ciency methodologies (data envelopment analysis, stochastic frontier ..... approach to measuring cost of equity capital based on AM Best ratings (see Cum- ..... Overall, it appears that developed countries in Asia and Europe achieve higher.
FRONTIER EFFICIENCY METHODOLOGIES TO MEASURE PERFORMANCE IN THE INSURANCE INDUSTRY: OVERVIEW AND NEW EMPIRICAL EVIDENCE Martin Eling, Michael Luhnen∗ JEL Classification: D23, G22, L11 ABSTRACT The purpose of this paper is to provide an overview and new empirical evidence on frontier efficiency measurement in the insurance industry, a topic of great interest in the academic literature during the last several years. In the first step, we review 83 studies and put them into a joint evaluation of efficiency measurement in the field of insurance. In the second step, an efficiency comparison of 3,710 insurers from 37 countries is conducted using a dataset that has not yet been subject to efficiency analysis (the AM Best Non US database). We compare different methodologies, countries, organizational forms, distribution systems, lines of business, and company sizes. We find a steady efficiency growth in international insurance markets from 2002 to 2006 with large differences across countries. Furthermore, we find evidence that stock companies are more efficient than mutuals as well as for economies of scale, but no evidence for economies of scope in the insurance industry. Only minor variations are found when comparing different frontier efficiency methodologies (data envelopment analysis, stochastic frontier analysis) and different distribution types (agency-based vs. direct writers). Our results give valuable insight into the competitiveness of insurers in various countries. (At the end of the paper there is an outline of further research that we would like to integrate in a revised version of the paper to be presented at the conference on “Performance Measurement in the Financial Services Sector.”)
1. INTRODUCTION In recent years, efficiency measurement has captured a great deal of attention. The insurance sector in particular has seen rapid growth in the number of studies applying frontier efficiency methods. Berger/Humphrey (1997) and Cummins/Weiss (2000) surveyed eight and 21 studies, respectively; now, less than ten years after the Cummins/Weiss survey, there are 83 studies on efficiency measurement in the insurance industry. Recent work in the field has refined methodologies, addressed new topics (e.g., market structure and risk management), and extended geographic coverage from a previously US-focused view to a broad set of countries around the world, including emerging markets such as China and Taiwan. ∗
Martin Eling is with the University of Wisconsin, School of Business, 4279 Grainger Hall, 975 University Avenue, Madison, WI 53706, USA. Michael Luhnen is with the University of St. Gallen, Institute of Insurance Economics, Kirchlistrasse 2, 9010 St. Gallen, Switzerland. We are grateful to Mareike Bodderas, Thomas Parnitzke, Hato Schmeiser, and Denis Toplek for valuable suggestions and comments.
The first aim of this paper is to survey these 83 studies. We provide a comprehensive categorization of this rapidly growing body of literature and point out new developments. Frontier efficiency methodologies have been used to analyze many important economic questions, such as comparison of countries (see, e.g., Fenn et al., 2008), organizational forms (see, e.g., Cummins/Weiss/Zi, 1998), and scale economies (see, e.g., Fecher/Perelman/Pestieau, 1991). Other research questions have involved the efficiency effects of mergers and acquisitions (see, e.g., Cummins/Tennyson/Weiss, 1999) and the comparison of different frontier efficiency methodologies (see, e.g., Cummins/Zi, 1998). In our overview, we will review these fields of application, summarize the methodologies, and highlight recent developments in the field. Existing cross-country comparisons of efficiency in the insurance industry provide valuable insights into the competitiveness of insurers in different countries. However, the geographic coverage of these studies is limited to certain countries or regions. Weiss (1991b) compares the United States, Germany, France, Switzerland, and Japan; Donni/ Fecher (1996) analyze 15 OECD countries. Both authors were restricted to using aggregated economic information instead of individual company