4.1 Underwriting and Big Data

The essence of the use of Big Data and analytics within the insurance underwriting function is to provide the underwriter with greater visibility of the potential risks on which to base their assessment of exposure and price. It does not automatically follow that an underwriter will be able to make better decisions with more information – the information has to be relevant, reasonably accurate and appropriate. One fundamental aspect of insurance is that it is an industry which is not based on certainty but rather on probability. By this we mean the probability of an event which may or may not happen, the probability of certain costs or consequences arising and the vulnerability of a property or person to the effect of that event. Insurers limit their exposure through the wording of the insurance contact but also by forcing the policyholder to take certain mitigating actions. For example in a commercial environment underwriters will want to ensure that the factory has a fully maintained water sprinkler system, or in a domestic situation that an intruder alarm is fitted.

Having access to all available information may not be the absolute key to prediction. Taleb's theory of ‘black swans’² recognizes that major unforeseen events can happen which, with hindsight, might have been predictable. Major catastrophic loss can always occur but it is the intention of the underwriter to ensure that if the worst does happen, then appropriate protections are in place such as reinsurance arrangements. Even then, there may be legal issues to consider in the interpretation of the insurance contract in place. For example, in the case of the World Trade Center, for the purposes of insurance cover the courts needed to decide whether the 9/11 disaster was only one attack or two.

The dominant effect of the data and analytics era is to make the information more relevant so that insurers are underwriting to a more granular or personalized level. What this means is that professions, locations, age and gender will continue to serve as indicators, but more information will become available about the individual, and that this more detailed information will become the dominant decision driver. There are interesting moral issues to consider. In their paper ‘Social Justice and the Future of Flood Insurance’³ the Joseph Rowntree Trust comment on the fact that there are two contrasting types of insurance emerging:

• Individualist, risk-sensitive insurance provided through a market in which individuals' payments are proportional to their level of risk.
• Solidaristic, risk-insensitive insurance in which those at lower risk contribute to the support of those at higher risk.

It is Rowntree's view that the UK for example is moving towards the former where insurance is being rated at an individual level whereas other countries are moving towards the latter. This is probably a generalization and arguably one which is incorrect as the spread of ‘user-based-insurance’ (UBI) becomes more dominant in North America and Western Europe. Even so-called ‘emerging markets’ such as China are looking at UBI with increasing interest.

The UK-based Chartered Institute of Insurance (CII) also comments on the ‘social benefit of insurance’ in which they say that insurance:⁴

• ‘Efficiently protects the public through innovative risk management techniques.
• Frees up businesses and professionals from everyday risks and encouraging innovation and competition.
• Relieves the burden from the state and providing comfort to individuals by providing safe, effective and affordable pension savings, (and) protection . . .’

The use of individualist, risk-sensitive and more granular underwriting does not appear to be at odds with the CII's view on the social benefit of insurance. Insurers are not in business to lose money – they never have been – and pricing a policy relative to risk is not unreasonable. In fact, this approach has been in existence since marine insurance started in the 17th century when different ships were rated according to their cargo, size and ownership.

As a corollary to the issue of the use of data at a granular level and whether it is correct or not to consider someone or something uninsurable is the issue of the ethical use of data. In their article ‘What's Up with Big Data Ethics’⁵ (Forbes magazine, March 28, 2014) Jonathan King (and others) provide a reminder of the ethical issues surrounding the use of data covering topics such as ‘privacy, confidentiality, transparency and identity.’ They suggest that Big Data is more than smart algorithms but is about ‘money and power’ relating to the profitability of the insurance carrier – and perhaps going forward, also to other related organizations which might become aligned to those insurers.

Whilst the article refers to the legal obligations which attach, organizations must give evidence of their trustworthiness by demonstrating transparency, honesty and confidentiality. The underwriting challenge may prove to be that although information is available to an underwriter, they may be precluded from using it. As insurers depend on underwriting guidelines to set the standards and ensure consistency, then it is only a matter of time before these guidelines will need to recognize the ethical position, and to draw lines which the underwriter may not cross in the search for better and more granular information.

