A challenge to insurance leaders: overcome asymmetric information

The theory of markets with asymmetric information

On a brisk Autumn night, in 2009, I dragged my then girlfriend to the Kensington Conference and Events Centre in London to see a lecture by Joseph Stiglitz. Stiglitz was an economic adviser to the Clinton administration, he also served as Chief Economist to the World Bank, and in 2001 jointly won a Noble Memorial Prize in Economics for laying the foundations of the theory of markets with asymmetric information. In 2009 I was living in London, studying Economics & Politics and Globalisation and its Discontents (Stiglitz, 2002) had featured in my macroeconomics reading list. The lecture fascinated me, and the concepts put forward have held their own in my consciousness these past seven years. 

Stiglitz was explaining the fundamentals that sat behind his work on asymmetric information. He explained that the research he and his colleagues had conducted showed that wherever imperfect or asymmetric information existed ‘externalities’ would arise; and the presence of ‘externalities’ did not allow markets to function efficiently. In this context, ‘asymmetric information’ is where one economic actor has relevant information and contracts with another economic actor who does not have this information. The ‘externalities’ are where the actions of an individual has an impact on others which they do not pay for creating ‘moral hazard’. Whilst this may seem obvious today, at the time this was being propagated it was challenging one of the fundamental assumptions of classical economics laid down 220 years earlier by Adam Smith in the Wealth of Nations (1776). The concept that each individual’s motive for profit will allocate resources across the economy efficiently (by ‘invisible hand’) and that governments only need interject in specific instances of market failure. 

In 2010 I graduated university and followed my father into the insurance industry. I was now observing, in my own daily life, a market that operated under the exact conditions Stiglitz had explained to me (and the rest of the auditorium) the prior year.

Do insurance markets function efficiently? Yes, If you have access to data and legal stability.

Stiglitz’s work is used to characterise dysfunctional agent-principal arrangements where those with an incentive to cheat will do so. If this is extrapolated across an entire economy the theory is that it will allocate resources inefficiently (for the economists among you, it will not be ‘Pareto optimal’). For those of us who work in insurance, moral hazard is a concept we are trained to understand. The foundation of any insurance contract is that the proposer knows everything, and the insurer knows nothing. There is therefore, inherent in insurance, an economic incentive to cheat. This action is known as insurance fraud. Notwithstanding this, the insurance market is not chaos, it runs (reasonably) efficiently. Fraudulent claims are kept to a tolerable minimum by legislation that protects the insurers rights, and is enforceable in a competent court of law. The market, and the law, has created an environment where the proposer could lose much more than they gain by committing insurance fraud (declined claim, legal prosecution resulting in fines and/or jail). Who’s going to lie for a cheaper premium, or to receive a larger claim when you could face jail time? Sadly the answer tends to be those in desperate situations with little to lose. In the developed world, this is statistically a small proportion of the population. 

It comes therefore as no surprise that insurance penetration in the poorest countries is low. Insurance penetration is a statistic that measures total premiums as a percentage of gross domestic product (GDP). In the chart below, you can see five developing economies benchmarked against Australia, United Kingdom and the USA in 2014/15.

In addition to the challenges of high fraud rates, insurance remains immature in the developing world due to poor access to underwriting information. This is another example of asymmetric information; if insurers do not have access to reliable claims data then it is impossible for actuaries to price products for those markets. This is why insurance markets are functional in developed economies; data and legal stability. These are conversely, the same two reasons insurance development and supply is so limited in the developing world. You will find published commentary containing other reasoning such as political instability or lack of local talent. Whilst these are indeed challenges, they are not (in my opinion) prohibitive. A lack of underwriting data and poor contractual enforceability are. Asymmetric information is limiting the availability of insurance in the developing world. The poignant irony here, is that it is individuals in the poorest countries who need insurance the most. 

Why should our industry invest in overcoming asymmetric information?

I believe that insurance is, fundamentally, a good thing. It helps smooth life’s shocks and gives people the confidence to invest, grow, and improve their community’s standard of life. Individuals in developing economies have less reliable government support, which makes the immaturity of private insurance not just an economic, or commercial issue, but a social welfare issue also. There is an argument that insurance is only required once an individual has a stock of wealth worth protecting. I believe this to be a flawed notion; from the day every single person is born they have an insurable asset (their health). Those who have children or other dependants accrue a second insurable asset (their life). In addition to these universal assets, the number of individuals in developing economies who are the owners of tangible assets such as vehicles, homes and businesses is increasing every year. I’ll come back to the social impacts of overcoming asymmetric information later on. For now, let’s focus on the economic and financial impact.

