Sunteți pe pagina 1din 16

The Journal of Risk Finance

Analysis of multinational underwriting cycles in property-liability insurance


Chao-Chun Leng Ursina B. Meier
Article information:
To cite this document:
Chao-Chun Leng Ursina B. Meier, (2006),"Analysis of multinational underwriting cycles in property-liability
insurance", The Journal of Risk Finance, Vol. 7 Iss 2 pp. 146 - 159
Permanent link to this document:
http://dx.doi.org/10.1108/15265940610648599
Downloaded on: 18 March 2017, At: 00:56 (PT)
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

References: this document contains references to 24 other documents.


To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 1111 times since 2006*
Users who downloaded this article also downloaded:
(2006),"Business cycles in insurance and reinsurance: the case of France, Germany and Switzerland", The
Journal of Risk Finance, Vol. 7 Iss 2 pp. 160-176 http://dx.doi.org/10.1108/15265940610648607
(2006),"Multi-national underwriting cycles in property-liability insurance: Part I some theory and empirical
results", The Journal of Risk Finance, Vol. 7 Iss 1 pp. 64-82 http://dx.doi.org/10.1108/15265940610637816

Access to this document was granted through an Emerald subscription provided by emerald-srm:273599 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.

*Related content and download information correct at time of download.


The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1526-5943.htm

JRF FOCUS ON INSURANCE CYCLES: PART II


7,2
Analysis of multinational
underwriting cycles in
146
property-liability insurance
Chao-Chun Leng
Towers Perrin, Philadelphia, Pennsylvania, USA, and
Ursina B. Meier
UBS AG, and University of Zurich, Zurich, Switzerland
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

Abstract
Purpose The paper sets out to use the loss ratio series of Switzerland, Germany, the USA and
Japan, to test whether underwriting cycles still exist internationally and to identify possible structural
changes.
Design/methodology/approach Based on financial theory and insurance pricing theory,
co-integration analysis was performed to check possible causes of structural changes.
Findings All four countries have breaks in different years. This result leads to the hypothesis that
the factors affecting underwriting cycles are mainly country-specific, such as economic environment
and regulations, rather than global/international. Although the financial theory and the insurance
pricing theory suggest that the loss ratio series should be co-integrated with the interest rate series
with co-integrating coefficient 21, the empirical results do not support the theories.
Originality/value More detailed analysis for the time series characteristics for countries other
than the USA is presented to investigate the possible existence of underwriting cycles.
Keywords Underwriting, Property, Insurance, Switzerland, Germany, United States of America, Japan
Paper type Research paper

Introduction
Underwriting cycles have been studied by regulators, consumer groups, academics,
and insurers in order to set appropriate premiums and mitigate fluctuations in
underwriting results. Similar to business cycles, underwriting cycles consist of
alternating phases of hard and soft markets. In a hard market, premiums for insurance
rise, insurers limit their policy renewals, new policies are difficult to get, and available
policies have higher deductibles and lower policy limits. Thus the affordability and
availability of insurance become an issue for consumers. Consumers may benefit from
the soft market, but insurer insolvencies create social externalities.
Several theories are used to explain underwriting cycles. Venezian (1985) shows that
underwriting results followed an AR(2) process due to the ratemaking procedures
adopted by insurers. Cummins and Outreville (1987) show that data reporting,
accounting, and regulatory delay cause the second order of auto-regression even under
The Journal of Risk Finance the assumption that the insurers have rational expectations. They conclude from
Vol. 7 No. 2, 2006
pp. 146-159 empirical estimations that underwriting cycles exist internationally with cycle lengths of
q Emerald Group Publishing Limited
1526-5943
six to eight years. Two later studies, Lamm-Tennant and Weiss (1997), and Chen et al.
DOI 10.1108/15265940610648599 (1999), followed Cummings and Outrevilles method and confirmed the results. Doherty
and Kang (1988) use an equilibrium model between demand and supply to explain Multinational
underwriting cycles. Gron (1994) and Winter (1988) apply the theory of capacity underwriting
constraints to insurance markets. Niehaus and Terry (1993) were the first to apply vector
auto-regression (VAR) for underwriting cycles. Fung et al. (1998) test the existing cycles
theories for underwriting cycles by VAR and found that no single theory can fully
explain underwriting cycles. Leng (2000) uses a longer data period and more recent data
and finds that combined ratio of property and liability insurance for the USA has a break 147
in 1981. Before the break, the combined ratio was stationary and followed an AR(2)
process. After 1981, it is non-stationary but behaves as financial theory predicts for a
competitive market. Studying more recent and longer time series than Cummins and
Outreville (1987), Meier (2001) shows that underwriting cycles for four countries
(Switzerland, the USA, Germany, Japan) have either longer cycle lengths than those from
Cummins and Outreville (1987) or for some specifications no more cycles at all.
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

Underwriting results from four countries, Switzerland, Germany, the USA and
Japan, are studied in this article. Cummings and Outreville brought underwriting
cycles to an international level. However, some questions need to be answered, such as
whether underwriting results are still predictable, or whether the pattern of the
underwriting cycles or the coefficients in the AR(2) process or the cycle lengths have
changed over time. Therefore, tests are imposed to check whether international data
series have breaks, and if there are breaks, what the cause is, and how the fluctuations
of underwriting profit behave before and after the breaks.
The remainder of the article is organized as follows. In the next section, some
background information about the four countries property-liability insurance industry
is given, followed by the data section with data sources, characteristics of the loss ratio
series for each country, and comparisons of the loss ratio series among the four
countries. Then follows a section on hypotheses on the loss ratio series and their test
procedures. We proceed with a section on the empirical test results for our hypotheses,
followed by a section with a discussion of the results and possible explanations. In the
further analysis section, results from extended tests for issues addressed in the
discussion section are reported. The article closes with the conclusions.

