I have found that interviews can often produce interesting results and insights. So, this month, I decided to sit down with Tony King, one of our actuarial analysts, and discuss the popular topic of loss development factors. The conversation has been edited for clarity, but I think you will find it a good review with maybe some new information on loss development factors.
Timothy: Tony, thanks for taking time to be interviewed. We are talking about loss development factors today. So how would you define loss development factors?
Tony: In simple terms, a loss development factor is just a factor that an actuary applies to current losses to estimate ultimate losses. It is used to develop reported losses as of a certain date to what we can expect the ultimate loss amount will be when all claims are closed. There are incurred loss development factors which are applied to incurred losses and paid loss development factors which are applied to paid losses. Both factors are generally greater than 1 as the ultimate losses are expected to be more than the current losses. Furthermore, paid factors are higher than incurred factors because paid losses are initially less than the incurred losses by the amount of case reserves.
Timothy: Ok and those losses could be the losses for the year that just ended or it could be losses that ended five years ago.
Tony: Right, there is actually a table of factors for each line of coverage. Generally, the table is sorted by the age of the policy period in months, so if you have a 1/1/2012 policy at 12/31/2012 you would use the 12 month factor. The actuary will choose the appropriate factor for each policy year based upon that policy’s age since inception. Sometimes there can be confusion about a given loss development factor – you have to make sure you have the appropriate loss development factor for the policy that you are applying it to.
Timothy: If we are looking at a report that shows incurred losses for past years, then this report really doesn’t tell us what those losses will ultimately be? Is that correct?
Tony: Right. The total at this time is what’s been paid and what the adjusters have put on top of that for their expectation, but there is always IBNR on top of that. I should clarify what IBNR is. There is always more development on existing open claims. And, there is also true IBNR, that is, there are claims that have been incurred but not been reported yet, so those claims will be coming in and you have to account for those claims in your development as well.
Timothy: So you use loss development factors to estimate the difference between the adjuster’s estimate of total incurred and some future amount that will truly represent the ultimate incurred losses.
Tony: Right.
Timothy: What do you do if the client cannot provide the historical data necessary to develop a loss development triangle?
Tony: Typically, if a new client does not have historical data, we use “benchmark loss development factors.” For workers compensation, then those factors may be the NCCI published factors. The problem with the benchmark NCCI factors is that they are not unique to that company’s loss development history. Also, benchmark factors often don’t take into account a retention limit.
Generally speaking, the benchmark factors work pretty well for paid losses because variability is low, but on the incurred side (which is generally a better indicator of what the ultimate loss can be) it can be more of a challenge.
Some organizations vary quite a bit from the benchmark. In these cases we really need their loss history so we can compute unique loss development triangles.
Timothy: For those clients that do have historical loss information, what do you need?
Tony: Ideally, if they are doing 2013 year end reserves we would hope to have a loss run from the year 2012, 2011, 2010. Three years of data is the absolute minimum and we prefer five or more years - the more data you have the more credible it is.
Once we have the client’s data, we still start by looking at the benchmark factors. This is because the benchmark factors represent a large database of data from many companies. Then, we look at how the client’s unique data compares to the benchmark. This is where we use our professional judgment on how best to blend the benchmark data and the client’s unique data.
Timothy: If a client does provide loss history, how do you decide whether to use their data and develop unique triangles or not to use their data?
Tony: The larger their loss portfolio, the more credibility they have. However, we often see situations where the client has six years of history but the first couple of years were less than $100,000 of losses. Then year three is when they really start to grow and they show a very significant increase in losses. In those cases, only those last three or four years have good credibility. We will generally include the older years in the triangle in our report, but we give them little if any weight because of their limited size.
Timothy: You mentioned using your judgment during this process. How does that work?
Tony: Creating a triangle is a mathematical process. You use historical loss information and compute age to age development factors (See our Actuarial Advantage booklet for a deeper explanation). Putting the triangle together is the science of it. The judgment or art of the process comes into play when the computed factors differ from the benchmark. Then we consider specific business issues, the amount of data we have, and experience with other similar accounts.
Timothy: When you are looking at different retention levels, say you’ve got a client who’s looking at a $100,000 retention, a $250,000, a $500,000 retention. Does that come into play when you are developing the triangles?