data. Diacon/Starkey/ O’Brien (2002) and Fenn et al. (2008) do use individual company data, but concentrate on European countries (15 and 14, respectively). Rai (1996) takes a look at nine European countries, Japan, and the United States, but considers a relatively small dataset of 106 companies. What is missing is a broad comparison of efficiency at the international level that incorporates a large number of countries and companies. The second aim of this paper is thus to contribute to the growing body of literature on frontier efficiency at the international level by answering key research questions based on a large number of countries and companies. We therefore consider a broad international dataset that has not yet been the subject of efficiency study—the AM Best Non US database. Our cross-country analysis uses data on 3,710 insurance companies from 37 countries, which gives our study—to our knowledge—the largest sample ever analyzed for the insurance industry. We consider six main aspects: (1) methodologies, (2) countries, (3) organizational forms, (4) distribution systems, (5) lines of business, and (6) company size. These six aspects allow us to address many of the important economic questions set out in the first survey part of the paper. Another important contribution of this paper is that we determine and compare efficiency for a large number of countries that have not been considered, at least to our knowledge, in the literature to date: the Bahamas, Barbados, Bermuda, Brazil, the Czech Republic, Hong Kong, Hungary, Indonesia, Lithuania, Mexico, Norway, Poland, Russia, Singapore, and South Africa. Thus, the main contribution of this paper is to provide a broad evaluation
2
of efficiency measurement in the insurance industry with a special emphasis on an international comparison of efficiency. Our main empirical findings can be summarized as follows: (1) We find a steady efficiency growth in international insurance markets during the sample period (2002–2006), with large efficiency differences between the 37 countries. The highest efficiency is found for Japan; the lowest for the Phillipines. (2) We find evidence for the expense preference hypothesis (i.e., stock companies are more efficient than mutuals) and for economies of scale (i.e., larger companies are more efficient than smaller companies). (3) We find only minor difference between distribution types (agency-based vs. direct writers) and very little evidence for economies of scope (i.e., multi-line insurers are not necessarily more efficient than mono-line insurers). (4) We find that there is not much difference between the two frontier efficiency methodologies data envelopment analysis and stochastic frontier analysis. The remainder of the paper is organized as follows. Section 2 contains an overview of 83 studies on efficiency measurement in the insurance industry. The empirical examination is performed in Section 3. The results of the study are summarized in Section 4. 2. OVERVIEW OF EFFICIENCY MEASUREMENT IN THE INSURANCE INDUSTRY In this section we provide an overview of 83 studies on efficiency measurement in the insurance industry. Our overview builds upon and significantly extends two earlier surveys of efficiency measurement in the financial services industry: one by Berger/Humphrey (1997), which focuses on banks and provides a valuable criteria framework that we use to analyze the studies in the field of insurance, the second by Cummins/Weiss (2000), which focuses on the insurance industry and covers 21 studies that had been published by the year 1998. Table 1 summarizes the main features of the 83 studies. Although many of the studies make contributions to more than one topic, in order to categorize them, we have arranged them according to their primary field of application (first column). The ten application categories were chosen and refined based on the earlier overview by Berger/Humphrey (1997). Columns 2 and 3 show the country(ies) under investigation and the efficient frontier method used; the fourth column lists the authors and year of publication of the study. More detailed information, such as input and output factors, types of efficiency analyzed, sample periods, lines of business covered, and main findings, can be found in the Appendix. In the following we discuss these studies, particularly focusing on recent developments.