In the table below you will find some readily available data that I have collated. It shows the non-life insurance penetration levels for 10 countries in Africa. 

As you can see, each of these countries has an insurance penetration of between 0.50% and 2.00%. The summation of their 2015 non-life premiums is just under U$ 4bn. You will recall from the bar chart above, that Australia’s non-life insurance penetration is 2.20%. This is actually just below the global average of 2.60% (Tripe & Kelly, 2015). After collecting this data, I wanted to test how much additional premium-to-market would be generated if each of these country’s 2015 insurance penetration was equivalent to that of Australia (2.20%). I have catalogued the results below.

As you can see, if each of these country’s insurance penetration was equivalent to that of Australia (2.20%), then this would generate a U$ equivalent of 12.5bn of additional non-life gross written premium (GWP). Remember, insurance penetration is a relative measure, not an absolute measure. These countries would not need to be as wealthy as Australia, just have the same proportion of premium to GDP. To put this in context, I have measured this U$12.5bn against some global insurers 2015 GWP (U$ equivalent).

As you can see, the additional non-life premium that would be generated is greater than the 2015 GWP of XL Catlin, Allianz and QBE. What I’m trying to demonstrate here, is that the rewards for conquering asymmetric information in the developing world are not insignificant. Furthermore, you will appreciate that the U$12.5bn is a number that does not take into consideration;

• Growth - that these economies are growing at 2-3 times the rate of developed economies, and what is U$12.5bn today, is substantially more tomorrow
• Life - that this number does not include life insurance, which, arguably based on OECD statistics, has more potential scale than non-life
• Profitability - that these countries have few market participants and that first movers (assuming asymmetries are overcome), will be able to generate combined ratios lower than in the developed world
• Scaleability - that this is a sub-set of 10 countries in Africa, and if scalable, there are 155 developing countries with similar characteristics (according to the IMF’s World Economic Outlook Report, 2015)

The work that could be done on adjusting the potential windfall for the above factors is enormous, and beyond the means of this commentator. Notwithstanding this, if I wanted to be reckless with my maths (and I do), and assumed a total GDP growth in developing countries of 30% over a 7 year period, and an ability to scale to 10x across 155 countries, then the 2024 prize for cracking a scalable solution to asymmetric information is over U$160bn of non-life GWP (or U$24bn EBITDA at an 85% combined ratio). I wont bore you with another chart, but if you wanted to contextualise this number, it is greater than the 2015 GWP of AIG, Chubb (inc. ACE), Swiss Re, Munich Re, XL Catlin, Allianz & QBE…combined.

How do we overcome asymmetric information in the developing world?

Technology. You would think that a U$160bn question would have more than a one word answer - but it doesn’t. It might be easiest to lead off with an example. 

At the 2015 World Economic Forum, Blue Marble, a microinsurance company backed by a consortium of global insurers launched a crop insurance product targeted at developing economies. These inexpensive policies use meteorological data gathered from satellites and weather stations, and are triggered when a pre-determined weather position is reached. For example, a policyholder might become entitled to a claim if there were 60 consecutive days without rain in their region. The policyholder is not able to make a claim themselves, and therefore, there is no scope for insurance fraud. In addition, the policies are rated on global weather patterns, for which we have decades of reliable data. There is also very little scope for legal contest given the claim’s validity is determined by non-subjective meteorological data that cannot be manipulated by either the insurer or the policyholder; so contractual enforceability is no longer a prohibitive factor. This is one example of how technology has created a means of overcoming asymmetric information. Although once we move away from weather related triggers (weather is easy to measure, and affects everyone), and we actually need personal data on policyholders, things become more challenging. 

