An inquiry into the existence of underwriting cycles in the South African reinsurance market 1964-2005.

List of FiguresFigure 1 ‘The stabilising effect of reinsurance on French non-life insurers’ underwriting results’ Figure 2 ‘Stages of the short-term insurance pricing cycle’ Figure 3 ’Un

An inquiry into the existence of underwriting cycles

in the South African reinsurance market 1964-2005.Tebogo Leshilo

A research report submitted to the Faculty of Commerce, Law and Management

of the University of the Witwatersrand, Johannesburg in partial fulfilment of the degree Bachelor of Commerce (Honours)

December 2007

I hereby declare that this is my own unaided work, the substance of or any part

of which has not been submitted in the past or will be submitted in the future for

a degree in to any university and that the information contained herein has not been obtained during my employment or working under the aegis of, any other person or organization other than this university

-

-(Name of candidate) Signed

Signed this - day of - 2007 at Johannesburg

I would like to thank the people instrumental to this report:

1 My mother and family for their understanding and support Without them, I would not have made it this far

2 Mr Frank Liebenberg, my supervisor who has been a beacon of light throughout this process Thank you for the time and effort you have devoted

to this research report and most importantly for letting me realise that I always had it within me to succeed

3 Prof R.W Vivian for his continued help and support

3.2 International underwriting cycles

3.3 Factors that can influence underwriting cycles

3.4 Reinsurance underwriting cycles

Reference list

List of Figures

Figure 1 ‘The stabilising effect of reinsurance on French non-life insurers’ underwriting results’

Figure 2 ‘Stages of the short-term insurance pricing cycle’

Figure 3 ’United States Combined Ratio: 1957-2001’

Figure 3 ‘United States Combined Ratio: 1957-2001’

Figure 4 ‘Underwriting result of U.S Property-Liability Insurers 1915-2000’ Figure 5 ‘South African Underwriting Cycle for the period 1974 to 2005’

Figure 6 ‘UK Underwriting Results (1983-2001)’

Figure 7 ‘Underwriting Margins of Hong Kong direct insurers (1992-2004)’ Figure 8 ‘Gross Combined Ratios in China, 1999 to 2005’

Figure 9 ‘Combined Operating Ratio for Australia (1977 to 2004)’

Figure 10 ‘U.S Direct Insurers’ Underwriting Profit or Loss 1991-2003’

Figure 11 ‘The Comparison of Loss Ratios from Four Countries’

Figure 12 ‘U.S reinsurance cycle and its influencing factors’

Figure 13 ‘Underwriting Result - U.S Reinsurers from 1980 to 2003’

Figure 14 ‘Unsmoothed South African Reinsurers’ Loss Ratio’

Figure 15 ‘3 Year Smoothed South African Reinsurers’ Loss Ratio’

Figure 16 ‘5 Year Smoothed South African Reinsurers’ Loss Ratio’

Figure 17 ‘South African Reinsurers’ Underwriting Profit as a Percentage of Net Written Profits’

Figure 18 ‘Comparison of Gross and Net of Retrocession Underwriting Profits for S.A Reinsurers’

Figure 19 ‘Comparison of South African Direct Insurers’ and Reinsurers’ Underwriting Profits as a Percentage of Net Written Premiums’

Figure 20 ‘Comparison of U.S Underwriting Losses and S.A Profit Ratio (1980-2005)’

Figure 21 ‘Comparison of U.S Underwriting Losses and S.A Loss Ratio 2005)’

give a review of the factors most frequently cited as the causes of underwriting cycles in short-term insurance markets

The causes of underwriting cycles are based on several different theories: (1) the extrapolation hypothesis by Venezian (1985); (2) the rational-expectations/institutional-intervention hypothesis by Cummins and Outreville (1987); (3) the fluctuations-in-interest-rates hypothesis suggested by Doherty and Kang (1988) and Doherty and Garven (1992); the capacity-constraint hypothesis

by Winter (1988,1989), Cummins and Danzon (1991) and Gron (1994); and the changes-in-expectations hypothesis by Lai and Witt (1990,1992) Additional factors related to disequilibrium between supply and demand, external shocks and general business influences are also cited as causes of the underwriting cycle

Despite the significant amount of research conducted about insurance cycles, no single factor in isolation has yet been attributed as the cause of underwriting cycles

Underwriting cycles in direct insurance markets have been demonstrated in several international markets other than the United States, such as Europe, Asia, Australia and Africa In spite of this, the available literature draws heavily upon research conducted in the United States The lack of research about reinsurance underwriting cycles, limits the scope of this report and the causes or otherwise of possible South African reinsurance cycles are not examined

This report begins with an overview of the forms of insurance offered within the industry, in order to gain an understanding of the market It is then followed by

an analysis of underwriting cycles and the factors believed to cause them The next section studies reinsurance underwriting cycles, followed by an analysis of data from the South African short-term reinsurance market in order to establish

if reinsurance cycles exist This report is then concluded and recommendations made for future research

The following section provides insight into the activities and functions of the risk carriers in the insurance industry

