ERM Saves the Day for the CEA
Abstract. Since its emergence in the 1990s enterprise risk management (ERM) has struggled to define itself and to prove its value. Numerous statistical studies suggest ERM does add value, but at the same time, the need for such studies belies their conclusion. This case study complements the statistical analyses. It describes a situation where ERM had a clear, positive and unambiguous impact on a material business decision. Using public information, it shows that a decision not to join the California Earthquake Authority (CEA) was rational from a silo-based perspective but that a holistic enterprise-wide analysis comes to the opposite conclusion. The analysis described is close to that undertaken at the time by a particular multi-line insurer whose revised decision was critical to the CEA satisfying certain legal requirements and becoming operational. The study identifies several difficulties the insurer faced during the process that are still relevant to ERM practice today and that would not necessarily be solved by a highly rated ERM program.
JEL Codes: G22, G24, H84, K23, C65
Keywords: Enterprise Risk Management, catastrophe insurance, reinsurance, market organization, dynamic financial analysis, California, earthquake, CEA, ERM
Introduction
ERM and the CEA both originated in the mid-1990s and ERM precepts were important to market participants as the CEA design evolved between 1994 and 1996. This case study considers how one multiline insurer used an ERM approach in its decision to join the CEA. Relying on public information it explains how the insurer’s initial silo-based analysis indicated it should not participate, but how, after subsequent events prompted a re-evaluation, an enterprise-wide analysis determined participation was optimal. ERM has always struggled to define and justify itself, to be regarded as more than regulatory compliance, and to prove that it “adds value”. This case study adds to the literature supporting ERM by providing a clear example where it added value by changing a material decision, improving the enterprise’s risk-return position and providing a boost to the CEA becoming operational. A subsequent study, [1], will consider the CEA more broadly within the industry context of the early 1990s and describe what we can learn from how other insurers evaluated their participation.
ERM is difficult to define and many papers try. [2] describes the origins and foundation of ERM. [3] discusses implementing ERM in an insurance enterprise. [4] summarizes definitions of ERM from academic, industry and professional perspectives. It states the consensus view of ERM is based on a belief it is more efficient to manage risk at the corporate than the subsidiary level; that ERM should take a broad view of risk to include strategic risk; and that ERM should not look at risk as a problem to mitigate but rather as a place firms where can seek to profit from a competitive advantage.
[5] identifies ERM adoption through the appointment of CROs and finds more leveraged firms more likely to be adopters. [6], [7], [8], [9], and [10] discuss implementing and operating an ERM program. [11], [12] links ERM to valuation and suggests the need link pricing to maximize value. [13] treats ERM as quantitative optimization problem. [14] considers which aspects of ERM help improve performance and identifies the existence of an economic capital model and dedicated risk managers as beneficial.
Several studies have shown ERM adds value. [15] finds a positive relation between firm value and the use of ERM, as do [16], [17] and [18]. [Berry-Stolzle2016?] finds ERM adoption significantly reduces firm’s cost of capital.
[19] finds evidence of a positive relationship between increasing levels of traditional risk management capability and firm value but no additional increase in value for firms achieving a higher Standard and Poor’s ERM rating. ERM’s scorecard was hurt by poor performance of several well-rated ERM programs during the 2008 financial crisis.
The literature includes a limited number of examples and compelling use cases outside of pure financial risk management, something to which this paper contributes. [20] discusses hedging amongst gold mining companies. [21] consider how airlines hedge the cost of aviation fuel and find weaker companies hedge less because of collateral constraints. [22] suggests insurers should only bear risks where they have a competitive advantage in risk evaluation and that they should hedge all financial risks at a fair market price.
Homeowners insurance is essential to the functioning of the economy because it is required to obtain a mortgage. It is the second largest line of P&C insurance, after personal auto. California is the largest homeowners market in the US and unique in having a tie to earthquake insurance through a mandatory offer law. In order to provide context for the financial analysis, the case study details the triggering Northridge event and explains why it caused such disruption in the California homeowners market. It then constructs a financial model an an insurer could use to decide whether or not to join, and shows that for multiline writers it was optimal not to participate. Trade journals reported November 25, 1996 that the CEA did not have enough industry support to become operational. They also reported Berkshire Hathaway had, in a surprise move, committed to write a substantial upper layer of CEA protection originally envisioned as a cat bond.
At that time I worked in the personal lines product management department at CNA, a large multi-line insurance company. CNA had originally decided not to join the CEA. I was asked by the company CEO to review that decision in the light of subsequent information, particularly the Berkshire announcement. As a result of taking an enterprise-wide view of CEA membership, rather than just considering it as a personal lines problem, and of understanding the economics of CEA reinsurance the company decided it should, in fact, join, subject to certain conditions. This re-evaluation provides a wonderful example of making an important, material and economically beneficial decision using the precepts of the (then yet to be christened) enterprise risk management, ERM. The decision was important to both the CEA, which was able to announce on November 26 that all its statutory conditions for existence had been met and to start operations on December 1, 1996, and to CNA, which radically improved its risk/return profile for California earthquake exposure.
The rest of the paper runs as follows. Section 1 describes the Northridge event and explains why it had such a profound impact on the industry as well as aspects of the CEA solution. Section 2 constructs a financial model of the CEA and residential earthquake insurance in California and shows why a silo-based model does not support CEA participation. Section 3 gives background on CNA, recaps press reports from Fall 1996 as the CEA struggled to meet its statutory requirements to become operational and describes a revised analysis used by CNA, based on holistic enterprise-wide principles, to decide it should participate in the CEA. Section 4 examines difficulties encountered during the enterprise decision making process and draws some conclusions for ERM.
References
The following shorthands are used for common references. The California Insurance Code is quoted by section: e.g. §10088, sometimes with additional subsection details. NU refers to the National Underwriter and BI to Business Insurance, both accompanied by publication date. The CEA Annual Report to the Legislature is ARL and AFS their audited financial statements. Both are available on the CEA’s website https://www.earthquakeauthority.com/. CADOI points to a public California Department of Insurance rate filing, available through the department’s web access to rate filing forms site https://interactive.web.insurance.ca.gov/warff/front. Finally, Best’s Reports refers to company reports in A. M. Best’s Insurance Reports, Property-Casualty United States, 1994-1996.
Acknowledgments
I would like to thank various folks.
Northridge, Its Impact, and The CEA
Prior to 1992 the industry had only experienced one USD 1 billion insured loss catastrophe event: Hurricane Hugo in 1989. Between Hurricane Andrew in August 1992 and the Northridge Earthquake on January 17, 1994, less than 18 months later, that changed profoundly. Andrew began an industry-wide re-evaluation of catastrophe loss potential and changed the industry’s opinion of homeowners insurance. State Farm, for example, had already begun to manage exposure in California prior to Northridge1. Catastrophe modeling suddenly moved to the top of the agenda. Data that had been impossible to obtain suddenly became available as reinsurance price differentials made gathering it an economic necessity.