To some degree these ethical guidelines may not be optional but rather might be enshrined in legislation. At the time of writing, the US-based National Association of Insurance Commissioners (NAIC) Casualty Actuarial and Statistical Task Force is looking in more detail at the question of rate calculations, and the topic of so-called ‘price optimization’ where rates are flexed dependent on matters not directly related to the physical risk, but rather related to ‘elasticity of demand.’ In other words, underwriters might consider the pricing of a risk, i.e., the premium, as being linked to the policyholder's propensity or need to buy, or to change insurer. Since December 2014, Maryland, Florida, Ohio and California have placed insurers on notice that they will not agree rate filings that use such a practice. ‘Rate filings’ are the process by which a State reviews insurance policy forms and their accompanying rates to ensure that consumers receive the protections and benefits required under local State law. The majority of filings fall into the categories of life, annuities, health, personal auto, homeowners and workers' compensation.

As underwriters come to terms with the Big Data and analytical environment, it is worth considering the additional information which might come into their hands, and the consequences (Table 4.1).

4.2 Underwriting for Specialist lines

There is a temptation to consider that analytics is for the ‘volume’ market, and to some degree that is true in so far as statistical analysis is usually based on the laws of big numbers. Notwithstanding, there are still important elements of the Big Data and Analytics agenda that apply to specialist lines, typically terrorism, High Net Worth, agriculture and transit, and these may apply more to the role of underwriter than perhaps to any other function in the specialist insurer segment (Table 4.2).

4.3 Telematics and User-Based Insurance as an Underwriting Tool

Since 2010, the use of telematics in auto insurance has grown and it is no longer seen as a niche underwriting product for hazardous drivers such as the young, whose premiums are prohibitive, but rather as a mainstream alternative to ‘normal’ types of cover. The term extends to cover not only the black box technology but also the collection of data, its storage, analysis and ultimately how that translates into actionable insight about the risk relating to any particular driver, in a specific place and time and in a particular vehicle.

All this represents a genuine evolution in the pricing of risk which traditionally has taken ‘class variables’ as an indicator of risk or ‘proxy,’ whereby behaviors provide a much greater insight into risk. Historically insurers have tried to reflect behavior by reviewing accident history, no-claims bonuses and even in some early cases the distance which has been driven. This has evolved beyond ‘pay as you drive’ to ‘pay how you drive.’

Other typical indicators might include:⁶

• How you drive
• When you drive
• What you drive
• What color is your car
• Where you drive to
• What else you drive.

Telematics is an example of the use of integrated technology at its best. The topic itself goes beyond technology but also extends to issues of culture, pricing, data ownership and risk management. Used effectively it also potentially allows insurers to move from a reactive mode where they are examining behavior and characteristics and retrospectively pricing the risk, to a potential position to be able to influence personal actions and in doing so to be proactive. One step towards this is the availability of data to the policyholder usually viewed through a portal. The consequence of this is the increased emergence of insurers as ‘risk managers’ rather than risk evaluators.

The access by underwriters to more data provides underwriters with considerable flexibility and control in setting premiums. The advent of systems which provide more data is likely to give underwriters considerably more insight. At the moment insurance underwriters seem only just to be coming to terms with their new-found capabilities. They seem in many situations to be viewing telematics (according to one recent UK job advertisement) as simply a new tool for collecting data, and from this to think about ‘managing business volumes, gaining insight into profitability, product structure and rating, together with product development and distribution.’ In effect the particular role appears to be as much about telematics product development as about better underwriting. Other insurers seem to view the role as some form of ‘underwriting data scientist’ which reinforces the point that data and analytics appear to be not only influencing existing professions and job roles but also creating new hybrid roles.

What becomes clear from the additional data available is that underwriters have access to much more data than before, but the real test is what they choose to do with it. It is obvious that they will benefit from greater segmentation and will be able to apply more dynamic pricing. This will also help with their product and segmentation ‘mix’ and ultimately their ‘go to market’ strategy. Underwriters can also improve the way that they apply discount rates and other incentives dependent on market conditions and the competitive landscape. Because of these variables and the differing risk appetite between insurers, premium rates can increasingly vary significantly between insurers for the same type of cover.

