Dr. Coenraad Vrolijk has been studying automotive safety statistics for nearly 20 years. Within this span, he has witnessed the application of countless safety developments and driver regulations, as well as the advent of autonomous vehicles, and followed the effect of each on the insurance industry.
A Doctor of Economics, his positions have included Allianz’s Regional CEO of Africa, CEO of the Rosewood Insurance Group, an MD at Blackrock, and a partner at McKinsey.
In this discussion, Dr. Vrolijk shares his research-based analysis and projections on motor insurance and car safety. It is a fascinating study on one of the most important and prevalent parts of our livelihood, and I found the prognosis overwhelming.
This interview refers to research performed by Coenraad—the preliminary findings can be found here.
This is a must read for anyone involved in motor insurance.
A Quote from Coenraad
“It is highly probable that over this coming decade there will be a substantial decline in the price of car insurance in developed markets, due to cars not letting themselves be part of accidents. This will impact the car insurance premium component of insurance companies in developed countries. Motor insurance, and therefore retail P&C insurance, is at a critical inflection point.”
David Schapiro (DS): Could you please tell us a bit about yourself and your career?
Coenraad Vrolijk (CV): Although I am Dutch by birth, I grew up in Venezuela, the UK and Belgium—a very international upbringing! After that I studied economics, statistics, and computing, and went on to do a very quantitative PhD in economics, part of which I did through an internship at the International monetary Fund (IMF). Initially I had planned on going academic, but the 1990s were just when professional consulting firms were expanding beyond their MBA-only-sourcing and McKinsey took me on.
DS: How did you get involved in the insurance industry—was it a proactive decision or an opportunity?
CV: At McKinsey, one of my first assignments was to improve the profitability of a 10,000-barge inland marine hull portfolio. For that I built a regression model (for which I used my grad school university mainframe account) and demonstrated that my statistically founded model performed much better out of sample on the portfolio than the manual underwriting that was done at the time. The rest is history. I went on to work on the first motor insurance rating models following EU motor tariff deregulation in Italy, Spain, Germany, Belgium, France, and the Netherlands. So, I fell into insurance based on my maths and stats background.
DS: Cars and motor insurance are such an important part of our livelihood. You have been researching this subject since the early 2000s; how did this journey start?
CV: I struggled squaring the motor insurance loss experience with the casual observation that cars were getting safer with each new model. In 2005, when I was at McKinsey, I looked at this for the first time using a Swiss dataset and found that loss experience was only going up because the bodily injury costs were rising faster than motor frequency was dropping. I have since performed this analysis with the latest 2020 data. And I also did the analysis on UK, Bavarian and U.S. datasets—all consistent with the empirical evidence on the Swiss data set.
DS: Vehicle, driver, and driving laws, and associated insurances are different but related subjects. How did you address the research from these perspectives?
CV: Every country runs its own set of laws and rules, and these are amended over time. While they are largely similar, obviously there are differences between countries. Car technology on the other hand is global in nature—safety improvements in one market port across all markets. My research was consistent across markets and, more importantly, over time. In fact, the Swiss dataset covers 1950–2020 and, although there is statistical noise year on year, each ten-year period is stubbornly consistent. This would suggest that idiosyncrasies are not causal factors here; rather it’s the overall changes in parameters and not the specific parameter levels themselves. I would classify three broad groups of safety improvements over these past 70 years:
- Physical (e.g., seatbelts, divided highways, roundabouts, safety cells, tire adhesion technology, etc.)
- Safety systems (e.g., anti-lock braking systems, anti-slip regulation, airbags, ‘smart’ traffic lights, etc.)
- Legal (e.g., alcohol limits, driving tests, enforcement, etc.)
It is interesting to note that speed—often used as a bugbear for road safety—is not particularly meaningful: average driven speed is largely unchanged (and even increasing over time, notwithstanding traffic jams). And although speed contributes to an accident’s severity, it is only a very modest cause of accidents in the first place. The single largest cause is only growing as a proportion: inattentive driving.
DS: Could you please deep-dive into the ongoing development of the car safety measures and their influence on car accidents?