The Northridge earthquake was a 6.7 moment magnitude event that occurred in the early morning of January 17, 1994. It stuck the densely populated San Fernando Valley region of Los Angeles, killing 57 people and causing more than 8000 injuries. See [23] for these and related facts. It is estimated to have caused economic losses in excess of USD 40 billion and a PCS insured loss of USD 12.5 billion when it occurred. Losses from Northridge cost the industry more than the total California earthquake premium written since 1972. Nearly 50 percent of the economic losses and over 70 percent of insured losses from Northridge were residential. Insurance covered just under half the residential economic loss, consistent with estimated earthquake insurance take-up rates in the mid-40 percent range.
Northridge was a very difficult loss to estimate, adding to the industry’s unease about earthquake risk. Petak says that while the earthquake was relatively small in terms of seismic intensity its economic losses were extreme, and they exceeded all previous predictions. Contemporary estimates illustrate the difficulties. In its first report on January 24, 1994 the National Underwriter headline proclaimed Calif. Quake Insured Loss Could Hit USD 1.9 Billion. It said EQE International, a cat modeling firm, estimated total quake damage at USD 15.1 billion and insured loss of between USD 1.5 and 1.9 billion.
California homeowners is the largest homeowners market in the US accounting for 11 percent of premium in 1994. It was obviously immune to wind losses: a huge attraction in 1994. Its five year combined ratio 1986-90 was 102.0 percent, which looked favorable compared to the countrywide average 105.4 percent. However in 1991 and again in 1993 loss ratios were adversely impacted by major brush fires. In addition there were major losses from the 1992 Los Angeles riots. By 1994 the five year average combined ratio had increased to an alarming 130.0 percent. Nonetheless, since homeowners does not include earthquake and since “it never rains in Southern California”, carriers remained confident they could quantify and price homeowners. In retrospect their confidence was well-founded. 1996-2000 all produced an underwriting profit and by 2000 the five year average combined ratio was 91.0 percent. Today, obviously, the incidence of devastating wildfires is again stressing the market. In summary: California was a large and important homeowners state, regarded as potentially profitable, and with low catastrophe risk. Insurers were highly motivated to write in the state, despite Proposition 103.
The industry’s reaction to the uncertainty caused by Northridge was fast and broad. By June 6, 1994 NU reported Calif. Homeowners Market Shrinking Fast, Agents Say with some companies refusing to sell new policies and others cutting back. The Independent Insurance Agent and Brokers of California (IIABC) said two thirds of its members reported restrictions, including termination of agents and a ban on new policies.
The reaction to Northridge was on homeowners insurance availability. Homeowners insurance is a critical part of the modern economy because it is required in order to obtain a mortgage. By contrast, and perhaps surprisingly, earthquake insurance is not required for a mortgage, although flood insurance is required for properties located in flood plains. How did earthquake losses spill over to homeowners coverage?
The California Insurance Code has a unique mandatory offer law created in 1984 that requires homeowners insurers to offer their policyholders an earthquake policy every two years (§10081). The policy must provide certain minimum coverages. The mandatory offer was introduced as a quid pro quo for the abrogation of concurrent causation as applied to earthquake (§10088). Insurers had seen earthquake exclusions successfully challenged in litigation and were very concerned about the possibility of implied coverage, especially following a major event. The mandatory offer exposes insurers to two undesirable effects. First, they have no control other their exposure. Penetration of earthquake insurance typically spikes after an event owning to heightened insured awareness. Second, insurers can be adversely selected against, with only high-risk (location or type of structure) insureds accepting the offer.
At the height of the crisis in summer 1996 the Department of Insurance reported that 95 percent of the homeowners writers in California had either stopped writing or had severely restricted the writing of new homeowners insurers (ARL1997). In addition, many insurers were actively non-renewing existing homeowners customers.
Homeowners are voters and elected representatives could not tolerate a frozen homeowners insurance market. Through a series of initiatives and bills they set about solving the problem, eventually leading to the creation of the CEA in November 1996. [24] provides a detailed, insider’s description of the founding, history and operation of the CEA. [1] and [25] also discuss the design and operation of the CEA. Here we provide an overview, just identifying points of the CEAs operation needed to build and understand a financial model.
Regulators and law makers and the State of California needed a solution that maintained the mandatory offer,made no call on the public purse, and was not seen as “bailing out” the insurance industry. They felt repealing the mandatory offer could end residential earthquake availability in the state, which was deemed unacceptable.
All parties were concerned that any new entity be credible, solvent and sustainable. It needed to be financially robust enough that it was unlikely to fail in the short term. No one wanted to be faced with the same problem again in the near future.
Everyone was concerned about the “over the top” (OTT) exposure of liabilities in excess of assets: if the new entity failed who paid? Deep pocketed insurers were wary of being forced to pay. The state was sensitive that if it were responsible for OTT losses it would strain public finances at a bad time and be seen as bailing out insurers. Because of the uncertainties introduced by Northridge, as well as a general post-Andrew caution, no one felt confident about estimates of the OTT exposure. [26] suggest earthquake losses could be distributed according to a power curve with \(\alpha\le 1\), distributions that fail to have a mean. As a result ultimate responsibility for “the big one” was a particularly sensitive problem and some insurers lobbied that, whereas (wind-free) California homeowners risks were well understood, earthquake risk was too high and too poorly understood. Stung by Andrew and surprisingly large Northridge losses some wanted an absolute cap on their earthquake exposure, including any liabilities to new or existing state entities.
To address the OTT exposure the CEA has an explicit right to pro rate claims it if determines it has insufficient funds to pay 100 cents on the dollar (§10089.35(a)). The Commissioner does not have the right to place the CEA into conservatorship and it is not subject to normal bankruptcy proceedings. As a result neither the state nor the insurance industry is responsible for losses in excess of claims paying capacity. The state’s liability is explicitly disavowed in §10089.35(a): the state has no liability and no “obligation whatsoever” for payments in excess of funds available. Removing the OTT exposure was a very important design consideration for both regulators and insurers. As [25] points out it is nearly impossible to conceive of a private market having the contractual ability to pro rate claims, especially on the basis of a market loss.
Insurers were also concerned with indirect exposure to earthquake loss via guaranty fund assessments and §10089.34 explicitly disavowed their liability.
