere potential of viable private alternatives will chip away from the broad citizen support for welfare states. At least some – and possibly many – citizens would prefer to take out private insurance. Just like many safe drivers are willing to allow car insurance companies to track their driving behavior, healthy people may be willing to allow health insurance companies to track their dietary, sleeping, and activity behavior, and even their DNA.

But this would set in motion similar dynamics as in the car insurance markets. Once the best risks leave the risk pool, premiums for everybody else will increase. This pulls even more good risks towards private alternatives, and so on. Left over are bad risks – which are often uninsurable in private markets (or only insurable at unaffordable premiums). In other words, when private insurers can skim good risks from the social insurance pool, the public pool worsens. In the extreme, the government would be left with those that no private insurer wants to cover. Actuarially fair premiums for ‘person insurance’ (calculated in proportion to real risk) may be possible in the near future, thanks to big data. This will pose profound challenges to existing welfare states.

  • This essay builds on joint work with Torben Iversen.
reagissez !

Ajouter un commentaire