Insurance environment
Switzerland
Most insurance companies in Switzerland are stock companies. Insurance is usually
distributed to customers directly by insurance companies. Independent agents and
brokers also exist. As contracts are often signed for several years, customers in
Switzerland do not often switch to alternative companies. In the late 1990s, some
collaboration between insurance companies and banks emerged. Until the early to
mid-1990s, the Swiss insurance market was quite intensively regulated. The profit
margin was limited by the Federal Office of Insurance Supervision, usually to around 3
percent, and there were strict rules on solvency requirements, but the insurers were
still able to raise and lower reserves to some extent to smooth profits. Until 1993, the
European Community (EC) and Switzerland reached an agreement to apply the 3rd
EC directive liberalizing the insurance industry between Switzerland and the EC. The
effect was to remove tariffs and impose more rigorous solvency rules. The Swiss
insurance market is highly concentrated and very saturated; for many years
Switzerland has had the highest insurance density worldwide. After several recent
JRF mergers, the market has become even more concentrated as well as more international.
7,2 Since the Swiss insurance market is comparatively small, Swiss insurance companies
must seek growth abroad.

Germany
The German insurance market has long been organized along similar lines to the Swiss
148 one. However, as a member state of the European Union (EU), Germany has to follow EU
regulations and the German authorities have correspondingly less freedom to impose
country-specific rules. Until the early 1990s, the German insurance market was strongly
nationally oriented, but it has become less intensively regulated with the liberalization of
the insurance industry within the EU. This has paved the way for German insurers to
move into other EU states and for foreign companies to enter the German market.
Most German insurance companies are stock companies, and the distribution
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

system is the direct writing system. However, independent agents also exist. Owing to
the regulated profit margin, it is difficult for new companies to enter the market. Also,
the market is becoming more concentrated owing to mergers and acquisitions over
time.

USA
There are four kinds of ownership structures for insurance companies in the USA.
They are stocks, mutuals, reciprocals, and Lloyds. The two most common ownership
forms are stocks and mutuals. Stock insurers business can be seen in a wide range of
lines and with low geographical concentration. On the other hand, mutual insurers
write more standard business, which requires less managerial discretion.
There are three distribution channels: brokerage, independent/American agency
system, and direct writing system. Among these systems, the independent agency and
the direct writing system are dominant in the market. Historically, the independent
agency system and brokers were the dominant sales systems. After the Second World
War, direct writers started gaining market shares by selling policies with lower rates,
especially in personal lines of business.
The US insurance market has long been regarded as competitive with a large
number of companies and low concentration. Since the capital requirement is moderate,
entry barriers traditionally were considered to be not very high. However, the
distribution system may act as a barrier to entry. The entry barrier is low when a new
entrant chooses to distribute through the American Agency System. If a new insurer
wants to get into the insurance market through the direct writing system, it has to
invest heavily in advertising to make its presence known to consumers, which
constitutes a de facto barrier to entry.

Japan
There are four kinds of insurance companies in Japan. They are the horizontal (financial)
keiretsu system, the vertical keiretsu system, independent companies, and foreign
companies. Financial keiretsu and vertical keiretsu systems dominate this industry. A
financial keiretsu has a commercial bank, a trust bank, a life insurance company, and a
non-life insurance company. A vertical keiretsu is usually related to a large industrial
company, e.g. Toyota or Hitachi. The relationship between contractor and sub-contractors
is long-term. Usually, a keiretsu insurer insures its own keiretsu members.
The Japanese distribution systems for non-life insurance are direct sale and the Multinational
independent agency system. The brokerage system had just started when the New underwriting
Insurance Business Law was enacted in 1996. In Japan, selling insurance is
labor-intensive because insurers hire large numbers of part-time sales people to sell cycles
their products. The independent agents represent almost all insurers but, unlike the
independent agents in the USA, the agents do not own their client lists. In addition,
commissions are higher in Japan than in the USA. By combining high commissions with 149
labor-intensive selling techniques, Japanese non-life insurers operate with high
underwriting expenses. Offsetting their high expense ratio, as we point out later,
Japanese insurers operate with the lowest loss ratio among the four countries in our study.
The total number of non-life insurance companies in Japan was only 54 in 1995. The
top four companies, which are all financial keiretsu, account for about half the market.
Therefore, the Japanese non-life insurance market can be classified as an oligopoly.
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

Due to regulations, all insurers follow a price schedule and sell standard policies, which
are highly controlled by the government. There was no price competition until the New
Insurance Business Law was introduced in 1996. This cartel price system provides
insurers with a very stable underwriting operating environment[1]. Due to a lack of
product differentiation and price competition, Japanese insurers compete by recruiting
and retaining more sales people and by offering better service, which increases
underwriting expenses.

