The Chairman (Mr W. J. Baker, F.I.A.): Welcome to our discussion on “Adverse selection in a start-up long-term care insurance market”.

Dr Les Mayhew is going to do a scene setting for the topic and then Dr Craig Adams will present his paper.

Dr Les Mayhew is Professor of Statistics at Cass Business School. He is an Honorary Fellow of the Institute and Faculty of Actuaries and a member of the Royal Economic Society. He is a former Whitehall senior civil servant. He is managing director of Mayhew Harper Associates Ltd., a research consultancy specialising in the use of large administrative data sets, and a member of the Office of National Statistics Expert Panel advising on population projections.

He has widely published with research interests including health and social care and pensions. His current research interests include paying for long-term care.

Prof L. Mayhew, H.F.I.A. (presenting the paper): It all started with the Dilnot Commission a few years ago. The key item in the commission accepted by the Government was to have a cap on individuals’ catastrophic costs of care. This is for England; the situation in Scotland is different. I am assuming that most people here would look to the English policy on this rather than the Scottish policy.

Secondly, there would be a revision to the means testing thresholds to extend State support to more people. In their argument they said this may prevent some people having to sell their homes at the point of need.

Next, new eligibility criteria, which would be portable between local authorities, would be introduced to ensure greater consistency and fairness across the system, so that two people in a similar situation but living in different parts of the country would be treated equally and fairly. In fact, it has not quite turned out like that, but that was the recommendation.

People who have entered adulthood with care and support needs will be eligible for free State care immediately. We are not going to talk about those people today.

This year the Care Act, which is the culmination of all the political and other discussions that went on about the Dilnot recommendations, was enacted. It has features that some think go beyond Dilnot but were deemed necessary to make it work. Key among these was providing people with information and advice relating to care and support, given that it is such a complex act with many traps and pitfalls. Another feature is to undertake the assessment process as to whether care is needed and, if so, what those needs are. This will be extended to carers. However, just being assessed for care does not mean to say that you receive it for nothing. It depends on your income, and your assets.

It also only applies to something called “eligible needs”. As part of that process, the local authorities will be required to keep an up-to-date record of the accrued costs of an individual adult’s care, a “care account”, to make sure the cap is not exceeded, or when it is, other arrangements will start.

As part of that, they will undertake a financial assessment, if required to do so. It might not be necessary in every case. In some cases, it will be obvious that a person does not have enough income or assets. In other cases, people will be wealthy and therefore out of the range of any potential State support.

The final option would allow costs to be deferred. In other words, the local authority would pay those care costs for the individual. I am not talking about self-funders here, obviously, but about people typically income-poor but asset-rich. This would theoretically give them time to decide how and when to dispose of their property, either when they die or beforehand, but as those costs accrue the local authority would charge interest on that loan.

The new fly in the ointment is the publication in the past couple of months by The King’s Fund. This publication combines a number of thoughts, some of which have been pursued quite vigorously by the Healthcare Select Committee. One of these is the idea of moving to a single ring-fenced budget for the NHS and Social Care, with a single commissioner for local services.

The complications of how that works and the democratic consequences are significant. It also proposes introducing a new care and support allowance, which I understand would replace something called “attendance allowance”. This would be available to people with low to moderate care needs. That would be at an earlier time than they would be eligible for local authority support, and would be paid through the benefits system.

There is a focus on fairness again: equal support for equal need, independent of how much you can afford.

The most important recommendation, of which the actuarial profession needs to take note, largely rejects private insurance options in favour of public funding. This is a key aspect of the Dilnot Commission because it says that putting a cap on long-term care costs would make it much easier for the insurance industry to insure up to that cap.

The King’s Fund’s work largely rejects that, and says, “There are no signs that private insurance products will emerge to cover the costs of care up to the Dilnot cap and this underlines the need for a predominantly public solution to care adding that insurers show little appetite for producing the new products that would be needed”.

That is one of the key topics for discussion after we have heard from Dr Adams.

Finally, insurance options are just one of several ways of dealing with this problem, and not necessarily the best way.

We produced this map a few years ago.

It is widely used in the industry and elsewhere. It is an asset–income map of the population aged 65 and over. The contours represent people in equal circumstances, based on their assets and their income.

The variation in people’s post-retirement income is quite small relative to the variation in their assets, where assets might consist of housing wealth and pension wealth, and so on.

This introduces a new factor because it means that there are different ways of setting the long-term care cap.

Another thing that we have been working on, which has received much publicity, is different ways of meeting the needs of people in different parts of this income/asset spectrum.