In many cases, current pricing for insurers seems to be restricted to retrospective review of telematics data, perhaps annually or at 90-day intervals. Discounts are then offered against the base rate and some insurers alternatively offer ‘free miles’ of cover rather than a financial discount.

As the use of telematics grows and the first signs of user-based insurance are emerging in other business areas, the indications are that the proverbial UBI ‘genie’ is out of the bottle as far as this approach is concerned. What has been demonstrated is that the technology has mainly proved to be robust, and that whilst some challenges might still exist it will be possible to resolve them. The growth in use suggests that, at least for some customers, this is an attractive proposition which will apply not only to the younger driver but also to the older driver or those who do less mileage. Experts anticipate significant continued impact on products and pricing.

It is clear that there is more development yet to happen in this space. If underwriting is ultimately to be at an individual, granular level then insurers will also need to think about how this extends to multicar policies, where a ‘second’ car is only used on a very occasional basis.

4.4 Underwriting for Fraud Avoidance

With insurers increasingly using analytics at the point of claim to root out fraudulent claimants who are either opportunistic or organized, there is a natural tendency to consider whether these same tools can be used earlier in the insurance value chain process. At the point of claim the decision to investigate will generally depend on a number of factors, usually a combination of the event circumstances, the timing of the incident, the claims history of the policyholder and often also the demographics of the immediate environment.

As insurers look forward there may be increasing pressure on the current or prospective policyholder to allow the insurer to have access to all available information, much as a mortgage lender is entitled to check credit details before making a final decision on providing a loan. In fact, one insurer has recently started to consider creditworthiness as a key element of the underwriting process. The context of this is unclear and has stimulated some debate, it being argued that an individual's ability to pay a bill on time has nothing to do with their behavior behind the steering wheel. A contrary argument might be that the way an individual manages their financial affairs could be an indicator of their approach to personal risk and as such is a valid albeit slightly controversial metric.

Predicting fraud at inception opens up the possibility of an insurer ‘pricing out’ a prospective or existing customer based on what the policyholder ‘might’ do, rather than what they have done. There is nothing strange in this, in the same way that an underwriter might want to consider, for example, the potential for a teenager to have a crash in a high-powered sports car. In the case of predicting fraud, the underwriter not only anticipates the possibility of a fraudulent claim at some time in the future but also anticipates potential criminal behavior on the part of the prospective policyholder.

Naturally, were an underwriter to openly decline to provide cover on the basis of anticipated fraud, this would be a serious issue. A policyholder who has cover declined by an insurer is obliged to disclose this to other insurers when seeking new cover, as this is a question usually explicitly asked by the insurer at the time of proposal. For the potentially fraudulent policyholder to deliberately mislead is a matter of non-disclosure which also entitles the insurer to void the policy.

At the time of proposal, an insurer is not obliged to give any reason why cover is not offered but in the absence of any obvious material cause for declining insurance, the reason may be implicit. At what point will a potential customer be entitled to see the data on which they are being penalized? Will they understand it? Similarly, were a customer to discover that the offer made to them by an insurer at renewal or inception was deliberately uncompetitive, would they have a cause of action and if so what might that action be? Are there potential issues of discrimination to be considered? Will regulators view this as a matter of treating the customer unfairly or some other breach of consumer regulation?

4.5 Analytics and Building Information Management (BIM)

If the insurance industry has to date focused on analytics relating to finance, risk and the customer then it will soon need to prepare itself for the next wave of information, that which originates from Building Information Modeling (sometimes also termed ‘Building Information Management’) but more commonly known as ‘BIM.’

