The risk of reinsurance: the Dutch case
Manuel Kampman1 February 2008
Abstract:
This study gives an assessment on how reinsurance claims could affect the stability of the Dutch insurance market. A scenario analysis is performed to quantify contagion risks. No evidence of systemic risk is found. Even when multiple reinsurance companies fail simultaneously (e.g. in a certain region) no contagion occurs. Next, it is analysed how many, and to what extent, individual primary insurers are affected following a particular shock. The life insurance industry is hardly affected by reinsurance failures, whereas the non-life industry is more vulnerable: a crisis in the European reinsurance market will cause a large balance sheet deterioration of several non-life insurers.
Keywords : reinsurance, contagion, simulation JEL Codes: G20, G22
1. Introduction
The proper functioning of a financial system is crucial in facilitating economic activities. When a financial crisis occurs, the cost can be severe and can lead to periods of low growth and recession (Allen and Gale (2000)). A sector that merits particular attention is the insurance industry, which can make important contributions to the economy. First, the insurance sector reduces the risk for economic agents, and encourages them to transact. Some transactions might not take place without the use of insurance. Second, the insurance sector fulfils some functions as a financial intermediary. Its main contribution lies in the ability for life insurance companies to obtain long-term savings (in contrast to banks, which obtain relatively more short-term savings). These long term savings increase the potential for long-term investment in the economy (Nagar (2005)).
Insurance companies aim for a delicate balance. On the one hand, risks are accepted from market participants, in exchange for a premium. When the insured event occurs, the insurance company is obliged to pay according to the terms of the contract. Payments have to come out of the (invested) premiums. Insurance
1
companies thus need to carefully balance their risks. If too much risk is accepted, this could lead to financial distress in times of high losses. On the other hand, part of the risk that insurance companies bear, is often transferred to other market participants.
One way to transfer risks to other market participants is to insure these risks with a reinsurance company. The reinsurer stands secure for part of the risk, in exchange for a premium. In case the insured event occurs, the primary insurer can claim the reinsured part on the reinsurer. The reinsurance industry thus acts as an insurer for the primary insurers. A second way an insurer can transfer (part of) the risk, is through securitization2. However, the analysis of this channel goes beyond the scope of this paper.
The reinsurance sector therefore plays an important intermediary role on a macroeconomic level. First, risks are transferred to other market participants improving the level of diversification (Group of Thirty (2006)). This reduces the impact of adverse shocks on the primary insurance companies, and limits the volatility of earnings. Second, the reinsurance market acts as an intermediary by balancing risk over, and across time. This increases the efficient allocation of risks. Third, reinsurance companies increase the overall underwriting capacity: Part of the liabilities are covered by the reinsurance company. This will increase the amount of capital effectively available by freeing equity that was previously tied up to cover risk. The latter will improve the balance sheet of the primary insurer.
Clearly, the risk transfer through the reinsurance sector provides advantages, but it also introduces several risk factors that could threaten financial stability (DNB (June, 2006)).
First, reinsurance introduces credit risk for primary insurers: the risk of reinsurance companies not being able to meet their liabilities. A failure of a single participant could have a large impact on other participants. If a reinsurer is not able to meet its liabilities, the primary insurer will have to honour its contractual
2
obligation to the policy holder itself, and thus face a loss3. In addition to the loss of the claim, the primary insurer also looses the premium already paid for the rest of the contract period, as most premiums are paid in advance. Acquiring new coverage will lead to additional costs. Credit risk also introduces possible contagion and systemic risk, especially in the life sector. Here the majority of claims is reinsured with Dutch counterparties (65 percent of total life, and 14 percent of total non-life claims, is reinsured at Dutch primary insurers). The failure of a Dutch insurer can thus result in the failure of other Dutch insurers through the claim outstanding from reinsurance.
Second, there is the risk stemming from retrocession: the risk arising from the risk transfer between reinsurance companies. When a particular (re)insurance company fails, this could spread to other (re)insurance companies, ultimately affecting the whole economy.
A third risk factor is a sudden increase in premiums. In times of limited losses, profits of the incumbent reinsurance companies will rise, attracting new firms. This will lead to a decrease in the general premium level. In case of large losses the opposite holds, and the general premium level will increase. These fluctuations in premium levels are called the “underwriting cycle”. In times of a high premium level, market participants find it hard to find coverage for a cost-effective price. In times of a low premium level, the insurance companies find it hard to survive, making multiple failures within a short time span more likely. Furthermore, a high volatility of general premium prices creates an unstable economic environment, and could lead to lower growth (see Lamoen (forthcoming)).
All in all, the existence of reinsurance contributes to the smooth functioning of the economy. It facilitates diversification, and an efficient allocation of risk over, and across time. It also increases the underwriting capacity of primary insurers. However, the use of reinsurance brings about several risks that could be a threat to financial stability: credit risk, contagion and possible systemic risk. In addition, there is the risk of a sudden increase in general premium level and retrocession.
3
In this paper we will focus on the credit risk arising from reinsurance, and investigate possible contagion effects. Market features like concentrated (re)insurance markets, risks that are becoming increasingly related, and the “nature” of the reinsured risks reinforce credit risk. In the next section we describe these factors in more detail. Both the insurance and the reinsurance market are highly concentrated. These markets show increasing concentration over the last couple of years (see e.g. Group of Ten (2001), Cole and McCullough (2006)). The global reinsurance market is dominated by a few large reinsurance companies, serving a big part of the global market. The top 5 insurance groups had a market share of 57 percent in 2001, and the top 10 had a market share of 77 percent (Sigma (2003))4. The Dutch non-life insurance market is fairly concentrated; in 2005 the four largest firms served about 50 percent of the non-life insurance market. The non-life industry is somewhat less concentrated, as the four largest companies served 27 percent of the life insurance market in 2005.