For example, at the bottom of the map (A) are people on low income with only moderate assets. How do you get them to put money aside for their long-term care needs? One of the ideas we came up with was “personal care savings bonds”, which has been published and presented to ministers. The basic idea is that it is like a lottery. People buy bonds similar to Premium Bonds regularly throughout life. These bonds, unlike Premium Bonds, earn interest, but they also pay prizes. You cannot encash those bonds until such time that you are assessed as needing care. We found that was a good way of engaging people in that part of the asset spectrum into putting money aside.

These people are not going to be home owners, so they are not going to have the wealth of the second group, B. These people have little income to fairly substantial income. Most importantly, they have fairly substantial wealth, certainly above the average.

There are two hotspots in that income–asset spectrum. The one at the bottom represents the people most dependent on the welfare system. The hotspot in the middle is the home owners, and assets here are calculated on an individual basis rather than a household basis.

The people in B have substantial housing wealth, which offers the potential for conversion into products such as an income stream from equity release or, more normally at the moment, immediate needs annuities, which are available for people at the point of need if they want to go down that route.

We called category C “disability-linked pensions annuities”. Annuities are in the news – there is much debate about whether the annuity market will decline or change. “Disability-linked pension annuities” is a term used by my colleague, Mr Ben Rickayzen.

The idea is that if you fail one or more activities of daily living (ADLs), then your annuity increases by an amount and that extra amount could be theoretically used to pay for some of your care needs. The problem with this is that, for most people, pension funds are relatively small. The average is roughly about £25,000–£35,000, which is not going to buy you a big annuity.

If you converted this amount into a disability-linked annuity, you might struggle. But there will be people on that pension spectrum for whom this could be a very good product.

Then we come to insurance products (D), which are there to protect assets, possibly housing wealth, or other assets The products for long-term care are quite expensive, so you need to have a fairly substantial income in retirement to be able to afford them.

Then there is the final group (E), the people that have considerable incomes and who may be able to self-fund, who are not affected by any of these other considerations.

There is a further complication: the means test. The means test boundary introduces all sorts of incentives and behaviours into the market and whether or not people will enter into an insurance contract or otherwise.

The basic problem is that if you enter at some point in your care journey, beyond that boundary you might be ineligible for some State support. So you may cut off your options if you enter into the wrong contract with a provider.

What has become clear from our work is that there are multiple product ideas for different segments of that market. This year, Mr Thomas Kenny and colleagues produced a paper with many ideas about how to use the value in your pension to help pay for your social care.

(Author: Products Research Group of the Pensions and Long Term Care Working Party

Source: How Pensions Can Help Meet Consumer Needs Under The New Social Care Regime

Publication date: 07 May 2014)

There are other product types mentioned in publications. There is a piece of work with DEMOS that we published earlier this year with a couple of new ideas.

Because insurance is expensive, one idea is that you might use some of the equity in your home to pay a one-off premium at some point, maybe at 60–65 years of age. If that insurance was triggered, the insurance company would pay your care costs. If you died without needing any long-term care, then the insurer would receive that proportion of equity that you had committed to buy that product.

In modelling, we showed that the amount that you would have to give up is reasonably small relative to house prices today.

Another idea was for people to ring-fence some of their assets into what we called “care accounts” at some point near or just after retirement in return for entitlement to more State support if care were needed. The idea here was to induce better forward planning, which is currently lacking.

Also it may be possible to expand significantly the equity release arrangements that exist. The market is much smaller than its potential. We saw a role here for the State in providing a different and perhaps fairer playing field for the kind of product in which housing equity would be exchanged for additional income in old age to supplement the State and other pensions that people had forgotten to save for.

Then there are some familiar products: immediate needs annuities, which companies like Partnership and Just Retirement offer. The idea here is that on going into care, you can exchange a lump sum, perhaps the value in your home, for an annuity that will pay your care costs until death.

I have already mentioned personal care savings bonds, which were launched by the ILC about a year ago.

What are the expected personal costs? The table from Mr Kenny and his colleagues’ paper shows different costs based on where one lives.

This does not include hotel costs, which tend to more than double the figures. It shows that the cost depends on how long you live once you go into care and is a skewed distribution with the costs increasing the longer in care.

Another chart taken from the same paper shows the approximate probability by age and gender of reaching the cap. Probability declines with age and is heavily skewed towards females.

What is happening to life expectancy? This is a new way of thinking about life expectancy and how we can foresee the future.

One of the problems of projecting life expectancy and mortality in general is that the projection methods tend to be unbounded, that is, if you extend your regression line it goes off into infinity and the other way goes through 0.

We have used a concept that we have refined and developed based on what we call 10-year partial life expectancy at 10-year age intervals. For this method life expectancy cannot exceed a 10-year limit. You can project life expectancy over the next 10 years in any age interval and see to what extent it is likely to hit that limit over a certain period. It turns out that these logistic curves neatly fit the time series.

