Attachment 1
Unpaid Losses and Loss Adjustment Expenses
Overview
One of the most significant estimates made by management in the preparation of our consolidated financial statements is our liability for unpaid losses and LAE, also referred to as “loss reserves.” Unpaid losses and LAE are estimates of future amounts required to pay losses and LAE for reinsured claims for which we are liable and that have occurred at or before the balance sheet date. Unpaid losses and LAE include estimates of the cost of claims incurred that were reported but not yet paid, generally referred to as “case reserves.” Unpaid losses and LAE also includes estimates of the cost of claims incurred but not yet reported, generally referred to as “IBNR.”
Our actuaries prepare estimates of our ultimate liability for unpaid losses and LAE based on various actuarial methods including the loss ratio method, the Bornhuetter-Ferguson method and the chain ladder method, which are discussed below. We believe that the quantitative actuarial methods used to estimate our liabilities are enhanced by management’s professional judgment. We review the actuarial estimates of our liability and determine our best estimate of the liabilities to record as unpaid losses and LAE in our consolidated financial statements. We use the same processes and procedures for estimating unpaid losses and LAE for annual and interim periods.
We do not establish liabilities until the occurrence of an event that may give rise to a loss. When an event of significant magnitude occurs, such as a property catastrophe event that affects many of our ceding insurance companies, we may establish liabilities specific to such an event. If an event has occurred but has not resulted in reported losses before the balance sheet date, our actuaries will generally estimate the impact of the event and take it into consideration when estimating our liability for unpaid losses and LAE, including potentially separate evaluation as a large or catastrophe loss. Estimated ultimate losses related to a catastrophe event may be based on our estimated exposure to an industry loss and may rely on the use of catastrophe modeling software.
We receive information from ceding companies regarding our liability for unpaid losses and LAE. This information varies but typically includes information regarding paid losses and case reserves and may include a ceding company’s estimate of IBNR. We may increase or decrease case reserves based on receipt of additional information, including information received from ceding companies. Adjustments to reported case reserves are generally referred to as “additional case reserves”.
Unpaid losses and LAE represent management’s best estimate, at a given point in time, of the ultimate settlement and administration costs of claims incurred, and it is possible that our ultimate liability may differ materially from such estimates. We review our estimates of unpaid losses and LAE quarterly. Any adjustments of unpaid losses and LAE are accounted for as changes in estimates and are reflected in our results of operations in the period in which they are made.
The liabilities recorded on our consolidated balance sheets as of December 31, 2009 and 2008 for unpaid losses and LAE were $[_____],000,000 and $2,463,506,000, respectively. These amounts exclude any future recoveries from our retrocessionaires for coverage we purchased, which are reflected as assets on the consolidated balance sheet. The following table sets forth a breakdown between case reserves, additional case reserves and IBNR by segment as of December 31, 2009 and 2008 ($ in thousands):
| | | Property and Marine | | | Casualty | | | Finite Risk | | | Total | |
| | | | | | | | | | | | | |
| December 31, 2009 | | | | | | | | | | | | |
| Case reserves | | $ | [____] | | | | [____] | | | | [____] | | | $ | [____] | |
| Additional case reserves | | | [____] | | | | [____] | | | | [____] | | | | [____] | |
| IBNR | | | [____] | | | | [____] | | | | [____] | | | | [____] | |
| Total unpaid losses and LAE | | $ | [____] | | | | [____] | | | | [____] | | | $ | [____] | |
| | | | | | | | | | | | | | | | | |
| December 31, 2008 | | | | | | | | | | | | | | | | |
| Case reserves | | $ | 255,468 | | | | 364,321 | | | | 56,638 | | | $ | 676,427 | |
| Additional case reserves | | | 4,591 | | | | 25,600 | | | | – | | | | 30,191 | |
| IBNR | | | 281,573 | | | | 1,378,000 | | | | 97,315 | | | | 1,756,888 | |
| Total unpaid losses and LAE | | $ | 541,632 | | | | 1,767,921 | | | | 153,953 | | | $ | 2,463,506 | |
Since we rely on information regarding paid losses, case reserves and sometimes IBNR provided by ceding companies in order to assist us in estimating our ultimate liability for unpaid losses and LAE, we perform certain procedures in order to help determine the completeness and accuracy of such information. Periodically, management assesses the reporting activities of these companies on the basis of qualitative and quantitative criteria. These procedures include conferring with ceding companies or brokers on claims matters. Our claims personnel may also conduct periodic audits of our ceding companies to: (1) review and establish validity of specific claims, (2) determine that case reserves established by the ceding company are reasonable, (3) assure that there is consistency in claim reporting from period to period, and (4) assess the overall claims practices and procedures of the ceding company. We also monitor the claims handling and reserving practices of ceding companies in order to help establish the proper reinsurance premium for reinsurance agreements.