3
Table 1: Studies on Efficiency in the Insurance Industry Application
Country Ukraine Deregulation, regulation Korea, Philippines, Taichange wan, Thailand Spain Germany, UK Germany Germany Austria Germany, UK US Distribution systems US US US UK UK Financial and risk manageUS ment, capital utilization US US General level of efficiency and Portugal productivity, evolution of Nigeria efficiency over time Canada Netherlands US Tunisia Italy US France US Taiwan Taiwan UK China Germany China Malaysia Greece China India US Australia China Intercountry comparisons France, Belgium 6 European countries 15 European countries 15 OECD countries Germany, UK Austria, Germany 11 countries Japan, France, Germany, Switzerland, US Market structure US Austria 14 European countries Mergers US US US 8 European countries
Method DEA DEA
Author (date) Badunenko/Grechanyuk/Talavera (2006) Boonyasai/Grace/Skipper (2002)
DEA DEA, DFA DEA DEA DEA DEA DFA DFA DEA DEA SFA SFA DEA SFA DEA DEA DEA Index SFA DEA DFA, SFA DEA SFA DEA, SFA DFA DFA DFA SFA SFA DEA DEA DEA DEA DEA DEA SFA DEA DEA DEA, SFA DEA DEA DEA DEA DEA DFA, SFA Index
Cummins/Rubio-Misas (2006) Hussels/Ward (2006) Mahlberg (2000) Mahlberg/Url (2000) Mahlberg/Url (2003) Rees et al. (1999) Ryan/Schellhorn (2000) Berger/Cummins/Weiss (1997) Brockett et al. (1998) Carr/Cummins/Regan (1999) Klumpes (2004) Ward (2002) Brockett et al. (2004) Cummins et al. (2006) Cummins/Nini (2002) Barros/Barroso/Borges (2005) Barros/Obijiaku (2007) Bernstein (1999) Bikker/van Leuvensteijn (2008) Cummins (1999) Chaffai/Ouertani (2002) Cummins/Turchetti/Weiss (1996) Cummins/Weiss (1993) Fecher et al. (1993) Gardner/Grace (1993) Hao (2007) Hao/Chou (2005) Hardwick (1997) Huang (2007) Kessner/Polborn (1999) Leverty/Lin/Zhou (2004) Mansor/Radam (2000) Noulas et al. (2001) Qiu/Chen (2006) Tone/Sahoo (2005) Weiss (1991a) Worthington/Hurley (2002) Yao/Han/Feng (2007) Delhausse/Fecher/Pestieau (1995) Diacon (2001) Diacon/Starkey/O’Brien (2002) Donni/Fecher (1997) Kessner (2001a) Mahlberg (1999) Rai (1996) Weiss (1991b)
SFA SFA SFA DEA DEA DFA DEA
Choi/Weiss (2005) Ennsfellner/Lewis/Anderson (2004) Fenn et al. (2008) Cummins/Tennyson/Weiss (1999) Cummins/Xie (2008) Kim/Grace (1995) Klumpes (2007)
4
Table 1: Studies on Efficiency in the Insurance Industry (continued) Application Country Method Author (date) Methodology issues, compar- US DEA, DFA, Cummins/Zi (1998) ing different techniques or FDH, SFA assumptions Spain SFA Fuentes/Grifell-Tatjé/Perelman (2001) Japan DEA Fukuyama/Weber (2001) Taiwan DEA Hwang/Kao (2008) US Index Weiss (1986) Canada DEA Wu et al. (2007) Canada DEA Yang (2006) Organizational form, corpoUS DEA Brockett et al. (2005) rate governance issues Spain DEA Cummins/Rubio-Misas/Zi (2004) US DEA Cummins/Weiss/Zi (1999) Germany DEA Diboky/Ubl (2007) Belgium FDH Donni/Hamende (1993) US DEA Erhemjamts/Leverty (2007) Japan DEA Fukuyama (1997) US SFA Greene/Segal (2004) UK DEA Hardwick/Adams/Zou (2004) Japan DEA Jeng/Lai (2005) US DEA Jeng/Lai/McNamara (2007) Scale and scope economies US SFA Berger et al. (2000) US DEA Cummins/Weiss/Zi (2003) France SFA Fecher/Perelman/Pestieau (1991) Spain SFA Fuentes/Grifell-Tatjé/Perelman (2005) Japan SFA Hirao/Inoue (2004) Ireland DFA Hwang/Gao (2005) Germany DEA Kessner (2001b) US DFA Meador/Ryan/Schellhorn (2000) Finland SFA Toivanen (1997) US SFA, TFA Yuengert (1993) Abbreviations: DEA: data envelopment analysis; DFA: distribution-free approach; FDH: free-disposal hull; Index: index numbers; SFA: stochastic frontier approach; TFA: thick frontier approach.