There is a number of things that come from this. There is potential for life expectancy to grow further and faster, particularly in the age group 70–90, which in theory ought to make the market for long-term care insurance and such products more attractive, but also potentially more expensive in some cases.

In the short term, as I mentioned, most advances would be in the 70–90 age bracket and, not shown in the graph, men will converge with women in terms of life expectancy, which has its own implications, because it means that households and partners should stay together longer.

We call this the “jam jar” model of life expectancy.

The process can be likened to a series of jam jars that gradually fill up over time. Since 1990, they have been filling up much more rapidly. The colourless part of the columns shows what we call “available years”. What is striking is that there is not much movement in the centenarian age group. All the key action occurs between 70 and 90. All the concerns about being overwhelmed by centenarians may not come to pass based on this model.

We are focusing on issues facing the long-term care insurance market. Dr Craig Adams is covering the issue of adverse selection and the genetic predisposition of individuals, and the impact of those features on premiums.

The Chairman: Dr Adams has a degree in Actuarial Maths and Statistics from Heriot-Watt. He then went on to work for Friends Provident, working on their protection business. He joined Scottish Widows in 2008. In 2010, he went back to Heriot-Watt to study for a PhD, passing with a thesis titled “Adverse selection insurance markets with genetic information”.

His first paper “Impact of known breast cancer polygenes on critical illness insurance” was published in the Scandinavian Actuarial Journal in 2013. To finish the actuarial exams, he decided to go for SA0 rather than SA2, with a dissertation based on part of his PhD leading to this talk. He is now working at Aegon, performing experience analyses to produce best estimate assumptions, and has passed his remaining actuarial exam.

Dr C. J. Adams (introducing the paper): Genetics is a sensitive subject, particularly in insurance, with fears that it might lead to a genetic underclass of uninsurable lives. I am not talking about the ethics today but getting a quantitative measure.

The UK has a moratorium on the use of genetic information, so it cannot use it in underwriting apart from Huntington’s disease and that is only for life policies over £500,000.

With this there is the potential for asymmetric information, which could lead to adverse selection. This presentation is about the costs of that adverse selection. It is not the first study of it: there have been many previously. There has also been a study connected with genetic information in long-term care insurance. That was in an established market so the adverse selection had already been absorbed into the premiums. This is in the context of a start-up market, so the premiums will gradually come into it.

Long-term care insurance is not a product which is sold, so here is some background. It is a product that covers the cost of care in old age and is to indemnify for the costs that have been paid out, which might be limited by an annual payment or overall sum assured.

The Government has proposed a cap on lifetime care costs in order to remove some of the tail risk. But it is covering only the care costs and not the hotel costs, so there is still some tail risk.

The reasons for triggering a claim are twofold: functional disability – measured in ADLs unable to be performed – and cognitive impairment (dementia).

The main cause of dementia is Alzheimer’s disease that has a strong genetic component. There are several known genes now, but only one has sufficient epidemiology from which we can gain any sort of modelling – the apolipoprotein E (APOE) gene, which has been established as being the largest genetic factor. There are other causes not related to Alzheimer’s disease. We group them all together as “non-Alzheimer’s disease”.

Some previous work that I draw from is by Macdonald and Pritchard, who studied the adverse selection in long-term care, looking at dementia and, in fact, only at Alzheimer’s disease, including the genetic aspect.

They had very limited data from the post-dementia transitions, so Colgan looked at other aspects building on Macdonald’s and Pritchard’s transition intensities using a larger data set. This still had no functional disability.

Akodu’s thesis in 2007 was not about a model of insurance. It was about modelling disability into the future, but he had included functional ability based on CFAS 1 (CFAS 2 had not yet been released). The thesis does not mention genetics or split out dementia.

If you put them together, you have a map of the model (see figure 1 in the paper). The numbering is as follows. The first digit indicates the cognitive function type and the second digit denotes the functional ability type.

The assumptions are that it is Markovian, and the transition between cognitive function types is independent of functional ability.

There is no recovery from cognitive decline, but there is recovery from functional disability, which is often ignored. In Pritchard’s later work he found that it does make quite a difference to the cost, so we are including the recoveries.

This model is not about trying to reflect old age perfectly. The aim is to get a baseline model to see how adverse selection causes a divergence from that baseline, and how the divergence creates a difference in the liabilities and premium income.

For each cognitive type we have a base mortality, then we have a relative risk of mortality that follows the functional disability. That was fitted to Akodu’s thesis. He had separate mortality; we are just simplifying it.

For mortality improvements, we used the CMI 2011 model, which was available at the time. That was for projecting 1-year probabilities. We had to derive a method to work with force of mortality. It is entirely consistent.

