In a common captive situation, a fronting insurance carrier issues a policy or policies to the captive’s parent company and reinsures some or all of the risk to the captive...
In a common captive situation, a fronting insurance carrier issues a policy or policies to the captive’s parent company and reinsures some or all of the risk to the captive. While the structure is quite straightforward, the rules for determining whether the policy qualifies as reinsurance for accounting purposes can be complex and difficult to apply.
For example, under US GAAP, specifically FASB Statement 113, a contract must transfer significant insurance risk (both underwriting and timing risk) and the contract must subject the insurer to the reasonable possibility of a significant loss in order to qualify as insurance or reinsurance. In a similar vein, the National Association of Insurance Commissioners’s (NAIC) Chapter 22 and some international accounting standards have comparable ideas but use slightly different language. From an actuarial perspective, the concepts are similar as they relate to distinguishing strictly financial deals from insurance contracts.
In the practical application of determining risk transfer, disagreement or lack of consensus among the parties responsible for determining ultimate loss and risk transfer may arise. In these situations, an actuary may be called in to offer a more technical viewpoint.
So how can there be disagreement about something as fundamental as whether or not an insurance contract contains enough ultimate loss and risk transfer to be treated as an insurance policy? Who makes the final decision?
Officers and accountants of the captive insurance company must sign statements attesting to the accuracy of financial reporting. A large part of that financial reporting relates to the correct recording of premium and loss for the insurance operations of a captive. In essence, those individuals attesting to the accuracy of financial reporting should make the final decision on risk transfer.
All of the guidance included in SFAS 113 and its cousins are principle-based guidance. There are no ‘bright-line’ boundaries or prescribed rules for determining whether or not a contract contains enough risk transfer to be eligible for insurance accounting.
A direct result of this principles-based guidance is that preparers, auditors, and regulators may reach dissimilar conclusions regarding risk transfer given the identical facts on a contract. If we examine some of the wording in the financial guidance, we can see how this uncertainty can arise.
For ease of reference, Paragraph 9 of SFAS 113 is reproduced below:
“Indemnification of the ceding enterprise against loss or liability relating to insurance risk in reinsurance of short-duration contracts requires both of the following, unless the condition in Paragraph 11 is met: (a) The reinsurer assumes significant insurance risk under the reinsured portions of the underlying insurance contracts. (b) It is reasonably possible that the reinsurer may realise a significant loss from the transaction.”
“A reinsurer shall not be considered to have assumed significant insurance risk under the reinsured contracts if the probability of a significant variation in either the amount or timing of payments by the reinsurer is remote. Contractual provisions that delay timely reimbursement to the ceding enterprise would prevent this condition from being met.”
Despite all of the words, the concept seems clear—there must be sufficient uncertainty in the amount of payments for losses and how these payments will occur over time.
But if we read the guidelines closely, there are words which are not ‘precise’ or defined. Examples of such words are “significant”, “timely”, “reasonably possible”, and the central phrase “probability of a significant variation”.
This is the crux of the potential disagreements. Differences in interpretation of the words can lead to the differences in risk transfer conclusions. Similarly, the concepts of probability of significant variation bring to bear concepts of the probability and variability of loss, which are usually handled by actuaries.
In order to get more clarity, a historically referenced ‘rule of thumb’ is the ‘10-10 rule’, that there is risk transfer if there is at least a 10 percent probability of at least a 10 percent loss. As with most rules, there are plenty of times when it leads to further questions.
Consider traditional insurance in areas of very low claim frequency, but extremely high severity of loss, should one occur. For some catastrophe covers, there may be a loss event only one in a hundred years or even one in a thousand. But when a loss does happen, it can cost a tremendous amount.
For example, if the limit of coverage is $1 million for an infrequent event, where the expected loss frequency is one in 20 years, the expected loss would be $50,000. However, 95 percent of the time, there would be no loss.
This coverage would not pass the ‘10-10 rule’. yet if you were the reinsurer you would probably be concerned about the uncertainty in the widely divergent outcomes. If the premium charged was just enough for expected loss, expenses and a reasonable profit, many observers would agree that this coverage has enough uncertainty to qualify as transferring risk.
At the opposite end of the spectrum, the quota share reinsurance of a personal auto liability book of business may be large and stable enough so that this reinsurance would not pass the ‘10-10 rule’. It makes little sense, however, that this arrangement should receive other than insurance accounting treatment.