Data
For insurance data we use direct premiums written, incurred losses, and internal
capital[2] from the Federal Agency of Private Insurance for Switzerland and from
Swiss Re for the other countries. The data period for premiums written and incurred
losses for Switzerland and the USA is from 1955 to 1997, for Germany from 1955 to
1991[3], and for Japan from 1968 to 1997.
For internal capital, the data period is 1974 to 1997 for Switzerland, 1975 to 1987 for
Germany, 1967 to 1997 for the USA, and 1974 to 1989 for Japan. As macroeconomic
data, we use GDP, CPI, and interest rate from the International Financial Statistics
Yearbook, OECD, and from the Bank of International Settlement. GDP is in local
currency and not inflation adjusted. The base year of CPI is 1995. The interest rate is the
Government Bond Yield, which is a long-term interest rate for either ten or 20 years.
Figure 1 shows the loss ratios[4] of the four countries. We can see the cyclical
patterns and the level of the loss ratio series from this figure: Switzerland and Germany
exhibit very similar cyclical patterns but the loss ratio of Germany is about 8 to 10
percentage points higher than that of Switzerland. The patterns of the loss ratio series
of the USA and Japan are very different from each other, as well as from those of the
two European countries. The USA has the highest loss ratio and Japan has the lowest
one. However, this does not imply that insurers in the USA have the lowest operating
profit since loss ratios do not include data for underwriting expenses and investment
income, which are very different among the countries due to the regulatory and
economic environments[5]. Table I shows the correlation coefficients of the loss ratios
among the four countries. Germany (D) and Switzerland (CH) are highly correlated and
both of them are correlated with the USA. Japan (JP), however, is negatively correlated
with the other three countries. This result confirms that international operations have a
diversification effect to lower the fluctuations of the underwriting results.
JRF
7,2

150

Figure 1.
Comparison of the loss
ratios from the four
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

countries

CHLR DLR JPLR USLR

Table I. CHLR 1 0.8905 2 0.1934 0.6391


Correlation coefficients of DLR 1 2 0.3150 0.5646
loss ratios among four JPLR 1 2 0.4059
countries USLR 1

Two situations may cause high correlation of loss ratio series among different
countries. The first situation is that the insurance markets in these countries are closely
tied. The second situation is that their economies are closely tied. The first situation
can be seen after catastrophic events in which the USA insurers suffered huge losses,
but a large portion of the losses was covered by reinsurance. Most re-insurers are
European companies. Therefore, we expect to see that the loss ratio series between
European countries and the USA are highly correlated[6]. An example of the second
situation is the close economic relationship between Switzerland and Germany. When
one decides to raise the interest rates, for example, the other is likely to follow. This
dynamic movement between the two countries can be seen from the correlation of
macroeconomic variables, such as interest rate, GDP, and CPI.
Table II shows the correlation coefficients of interest rates among the four countries.
Germany and Switzerland again have the highest correlation coefficient, and
Switzerland and the USA have the lowest one. This shows that the economies of
European countries have closer ties with each other than with either the USA or Japan.
Further analysis is needed to find out whether insurance market ties or economic ties
underlie highly correlated underwriting results between two countries.

CHI DI JPI USI

Table II. CHI 1 0.7877 0.5555 0.1174


Correlation coefficients of DI 1 0.7485 0.4855
interest rates among four JPI 1 0.4828
countries USI 1
Hypotheses Multinational
There are four sequential hypotheses. If the test results support a hypothesis, we reach
conclusions and stop. Otherwise, we go on to the next hypothesis.
underwriting
cycles
H1. The loss ratio follows an AR(2) process.
Venezian (1985) and Cummins and Outreville (1987) show that underwriting losses/profits
follow a second order autoregressive model[7]. We test whether this is still true by using 151
more recent data. If this hypothesis is true, that is, if the AR(2) process for the loss ratio
has significant coefficients and a high R 2, then we find cycle lengths and compare them
with previous studies. If this hypothesis is not true, we go on to the next hypothesis.
H2. The loss ratio series are stationary.
The autocorrelation function (ACF), partial autocorrelation function (PACF), and
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