The onset of dementia is fitted to a European-wide study. It is a pooled analysis of various studies, which all used the same criteria. In these EURODEM studies, the largest coverage was British, so it should be a reasonable fit for the purpose for which we are using it.

They set it up with exposures and occurrences, so we were able to use a maximum likelihood method, which is an improvement on what was done in Macdonald and Pritchard’s work, which extrapolated from some moments.

The genetic model that we are using is from Macdonald and Pritchard (2001). They have a model of the different genotypes – what impact they have – and it is age dependent.

There are at least three variants of APOE, which give rise to six potential genotypes – you have one gene from each of your parents. We have a specific relative risk for each of those genotypes. Actually, it is five when we group two of them together, as one of the groups is very small. That gives a means of calculating the overall dementia onset.

Now we change the model a little. We insert a state where the initial signs of dementia have been noticed by the customer or perhaps their partner. They maybe have memory loss or more difficulty in understanding and reasoning in daily life. This is before they have been clinically diagnosed. This is to give the possibility of adverse selection from a life, being able to tell that they have initial signs. At this stage it would not be picked up by an underwriter in an insurance company.

That gives us a problem in fitting the intensities. We have people who have been diagnosed. We have to try to go from a two-state to a three-state model, although for mathematical simplicity we have assumed the rate into “initial signs” is the same as the rate out. That reduces it down to a differential equation. It will actually explode at time 0. Assuming that it is equal over very small time period h, we get the initial condition and then we can solve it.

Now we apply it to insurance. An insurer, when underwriting, is going to group everyone. The insurer does not know what state of health the customer has. As it experiences some claims, it can gain some knowledge from them.

There are three key events – inception, a claim for functional disability and a claim involving dementia – and death. At inception, the insurer just knows that they are in one of a group of, say, four states. It does not know which one. It assumes, based on the pricing basis, what the general mix is.

After there has been a claim for functional disability, the insurer knows that they have failed at least two ADLs. Because it is a Markov process, it can model, from that point on, the functional ability type, moving up and down through the rows.

It does not know still what cognitive type they are. It could be one of three. There is the genotype as well. Once it has a claim for dementia, that information is irrelevant. It does not matter what genotype they had or what state they started in from the point of the dementia claim. Everything is irrelevant. It is just about modelling forwards from now. It is a Markov process so it can do that perfectly. In reality it would be different.

By way of an example, and referring to figure 6 in the paper, consider a life which started off in state 5. Let us say that the insurer was underwriting so that states 0–6 were all in one underwriting class. The insurer sees it as one block – it does not know where the life is in that block.

But then this life claims and the functional ability type is state 7, at that stage has moved into period C. When they recover, they go about their life and then they claim again in state 7.

That does not give us any additional information because we already know about the functional ability – we are modelling that already. When we move into the dementia-type claim, it gives us the knowledge that the life is now in period D.

The payments under a policy, as I mentioned earlier, could be restricted by some annual limit or an overall sum assured. The Government is imposing a cap on the care costs but not the hotel costs, the hotel costs being those for room and board.

Because it is separating the Markov nature of moving around and the care needed from the actual payments, it is now part dependent, so it becomes difficult to solve. What I have suggested doing is simulating life histories. We simulate moving around the health model. We can then attach payments to whatever state the life is in, paying a premium or receiving a benefit, so we can then find the present value of them. As we learn what period they are in, we can adjust the premiums accordingly.

In order to do that, we record all the benefits that are in the period P and later periods. Anything in period C is just the functional disability, and then anything beyond that is in the dementia period, and then the period D, which ends in death.

If you are simulating a large number of lives, you take the average of that to get an expected value.

In another model, we have a life that starts off untested and uninsured. They then become tested and find they have a genetic predisposition for Alzheimer’s disease. There would therefore be a different rate into the insured status than before they were tested. Then we have a state for being insured.

Premiums are calculated conditionally on having insurance, so we get a conditional probability. This is determined by the rate of insurance purchased according to whatever state you are in and what genotype you have. That therefore forms the experience basis.

We then obtain a base premium. It will be adjusted because, due to the European Courts of Justice ruling, we must have a unisex premium. Assuming we recognise the ratio of males to females immediately, we can scale up for the proportions of males and females. Then you have a summation of the benefits, otherwise the premium would be on the pricing basis.

These are dynamic premium rates. The insurer observes its claims experience and through that it can better estimate who is buying the product, or it can adjust the premiums by some amount based on what it is seeing in the claims.

The insurer will be doing that by looking at what it was charging at the outset and, given this new experience, what it thinks it should have been charging. So it is changing its expected future amounts. It can see what the actual benefits are and change its future expectations.