CV: I firmly believe that the current generation of car safety improvements are software driven, with the car’s internal software systems avoiding accidents or, at a minimum, proactively reducing their severity through active steering and braking to reduce impacts—including with pedestrians. The Defense Advanced Research Projects Agency (DARPA)-driven first generation autonomous vehicles from the 2000s have gone commercial—Waymo since 2018, Cruise since 2020, and Tesla’s full self-driving (FSD) beta are all part of that trend. One of the main things to note in these car safety features is that the sensors and tools (e.g., cameras, radar, airbags, ABS, cruise control) have been pretty much the same the last decade—what is advancing rapidly now is the software that is working with this data and toolset.
Electric vehicles are accelerating this trend. Although there is nothing stopping recent internal combustion engine (ICE) vehicles getting software updates, it’s the new generation of electric vehicles (EVs) that systematically receive over-the-air software updates, just like consumers are used to on smartphones and tablets. And these software updates include substantial safety improvements. A 10-year-old Tesla can—with only very modest hardware changes—get the latest 2022 FSD version even though that did not exist at all at its outset, which would upgrade it to the latest safety levels.
Automobile parks are undergoing a very fast turnaround; EV roll-out is happening much faster than initially anticipated, and industry insiders reckon >90 percent of new sales across most of Europe will be EV by 2026. This change, coupled to the ability of most EVs to retroactively implement the newest safety software, will result in a one-off doubling of the road safety improvement trend for the coming decade.
It’s not that we are waiting for the full rollout of autonomous vehicles for the bulk of the impact to start happening. All the latest driver aids, including lane assist, collision detection & avoidance, and active braking, do not require autonomous vehicles—yet they significantly reduce accident and severity rates.
DS: Could you please share the key findings of your research, and how they have developed over the last 15+ years?
CV: The underlying parameter I have been looking at is deaths and injuries per driven kilometer. This represents the underlying risk best. Deaths per kilometers driven in Switzerland (and for developed countries in general) drops at approximately six percent, per year, every year since my dataset started in 1950. That means a drop of 98 percent over the past 70 years. This trend is consistent for any 10-year period (e.g., 1950–60, 1960–70, 1970–80, etc.) and continues its trajectory over the most recent 15-year period until 2020. I cannot emphasize enough that this trend is absolutely consistent over a very long period of time, including the most recent years.
There are some nuances, but they don’t change the message. Key nuances include deaths in cars dropping annually at 6.8 percent, faster than road deaths overall (including pedestrians, motorbikes, and bicycles) at 5.7 percent. Major road-related injuries (defined as those involving hospitalization) are dropping at 3.8 percent and all injuries per kilometer are dropping at four percent annually since 1950.
Car and car-occupant safety have increased dramatically, but two-wheelers lag reducing at only 3.9 percent, per year. In fact, by 2020, cars and pedestrians were only involved in 26 percent of all road accidents, with motorbikes, e-bikes, and bicycles representing 70 percent (compared to 1992, when cars and pedestrians were involved in 60 percent of road accidents).
The improvements in car safety are phenomenal and continue their asymptotic trend toward nil.
DS: Have you begun to see the implications of these findings on the motor insurance industry?
CV: The impact of the continually reducing accident rate has been impacting motor insurance since its outset, with perpetually dropping frequency per kilometer. The frequency drop has been offset by increase in mileage driven, both through population growth as well as higher mileage per car, but most importantly by severity, in particular the inflation rate of health costs (the underlying cost of bodily injury).
Yet these factors are also dampening. Mileage growth was very high 1950–1970, but since 2000 (and especially now with the emergence of home-working post-COVID) this growth is much lower. And the pressure on health systems to reduce costs and temper health cost growth is immense across all developed markets, even though in the highest cost health market—the U.S.—the share of health costs in the economy has tripled from 1970–2020 (i.e., an annual growth rate of 7.4 percent in nominal terms per capita, just offsetting the four to seven percent frequency drop rate). European markets are more modest with 4.5 percent nominal health cost growth rates over the same time.
What this results in is dropping pure premiums for motor insurance, starting with the most competitive markets. Since 2011, following decades of growth, the UK has been experiencing a persistently dropping real insurance premium price (at -4.3 percent, per year). Less competitive markets such as Switzerland (where only since the start of 2022 are consumers able to cancel a motor insurance contract after a minimum of 3 years) have simply seen motor insurance margins rise as claims fell without changes in premium levels. And Swiss motor third-party liability (MTPL) in 2019 had a market loss ratio of just 22 percent.
