Written by Fouad Husseini, founder at The Open Insurance Initiative and at Robosque.
Prediction markets (PMs) as a concept is not new. In relying on the collective wisdom of the crowd, they have been used as tools for generating accurate forecasting information in order to hedge against real-world risks and for making trading gains. Their most common use has been in forecasting sport, political or economic events. Examples of such markets include PredictWise, The Iowa Electronic Markets and BetFair.
The futures and options financial markets are similar in concept but there are distinct differences. Futures and options are treated as financial derivatives, exchange-traded and created for the purpose of managing risks of currency and interest rate swings i.e. financial uncertainty. These markets are hitherto tightly regulated by the relevant financial authorities. Brokers and market makers dealing in such instruments, operate within centralized markets and companies.
With the explosion of interest in blockchain technology few years ago, prediction markets (PMs) gained then attention as a result of various blockchain startups trying to build decentralized equivalent of prediction markets including that for insurance products. The transparent and disintermediating qualities of blockchains would allow anyone to trade outcome tokens (shares) taking advantage of one’s personal expertise and deep knowledge of insurance and risk management. They would offer wagering opportunities on a far wider range of events.
Uses of insurance prediction markets
By their nature, insurance prediction markets are solely used to predict the outcome of future events. Contracts are established when people place bets on the probability of an event happening and therefore cause a market price. The price reflects the probability of the outcome.
They are literally used to answer questions or predict outcomes. For instance, “will an earthquake with magnitude greater than 5 on Richter scale hit Turkey in 2020?”
Those who believe they have more knowledge than the market can bet and potentially earn a reward for their superior knowledge. The market price adjusts in response to what the crowd thinks the probability of the event is over time. Betting and payouts are normally made using tokens mined by the exchanges traded on (tokenization is a big subject in this respect but not relevant to this study).
If the shares of the earthquake insurance prediction market were priced at 5 cents this translates to a 5% probability of the event happening. Similarly, if the price reaches 90 cents this translates into high confidence in the event occurring.
In that sense, prediction markets have been proven to be reliable and valuable as a way for forecasting outcomes if, a sufficiently large enough number of participants bet on the outcome.
Why would an insurance prediction market be a good predictor?
Perfect predictions would be an act of god. This would translate into 0% probability of an event not happening or 100% probability of an event happening.
Predicting the outcome of an earthquake, is complex and no insurance, reinsurance company or expert individual would have a perfect model to make a safe prediction. But if the diverse forecasting models of all these companies and experts are aggregated, we may generate a better prediction.
The possibility of making a profit lies in deviating or disagreeing with consensus and therefore experts, actuaries or companies with advanced prediction capabilities are encouraged to participate to the disadvantage of ill-informed opinions. The possibility of making a loss or a profit motivates people to exercise more caution and become more objective with their opinions and decision making.
Trading outcomes on exchanges
Predictions or outcomes would be listed in a manner resembling futures exchanges and bets are made by buying and selling shares or outcome tokens. If the outcome has not been fulfilled, share prices will fluctuate according to supply and demand. Shares expire when the correct outcome is determined. In the case of our earthquake example, this would be 31.12.2020 or an earthquake occurrence.
Buying and selling shares is very similar to other exchanges. The participant must specify the number of shares and the maximum price. Everyone is free to sell or buy shares at any price.
The advantages of insurance prediction markets lie in their flexibility to allow for quicker development of new products which can happen in hours in contrast to conventional insurance products which could take several months and involve reinsurance, pricing actuaries and underwriters. Similarly, the types of products that could be developed are infinite.
We will demonstrate how they work with a simple example. Let’s say a market user registers an event question asking, “how many hurricanes will hit Hawaii in 2019?” The user includes the answer options: 1, 2, 3, 4 and Other. Other must be included to ensure that the probability of all answers adds up to 100 percent. If Other was not included, there would’ve been a possibility that none of the answers end up being correct.