2.1. FRONTIER EFFICIENCY METHODOLOGIES The two main methodologies in efficient frontier analysis are the econometric (parametric) and the mathematical programming (nonparametric) approach, which we discuss only briefly here. For a more detailed overview, the reader is referred to Banker et al. (1989), Greene (1993), Charnes et al. (1994), Berger/Humphrey (1997), Cummins/Weiss (2000), and Cooper/Seiford/Tone (2006). Econometric (parametric) approach The econometric approach specifies a production, cost, revenue, or profit function with a specific shape and makes assumptions about the distributions of the inefficiency and error terms. There are three principal types of econometric frontier approaches. Although they all specify an efficient frontier form (usually translog but, increasingly, alternative forms such as generalized translog, Fourier flexible, or composite cost), they differ in their distributional assumptions of the inefficiency and random components (see Cummins/Weiss, 2000). The stochastic frontier approach (SFA) assumes a composed error model where inefficiencies follow an asymmetric distribution (e.g., half-normal, exponential, or gamma) and the error term follows a symmetric distribu5
tion, usually normal. The distribution-free approach (DFA) makes fewer specific assumptions, but requires several years of data; efficiency of each company is assumed to be stable over time, and the random noise averages out to zero. Finally, the thick frontier approach (TFA) does not make any distributional assumptions for the random error and inefficiency terms, but assumes that inefficiencies differ between the highest and lowest quartile firms (see Berger/Humphrey, 1997). Mathematical programming (nonparametric) approach Compared with the econometric approach, the mathematical programming approach puts significantly less structure on the specification of the efficient frontier and does not decompose the inefficiency and error terms. The most widespread mathematical programming approach is data envelopment analysis (DEA), which uses linear programming to measure the relationship of produced goods and services (outputs) to assigned resources (inputs). As an optimization result, DEA determines an efficiency score, which can be interpreted as a performance measure. DEA was introduced by Charnes/Cooper/Rhodes (1978) and builds on the method suggested by Farell (1957) for computation of technical efficiency. DEA models can be specified under the assumption of constant (CRS) or variable returns to scale (VRS) and can be used to decompose cost efficiency into its single components—technical, pure technical, allocative, and scale efficiency (see Cummins/Weiss, 2000). The free-disposal hull (FDH) approach is a special configuration of DEA. Under this approach, the points on the lines connecting the DEA vertices are excluded from the frontier and the convexity assumption on the efficient frontier is relaxed (see Berger/Humphrey, 1997). Although it is not the focus of our overview, we will briefly mention total factor productivity, since many of the reviewed studies calculate this measure. The Malmquist productivity index is the most common measure of total factor productivity. It is efficient-frontier based (mostly DEA) and able to split total factor productivity growth into a technical efficiency and a technological change component (see Grosskopf, 1993). Another approach to total factor productivity measurement is the calculation of index numbers (IN), which are, however, not frontier based, and have to do with the difference between input and output growth (see Cummins/Weiss, 2000). Both these approaches—econometric and mathematical programming—have their supporters and there is no consensus in the field as to which method is superior (see, e.g., Cummins/Zi, 1998; Hussels/Ward, 2006), most likely because both approaches have their advantages and disadvantages. While the econometric approach uses strong assumptions regarding the form of the efficient frontier, and therefore implies a certain economic behavior, the mathematical programming approach has the advantage of assuming less structure for the frontier; however, it does not take into account a ran6
dom error and therefore runs the risk of mistaking a true random error for inefficiency (see Berger/Humphrey, 1997). Looking at the overview of methods used in frontier efficiency studies as set out in Table 1, it becomes obvious that the nonparametric DEA approach has been most frequently used: out of the 83 surveyed studies, 46 use DEA, 18 SFA, eight DFA, one FDH, and three use index numbers. Additionally, total factor productivity using the Malmquist productivity index has been calculated by 22 studies. For DEA, the most widely used specification has been under the assumption of VRS. Only seven studies follow the advice given by Cummins/Zi (1998) to consider multiple approaches, ideally from both the econometric and mathematical programming sides. However, most of those seven studies find highly correlated results across different approaches (see, e.g., Hussels/Ward, 2006; Fecher et al., 1993; Delhausse/Fecher/Pestieau, 1995). Often, the choice of methods is determined by the available data, e.g., if the available data are known to be noisy, the econometric approach, featuring an error term, may