2.1 Direct Insurance

Stettler, Eugster and Kuhn (2005:13) define insurance as “an operation by which one party, the insured, obtains from another party, the insurer, the promise to indemnify the insured or a third person in case of a loss The payment for this

service is called premium The insured accepts a totality of risks and compensates the insured or a third person according to statistical laws”

An insurance policy cannot protect an individual or corporate entity against fire,

a traffic accident, potential legal liability or the loss of a motor vehicle; rather the subject of insurance is a person, benefit or property exposed to loss or damage or some potential legal liability the insured may incur Direct insurers offer their services through independent agents or brokers, insurance consultants or banks.The insurer is a commercial entity who will take some part of the insured’s risk and bear the financial loss sustained by such events, subject to the terms and conditions of the insurance policy This provides the insured with greater security for which they agree to pay the insurer a premium (Stettler et al., 2005)

Risk management, a managerial function involving measuring risk and developing strategies to control the risk, is aimed at protecting the insurer against the financial consequences of event risk Reinsurance is a risk management tool employed by insurers to transfer some or all of an insurance risk to another insurer, namely the reinsurer There are several motivations for primary insurers to reinsure: professional reinsurers have specific skills in risk management; the direct insurer has undiversified risks; the direct insurer may realise some tax advantages by ceding premiums, and reinsurance can increase a primary insurer’s surplus thus enabling them to write new policies (Chen, Hamwi and Hudson, 2001; Stettler et al.,2005)

2.2 Reinsurance

Reinsurance is a mechanism used by the direct insurers to spread the risks and hazards assumed from policyholders; it serves as insurance for direct insurers Since reinsurance covers part of the risks assumed by direct insurers, the direct insurers’ exposure to liquidity problems or the threat of financial ruin after a catastrophe is reduced Reinsurance absorbs the direct insurance industry’s losses and distributes them among a group of companies so that no single insurer is overburdened by the financial responsibility of offering coverage to its policyholders Reinsurance moderates the effect of primary insurer losses in unprofitable years; however, reinsurers expect to receive adequate compensation

in years with favourable results The portion of risk that primary insurers transfer to reinsurers is that which exceeds their underwriting capacity and which would negatively affect the balance of their retained portfolios (Enz, 2002; Baur and Breutal-O’Donoghue, 2004; McIsaac and Babel, 1995)

A reinsurance treaty is an insurance contract by the reinsurer to the direct insurer, usually referred to as the ceding company, cedent or reinsured Based on the terms of the reinsurance treaty, the reinsurer agrees to indemnify the cedent against all or part of a loss that may be incurred under a policy initially issued by the cedent The reinsurer is only liable to the direct insurer, as no legal contractual relationship exists between the reinsurer and the policyholder The ceding company is fully liable for claim payments to its policyholders even if its reinsurance contracts cannot be enforced Although reinsurance protects the cedent against frequent or severe losses, it does not change the original policies

or any obligations created by those policies (Stettler et al., 2005; McIsaac and Babel, 1995; Chen et al., 2001)

A primary insurer is willing to forego some of its expected profits to a reinsurer

in exchange for several benefits Reinsurance enables the primary insurer to expand its underwriting capacity and flexibility in underwriting risks that would normally be too large or for types of business that would normally not be written; i.e it increases the insurer’s capital Adequate reinsurance stabilises earnings and limit fluctuations in underwriting results by significantly reducing the probability of a capital depleting loss event Access to reinsurer’s advice in the areas of product development, pricing, underwriting and claims management gives a direct insurer a competitive edge in the market Primary insurers use reinsurance as a financial risk management tool to diversify risks, optimise capital allocation, reduce capital cost, maintain solvency, manage investment risks, realise tax advantages and circumvent international insurance trade barriers (Baur and Breutel-O’Donoghue, 2004; Stettler et al, 2005; McIsaac and Babel, 1995, Meier and Outreville, 2006; Chen et al, 2001; Thomas, 1992)

Figure 1: The stabilising effect of reinsurance on French non-life insurers’ underwriting

results

Source: Baur and Breutel-O’Donoghue, 2004

One of the fundamental reasons that insurers purchase reinsurance is to protect their capital base against large deviations from expected losses, especially in the event of major catastrophes The smoothing effect of reinsurance for an individual insurer can amount to at least 50% of its premiums Figure 1 demonstrates the stabilising effect of reinsurance on the net underwriting results (after reinsurance) of French property and casualty insurers from 1990 to 2003 The effect of reinsurance was most pronounced in 1999, the year of the Winter Storm Lothar; reinsurance cover protected the French insurance industry from severe distress and possible insolvencies It can be seen from figure 1 that during the Winter Storm Lothar in 1999, the underwriting result before reinsurance was

8 percent but it decreased to 4.25 percent after the introduction of reinsurance (Baur and Breutel-O’Donoghue, 2004)

Should a reinsurer not want to keep the full share of ceded risk or have inadequate capacity to retain the risks, it is possible to reduce underwriting and investment risks by transferring the risks outside of the company using retrocession (Fitt, 1982; Baur and Breutel-O’Donoghue, 2004)

2.3 Retrocessions

The insurance ceded to a reinsurer is called the cession Should the reinsurer accept more of that risk than it desires, the reinsurer may reinsure a portion of that risk with yet another reinsurer (McIsaac and Babel, 1995) Retrocession is the transfer of ceded premiums by reinsurers to other reinsurers or insurers

