G M ’ T MS E ( EWM077A20) U

M

’

T

MS

E

(

EWM077A20)

U

G

Under supervision from the University of Groningen of

Prof. dr. K.H.W. Knot (University of Groningen, De Nederlandsche Bank)

Under supervision from De Nederlandsche Bank of

Prof. dr. J.A. Bikker (Utrecht University, De Nederlandsche Bank) Dr. J.K. Gorter (De Nederlandsche Bank)

The Impact of Business Cycle Developments on

Profitability in the Non-Life Insurance Sector

Britt Rijnbeek

May 2009

Abstract

1 Introduction

Financial institutions fulfil a pivotal role in the financial sector. As a result, supervision on these institutions is desirable. With the current financial and economic crisis one could wonder whether insurance companies would suffer and hence be of possible concern to supervisors. As pointed out by Cutler and Ellis (2005) and the world’s second largest reinsurer Swiss Re (Sigma, No. 5 / 2001 and No. 4 / 2002), during the last 15 years – apart from the economic downturn in 2000/2002 – property-liability insurers could rely upon net investment income to offset operating losses with the exceptional performance of global stock markets and high interest rates. But could a financial crisis – along with a slowdown in the real economy – inducing lower stock markets and low bond yields, endanger insurers’ performance?

Cyclicality of underwriting profit rates and insurance prices in the property-liability insurance industry is a well known phenomenon. Since the post war period, profits have tended to rise and fall in fairly regular patterns lasting between six and eight years from peak to peak; this phenomenon is termed the underwriting cycle. The phases of the underwriting cycle are typically referred to as a “hard” market, where prices and profits are high and coverage is restricted and a “soft” market, where prices and profits are low and coverage is expanded (Niehaus, 1993).

The explanation of the observed cyclicality of underwriting performance is one of the most debated topics within the insurance literature. Most literature investigates the influence of industry-specific variables, such as policyholder surplus, net investment income, underwriting profit or specific economic variables such as interest rates, on insurers’ underwriting performance, underwriting margins or premium prices. In these studies the effects of macroeconomic variables are scarcely investigated. Unsoundness due to cyclical factors increases the probability of a deterioration of an economic downturn since all insurers are more or less exposed to the same conditions and hence monitoring macroeconomic indicators for macro prudential purposes is very relevant.

long-run relationship, but the effects of shocks to the general economic variables have little effect on the combined ratio. The combined ratio of property-liability insurance is the sum of the loss ratio and the expense ratio. The loss ratio is the sum of claims plus loss adjustment expenses divided by the sum of net premiums earned. The expense ratio is the sum of operating expenses divided by annual net premiums written (Cutler and Ellis, 2005).

Cutler and Ellis (2005), including similar independent economic variables, also find that macroeconomic variables have little impact on the property-liability insurance loss ratio. The data used in both analyses were obtained from Best’s Aggregates and Averages, which are published by the A.M. Best Company. The Aggregates and Averages series details current and historical statistics on the United States and Canadian property-liability insurance industries.

Although it appears that the underwriting cycle is mainly caused by industry-specific factors, the profits of non-life insurers are possibly influenced by disturbances in the overall economy, due to the nature of their business. Insurers’ reliance on their investment income could turn out dangerously in this respect when potential sources of distress rooted in business cycle developments surface. As a result, investigating the influence of macroeconomic variables solely on underwriting performance and not on revenues from investment activities, gives possibly not an adequate picture of profitability in general. Our work is motivated by the two papers mentioned above but our objective is to evaluate the influence of the business cycle not only on underwriting performance but also on investment income and on total profitability of non-life insurance companies worldwide. The current analysis differs from previous studies in two important ways. First, this study is unique in assessing the influence of external

This paper aims to assess the effects of business cycle developments on non-life insurers’ profits. It seeks to assess whether different phases of the business cycle could threat stability in the insurance system. The underlying theory is that insurance companies offset underwriting losses with investment income which is assumed to be dependent on the state of the economy. Insights into the developments of underwriting profits indicate that among others higher unemployment rates, a decreased demand for houses and other durable goods and an increase in fraud could negatively influence insurance demand and claims paid. Although the studies by Grace and Hotchkiss (1995) and Cutler and Ellis (2005) do not find a statistically significant relationship between macroeconomic variables and underwriting performance, possible influences of business cycle developments on underwriting results cannot be ignored.

The remainder of this paper is structured as follows. Section two investigates the existing literature concerning macroeconomic variables influencing insurance companies’ performance. The next section describes the data. The fourth section explores the empirical model, the hypotheses and the regression results. The last section concludes.

2 Literature overview

The focus of the existing literature has been on explaining the cyclical behaviour of premiums, or underwriting performance measured by the combined or loss ratio. A contribution to overall performance of non-life insurance companies has been neglected in these studies. Profits in the property-liability insurance sector are determined first by underwriting performance and second, by investment performance. In assessing the profitability of insurance companies it is necessary to recognize the investment income opportunity that is related to the underwriting activity. Thus, one needs to analyze the underwriting result in conjunction with investment income. Since we are interested in the impact of the business cycle on the general profits of non-life insurance companies it is important to consider through what channels the business cycle could possibly influence the underwriting result on the one hand and investment income on the other hand.

measured by the combined or loss ratio (see Venezian, 1985; Cummins and Outreville 1987; Doherty and Kang, 1988; Cummins, 1990; Niehaus and Terry, 1993; Haley, 1993; Gron, 1994; Cummins and Danzon, 1997; Fung et al., 1998). However, Grace and Hotchkiss (1995) were the first to include macroeconomic variables besides the interest rate. Grace and Hotchkiss examine the long-run and short-run relationship between fluctuations in the national business cycle and underwriting performance. Underwriting performance is measured by the combined ratio and macroeconomic variables included are real gross domestic product, the interest rate on 90-day US Treasury bills, and the Consumer Price Index. For estimating the long-run relationship they use cointegration techniques and for the short-run relationship they use vector autoregression techniques. They provide evidence that a long-run relationship exists between general economic changes and underwriting performance, but that short-run unanticipated general economic fluctuations have little effect on the combined ratio. The effects on profitability are greater. They find that an increase of one percent from the mean level of real gross domestic product causes a decrease of 0.70 percent in profits. However, this is only an indication of the effect on profits of a one-period response to a shock using some stylized facts.