Non-Catastrophe Reserves
Non-catastrophe reserves were $[_____],000 as of December 31, 2009 representing [__]% of our unpaid losses and LAE. When estimating unpaid losses and LAE, we segregate the business into classes by reinsurance subsidiary, by type of coverage and by type of contract (resulting in approximately 133 classes). Within each class the business is further segregated by Underwriting Year, starting with 2002, our first year of operations.
Our actuaries calculate multiple point estimates of our liability for losses and LAE using a variety of actuarial methods for many, but not all, of our classes for each Underwriting Year. We do not believe that these multiple point estimates are or should be considered a range. Our actuaries consider each class and determine the most appropriate point estimate for each Underwriting Year based on the characteristics of the particular class including: (1) loss development patterns derived from historical data, (2) the credibility of the selected loss development pattern, (3) the stability of the loss development patterns, and (4) the observed loss development of other underwriting years for the same class. Our actuaries also consider other relevant factors, including: (1) historical ultimate loss ratios, (2) the presence of individual large losses and (3) known occurrences that have not yet resulted in reported losses.
We believe that a review of individual contract information improves the loss estimates for some classes of business. For example, individual contract review is particularly important for classes of business within the Finite Risk segment and for the accident and health class within the Casualty segment. Our actuaries make their determinations of the most appropriate point estimate loss for each class based on an evaluation of relevant information and do not ascribe any particular portion of the estimate to a particular factor or consideration. These estimates are aggregated for review by management and, after approval, are the basis for our liability for unpaid losses and LAE.
Generally, estimates of ultimate losses that are not related to a specific event are initially determined based on the loss ratio method applied to each Underwriting Year and to each class of business. The selected ultimate losses are determined by multiplying the initial expected loss ratio times the earned premium. The initial expected loss ratios are key inputs, involve management judgment and are based on a variety of factors, including: (1) contract by contract expected loss ratios developed during our pricing process, (2) our historical loss ratios and combined ratios (loss plus acquisition cost ratios), and (3) when available, updated and appropriately adjusted, the historical loss ratios of St. Paul Re. These judgments take into account management’s view of past, current and future factors that may influence ultimate losses, including: (1) market conditions, (2) changes in the business underwritten, (3) changes in timing of the emergence of claims and (4) other factors that may influence ultimate loss ratios and losses.
Over time, as a greater number of claims are reported, actuarial estimates of IBNR are based on the Bornhuetter-Ferguson and the chain ladder techniques. The loss development pattern is a key input to these techniques. The Bornhuetter-Ferguson technique utilizes actual reported losses, a loss development pattern and the initial expected loss ratio to determine an estimate of ultimate losses. We believe this technique is most appropriate when there are few reported claims and a relatively less stable loss development pattern. The chain ladder technique utilizes actual reported losses and a loss development pattern to determine an estimate of ultimate losses that is independent of the initial expected ultimate loss ratio and earned premium. We believe this technique is most appropriate when there are a large number of reported losses with significant statistical credibility and a relatively stable pattern of reported losses. The determination of when reported losses are sufficient and credible to warrant selection of an ultimate loss ratio different from initial expected loss ratio also requires judgment. We generally make adjustments for reported loss experience indicating unfavorable variances from initial expected loss ratios sooner than reported loss experience indicating favorable variances. This is because the reporting of losses in excess of expectations tends to have greater credibility than an absence or lower than expected level of reported losses.
While we commenced operations in 2002, the business we write is sufficiently similar to the historical reinsurance business of St. Paul Re such that we review the historical loss experience of this business when we estimate our own initial expected loss ratios and loss development patterns. The historical loss experience of St. Paul Re is updated quarterly by St. Paul Re and is available to us. These loss development patterns can span more than a decade and, given our own relatively limited history, the availability of the St. Paul Re data is a valuable supplement to our own and industry data.