California’s unique mandatory offer specifies the coverages insurers must offer residential property policyholders. In 1995 the Legislature scaled back the mandatory coverages by creating the minipolicy (§10089). It included a substantial 15 percent of TIV deductible, limited coverage to the dwelling and excluded outbuildings, capped contents coverage at USD 5,000 and additional living expenses at USD 1,500. The minipolicy design incorporated lessons from Northridge, for example reflecting the high level of damages to appurtenant structures (Coverage B) such as swimming pools, detached patios and garages, [27]. Companies that elected not to participate in the CEA were allowed to file their own mini-policies, which most did.
The final CEA design combined a robust financial structure, Federal tax free status and an explicit right to pro rate claims above its financial resources. It is privately financed and publicly managed entity, governed by a Board consisting of the Governor, the Treasurer and the Commissioner. Its creation allowed law makers to maintain the mandatory offer, albeit subject to the lower minipolicy provisions.
The proposed structure gave the CEA the ability to pay claims up to USD 10.5 billion if all insurers participated. This represents substantial protection compared to re-estimated minipolicy losses of USD 4 billion for Northridge (REF). The 100 percent financial tower is shown in . All except Layer 4 is pre-funded, a deliberate design choice to avoid the possibility of substantial debts early on at the expense of slower capital build-up.
| Layer | Abbreviation | Description | Cover at 100 Pct |
|---|---|---|---|
| Over the Top | OTT | Pro Rate Claims | unlimited xs 10.5 |
| Layer 5 | 2IAL | Second Industry Assessment Layer | 2.0 xs 8.5 |
| Layer 4 | Re2 | Securitized Cat Bond | 1.5 xs 7.0 |
| Layer 3 | Debt | Policyholder assessment | 1.0 xs. 6.0 |
| Layer 2 | Re1 | First Reinsurance Layer | 2.0 xs 4.0 |
| Layer 1 | 1IAL | First Industry Assessment Layer | 3.0 xs 1.0 |
| Working capital | Capital | New Member Capital Contributions | 1.0 xs 0 |
| Total | 10.5 xs 0 |
Participant capital contributions are based on residential earthquake premiums at January 1, 1994. [25] describes the capital contribution as an “exit tax”. Liability for the industry assessment layers (IAL) is based on actual writings of member companies in the CEA. The IALs had various phase outs, with the first IAL eroding as retained capital accumulated.
The first reinsurance layer was a two year, USD 2 billion aggregate protection. The CEA was required to obtain reinsurance contracts in an amount of at least 200 percent of the total initial cash contributions from participating insurers (§10089.14).
The policyholder assessment layer is provided by debt issuance or a letter of credit repayable by policyholders assessments.
Layer 4, the second reinsurance layer, was proposed to be a securitized catastrophe bond transaction. At the time the bond would have been by far the largest cat bond issued and it would have been a huge boost to the emerging cat bond market. However, as events transpired, it was placed as traditional reinsurance with Berkshire Hathaway.
The CEA is exempt from state premium tax. Amounts of premium tax otherwise due, amounting to 2.35 percent of gross premium, are included in rates and booked as a capital contribution by the state to the CEA.
The CEA legislation specified three conditions for the CEA to become operational (§10089.14): the Internal Revenue Service determines the authority is exempt from federal income tax; insurers whose residential earthquake insurance market share is more than 70 percent as of January 1, 1994 commit to participate and agree to make an initial capital contribution (up to USD 1 billion); and the authority secured reinsurance limits of at least 200 percent of the total capital contributions of participating insurers (up to USD 2 billion). The second hurdle, requiring 70 percent participation, proved to be the most difficult to satisfy.
Financial Model and Silo-Based Analysis
Next we build a financial model for California residential earthquake. The model will use an accounting view, similar to the approach taken at the time, and measure risk using an earthquake PML. The model estimates the underwriting results of the minipolicy as written by the CEA and adjusts them to show how they would appear to NPIs. First we determine point estimates, and then introduce a frequency severity loss model to determine expected losses in different layers of the CEA claims paying tower.
The model shows the minipolicy was profitable for a multiline insurer like CNA. It makes no sense to pay an assessment for the right not to write profitable business, particularly for companies with experience writing and pricing earthquake. This is what CNA Personal Lines determined in the summer of 1996.
In order to understand CEA minipolicy economics we work backwards from a knowledge of how their rates were set. We estimate the key expense components and end up with a breakdown of the overall CEA combined ratio into underwriting expenses, gross losses and cost of reinsurance. The analysis is based on the CEA’s first public filing, CADOI 96-88342.
Ins. Code §10089.40(a) requires CEA rates to be ``actuarially sound’’. Marshall (p. 110-111) describes that this to mean premium must cover expected losses, the net cost of reinsurance and CEA expenses. Normally an insurer would require a cost of capital for its shareholders, but the CEA does not have any shareholders and it chose not to load rates with an explicit underwriting profit provision. It did, however, assume an implicit margin because rates did not include an offset for investment income. Thus it wrote to a 100 percent combined ratio but an to operating ratio below 100 percent by the amount of investment income.
It could be argued that the IALs should include a cost of capital adjustment because they expose (costly) participant capital. Indeed in CADOI 05-6848 (p.131) the Personal Insurance Federation of California raises exactly this point saying insurers have “real exposures to loss and many buy reinsurance for their exposure.” They also point out that if insurers did not provide the capital then the CEA would have to buy protection “in the open market” and would incur a cost of capital charge. They calculate the industry’s expected annual assessment is USD 33 million (compare to current estimates). However, the cost of the IALs is offset by the OTT cost participants avoid and so no additional loading is included in the rates.
In order not to be distracted by timing issues, earning commissions etc., we will take a policy year view.
Based on this interpretation actuarially sound rates implies that premium satisfies \[\begin{equation}\label{eq:pricing} P= \frac{L + R^*}{1-e} =\frac{(L-L^*) + R}{1-e} \end{equation}\] where
- \(P\) is total CEA premium,
- \(L\) is expected loss payments limited by CEA claims paying ability of USD 10.5 billion,
- \(R^*\) is the net cost of reinsurance, that is reinsurance premiums less expected reinsurance recoveries,
- \(R\) is reinsurance premium and \(L^*\) is ceded losses, and
- \(e\) is the expense ratio excluding reinsurance, so \(P(1-e)\) gives premium net of expenses.
By definition \(R=L^*+R^*\) and net losses are \(L-L^*\) giving the second equation. The cost of reinsurance to the CEA is not the same as profit to the reinsurers because of internal reinsurer expenses. The CEA included reinsurance brokerage as a separate expense.
can be re-written to give the underwriting combined ratio \(u\) by dividing by \(P\) \[\begin{equation}\label{eq:pricing-u} u = 1 - e - l - r^*. \end{equation}\] where \(l = L/P\), etc. Because of the ratemaking formula \(u=0\).
In all the calculations we assume that total premium for the CEA is USD 1.2 billion at 100 percent industry participation and denote it by \(P\). The premium estimate is based on CAL Update document. It is broadly consistent with the CEA maintaining earthquake penetration at its 1996 levels, in the mid-30 percent range.