Hence, based on these two papers, it seems that underwriting results do not show a considerable response to business cycle developments. Besides the empirical results of these two studies, many research studies focus on the question whether insurance is a normal good and hence whether demand for insurance would increase when income increases. Theoretically two different effects play a role concerning the influence of income on an individual’s purchase of insurance (Foncel and Treich, 2007). First, insurance consumption is hypothesized to be positively correlated with the level of risk aversion (Browne, 2000). A widely-used measure of risk aversion is the Arrow-Pratt measure of risk aversion. According to Campbell (1980) the most plausible theoretical form of the Arrow-Pratt measure of risk-aversion is a decreasing function of wealth. This hypothesis is called the hypothesis of Decreasing Absolute Risk Aversion (DARA), according to which absolute aversion to risk declines at higher levels of financial resources. Following this hypothesis we consequently expect wealthier people to demand less insurance. Second, wealthier people buy more valuable goods and thus are likely to demand more insurance because the risk of loss increases. Empirical evidence should disclose which effect is the dominating effect.

Beenstock et al. (1988) and Enz (2000) find an empirically positive relationship between per capita income and spending on property-liability insurance. Outreville (1990) finds a positive relationship between property-liability premiums per capita and GDP per capita for developing countries. These empirical results indicate that the risk of loss effect dominates the absolute risk aversion effect, or it may mean that the risk aversion is increasing rather than decreasing. Browne (2000) finds that GNP per capita is positively related with non-life premium density, which is consistent with the results found by Beenstock et al. (1988), Enz (2000) and Outreville (1990) that income is positively correlated with insurance consumption. Increased spending on insurance as a result of an increase in income could suggest greater profitability for the insurance companies. One should keep in mind though that positive developments in the business cycle do not necessarily imply an increase in income, which is in general defined as an increase in GNP per capita. However, in lack of previous theoretical or empirical literature on the effects of business cycle developments on underwriting profits we believe that understandings of the relationship between GDP per capita and insurance demand can provide insights into a possible impact of the business cycle on underwriting results.

yields, dividend income and capital gains (Sigma, No. 5 / 2001). During the ten years preceding the year 2000 insurance companies could rely, and in fact did, on high investment income. The equity markets were strong and declining interest rates produced capital gains on bond portfolios. This period is known as the dot-com bubble. During this stock market boom, insurance companies were increasing their investment risk by moving capital into equities and corporate bonds. By mid-2000 however, the dot-com bubble came to burst. Stock markets and interest rates were declining, resulting in a severe bear market and the real economy experienced a severe slowdown. This changed the investment environment and insurers became more conservative in their investment allocation. They started shunning equities and riskier investments and instead increasing the share of fixed-income investments in order to lower their investment risk (Sigma, No. 4 / 2002).

3 Data sources and description

In our regression analysis we try to estimate the effects of developments in the business cycle on non-life insurance companies’ overall performance. Our data set is unique since it not only covers insurance companies in North America as is the case with Best’s Aggregates and

Averages on which Grace and Hotchkiss (1995) and Cutler and Ellis (2005) base their study

but also insurance companies in South America, Europe, Asia, Australia and Africa. We use income statement and balance sheet data of non-life insurance companies from the ISIS database provided by Bureau van Dijk. The panel data set includes more than 6.000 non-life insurance companies in 125 countries during the years 1991 – 2007, comprising over 50.000 individual company observations. Over 21.000 observations concern however US companies, which is not surprising given the size of this market (North America is the largest market in the world with a world market share of 42 percent in premiums in 2007; Sigma, No. 3 / 2008). In our analyses distinctions are made and differences evaluated for G7 countries, Eurozone countries, countries with a real gross domestic product under $6.000 per capita labelled as Poor countries, and the United States of America. A list of the countries included in the several regional categories appears in the Appendix A Table 1.

In defining the dependent, we do not follow Grace and Hotchkiss (1995) or Cutler and Ellis (2005). They select the combined ratio and the loss ratio respectively as the dependent variable. We consider profits to be a better estimate of the general performance of a non-life insurance company than the loss or combined ratio since it comprises the connectedness between underwriting performance and investment returns. In our regressions profits is defined as the sum of the underwriting result and net investment income and therefore other gains and losses are excluded from profit estimates. If the business cycle would influence profits through other gains and losses, it would be difficult to draw implications of this impact since the origin of other gains and losses is unclear. To take into account size differences of companies the dependent is defined as underwriting result plus net investment income divided by net premiums earned. To analyse possible influences of the business cycle more explicitly we take net investment returns and underwriting results per net premium earned separately as dependents as well. In analysing the underwriting result, the ratio of underwriting result over net premium earned is considered and in analysing the net investment income the ratio of net investment income over total investments is considered.