Next to the concentration in the insurance and reinsurance market, the risks insured are becoming increasingly related. The rising population density leads to more and more simultaneously occurring health and property risks. This will possibly increase the volatility of insurance shocks (Krenn and Oschischnig (2003)).
Furthermore, it is observed that the demand for reinsurance is declining because of international consolidation of insurers. As consolidated insurance companies have many kinds of risks spread over various regions, risks are already well diversified within each company. This leaves the reinsurance companies with relatively more large risks which primary insurers are wary to accept. As a result, the volatility of the shocks that reinsurance companies face is increased.
To sum up, a single (re)insurance failure may lead to large losses, both for the reinsurance and primary insurance market. This may affect the economy as a whole as well. Given the market features described above, the size of such an impact is of interest. Therefore, the central question in this paper will be:
What is the risk of reinsurance failures for financial stability in The Netherlands?
4
The paper is structured as follows. Section 2 provides a literature review that frames our empirical analysis. We will explain the various views on why reinsurance companies use reinsurance, and in addition explain how reinsurance can cause contagion. Previous studies relating to systemic risk and reinsurance will be reviewed. Section 3 outlines the methodology. In section 4 we give a description of the data used, section 5 describes the data characteristics, while the results are displayed in section 6. No contagion effects are detected, therefore we analyse the impact on individual insurers of several scenarios of reinsurance failures. In section 7 we describe the policy implications that emanate from our results. The final section presents the conclusion.
2. Literature review to frame the empirical analysis
We first explain why insurance companies use reinsurance. Then we describe how the use of reinsurance could pose a risk to financial stability, and review some recent studies concerning reinsurance and contagion.
2.1 Reinsurance
There are several factors that can explain the use of reinsurance. Roughly two views can be distinguished in the literature: reinsurance as an optimal risk sharing device and reinsurance as a capital structure decision. The combination of the two views emphasizes the dual nature of the use of reinsurance.
sell part of the risks to each other, and aim for pooling the aggregate risk at an equilibrium. However, in practice it is observed that primary insurers reinsure the majority of the risk to “specialized” reinsurers (i.e. only active in reinsurance activities). Another example is provided by Mayers and Smith (1990). Mayers and Smith (1990) observed that small and geographically undiversified firms reinsure relatively less. Optimal risk sharing theory would predict a more extensive use of reinsurance for these particular firms.
Alternative explanations can be found in the second view, which is based on corporate hedging theory. Motivations for the use of reinsurance are similar to reasons why non-financial firms use insurance: the use of reinsurance effectively reduces leverage. Therefore, reinsurance can be interpreted as a capital structure decision (Garven and Lamm-Tennant (2003)). For this capital structure decision to make any sense, capital cost frictions like e.g. taxes, agency cost and bankruptcy cost need to be introduced.
For example, a progressive tax code can create a demand for reinsurance. Reinsurance effectively lowers earnings volatility which in turn lowers expected tax liabilities. Smith and Stulz (1985) and Mayers and Smith (1990) find evidence that reinsurance lowers the expected tax liability. A similar mechanism applies within insurance groups. Financial reinsurance contracts could be used to optimally transfer profits between group members5.
Garven (1987) argues that the theory of finance can be applied to the insurance firm. Similar to non-insurance financial firms, agency costs increase with higher leverage. As shareholders have a call option on the residual cash flows of the firm, a higher leverage will be an incentive for the firm’s management to pursue selfish strategies. The management of the firm maximizes the expected return on the call option, and will therefore have an incentive for example to, undertake more riskier projects.
Carson and Hoyt (1995) seek to explain life insurance insolvencies, using among other things bankruptcy cost. The idea is that bankruptcy cost are first reduced when purchasing reinsurance: this will limit the annual fluctuations of losses that the insurer has to bear on his own account. However, at a certain point,
5
too much reinsurance increases bankruptcy cost again as a large part of the cash flow is reserved for reinsurance payments. These small losses that are not reinsured might lead to bankruptcy. To measure the potential effect of leverage on the demand for reinsurance, a variable measuring a firm’s gross premium written to equity is included. As reinsurance is a substitute for equity, a positive relationship is expected between direct premium written to equity and the demand for reinsurance: firms with a relatively large premium to equity ratio will face a higher probability of insolvency because equity acts like a buffer for unexpected losses. A high ratio of premium written to equity will increase the demand for reinsurance as the probability of bankruptcy increases. Empirical evidence suggests the existence of an optimal level of leverage: if the level of leverage exceeds the optimum, it actually decreases firm value.
A more recent study of Cole and McCullough (2006) estimates the demand for reinsurance by U.S. insurance firms using more traditional factors (e.g. tax, agency cost, and bankruptcy cost) and by incorporating the state of the international reinsurance industry into the model. Factors like the price and the amount of liquidity available in the reinsurance market are used to represent the state of the reinsurance market. When reinsurance prices are high and the amount of liquidity available is low, the demand for reinsurance is expected to be low (i.e. an underwriting cycle). Cole and McCullough (2006) show that the price of reinsurance is negatively correlated with the demand of reinsurance: a increase in the cost of coverage will decrease the demand for reinsurance.