lead to more accurate results than the mathematical programming approach (see Cummins/Weiss, 2000). In recent years, there have been a number of proposals for the improvement of efficient frontier methodology in the insurance field. For the econometric approach, a major direction has been to find more flexible specifications of the functional form, such as the Fourier flexible distribution. This form approximates the real underlying distribution more closely than do other forms, such as the translog, and thus addresses the main drawback of the econometric approach (see, e.g., Berger/Cummins/Weiss, 1997; Fenn et al., 2008; Klumpes, 2004), which is the possible misspecification of the shape of the efficient frontier due to strong distributional assumptions. A further proposal has been made regarding the incorporation of firm-specific variables into the estimation process. Instead of using a two-stage approach, which first estimates inefficiency of sample firms and then examines the association of inefficiency with firm-specific variables (such as organizational form), a one-stage approach is suggested. The estimated frontier directly takes into account firm-specific variables and allows a joint modeling of inefficiency effects (see, e.g., Greene/Segal, 2004; Fenn et al., 2008). Another contribution has been made with regard to the Malmquist index of total factor productivity. Although this index is usually applied to nonparametric DEA for insurance companies, Fuentes/Grifell-Tatjé/Perelman (2001) develop a parametric distance function approach for the Malmquist productivity index calculation. The mathematical programming approach has traditionally been viewed as a strictly nonparametric approach; however, it has been shown that DEA can be interpreted as a maximum likelihood estimator, providing a statistical base to DEA (see, e.g., Banker, 7
1993). Nevertheless, efficiency estimates are biased upward in finite examples and therefore need to be corrected. That is why the bootstrapping procedure proposed by Simar/Wilson (1998) has been applied to and extended for the insurance industry to account for various kinds of efficiency (cost, technical, revenue) as well as CRS and VRS models (see, e.g., Cummins/Weiss/Zi, 2003; Erhemjamts/Leverty, 2007; Diboky/Ubl, 2007). A further innovation is the introduction of cross-frontier efficiency analysis, which estimates efficiency of one particular technology relative to a best practice frontier of an alternative technology (see Cummins/Weiss/Zi, 1999, 2003; Cummins/Rubio-Misas/Zi, 2004). Cross-frontier analysis has been mainly used to examine the efficiency of different organizational forms. Finally, Brocket et al. (2004, 2005) apply a RAM (range-adjusted measure) version of DEA to the insurance industry, addressing theoretical problems regarding the comparison set of each firm for which efficiency has been calculated. 2.2. INPUT AND OUTPUT FACTORS USED IN EFFICIENCY MEASUREMENT An important decision when employing frontier efficiency models involves the choice of inputs, outputs, and their prices, since the definition of these factors can significantly impact the results of a study (see Cummins/Weiss, 2000). Choice of input factors There are three main insurance inputs: labor, business service and materials, and capital. Labor can be further divided into agent and home-office labor. The category of business service and materials is usually not further subdivided, but includes items like travel, communications, and advertising. At least three categories of capital can be distinguished: physical, debt, and equity (see Cummins/Tennyson/Weiss, 1999; Cummins/Weiss, 2000). Since data on the number of employees or hours worked are not publicly available for the insurance industry, to proxy labor and business service input, input quantities are derived by dividing the expenditures for these inputs by publicly available wage variables or price indices, e.g., the US Department of Labor data on average weekly wages for SIC Class 6311 (home-office life insurance labor), in the case of a US study (see Berger/Cummins/Weiss, 1997; Cummins/Zi, 1998). Physical capital is often included in the business service and materials category, but debt and equity capital are important inputs for which adequate cost measures have to be found (see Cummins/Weiss, 2000). One example of how to derive equity cost is the three-tier approach to measuring cost of equity capital based on AM Best ratings (see Cummins/Tennyson/Weiss, 1999). There appears to be widespread agreement among researchers with regard to the choice of inputs (see Appendix): 55 out of 83 studies use at least labor and capital as inputs and most of them also add a third category (miscellaneous, mostly business ser8
