It is not enough to simply use data and the automated processes that may be driven by it – businesses need to demonstrate it is complete and it is being used ‘ethically’
Data adequacy, management and control are the currency of a successful business in modern times – and for the reinsurance market it is no different.
But as the market and its stakeholders become accustomed to just how important data is to insurance risk selection and management, so are there growing demands for business and legal accountability for the use and adequacy of such data.
It is becoming not enough simply to use data and the automated processes that may be driven by it but businesses also need to demonstrate the data is complete and is being used “ethically”.
At a more immediate level, executives and boards will be held to account for the impact data adequacy and use has on business profitability and, of course, ultimately return on equity; in other words, on underwriting performance.
If the human underwriter is to be replaced by an AI algorithm and that algorithm is now to determine if a prospective customer will be accepted for a risk and at what price, customers want to be assured this newly automated decision-making is the output of a fully comprehensive data capture.
Take climate risk, for example. Insurers and reinsurers now recognise the greatest challenges they face in managing climate risk is in respect of the data they hold and analyse. Principally, if climate risk is to be identified, managed and derisked by the market, can re/insurers be sure the data they hold and use is comprehensive enough?
For example, the market is looking at data to inform it of growing or changing risk in terms of frequency and geography with respect to climate inspired events such as hurricanes, storms, floods, drought and fires. Where will those claims hit? Which potential insured businesses will suffer from them and how will those liability exposures manifest themselves?
Property insurance is an obvious first focus in this respect, but climate change issues and associated litigation are not confined to the desks of property/ casualty underwriters.
Climate events can expect to generate more liability claims as property and persons become at more risk of property damage or bodily injury.
Growing stakeholder/investor oversight more generally can expect to translate into directors’ and officers’ liability claims, for example. In the US, lawsuits already exist where investors claim corporations have not disclosed fully or adequately to their investors the financial impact of climate change issues.
The issues are not just confined to the direct carriers. At a reinsurance level, there is an obvious aggregation risk to be addressed as more widespread climate-generated events produce more individual claims in potentially more geographically specific locations.
Bad business and underwriting decisions will be made if risk decisions or ultimate exposure analyses are made through incomplete or statistically dubious data; risks will be wrongly priced and ultimate exposures may be hidden.
Conversely, data inadequacy may mean decisions to decline risks or portfolios are made on a false belief they carry too much risk when that may not be the case – profit and earnings also risk being left on the table.
Data and its modelling has to factor in that risk in the world is interconnected and climate risk has the propensity to generate new exposures or exposures in new geographical locations.
The importance of data integrity extends into other newly important areas for the market, such as disruption technologies and artificial intelligence (AI).
Telematics, smart technologies, algorithmic analysis – all provide exciting new ways for the market to access data and then to process it, to improve efficiency, manage cost and open markets to new customers. As with the focus on climate risk, the successful delivery of these objectives – the benchmark for underwriter and executive accountability – using these new data-driven tools will only happen if the data that is compiled and used to drive them is thorough and complete.
However, there is a growing trade-off now being demanded by customers in return for their acknowledgement and acceptance for these disruptor technologies, namely that the market can show its data capture and use is “ethical”.
The signs are this trade-off will in future have legal/ regulatory underpinning. Algorithm regulation is very much on the table of all developed jurisdictions. Most have in very recent years produced AI ethics guidelines but now legislation is clearly coming.
For example, in February US lawmakers proposed the Algorithmic Accountability Act, which would require companies using AI to conduct critical impact assessments of their automated systems by reference to Federal Trade Commission regulations.
Additionally, the EU is attempting to create a legal regulatory framework for AI with its proposed legislation. In the UK, meanwhile, the Government published its AI regulation policy paper on July 18.
To the public, if the human underwriter is to be replaced by an AI algorithm and that algorithm is now to determine if a prospective customer will be accepted for a risk and at what price, customers want to be assured this newly automated decision-making is the output of a fully comprehensive data capture.
The public want any automated decision-making process to be free from inherent biases. Customers and Governments/regulators want to be assured in this modern world of operating that automated decision-making is free from demographic or gender bias.
No one wants to find out that because of incomplete or statistically misunderstood odd data capture, automated processes have crossed that line. Get that wrong and businesses face the perfect storm of brand damage, regulator intervention and also customer/ class legal action.
If investors and shareholders hold, as they will, businesses to account for their decisions and performance going forward, checks need to be built in to ensure and demonstrate the integrity of the data being relied upon to substantiate those fundamental risk and strategic decisions being made.
That may not be so easy as it sounds because many in the market are reliant on third-party modelling and data provision.
However, the market needs to find a way. This is not just because data-driven sub-optimal performance will cost underwriters and board members their jobs, but as can be seen around the world, Governments and regulators will demand a way be found.
For more information, please contact Stephen Netherway.