Loss development patterns are determined utilizing actuarial analysis, including management’s judgment, and are based on historical patterns of paid losses and reporting of case reserves to us, as well as industry patterns. Information that may cause future loss development patterns to differ from historical loss development patterns is considered and reflected in our selected loss development patterns as appropriate. For property and health coverages these patterns indicate that a substantial portion of the ultimate losses are reported within two to three years after the contract is effective. Casualty loss development patterns can vary from three years to over twenty years depending on the type of business.
In property lines the loss development patterns are based on historical reported loss data. For all lines, historical data by effective date and business type is used to determine loss development patterns that reflect each year’s reinsurance contract inception date distribution and the distribution of underlying business written on a losses occurring versus on a risk attaching basis. In marine lines the loss development patterns are primarily based on historical reported loss data. Loss development patterns are analyzed for various reinsurance sub-classes and an overall pattern is determined by the mix of business within each Underwriting Year.
In the North American casualty excess of loss classes, the loss development patterns are primarily based on our historical reported loss data and that of St. Paul Re, both of which are supplemented by industry data from the Reinsurance Association of America (“RAA”) and Insurance Services Offices, Inc. (“ISO”). Due to the long loss development pattern in general liability, various sources are used to estimate the end of the loss development pattern referred to as the “tail”. To estimate the tail we supplement our historical data and that of St. Paul Re available to us, with industry data, generally from the RAA.
We analyze historical loss development patterns and may adjust them for observed anomalies. For example, we observed that loss development patterns were much slower in Underwriting Years that were characterized by especially intense competition, known as the “soft market”, particularly in the North American excess-of-loss claims made class. We believe this is due to multiple year policies written by ceding companies and the deterioration in underwriting standards during these periods. In determining our loss development patterns for certain classes we may exclude certain historical data from the soft market years as none of our business was written in these soft market periods. One of the risks of excluding some of the years is that we could be obscuring trends in loss development patterns. Our actuaries consider this when determining the credibility of indications that use these patterns. For a small number of reinsurance contracts appropriate historical loss development patterns must be developed from ceding company data or other sources.
In finite casualty classes, expected loss development patterns for the largest contracts are based on ceding company data. We use appropriate patterns from our data or the RAA for other contracts when ceding company data is unavailable.
Catastrophe Reserves
Generally, an event must cause more than $1 billion of property losses to the insurance industry or $10 million of property losses to the Company to be considered and tracked as a major catastrophe. Unpaid losses and LAE related to major catastrophes total $[_____],000, which represents [__]% of our total unpaid losses and LAE as of December 31, 2009.
Our underwriters will typically prepare an initial estimate of our ultimate losses for a catastrophe event on a contract-by-contract basis. This estimate is typically based on estimates of losses for the insurance industry as a whole, estimates of losses prepared by ceding companies, estimates of market share of our ceding companies and, in certain cases, output from catastrophe models. Information is typically updated as it becomes available. Our actuaries and underwriters will also consider a variety of factors, including: (1) the credibility of ceding company estimates, (2) whether the ceding company estimates include IBNR, and (3) whether the ceding company information is current. After reviewing loss estimates and other information with our underwriters, our actuaries make an estimate of ultimate loss.
As losses from catastrophes mature, our actuaries consider losses reported to us relative to loss development patterns from prior catastrophe events. Our estimate of ultimate liability for losses and LAE related to a catastrophe event will generally be based on these development patterns after approximately twelve months following the event. However, since loss development patterns may be inconsistent between events, for the very large catastrophes, such as Hurricane Katrina in 2005, we will generally review information on a contract-by-contract basis for a longer period. Ultimate losses for a catastrophe event are typically reasonably well known within 12 to 24 months following the event, although ultimate losses from an earthquake may take longer to develop.
We have established specific reserves for the following 2008 major catastrophes: European storm Emma, Hurricane Gustav, Hurricane Ike and two other U.S. catastrophe events referred to by Property Claim Services (“PCS”) as Catastrophe 42 and 43. PCS is a division of ISO. We also have established specific reserves for catastrophe events prior to 2008, including Hurricane Katrina.