A retrocessionaire is defined as any insurer or reinsurer accepting retroceded risks, while a retrocedent is a reinsurer purchasing reinsurance for his portfolio Unlike primary insurers and reinsurers, retrocession companies as such do not exist; retrocession business is written either by direct insurers or reinsurers (Stettler et al., 2005) Therefore, the retrocession of risks occurs in a chain-like fashion, diversifying exposure to risks throughout the international reinsurance industry (McIsaac and Babel, 1995) Retrocession ensures that risks are spread among a large enough group of insurers and reinsurers that a single catastrophe can be absorbed better without causing insolvencies among those carrying the risks (McIsaac and Babel, 1995)

3 Underwriting Cycles

3.1 Definition

Short-term insurance markets alternate between rising and falling prices and supply of coverage in a persistent and pervasive process that is referred to as the underwriting cycle (Klein, 2003)

An underwriting cycle has been defined by Harvey Rubin in the Barron’s Dictionary of Insurance Terms (1995) as:

“The tendency of short-term insurance premiums, insurers’ profits and the

availability and quality of coverage to rise and fall with some regularity

over time A cycle can be said to begin when insurers tighten their

underwriting standards and sharply raise premiums after a period of

severe underwriting losses Stricter standards and higher premium rates

often bring dramatic increases in profits, attracting more capital to the

industry and raising underwriting capacity On the other hand, as insurers

strive to write more premiums at higher levels of profitability (following a

hard market), premium rates may be driven down and underwriting

standards relaxed in the competition for new business Profits may erode

and then turn into losses if more lax underwriting standards generate

mounting claims The stage would then be set for the cycle to begin again”

(Fitzpatrick, 2004:256)

Cyclicality is not unique to insurance; upturns and downturns in prices and profits, as well as variations in product supply and quality are present in many industries Fluctuations in the overall business activity are referred to by economists as the ‘business cycle’ (Markham, 2006) Stewart (1984) states that cycles in the property-casualty industry do not coincide with the general business cycle, nor are they reliably contra-cyclical According to Webb (1992), not only does the underwriting cycle not synchronise with the general business cycle, it is much more regular (Chen et al., 1999) Insurance cycles actually reflect more volatility than other business cycles, in that they display higher ‘highs’ and lower ‘lows’ It is this volatility that is responsible for the regular crises in insurance markets (Fitzpatrick, 2004)

Figure 2: Stages of the short-term insurance pricing cycle

Source: Parsons (2003)

The insurance underwriting cycle is divided into four phases based on movements in price, quantity and reported profits (Gron, 1994) The first stage, known as the ‘soft market’, is illustrated by several years of low profitability until it reaches a trough Prices and profits are relatively low, quantity is abundant and underwriting standards are loosened Thereafter, there is an abrupt transition to rapidly increasing profitability This second stage, called the

‘hard market’, is characterised by substantial price increases and restricted supply The rise in insurance premiums and the reduction in availability can be

so abrupt and severe that hard markets are often referred to as ‘liability crises’; which are crises in availability, adequacy and affordability The third stage sets

in, where although profitability stays high, it is no longer increasing This stage is characterised by relatively low availability, high premiums and high profitability In the final stage, profitability gradually declines as the industry

returns to a period of low profitability The decline in profitability during the fourth stage is accompanied by falling prices and eased availability restrictions (Gron, 1994 a, b; van Fossen, 2002) The phases of the insurance pricing cycle are summarised in figure 2.

It is widely believed that the American underwriting cycle spans a period of about six to eight years from peak to peak (Venezian, 1985; Cummins and Outreville, 1987; Doherty and Garven, 1995; Chen et al., 1999) There is no regularity to the underwriting cycle; the current cycle will not end simply because its 2001 starting point plus six equals 2007 (Stewart, 1980) The length of the cycle may be shortened somewhat by ease of entry and exit in the insurance market; while reserve management may extend the cycle by causing profits to be reported when the business is no longer profitable (Boor, 1998)

Underwriting results are most commonly expressed in a measure called the combined ratio or in a measure known as the loss ratio The percentage loss ratio

is calculated as claims incurred divided by net written premiums The percentage combined ratio, which measures the growth rate in premiums and the ratio of losses and expenses to premiums, is calculated as claims incurred plus expenses and commission divided by net written premiums Therefore, the combined ratio is an extension of the loss ratio that signifies underwriting profit

or loss A ratio above 100 for an operating period signals an underwriting loss, while a ratio below 100 signifies an underwriting profit The combined ratio decreases if premium growth exceeds the growth in losses and expenses (Long, 1986; Berger et al., 1992)

Simmons and Cross (1986) assert that risk managers ought to observe the underwriting cycle when developing risk management programmes In the rising phase (combined ratio increasing) of the cycle, increasing profitability means underwriters can relax underwriting standards and reduce effective rates The risk manager may depend more heavily on insurance, a risk financing tool