Additional information concerning explanatory variables for the demand for reinsurance (for instance, service efficiencies and investment incentives) can be found e.g. in Mayers and Smith (1990), Cole and McCullough (2006) and Lamoen (forthcoming).
2.2 Reinsurance failures and channels of propagation
Two channels can be isolated through which a reinsurance failure can influence the primary insurer: credit risk and information contagion. We will cover each of these in turn.
2.2.1 Credit risk and systemic risk
Credit risk arises from the possibility that a reinsurer does not pay the claim outstanding to the primary insurer, because of for example a dispute or a reinsurance failure (Ollodart (1998)). Reinsurance risk has been described in Cummings (1995), Das et al. (2003) and Krenn and Oschischnig (2003) as the largest component of the credit risk for an insurance company. According to Adiel (1996), three causes can be distinguished. First, the claim of the primary insurer on the reinsurer cannot be collected. However, the primary insurer still has an obligation to pay the policy holder according to a specified contract. This decreases the financial strength of the primary insurer as losses are subtracted from the companies equity. Second, the loss of financial strength can decrease the credit rating of one, or multiple, primary insurers. When credit ratings are adjusted downwards, the cost of capital increases and will lead to a deterioration of the financial position of the firm. Third, if a guarantee fund is in place, a failure could force the other insurance companies to pay for the claims of the failed reinsurer. If losses resulting from a failure of reinsurance company, are large enough (i.e. greater then the firm’s equity), a primary insurer could fail. This might result in the failure of other primary insurers, as some primary insurers also offer reinsurance.
and analysed for their risk potential. Sigma (2003) identifies three sources of risk and concludes that the risk potential resulting from reinsurance is low. First, based on total global premiums written, only a small percentage (6 percent) of total risk is reinsured. On the aggregate, reinsurance constitutes a small part of the balance sheet of primary insurers. Second, based on the credit ratings of reinsurance companies, the probability of a reinsurance bankruptcy is low. The latter implies that the risk of retrocession is also low. However, this could become a threat in the future. Third, the Sigma (2003) study notes that only a small number of reinsurance companies have failed in the past, and none of the failures showed any evidence of being a threat to systemic stability.
addition, many primary insurers offer both reinsurance and primary insurance products.
2.2.2 Information contagion
The second channel consists of information contagion. Information contagion occurs when market participants are imperfectly informed about the type of shocks hitting insurance companies (i.e. asymmetric information). The arrival of new information (e.g. a reinsurance failure) about a particular reinsurance company can be interpreted as information about rival reinsurance firms. The new information could signal an upcoming crisis in the reinsurance industry, indicating upcoming losses for primary insurers. Investors in primary insurance companies can react by selling their shares in the company, causing the financial position to deteriorate, ultimately leading to defaults. DNB (June, 2006), for example, mentions that the failure of a single reinsurance company could cause reputation damage for the whole industry. Note that a classical bank run (as in the Diamond and Dybvig (1983) model), where depositors withdraw their deposits, is not possible in the insurance industry. In the non-life industry, a premium is paid in advance to cover a particular risk, this premium can not be withdrawn during the contract time. In the life industry the premium also contains a savings component, however it is difficult to withdraw the amount saved on a short notice. Often savings can only be withdrawn subject to a fine.
Avila and Eastman (1995) follow a similar approach. The effect of information releases regarding the failure of four major U.S. life insurance companies is investigated in the period ranging from 1990 up to and including 1991. The response of the capital market is measured for three different groups of rival firms. These groups are classified according to the stock exchange at which they are quoted.
The contagion effect is modelled by using a dummy variable to capture the abnormal return on announcement dates. Weak evidence for information contagion is found at only one stock market group, the other two groups did not show any evidence for the existence of information contagion.
Another possibility for information to cause contagion is through financial conglomerates. In some cases, risk is reinsured with other conglomerate members. When one of the conglomerate members is in financial distress, investors might interpret this signal as trouble for the whole group. As a result, confidence in all group members will decrease, and consequently financially sound group members can, also run into trouble. Nagar (2005) emphasized that the links between insurance companies and banks are the most important channel which insurance failures could cause systemic risk. Through e.g. mergers or joint business ventures, banking and (re)insurance companies are becoming increasingly interconnected. Information asymmetry could undermine the trust in the whole group and result in depositors withdrawing their deposits. This could result for example a run on the bank and thus cause contagion. Investors could also decide to sell their shares in other group members.
3. Methodology
3.1 Insurance Matrix
To represent the system of reinsurance claims we make use of the same methodology used to describe the interbank linkages in the banking market (see e.g. Upper and Worms (2004), Van Lelyveld and Liedorp (2006) and Elsinger et al. (2006)). Note that this way of describing and analysing the system of (re)insurance claims has not been used before in the existing literature and offers a new perspective on how to analyse such problems.
To represent the reinsurance claims outstanding to primary and reinsurers, we use a matrix X (figure 1), where “P” represents a primary insurer and “R” stands for a pure reinsurer.