Uncertainty of Estimates
The ultimate liability for unpaid losses and LAE may vary materially from our current estimates for several reasons. Our estimates of ultimate loss are the basis of unpaid losses and LAE and are inherently uncertain because our estimates are affected by factors that are highly dependent on judgment. There are numerous other factors that add uncertainty to our estimates of losses, the more significant of which include: (1) our estimates are based on predictions of future developments and estimates of future trends in claim severity and frequency, (2) the reliance that we necessarily place on ceding companies for claims reporting, (3) the associated time lag in reporting losses, (4) the need to estimate an initial expected loss ratio before significant loss experience is reported, (5) the low frequency/high severity nature of some of the business that we underwrite, and (6) the varying reserving practices among ceding companies.
Our estimates are based on assumptions that historical loss development and trend are reasonably predictive of how losses will develop in the future when reported. New or updated information or loss data may impact our selection of ultimate loss ratios in subsequent periods. There are various elements of updated loss data and related information that may result in a materially different estimate of ultimate losses. The four most significant inputs into our loss estimation process are: (1) the initial expected loss ratio, (2) the loss development patterns, (3) earned premium and (4) reported losses to date. The frequency and severity of reported losses relative to anticipated frequency and severity of losses is one of the most influential factors and is largely dependent on the loss experience of our ceding companies. Reported loss experience is a key input to our loss estimation process and, if loss experience reported in periods subsequent to estimating the ultimate losses are materially greater or less than anticipated, we may adjust the ultimate loss ratio accordingly. Adjustments to increase or decrease a prior year’s ultimate loss ratio are generally referred to as unfavorable or favorable loss development.
The initial expected loss ratios are key inputs to our loss estimation process, are derived from historical data and involve a high degree of management judgment. The selection of the initial expected loss ratios also take into account management’s view of past, current and future factors that may influence expected ultimate losses. Because of the high degree of judgment required in establishing initial expected loss ratios, there is uncertainty in the resulting estimates.
The loss development patterns are also key inputs to our loss estimation process. Loss development patterns reflect the time lag between the occurrence and settlement of a loss. The time lag in reporting can be several years in some cases and may be attributed to a number of reasons, including the time it takes to investigate a claim, delays associated with the litigation process, and the deterioration in a claimant’s physical condition many years after an accident occurs. As a predominantly broker market reinsurer for both excess-of-loss and proportional contracts, we are subject to a potential additional time lag in the receipt of information as the primary insurer reports to the broker who in turn reports to us. As of December 31, 2009, we did not have any significant back-log related to our processing of assumed reinsurance information. All of the foregoing factors can impact the loss development pattern. A key assumption is that past loss development patterns are reasonably predictive of future loss development patterns. It may be difficult to identify differences in business reinsured from Underwriting Year to Underwriting Year and how such differences can affect loss development patterns. This difficulty adds to uncertainty in estimates that use these patterns.
In property classes, there can be additional uncertainty in loss estimation related to large catastrophe events. With wind events, such as hurricanes, the damage assessment process may take more than a year. The cost of rebuilding may be subject to increase due to supply shortages for construction materials and labor. In the case of earthquakes, the damage assessment process may take several years to discover structural weaknesses not initially detected in buildings. The uncertainty inherent in loss estimation is particularly pronounced for casualty coverages, such as umbrella liability, general and product liability, professional liability and automobile liability, where information, such as required medical treatment and costs for bodily injury claims, emerges over time. In the overall loss estimation process, provisions for economic inflation and changes in the social and legal environment are considered.
Loss development patterns can vary significantly from event to event adding further uncertainty to estimates. In the table below, the paid and reported losses are shown as a percentage of the current estimated ultimate loss for Hurricane Ike in 2008, our most significant catastrophe in 2008, and compared with losses from catastrophe events of 2004 and 2005 at nearly similar points of their respective development:
| | | Paid % | | Reported % |
| | | | | |
| Hurricane Ike losses as of December 31, 2008 | | [__]% | | [__]% |
| | | | | |
| 2005 hurricane losses as of December 31, 2005 | | [__]% | | [__]% |
| | | | | |
| 2004 hurricane losses as of December 31, 2004 | | [__]% | | [__]% |
Paid losses for Hurricane Ike are a higher percentage of estimated ultimate loss than for the 2004 and 2005 Hurricanes at similar points of their respective development. This reflects the advance payments on certain large catastrophe covers for Hurricane Ike. Reported losses for Hurricane Ike are a lower percentage of estimated ultimate loss than for the 2004 and 2005 Hurricanes at similar points of their respective development. While we review this type of data to help us determine the reasonableness of our estimate of ultimate loss we recognize that the pattern of paid and reported losses varies significantly by event.