In the downward phase (combined ratio decreasing) of the cycle, underwriters tighten underwriting standards and raise effective rates, whilst relying on risk retention and utilising captives The use of these non-insurance tools may impact the cycle The use of captives during the rising phase of the cycle can increase the steepness (increase combined ratio) of the cycle by reducing the demand for insurance The industry’s increased capacity together with the reduced demand leads to even more lenient underwriting standards and greater reductions in rates; resulting in even higher combined ratios As the combined ratio reaches the peak of the cycle, the industry’s low profitability leads insurers to tighten underwriting standards and raise rates This causes the combined ratio to decline until it reaches the trough again and the cycle begins again (Simmons and Cross, 1986)

3.2 International underwriting cycles

Although the characteristics of the cycle such as its length and amplitude vary between lines of insurance, geographical markets and over time, cycles are a universal feature of insurance markets (Schnieper, 2002) Cycles are not limited

to the United States; they are also observed in many countries (Cummins and Outreville, 1987; Lamm-Tennant and Weiss, 1997; Chen et al., 1999) The growth

of international reinsurance services and the deregulation of global financial

markets suggest that insurance cycles will be present in parts of the world other than the United States and developed countries (Cummins and Outreville, 1987; van Fossen, 2002) Underwriting cycles exist in South Africa (Markham, 2006), Morocco, Tunisia, Kenya, Egypt, Nigeria (Ambaram, 2002), Asia (Chen et al., 1999), Switzerland, Japan, Germany (Leng and Meier, 2002), and Australia (Australian Competition and Consumer Commission, 2003; Chidgey et al., 2005)

Figure 3: United States Combined Ratio: 1957-2001

Source: Klein (2003)

Figure 3 presents the United States underwriting cycle depicted by the combined ratio from 1957 to 2001 Six distinct cycles can be observed when measuring from peak to peak: 1957-1964; 1964-1969; 1969-1975; 1975-1984; 1984-1992; and 1992-

2001 One can observe that the individual cycles range between 5 to 9 years, however on average, the length of the cycles is equal to 7 years The United States short-term insurance industry reported its first underwriting profit in

2004, twenty five years after the last reported profit in 1978 (Pressman, 2005)

Figure 4: Underwriting result of U.S Property-Liability Insurers 1915-2000

Source: Ambaram (2002)

It can be seen in figure 4 that cyclicality in United States underwriting results dates back to the early 20th century when data first became available (Long, 1986) According to Ralph (1991), the insurance cycle appears to be increasing in amplitude; it can be seen in figure 4 that the cycle has become more pronounced over the time (Ambaram, 2002)

Figure 5: South African Underwriting Cycle for the period 1974 to 2005

Source: Markham (2006)

Figure 5 depicts the South African underwriting cycle expressed as a percentage

of earned premiums for the period 1974 to 2005 Six distinct cycles can be observed from figure x: 1978-1982; 1982-1987; 1987-1991; 1991-1995; 1995-2002 The South African cycles appear to be of a shorter length than the U.S cycles, with an average length of 4 years Ambaram (2002) provides evidence that although the South African cycle generally followed the same trend; it lagged behind that of the international cycle

Figure 6: UK Underwriting Results (1983-2001)

Source: Sears (2004)

Figure 6 depicts the underwriting results in the United Kingdom insurance market Underwriting cycles of differing lengths and frequencies are depicted for the property, accident and health, and motor lines for the period 1983 to 2001

Figure 7: Underwriting Margins of Hong Kong direct insurers (1992-2004)

Source: Chye (2005)

Figure 7 provides evidence of the underwriting cycle in Hong Kong which has been through one complete cycle of underwriting profit-loss-profit in the past decade The Hong Kong cycle has been found to be correlated with the economic cycle; economic downturns are characterised by slower premium growth, lower premium rates and increased competition among insurers in the insurance market (Chye, 2005)

Figure 8: Gross Combined Ratios in China, 1999 to 2005

Source: Tucci and Baker (2007)

Figure 8 depicts the gross combined ratios1 for China from 1999 to 2005; it appears that the overall underwriting performance of the Chinese insurance industry has been quite profitable in recent years However, Tucci and Baker (2007) state that the combined ratios shown suffer from deficiencies; if corrected, the estimated true combined ratios would be higher than shown This would translate to underwriting losses instead of profits for 2003 and 2004

Figure 9: Combined Operating Ratio for Australia (1977 to 2004)

Source: Chidgey et al (2005)

Figure 9 depicts the Australian underwriting cycle expressed as the combined operating ratio The industry experienced underwriting losses between 1997 and 2000; managing to write business profitably at the underwriting level only after

2001 (Chidgey et al., 2005)

1 A gross combined ratio is defined as the gross claim payment ratios adjusted to include an estimate for acquisition and operating expenses.

