Canadian insurers should study their methods of analyzing data credibility in depth, a survey of two actuary associations finds.
In Canada, insurers mainly use the Limited Fluctuation Credibility Theory (LFCT) method to assess the credibility of experience data, a report called The Application of Credibility Theory in the Canadian Life Insurance Industry concludes.
A joint initiative of the Canadian Institute of Actuaries and the Society of Actuaries in the United States, this survey of the 11 main life insurers in Canada explored the credibility methods used by actuaries at these life insurers.
The application of credibility theory is often needed to evaluate the appropriateness of assumptions such as mortality and lapse levels for a company’s block of business. In addition to LFCT, insurers also mentioned using the Greatest Accuracy Credibility Theory (GACT).
Quest for simplicity
The main finding: nine of the 11 insurers surveyed said they use the LFCT method to determine mortality credibility. The researchers noted that most of the firms surveyed favour this method because of the availability of CIA guidelines and because of its simplicity. One insurer said it still uses this method because it has insufficient internal resources to implement new ones.
In fact, this simplicity masks a complex universe. Most of the insurers said that they apply the “by number” (or by policy) approach to calculate credibility factors, while a small number of companies opt for the “by amount” approach. Four of the insurers surveyed said they use the normalized method to calculate credibility factors by subcategory.
Ideal and real world
What is credibility theory used for? In an ideal world, an insurer would be able to rely entirely on its experience studies to establish actuarial assumptions like those concerning mortality and lapse (policy cancellation), the report authors explain.
Yet the picture is quite different for insurers whose experience is not available or sufficient. These firms need external data sources or their own judgment to make reliable assumptions. This is where credibility theory comes in: it helps them determine whether their experience data are “fully credible” or “100 % credible.”
If the data is fully credible, insurers can use it to formulate assumptions or create tables. Otherwise, credibility theory can combine the insurer’s experience with that of the industry, for example in the form of a prescribed valuation table for mortality rates. This table helps insurers make a more accurate estimate.
A flawed method
The Limited Fluctuation Credibility Theory method nonetheless has drawbacks. The main one, the report says, it that lacks a solid theoretical base. On top of that, the LFCT method requires industry experience that is not always available. Insurers then have to turn to other sources of information or an actuary’s judgment rather than apply credibility theory to partly credible data.
The report also points out that in some cases, industry data is not fully credible and the insurer must then apply more weight to its internal data. In other cases, subpopulations (cohorts of insured) on which the assessments rest lack credibility and have to be normalized using a larger population, the researchers explain.
Laws of large numbers
Four insurers said they use 3,007 deaths as the criterion for full credibility, versus 6,014 for another insurer, which maintains that the 3,007 number assumes homogeneous policies, and may thus underestimate true full credibility. This insurer adds that the CIA approach uses 6,014 deaths as full credibility in order to recognize variability in coverage amounts and age distribution. Other insurers say they are concerned that the number 3,007 would be inadequate when comparing non-homogeneous lives.
One insurer does not apply a credibility method. It thinks that full credibility standards are not high enough. Using the LFCT method places too much weight on the insurer’s data for small numbers of deaths.