Augmented Dickey-Fuller (ADF) test are used to check whether the loss ratio series
have unit roots or not. If the hypothesis of unit roots is rejected, we should use vector
auto-regression (VAR) and impulse-response functions to determine the relationship
between the loss ratio and the macroeconomic variables. If the loss ratio series are not
stationary, we should check whether these series have breaks because a break in a
series may cause incorrect acceptance of unit roots.
H3. The loss ratio series do not have breaks.
The Chow test and switching regression are used to test for breaks. If this hypothesis is
true, we take the first difference of loss ratio series to run an AR(2) process[8]. If the loss
ratio has a break, we look for the year of the break for each country.
H4. Loss ratio and interest rate are cointegrated before and after the break.
From the Capital Asset Pricing Model (CAPM, see Fairley, 1979; Hill and Modigliani,
1981), an insurance policy is treated as if an insurer borrows a lump sum from its
policyholder and returns a certain amount of payment if an insured event occurs
during the insured time period. In other words, an insurance policy is a debt-like
contract. Underwriting return (r u ) should be:

r u 2r f bu r m 2 r f ; 1

where r f is the risk free rate, and r m 2 r f is the market risk premium.
The assumptions for Equation (1) are that the tax rate is zero and insurers invest the
entire premiums earned for a year[9]. The relationship between underwriting return
and loss ratio is:
1 
ru P 2 L 2 E 1 2 LR 2 ER ) LR r f 1 2 bu r m 2 r f 2 ER;
P
where P is premium,
 L is losses, E is expenses, LR is loss ratio, and ER is expense ratio.
If bu r m 2 r f and ER are constant, then Equation (1) is:
LR r f c ) LR 2 r f 2 c 0; 2
where c is a constant. Equation (2) shows that the loss ratio and the risk free interest
rate are co-integrated and positive correlated with co-integrating coefficient 2 1.
JRF From insurance pricing theory (Doherty and Kang, 1988; Haley, 1993), the price of
7,2 insurance should reflect the investment income by discounting expected losses because
insurers invest premiums from the time when the premiums are received to the time
when the losses are paid. Taking i as the rate of return on investment, we get the
following:

152 E Lt E Lt
Pt ) LR 1 i ) LR 2 i 2 1 0:
1 i Pt
The result from insurance pricing theory is consistent with the one from the CAPM.
Therefore, from a theoretical point of view, the loss ratio series should be co-integrated
with the interest rate. If a break in the loss ratio series is caused by a break in the
interest rate series, two series should be co-integrated regardless of the timing of the
break. Otherwise, financial theory and insurance pricing theory cannot fully explain
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

the fluctuations of the underwriting results.

Results
AR(2) process with and without a time trend
Tables III and IV show the results of AR(2) processes for the loss ratio series with and
without a time trend. The coefficients of the second lag for the AR(2) process with or
without time trend for all countries are not significant at the 5 percent level.

Autocorrelation function (ACF) and partial autocorrelation function (PACF)


Figure 2 shows the ACF and PACF for the loss ratio of Switzerland[10]. The figure
suggests that this series is not stationary since the ACF decays slowly and the PACF is
significant at one lag only.

C LR(2 1) LR(2 2)
Coef. t Coef. t Coef. t Adj. R 2 F

Switzerland 7.070* 1.753 0.882** 5.437 2 0.005 2 0.032 0.786 74.499


Germany 22.107** 2.749 0.758** 4.044 2 0.072 2 0.421 0.544 18.864
USA 10.518* 1.955 0.978** 5.915 2 0.118 2 0.746 0.783 72.957
Table III. Japan 9.647* 2.145 0.981** 5.378 2 0.222 2 1.220 0.683 31.120
Loss ratio following an
AR(2) process without Notes: * is for 5 percent significance level; ** is for 1 percent significance level
time trend LR(21), LR(22): coefficients of the first and second lag in the respective AR(2)-processes

C LR(2 1) LR(2 2) T
Coef. t Coef. t Coef. t Coef. t Adj. R 2

Switzerland 22.553* 3.186 0.746* 4.660 2 0.218 21.299 0.169** 2.582 0.814
Germany 33.511* 3.005 0.695* 3.676 2 0.202 21.064 0.119 1.447 0.561
USA 27.661* 2.789 0.860* 5.076 2 0.291 21.669 0.197** 2.026 0.799
Table IV. Japan 17.622* 2.388 0.911* 4.872 2 0.295 21.576 20.144 2 1.351 0.692
Loss ratio following an
AR(2) process with time Notes: * is for 1 percent significance level; ** is for 5 percent significance level
trend LR(21), LR(22): coefficients of the first and second lag in the respective AR(2)-processes
Multinational
underwriting
cycles

153
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

Figure 2.
ACF and PACF of the loss
ratio for Switzerland

Augmented Dickey-Fuller (ADF) test


We check whether the loss ratio series have unit roots by applying the ADF test[11].
The results in Table V lead us to accept the null hypothesis that all those series have a
unit root. The results of the ADF test are confirmed by those from the ACF and PACF.