vices). Out of those 55 studies, 15 differentiate between agent and nonagent labor. Also, the number of studies differentiating between equity and debt capital is quite low; only ten papers do so. Looking at the remaining 28 contributions that do not follow the standard input categories, it becomes obvious that 15 of them incorporate broader expenditure categories as inputs (e.g., total operating expenses) without decomposing them into quantities and prices (see, e.g., Rees et al., 1999; Mahlberg/Url, 2003). There are a further nine studies that do not cover capital explicitly, i.e., they consider labor only or labor and an additional composite category. Finally, four studies that focus on financial intermediation consider only capital-related inputs (see, e.g., Brocket et al., 1998). The choice of input prices is mainly determined by the data that are publicly available in the countries under investigation. Choice of output factors There are three principal approaches to measuring outputs in the financial services industry: the asset or intermediation approach, the user-cost approach, and the valueadded approach (see Berger/Humphrey, 1992). The intermediation approach views the insurance company as a financial intermediary that manages a reservoir of assets, borrowing funds from policyholders, investing them on capital markets, and paying out claims, taxes, and other costs (see Brocket et al., 1998). The user-cost method differentiates between inputs and outputs based on the net contribution to revenues: if a financial product yields a return that exceeds the opportunity cost of funds or if the financial costs of a liability are less than the opportunity costs, it is deemed a financial input; otherwise, it is considered a financial output (see Hancock, 1985; Cummins/Weiss, 2000). The value-added approach counts outputs as important if they contribute a significant added value based on operating cost allocations (see Berger et al., 2000). Usually, several types of outputs are defined, representing the single lines of business under review. In line with the theoretical literature on insurance pricing, the price of insurance outputs is calculated as the sum of premiums and investment income minus output divided by output (see Cummins/Weiss, 2000). The value-added approach to output measurement is clearly dominant in the insurance industry: 69 out of 83 studies apply this approach (see Appendix). However, there is an intense debate among those using the value-added approach as to whether claims/benefits or premiums are the most appropriate proxy for value added (see, e.g., Cummins/Weiss, 2000). Out of the 69 articles, 32 specify output as either claims/present value of claims (property-liability) or benefits/net incurred benefits (life), following the arguments of Cummins/Weiss (2000). However, 35 studies specify output as premiums/sum insured, most likely because these measures are more readily available for most countries. Two of the 69 studies applying the value-added approach use neither of the two main proxies (e.g., Yuengert (1993) uses re9
serves/additions to reserves to proxy value added). There is no recognizable trend over time as to whether either of the two main proxies is gaining more of a following among researchers. Regarding the other two approaches for output measurement, five studies employ the intermediation approach, e.g., taking ROI, liquid assets to liability, and solvency scores as outputs (see Brockett et al., 2004, 2005). As argued by Cummins/Weiss (2000), this approach is not optimal because insurers provide many services in addition to financial intermediation. None of the studies reviewed uses the user-cost approach, because this approach requires precise data on product revenues and opportunity costs, which are not available in the insurance industry (see Klumpes, 2007). Four studies use both the value-added and intermediation approaches (see, e.g., Jeng/Lai, 2005). One study proposes a new relational two-stage approach to output measurement (see Hwang/Kao, 2008): the authors construct a series relationship between the first output process (premium acquisition), providing the inputs for the second output process (profit generation). Four studies apply physical outputs, e.g., Weiss (1986) and Bernstein (1999) use number of policies as insurance output. 2.3. FIELDS OF APPLICATION IN INSURANCE EFFICIENCY MEASUREMENT AND SELECTED RESULTS Frontier efficiency methods have been applied to a wide range of countries (33 countries according to our survey) as well as to all major lines of business. Furthermore, frontier efficiency methods have been used to investigate various economic issues, including risk management (see, e.g., Cummins et al., 2006), market structure (see, e.g., Choi/Weiss, 2005), organizational forms (see, e.g., Jeng/Lai, 2005), and mergers (see, e.g., Cummins/Xie, 2008). However, it should be noted that findings regarding the same economic issues often vary depending on country, line of business, time horizon, and method considered in the different studies. In the following, we analyze the 83 studies of our survey according to their field of application and selected main results. For this purpose, we consider ten application categories (see Table 1). Deregulation and regulation change There has been significant market deregulation of financial services in many countries (e.g., European Union Third Generation Insurance Directive, 1994). The aim of deregulation in the financial services sector is to improve market efficiency and enhance consumer choice through more competition. Consolidation, in particular, has the potential to improve efficiency by efficient firms taking over less efficient ones and realizing economies of scale (see Cummins/Rubio-Misas, 2006). However, the evidence on efficiency gains due to deregulation have been mixed. Rees et al. (1999) find modest efficiency gains from deregulation for the United Kingdom and Germany for the period from 1992–1994; however, Hussels/Ward (2006) do not find clear evidence for 10