Figure 1 : Matrix of insurance and reinsurance claims
P1 Pj Pn R ∑j P1 P11 P1j P1n A1 Pi Pi1 Pij Pin empty Ai X= Pn Pn1 Pnj Pnn An Rn+1 R11 R1j R1n Bn+1 Ri Ri1 Rij Rin not known Bi Rm Rn1 Rnj Rnn Bm ∑i I1 Ij In
Pij (Rij) represents the amount that primary insurer j has reinsured with insurer
(reinsurer) i. Hence, the total amount of reinsurance claims outstanding of insurer
j on other companies is represented by column total Ij. The rows represent the
claims on an insurance (reinsurance) company Pi (Ri). The total amount of claims on an insurer (reinsurer) following from reinsurance contracts is represented by Ai
(Bi). Therefore, Bi represents the total amount that a reinsurer has insured for
other reinsurers are not public (i.e. “unknown”). The cells on the main diagonal are all zero’s as individual institutions can not reinsure at themselves.
Note that Pij represents a reinsurance claim outstanding from one primary insurance company to another. As described in the introduction, 14 percent of total non-life and 65 percent of life reinsurance claims are reinsured at other primary insurers. This implies that some primary insurers are “hybrid insurers”: primary insurers that also act as a reinsurer.
3.2 Scenario analysis
Based on the insurance matrix X, shocks can be applied to the insurance system. In the first three shocks we assume that a particular set of (re)insurance companies fail in turn as the result of some exogenous shock. The set of reinsurance companies that fail are represented by the following shocks:
1. All single reinsurance companies fail in turn. 2. All Dutch non-life insurers fail in turn. 3. All Dutch life insurer fails in turn.
In shock one we exclude the failure of Dutch hybrid insurers (primary insurers that also operate as a reinsurer), as they are included in shock two and three. Note that complete idiosyncratic shocks are quite rare. It is more likely that multiple insurance companies are affected by a shock. However, operational risk is a different matter. For instance, Kobe Reinsurance (Belgium, 1995), Dai-Ichi Kyoto Re (Belgium, 1995) and Ardenia S.A. (Luxembourg, 1993) all failed as the result of fraud. These failures therefore affected only single institutions and reflect an idiosyncratic shock.
In shock four till seven, we let one particular group of re(insurers) fail simultaneously as the result of an exogenous shock.
4. In a particular region, all reinsurance companies fail simultaneously. 5. All Dutch life insurance companies fail simultaneously.
Again, similar to shock one, we exclude hybrid insurers from shock four. The first three shocks measure concentration risk, i.e. whether a single reinsurance failure has a large impact on primary insurers. In addition, in the second and third scenario we measure the impact of the linkages between the life and non-life sector. This is especially relevant since many primary insurance companies reinsure with other Dutch primary insurers. Shocks four through six represent problems in a certain region, or at a group of primary insurers caused by for instance retrocession (see e.g. the London Market Excess of Loss ("LMX") spiral). Shock seven is quite unrealistic, and will be used for theoretical purposes only.
3.3 Contagion and the impact on an individual level
First, we analyse the impact of (re)insurance failures on the financial stability of the system. One, or multiple, failures of primary insurer(s) could lead to the failure of other primary insurers, and cause systemic risk (i.e. domino effects/contagion risk). Second, we analyse the impact of reinsurance failures on a individual firms. Such an analysis might provide additional information as some shocks might not lead to (contagious) default(s) but might nevertheless have a significant impact in terms of for instance profitability or solvency.
3.3.1 Contagion risk, impact on an system level
All simulations involve a failure of a (re)insurer. Which claim(s) are initially lost by the failure of a (re)insurer is determined by each of the seven scenarios specified above. The failure of a primary insurer could lead to additional failures (systemic risk). A primary insurance company fails when the sum of the loss following the failure of one or multiple reinsurer(s), is larger than its equity.
θ*(
∑
N=1
j R ) > Eij j , (1)
matrix X by multiplying the row of claims of (re)insurer j by (1-θ). For instance, a failure with a loss rate of 100 percent transforms all the claims in the row of (re)insurer j to zero.
Note that we keep the loss rate constant over time because the counterparties can not react by reducing their exposure following a signal for possible trouble: counterparties are bound to contracts that can not be changed.
3.3.2 Impact on an individual level
To analyse the effect of (re)insurance failures on individual primary insurers, we make use of three different indicators; equity, solvency, and profit. Each of these indicators (described below) provides a different view on what might be called a significant impact.
Equity
One way to measure the impact of (re)insurance failures on primary insurers is to measure the percentage equity lost. Equity acts as a buffer, and the percentage equity lost reflects the loss in the buffer of the primary insurer. Note that if there is an equity change above 100 percent, the primary insurer becomes insolvent and will fail as a result.
Solvency
The solvency of an insurance company represents its ability to pay for its future liabilities and is defined as:6
Solvency =
∑
= − N j ij R provisions Technical Equity 1 (2)This ratio represents the leverage of the institution. When a reinsurance company fails, diminishing the second term in the denominator, the primary
6
insurer still has an obligation to its policy holder(s) (reflected by its technical provisions). As a result, the solvency ratio will decrease. We measure the percentage change in solvency as a result of the reinsurance claim lost. For instance, a drop of the solvency ratio from 60 percent to 30 percent, is represented by a change in solvency of 50 percent (30/60). A decreasing solvency ratio signals worsening financial health of the primary insurer.
Profit
To indicate the severity of a loss, we also measure the impact in terms of year’s profit. However, this method is likely to generate outliers: profits can fluctuate heavily from year to year and, ceteris paribus, the impact will increase if profits are low, and vice versa. This indicator is useful when the other two indicators contradict.