If you think of it as a stochastic process, then the claims experience becomes your filtration. Conditional on this filtration – your claims experience – you have the actual benefits and you have the expected future benefits. You can then work out what the actual benefits are. Obviously, this relies on good claims data. Because this is actual claims, we can use the experience basis to work out the functional disability mix. In the first period, period P, when there have not been any claims for this block of lives, the insurer just had the pricing basis to go on for the mix of business, but the volume of business will not follow that basis. That is going to be changing, so they have to scale it for the volume of business that it has.

Period C is more complicated because we have knowledge of the functional disability but not the cognitive function. We model from the point of the claim, but we do not know the mix of the functional disability type.

The functional disability is taken out of the final summation, scaling up according to the size of each of the functional groups.

In period D, everything is known. It is very simple. The calculation is done using the experience basis.

From these expected values, we can calculate our re-pricing adjustment factor. The make-up of the market will be changing over time so the older business will not be as relevant. The costs that have been incurred from adverse selection from that business will not be as relevant either. We have a simple weighting factor, which is just a reverse sigmoid-type function to attribute lower weight to the older business and larger weight to the newer business.

Summing up over everything, we can obtain the adjustment to the base premiums, which applies from after the 1st year. In the 1st year it is equal to 1.

We now apply it to the UK market. As has been mentioned, there is currently no deferred-needs cover for sale in the UK. The situation differs in England, Scotland and the rest of the UK.

In England, Wales and Northern Ireland, it is means tested, but care is paid for by the individual. The means test is currently at about £27,000, I think. Above that, you would probably have to sell your home. In Scotland, the Scottish Government took the advice of the Royal Commission in 1999 to pay for the care costs and does so through a fixed contribution. But it does not pay hotel costs.

The UK Government has proposed a cap. What I have used in my work was what was previously suggested, but it has changed since then. It was £75,000 from 2017. They have brought it forward a year to make it £72,000 in 2016. This work was all done using the £75,000 from 2017.

A table from the Office of National Statistics gives costs of care by region, weighted by population. The number of hours of in-home care provided was not so clear, so in the paper it is left as a scenario-specific parameter (H). Using H equals one did not really make a difference to the shape.

The graph on the top left of figure 9 of the paper shows the premiums that would be charged at the purchase time. The market starts at 2013, when I was doing the research, so this represents the number of years since 2013.

The graph on the top right shows the premium rates that would be charged if the insurer knew your genotype, initial state and from whom you are buying your insurance. The thick lines are what would be charged in a small market aggregating everything. It is what it would be if males and females were separate. The blue line is for unisex, which would be charged in the UK. These are only for those with no functional disability. There is more in the paper. There is a large discrepancy between “initial signs” and “healthy”. With initial signs you are obviously closer to a claim.

The premiums for females are substantially higher than those for males. Also, when you are healthy the genetic aspect does not make so much of a pronounced difference as it does when you are closer to developing dementia, which makes sense because you still have an extra step to go through in order to develop it.

It would be to your advantage to buy insurance early on when the premiums are low, if you have the test knowledge then.

If we have the adverse selection costs, what would be the premiums adjustment? And what adjustments should be made if we look at each source of adverse selection in isolation? The black line on the right-hand side at the top of figure 10 in the paper represents people with initial signs of dementia buying at an increased rate.

But that is not recognised very quickly on the left. The amount by which premiums are being adjusted is going up very slowly. If the insurer was not able to underwrite for functional disability for some reason, then quite a high adjustment would need to be made to the premiums.

If healthy lives were not buying the product, it is a similar level to the cost related to ADLs, but it is recognised a little slower. That is probably the outcome. If people do not value the product, they are not likely to buy it.

The two lines at the bottom relate to genetics. If you have low-risk genes – an E2 type of the APOE – and you do not buy the product because you think you are at low risk, it is not going to add much to the premium. If you have an E4 type variant of APOE, it will add something to the premium, but not as much as the others. Again, it is recognised slowly.

The progression of the costs of adverse selection is essentially the right-hand side divided by the left-hand side. The lives with initial signs and with one ADL buy at a higher rate (see figure 11 in the paper).

In the smallest market we have the largest costs, which makes sense because, relative to the size of the premium, you are not going to have many healthy lives. You are not expecting many people to be buying the product. Then you have a high rate of poor lives buying it. You have substantial costs that is recognised quickly and is the quickest to adjust.

If males are paying a male-only premium, the premium would be a lot lower and that gives rise to a larger adverse selection cost than it would do for the females.

We now add in the genetics (see figure 13), and the behaviour changing after a genetic test. Lives with the high-risk variants, the E4 type, buying at a higher rate, and lives for E2, choosing not to buy. This is before they reach initial signs or failure of one ADL, but it would cater for that as well. This reduces the costs somewhat. The reason for that is that lives with the higher risk genes would reach the initial signs ultimately, but they are buying earlier, so the costs associated with that are lower. That causes the adverse selection cost to come down.