A detailed analysis of the CEA’s 1999 rate filing3 gives the following accounting summary of the CEA’s minipolicy.
| Item | Symbol | Ratio |
|---|---|---|
| Loss Ratio | \(l\) | 49.0 |
| Net Cost of Reinsurance | \(r^*\) | 30.5 |
| Expense Ratio | \(e\) | 20.5 |
| Combined Ratio | \(e+l+r^*\) | 100.0 |
| Underwriting Result | \(u = 1-e-l-r^*\) | 0.0 |
At this point observe that all decisions are scale invariant in the sense that premium, expenses and the loss distribution are homogeneous variables that scale in proportional to market share, [28], [Tsanakas2016b?]. Therefore going forward we work on the level of the total market and consider results scaled by \(P\). All ratios should be multiplied by residential earthquake premium to convert to dollar amounts for a particular company.
Having established the baseline underwriting result for the CEA we next look at how it needs to be adjusted to show the underwriting results for an insurer like CNA writing its own minipolicy. The loss ratio, expense ratio and cost of reinsurance all need adjusting. We start with reinsurance.
CNA did not purchase any catastrophe reinsurance (Best Report QUOTE) and According to A. M. Bests Reports (1995, 1997 editions), “CNA chose not to purchase catastrophe reinsurance upon its June 1, 1993 renewal and has not purchased such protection since. The group views the capacity of catastrophe reinsurance available from financially sound companies as low and prices high, thus making it impractical for the group to purchase such coverage.” It re-entered the market in 1996. Therefore \(r'=0\): this difference is the source of the apparent minipolicy profitability.
Next consider expenses. The average NPI expense ratio can be estimated using the earthquake line in the Insurance Expense Exhibit (IEE). Average direct acquisition expenses for earthquake (commercial and residential, all states) are 18 to 21 percent and general expenses are 6 to 8 percent, with a total of 24 to 28 percent. California residential property was a distressed market and an earthquake policy could piggy-back off a homeowners policy and so it is reasonable to assume total expenses at the lower end of the range. Assume an NPI has \(e'=0.24\) implying a 3.5 point expense disadvantage compared to the CEA, . These estimates suggest the CEA was expense-efficient compared to commercial carriers.
Finally consider the loss ratio component, computed as expected loss divided by premium. Because the market is competitive, minipolicy premium must be based on the same rate level as the CEA. An NPI could charge a lower rate and try to cherry pick (a strategy adopted by some specialist writers) but it could not reasonably charge a higher average rate. To be adequate in the aggregate it needs to match the CEA’s rates. Expected losses need to be adjusted from CEA assumptions to reflect different levels of capitalization, i.e. the absence of right to pro rate claims. To compute this impact requires an earthquake loss model.
In 1996 the aggregate loss model was created using output from the EQECat catastrophe model. Since the late 1980s catastrophe computer simulation models have become the preferred approach to estimating losses. Catastrophe models are described in [29] and [30]. For each simulated event cat model output includes the annual rate of occurrence and the insured loss. The sum of individual event rates gives the total annual frequency for modeled earthquakes. Dividing the event rate by the total rate produces a severity curve, i.e. a conditional occurrence probability distribution. Aggregate losses are then computed assuming a compound Poisson process for each individual event. See [bowersEtAl?] Chapter 10 or [Klugman2004?] for standard facts about compound Poisson distributions.
The original EQECat distribution was based on an assumed distribution of policyholders. It was not put into the public domain as far as I know. In order to proxy the results we show results from three newer evaluations of California earthquake residential losses, each scaled to match the the original CEA loss cost of gross loss cost of USD 588.5 million , which we call models A, B and C (AIR, IF, RMS). None is perfect since the original data and models no longer exists. But their variety illustrates the realities companies grappled with in 1996: there were (and still are) a variety of models producing different results. Part of management’s task is to select between them.
shows aggregate losses from the three models on a linear scale laid against the claims paying structure.
Given a random variable \(X\) representing unlimited aggregate earthquake losses we can compute expected losses to different layers as follows. Let \(F\) and \(S\) be the distribution and survival function \(X\). A layer of width \(y\) in excess of an attachment \(a\) pays \(L=\min(\max(X-a, 0), y)\) when ground-up losses are \(X\). The expected value of \(L\) is given by \[\begin{equation} \label{eq:layer-cost} \mathsf{E}[L] = \int_a^{a+y} S(x)\, dx. \end{equation}\] This expression follows by integration by parts, see [31]. By applying this formula with diffferent values of \(a\) and \(y\), given by table XXXX, we can compute expected losses for the CEA claims paying tower and estimate the value of OTT losses. shows the results. It assumes the OTT layer covers claims up to USD 100 billion, which would proxy CNA’s level of claims paying ability at 100 percent4.
shows expected loss dollars by layer, computed using , as well as their ratio to premium. Total insured losses are calibrated to USD 589.5 million, giving a 49 percent loss ratio for all models. As is clear from the different models have different implications for losses by layer, even though they all agree up to USD 10.5 billion. The biggest differences are at the bottom of the program, in the Capital layer, with expected losses between 14.1 percent and 23.2 percent, and the OTT layer, between 5.4 percent and 30.7 percent. CEA’s 1999 rate filing quantifies the OTT benefit as only 1 percent of unlimited (modeled) losses5.
shows CNA’s estimated profitability for the minipolicy ranges from an underwriting profit of 21.6 percent (combined ratio of 88.4 percent) to a small loss of 3.5 percent, assuming a 24 percent expense ratio.
Since the minipolicy is profitable under RMS and IF, and since participation involves a certain cost, it is clear that faced with this choice and insurer would not participate. But what about model AIR? To know for sure we need to estimate the cost of the IALs. These are shown : between 18.3 and 21.6 percent of premium, or, at a 1 percent market share, between USD 2.2 and 2.6 million per year. These costs are far in excess of the 3.5 percent loss from participating under AIR.
The metrics are expressed per dollar of premium, in terms of \(u_i\), the underwriting ratio for an NPI writing a minipolicy directly, \(v_i\), the expected annual cost of the two industry assessment layers, and \(b=1/1.2=0.83\) the cost to buy-into the CEA per dollar of premium.
The basic decision rule is to use expected net present value at a risk adjusted discount rate. Denote the insurer’s cost of capital by \(r\) and recall the present value of an annuity of 1 paid at the end of the year is just \(1/r\). Also it will be convenient to let \(b = B/P = 1/1.2 = 0.83\) denote the initial assessment per dollar of residential earthquake premium.
We now have all the components needed to decide whether a given insurer should participate.