a link between deregulation and efficiency for the same countries during the period 1991–2002. Mahlberg (2000) finds decreasing efficiency for Germany for the period of 1992–1996, but an increase in productivity. For Spain, Cummins/Rubio-Misas (2006) find clear evidence for total factor productivity growth for the period of 1989– 1998, with consolidation reducing the number of firms in the market. Boonyasai/ Grace/Skipper (2002) find evidence for productivity increases in Korea and the Philippines due to deregulation. On the issue of changing regulation in the United States, Ryan/Schellhorn (2000) find unchanged efficiency levels from the start of the 1990s to the middle of that decade, a period during which risk-based capital requirements (RBC) became effective. Distribution systems The effect of different distribution systems on the efficiency of insurance companies has been another field of investigation but in this case the results have been more consistent. Brockett et al. (1998, 2004), studying the United States, and Klumpes (2004), studying the United Kingdom, both find that independent agent distribution systems are more efficient than direct systems involving company representatives or employed agents. However, Berger/Cummins/Weiss (1997) find in their study of the United States that agent systems are less cost efficient, but equally profit efficient, as direct systems due to differences in service intensity that are offset by higher revenues. On a more general level, Ward (2002), in his study of the United Kingdom, finds that insurers focusing on one distribution system are more efficient than those employing more than one mode of distribution. Financial and risk management, capital utilization A relatively new field of application for frontier efficiency methods is that of financial and risk management. Cummins et al. (2006) were the first to explicitly investigate the relationship between risk management, financial intermediation, and economic efficiency of insurance companies. In their application to the US property-liability industry, they find that both activities significantly increase efficiency. In a similar vein, Brockett et al. (2004) find that solvency scores used as an output have only limited impact on efficiency scores and Cummins/Nini (2002) state that capital is used suboptimally in US property-liability insurance. General level of efficiency and productivity, evolution of efficiency over time This category is comprised of all those studies whose primary research objective is to show a country’s levels of efficiency and productivity and, in most cases, also to investigate the evolution of those measures over time. Many of these studies also look at other issues, such as scale or scope economies. Naturally, this category contains a large number of studies that represent a first application of efficiency frontier methods 11
to countries such as Nigeria (see Barros/Obijiaku, 2007), Tunisia (see Chaffai/ Ouertani, 2002), Malaysia (see Mansor/Radam, 2000), or Australia (see Worthington/ Hurley, 2002). Given the broad range of countries and time horizons employed, findings regarding efficiency and productivity are obviously mixed; however, nearly all studies note that there are significant levels of inefficiency in most countries with corresponding room for improvement, e.g., the Netherlands with 75% cost efficiency on average (see Bikker/van Leuvensteijn, 2008), average efficiency of 77% (non-life) and 70% (life) in China (see Yao/Han/Feng, 2007), and average efficiency of 65% for Greece (see Noulas et al., 2001). Intercountry comparisons Cross-country comparisons in the insurance industry can provide valuable insight into the competitiveness and efficiencies of insurers in different countries. This sort of study is especially interesting for the European market, where implementation of the single European Union (EU) insurance license in 1994 raised concerns about international competitiveness among insurers (see Diacon/Starkey/O’Brien, 2002). Consequently, there have been quite a few efficiency studies on this topic. For a sample of 450 companies from 15 European countries and for the period 1996–1999, Diacon/ Starkey/O’Brien (2002) find striking international differences in average efficiency. According to their study, the United Kingdom, Spain, Sweden, and Denmark have the highest levels of technical efficiency. Additionally, in a recent study involving 14 European countries for the period 1995–2001, Fenn et al. (2008) find increasing returns to scale for the majority of EU insurance companies, indicating that mergers and acquisitions, facilitated by the liberalized EU market, have led to efficiency gains. There are also cross-country efficiency studies looking at international country samples, e.g., Weiss (1991b), covering the United States, Germany, Switzerland, France, and Japan; Donni/Fecher (1997), comparing 15 OECD countries; and Rai (1996), analyzing 11 countries. Market structure Several contributions to the efficient frontier literature study the effects of deregulation and consolidation in general (see above), but only very few explicitly analyze the relationship between market structure and performance. In a recent contribution looking at the US property-liability market, Choi/Weiss (2005) introduce the efficiency measure into market structure analysis. They formalize the efficient structure hypothesis, which claims that more efficient firms charge lower prices than their competitors, allowing them to capture larger market shares as well as economic rents, leading to increased market concentration, and incorporate x-efficiency and scale-efficiency into their analysis. Results suggest that regulators should be more concerned with efficiency rather than market power arising from industry consolidation. Further contributions to 12
