Later today I will be meeting with the students in Gary Chou’s Entrepreneurial Design SVA class (which Gary and Christina started together last year). We will talk about what’s happening in the labor market broadly and why that puts a premium on being a (micro) entrepreneur. As part of the preparation, I watched and read the materials that Gary had assigned for the class which included Max Levchin’s keynote from DLD.

Max is super smart and working on some fascinating ideas around making more of the analog world digitally accessible through cheap sensors. He ends his talk with a hypothetical example of what this kind of sensor data could do for car insurance which includes the following quotes:

On a Sat morning, I load my two toddlers into their respective child seats, and my car’s in-wheel strain gauges detect the weight difference and reports that the kids are with me in a moving vehicle to my insurance via a secure message through my iPhone. The insurance company duly increases today’s premium by a few dollars. […]

But In a few hours, my car’s GPS duly reports to my insurer that I only drove two miles to the park, never sped and, and observed all traffic signs. My phone now chirps happily: not only has my rate been discounted, several companies are offering me a deal on insurance!

It’s true that already today you can get somewhat better rates if you are a more defensive driver and there are even insurance companies that already offer this based on sensor data. But there is an important rub that a lot of people don’t fully appreciate (Max may, but he doesn’t mention it in his post): If you discriminate risk classes too finely insurance breaks down. 

Let me illustrate with an example that I was reminded of when I saw the picture of Lindsay Vonn being airlifted by heli after a devastating fall in yesterday’s Super G race in Schladming. I was just away for a few days skiing in Verbier which offers insurance for 3 Euros per day that pays for being airlifted off the mountain. Verbier on a winter day probably sees some 20,000 skiers. Let’s say that it costs 30,000 Euros per day to have the rescue personnel and chopper available. They can cover that nut if about half the people (i.e. 10,000 folks) buy 3 Euros of insurance each.

Now imagine a scheme in which you give every skier an app for their phone that tracks how fast you ski and whether you go out of bounds or stay on the groomed slopes. You charge the safe skiers only 1 Euro and 80% of skiers are safe. You now make only 8,000 Euros on those safe skiers and need to cover 22,000 Euros from 2,000 risky skiers which winds up being 11 Euros per day. Suddenly you may find yourself in a place where the risky skiers stop buying the insurance because they no longer think it’s worth it! If only half of them buy it, the ski resort is 11,000 Euro short of the operations budget and the entire insurance scheme has collapsed!

This problem exists for most types of voluntary insurance, which is why mandatory insurance is often an appealing solution. In the above example in fact if you simply made the insurance mandatory every skier would have to pay only 30,000 / 20,000 = 1.50 Euros! What is the lesson here then? In voluntary insurance schemes you cannot discriminate risk classes too finely because it will result in a collapse of the insurance scheme. Also, the true gains come if you can use the sensor to reduce the total cost incurred by the insurance scheme. For instance, earlier in Max’s car insurance example, if the sensor that detects the toddlers changes the car’s driving characteristics so that you can only drive up to a certain speed that could be a way to actually reduce the number of car accidents involving toddlers. That results in savings some of which (with care) you can allocate to the folks with the sensors without the insurance scheme collapsing.