For a PI the cash flows are:
- On electing to participate pay a one-time capital assessment equal to USD 1 billion times its 1994 earthquake market share.
- Assume a liability for the two contingent IALs, triggered in the event of CEA losses above its capital (USD 1 billion at 100 percent) and at the top of the program (above USD 8.5 billion). The PIs exact liability is based on their CEA market share. The annual expected amounts are \(v_A, v_E\) on an accounting and risk adjusted basis. They are independent of characteristics of the insurer.
In return the CEA writes residential earthquake policies directly for any PI policyholder who does not reject the mandatory offer.
In general amounts due under the two IALs are subject to the participant’s catastrophe reinsurance. This could drive a perceived cost to the PI closer to the economic view than the accounting view.
A PI benefits from not being liable for earthquake related CIGA assessment liabilities. We ignore CIGA assessments as a second order effect.
We treat \(v\) as fixed over time even though it changes each year as the CEA builds capital, the IALs are eroded, reinsurance pricing changes, and minipolicy rate adequacy changes. In reality the CEA’s capital built up quite slowly. By year end 2000 surplus had increased to USD 948 million compared to contributed capital of USD 700 million. Reinsurance and minipolicy pricing moved in tandem, acting as hedges. Our simplified model is adequate for illustrative purposes, particularly because in most cases the decision is clear.
Therefore the expected net present value of cash flows for a PI are \(b + v_i / r\), \(i=A,E\) and regarding the assessments as a perpetuity.
For an NPI the cash flows are:
- Collect premium and pay losses and expenses, and possibly reinsurance costs, on the minipolicy, modeled as \(u_A, u_E\) in and as a function of \(e',s', r', Q\) the insurers expenses, safety level, cost of reinsurance (accounting view) or risk adjusted pricing measure \(Q\) (economic view).
A non-participant insurer must make an earthquake insurance offer to all residential property policyholders every two years and write earthquake policies directly for insureds who do not reject the offer. It can manage its earthquake exposure through internal risk management processes, including reinsurance, but it has limited control over take-up rates and premium volume. It is liable for large earthquake losses up to its total capital and surplus: as always, all capital stands behind all policies.
An NPI is liable for CIGA assessments to pay covered claims of insolvent insurers but as for PIs we ignore the impact of potential CIGA assessments.
The insurer should participate in the CEA if the expected net present value of doing so is greater income than writing the minipolicy directly, that is \[\begin{align}\label{eq:participate} \text{\bf participate} &\Longleftrightarrow \text{EPV(NPI cash flows)} \le \text{EPV(PI cash flows)} \\ &\Longleftrightarrow u_i / r \le -b + -v_i/r \ge u_i/r \\ &\Longleftrightarrow u_i \le -br - v_i \ge u_i < 0 \label{eq:participate3} \end{align}\] where \(i=A,E\) for the accounting or economic views. The PI cash flows are expenses, hence the signs. Since \(-br-v_i<0\) it follows that if \(u_i \ge 0\), i.e. if the insurer believes it can write the minipolicy profitably, then it will never participate no matter its cost of capital \(r\) nor its assessment of \(v_i\). It also follows that in most situations the incremental cost of the initial capital contribution, \(-br\), will not be material.
shows the results of performing these calculations across the three models for a reasonable selection of parameters. All panels assume the insurer’s expenses for earthquake are 24 percent.
For all views the decision metric, comparing \(u\) to \(-br-v\) depends on \(r\), the cost of capital. The exhibit shows results for \(r=0.05, 0.15, 0.25\) bracketing a reasonable range.
The value of \(u\) does not depend on the cost of capital and so is shown only once for each block. The value of \(v\), which is the expected value of the two IALs does not vary by cost of capital or the insurer’s safety level because it is just determined by CEA performance. It is shown only once for each block.
The decision compares \(u\) to \(-br-v\), . Here \(-br-v\) obviously varies the the cost of capital \(r\) but not with the insurer’s safety level. Finally the decision is NPI (do not participate) if \(u>-br-v\) and PI (participate) otherwise. For example in the first table looking at the first row \(u=0.271>0\) which always drives NPI. In the seventh row for A safety level \(u=-0.295\) is a worse result that \(-br-v=0.258\) at \(r=0.05\) resulting in PI; but at \(r=0.15\), \(-br-v=0.341\) resulting in NPI.
Look first at the accounting \(P\) view. There is unanimity not to join, as expected.
The graphs show the results scenario by scenario for a participant vs. a non-participant. Scenarios have been sorted from smallest to largest loss and are identified with their percentile on the \(x\)-axis. The right hand column plots the same data as the left hand column but with a different scale to reveal the detail. The \(x\)-axis scaling uses the distortion calibrated by the first reinsurance layer to better reveal details in the worse outcomes. The blue lines show the result for an NPI from writing the minipolicy, assuming a 1 percent market share. These show the insurer will make money between 80 and 90 percent of the time, when there are no events or only small events. Then the NPI stands to pay considerable losses. The orange is zero until CEA losses exceed its capital, when the first IAL activates. Losses increase through the first layer. Then there is a period where the debt and reinsurance layers respond and the PI pays no more. Finally, the second IAL begins to respond. In all cases the most the PI pays is its market share times USD 5 billion. The compare line is the difference, with positive values indicating participation is better.
Insurers who opted not to participate are revealing a risk preference for the green line, which has a positive expected value, . The green line is typical of the risk-return profile assumed by insurers: a large chance of little or no claims on an individual policy but a small chance of a large loss.
CNA had filed its own minipolicy. Since CNA had no earthquake catastrophe reinsurance costs, based on it expected at least a nearly a 20 percent underwriting margin, after allowing for higher expenses, especially if it ignored its higher safely level. Relative to the homeowners book, with a five year average pure loss ratio of 84.1 percent this appeared very profitable.
Against this backdrop, how did companies analyze whether or not to participate in the CEA? We begin by giving some qualitative advantages and disadvantages to participation. Then we present a quantitative financial model.
Throughout PI is a participating insurer and NPI a non-participating insurer.
The principle advantage of CEA participation for some insurers was to provide a definite cap on their exposure to earthquake losses. Insurers were particularly sensitive to their total exposure for obvious solvency reasons, but also because of their recent experiences in Andrew and Northridge suggested peak exposures were very difficult to estimate reliably.
The value of a cap on varies by type of company. A typical multi-line insurer had several potential sources of earthquake exposure:
- Many carriers voluntarily wrote commercial lines earthquake.
- All property writers had exposure through fire policies. California is a Standard Fire Policy state and insurers have to cover fire following earthquake even if the policyholder has no earthquake coverage.
- Similarly, it is impossible for workers compensation writers to exclude injuries at work caused by earthquake.
- Some wrote third party reinsurance assumed and these would generally participate in the earthquake market.