Changes in estimates of prior years’ earned premiums can also affect prior years’ ultimate losses. Our actuaries consider factors affecting all key inputs to actuarial techniques when determining the credibility of indications.
The current estimate of unpaid loss and LAE is a central estimate that reflects many reasonable possible outcomes. The range of reasonable alternative estimates is necessarily smaller than a range of reasonably possible outcomes. In the following two sections we discuss two types of uncertainty with respect to loss estimation. Under Variability of Outcomes we discuss how estimates change over time as new information or loss data develops. Under Sensitivity of Estimates we demonstrate that alternative reasonable estimates can be made with current information.
Variability of Outcomes
The liability for unpaid losses and LAE as of the balance sheet date represents management’s best estimate of the ultimate liabilities as of that date. The actuarial techniques used by our actuaries in estimating our liabilities produces a central estimate of ultimate losses and LAE for each class and underwriting year. These techniques do not produce a range of reasonably possible outcomes. For some classes the ultimate value of the liability for unpaid losses and LAE will not be known for many decades. We expect that the ultimate value will differ from current estimates as losses are reported and paid and that difference could be material as reported losses reflect the actual emergence of factors that influence claim costs. Each quarter we re-estimate ultimate losses and LAE and reflect updated information in those estimates.
During the years ended December 31, 2009 and 2008, we experienced net favorable development of $[_____],000 and $159,936,000, respectively. This net favorable development was attributable primarily to a level of losses reported to us by our ceding companies that was lower than expected and that, in our judgment, resulted in sufficient credibility in the cumulative loss experience to adjust our previously selected ultimate loss ratios. During 2009 and 2008, approximately $[_____],000 and $[_____],000, respectively, of the total net favorable development was attributable to lower reported loss experience than we expected. During the years ended December 31, 2009 and 2008, changes in the initial expected loss ratio and the estimated pattern of reported losses resulted in net favorable development of losses of $[_____],000 and $[_____],000, respectively. Conditions and trends that have affected development of reserves in the past may not necessarily occur in the future. We believe, however, that as a greater percentage of losses are reported, the likelihood of material changes to ultimate losses declines. The factors that may result in differences between our current estimates of loss liability and our ultimate loss liability are set forth above under “Uncertainty of Estimates”.
Sensitivity of Key Inputs to Actuarial Techniques
Initial expected loss ratios and loss development patterns are key inputs to our loss estimation process. We exercise judgment in establishing key inputs at the beginning of an accident or underwriting year and as we modify them, as appropriate, throughout the loss development period. We have selected the initial expected loss ratio and the estimated pattern of reported losses for sensitivity analysis. Ultimate loss estimates for the North American casualty excess of loss classes of business, which generally have the longest pattern of reported losses, have a higher degree of uncertainty than other reserving classes. IBNR for these classes as of December 31, 2009 was $[_____],000, which was [__]% of the total IBNR at that date. The estimates of unpaid losses and LAE related to North American casualty excess of loss classes of business have a higher degree of uncertainty and, consequently, reasonable alternatives to our selected initial expected loss ratios and loss development patterns could vary significantly. For example, if we increased the initial expected loss ratio for these classes from 68% to 73%, we would increase the IBNR for these classes by $[_____],000, which represents approximately [__]% of unpaid losses and LAE for these classes as of December 31, 2009, or if we increased the initial expected loss ratio for these classes from 68% to 78%, we would increase the IBNR for these classes by $[_____],000, which represents approximately [__]% of unpaid losses and LAE for these classes as of December 31, 2009.
As another example of key assumption sensitivity, if we accelerated the estimated pattern of reported losses related to North American casualty excess of loss classes by 5%, we would decrease the IBNR for these classes by $[_____],000, which is less than [__]% of unpaid losses and LAE for these classes as of December 31, 2009, or if we accelerated the estimated pattern of reported losses by 10%, we would decrease the IBNR for these classes by $[_____],000, which is less than [__]% of unpaid losses and LAE for these classes as of December 31, 2009.
The sensitivity analysis illustrates how a reasonable alternative assumption could affect the current estimate of our ultimate loss liability. It is not intended to present a range of reasonable possible settlement values in the future. Actual settlement values could be materially different from the current estimates. Over time changes to the initial expected loss ratio and loss development patterns may vary by more than the sensitivity analysis above as new loss information and data emerges.