Note that in upcoming results we only display the percentage solvency change. The percentage equity loss and the impact in terms of years` profit give similar results and did not contradict. These indicators did not provide additional information and are therefore not shown.
4. Data
the reinsurance company is a pure reinsurer or a primary insurer. Data is available for 2003, 2004 and 2005, but we only use 2005 as reported reinsurance claims are very stable over time.
The primary insurance market
In 2005, the Dutch insurance sector consisted of 315 firms in total with 77 life insurance firms (excluding pension funds) and 238 non-life firms. In both the life and non-life industry the number of firms decreased over the years. The total number of firms dropped from 361 in 2003 to 315 in 2005 with a slightly larger decline in the non-life compared to life insurance market (-7 percent vs. -14 percent). The decline in the total number of firms is mainly the result of a consolidation process.
The life insurance sector is more concentrated compared to the non-life insurance sector. In 2005, the four largest life insurers, served 81 percent of the life market (measured in total premiums), where the ten largest firms had a market share of nearly 93 percent. The four largest non-life insurance companies served 43 percent of the non-life market, where the ten largest firms have a market share of nearly 69 percent.
The reinsurance market
Considering the reinsurance claims outstanding from Dutch insurers, we note that there are a few large reinsurers that serve a big part of the Dutch market. The four (ten) largest reinsurance companies hold approximately 37 (61) percent of all outstanding reinsurance claims.
Geographical spread
Figure 2 displays the percentage of total claims reinsured in each region.
Figure 2: Percentage reinsured by region
Dutch European
RoW
life insurance sector
Dutch
European RoW
non-life insurance sector
In the left panel of figure 2, it is observed that life insurance companies reinsure almost all their risk in the Netherlands (88 percent). The non-life insurers (right panel of figure 2) reinsure their claims on average more equally to the discerned regions: 42 percent in Europe, 24 percent in the Netherlands and 34 percent in the rest of the world. The dataset also allows us to see whether the reinsurer is a pure reinsurer (only active as reinsurer) or a hybrid reinsurer (both active as insurer and reinsurer). There is a remarkable difference between the regions: on aggregate, 78 percent of reinsured risk in the Netherlands is carried by Dutch primary insurers. In Europe and the rest of the World, the opposite holds: almost all the risk (in Europe and in the rest of the world 86 and 99 percent) is carried by pure reinsurance companies.
Percentage risk reinsured
the total technical provisions. In the left panel of figure 3, the percentage of total risk that is reinsured is displayed and in the right panel of figure 3, the percentage reinsured is reflected in terms of total technical provisions. On the horizontal axis, firms are divided into three groups: small, medium and large. In addition, we also differentiate between life and non-life companies. The vertical axis displays the percentage of reinsured risk. The shaded part of the box plot represents 50 percent of all observations, the upper/lower boundary of the plot represent the largest/smallest non-outlier observation7. The dots represent outliers.
Figure 3: Percentage of total risk reinsured
0 2 0 4 0 6 0 8 0 1 0 0 % Non-life Life
Small Medium Large Small Medium Large In terms of premium 0 2 0 4 0 6 0 8 0 1 0 0 % Non-life Life
Small Medium Large Small Medium Large In terms of technical provisions
Note: each box-and-whisker plot is constructed using the number of companies in a particular class.
Some interesting features are observed in figure 3. First, life insurers only reinsure a small part of total risk, non-life insurers reinsure relative more. However, some small life insurers do reinsure a significant part of their total risk. Generally in the non-life industry smaller insurers reinsure relatively more compared to larger firms. The large differences of percentage risk reinsured between the life and non-life insurers could be explained by the nature of life insurers: life insurers face less uncertainty because of relatively long-term contracts and in addition, the risk insured is more predictable. Therefore the need for reinsurance is less.
7
Data on consolidated level.
In the dataset there are numerous (re)insurance companies that belong to the same conglomerate. We observed that 37 percent of all claims outstanding are reinsured between groups members. In this paper we use this information in two ways. First, we use group data to investigate whether some conglomerates are hit more frequent, and more severe compared to other conglomerates. Second, when the instability of the system is caused by reinsurance claims between group members this should be taken into account when considering policy implications.
To sum up, based on the description of the data above, it is observed that Dutch life insurers only reinsure a small part of premium received. Contagion risk threatening the life insurance sector is likely to originate from Dutch primary insurance companies. The impact on the non-life sector is expected to be larger, especially for small firms who on average reinsure the most relatively to other groups. Reinsurance failures in Europe and the rest of the world will probably have the largest impact on non-life insurers.
5. Description of the matrix
In 2005, 229 primary insurers out of a total of 315 use reinsurance. These insurance companies have 1329 claims outstanding on a total of 323 reinsurance companies. This results in a square matrix X with 552 dimensions. The cells on the diagonal are all zeros, as individual insurance companies do not reinsure with themselves. The upper right quadrant of matrix X is empty because reinsurance companies generally do no reinsure at primary insurers. The bottom right quadrant is unknown, as there is no data about claims outstanding between reinsurance companies. However, Sigma (2003) finds that reinsurance companies reinsure approximately 21 percent at other reinsurance companies, possibly imposing a threat to financial stability. The risk of retrocession is left for future research.