Chow test and switching regression


The methods we use to check for structural changes are the Chow test and switching
regression[12]. The results are reported as the F-statistic for the Chow Test and the Log
Likelihood Ratio (LLR) for the switching regression in Table VI. Figure 3 is the plot of
the LLR for Switzerland. It shows that the structural change happened gradually and
reached the highest LLR in 1975.
By taking the most significant values in both methods, we found that Germany also
has a break in 1975, the USA in 1986, and Japan in 1985. In Figure 1, the loss ratio of

Switzerland Germany USA Japan

LR(21) 2 0.1231 2 0.3135 20.1399 2 0.2244


t-statistic 2 1.6936 2 2.7178 21.9303 2 2.0509
Table V.
Note: The MacKinnon critical value is 2 2.9339 for the 5 percent significance level ADF test for the loss ratio
JRF
Switzerland Germany USA Japan
7,2 F-statistic LLR F-statistic LLR F-statistic LLR F-statistic LLR

1960 0.4038 1.3951 0.6231 2.1332


1961 0.2821 0.9796 0.6797 2.3217
1962 0.7216 2.4607 0.8090 2.7489
154 1963 0.6949 2.3721 0.8558 2.9024
1964 0.3303 1.1445 0.5422 1.8624
1965 0.5285 1.8164 0.4229 1.4598
1966 0.9325 3.1527 0.4441 1.5319
1967 1.5233 5.0315 0.9203 3.1130
1968 1.2206 4.0797 0.7644 2.6020
1969 1.5448 5.0983 0.6883 2.3502
1970 1.8103 5.9140 0.8980 3.1725 0.6129 2.0994 1.8854 6.3226
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

1971 0.7954 2.7042 0.3413 1.2444 0.8648 2.9318 2.2699 7.4677


1972 1.6835 5.5266 1.5674 5.3427 1.6044 5.2829 1.8319 6.1601
1973 1.9637 6.3781 0.9241 3.2602 2.2334 7.1813 1.2670 4.3877
1974 4.8762* 14.3180* 1.6857 5.7110 2.0809 6.7292 1.0753 3.7624
1975 4.9980* 14.6187* 2.4210 7.9064* 1.1673 3.9096 1.7553 5.9255
1976 1.4836 4.9079 2.1613 7.1487 1.1934 3.9931 1.6773 5.6851
1977 3.5371** 10.8570** 1.6392 5.5667 1.3329 4.4353 1.9714 6.5826
1978 2.4211 7.7315 1.8212 6.1272 2.5279 8.0411** 2.0570 6.8389
1979 2.8543 8.9730** 0.7425 2.6460 3.0213** 9.4421** 2.0816 6.9123
1980 2.8495 8.9595** 1.0871 3.8011 2.6457 8.3798** 2.0425 6.7957
1981 2.8717 9.0223** 0.8162 2.8967 2.8616 8.9937** 2.1531 7.1243
1982 2.7764 8.7526** 0.5442 1.9612 2.8646 9.0022** 2.7120 8.7341**
1983 2.5352 8.0622** 0.4497 1.6292 2.4170 7.7195 3.7040** 11.4002*
1984 2.5388 8.0724** 0.5006 1.8084 2.3232 7.4455 4.0146** 12.1900*
1985 1.4785 4.8921 0.4943 1.7863 2.6860 8.4953** 4.1556** 12.5420*
1986 1.5292 5.0500 0.6721 2.4045 4.0817** 12.3001* 3.6948** 11.3766*
1987 1.4654 4.8511 0.5330 1.9220 2.5980 8.2430** 1.9464 6.5073
1988 1.5789 5.2040 0.4573 1.6560 3.0800** 9.6055** 1.9169 6.4183
1989 1.9268 6.2669 2.8740** 9.0287** 1.2504 4.3339
1990 2.5233 8.0277** 2.7712 8.7379** 2.0069 6.6891
1991 1.6298 5.3614 3.4398** 10.5939** 4.5257** 13.4473*
1992 0.6806 2.3248 3.4562** 10.6385** 1.8046 6.0766
1993 0.3811 1.3178 4.1916** 12.5851* 0.0717 0.2657
Table VI. 1994 0.3852 1.3317 0.8429 2.8601 0.1478 0.5451
Results from the Chow 1995 0.9841 3.3203 0.5116 1.7595 0.1109 0.4099
test and switching
regression Notes: * is for 1 percent significance level; ** is for 5 percent significance level

Switzerland was consistently lower than 55 percent before 1975 but higher than 55
percent after 1975. The loss ratio for Germany is lower than 70 percent and rather
volatile in the first sub-period (before 1975), but above 70 percent in the second
sub-period. The Japanese loss ratio is above 40 percent in the first period but lower
than that afterwards. Using combined ratio, Leng (2000) found a break for the USA in
1981. Apparently, the years of the breaks are different for the combined ratio and for
the loss ratio in the USA because the expense ratio is not constant. To reflect the
fluctuations of the underwriting results, combined ratio is more suitable for studying
structural change than the loss ratio. However, underwriting expenses are only
available for the USA. Different variables with different years of breaks for the USA
Multinational
underwriting
cycles

155

Figure 3.
Log likelihood ratios from
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

switching regressions for


Switzerland

help us to be aware of a possible bias from using the loss ratio as a variable in the
analysis for the other countries.

