How’s that for a provocative topic? Yes, just as usage based auto pricing has become mainstream, it has already peaked and nobody has realized it yet. Think of it like the Time magazine effect. When everybody recognizes the trend, it’s time to “sell the news”.
Now, at the risk of my stealing my own thunder, I will say that I don’t think UBI is suddenly going away. It will continue to have a place. But, in math terms, it’s second derivative has turned negative. It’s capture of share is peaking and will flatten out and possibly decline (i.e. the first derivative may turn negative too).
What makes me so confident? Well, to be honest, I wouldn’t say I’m confident. What I would say is the belief in UBI is “overpriced” and there are greater risks to its outlook than commonly perceived. Since nobody seems to think there is any risk, that makes me a UBI bear.
The better way to describe my position is that the outlook for the future of UBI penetration in the auto market has become very clouded.
The Mainstream View
Before I get into my concerns, let’s address the consensus on UBI. In the beginning, there was Progressive.
Progressive developed its first version of usage based pricing in the mid 2000s (Snapshot launched in ’08). Nobody else cared. GEICO ignored them. State Farm ignored them. So did most of the other majors.
As an aside, this is a good illustration of how long it can take for things to go mainstream. One of the largest auto insurers has been pushing it for 15 years and it’s finally widely accepted. Insuretech CEOs may want to file that nugget away.
Anyway, five years later, Progressive had $2B of premium on Snapshot. That’s more than all the startup UBI auto insurers (and Tesla) today combined. And that was eight years ago.
So yeah, Progressive was way, way ahead of everyone. And yet nobody cared. Allstate and Safeco started dabbling, but GEICO specifically missed the boat and even Warren Buffett eventually had to admit his error in overlooking UBI.
Things have changed so much in recent years that not only do we have a smorgasbord of UBI centric startups (including mileage based since that is a subset of usage based), but now the auto manufacturers, most prominently Tesla, want back in the insurance business, ignoring the long history of prior OEM failures.
Even Cathie Wood is bullish on UBI calling for Tesla alone to have $23B in premium just four years from now (and at a 60% CR!!!!). That would make it the fifth biggest auto insurer in the country!
And now Allstate is so enthusiastic about UBI that it is pushing regulators to allow wider use of it as an alternative to traditional metrics like credit score. Even GEICO is finally getting in on the act (though they apparently recently let go many of their data scientists which led to some interesting social media back and forth)!
The Contrarian View
And now it’s time for me to rain on everyone’s parade. Sorry.
The simple thesis is UBI works best when things are “normal”. UBI was a way of pushing the frontier beyond segmented pricing models which had begun to reach the point of diminishing returns as they had become ubiquitous.
And that premise made a lot of sense. Now that we know “everything” about the difference between a red 2016 Accord in Atlanta who driver has a 720 credit score and a blue 2017 Camry belonging to a 750 credit score driver in Phoenix, the only way to differentiate was to move beyond proxies for driving behavior to actual driving behavior.
But things aren’t normal. The world is changing a lot more than usual and particularly how and when people drive and how and when they might have an auto claim (hint: notice I didn’t say accident) and how much said claim will cost to remediate.
Am I referring to my old friend stationarity again? Sort of, but that’s secondary this time. I’m not really focused on the fact that UBI may not be as predictive of driving behavior as it used to be. It’s more that driving behavior itself is less important to predicting claims than it used to be. This is a big deal!
The Predictiveness of Driving Behavior
Before anyone gets all worked up, read what I wrote again slowly. I didn’t say driving behavior is no longer predictive. I said it’s less predictive. Big difference. What do I mean by that?
A number of things. Let’s start with an easy one. UBI only predicts collision claims. How hard I brake says nothing about whether my car is likely to be broken into or get flooded or catch fire or have falling debris crush it.
Why does that matter? Because we’re seeing an increase in comprehensive claims relative to collision. Typically, collision claims happen about twice as often as comprehensive. That mix obviously changed a lot during Covid, but, interestingly, while collision claims were down significantly last year, comprehensive was largely unchanged.
Auto insurers have begun to report increased incidence of comprehensive losses recently. Increased theft has played a large role due to societal changes. As floods become more frequent, those lead to comprehensive losses as we saw this summer (and again this week). Increased wildfires similarly cause more comprehensive claims.
None of these can be helped with better UBI. But that’s just comprehensive. What other issues have changed?
As I wrote about recently, severity has been on an upswing that may be with us to stay for quite some time. UBI is largely helpless at managing severity.