Table 1 presents the descriptive statistics for the different group sizes. Small non-life insurers use on average a smaller number of reinsurance claims (4.4) compared to the medium and large insurers (14.5 and 8.8). In addition small insurers have on average smaller claims outstanding (€1,8 Million) compared to medium and large insurers (€8,8 and €122,8 Million). In the life industry a similar situation applies: smaller life insurers have on average a smaller number of reinsurance claims (3.7) outstanding compared to medium and large life insurers (7.6 and 13.0). Surprisingly, small insurers have a larger average claims size (€25,3 Million), compared to medium life insurers (€12,3 Million). On aggregate, life industry reinsurance claims make up 3.4 percent of total equity and less then 1 percent of the total assets. In the non-life industry total reinsurance claims make up 11.6 percent of total equity and 3.1 percent of total assets. In the life industry, total accounting profits make 51 percent of total claims whereas in the non-life industry this percentage amounts to 92 percent.
6. Results In the first part of the results, we quantify systemic risk arising from reinsurance. Based on the matrix a scenario analysis is run to detect whether there are contagion effects In the second part of the results, we quantify the impact on individual insurers. This is done by analysing the change in solvency following a particular shock.
6.1 Systemic risk
To quantify possible systemic risk, each scenario as described in paragraph 3.2 is executed using a loss rate of a 100 percent. First, we let each (re)insurance company fail in turn (shock 1 through 3). Second, we let a particular group of (re)insurers fail simultaneously (shock 4 through 7). None of the scenarios resulted in second round effects, only first round effects were detected. Even if all reinsurance companies would fail simultaneously (shock 7), we only observe insurance failures in the first round and these failures did not cause other primary insurers to fail as well. The reinsurance claims are not large enough to cause systemic risk.
6.2 Impact analyses on individual insurers
In this subsection additional information is provided by running several scenarios and analysing the impact on individual insurers (in contrast to the impact on the stability of the system). The impact of one of these shocks might have a significant impact (in terms of e.g. solvency) on the industry. The aim is to identify how many primary insurers are affected following a particular shock, and to what extent. In this way it we can observe whether particular scenario outcomes might threaten the Dutch insurance industry.
is used to investigate whether there are some conglomerates that are hit harder, and more frequent than other conglomerates.
Failures of reinsurance companies in a particular region
Figure 4 presents the impact on primary life insurance companies from a simultaneous failure of reinsurance companies in a particular region. The horizontal axis displays three scenarios: “Dutch” represents a simultaneous failure of all Dutch reinsurers, while “European” represents the simultaneous failure of all European reinsurance companies. The simultaneous failures of all reinsurance companies in the rest of the world is labelled “RoW”. Within each scenario, we differentiate between size. On the vertical axis the decrease in percentage solvency is displayed. Note that a log10 scale is applied, and represents only decreases in solvency (e.g. a deterioration of 0.01, 10 and 100 percent is represented by -2, 1 and 2). This type of graph will be used in the remainder of this paper.
Figure 4: Impact on life insurers, 100% loss rate.
-4 -2 0 2 % S o lv e n c y c h a n g e (L o g 1 0 )
Dutch European RoW
Small Medium Large Small Medium Large Small Medium Large
percentage solvency change of below two percent. In each of the three scenarios, some small companies face a large balance sheet deterioration. The largest impacts could occur when Dutch reinsurance companies simultaneously fail, this will lead to a decrease in solvency of 103.3 and 83.2 percent respectively. However, the general impact on small companies is low.
The small impact on the life insurance firms was to be expected as life insurance companies only reinsure a small part of their total risk. It is also not surprising that the main risk originates from Dutch reinsurers, as the largest part of the reinsured risk is reinsured by Dutch reinuinsurers.
In figure 5 (similar to figure 4), the impact of the simultaneous failures in a certain region on primary life insurers is presented.
Figure 5: Impact on non-life insurers, 100% loss rate.
-2 -1 0 1 2 % S o lv e n c y c h a n g e (L o g 1 0 )
Dutch European RoW
Small Medium Large Small Medium Large Small Medium Large
50 up to and including 228 percent. Medium and large non-life insurers do not face a large balance sheet as a result. The simultaneous failure of all European reinsurers (in the middle of figure 5) would result in a more significant impact. One large primary insurer will face a decline in solvency of 93 percent; two medium sized non-life insurers will see a drop in solvency of 107 and 50 percent. Other large and medium sized firms only have minor changes in solvency. In addition, 11 small non-life insurers will experience a large balance sheet deterioration presented by a decline in solvency ranging from 223 up to and including 50 percent. A simultaneous failure of all Dutch reinsurance companies (in the left of figure 5) would only lead to large balance sheet deteriorations of small non-life firms: 11 firms would experience a decline in solvency ranging from 239 up to and including 39 percent. Large and medium sized firms face only moderate balance sheet deteriorations.
The overall impact on the non-life insurers is larger compared to the life industry. Especially in case all European reinsurance companies fail simultaneously, there are significant balance sheet deteriorations at multiple Dutch non-life companies.
Failure of the two largest life and non-life reinsurance companies
Similar to the Group of Thirty (2006) study, we assume that a percentage of reinsurance industry capacity is suddenly lost. Here a drop in capacity is presented by the simultaneous failure of the two largest life, and non-life reinsurers. In the life insurance industry, such a failure accounts for 62 percent of total claims outstanding whereas in the non-life industry, this accounts for approximately 24 percent of total claims outstanding.