As a proxy variable of the business cycle we choose annual percentage deviations from the 1991 – 2007 country average of real gross domestic product. To be precise, we take the first difference of the logarithm of gross domestic product in current prices in million USD, essentially the percentage change in gross domestic product. The deviation of this first difference from the mean of these first differences of the logarithm of gross domestic product over 1992 – 2007 yields our dependent. While Cutler and Ellis (2005) take the first difference of the logarithm of real gross domestic product, we select this particular proxy since we investigate pure cyclical movements and differences of long term growth trends between countries are controlled for when the business cycle is expressed by the deviation of GDP growth from its country specific average. The data for real GDP and real GDP per capita – used to select the countries in the category “Poor countries” – are coming from the World Development Indicators database from The World Bank.

which macroeconomic developments affect the profitability of non-life insurance companies, we add several control variables. Control variables included are inflation, total percentage returns on the MSCI World index, and the change in interest rate on five-year US Treasury notes. Inflation is expressed as the annual change in the USD deflator for constant prices, with the year 1991 equalized to unity. For the five-year US Treasury notes the yearly published five year Constant Maturity Treasury Middle Rate is selected. This index is an average yield on the most recently auctioned United States Treasury securities adjusted to a constant maturity of five years, as made available by the Federal Reserve Board. The US Treasury yield is incorporated in the model as a first difference variable since a negative change in interest rates yields capital gains with respect to investment income. The data for the control variables inflation, total returns on the MSCI World index, and the interest rate on five-year US Treasury notes originate from the International Financial Statistics of the International Monetary Fund, from MSCI, and from the Department of the Treasury respectively and are extracted from DataStream.

These variables are related to those used by Grace and Hotchkiss (1995) and Cutler and Ellis (2005). By selecting the five-year US Treasury notes we oppose Grace and Hotchkiss selecting the 90-day US Treasury bill and we follow Cutler and Ellis. From A.M. Best we know that American Property-Casualty insurers’ bond portfolio was for 30.5 percent invested in 1-5yrs bonds, 31 percent in 5-10yrs bonds and only 15 percent in one-year bonds or less over the period 1999 – 2004E. As such we consider this interest rate to be more appropriate in investigating changes in interest rates on investment income. Moreover, Cutler and Ellis select this interest rate because “the issuance period captures the loss tail of most property-casualty claims” (2005, p. 132).

An accounting breakdown of the underwriting components and an overview of the descriptions of the dependent and independent variables can be found in the Appendix A, Table 2 and 3 respectively.

We apply a number of selection rules to the most important variables. Negative net premium earned and net premium earned equal to zero are considered to be dubious and hence these observations (1.995 in total) are excluded. An underwriting result exceeding net premium earned seems doubtful since the underwriting result consists of net premium written (which differs from net premium earned by movements in unearned premium) minus underwriting expenses (net claims, movements in insurance funds, commission and management expenses). As a result we eliminate 277 observations with a ratio exceeding unity. The final sample consists of 51.144, 46.305, 46.174 observations for the investment ratio, underwriting result ratio and the profit ratio respectively.

In Appendix A, Table 4 provides descriptive statistics for the dependent and independent variables for both the worldwide sample and several geographical subcategories. The underwriting result ratio has a weighted mean of -0.12, implying that on average a loss of $0.12 per $1 net premium earned is generated in underwriting activities. When investigating regional differences we find particularly high underwriting losses per net premium earned in the US (experiencing a mean underwriting result ratio 1.7 times higher than the worldwide mean) and in the – by the US dominated – G7 (1.25 times as high). In the Eurozone and Poor countries the underwriting losses per net premium earned are considerably lower than the worldwide mean (0.4 and 0.5 times the worldwide mean respectively).

The mean of 0.05 of the net investment income ratio indicates a return of $0.05 of net investment income on one dollar invested. Investigating regional differences we find a net investment ratio close to the worldwide mean for G7 countries, the US and Eurozone countries but a considerably higher net investment ratio for Poor countries (1.6 times higher than the worldwide mean). The spread of the ratio, as derived from the standard deviation, is noticeably bigger in the worldwide sample, the G7 and in the US than the spread of the ratio in the Eurozone and in Poor countries.

worldwide weighted average. The US and Poor countries know a mean below the worldwide average and the Eurozone knows a mean above the worldwide average.

By construction the mean of deviations from the average GDP growth rate – which is 5.6 percent worldwide over the period 1992 until 2007 – is zero. The majority of average national growth rates are between zero and ten percent and the majority of the deviations are within 25 percent above or below the average growth rate. The US knows a relatively stable growth rate with 95 percent of the deviations within 2 percent from the average growth rate of 5.3 percent. Poor countries on the contrary know a relatively volatile growth rate with 95 percent of the deviations within 32 percent from the average growth rate being 7 percent over 1992-2007. Over the period 1991 – 2007 US inflation shows a mean of 2.27 percent, a minimum of 1.19 percent in 2002 and a maximum of 4.19 percent in 1991. Total returns on the MSCI World index over the period 1991 – 2007 average 10.88 percent, showing a minimum of -19.5 percent in 2002 and a maximum of 33.8 percent in 2003. The five year Constant Maturity Treasury Rate averaged 4.99 percent over the period 1991 – 2007 with a minimum of 2.78 percent in 2002 and a maximum of 7.83 percent in 1994.

While non-life insurance companies in Poor countries show a better underwriting result per premium earned and better investment returns than the weighted average of the worldwide ratios, the mean for the profit before tax ratio in this region is lower than the profit before tax ratio for the worldwide sample possibly because they have relatively small total investments in comparison with net premiums earned in contrast to other regions. In comparing the Eurozone with the US, insurance companies in Eurozone countries experience much better underwriting results for every dollar premium earned but investment returns are slightly below the investment returns in the US. Profits per dollar net premium earned are $0.23 on average in the Eurozone and $0.15 in the US. Comparing insurance companies in Eurozone countries with the G7 yields fairly similar results.