Originating in the USA, specifically in the US Army Corp of Engineers (‘USACE’), the main intention of BIM is to provide an accurate record of data relating to the construction and maintenance of new buildings.⁷ Already used in many USACE projects and missions, it is described as ‘Smart 3-D,’ differentiating it from the 3-D ‘autocad’ drawing capabilities already used by many architects. One of the criticisms of BIM is that whilst it provides a digital plan of the proposed construction, it is not always representative of the ‘as-constructed’ version due to construction changes and general modifications where records have not adequately been captured.

There is a useful (although slightly tenuous) comparison between BIM and the Solvency II program. Solvency II was intended to be a principle-based approach to the management of risk capital. As insurers increasingly asked questions of the regulators due to the uncertainty of the requirement, it became necessary to provide more detail to a point where the so-called Red Book became especially overcomplicated. The impact of this was to add cost and complexity. The same issue may potentially occur with BIM where the general principles of digitalization may well be overtaken by the degree of detail called for by practitioners.

Increasingly, major engineers and consultants are adopting the BIM methodology, specifically the PAS 1192 suite of information. PAS 1192 is the UK specification for information management for the capital/delivery phase of construction projects using building information modeling. One of BIM's strongest evangelists, Anne Kemp, a Fellow and Director at Atkins Engineering Consultancy, encourages her colleagues to look for the data points at each part of the construction process and capture information accordingly, a process which she describes as the ‘digital plan of work.’

In reality, BIM has been relatively slow to take on in Europe but the decision by the UK Government to enforce the adoption of BIM in the construction of public buildings from 2016 will create an additional catalyst for change. The UK Government Strategy is to encourage ‘fully collaborative 3-D BIM with all project and asset information, documentation and data being electronic.’ The aim is for the UK Government to drive improvements in cost, value and carbon performance through the use of open sharable asset information.⁸

Whilst for many this will be viewed as another layer of bureaucracy, and for engineers it might be viewed as an unnecessary evil, it is nevertheless an interesting and important step in the digitalization of the Built Environment, and will be of enormous value to insurance underwriters. Skeptics who view this with suspicion must take into account the rapid development of the Internet of Things (‘IoT’), and the ability to increase the amount of data through an increase in the number of devices and from this gain meaningful information. One useful definition of a device or transmitter contributing to the IoT is ‘A thing, in the context of the Internet of Things (which) is an entity or physical object that has a unique identifier, an embedded system and the ability to transfer data over a network.’⁹ Over the past five years, the speed of connection has increased by 200 times, the number of machines connected to each other by 300 times and the price of remote devices collecting information has fallen by 80%.

If BIM is likely to take a decade to embed itself in public buildings in the UK and beyond, and provide meaningful information from any data, then how long will it take to embed into commercial and residential properties? Changes often occur more quickly where they are least expected, or sometimes more slowly when there appears to be an absolute certainty of change. The speed and degree of engagement of the construction industry will be critical to change. It is only a matter of time before new construction reaches a ‘tipping point’ where all new buildings and their contents are digitized.

This BIM information naturally will be in a structured form but it will become more valuable as it is combined with unstructured information, typically climate conditions, weather and building usage such as customer footfall (for example in a retail scenario). Publicly funded building construction has the potential to provide the necessary proof of technology which leads into the digitalization of all new and, eventually, all existing buildings.

The absolute impact on insurers and underwriters is difficult to gauge at the present time. Clearly additional information will become increasingly available. This will help not only in underwriting decisions by providing greater levels of granularity but also in the claims process where investigations into policy conditions and warranties will perhaps also become more automated. BIM will also be one of the catalysts for the ‘connected (or “Smart”) building.’

The use of information collected through multiple data points will also help insurers gain a better understanding regarding building usage, behavior of visitors, physical conditions within the perimeter (and perhaps also in the surroundings), comparison with other similar buildings, and all in the context of codes and regulations. This will all lead to deeper insight and granularity regarding classes of risk. BIM when coupled with other data will also provide better insight with regards to compliance with warranties and policy conditions, and may even trigger alerts to building owners and managers that they are in breach of coverage requirements. Perhaps the use of BIM will also be another catalyst in the transformation of insurers from ‘compensators’ to ‘risk managers.’