A failure of a particular reinsurance company can lead to losses in both the life and non-life industry. Here the life and non-life sectors are analysed separately. When we would only select reinsurers based on total claims outstanding, then reinsurers that primarily reinsure life insurance risk would be overrepresented (because of their relatively large claims). If reinsurance companies both reinsure for life and non-life companies, claims are separated into a life and non-life share8.
8
Figure 6 contains two scatter plots. The left (right) scatter plot represents the failure of the two companies that reinsure the largest amount of claims to life(non-life) insurers. On the horizontal axis, the size of the institution hit is represented in terms of total assets. The vertical axis represents the impact in terms of the change in solvency.
Figure 6: Impact on life and non-life insurers, 100% loss rate.
-2 -1 0 1 2 C h a n g e i n S o lv a b ili ty , L o g (1 0 ) 8 9 10 11 Assets, Log(10) Life -2 -1 0 1 2 C h a n g e i n S o lv a b ili ty , L o g (1 0 ) 6 7 8 9 10 Assets, Log(10) Non-Life
In the life industry, only a few institutions are affected. Although 62 percent of total outstanding claims are lost, the life insurance sector as a whole is not seriously damaged as only two primary insurers face a large balance sheet deterioration. The non-life insurance sector is affected relatively more by the simultaneous failure of the two largest non-life insurers as more institutions are hit. However, such a shock would -in absence of strong information contagion effects- most likely not threaten financial stability. Only two non-life firms really run into trouble, the remaining firms face moderate balance sheet deteriorations (i.e. a decline in solvency of 40 percent or less).
Consolidated data
three indicators). Subsequently, all median impacts of each group member are selected, and from this set the median is taken once more. In this way a group median is obtained which is done for each group. Groups of less then 3 primary insurers are excluded, as these groups could give a distorted image of the severity of the impact (i.e., the group median and mean would in that case obtained by considering only one or two observations). Note that groups members only include reinsurers and primary insurers that engage in reinsurance activities in the Dutch market.
On the horizontal axis of figure 7, total group assets are shown. Total assets are obtained by simply aggregating the assets of each group member. On the vertical axis, the median group impact is shown. The same analysis is conducted using the average group impact. This analysis generates similar results and are therefore not shown.
Figure 7: Frequency and size of group impact, 100% loss rate.
-3 -2 -1 0 1 2 (M e d ia n ) C h a n g e i n S o lv a b ili ty , L o g (1 0 ) 6 8 10 12 Assets, Log(10)
7. Policy implications
In this section some remarks regarding future policy implications are presented. Furthermore, some possible threats are addressed. In the analyses it is observed that especially the non-life insurers reinsure a considerable part of their risk with foreign counter parties. European and reinsurers from the rest of the world operate under different legal systems and deal with different rules regarding e.g. capital requirements. Described in Nagar (2005) and IAIS (2006), the global nature of the reinsurance business generates some threats to the system as a whole. Large reinsurance companies operate on a world wide scale: risks are transferred to, and within, large groups. It may not be clear who bears the final risk and whether this party is able to pay out in case of an large loss event. In some countries there could be little supervision, or the supervision regime could be very accommodating. For instance, Bermuda is considered as an offshore centre where rules and supervision are relative less strict. To reduce the problem of (the lack of) supervision, one could pursue more harmonization of supervisory rules and increasing cooperation between different supervisory authorities. This could be combined with more transparency concerning who bears the final risk. This could for example be achieved by imposing an obligation for registering certain large exposures. However, the volume of risks sold on the public market (through e.g. securitization) is rapidly increasing. This will make more transparency harder to achieve.
In the Netherlands, a considerable part (37 percent) of the accepted risk is reinsured between group members. Analysis showed that especially small groups are vulnerable to reinsurance failures. Supervision so far has focussed primarily on single institutions, however risks are managed on a group level and should therefore be supervised according to this perspective.
8. Conclusion
a new perspective on how to assess systemic risk originating from reinsurance claims.
Based on several scenarios, we find no evidence of systemic risk. Even when multiple reinsurance companies fail simultaneously (e.g. in a certain region) no contagion occurs. Only single primary insurance failures are observed. The failures did not cause other primary insurers to fail as well. The system as a whole is not threatened by reinsurance failures.
In addition it is analysed to what extent the balance sheets of individual primary insurers deteriorate as a result of reinsurance failures. The life industry does not face the risk of large financial losses at multiple firms. Only when reinsures failed simultaneously in a particular region, a few small firms were faced with a large balance sheet deterioration. The non-life industry is more vulnerable for the risk from reinsurance. When for example all European reinsurance firms would simultaneously fail, one large, two medium and several small sized firms will probably run into financial distress. This will probably not lead to a total disruption of the financial system as policies are likely to be transferred to other primary insurance companies. However, such an event could increase uncertainty in the market and cause reinsurance premiums to rise, increasing insurance costs for policy holders and in this way have an effect on the real economy. As described in the literature review, such a large shock could lead to information contagion.
In this paper we only consider risk originating from credit risk. However, there are additional risks that should be incorporated into the analysis to get a more complete risk assessment. First, one should also consider cross-participations: part of the capital of an insurer is invested into bonds and shares of rival companies. Second, there are some events that can cause financial deterioration at multiple primary insurers. It is for instance not clear whether the insurance sector is able to handle large loss events that affect multiple primary insurers (Nagar 2005). Third, one should take into account possible information contagion. Information concerning other reinsurers, or group members can trigger for example a drop in share prices.