- Indirect losses from the FAIR plan and CIGA assessments. Although FAIR plan assessments can and CIGA assessments must be recouped in rates insurers were exposed to timing risk, between paying an assessment and re-couping it in premium. They were also subject to Proposition 103 prior approval laws, which created significant uncertainty about whether assessments could be fully reflected in rates. A surcharge required a filing that could lower profitability: if rates were very profitable they would have to be lowered to add the assessment, making the insurer worse off.
As a result an exposure cap was more valuable to predominantly personal lines companies than multi-line companies because it addressed a larger proportion of their total earthquake exposure. For CNA the cap was not a material benefit.
A second advantage of participation was to eliminate risks associated with developing, maintaining and filing a voluntary residential earthquake policy form. Insurers were allowed to file their own mini-policies, and many did so. But their filings exposed them to Prop 103 prior approval scrutiny, as well as potentially unwanted public scrutiny of fairness, soundness and affordability. CNA had a large and robust department dedicated to creating and filing rates, and was small enough its filings did not attract undue attention. Again, this was not a material benefit for them.
Third, participation meant insurers did not need to purchase their own reinsurance for residential earthquake losses. Placing reinsurance is an expensive and time consuming process. It exposes the buyer to reinsurance market price swings, which can be extreme. Like CIGA and FAIR plan assessments, the net cost of reinsurance can in theory be passed along in filed rates. But California prior approval regulation made companies wary they would be able to do this quickly and completely. CNA NO RE.
Fourth, participation meant insurers no longer need be concerned with the impact of the mandatory offer, which made it hard to manage exposures and opened them to adverse selection. Participation for a personal lines writer was essentially equivalent to exiting earthquake altogether, allowing them to focus on their core homeowners and auto businesses.
Overall, the qualitative advantages point to the CEA being very attractive to large personal lines writers, and Allstate and State Farm were both vocal advocates. On the other hand, large multiline carriers such as CNA, with a commercial lines focus, had much less to gain.
To join the CEA insurers had to pay an assessment to fund its working capital. The assessment was computed as USD 1 billion times the participants January 1, 1994 residential earthquake market share. A company with a 1 percent market share would pay an assessment of USD 10 million. DUP BEST HERE?
In addition participants were exposed to two Industry Assessment Layers, for USD 3 billion excess CEA capital and USD 2 billion at the top of the capital structure. Their assessments were proportional to their share of CEA premiums.
Enterprise-Wide Analysis
CNA Situation
In 1996 CNA was the fourth largest property casualty writer in the US based on direct written premium. It was a multi-line insurer, including personal, commercial and specialty lines business and a professional reinsurer. Its 1994 merger with Continental Insurance6 was motivated in part by a desire to boost personal lines. Following the Jack Welsh strategy CNA looked to be a top 3 player in lines where it competed, and so growth was essential. CNA had a 1.4 percent residential earthquake market share in 1994, which by 1996 was about 1 percent (SOURCE). It had a substantial loss from Northridge, caused in part by the inclusion of earthquake coverage on contents for some of its premier homeowners policies. Its countrywide homeowners loss ratio in 1994 was 110.7 percent compared to 98.5 percent in 1992 and 70.9 percent in 1993.
List many large companies in the early 1990s CNA included a professional reinsurer, CNA Re. Over the next decade all of these were shut down or sold off, as multi-lines came under attack from more focused competitors, often based in tax advantaged jurisdictions.
Essentially risk-neutral.
SBUs. NEV analysis.
Not a buyer and seller of reinsurance.
The CEA was frequently in the trade press during the Fall of 1996. This section presents extracts from Business Insurance (BI), the National Underwriter (NU) and other sources to give a flavor of how events unfolded. It focuses on press treatment of the CEA’s reinsurance program.
A Call from Quakenbush
September 27, 1996: Governor Wilson signed the CEA package of bills
October 14 “CEA Law Prompts Allstate To Write Again In Calif.” (NU). A representative stated the creation of CEA “which we supported all along” will allow us to reduce our cat exposure “to the $1 billion range”. Allstate had stopped renewing policies shortly after Northridge.
Insurers representing 95 percent of the market in California stopped selling earthquake protection to new customers after the Northridge earthquake, according to the insurance department.
October 28 “CEA Gears Up For Kickoff on Dec. 1” (NU). Commissioner Quakenbush attended the annual Baden Baden reinsurance meeting. He “wanted to assure that reinsurance commitments were sill in place, given the fact the startup of the CEA was delayed for months” by the legislature. There is broad reinsurance market support for the CEA, with Continental European reinsurers committing USD 600 million, Lloyd’s USD 200 million the remainder of the London market USD 100 million, Bermuda USD 575.5 million and US/Canadian reinsurers USD 400 million, for a total of USD 1.875 billion. At least three companies, AXA Re, Swiss Re and Munich Re, authorized USD 100 million lines.
November 4: “California Quake Authority Suffers Growing Pains” (NU). Despite split votes on the three person board over rates and the CEO search, Commissioner Quakenbush says “[CEA] is on track and heading for a Dec. 1 kickoff”. He characterized the votes as “stylistic differences”. Thus far, carriers representing 48 percent of the market have signed on. The largest company that has not yet joined is Farmers, but they are expected to sign up by Nov. 8.
November 4: “Bermuda Writes $700M For Calif. Quake Facility” (NU). After a successful marketing trip Quakenbush reports Bermuda has the potential to become “the center of catastrophic reinsurance in the world if it chooses to move in that direction”. All eight property cat markets signed-on, with Ren Re the first to quote. He continued, “The fact that Bermuda came on so quickly motivated the rest of the markets to come on board.” The CEA USD 2 billion was program over-subscribed after the marketing trip. Quakenbush and his team are now “turning their efforts to raising the USD 1.5 billion of coverage which the CEA’s architects plan to tap from the U.S. capital markets”.
November 11: “CEA Creation Fueling Reentry of Large Insurers” (NU). Several large insurers have announced plans to reenter the homeowners market. Allstate and State Farm confirmed they will lift their moratoriums. Allstate “began selling policies last week” after a 24 month moratorium. “We did not want to have a moratorium, but because of everybody else leaving the market, we couldn’t be the only game in town” said a senior director at Allstate. Prior to Northridge State Farm had begun to cut exposure because of Andrew and other property losses. Farmers, which has had a moratorium since June 1994, is “developing [a] managed reentry.”
November 18: “California Earthquake Bonds Slated For Sale In Early ’97” (NU). Morgan Stanley, Bear Stearns and Goldman Sachs are set to bring a USD 2 billion offering to market early next year.
There has been softening in the reinsurance market, with “premium levels down 20-30 percent for the year”. In contrast CEA pricing as “remained firm” and as a result “companies are dedicating higher limits than they were last February to the CEA”. The CEA program is expected to be over-subscribed. REFERENCE THIS!