Discussion
The breaks for Switzerland and Germany coincide with a serious recession in these two
countries. If the recession did cause the break in the insurance industry, the loss ratios
should be cointegrated with GDP or/and CPI for the whole period regardless of the time
of the break. If the break is caused by changes of regulations or competitiveness in the
property-liability insurance industry, the relationship between the loss ratio and the
interest rate should change after the break. This may be the explanation for the break
in the USA since regulations changed pricing rules from bureau rating to competitive
pricing. In addition, increasing competition in the property-liability insurance market
forced insurers to take investment income into account in the ratemaking process from
the end of the 1970s to the beginning of the 1980s[13]. It is also possible that the
liability crisis from 1984 to 1985 effected tougher underwriting standards. If this is the
case, the break should only affect the liability part of business. However, most lines of
business include both property and liability, which are difficult to separate. Based on
the available data, the loss ratio should be co-integrated with the premium to surplus
ratio, which is used to measure the insurers risk taking behavior[14].
In Japan, the loss ratio behaves different from the loss ratios in the other three
countries: it falls after the break. Therefore, the cause of the break in Japan is
apparently different from those in the other countries. In the 1980s, savings-type
insurance policies started to become popular in the Japanese non-life insurance market.
From 1985 to 1994, the savings portion reached a share of 30 to 45 percent of the
non-life insurance premiums. The premiums for savings-type insurance contain a risk
premium and savings, which accounts for more than 90 percent of the premium. If no
covered incident occurs during the policy period, insurers refund the savings portion of
the premium and guarantee the payment of interest to the policyholder after the policy
matures. If a covered incident occurs, the claims are determined by the policy
agreement. Since internal capital includes a savings portion of the premiums, it should
have increased by more than the loss ratio when savings-type insurance became
JRF popular. As a corollary, the relationship between the loss ratio and internal capital
7,2 should have changed. It is also worth noting that Japanese insurers invest much more
of their assets in stocks, loans, and real estate than US insurers. Therefore, the
condition of the economy should have a stronger effect on the income of Japanese
insurers than interest rates do[15].

156
Further analysis
In this section, we test for the possible reasons for the cause of the break for each
country and whether these are country-specific factors. To do so, we apply
co-integration analysis for the loss ratios and the macroeconomic variables as in Leng
et al. (2002, 2004). For Switzerland, the interest rate has a break in 1975 as the loss ratio.
The results of the co-integration analysis are in Table VII. They show that loss ratio
and interest rate are co-integrated after the break, which implies that the break is most
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

likely caused by regulatory changes. The loss ratio is also co-integrated with GDP, but
this relationship changed after the break. Thus, it supports our assumption that
economic recession had an impact on the insurance industry.
For Germany the loss ratio and the interest rate are co-integrated only after the
break, but the co-integrating coefficient is not 2 1, which is what insurance pricing
theory would suggest. Unlike Switzerland, the loss ratio in Germany is not
co-integrated with GDP. This implies that underwriting results from the two countries
are highly correlated not because their economies are closely tied but only because
their insurance markets are tightly connected.
For the USA, the loss ratio is neither co-integrated with interest rate nor with the
premium to surplus ratio. This may be caused by the small sample in the second
period. However, the loss ratio is co-integrated with internal capital, which is consistent
with the capacity constraints theory. In addition, internal capital in the second period is

Country Variables Period Coef. 1 Coef. 2 Constant

Switzerland LR&GDP Before break 1 2 0.1342 2 41.3617


After break 1 0.0047 2 60.8369
LR&I After Break 1 2 2.7396 2 46.8807
Germany LR&I Whole 1 2 16.6483 55.9011
After break 1 2 1.9606 2 57.6183
USA LR&IC Whole 1 2 0.0004 2 40.2437
Before break 1 2 0.0004 2 61.8495
After break 1 2 0.0001 2 72.1419
P&IC Whole 1 2 2.0462 47878.96
Before break 1 2 1.4257 2 21671.73
After break 1 2 0.7741 2 98889.13
Japan LR&I Before break 1 7.2962 2 101.60
LR&GDP Whole 1 2.91E-05 2 49.5584
Before break 1 8.50E-06 2 47.0896
Table VII. After break 1 2 8.47E-06 2 31.5038
Co-integration equations
for before and after Notes: Listed co-integration equations are at a 5 percent significance level. The first coefficients are
breaks normalized to be 1. An example of the first co-integration equation is LR 2 0.1342 *GDP 2 41:3617 0
twice as high as in the first period for the same amount of premium. This may be Multinational
caused by new regulatory requirements or by an increasing portion of liability underwriting
coverage.
For Japan, the co-integrating relationship between the loss ratio and GDP changed cycles
due to the break[16] and the one between loss ratio and interest rate only exists before
the break. This implies that the break is the result of the joint effect of the economic
situation and regulations. 157

Conclusions
The loss ratio series from the four countries are neither stationary nor stable and all
four countries have breaks in different years. This shows that even though
underwriting cycles are an international phenomenon, they depend heavily on regional
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

or country-specific effects, such as the economic environment and regulations rather


than solely on equivalent international/global effects. This assumption is supported by
the fact that the long-term relationship between loss ratio and the GDP does not hold
for every country at all times.
Financial theory and the insurance pricing theory suggest that the loss ratio and the
interest rate series should be co-integrated and this relationship should not be affected
by the presence of breaks. However, these two theories are not well supported
empirically. On the other hand, loss ratio and interest rate are co-integrated either
before or after the break in most of the countries, which implies that changes in the
regulations play an important role in the insurance industry.