3.3 Factors that can influence underwriting cycles

During the past decade or more, a substantial body of literature has developed in explaining the cyclical nature of insurance rates and profits in the property-liability insurance market Although Meier and Outreville (2006) acknowledge that there is no generally accepted view of what the causes of the underwriting cycle may be, they consolidate the literature into three main schools of thought:

1 disequilibrium between supply and demand;

2 external shocks; and

3 general business influences

These three views are not mutually exclusive as the rational expectations hypothesis implies that prices are influenced by many factors other than the present value of expected losses Therefore, one cannot test the theories against each other, nor can a single theory explain the cause of underwriting cycles (Meier and Outreville 2006) Some theories do not explain the cause of the cycle itself, but merely the symptoms thereof; hence, some theories are less valid than others However, once the cycle starts its depth and length can be affected by the response of insurers, such as increasing supply when premium profits are rising (Rose et al 2004; Markham 2006)

3.3.1 Disequilibrium between Supply and Demand

Stewart (1980) explains disequilibrium as the imbalance between the supply of insurance and the demand for insurance Demand for insurance varies over time

as a result of fluctuations in the flow of business directed to insurers, such as

changes in the marketing or distribution strategy However, one must remember that insurers are price takers rather than price makers; should there be excess supply within the market, prices will face downward pressure towards the equilibrium price level (Markham, 2006)

3.3.2 External Shocks

a Interest Rates

Myers and Cohn (1987) and Cummins (1991) suggested that in a rational expectations framework, insurance premiums reflect the present value of expected losses and expenses Thus, higher discount rates imply lower premiums, other things being equal This prediction is consistent with Doherty and Kang (1988) who hypothesise that fluctuations in interest rates cause insurance price cycles Unexpected changes in interest rates may generate external shocks which could stimulate an underwriting cycle (Doherty and Kang, 1988; Fields and Venezian, 1989; Doherty and Garven, 1992; Fung et al., 1998, Lamm-Tennant and Weiss, 1997)

The intuition behind Doherty and Kang’s (1988) hypothesis is that higher interest rates generate greater investment income, which lowers premiums; the converse

is true As a result, premiums are inversely related to interest rates (Fung et al., 1998) Insurers suffering from diminishing investment returns will have to increase underwriting margins by increasing prices or toughening underwriting standards so as to generate earnings Thus, interest rates can significantly contribute to the ‘hardening’ phase of the underwriting cycle Conversely,

unexpectedly robust investment returns can extend a soft phase of the cycle by propping up insurers with poor or negative underwriting margins (Fitzpatrick, 2004)

Fields and Venezian (1989) found evidence of a correlation between unanticipated interest rates and the loss tail Doherty and Kang (1988) state that changes in interest rates have a greater impact on insurance prices in long tail lines than short tail lines The longer the loss tail, the greater portion of their portfolios than can be invested, thus a greater impact of interest rates on profitability Lines with more investable funds display a greater sensitivity to fluctuations in interest rates High interest rates encourage price cutting due to greater investment income (Fields and Venezian, 1989; Neale, 2004; Skurnick, 1993)

Doherty and Kang (1984) rely on capital-asset pricing theory to show that supply

is considered to be a function of interest rates and expected profits Interest rates are preceded by positive sign; hence when interest rates rise, insurers will increase supply in order to obtain funds for investment They hypothesise the demand for insurance to be a function of price – the inverse of the loss ratio – and aggregate income Equating the quantity of insurance demanded with the quantity supplied determines the equilibrium price This approach contributed towards developing a viable explanation of insurance prices and underwriting cycles and serves as the founding belief of the rational expectations model hypothesised by Cummins and Outreville (1987)

The underwriting profit represents a measure of the average price of the contracts traded Insurance pricing models based on financial theory

unanimously show that competitively determined insurance premiums are inversely related to interest rates and will therefore change as interest rates change Using this model, together with the rational expectations and lag features of the Cummins and Outreville model, Doherty and Kang (1988) show that the intertemporal behaviour of underwriting returns in insurance markets is quite well explained as a market clearing process in which equilibrium prices change in lagged response to changing interest rates In other words, the underwriting cycle may be viewed as a series of converging responses to changing spot prices (Doherty and Kang, 1988; Doherty and Garven, 1995)

It has yet to be proven that interest rates themselves exhibit cyclical behaviour; however the unexpected change in interest rates may create external shocks which could stimulate an underwriting cycle through the adjustment lags in the insurance market (Doherty and Kang, 1988; Meier and Outreville, 2006) Doherty and Garven (1995) determined that changes in interest rates affect insurance prices through the discounting process as well as by changing the insurer’s capital structure The effect of interest rates on capital structure is evident in the decrease in the market value of equity when interest rates rise They found evidence supporting the hypothesis that insurance prices respond with greater sensitivity to falling interest rates than to rising interest rates The results imply that changes in interest rates, in conjunction with changes in equity values, could initiate the shift from a hard market to a soft market (Klein, 2003)

b Lags

Institutional lags attributed to data collection, accounting rules and policy renewal periods are believed to cause underwriting cycles Insurance prices are

based on annual data which is not available for use until several months after the close of the experience period Delays arise in the tabulation and analysis of this data as well as the slow emergence of information on losses in long-lines Policy renewal lags arise because premiums cannot be adjusted simultaneously to reflect information as soon as it becomes available Most property-liability insurance policies have a set premium for the entire policy period Additionally, when new rates are approved in a rate regulated market, a lag occurs in changing to the new rate level (Lamm-Tennant and Weiss, 1997; Meier and Outreville, 2006) These time lags can stimulate an underwriting cycle despite rational business behaviour on the part of insurers (Rose et al., 2004).