I am not a futurist. I am not qualified to predict where science will lead us (If you have checked my LinkedIn you will know I’m barely qualified for the opinions in this paper). That said, for what it’s worth, there are those whose predictions I hold in high regard. One of these predictors is Jeff Immelt of General Electric (GE). In November 2012 Immelt invested U$1bn in developing technology and systems for his ‘industrial internet’ ecosystem. This had built on earlier investments that had been made into the development of sensor technology, which was deemed as one of the ‘unstoppable trends’ his team identified in the post-Welch GE era. In June 2013 GE invested a further U$2bn in their industrial internet ecosystem, this time focused at health care. In 2014 GE announced that they would exceed U$1bn of annual revenue by selling the insights they were collecting from the data produced by their industrial sensor technology. I’ve used this example to demonstrate that where we are on the technological development curve, investments in sensor technology pay back quick.

Now, here is the challenge I put to the insurance industry leaders. Turn your attention to the development of sensor technology. Direct some of your resources and mental faculties to answering the following six questions;

1. How can a sensor tell me if a piece of equipment has been genuinely stolen, damaged or destroyed
2. How can a sensor tell me if livestock is healthy, sick, or dead
3. How can a sensor tell me what contents were in a home at the time of a fire
4. How can a sensor tell me if a vehicle has been damaged or destroyed by a fortuitous cause
5. How can a sensor tell me if a worksite is compliant with safety standards
6. How can a sensor tell me if a building’s sprinklers and alarms are on and working

If a sensor can tell an insurance company these things in real time, then you have equalised the effects of asymmetric information. The answers to these six questions ought to enable insurance companies to maintain fraud at a tolerable level (sensors don't have an economic incentive to cheat, so they tell the truth). They also ought to enable insurance companies to collect meaningful underwriting data at a personal level, allowing for the profitable pricing of products. I am not a scientist. I have no idea how a sensor can do any of those things, but I am aware that the cost of sensor production is reducing. In fact, ATLAS have mapped the cost of IoT sensors and the forecast cost for a sensor in 2020 is U$0.38. Skeptics, use your imagination! Answering these six questions could result in something that can actually scale viably.

Social welfare considerations

As we approach the end of my challenge to insurance leaders I want to leave you with a final set of data. It is the same 10 countries in Africa we reviewed earlier, although showing some different metrics.

As you can see above, the 10 countries we reviewed earlier also collectively have 250 million people living below the international poverty line. I’m going to stray from hard facts now, just to illustrate a point. 

Try to imagine the number of people you know who have suffered some form of economic hardship in their life. Perhaps they lost their home in a fire, wrote their car off, got critically sick or had a relative pass away. For arguments sake, lets say that you can think of 1 in every 10 people you know who have suffered some form of hardship like this. In the developing world, with no compensation, an event like this puts an individual below the poverty line. Now, let's say that 50% of the population bought insurance, statistically that’s 1 in every 20 people who will suffer a hardship and receive insurance proceeds. If the perpetual stock of people in poverty, just in these 10 countries, were to reduce by 1 in 20 (5%), that’s 12.5 million people. This is just a loose illustration of (but hopefully demonstrates) the potential social welfare impact higher insurance penetration could have on the developing world.

Summation 

I will now finally address the question you have invariably all been wondering…yes, the girl who I dragged to Stiglitz’s lecture on asymmetric information did eventually forgive me for the dullest 90 minutes of her life (and actually, against all good sense, agreed to marry me a few years later). 7 years on from that night and Stiglitz’s work still causes me to wonder why more insurance executives don't have a budget for this kind of targeted R&D. If you consider that we live in a world where insurers are acquiring and consolidating at between U$1 - U$2 of investment per U$1 of GWP growth (based on metrics from major mergers in the last 3 years). Given this, you could argue that the insurance market would value U$160bn of annual GWP growth at an acquisition price of ~U$240bn (give or take U$50bn…). Remember, Jeff Immelt is revolutionising the industrial world with an R&D spend of ~U$3bn. If you could solve the problem of how can technology help insurance overcome asymmetric information in the developing world? for the same U$3bn, then you would arguably obtain ~U$240bn of economic value for a 1.2% R&D outlay. U$160bn of GWP, U$24bn EBITDA, and…you never know, you may also inadvertently lift twelve and a half million people out of poverty along the way. 

Sources & Notes

• African Insurance penetration data sourced from Peter Tripe & Norman Kelly’s presentation to the Actuarial Society of South Africa’s 2015 Convention: “Doing Insurance in Africa – a brief snapshot”
• OECD insurance penetration data sourced from stats.oecd.org
• GDP, population & poverty line data sourced from the CIA World Factbook online
• This has been written in my personal capacity and the views and opinions above are not necessarily those of my employer