One of the central problems in the ‘10-10 rule’ is that it looks at essentially one point in the distribution of possible results—the point being that at which there is a 10 percent chance of loss. What about all the other possibilities where there might be a loss?
Weaknesses of the ‘10-10 rule’ were analysed in an actuarial paper, Risk Transfer Testing of Reinsurance Contracts: Analysis and Recommendations, published by the Casualty Actuarial Society back in 2006. Contributors to the paper went so far as to subject then recent S&P 500 results to the ‘10-10 rule’ and found that “a quota share reinsurance that has the same volatility characteristics ascribed to the S&P 500 by the options market over the period since 1990 would have been considered risky only about half the time”. No rational person would suggest investing is a risk free activity.
That same paper suggested a straightforward and understandable solution to the many problems of the ‘10-10 rule’. Why not use all information about the loss potential of a contract, not just those above the 90th percentile or loss greater than 10 percent of income?
The paper suggested a measure called the expected reinsurer deficit (ERD). Simply stated, it is the expected cost of all present value underwriting loss scenarios. The measure incorporates information about both the frequency and severity of a contract’s downside to the reinsurer over all loss scenarios.
The threshold for risk transfer suggested by the paper is for an ERD greater than or equal to 1 percent of premium. One of the beauties of this measure is that if a contract passes the ‘10-10 rule’, it will pass the ERD test. Frequency of loss equal to at least 10 percent multiplied by severity of loss equal to at least 10 percent yields an ERD of at least 10 percent x 10 percent or 1 percent.
In many instances when the management of a captive wants such probabilities determined as an aid to decision making about risk transfer, they seek an actuarial analysis of the contract’s loss probabilities.
The first step in the actuarial analysis will be the estimation of what the ultimate losses could be under the contract. This process will typically require historical unlimited loss and exposure information from the insured and may reference outside data to which the actuary has access.
Unless the insured’s loss experience is substantial, there is often a great deal of variability, and hence uncertainty, in the insured’s own loss history.
In order to compensate for this lack of predictable experience, ‘industry data’ is often referenced in actuarial analysis. The ‘industry’ could be insurance industry data, governmental data, proprietary data collected by the actuary or the firm engaged in the work, or other sources.
Using this history, the actuary will typically analyse how losses have emerged historically and, from this emergence pattern, estimate ultimate claim counts and loss amounts. The actuary will try to find probability distributions that fit well with the estimated ultimate historic loss frequency and severity data.
Often a Poisson distribution is used to model claim frequency whereas a log normal or other skew distribution is used to fit claim severity. Skewed distributions are most likely to be used for severity because most claims will be small or near the average, but some claims that are very costly and cause the distribution to be stretched out or skewed toward the high end of costs. After appropriate distributions are mathematically fit to the data, the actuary may be able to solve for the ERD algebraically.
In most cases, it is much easier to estimate the interaction between frequency and severity by constructing a stochastic model, usually referred to as a ‘Monte Carlo’ model. With this type of model, an actuary uses the mathematical formulations of the frequency and severity distributions to produce 10,000, 100,000 or one million random events, run these events through the structure of the insurance contract, and then reproduce thousands of randomly simulated years of experience to see how much loss is passed to the insurer.
All of the premium and loss flows are discounted back to inception of the contract using a risk-free rate. As stated in the accounting guidance, all cash flows are to be present valued, but interest rate risk is not a part of the insurance risk calculation. Hence, a current treasury rate or sovereign rate of appropriate duration is often used.
Even after such analysis through various models there can still be differences in opinion. One decision that might cause differences involves the data used in the analysis. Was external (eg, industry or insurer) data used and was it appropriate? Were enough years of data used to give a reliable answer? Other points that may give rise to differences in opinion may concern what threshold to use in the risk transfer test, and which policies to consider. The derived ERD should be treated as one element of the process in helping to determine risk transfer.
It is ultimately up to the management of the captive to decide on what degree of risk transfer is contained within any contract. A detailed actuarial analysis can provide management with guidance on how much risk transfer exists in the contract, at least from a quantitative basis.
There are many professions weighing in on insurance and captive management issues: legal, tax, accounting, underwriting, claims, and actuarial. At AIG Global Risk Solutions, we believe it is essential to have all disciplines represented and available to help resolve any issues that may arise. Working with an experienced team of captive managers and other insurance professionals can help companies make better decisions in this complex area.