Outreville (1990) provides empirical evidence that price approval regulation in the United States exacerbates any cyclical behaviour in underwriting results by delaying the rapid adjustment of prices to new information (Lamm-Tennant and Weiss, 1997) Rate regulation does not exist in South Africa, rendering this explanation invalid for the South African insurance market (South African Insurance Association, 2007)

c Catastrophic Losses

Empirical evidence suggests that prices may increase by more than the discounted value of expected costs if large underwriting losses are unexpectedly reported Unexpected claims payments on existing policies are a shock to capital and could deplete the insurer’s capital and reserves Should it not be possible to raise sufficient capital at a relatively low cost, insurers have the option of increasing the probability of insolvency or reducing the amount of coverage supplied at a given price This situation is temporary and adjustments create

shocks that could induce an underwriting cycle (Harrington and Niehaus, 2000; Meier and Outreville, 2006)

Figure 10: U.S Direct Insurers’ Underwriting Profit or Loss 1991-2003 2

Source: Hartwig (2003)

While catastrophic losses may not be the cause of underwriting cycles, they often serve to set off or at least hasten the arrival of a hard phase of an underwriting cycle (Winter, 2003; Fitzpatrick, 2004) The September 11, 2001 attacks in the United States are the prime example of this phenomenon Industry observers acknowledge that while the hard market cycle of 2001-2002 was already in progress prior to the attack, the catastrophic losses suffered on that day served to accelerate and heighten the peak of that portion of the cycle (Fitzpatrick, 2004) The U.S property-casualty insurance market appears to be softening as premium increases have decelerated (Grace and Klein, 2003) The effect of the 9/11 attacks

on the U.S underwriting profits and losses can be seen in figure 10

In the United Kingdom, the hardening of the insurance market was well under way in 2001 before the catastrophic loss shocks of 9/11 attacks (see figure 6) However, this event is considered to have contributed to the existing trend of

2 Data originally expressed as the combined ratio, but converted to depict underwriting profit or loss (100-combined ratio)

rising rates The claims arising from the attack served to increase the cost and decrease the availability of reinsurance This in turn has impacted on the cost of direct insurance (Parsons, 2003)

From figure 7 depicting the Hong Kong underwriting results, it can be seen that the years following 1999 were characterised by the continuous hardening of direct insurance and reinsurance premiums and improved underwriting results until 2003 Chye (2005) attributes the increased rates and tightened terms to the 9/11 attack

Figure 11: The Comparison of Loss Ratios from Four Countries

Source: Leng and Meier (2002)

Figure 11 provides evidence that the September 11th attack caused high correlation of loss ratio series (i.e incurred claims divided by net written premium times 100) among different countries, namely the U.S, Switzerland and Germany As a result of the attack, the U.S insurers suffered catastrophic losses,

a large portion of which were covered by reinsurance The close ties between these countries’ insurance markets together with the fact that most reinsurers are

European companies, imply a high correlation of the loss ratio series between European countries and the U.S (Leng and Meier, 2002).

From the figures presented above, it can be seen that underwriting cycles are an international phenomenon with features particular to each country

3.3.3 General Business Influences

a Competition driven prices

According to many authors, the underwriting cycle is caused largely by price competition due to the standardisation of the property-liability business Price competition can also be regarded as irrational behaviour on the part of insurers, due to their aim of maintaining or gaining market share Following information

on the anticipated behaviour of competitors, insurers deviate from the theoretical model in their pricing practices As competition intensifies, insurers decide at some point to reduce prices or engage in excessive risk-taking in order to gain market share; the result of which is negative results and subsequent reductions

in supply, which drive prices up (Berger, 1988; Harrington and Danzon, 1994)

Harrington and Danzon (1994) proposed a moral hazard hypothesis that predicts that insurers with weak safety incentives will charge inadequate prices but grow more rapidly than firms with higher target safety levels Moral hazard among insurers arises from limited liability, risk-insensitive guarantee programmes and uniformed or unconcerned customers As a response to the underpricing by some insurers, other insurers may reduce prices to preserve market share The

result of this is a subset of insurers causing prices for other firms to fall below costs, leading to subsequent reductions in supply as a means to restore adequate prices and improve performance Harrington and Danzon (1994) conclude that other insurers’ price responses to low prices due to moral hazard cause variations in insurance prices

However, Cummins and Outreville (1987) were sceptical about the assumption that at some point, insurers will ‘decide’ to raise prices and reduce supply Cummins and Outreville (1987) were critical of the competition-driven prices view as it did not specify the causal mechanism through which market reversals take place; they called for more concrete justification for this view (Chen et al., 1999)

b The Winner’s Curse

The fundamental factor that distinguishes the insurance market from other markets is its inability to calculate its ‘cost of goods sold’ at the time that they price their products This issue has far-reaching implications as it guarantees that pricing volatility and periodic constrictions of supply will be inevitable, as insurers react to unexpected changes in the underlying liability environment that affect policies written in earlier periods or incorrect guesses in their pricing in a stable liability environment (Fitzpatrick, 2004)