If we say that functional disability can be underwritten, as is likely, we change the pattern a little more (see figure 15). The slight hump is now more pronounced and the costs have increased. You have fewer people, but the costs have increased. That is because the premiums will have decreased because you do not have to include those people who are close to a claim in the premium. You now have the adverse weighted over a smaller premium, which puts the cost up. The hump is because as the youngest lives become older they are more likely to be in the initial signs and it makes them more pronounced.

Professor Mayhew has already mentioned the probability density function for the care costs. In reality it could be different. The proportion of lives reaching the cap is fairly low. It is not benefiting a huge number of people.

Here are my conclusions. Adverse selection costs associated with the initial signs of dementia are the highest and the slowest to emerge so that is going to be where the greatest risk will be.

On its own, the inability to use genetic tests brings some adverse selection costs, but they are small relative to those associated with the initial signs of dementia. When genetic testing occurs with other sources, it brings adverse selection forward to the point where the cost of it is lower.

Finally, the Government’s cap does not benefit most people. But it could limit some of the tail risk. We would have to analyse that. That would be another study. This is just looking at the costs of adverse selection, not the risks.

Mr P. J. Banthorpe, F.I.A. (opening the discussion): I am unclear about some of the assumptions, because, when testing the assumptions, it is not clear whether there is an overstatement or an understatement of anti-selection costs. One assumption was that the model is Markov with dementia being a degenerative disease. How does that stand up to reality?

Also, the authors correctly identified that insurance companies are not the best with their data with some real issues around calendar reporting lags. You do not get data immediately. Does that allow the anti-selection problem to grow bigger before you can deal with it?

My final point is about lapses. Are you assuming that this is a single premium product so that the effect of anti-selective lapses is not material?

Dr Adams (responding): On your last point, in the paper I had single and regular premiums. A lapse and re-entry would be perfect for regular premiums. A single premium would be very large.

In the model, the Markov assumption simplifies the situation quite substantially. The reality is that we would expect something slightly different. I do not know how it would impact upon the adverse selection costs. This was about getting a baseline, so we can say the genetics is not going to be that much of a big deal compared to the initial signs. I would not suggest using this as a pricing model or using it for reserving. It was to bring the genetic component into the debate to see whether it should be usable in underwriting.

In respect of the data lags, it would be nice to have some sort of IBNR adjustment. It would delay other adjustments being made and increase the costs.

If we have adverse selection occurring in the market, the policyholders are likely to submit their claims. How much IBNR would be, I do not know.

The Chairman: I am sure with all businesses there are late-reported claims. Sometimes it is down to the individual and sometimes it is down to the administration staff.

Mr C. Harvey, F.I.A.: You indicated that there is a significant genetic predisposition to some of the dementia conditions. To what extent is underwriting on family history any protection against anti-selection for those conditions?

Dr Adams (responding): Family history is unlikely to be sufficient. The knowledge of genetics in respect of Alzheimer’s disease is not well known. There are many different genes expected to be at play. Parents do not necessarily pass on their combination of genes. They pass on half of the combination, but because there are so many different genes, one plays off against another.

The Chairman: May I ask, as chairman, the audience a question about individuals being able to select against an insurer? How good do we think insurers can be at spotting anti-selection compared to an individual? Is an individual going to be in a better position to spot the early signs of dementia, or is an insurer going to be able to come up with appropriate underwriting tools and tests to enable it to spot it?

We certainly know that sometimes, in terms of life insurance, insurers will spot problems before the individual does. Does anyone have any views on that?

Insurers have attempted to deal with cognitive impairment by asking memory recall questions. A classic way used in the UK and in North America is to ask somebody to remember five words and then have a conversation with them and ask them to repeat the five words afterwards. Do you think individuals will spot when they have difficulties?

Mr J. Parmar, F.I.A.: My understanding of current research on dementia is that there is a great deal of effort going into testing for early onset, with the emphasis on looking for chemical changes ahead of the behavioural ones.

Cost is going to be a factor, and how easy it is for individuals to have access to the testing compared to insurance companies and their difficulties in using it for underwriting or adverse selection.

The Chairman: Under that scenario, the insurer may be in the better position to judge the early signs if they were able to ask for the test and to interpret the result rather than the individual.

Mr J. B. Hurley, F.I.A.: The products that are sold at the moment are immediate needs annuities, in which case, the onus is on the customer to “prove” that they have a medical impairment and how severe it is.

In terms of whether customers will, I agree with Mr Parmar. It is going to depend on the cost and how easy it is to get hold of that information.

There is a risk that insurers should be aware of, and I suspect they will not react as quickly as they probably should.