But even in these less competitive markets, where profit overshooting is now peaking, new policy premium levels are also beginning to drop.
What this means is motor insurance market size and margins have recently peaked (in competitive markets like the UK) or are peaking (in less competitive markets like Switzerland) across developed markets. It’s worth noting that we had the same conclusion 15 years ago, but at the time the peak point was clearly in the future (we estimated it to be 2015 at the time). Now we are at or past that peak point.
DS: Based on your research and looking into your ‘crystal ball,’ how would you see the future of motor insurance over the next 5-10 years?
CV: There is no way to avoid this trend. The whole motor and software industry are continuing their research to remove car crashes from society (e.g., Volvo’s motto that nobody will die in one of their latest cars five years from now). As I laid out earlier, the ability of over-the-air updates to retroactively improve safety performance for the latest generation of EVs will accelerate the safety pickup for the coming 7-10 years.
For those insurers heavily dependent on motor insurance, this is their Kodak moment. These insurers will have been experiencing significant shares of the overall profit coming from motor insurance, and even increasing over the past few years as claims undershoot premium ever more significantly, which will have reduced their appetite to find alternatives to motor insurance. In fact, motor insurance in Switzerland has grown from 25 percent of market profit to 39 percent from 2009–2019.
But the Kodak moment is here, now. Ultimately, as regulators put in place mechanisms for insuring self-driving vehicles, motor insurance as we know it today becomes cosmetic and the liability component becomes a small manufacturers liability product, most likely bundled in with the price of the car.
As the premium levels in motor insurance erode, portfolios may be sold (but likely at discounted price levels, given the shrinkage) and larger insurers will be freeing up the capital allocated to motor insurance. Where this goes is their choice, but it won’t find much use in the shrinking motor line.
It’s interesting to note that insurer share prices are not yet reflecting this outlook—just like Kodak ten years after the first sales of digital cameras. Kodak was priced at $106 in 1975 (the year Kodak built the first digital camera), was still in the $50-60 range in the 1990’s as digital camera growth took off, until bankruptcy in 2013. For car insurance, we are in 1999—Kodak was at $66 in end 1999.
It is highly probable that over this coming decade there will be a substantial decline in the price of car insurance in developed markets, due to cars not letting themselves be part of accidents. This will impact the car insurance premium component of insurance companies in developed countries. Motor insurance, and therefore retail P&C insurance, is at a critical inflection point.
DS: Are there any final thoughts you would like to share?
CV: Even in a shrinking game, there will be success cases. But these will be few and far between, with very efficient low-cost operating models.
Coenraad Vrolijk – Bio
From 2015-2022 Coenraad was at Allianz, one of the world’s leading insurance firms, as the Regional CEO Africa. He led the expansion of Allianz’s footprint across Africa, building leadership positions in a dozen key African countries. During 2013-2014, Coenraad was CEO of a startup, Rosewood Insurance Group, where he developed partnerships in Pakistan, Rwanda, and Nigeria. Prior to Rosewood, he was a managing director at Blackrock, the world’s largest asset management firm, where he worked from 2011-2012, leading and building the EMEA Financial Markets Advisory team. Coenraad joined BlackRock after more than 13 years at McKinsey, where he worked on hundreds of projects primarily in the insurance sector. His work covered countries across Europe, as well as Brazil, the Caribbean, India, and the United States.
Coenraad has had extensive Board experience, having served on the Boards of AfricaRe (Africa’s largest and highest rated reinsurer), BIMA (the mobile insurer in emerging markets), the International Baccalaureate, as well as many of Allianz’s subsidiaries in Africa.
Before McKinsey, he was briefly at the International Monetary Fund, where he authored a report on the impact of derivatives on monetary policy. He was also a lecturer in economics at Brown University. Coenraad studied at Brown, completing a PhD and a Master’s in Economics. He also holds a Bachelor of Science degree from the University of Bath, UK, in economics with computing and statistics. He is an IB graduate from the United World College of the Atlantic.
Currently living in Geneva, Coenraad is married with two daughters and a son.