Pricing uncertainty and hence variability of supply, are built into the very nature

of insurance A favourable outcome of the pricing problem is when insurers

repeatedly rob the Peter of their present risk pool to pay the Paul of some prior year’s pool whose premiums were not sufficient to fund his liabilities This process also works in reverse; in the event that the insurer’s actual experience is more favourable than anticipated, they are able to release excess reserves to offset poor results experienced in later periods According to Harrington and Danzon (1994), “the risk spreading function of insurance has a temporal aspect

as well as its more familiar ‘horizontal’ one” (Fitzpatrick, 2004)

Harrington and Danzon (1994) hypothesised that insurers form estimates or expectations of future losses based on public and private information These expectations are considered rational in the sense that they are correct on average However, due to heterogeneous information, in any given period some insurers with optimistic private information on future claim costs may price too low relative to full-information conditional expectations In a competitive market, these insurers will gain market share at the expense of insurers whose loss estimates are too high As a result of the low loss estimates, these insurers grow rapidly but fail to earn a fair return on equity and subsequently incur excessive underwriting losses Inadequate prices as a result of heterogeneous information would be especially likely for inexperienced firms This problem is called the

‘winner’s curse’ (Klein, 2003; Harrington and Danzon, 1994)

The winner’s curse is an economic theory that postulates that the winning participants in an auction will typically pay too much for the auctioned item – or

in the insurance context, charge too little to win a customer This is because the nature of an auction is to favour the bidder with the most optimistic view of the value of the underlying asset, namely future loss estimates of the insured The

insurance market- especially in ‘long tail’ lines of business- is particularly fertile ground for occurrences of the winner’s curse (Fitzpatrick, 2004)

The winner’s curse is exacerbated by adverse selection and information asymmetries which are displayed in ‘bidding’ models An insurer may gain a high market share by attracting customers through offering low premiums This may have been achieved simply because the insurer was optimistic about the prospect of losses Recognising this, the insurer will cut back by the amount of insurance offered and attempt to reduce winner’s curse effects (Winter, 1991) Wilson (1977) and Milgrom (1979) postulate that the winner’s curse can be avoided if bidders adjust their bids given knowledge of the bidding processes used by other firms and the joint density of public/private information and expected loss costs Should all firms bid optimally, under certain assumptions the winning bid will converge to the true value of the object being bid for as the number of bidders increases Established firms have learned to adjust their forecasts to avoid the winner’s curse However, if inexperienced firms, such as new entrants, make inadequate adjustments, they will price too low and incur excessive underwriting losses (Harrington and Danzon, 1994)

Harrington and Danzon (1994) suggested that an additional contributing factor

to excessive price cutting was some insurers engaging in disproportionate risk taking such as ‘go for broke’ behaviour Insurers more likely to engage in excessive risk taking are those with low levels of intangible capital Established insurers may decide to cut prices below levels dictated by optimal bidding strategies in order to retain their market share and compete with those insurers who set their prices too low due to lack of experience or intentional risk taking This may exacerbate pricing errors and contribute to price/availability crises

c General Business Cycle

Grace and Hotchkiss (1995) examined the underlying relationship between the underwriting cycle and the general condition of the economy by testing for a long-run relationship between real gross domestic product (GDP), inflation and the short-term interest rate on the underwriting cycle as measured by the combined ratio Their findings reveal that while fluctuations in the property-liability underwriting cycle are generally linked to the long-run performance of the economy, the fluctuations in the cycle are not linked to the effect of short-run shocks on economic variables (Chen et al., 1999) Lamm-Tennant and Weiss (1997) also find evidence in support of the relationship between the underwriting cycle and changes in real GDP Webb (1992) suggests that the underwriting cycle

is not necessarily synchronised with the general business cycle; findings reveal that the underwriting cycle is a lot more regular than the general business cycle (Chen et al., 1999)

Leng and Meier (2002) provide evidence that despite the fact that underwriting cycles are an international phenomenon, they do not stem from the same international effects They find it more likely that the factors affecting underwriting cycles are country-specific, such as the economic environment and regulations

d Business Practices

Improved risk management practices can be expected to dampen the effects of shocks and the shifts in demand and supply that may cause variability in

underwriting results Evolutions of the regulatory environment at national and international level also have an impact on business, capital requirements and pricing policies (Cummins et al., 1991; Winter, 1991) This is evident in the South African market pre- and post 1974 when the Tariff system was abrogated and the insurance market changed from a voluntary self-regulated market to a non-regulated environment (South African Insurance Association, 2007).

Cross and Simmons (1986) suggest that risk managers ought to heed the insurance cycle when planning their risk management programmes Insurance as

a risk financing tool should be relied upon more heavily during the upward phase of the insurance cycle During the downward phase, risk managers should implement risk control tools such as avoidance and loss prevention and control

In addition, the risk manager may also use non-insurance risk financing transfers such as risk retention and captives; such a strategy may impact the severity of the fluctuations associated with the underwriting cycle (Cross and Simmons, 1986)

e Ratemaking Methods

Venezian (1985) proposed the loss extrapolation hypothesis based on the observation that insurers and rating bureaus set premiums to a certain extent by using regression results derived from past losses He suggested that premiums set by this method would create “a quasi-cyclical pattern” of underwriting profit margins That is, past losses may explain current premiums seeing that premiums are partially set by using regression results based on past losses (Fung, Lai et al., 1998)