Graphical representations of the developments of the dependent variables over time, including regional differences, are given in Figures 1-4 in Appendix B. However, one should note that since almost half of the observations are for American companies, the United States will dominate results of the worldwide sample and the G7 sample.

The weighted average of the net investment ratio of the worldwide sample has been fairly stable over time. Over 1991 – 2007 net investment income per dollar invested has been between 4.5 percent and 6.5 percent on average. Profit before tax over net premium earned has been positive since 1992 and is showing a positive change since 2002. In 2007 we observe a remarkable occurrence; underwriting results are taking a nosedive while investment income barely changes and profits before tax continue their rise. In the regional analysis we see that the fall in underwriting results is found in the US and G7 and not in the Eurozone and in Poor countries. The deterioration in underwriting results is the result of weakness in premiums (reflecting increased competition) and an increase in loss and loss adjustment expenses. In establishing a distinction between regions, underwriting results have been better for companies in Eurozone and Poor countries than for companies in G7 countries and the United States since 1995.

Investment returns for companies in United States countries are close to the returns for companies in G7 countries. For the majority of the years the investment income over total investments for companies in Eurozone countries is lower than companies in the United States and the G7. The investment ratio in Poor countries stands out in being above the investment ratio in the other regions, but also more volatile.

For two countries we graphically investigate the relationship between the dependents and the independent of our main interest, deviations from average GDP growth; see Figures 5-10 in Appendix B. The United States and Germany are selected since these countries have the greatest data availability. The graphs show the developments of the yearly weighted means of the underwriting result ratios, of the net investment income ratios and of the profit before tax ratios with the developments of the deviations from the average GDP growth rate over the period 1992 – 2007. In the United States the impact of deviations from average GDP growth on profits seems the most considerable relationship while in Germany the impact of deviations from average GDP growth on the underwriting result seems most considerable. In both countries the correlation of deviations from average GDP growth with net investment income seems minimal. An overview of the numerical correlations between the weighted means of the dependent variables and deviations from average GDP growth is given in Table 6 in Appendix A. The results are contradicting with our suggestions based on the graphical representations and the correlation of deviations from average GDP growth with net investment income appears far more considerable than suggested. However, what is actually important are the correlations between deviations from average GDP growth and the dependent variables based on individual observations instead of on the weighted mean of the ratios per year. An overview of these correlations is given in Table 7 in Appendix A. Measured in this way the correlations seem negligible and therefore one should be cautious with the suggestions stemming from the graphical representations since based on these numerical correlations deviations from average GDP growth do not correlate with any of the dependent variables.

4 Empirical model and results

4.1 Empirical model

To investigate the empirical relationship of (a) the underwriting result ratio (b) the investment income ratio, and (c) the profit before tax ratio with the business cycle, we estimate the following equations:

(a) The underwriting result ratio

(b) The investment income ratio

(c) The profit before tax ratio

Here, the subscript i represents the firm, the subscript j represents the country and the subscript t represents time;

γ

θ

φ

investment income ratio and the profit before tax ratio respectively for firm i in country j at time t; α0 is the fixed intercept term and αi are firm-specific intercept terms; β1, …, β5 are the

slope parameters; and εit is the random error term. The control variables inflation, total returns

on the MSCI World index and the five-year US Treasury Rate are only dependent on time. The main interest of our paper is the correlation of the deviations of average growth rates with the profit before tax ratio.

4.2 Hypotheses

Literature concerning the insights into possible effects of macroeconomic developments on the profitability of non-life insurance companies is scarce. Grace and Hotchkiss (1995) and Cutler and Ellis (2005) find no relationship between macroeconomic variables and underwriting performance but theoretical insights are lacking.

workers’ compensation premiums. Housing demand and demand for durable goods like cars usually slump in a weakening economy. Especially in the current economic downturn exposure growth forecast for homeowners insurance is dim for 2009. The Insurance Information Institute estimates that for the United States each incremental 100,000 decline in housing starts costs home insurers $87.5 million in new exposure (gross premium) (Hartwig, 2009). The net exposure loss in 2009 versus 2005 is estimated at about $1.2 billion. Vehicle demand drives insurance demand. The weakening economy is taking its toll on vehicle sales affecting exposure growth in auto insurance.

Many people depend on their employer for their health and disability insurance. Fifty percent of all people in the US with health insurance coverage receive this insurance through their employer (Cawley and Simon, 2003). An additional nineteen percent receive it through the employer of a parent or spouse. A recession may increase the rate of people with no insurance coverage when the newly unemployed are losing health insurance which was provided by their previous employers. Overall, the study of Cawley and Simon (2003) indicate that forty-four percent of the people that fall into unemployment become uninsured as a result. Also, employers may cease offering health insurance to employees to cut costs in the face of falling profits. Consequently, premiums written would decrease when unemployment rates rise. According to a report by the Center for Economic and Policy Research (Schmitt and Baker, 2008), this year’s recession will cause at least 4.2 million people to lose their health insurance coverage in the United States.

On the cost side, the claims paid, an impact of an economic downturn could also be experienced. As the Fraud Prevention Service of the United Kingdom (the CIFAS) and the Association of British Insurers (ABI) suggest, economic slowdowns and rises in fraud go hand in hand. As the economic situation worsens, the opportunity to commit fraud, to obtain easy money, increases and could therefore result in relatively bigger costs in claims for insurers. However, there has been no solid statistical evidence. A leading provider of Social Security disability, Allsup, notes that disability insurance claims have generally increased during the seven recession periods which occurred since 1969. A possible explanation for this increase is that in periods when the economy contracts, people who have struggled with a disability are forced to look for alternatives they might not have otherwise considered, including disability insurance.