Mr J. A. Jenkins, F.I.A.: I would be concerned about this risk of selection. The individual in question may not spot that they are at risk, but their husband or wife might be well placed to spot that they are observing changes in their normal behaviour. They might be able to take out a policy and therefore select against the insurance company before it is able to pick it up with whatever tests it has available to it.

Mr Baker can you summarise what is the state of interest and activity in the UK insurance industry and the reinsurance industry in terms of this sort of business? Other than the impaired life annuities and immediate needs annuities, is anybody writing this business at the moment, and underwriting these risks, or is it just a product that we think will come of age at some time in the future?

The Chairman: As far as the UK market is concerned, to the best of my knowledge, nobody is selling a pre-funded product today, but they did so back in the 1990s. They stopped selling it because of the difficulty of distribution, not because of the difficulty of experience.

However, I believe one of their early pricing assumptions was that interest rates would be around 6% or 7%, so perhaps over 20 years that has not quite worked out exactly the way they expected.

But pre-funded contracts are widely available in North America. They are also available in France and Germany. There are other markets where these products are available.

My view is that the problem with pre-funded products is getting the right distribution. That is one of the reasons why the product has disappeared.

Mr Hurley: The way I see it, and Professor Mayhew hinted at it when he was talking about The King’s Fund, they are already changing the regulations. Would you really expect insurers and reinsurers to develop products and people to sell products in the long term, given the political risks and how quickly the regulations could change?

There is no tax incentive. Products are inflexible. There are plenty of other products around that would be more flexible and more tax advantageous for the customer to use for saving.

The Chairman: That is a good point. In some of the markets where it is successful, it is compulsory or there is some other push from the state to require you to do so.

Prof Mayhew: I do not believe that long-term care, and periods of social care leading up to periods in institutional care, will ever be completely free under any of The King’s Fund scenarios. If you read the small print, they were talking about only those people with the most critical and possibly substantial needs. For low and moderate needs, they clearly see that people should continue to support themselves one way or another.

It would put a big brake on our thinking about this market and which would be the best products to meet all possible needs. It is clear that the system needs more private money injected into it because public funds are very tight and budgets are being cut everywhere.

It may be that we have to think more carefully about the products that would be most suitable and about the potential interactions with what the State will provide.

For example, you do not want to take money off people who have saved for care because the means test says that they deduct, pound for pound, against their income. That would be a big mistake. It will take all the wind out of any market for private long-term care products.

Mr M. Martin, F.I.A.: I have a couple of questions. In the application form, could we ask the spouses if they have noticed any difference in the behaviour of the other party? Given the size of these premiums, is there not a case for face-to-face underwriting where, rather than simply an application form, we can test maybe their physical ability, memory loss and suchlike. There are probably sufficient margins in the premium to justify such an approach compared to the available expense loadings in a £20 term assurance policy.

The Chairman: In markets where it is sold, they have telephone interviews. They do not provide it purely on the basis of a written application form, which you would be happy to do with many other products.

Any thoughts on why not use a nurse and assess the risk? Could you manage the adverse selection sufficiently?

Dr Adams: That may put off people. If it is already a product which is difficult to sell, and you are adding this layer of intrusive underwriting, then it is going to put them off even more.

The Chairman: If you are 70, would you not like somebody to come round and have a cup of tea with you?

Prof Mayhew: I do not know how it works in the US market, but I have looked at websites where you can apply for insurance online. I do not know whether that is the beginning and end of it or whether there are other hoops you have to jump through before you get a quote. Does anybody happen to know? Or have you investigated any of these websites?

The Chairman: In the US market, it is quite common to have a medical examination for standard term insurance. They rely more on blood testing, and you cannot do that over the phone or online. So they have people that will come and visit you and take samples.

Mr T. P. J. Kenny, F.I.A.: In the US, I understand that they do nurse screening for pre-funded long-term care. They will send someone along and they will take you through a variety of tests, including a dementia test. They will probably take a few samples as well.

The Chairman: And they have that nurse screening system set up for even basic term assurance, so they already have the infrastructure in place. Obviously, it is a different set of tests.

Mr Kenny: The risks involved in pricing them accurately, given their long-term nature, are quite substantial. The costs also seem very high.

I can see pre-funded long-term care being a rider on some other products where it would provide some sort of top-up insurance. But I cannot see it being pre-funded and providing indemnified cover or anything approaching the true cost of care in the UK. In the US, the pre-funded market, as modelled here, is declining. The cost of insurance has gone up. There are only a small number of providers in the US that offer it now.

The products that they are offering there now tend to be combination products, combining savings with accelerated claims, so if you develop a condition that qualifies, it may pay out an earlier benefit, like accelerated critical illness or accelerated long-term care.