Cummins and Outreville (1987) proposed the rational expectations with institutional lags hypothesis, which suggests that underwriting cycles may be due to the presence of data collection lags, regulatory lags and policy renewal lags in a rational expectations framework They further suggest that in such a framework, premiums are set based on future expected losses and expenses Niehaus and Terry (1993) argue that any measure of expected losses may be subject to measurement error which may be correlated with past losses Therefore, it can be stated that past losses may predict premiums in a rational expectations framework with various lags and institutional factors due to measurement errors (Chung, Fung et al., 1993)

Venezian (1985) proposes that underwriting cycles may be caused by imperfect regulatory and accounting systems and that these imperfections allow errors to creep into the firm’s decision making process (Grace and Hotchkiss, 1995) Skurnick (1993) adds that many businesses change to more favourable accounting treatments during a ‘downturn’ in the cycle; the practice of under-reserving during bad years and strengthening reserves in good years serves to moderate the cycles and possibly extend them

Grøn (1992) implies that premiums are not the best predictors of actual losses as past surplus affects premiums This is in corroboration with Venezian’s (1985) model which shows that premiums are mechanically set based on the realisation

of past losses Accordingly past loss experience affects premiums (Niehaus and Terry, 1993) Venezian (1985) is of the opinion that the underwriting cycle is caused by nạve forecasting procedures; a belief held by Outreville (Cummins and Outreville, 1987)

Additionally, Venezian’s hypothesis implies a degree of irrationality on the part

of insurers; by extrapolating past loss trends into the future in a rather mechanical way, other potentially relevant information is de-emphasised or disregarded (Cummins and Outreville, 1987) Extrapolative forecasting is said to destabilise insurance markets, which results in rates considerably higher or lower than competitive levels due to errors in estimates of losses or investment income (Lamm-Tennant and Weiss, 1997)

Venezian (1985) hypothesises that underwriting results follow a second-order autoregressive process due to the ratemaking procedure used by insurers (Leng and Meier, 2002) Smith (1984) found results consistent with Venezian’s hypothesis (Cummins and Outreville, 1987) This process is thought to cause the cycles in underwriting results

f Insurers’ Expectations

Rational-expectations/institutional-intervention

This theory hypothesises that current premiums are informationally efficient predictors of future losses Several financial models applied to insurance pricing suggest that premiums reflect the present value of expected losses and expenses

in a rational expectations framework that ignores or assumes away institutional intervention (Chung, Fung et al., 1993)

While Venezian’s (1985) hypothesis implies a degree of irrationality on the part

of insurers, an alternative explanation for underwriting cycles may be found in the rational-expectations model when institutional lags and reporting practices

are taken into account This model implies that economic agents forecast economic variables without systematic error; that is; the expected values are the same as the actual values conditional on all available information at the time that the forecasts are made This condition implies that price determination is a rational process In constructing this theory, the authors hypothesised that both demand and supply play a role in the determination of insurance premiums and assumed that insurance markets are competitive and rational (Cummins and Outreville, 1987).

The rational-expectations hypothesis implies that the insurance market’s evaluation of relevant economic data is rational, even though insurance profits seem to follow “irrational” cycles This irrational behaviour is due to intervening factors filtrating rational prices Institutional and regulatory lags prevent insurance prices from adjusting promptly to changing economic conditions, while reporting practices tend to average together prices from different periods and exaggerate any autocorrelation present in the price relationships The model hypothesises that in the absence of intervening factors, the rational-expectations hypothesis would be inconsistent with the existence of underwriting cycles Therefore, institutional and regulatory lags, combined with accounting practices, are responsible for cycles in underwriting profits (Cummins and Outreville, 1987; Lamm-Tennant and Weiss, 1997)

Questions have arisen as to why these practices are not discarded, because they are partially responsible for cyclicality in underwriting profits The answer is that adverse selection and underwriting costs may explain why insurance prices do not change more rapidly (Cummins and Outreville, 1987)

Additionally, Cummins and Outreville (1987) state that historical information is used to develop a rational-expectations framework with institutional lags Underwriting premiums reflect the presence of lags; subsequently premiums are based upon expected future losses and expenses Niehaus and Terry (1993) suggest that measures of expected losses might be subject to measurement error, which may be correlated with past losses Hence, historical data is used to develop a rational-expectations hypothesis with institutional lags (Fung, Lai et al., 1998)

A test of the rational expectations/institutional intervention hypothesis provides empirical evidence suggesting that institutional lags and reporting practices cause not only domestic (U.S.) but international underwriting cycles (Lamm-Tennant and Weiss, 1997) Mamatzakis and Staikouras (2006) present supporting evidence with their findings that claims are used to set premiums within the underwriting cycle observed in the UK insurance market, as derived from the rational expectations and institutional rigidities hypotheses

Change in expectations

Lai and Witt (1990) constructed a pricing model under uncertainty where premiums are functions of variance in the underwriting expenses in addition to the expected values of the underwriting expenses The model suggests a positive relationship between underwriting expenses and premiums; when insurers’ expectations about future underwriting expenses are high then premiums will tend to be high, other things being equal (Fung, Lai et al., 1998)