A possible positive influence of economic downturns on claims paid can result from workers’ compensation coverage. Workers’ compensation data indicates that frequency fell or declined faster during the three most recent recessions. Inexperienced workers are much more likely to be injured on the job. As the economy moves into recession, employers typically lay off their newest hired, least experienced workers.

Overall, due to increasing unemployment during negative developments in the business cycle and a decreased demand for durable goods which are in many cases the underlying exposure base of non-life insurance we expect a negative relationship between business cycle developments and premiums written. Also, claims paid could relatively increase because of among others fraud, an increase in burglary and bigger efforts to collect insurance coverage in times of economic hardship. With lacking statistical evidence we either expect an insignificant or a positive relationship between deviations from average GDP growth and underwriting profitability.

The expected relationship between the profit before tax ratio and the business cycle is a priori ambiguous. Profit before tax is defined as the underwriting result plus net investment income and hence the sign and possible significance depends on the impact of the business cycle on investment income and the underwriting result. Most in line with expectations are a positive or an insignificant coefficient.

The independent variable inflation is incorporated as a pure control variable and only a potential negative impact is expected on the underwriting result ratio. Underwriting results may worsen when facing a period of relatively high inflation since net premium earned tends to be less easily adjusted for inflation than losses paid. As a result inflation adjusted underwriting results look worse with periods of high inflation since corrected income is understated (Hodes, 1999).

4.3 Empirical results

This section presents the regression results for the worldwide data set and for the following regions: the G7, the United States, the Eurozone countries, and Poor countries. We report the regression results for the ratio of underwriting result over net premium earned, the ratio net investment income over total investments, and for the ratio profit before tax over net premium earned as dependent variables. We have found that heteroskedasticity could be a problem in some of the regressions and adjusted for it accordingly by using the standard heteroskedasticity-robust estimator. Multicollinearity is not a problem since none of the correlations between the independent variables exceeds 0.35 as can be seen in the correlation matrices in Table 8 in Appendix A. The correlation matrices for the different regions show no multicollinearity with an exception of a higher correlation (0.8) between deviations from the average GDP growth rate and lagged returns on the MSCI World index in the United States. The results for the worldwide sample and the several geographical subcategories are displayed in Table 9 and 10 respectively in Appendix A. The results of the impact of the independent variables on the dependents are discussed per independent below. In the worldwide sample, the underwriting result ratio is the only dependent significantly affected by business cycle developments as measured by deviations from average GDP growth. The coefficient is rather small though. Besides deviations from average GDP growth underwriting results are positively influenced by inflation. Investment returns are positively affected by the variable total returns on the MSCI World index and negatively by changes in US Treasury rates which is in line with expectations.

Deviations from average GDP growth

Another candidate explanation for this relationship could be that insurance is a normal good in that an increase in the real income of an individual or economy increases insurance demand. This relationship is found in previous studies where GDP per capita is included as the independent variable. However this explanation has to be accompanied with the assumption that an increase in wealth is positively correlated with deviations from the average GDP growth rate in order to be a possible explanation for our findings.

The insignificance of the impact of the business cycle on the net investment income ratio could indicate prudent investment strategies of insurance companies and hence that investment returns are less influenced by the business cycle developments. However, the channels through which we expect the business cycle to affect net investment income are through changes in stock and bond markets. These developments generally precede the business cycle. Nevertheless, taking a lag of the independent variable deviations from average GDP growth does not bring better results.

By investigating possible regional differences between G7 countries, the United States, Eurozone countries, and Poor countries we find significant coefficients for all regions but the Poor countries, see Table 10 in Appendix A. As in the worldwide sample we find a positive relationship between deviations from the average GDP growth and underwriting results in the Eurozone. A 10 percent deviation from the average GDP growth would result in an increase of 0.028 dollar of underwriting result per dollar net premium earned. Furthermore, deviations from average GDP growth have a fairly equal negative impact on net investment results in the G7 and the Eurozone. The coefficient is very small though. A 10 percent deviation from the average GDP growth would result in a decrease of 0.3 percent in investment returns. It is rather difficult to provide an explanation for this negative relationship. Lastly, deviations from the average GDP growth rate influence the profit before tax ratio in the US. A 10 percent deviation from the average GDP growth rate would result in an increase of 0.3 dollar of profits per dollar net premium earned. Taken into account the positive relationship between deviations from average GDP growth and the underwriting result ratio found in the worldwide sample, this positive relationship could result from the fact that positive/negative deviations from average GDP growth cause higher/lower underwriting results and therefore higher/lower profitability.

Inflation

in the worldwide sample. A significant positive relationship between inflation and the underwriting result is found in the regional analyses for G7 countries and the United States. This result however, is hard to explain theoretically since the only existing literature stating suggestions for the relationship between underwriting result and inflation indicates that underwriting results worsen facing inflation since net premium earned is less easily adjusted for inflation than losses paid (Hodes, 1999). As a result inflation adjusted underwriting results look worse in periods of high inflation since corrected income is understated. However, this indicates a negative relationship in contrast with the positive relationship found in our regression analyses. Another significant coefficient is found when the distinction is made between the different regions. Inflation appears to have a positive impact on net investment income in G7 countries, the United States and Eurozone countries and at the 10 percent significance level in Poor countries. A candidate explanation could be that investments are shifted from fixed income securities towards equities in periods of high inflation, leading to a higher risk higher return portfolio. A rate of inflation of 10 percent in comparison with a rate of inflation of zero increases investment returns between 0.03 and 0.04 percent.