Notes
1. Between 1968 and 1997, the data period studied in this article, only one insurance company,
Nissan Life, went bankrupt in 1997. Later, two other life insurance companies, namely
Chiyoda and Kyoei, went bankrupt in December 2000. In November 2001, a non-life
company, Taisei Fire and Marine, went bankrupt due to the reinsurance liability exposure
from 11 September 2001.
2. Internal capital is the sum of paid-in internal capital, reserve allocation, and balance carried
forward to the next financial year.
3. Due to the German reunification in 1989, data for former West Germany is available up to
1991 only.
4. Some ratios often used in the insurance literature need to be mentioned: (1) Pure loss ratio is
the ratio of incurred losses to premiums earned; (2) Loss ratio (LR) is the ratio of incurred
losses plus loss adjustment expenses to premiums earned; and (3) Combined ratio (CR) is the
loss ratio plus the expense ratio (ER). Expense ratio is the ratio of underwriting expenses to
premiums written. Combined ratio is often used to show insurers underwriting results. If the
CR . 100 percent, insurers suffer underwriting losses. Due to the data limitation, pure loss
ratio is used but we call it loss ratio throughout the article.
5. In 1995, the combined ratios for Switzerland, Germany, USA and Japan are 107 percent, 99
percent, 107 percent, 96 percent, respectively. In the same year, expense ratios for these
countries are 34 percent, 27 percent, 30 percent, and 46 percent. The Japanese distribution
system does have higher underwriting expenses.
6. However, a counter argument can be made as well. For example, there was a liability crisis in
the USA in 1984 to 1985, but the loss ratios of Germany and Switzerland did not have a peak.
JRF 7. Cummins and Outreville (1987) included a time trend in AR(2) process to adjust for the
downward trend of underwriting expenses.
7,2 8. We should take the first difference of LR when it is a difference-stationary (DS) process. LR
series is not a trend-stationary (TS) process because LR cannot go up or down unlimited.
Nelson and Plosser (1982), and Stock and Watson (1988) discuss these two processes in
detail.
158 9. The assumptions are set due to the data limitations. Usually, taxes on premium and
underwriting profit are not zero. Also, for liquidity purpose insurers do not invest all their
premiums and they do not invest only on a yearly basis. Especially, insurers for long tail
lines would invest with a focus on more than a year. Biger and Kahane (1978) developed a
fund-generating coefficient to measure the proportion of the premiums invested and how
long insurers invest their funds. Kahane (1978) and Leng (2001) show that this coefficient is
not equal to 1.
10. The ACF and PACF for the other three countries are similar as those for Switzerland. They
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

are available on request from the authors.


11. The number of lags for the lagged difference term in the test is determined by the Akaike
Information Criterion (AIC) and the Schwarz Bayesian Information Criterion (BIC). In our
case, one lag is included.
12. Judge et al. (1985) describe the method of switching regression, which seeks to identify the
switching point or year of structural change. Brown et al. (1975) call this method Quandts
log-likelihood ratio technique because it was originally developed by Quandt (1958).
13. Self-insurance and captives are the alternatives to insurance. Insurers use price competition
and reflect investment income into premiums to decrease consumers switching to alternative
methods. However, self-insurance and captives gain their popularity in a hard market.
14. The premium to surplus ratio is one of the tests used in the Insurance Regulatory
Information System (IRIS) to predict insurers financial strength to prevent insolvency. If an
insurer has a premium to surplus ratio of more than 3, which is considered as a high risk
underwriting practice, the insurer fails this test.
15. In 1995, US insurers invested 60.7 percent of their assets in bonds, but Japanese insurers
invested only 18 percent of their assets in bonds.
16. Before the break, loss ratio and GDP move into opposite directions, but after the break, the
two variables move into the same direction.

References
Biger, N. and Kahane, Y. (1978), Risk considerations in insurance ratemaking, Journal of Risk
and Insurance, Vol. 45, pp. 121-32.
Brown, R.L., Durbin, J. and Evans, J.M. (1975), Techniques for testing the constancy of
regression relationships over time, Journal of Royal Statistical Society, Series B, Vol. 37,
pp. 149-72.
Chen, R., Wong, K. and Lee, H. (1999), Underwriting cycles in Asia, Journal of Risk and
Insurance, Vol. 66, pp. 29-47.
Cummins, J.D. and Outreville, F. (1987), An international analysis of underwriting cycles in
property-liability insurance, Journal of Risk and Insurance, Vol. 54, pp. 246-62.
Doherty, N. and Kang, H. (1988), Interest rates and insurance price cycles, Journal of Banking
and Finance, Vol. 12, pp. 199-214.
Fairley, W.B. (1979), Investment income and profit margins in property-liability insurance:
theory and empirical results, Bell Journal of Economics, Vol. 10, pp. 192-210.
Fung, H.G., Lai, G., Patterson, G. and Witt, R.C. (1998), Underwriting cycles in property and Multinational
liability insurance: an empirical analysis of industry and by-line data, Journal of Risk and
Insurance, Vol. 65, pp. 539-62. underwriting
Gron, A. (1994), Capacity constraints in property-casualty insurance markets, Rand Journal of cycles
Economics, Vol. 25, pp. 110-27.
Haley, J. (1993), A cointegration analysis of the relationship between underwriting margins and
interest rate: 1930-1989, Journal of Risk and Insurance, Vol. 60, pp. 480-93. 159
Hill, R.D. and Modigliani, F. (1982), The Massachusetts model of profit regulation in nonlife
insurance: an appraisal and extensions, paper presented at the Massachusetts
Automobile Hearings, reprinted in Cummins, J.D. and Harrington, S.A. (Eds), Fair Rate
of Return in Property-Liability Insurance, Kluwer-Nijhoff, Boston, MA (1986).
Judge, G.G., Griffiths, W.E., Hill, R.C., Lutkepohl, H. and Lee, T. (1985), The Theory and Practice
of Econometrics, 2nd ed., John Wiley & Sons, Chichester.
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