While it can tell us that an aggressive driver will have crashes at higher speeds and those accidents cause more damage than fender benders, that only addresses relative severity.
If severity is going up across the board because of inflationary pressure when a repair is needed, UBI doesn’t help find the missing parts stuck in a container outside Long Beach or find extra workers for the body shop. It also doesn’t impact the decision whether total or repair.
Granted, in a perfect world, where you can use UBI to attract only the safest drivers to your company, you can have fewer total claims and avoid the adverse severity inflation.
But that’s not a sustainable outcome. Perhaps you can do that for a short period of time, but eventually equilibrium is reached where risky drivers are no longer subsidizing safe drivers and every company has a “normal” spread of driving risk in its book.
But there’s an even bigger reason than changes in comprehensive and inflation that are weakening the utility of UBI. People’s driving patterns have permanently changed due to changes in work patterns.
The initial use case for UBI was the desire to know when people were driving, i.e. if Sue drove 300 miles a week but they were all in the middle of the day and weekends and Jim drove 300 miles but it was all during rush hour commute, insurers wanted to be able to price them differently because the risks were far different.
While UBI evolved to also capture aggressiveness of individual drivers, it still looks very much at time of day and congestion (hard breaking is really more an indicator of rush hour driving than aggressiveness).
If many people are no longer commuting to work, then the value of knowing who drives in rush hour and who doesn’t is diminished. There is less hard breaking to detect and it is thus more difficult to know who is a good or bad driver.
Of those who still do go to work, some may have flexibility to go in at different hours and avoid rush times. Others may have given up public transportation and started commuting, even though their pricing for their UBI test drive may have been done while they still mostly commuted.
In other words, the models that all the pricing data was calibrated off...you know, the billions and billions and billions of miles these companies all brag about…they’re kind of out of date. So, yeah, stationarity. It’s a problem.
The Longer Term
But wait, there’s more…there are still the two giant elephants in the room that were going to kill UBI somewhere out in the future…ride sharing and self driving cars. Pretty obvious UBI doesn’t matter if you’re not driving and, over time, you won’t drive as much because someone (or something) will do it for you.
Let’s face it, a car is a really bad investment and a big financial drain for many families given it spends most of its time idle. Fractional car ownership will eventually come and only the ultra wealthy will still own individual cars as status symbols.
But before we get there, one more obstacle may emerge first. The regulators. It’s ironic Allstate is pushing regulators for more UBI given the risks involved.
First, there is the one I’ve complained about forever, which is it defeats the concept of pooling. It is OK to price individuals with other people who behave like them. It is not OK to price you individually as your own risk asset. Eventually, regulators will wake up to this.
The likely cause will be when it is realized that UBI discriminates against some group. Let’s imagine that utopian future that’s in all the pitch books. You know, the one where the UBI based insurers have all the good drivers and all the risky drivers end up at the crummy non-standard companies.
Now, let’s say we learn that all the “safe” drivers are higher income, older, married people who no longer have to commute to work because they can comfortably work from home. Then, we realize all the “risky” drivers are lower income, younger, renters who don’t have the means to work at home or their jobs require them to be there in person.
Uh oh. It may be that UBI creates disparate impact? Too soon to tell, but would it surprise you? Can you not imagine a future where UBI is considered worse than using credit score?
People tend to say they care more about data privacy than they act like they care about data privacy. As evidence, all these people signing up for UBI don’t seem to worry about how that data may be used against them.
But forget, the pricing element of this. Let’s get a little creative. Now that most UBI is done through a cell phone app rather than a dongle, how long is it before someone hacks it?
I’m not sure what the bad guys would want to do with your driving data, but imagine when some malcontent, just for fun, puts in some code that make you look like a much riskier driver running up big charges for you. Maybe it’s someone aggrieved that the algo gave them a bad driving score so they seek revenge???
I have no idea how hard or easy this is to do, but given many UBI providers still struggle to determine if you are the driver or passenger in the car, I’ve got to think it isn’t too hard to trick the phone into thinking you’re driving when you’re actually sitting on your couch.
Once people realize that there is a downside to their phone watching them drive, there is a good chance they will abandon the platform or demand changes to how the data is used.
So there you go, it wasn’t too much effort to come up with 2000 words on risks to the future of UBI. I’m guessing with some more effort I can add even more to the list.
That doesn’t mean these will all emerge as problems, but, really, only some of them have to do so to change the adoption rate of UBI materially. This will admittedly be a slow moving story so this prediction will take some time to bear fruit, but when the turn eventually begins, try to remember who warned you first!