Returns on the MSCI World index

A strong expectation concerning the influence of Returns on the MSCI World Index is the expected positive relationship with net investment income. Although small, we do find a positive significant relationship in the worldwide sample. Total returns of 10 percent increases investment returns with 0.1 percent in comparison with returns of 0 percent on the MSCI World index. The MSCI World index is possibly not very representative for investment allocations in equities for non-life insurance companies. Another more likely explanation could be the dominance of investments in fixed income securities in comparison with stock investments.

Changes in the five-year US Treasury yield

For this independent variable there is a strong expectation concerning the influence of the change in Treasury rates, namely the negative relationship with net investment income. Falling interest rates implies higher prices and therefore the market values of fixed income securities would increase. This negative coefficient is found in both the worldwide sample and in all separate regions, varying between -0.0010 and -0.0074. This implies that for non-life insurance companies in the United States for example a 1 percent decrease in the five-year US Treasury rate yields an increase in investment returns of 0.1 percent.

Industry-specific control variables

In regressing the underwriting result ratio we include several industry-specific control variables. Investment results and underwriting results are highly interdependent. We expect high investment returns in the previous period to have a negative impact on underwriting performance due to low and possibly insufficient premium levels in order to strive for market share without risking overall losses. Therefore, we expect the lagged versions of total returns on the MSCI World index and of the net investment income ratio in the regression for the underwriting result ratio to exhibit negative coefficients. The lagged change in the five-year US Treasury rate is expected to show a positive impact on underwriting results since decreasing interest rates result in capital gains and hence a strong investment performance which accommodates low price setting. However, the regression results indicate that none of the included industry-specific variables significantly influences the underwriting result ratio, with an exception of the coefficient for lagged returns on the MSCI World index in the Eurozone. The results are therefore inconsistent with the hypothesis that strong investment performance in the previous period results in aggressively lower pricing and in lower underwriting results as insurers strive for market share.

5 Conclusion

effects of macroeconomic variables on underwriting performance. In assessing the overall profitability of insurance companies however, it is necessary to recognize the investment income opportunity that is related to the underwriting activity. Using a worldwide panel data set covering 125 countries during the years 1991 – 2007, we find that of more than one third of the nonnegative profit observations, negative underwriting results are offset by investment income. This reliance on investment income could turn out dangerously assuming that investment returns are dependent on the business cycle.

Literature overview

Beenstock, M., G. Dickinson, S. Khajuria (1988) The Relationship Between Property-Liability Insurance Premiums and Income: An International Analysis. Journal of Risk and

Insurance 55, pp. 259-272.

Browne, M. J., J. W. Chung, E. W. Frees (2000) International Property-Liability Insurance Consumption. Journal of Risk and Insurance 67, pp. 73-90.

Campbell, R. A. (1980) The Demand for Life Insurance: An Application of the Economics of Uncertainty. The Journal of Finance 35, pp. 1155-1172.

Cawley, J., K. I. Simon (2003) The Impact of Macroeconomic Conditions on the Health Insurance Coverage of Americans. NBER/Frontiers in Health Policy Research 6, pp. 87-114. Cummins, J. D. (1990) Multi-period Discounted Cash Flow ratemaking Models in Property-Liability Insurance. Journal of Risk and Insurance 57, pp. 79-109.

Cummins, J.D., P. Danzon (1997) Price, Financial Quality and Capital Flows in Insurance Markets. Journal of Financial Intermediation 6, pp. 3-38.

Cummins, J. D., Outreville J. F. (1987) An International Analysis of Underwriting Cycles in Property-Liability Insurance. Journal of Risk and Insurance 54, pp. 246-262.

Cutler, D. R., P. M. Ellis (2005) A Simple Model to Predict Loss Ratios in the Domestic Stock Property-Liability Insurance Industry. Quarterly Journal of Business and Economics 44, pp. 129-139.

Doherty, N. A., H. B. Kang (1988) Interest Rates and Insurance Price Cycles. Journal of

Banking and Finance 12, pp. 199-214

Enz, R. (2000) The S-Curve Relation Between Per-Capita Income and Insurance Penetration.

Geneva Papers on Risk and Insurance 25, pp. 396-406.

Foncel, J., N. Treich (2007) Insurance as a Normal Good: Empirical Evidence for a Puzzle.

Working Paper Toulouse School of Economics

Fung, H., G. C. Lai, G. A. Patterson, R. C. Witt (1998) Underwriting Cycles in Property and Liability Insurance: An Empirical Analysis of Industry and By-Line Data. Journal of Risk and

Insurance 65, pp. 539-562.

Grace, M. F., J. Hotchkiss (1995) External Impacts on the Property-Liability Insurance Cycle.

Journal of Risk and Insurance 62, pp. 738-754.

Gron, A. (1994) Capacity Constraints and Cycles in Property-Casualty Insurance Markets.

RAND Journal of Economics 25, pp. 110-127.

Hodes, D. M., S. Feldblum, A. A. Neghaiwi (1999) The Financial Modeling of Property-Casualty Insurance Companies. North American Actuarial Journal 3, pp. 41-69.

Hartwig, R. P. (2009) 2008 – Year End Results. Insurance Information Institute April 2009. Niehaus, G., A. Terry (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 60, pp. 466-479.

Outreville, J. F. (1990) The Economic Significance of Insurance Markets in Developing Countries. Journal of Risk and Insurance 57, pp. 487-498.

Schmitt, J., D. Baker (2008) What We’re In For: Projected Economic Impact of the Next Recession. Center for Economic and Policy Research January 2008.

Swiss Re (2001) Profitability of the non-life insurance industry: it’s back-to-basics time.

Sigma 5 of 2001.

Swiss Re (2002) Global non-life insurance in a time of capacity shortage. Sigma 4 of 2002. Swiss Re (2008) World insurance in 2007: emerging markets leading the way. Sigma 3 of 2008.