November 25: “Berkshire Hathaway to reinsure CEA” (BI) The CEA arranged up to $1.5 billion in reinsurance from Warren Buffett’s Berkshire Hathaway Inc., abandoning original plans to finance excess layers through capital markets. Reinsurance was less expensive, cumbersome to arrange and more plentiful than expected, said aides to California Insurance Commissioner Chuck Quackenbush. The cover will earn USD 590 million in premium for paying losses USD 1.5 billion excess USD 7 billion during a 4 year period starting April 1, 1997.
November 25: “California Quake Authority Scraps Bond Proposal” (NU) “State authorities, struggling to put together the final pieces of the CEA, have scrapped a proposed bond issue in favor of a more traditional reinsurance arrangement.”
“While state officials along with key insurers publicly remain optimistic the CEA will become a reality, private doubts are being raised as to how soon.”
“At least two self-imposed dead-lines by the CEA’s staff and advisory board to announce the finalization of the program’s financial underpinnings—the participation of insurers representing at least 70 percent of the market—have been missed.””
There are “conflicting reports” of who has signed. “Companies representing 60-65 percent are committed to the CEA, although in most cases contracts have not been signed”, officials say.
Greg Butler, the CEO of CEA, said the 70 percent market share threshold is likely to reached either via the inclusion of the FAIR Plan (with 240,000 policies and USD 14 billion of insured value, more than two-thirds of it in Los Angeles county) or by permitting 20th Century Industries to re-enter the market7.
The week before Thanksgiving 1996 the CEA was struggling to meet its statutory 70 percent participation hurdle, as the press clippings make clear. Sometime during that week Commissioner Quakenbush called Dennis Chookaszian, CEO of CNA, and asked if he was sure CNA did not want to join. As a result, a request came down through several layers of management to me.
Enterprise-wide Analysis
laid out CNA’s analysis of the CEA. It was performed by Personal Lines8 from a Personal Lines perspective. It was a straightforward analysis given agreement on the profitability of the minipolicy.
Given CNA’s 1994 earthquake market share of 1.4 percent and 1996 approximately 1 percent share, participating in the CEA involved a payment of USD 14 million and assumption of contingent liabilities to the two IALs of approximately USD 2.5 million per year on an expected basis in order for the right not to write its own profitable minipolicy. REFS
Details of the CEA’s reinsurance had been reported in the press since the original review, particularly the surprise move by Warren Buffett and Berkshire Hathaway to write the second layer. Hearsay in the markets at the time (backed up by subsequent published articles, LANE/Sedgwick) suggested the reinsurance layers were very well priced. CNA’s businesses included an active and growing assumed reinsurance operation, CNA Re. Alerted to the potential profitability of the reinsurance layers I called CNA Re management to find out how much capacity they committed to the first reinsurance layer. Lacking a reinsurance business perspective, I was surprised to learn the answer: very little. They acknowledged the CEA layer was well priced but had already used all their ``California quake aggregate’’ (i.e. allocated capacity) writing existing clients. And for them the relevant pricing benchmark was not a minipolicy, it was other California catastrophe business written at broadly comparable rates to the CEA’s treaty. Unlike the Bermuda markets mentioned in NU October 28, CNA Re had been unwilling to “warehouse” capacity for the CEA. It is understandable CNA Re did not to want to cut back or non-renew existing customers to hold capacity for an organization that may never write business. Since the CEA did not become operational until December 1 the capacity needed to be warehoused for almost a full year, bearing in mind the prevalence of January 1 renewals. Deference to incumbents is important in reinsurance where reinsurers make long-term commitments to their customers, staying on programs over multiple years. Nonetheless, it was surprising to me that CNA Re was unable to participate more.
[32] reports that Berkshire’s stock market valuation jumped by almost USD 300 million in excess of the board market change on the day the deal was announced (November 18). The total premium was USD 423 million at placed percentages.
The magnitude of the profit differential between the minipolicy, roughly an 80 percent combined with no reinsurance cost, and the reinsurance, less than a 20 percent loss ratio and a 30 percent combined ratio, seemed like an opportunity. Because of the CEA’s approach to actuarially fair rates, their pricing only included market risk loads for the reinsurance layers, but their pricing drove market rates for the whole tower. Therefore NPIs faced diluted pricing. Writing the reinsurance allowed to capture all of the CEA’s risk margin without assuming any business with no risk load.
CNA thus faced a classic ERM problem. A business unit silo-by-silo analysis said
- Personal Lines should not participate in the CEA because doing so amounted to paying money to be barred from writing profitable business and
- Reinsurance could not participate on the CEA reinsurance owing to commitments to existing clients and its need to manage within its risk budget and agreed aggregate limits.
Looked at holistically, the solution was clear. If CNA Personal Lines participated in the CEA it would lower Personal Lines’ and CNA’s earthquake exposure. The 100 percent exposure would be replaced with a liability for just the two industry assessment layers meaning the net exposure is lowered by the impact of the remaining elements of the CEA’s capital layer: the capital, debt, and reinsurance layers, as well as the OTT cap, which was quite material for CNA, . As a result, CNA could increase its exposure, by writing reinsurance, and leave CNA’s risk, as a whole, in the same position. Or almost the same position: it would also have to pay USD 14 million to join.
is a summary of the accounting and risk adjusted expected values, with the latter calibrated as before, to CEA pricing \(Q\) and to a 40 percent reduction, \(Q-0.4\). Insured losses are computed based on a safety level of 83.3, i.e. USD 100 billion claims paying capacity. The first three rows of each block show the 100 percent expected losses, IALs and reinsured losses (DOUBLE CHECK AGAINST TABLES AT END). The PI and NPI results reflect a 1 percent market share and 8.6 percent participation on the two reinsurance layers, based on actual cessions in 2000, . The table shows a large accounting gain from the reinsurance. There is a much smaller gain under \(Q\) because it is calibrated to the first layer pricing. Thus there is only a gain from the second layer, which had slightly preferable pricing. Under \(Q-0.4\) both layers look attractive and the gain is larger.
shows actual cash flows under the NPI, PI and PI with reinsurance scenarios. Adding reinsurance means participating is preferred through the 90th to 95th percentile, depending on the cat model output. There is then a zone when the reinsurance will incur losses and the position is worse. For very large losses the OTT benefit dominates.
shows the CEA’s reinsurers at year end 2000. (The CEA was not an insurance company and did not initially file statutory statements. The 2000 statement is the oldest I have been able to obtain. Other NAIC reporting companies did not list their assumptions from the CEA in their Schedule F reports.) It shows that CNA, lead company Continental Casualty Company, was the second largest reinsurer of the CEA after National Indemnity (Berkshire Hathaway). Thus CNA was able to obtain significant participations across the CEA’s reinsurance programs. It is impossible to tell from these exhibits how CNA’s participations were split between the first and second reinsurance layers.