Kahane, Y. (1978), Generation of investable funds and the portfolio behavior of non-life
insurers, Journal of Risk and Insurance, Vol. 45 No. 2, pp. 65-78.
Lamm-Tennant, J. and Weiss, M. (1997), International insurance cycles: rational
expectations/institutional intervention, Journal of Risk and Insurance, Vol. 64, pp. 415-39.
Leng, C. (2000), Underwriting cycles: stationarity and stability, paper presented at the Annual
Meeting of the American Risk and Insurance Association, Baltimore, MD.
Leng, C. (2001), An examination of the fluctuations in underwriting profits of property-liability
insurers, PhD dissertation, Temple University, Philadelphia, PA.
Leng, C., Powers, M.R. and Venezian, E. (2002), Did regulation change the competitiveness in
property-liability insurance? Evidence from underwriting and investment income,
Journal of Insurance Regulation, Vol. 22, pp. 57-77.
Leng, C., Powers, M.R. and Venezian, E. (2004), The relationship between underwriting profit
margin and investment income: changes in competitiveness in property and liability
insurance, Journal of Insurance and Risk Management, Vol. 3, pp. 35-62.
Meier, U.B. (2001), Underwriting cycles in property-liability insurance: do they (still) exist?,
paper presented at the Annual Meeting of the European Risk and Insurance Economists,
Strasbourg.
Nelson, C. and Plosser, C. (1982), Trends and random walks in macroeconomic time series,
Journal of Monetary Economics, Vol. 10, pp. 139-62.
Niehaus, G. and Terry, A. (1993), Evidence on the time series properties of insurance premiums
and causes of the underwriting cycle: new support for the capital market imperfection
hypothesis, Journal of Risk and Insurance, Vol. 60, pp. 466-79.
Quandt, R.E. (1958), The estimation of the parameters of a linear regression system obeying two
separate regimes, Journal of American Statistical Association, Vol. 53, pp. 873-80.
Stock, J. and Watson, M. (1988), Variable trends in economic time series, Journal of Economic
Perspectives, Vol. 2, pp. 147-74.
Venezian, E. (1985), Ratemaking methods and profit cycles in property and liability insurance,
Journal of Risk and Insurance, Vol. 52, pp. 477-500.
Winter, R. (1988), The liability crisis and the dynamics of competitive insurance market, Yale
Journal on Regulation, Vol. 5, pp. 455-99.

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com


Or visit our web site for further details: www.emeraldinsight.com/reprints
This article has been cited by:

1. Frauke von Bieberstein, Jrg Schiller. 2017. Contract design and insurance fraud: an experimental
investigation. Review of Managerial Science . [CrossRef]
2. Chengyi Pu, Bismark Addai, Xiaojun Pan, Pangtuo Bo. 2017. Securitization product design for Chinas
environmental pollution liability insurance. Environmental Science and Pollution Research 24:4, 3336-3351.
[CrossRef]
3. Shinichi Kamiya. 2016. Credit Crunch and Insurance Consumption: The Aftermath of the Subprime
Mortgage Crisis. Journal of Risk and Insurance . [CrossRef]
4. Catherine Bruneau, Nadia Sghaier. 2015. Cyclicity in the French Property-Liability Insurance Industry:
New Findings Over the Recent Period. Journal of Risk and Insurance n/a-n/a. [CrossRef]
5. Dorina Lazar, Michel Denuit. 2011. New evidence for underwriting cycles in US propertyliability
insurance. The Journal of Risk Finance 13:1, 4-12. [Abstract] [Full Text] [PDF]
Downloaded by Universitas Gadjah Mada At 00:56 18 March 2017 (PT)

6. B. Elango. 2009. Impact of insurers' product variety on performance across underwriting cycles.
Management Decision 47:2, 359-374. [Abstract] [Full Text] [PDF]
7. Ursina B. Meier. 2006. Multinational underwriting cycles in propertyliability insurance. The Journal of
Risk Finance 7:1, 64-82. [Abstract] [Full Text] [PDF]
8. Ursina B. Meier. 2006. Multinational underwriting cycles in propertyliability insurance. The Journal of
Risk Finance 7:1, 83-97. [Abstract] [Full Text] [PDF]

S-ar putea să vă placă și