Appendix A

Table 1 Specifications of included countries per regional sub sample

G7 countries Canada France Germany Italy Japan United Kingdom

United States of America Eurozone countries Austria Belgium Cyprus Finland France Germany Greece Ireland Italy Luxembourg Malta Netherlands Portugal Slovenia Spain Poor countries Albania Algeria Argentina Azerbaijan Bangladesh Barbados Belarus Bolivia

Bosnia and Herzegovina

Botswana Brazil Bulgaria Cambodia Cameroon Chile China Colombia Croatia Czech Republic Dominican Republic Ecuador Egypt El Salvador Estonia Ethiopia Fiji Ghana Guatemala Honduras Hungary India Indonesia Iran Jamaica Jordan Kazakhstan Kenya Korea Lao Latvia Lebanon Lesotho Lithuania Macedonia Malawi Malaysia Malta Mauritius Mexico Morocco Mozambique Nicaragua Nigeria Oman Pakistan Panama

Papua New Guinea Paraguay Peru Philippines Poland Romania Russian Federation Serbia Slovakia Slovenia South Africa Sri Lanka Swaziland

Syrian Arabic Republic Tanzania

Thailand

Table 2 Accounting breakdown of underwriting components

Gross premium written

Premium ceded -

Net premium written

Net premium written

Movement in unearned premium -

Net premium earned

Gross claims paid

Gross change in provision for claims +

Gross claims

Gross claims paid

Reinsurer's share/claims paid +

Net claims

Net claims

Movement in insurance funds +

Commission expenses +

Management expenses +

Underwriting expenses

Net premium written

Underwriting expenses -

Underwriting result

Underwriting result

Net investment income +

Other income +

Table 3 Variable description overview

Unit Description

Dependent variables

Underwriting result Ratio Underwriting result / Net premium earned Investment income Ratio Net investment income / Total investments Profit before tax Ratio Profit before tax / Net premium earned

Independent variables

Business cycle Percentage First difference of log of real GDP – Mean of the first difference of log GDP over 1992-2007 Inflation Percentage Annual change in the USD deflator for constant

prices

MSCI World index Percentage Total returns on the MSCI World index Change in the 5y US

Treasury rate

Table 4 Descriptive statistics

* Underwriting result / Net premium earned ** Net investment income / Total investments *** Profit before tax / Net premium earned

Whole sample G7 US Eurozone Poor countries

_Standard deviation _Standard deviation _Standard deviation _Standard deviation _Standard deviation

Variable Mean Mean Mean Mean Mean

Dependent variables

Underwriting result ratio* -0.12 2.23 -0.15 2.74 -0.20 3.36 -0.05 0.54 -0.06 0.65

Net investment income ratio** 0.05 0.34 0.05 0.26 0.05 0.07 0.04 0.41 0.08 0.65

Profit before tax ratio*** 0.19 2.23 0.19 2.48 0.15 2.01 0.23 3.20 0.16 0.90

Independent variables

Deviations from average GDP growth 0.00 9.72 0.00 5.29 0.00 0.97 0.00 9.02 0.00 16.12

Average GDP growth 5.66 2.08 4.97 0.82 5.27 0.00 4.87 1.44 7.00 3.25

US Inflation 2.27 0.68

Total return on MSCI World index 10.88 15.18

Table 5 Relative contributions to nonnegative profits of investment income versus underwriting results

All G7 US Eurozone Poor countries

Total nonnegative 40.735 25.774 16.807 11.725 7.030

Positive because of net investment income 14.927 9.838 6.450 3.750 52 (36.6%) (38.2%) (38.4%) (32.0%) (0.7%) Positive because of underwriting results 233 30 8 44 128 (0.6%) (0.1%) (0.0%) (0.4%) (1.8%)

Table 6 Correlation matrix between the business cycle and weighted means of the dependent variables per year

Underwriting result ratio* Net investment income ratio** Profit before tax ratio*** United States Deviations from average GDP

growth 0.11 -0.21 0.21

Germany Deviations from average GDP

growth 0.32 -0.47 0.40

Table 7 Correlation matrix between the business cycle and the dependent variables based on individual observations

Underwriting result ratio* Net investment income ratio** Profit before tax ratio*** United States Deviations from average GDP

growth 0.00 -0.03 0.01

Germany Deviations from average GDP

growth 0.02 -0.02 0.03

Table 8 Correlation matrices of the independent variables For the underwriting result ratio

Deviations from average GDP growth Lagged total returns on MSCI World index Lagged change in US Treasury 5y yield Lagged net investment income ratio Inflation

Deviations from average

GDP growth 1.000

Inflation -0.023 1.000

Lagged total returns on

MSCI World index 0.048 0.301 1.000 Lagged change in US

Treasury 5y yield 0.035 0.342 0.354 1.000

Lagged net investment

income ratio 0.000 -0.001 -0.005 -0.007 1.000

For the net investment income ratio

Deviations from average GDP growth Total returns on MSCI World index Change in US Treasury 5y yield Inflation

Deviations from average

GDP growth 1.000

Inflation -0.007 1.000

Total returns on MSCI

World index 0.134 -0.106 1.000

Change in US Treasury 5y

yield 0.051 0.094 0.223 1.000

For the profit before tax ratio

Deviations from average

GDP growth Inflation Deviations from average

GDP growth 1,000

Table 9 Fixed-effects model parameter estimates for the worldwide sample

Dependent variables

Independent variables UWR* NII** PBT***

Intercept -0.3086 0.0513 0.2505

(0.000) (0.000) (0.000) Deviations from average GDP growth 0.0010 -0.0021 0.0013 (0.005) (0.148) (0.149)