As the hurricane season gets underway, the prices of the insurance prediction market will adjust accordingly. As soon as two hurricane occurrences materialize, prices for the “1” tokens will immediately drop to zero as traders rush to sell their worthless tokens. Tokens of “3” may command the highest price compared to other answer options.
At the end of the year, the insurance prediction market closes having (theoretically) witnessed two hurricanes only. Holders of “2” answer tokens claim their reward. Money could have been also made during the validity of the prediction market by profiting from price fluctuations.
The mechanics of an insurance prediction market
Two companies that have made attempts at commercializing decentralized insurance prediction markets are Gnosis and Aigang.
Gnosis’ thesis for an insurance pool is minimalistic-ly described in several online posts and a video describing how an insurance carrier could build a basket of “no” outcome tokens for various unrelated events such as earthquakes and floods being used as collateral for other events. This is obviously an attempt to counter potential event accumulations.
Aigang on the other hand, uses the example of farmer living in flood prone region purchasing “yes” outcome tokens to hedge against a flood event. Aigang’s foray into insurance smart contracts revolved around a contract to insure Android mobile phone batteries and they had chosen to link their prediction market development to the battery smart contract which involved:
Determining claim payouts is usually handled by smart contracts relying on the use of an external oracle to verify the occurrence or the outcome of the event. If your prediction comes true you would collect your winnings to pay for your losses.
The problem with insurance prediction markets
On the surface, much of the interest seemed to arise where traditional insurance coverage is not available or where traditional insurance is too expensive or where risk assessment of the peril requires hard to find expertise or where it relates to an emerging risk with little accident records.
In these scenarios, insurers and actuaries could use information generated by PMs to assess product pricing, risk accumulation and reinsurance needs.
However, rarely addressed issues by those that looked to develop insurance prediction markets tended to overlook:
- Potential volatility in token price action
Token liquidity and price is a problem, though both Gnosis and Aigang were built as decentralized applications (dapp) on the Ethereum blockchain and used ERC-20 tokens, it is shocking to see how prices have fluctuated since their debuts. Market manipulation by speculators is a major risk to consider.
For example, all time high of Gnosis’ GNU token was US$427.89 on Jan 6, 2018, the price has since collapsed and stands at US$16.52 on Feb 29, 2020.
- Regulatory hurdles
Navigating regulatory obstacles has been one of the major obstacles for the development of PMs, while few regulators may have been tolerant to token launches and blockchain based ventures others have been very resistant. This has a significant effect on liquidity.
- Potential lack of interest of participating experts (liquidity)
Without enough participants willing to bet and reflect aggregated information, the pricing of shares and their predictions will be incorrect.
- Most prediction markets until now have proved to be successful in predicting the results of clearly defined political questions or sports events. There has been little evidence of their success in science or insurance related context.
The propositions offered by Gnosis and Aigang relied on theories that proclaimed that the wisdom of the crowd could match or even replace actuarial models without really providing rigorous scientific evidence. There is little research to back the theory behind insurance prediction market projects not to mention their commercial viability.
So, is there enough demand for insurance prediction markets today?
Prediction markets have been around for many years and operated by using conventional technologies. This questions the added value that smart contract could add. Moreover, on the aspect of crowd sourced predictions, wouldn’t machine learning suffice and do the same job in a more traditional business environment without the risk that tokenization presents?
Regulators have raised significant concerns over the viability of tokenized businesses and perhaps rightly so. Many have hardly met the promises that they had made in their initial coin offering. Unfortunately, both companies, Gnosis and Aigang seem to have given up on the development of insurance prediction markets and Aigang appears to have ceased development and trading all together.
Though real-world practical use cases have been and are still largely experimental, prediction markets are an interesting product from an insurance perspective and deserves further exploration. We hope that the efficacy of these markets will develop in the correct context in the future.
However, a major stumbling block for adoption by incumbents and startups alike will be the requirement to conform to financial regulations. Many governments are concerned that tokens used for insurance predictions are nothing but gambling schemes.
This article first appeared on Robosque.