References:
Adiel, R. (1996), Reinsurance and the Management of Regulatory ratios and Taxes in the Property-Casualty Insurance Industry, Journal of Accounting and
Economics, Vol. 22 : 207-240.
Allen, F. and Gale, D. (2000), Financial Contagion, Journal of Political
Economy, Vol.108, No.1: 1-33.
Avila, S. and K. Eastman, (1995), The Market Effects of Life Insurer Insolvencies and the Implications for Regulators, Journal of Insurance Regulation, Vol.13: 301-329.
Borch, K. (1962), Equilibrium in a Reinsurance Market, Econometrica, Vol. 30, No.3: 424-444.
Das, U.S. and Davies, N. and Podpiera, R. (2003), Insurance and Issues in Financial Soundness, International Monetary Fund, Working Paper, No.138. Carson, J.M. and Hoyt, R.E. (1995), Life Insurer Financial Distress:
Classification Models and Empirical Evidence, The Journal of Risk and
Insurance, Vol. 62, No.4: 764-775.
Cole, C.R. and McCullough, K.A. (2006), A Re-examination of the Corporate Demand for Reinsurance, The Journal of Risk and Insurance, Vol. 73, No.1: 169-192.
De Nederlandsche Bank (2006), Prudential Supervision: Current Developments in the Financial Sector, Quarterly Bulletin (September): 22-30, Amsterdam. De Nederlandsche Bank (2006), Financial Stability, The Reinsurance Market, Quarterly Bulletin (June): 37-44, Amsterdam.
Diamond, D.V. and Dybvig, P. (1983), Bank Runs, Deposit Insurance, and Liquidity, Journal of Political Economy, Vol. 91: 401-419.
Elsinger, H. and Lehar, A. and Summer, M. (2006), Using Market Information for Banking System Risk Assessment, International Journal of Central Banking, Vol. 2: 135-165.
Garven, J.R. (1987), On the Application of Finance Theory to the Insurance Firm, Journal of Financial Services Research, Vol. 1, No. 1: 57-76.
Garven, J.R. and Lamm-Tennant, J. (1995), The Demand for Reinsurance: Theory and Empirical Test, Working Paper, University of Texas at Austin and
Villanova University.
Group of Ten (2001), Report on Consolidation in the Financial Sector.
Group of Thirty (2006), Reinsurance and International Financial Markets, Group of Thirty, Washington, DC.
International Association of Insurance Supervisors (2006), Global Reinsurance Market Report, November.
Krenn, G. and Oschischnig, U. (2003), Systemic Risk Factors in the Insurance Industry and Methods for Risk Assessment, Oesterreichische Nationalbank, Financial Stability Report, No.6: 62-74.
Lamoen, R. van, (Forthcoming), Go with the Flow, Theory and Practice of Underwriting Cycles in Insurance Sectors.
Lewis, R.E. (1998), Capital from an Insurance Company Perspective, FRBNY
Economic Policy Review, Page: 183-185.
Lelyveld, I. van, and Liedorp, F. (2006), Inter-bank Contagion in the Dutch Banking Sector: a Sensitivity Analysis, International Journal of Central
Banking, Vol. 2(2): 99-133.
Mayers, D. and Clifford, W. (1990), On Corporate Demand for Insurance:
Evidence from the Reinsurance Market, The Journal of Business, Vol. 63, No.1, Part 1: 19-40.
Nagar, W. (2005), The Insurance Sector and Financial Stability: An International Perspective and an Assessment of the Situation in Israel, The Geneva Papers, Vol. 30, No. 1: 65-71.
Ollodart, B.E. (1998), Uncollectible Reinsurance Reserves, American Actuarial LLC, Wallingford Connecticut.
Plantin, G. (2006), Does Reinsurance Need Reinsurers?, Journal of Risk and Insurance, Vol. 73, No.1: 153-168.
Polonchek, J. and Miller, R.K. (1999), Contagion Effects in the Insurance Industry, The Journal of Risk and Insurance, Vol. 66, No.3: 459-475. Smith, C. and Stulz, R. (1985), The Determinants of Firm’ Hedging Policies, The Journal of Financial and Qualitative Analysis, Vol.20, No.4: 391-405. Sigma, (2002), An Introduction to Reinsurance, Swiss Reinsurance Company,
Zurich.
Sigma, (2003), Reinsurance, a Systemic Risk?, Swiss Reinsurance Company, No.5, Zurich.
Upper, C. and Worms, A. (2004), Estimating Bilateral Exposures in the German Inter-bank Market: is there a Danger of Contagion?, European Economic
Appendix A : Determination of group accordingly to size
Based on the amount of gross premium received, all companies are grouped according to size because differences in size can lead to different firm behaviour. For example, relative larger firms might take on different risks, or larger risks, compared to small companies. Small firms, might need to reinsure more of their total risk as they might not be able to take on large risks.
All amounts in Billion euro.
Boundaries were selected such that the market features related to firm size are visible. Both the life and non-life insurance sector have a few large firms, some medium size firms, and numerous small firms.
Life Non-Life
Small 0 ≤ Gross premium < 400 0 ≤ Gross premium < 100 Medium 400 ≤ Gross premium < 1500 100 ≤ Gross premium < 750