US Inflation 0.0943 0.0012 -0.0079

(0.001) (0.753) (0.787) Total returns on MSCI World index (%) 0.0001

(0.026) Lagged total returns on MSCI World index -0.0007

(0.239)

Change in US Treasury 5y yield -0.0017

(0.000) Lagged change in US Treasury 5y yield -0.0114

(0.190) Lagged Net investment income ratio -0.0109 (0.103) Note: p-values are in parentheses

Table 10 Fixed-effects model parameter estimates per region

G7 US Eurozone Poor countries

Dependent variables Dependent variables Dependent variables Dependent variables

Independent variables UWR NII PBT UWR NII PBT UWR NII PBT UWR NII PBT

Intercept -0.4343 0.0428 0.2989 -0.5550 0.0444 0.2059 -0.0644 0.0344 0.4682 -0.0650 0.0568 0.1461 (0.000) (0.000) (0.002) (0.000) (0.000) (0.002) (0.000) (0.000) (0.052) (0.041) (0.001) (0.000)

Deviation from average GDP growth 0.0036 -0.0003 0.0030 0.0343 -0.0009 0.0302 0.0028 -0.0003 -0.0001 0.0001 -0.0031 0.0007 (0.002) (0.000) (0.289) (0.500) (0.096) (0.047) (0.000) (0.000) (0.958) (0.716) (0.165) (0.330)

US Inflation 0.1414 0.0037 -0.0245 0.1920 0.0031 -0.0076 0.0078 0.0045 -0.0849 0.0105 0.0117 0.0144 (0.000) (0.000) (0.550) (0.001) (0.035) (0.784) (0.413) (0.000) (0.429) (0.529) (0.068) (0.507)

Total returns on the MSCI World index 0.0000 -0.0001 0.0001 0.0004

(0.516) (0.003) (0.000) (0.073)

Lagged total returns on MSCI World index -0.0011 -0.0033 0.0008 -0.0006

(0.222) (0.290) (0.001) (0.139)

Change in US Treasury 5y yield -0.0011 -0.0010 -0.0015 -0.0074

(0.000) (0.010) (0.001) (0.051)

Lagged change in US Treasury 5y yield -0.0186 -0.0265 0.0034 0.0042

(0.125) (0.142) (0.294) (0.392)

Lagged NII ratio -0.0934 -0.3305 -0.0520 -0.0080

(0.693) (0.563) (0.513) (0.069)

Note: p-values are in parentheses

Appendix B

Figure 1 Developments in the dependent variables over time

-0,3 -0,2 -0,1 0 0,1 0,2 0,3 0,4 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Net investment income ratio Underwriting result ratio Profit before tax ratio

Figure 2 Regional developments in underwriting result ratio

-0,45 -0,4 -0,35 -0,3 -0,25 -0,2 -0,15 -0,1 -0,05 0 0,05 1991 1993 1995 1997 1999 2001 2003 2005 2007

Underwriting result ratio G7 Underwriting result ratio US Underwriting result ratio Eurozone

Figure 3 Regional developments in the net investment income ratio 0 0,02 0,04 0,06 0,08 0,1 0,12 0,14 0,16 1991 1993 1995 1997 1999 2001 2003 2005 2007

Net investment income ratio G7 Net investment income ratio US Net investment income ratio Eurozone

Net investment income Poor countries

Figure 4 Regional developments in the profit before tax ratio

-0,2 0 0,2 0,4 0,6 0,8 1 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Figure 5 Business cycle developments versus developments in the underwriting result ratio in the United States

-2,5 -2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 D e v ia ti o n s fr o m a v e ra g e G D P g ro w th -0,45 -0,4 -0,35 -0,3 -0,25 -0,2 -0,15 -0,1 -0,05 0 U n d e rw ri ti n g r e su lt r a ti o

Deviations from average GDP growth Underwriting result ratio

Figure 6 Business cycle developments versus developments in the net investment income ratio in the United States

-2,5 -2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 D e v ia ti o n s fr o m a v e ra g e G D P g ro w th 0 0,01 0,02 0,03 0,04 0,05 0,06 0,07 N e t in v e st m e n t in c o m e r a ti o

Deviations from average GDP growth Net investment income ratio

Figure 7 Business cycle developments versus developments in the profit before tax ratio in the United States

-2,5 -2,0 -1,5 -1,0 -0,5 0,0 0,5 1,0 1,5 2,0 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 D e v ia ti o n s fr o m a v e ra g e G D P g ro w th 0 0,05 0,1 0,15 0,2 0,25 0,3 P ro fi t b e fo re t a x r a ti o

Deviations from average GDP growth Profit before tax ratio

Figure 8 Business cycle developments versus developments in the underwriting result ratio in Germany -20 -15 -10 -5 0 5 10 15 20 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 D e v ia ti o n s fr o m a v e ra g e G D P g ro w th -0,1 -0,08 -0,06 -0,04 -0,02 0 0,02 0,04 U n d e rw ri ti n g r e su lt r a ti o

Figure 9 Business cycle developments versus developments in the net investment income ratio in Germany -20 -15 -10 -5 0 5 10 15 20 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 D e v ia ti o n s fr o m a v e ra g e G D P g ro w th 0 0,01 0,02 0,03 0,04 0,05 0,06 N e t in v e st m e n t in c o m e r a ti o

Deviations from average GDP growth Net investment income ratio

Figure 10 Business cycle developments versus developments in the profit before tax ratio in Germany -20 -15 -10 -5 0 5 10 15 20 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 D e v ia ti o n s fr o m a v e ra g e G D P g ro w th 0 0,2 0,4 0,6 0,8 1 1,2 1,4 P ro fi t b e fo re t a x r a ti o