Prof Mayhew: Adding something to that, Dr Adams do your premiums presume hotel costs as well as personal care costs? Would it not be much cheaper if you separated out the hotel costs and perhaps assumed that the individual can afford them out of their normal income in retirement?

Dr Adams: It does include hotel costs and care costs. The policy and type of cover does provide for those. I was trying to reflect what the product does. I did not consider stripping the hotel costs out.

I did do a version to reflect what is occurring in Scotland, where you have only the hotel costs, but I cannot recall the results.

Dr Adams: The graphs were of the first selection cost rather than the premiums. I cannot recall what the situation was.

Prof Mayhew: That kind of product may make it more affordable for more people. The average income in retirement for the people we are talking about is probably in the region of £20,000 a year, which would be sufficient based on figures that I have seen, and what the Government suggests hotel costs might be in the future, which is around £12,000 or £13,000, I believe.

Dr Adams: If it still meets the customers’ needs, then it sounds like it would be a viable option.

Prof Mayhew: But the top-up idea must be one of the ways to go with all of this because people will have multiple sources of income. They may have an informal carer, they may only need to cover holiday periods and so on during the year.

So there is scope for creative thinking. You might be able to launch a product that gives you, say, 100 hours of care per quarter.

The Chairman: It is a challenge. What do you think has changed so much that 15–20 years ago major household names – the big insurers – were quite happy to enter the pre-funded long-term care market and, to the best of my knowledge, they all withdrew simply because they could not sell the volume of business needed to justify that level of investment. None withdrew on the basis of claims experience. Are they all more risk averse today?

Dr Adams: There is much coverage in the news for the reform of the care sector, so people are quite unaware that the Government does not cover the costs. If they are unaware that there is a gap there, then they would not buy the products. However, I suppose the Royal Commission in 1989 would have received coverage.

The Chairman: The one thing that we have learnt over the last 15–20 years is that you cannot trust Governments. It was the Royal Commission in 1989 that persuaded some of the big insurers to enter into this market because they thought that they would sell volumes and that it would be a key line of business.

Mr Kenny: The other challenge with a pre-funded market is the level of political risk. The regime is changing and even with the regime that they are planning to introduce, there is still uncertainty around exactly how it will work. For example, the definition of eligible needs is still to be finalised. That is a potential political tool. The Government could play around with that definition at any point in the future. If you are trying to design a pre-funded product that will pay out in the future to someone deemed as having eligible needs, you want some certainty that that will not change dramatically, otherwise you are opening yourself up to complaints from policyholders saying “I bought this insurance thinking it would cover me in that situation. Now it does not”. That is a huge risk.

Now, compared to 15 years ago, we have Solvency II and risk-based capital. You have to understand the risks you are taking on. In offering a pre-funded product you have investment risk and a huge claims risk. Someone would buy this product at 60, and not expect to claim until they are 85. That is a long time. I am not aware of any particularly good data available in the UK on care.

It is one of the challenges that we are trying to address: having good data sets that can help insurers design and price products accurately.

Prof Mayhew: I will add two other reasons to try to answer Mr Baker’s question? Surveys show that people think that social care is free. It has taken many efforts – not successful efforts – on the part of the Government to change that perception. That is because the system is complicated. People receive free healthcare and confuse healthcare and social care.

The second reason is that it is expensive and it would only suit quite a narrow sliver of the population based on my analysis of the income/asset spectrum.

Mr Kenny is right. There is also a problem with these long-term contingent products because if the environment changes, then they could become a bad purchase and then the insurer would be blamed, not the Government.

There are four levels of eligible needs at the moment: critical, substantial, moderate and low.

Maybe somebody could correct me, but the Government is thinking of setting it at critical. There will be many local authorities that will offer substantial as well as critical. So there is an element of geographical variation around. Which of those levels insurers would recognise – critical or substantial – must also be an unknown and difficult to judge. I am not sure that these kinds of contingent products are the best way forward.

Maybe savings vehicles are the best way forward. It is much more difficult to take away your savings (except by inflation) than it is to renege on your commitments, whether it is your fault or not.

Mr C. Swatton, F.I.A.: On your point around the culture of thinking that social care is free, that is something that has changed in the last 20 years.

The people who are approaching retirement now are the people who have seen their parents living much longer lives in retirement and potentially having to sell their homes to fund care. There is the culture of home ownership in the UK that means there is a desire of people approaching retirement now to not have to sell their homes. They want to hold on to them to pass them onto the next generation, having seen their parents go through that situation of having to fund their long-term care, whereas for the previous generation 20 years ago that was perhaps a new situation coming through.

The Chairman: In earlier comments, there was discussion about pre-funded long-term care being popular in France, while it is no longer available in the UK. A possible explanation is that under French law the children have a legal responsibility to care for their parents.

On that point it remains for me to express my thanks to our speakers.