Tony: It depends on the history that you have. But ideally you would have a separate triangle at $100,000, at $250,000 and at $500,000. Certainly the higher the retention generally the higher the development. If losses are capped at $100,000 you probably do not have development in later years because all the development that’s happening that late is usually on a claim that is greater than $100,000.
Timothy: Ok, that makes sense.
Tony: Typically, loss development factors are always greater than one. Early on they may be a large number but they trickle down towards one because as you get fewer claims left open, fewer claims still developing, eventually you get to one where everything is done. Factors do go down over time quite a bit at the beginning and then slower as you move along.
Timothy: What are other uses of loss development factors and triangles?
Tony: Well, one great reason to have a triangle would be during collateral discussions. A lot of times in those collateral agreements the insurance provider will have a loss development factor built into the documentation. Some clients are probably okay if there is a factor like 1.5 in there. They sign the agreement not realizing that based upon their unique history they should be a 1.2, for example.
Timothy: Ok, that is interesting. Tell me more.
Tony: So obviously, if you have the triangle going into the initial negotiation that’s where you want to say where are you getting your 1.5? And they may say well that’s a benchmark. If you have a triangle prepared, you could then suggest that they use a factor more appropriate for your history.
Timothy: What would you say are some of the pitfalls when putting together a loss development triangle?
Tony: Well, one would be the process of making your selections. You don’t want to be either too aggressive or too conservative in making the selections. If you are always picking the max, or always picking the min, you’re going to be too high or too low probably. You want to be in the middle.
Also, on the reported side sometimes you will see some changes over time, especially if there has been a change in adjusters or even something internal; such as a new safety program that can change the way some losses are set up and reserved. And so you can get into a case where losses have always developed this way but now because we’ve got this new program going on or we’ve got a new adjuster that maybe sets up higher initial reserves than the old adjuster used to so the claims just don’t develop as much as they used to. Anything that has happened in your history that may affect development you’ve got to be careful with. Generally what an actuary would do is mark that diagonal and just see does it have an impact on development or not. And of course the problem is that when there is a line like that it takes three years to have enough history to see what it does but during that time you have to be careful that you are not just assuming things are going to develop as they always have, because they can change.
The other pitfall I was thinking of is retentions, you have to make sure you know what retention limit your triangle is. You can’t just take a $500k triangle and say I’m going to apply it to a $50k deductible or vice versa. You have to make sure you use the appropriate retention for the policy. If you’ve been at a $250k retention but then shift to $500k, that triangle is now no longer any good going forward. We would have to create new triangles at the new limit.
Timothy: So you would have to go back and get the unlimited losses and limit them to $500,000…
Tony: Correct, and recomputed the triangle because it is likely that $250k triangle doesn’t have enough development in it. Any claim above $250k was being capped before and now it won’t be, and so if I continue to use those triangles I would understate development.
Timothy: I understand.
Tony: It all depends on how many claims you have in that layer, if you’ve never had a claim above $250k then there’s no real impact. When a company changes the retention, we always look how many claims are in that layer and if there’s enough we have to restate the triangles.
Timothy: Are there any other uses of the unique loss development factors we develop for our clients?
Tony: Yes, there’s a couple things I would think of. Once you have a loss development triangle, those factors can be interpolated into monthly factors. From those monthly factors, there is an implied development each month. Once I do the year-end report for the next year, I can show the client what their implied development is for each month throughout the year. They can then compare their actual losses each month to the losses expected by the factors and determine if their losses are developing more or less than expected. So that’s a great use for it.
The other thing would be for the client to compare their loss development against others in the same industry. You certainly could easily compare yourself to a benchmark but it would be even better to compare yourself to competitors.
Timothy: Thanks, Tony. I think we will stop there. This has been a great conversation to provide an unstructured conversation about loss development factors. If our readers have questions, how do you suggest they contact you?
Tony: They can email me at Tony@SIGMAactuary.com or call me at 615-376-5110 x208
If you haven't already, please take a minute to register for our free resource portal at www.SIGMAactuary.com/resources.
© 2014 SIGMA Actuarial Consulting Group, Inc.