| Reinsurer | Assumed Premium | Pct Total |
|---|---|---|
| National Indemnity | 93,361 | 38.9% |
| Continental Casualty Company | 20,735 | 8.6% |
| Lehman Re, Ltd. | 16,721 | 7.0% |
| Swiss Re | 16,141 | 6.7% |
| Lloyds | 14,362 | 6.0% |
| AXA Re | 11,341 | 4.7% |
| Partner Re | 8,721 | 3.6% |
| Tempest Re | 5,742 | 2.4% |
| Munich Re | 5,476 | 2.3% |
| XL Global Re | 5,444 | 2.3% |
| Zurich Re | 5,426 | 2.3% |
| National Union Fire Ins. Co. | 5,342 | 2.2% |
| Numerous smaller shares | ||
| Grand Total | 239,758 | 100.0% |
Was any other carrier positioned to execute the same strategy as CNA? It required a reasonably large book of residential earthquake insurance and an active reinsurance assumed operation. Residential earthquake is hard to isolate from statutory filings but will be correlated with homeowners premium. Based on multiline carriers with assumed reinsurance operations and 1996 California homeowners premium Chubb, Fireman’s Fund, Hartford, Zurich, Kemper and TIG appear potential candidates. The first companies two specialized in high value homeowners business which was not a good fit with the low-limits minipolicy. The last two were caught up in the industry’s general transition from a multiline model to a specialist model: Kemper Re was bought by GE Re in 1998 and TIG was in the process of being sold by Xerox. Hartford is left as a potential candidate. Its statutory filings show it was more focused on assuming casualty business than property. Zurich lacked the homeowners premium in 1996. They bought the Farmers attorney in fact in 1998 but the accounting for reciprocals would still not have given them the same opportunity as CNA. We conclude that CNA was in a unique position in the industry being able to execute the participate-and-reinsure strategy.
Meeting on the Monday?
Lessons for Effective ERM and Conclusions
November 26: CEA Certified. The State of California Insurance Commissioner certified that all statutory conditions necessary for the CEA to become operational had been met, and the CEA began writing earthquake policies on December 1, 1996.
Thursday November 28: Thanksgiving
December 1: CEA accepted its first insurance risks (Marshall p. 91).
December 2: “California Starts Marketing Earthquake Facility Cover” (NU) “California’s first-in-the-nation earthquake insurance program, culminating two years of negotiations, won final approval last after week to begin marketing policies after reaching—just barely—the threshold of insures’ participation required by law.””
Early expectations were that the CEA would attract 85 to 90 percent participation. It became operational with 71.6% of the market [including 1.5 percent from the California FAIR plan].
Commissioner Quakenbush said the no-fills policy, with a 15 percent deducible will have rates averaging about USD 3.55 per USD 1,000 of insured value.
The CEA solved the homeowners availability problem in California, with the large writers quickly re-opening for business. The importance of the CEA in this process is clear from the comments of Allstate, Farmers and State Farm (ref Sec 6).
The CEA had a beneficial effect on commercial lines too, with contemporary reports of higher capacity availability and rates down 15 to 20 percent (Dec 23/30, BI “CEA success may create beneficial aftershocks in commercial lines”).
Today the CEA stands as a financial success. It has been rated A- by A. M. Best Co. since 2002 and has remained privately financed, receiving no money from the state.
ERM Lesssons
Managing Large Organizations
EPS vs. EVA approaches.
ERM is difficult to put into action
- Hard to get info cross silos
- Who makes the decision? What is their motivation?
- Other parts of company may not react as you expect for various reasons
- Cos often not good at opportunistic bets (short term, high risk, non recurring, outsize for material benefit = what Warren did)
- Analyze everything, not just your part! Cf impact of assets (pension assets in stocks)
- Take the enterprise view!
- ERM should prepare you to make these decisions; not ctees and reports but an awareness and openness to broad thinking
- Even today cos not making good decisions e.g….
This paper has described how CNA, with 1.4 percent market share, decided to join the CEA and made clear they were in a unique position. What about the other 68.6 percent? That is a story with interesting implications about the structure of the PC industry and the nature of corporate risk tolerances that I will return to in a subsequent paper.
Lessons
[33] [34] [4] three from start.
There was no one to perform the analysis. No one owned the corporate view. Even with central risk aggregation it would not have been uncovered—needed to see inside data; would have picked up Cal quake exposure.
There was no communications between the towers to identify that there was an opportunity. Same risk without comms. (Had decided not to buy and sell reinsurance at the same time.) Same risk with different returns.
No one owned the corporate decision. Hold Co BU str. execs don’t run the “businesses”
Re didn’t want the out-sized risk. Didn’t act as expected. Not set up to capitalize on an opportunistic bet, c.f. Ajit Jain’s decision at BH. (again, Stein)
Re couldn’t warehouse capacity.
Re got the benefit. Personal lines paid the cost.
Arrow / Stein on optimal structure. Many risks in insurance “soft”. Because of (new) cat models this was a hard risk.
Unique opp for CNA.
Future research: how other companies decided to join.
Accounting vs. risk adjusted view of cash flows.
ERM frameworks …risk appetite, risk aware culture, etc. would not solve most of these problems. Not ctees, risk registers, ID, etc.
References
References
Footnotes
NU, 6/6/1994, 11/11/1996.↩︎
Alas not available on-line, but obtained by the author.↩︎
The original rate filing, made in 1996 and California DOI code #96-8834 is no longer available because it is past the document destruction retention period. The 1999 filing #99-377 is available from the WARFF website. The detailed analysis is available from the author on request.↩︎
The CEA covered claims up to \(8.75=10.5 / 1.2\) times its premium. For CNA (Continental Casualty Company) claims paying capital (USD 4.6 billion) was about \(255\) times earthquake premium (USD 18.1 million). However the capital covers earthquake from all sources, including commercial lines and workers compensation, not just residential. In the exhibits total losses are computed using USD 100 billion claims paying capacity. That represents \(83.3=100/1.2\) times premium, and a reasonable middle ground.↩︎
See Exhibit 20 page 2 in CEA filing 99-159. The insurance department insisted the CEA lower the filed loss cost to reflect their ability to pro rate claims. the CEA’s original filing did not make this adjustment (S.A. personal communication).↩︎
Continental Insurance Company is distinct from CNA’s lead company, Continental Casualty Company.↩︎
In the end The FAIR Plan participated; Twentieth Century did not.↩︎
At the time I was as a pricing actuary in the product management department of CNA Personal Lines. I performed the first analysis leading to the “do not participate” decision.↩︎