the topic of market structure, but with a focus on the EU, have been made by Fenn et al. (2008) and Ennsfellner/Lewis/ Anderson (2004). Mergers Another relatively new field for the application of frontier efficiency methods is that of mergers and acquisitions. The few studies that have been done aim at understanding the motivations of merger activity from the angle of efficiency gains. An early study by Kim/Grace (1995) simulating efficiency gains from hypothetical horizontal mergers in the US life insurance industry indicates that most mergers would improve cost efficiencies, with the exception of mergers between large firms. Two other studies looking at the US market are Cummins/Tennyson/Weiss (1999) for life insurance and Cummins/Xie (2008) for property-liability insurance. Both studies conclude that mergers are beneficial for efficiency, for both the acquiring and the target firm. Also, financially vulnerable firms are found more likely to be acquired. In his application to the European insurance market, Klumpes (2007) tests the same hypothesis as did Cummins/Tennyson/Weiss (1999) and Cummins/Xie (2008) and finds that acquiring firms are more likely to be efficient than nonacquiring firms. However, he finds no strong evidence that target firms achieve greater efficiency gains than nontarget firms. He also implies that merger activity in the major European insurance markets seems to be mainly driven by solvency objectives (i.e., financially weak insurers are bought by financially sound companies) and less by value maximization, as found in the US. Methodology issues, comparing different techniques or assumptions There are a few studies whose primary objective is to solve methodological issues or compare different techniques or assumptions over time. Among the most important of these are Cummins/Zi (1998), comparing different frontier efficiency methods (DEA, DFA, FDH, SFA), and Fuentes/Grifell-Tatjé/Perelman (2001), introducing a parametric frontier approach for the application of the Malmquist index. For a discussion of these and other studies having methodological issues as their primary or secondary research objective, the reader is referred to Section 2.1 of this paper, covering different efficiency frontier techniques, and Section 2.2, covering different ways of treating inputs and outputs. Organizational form, corporate governance issues A very important and well-developed field of frontier efficiency analysis deals with the effect of organizational form on performance. The two principal hypotheses in this area are the expense preference hypothesis (see Mester, 1991) and the managerial discretion hypotheses (see Mayers/Smith, 1988). While the expense preference hypothesis states that mutual insurers are less efficient than stock companies due to higher perquisite consumption of mutual managers, the managerial discretion hypotheses pos13
its that the two organizational forms use different technologies and that mutual companies are more efficient in lines of business with relatively low managerial discretion (see Cummins/Weiss, 2000). Frontier efficiency studies have confirmed these hypotheses with the general finding that stock companies are more efficient than mutual companies in most cases (see, e.g., Cummins/Weiss/Zi (1999) and Erhemjamts/Leverty (2007) for the United States and Diboky/Ubl (2007) for Germany). However, in some areas, mutuals have been found to be more efficient. For example, in the study by Cummins/Rubio-Misas/Zi (2004) for Spain smaller mutuals dominate stock companies in the production of mutual output vectors. In another study covering Japan, Fukuyama (1997) rejects the managerial discretion hypothesis, finding that mutual and stock companies possess identical technologies. Other studies investigate efficiency improvements after demutualization (see, e.g., Jeng/Lai/McNamara, 2007) and, looking at corporate governance issues, the relation between cost efficiency and the size of the corporate board of directors (see Hardwick/Adams/Zou, 2004). Scale and scope economies Scale and scope economies are classic areas of research in frontier efficiency literature. So far, scale economies has been the more extensively researched of the two and is particularly important in the context of consolidation and the justification of mergers (see Cummins/Weiss, 2000). Although detailed results vary across studies, depending on countries, methods, and time horizons employed, many contributions have found, on average, evidence for increasing returns to scale, meaning that unit costs of production decline as firm size increases (see, e.g., Hardwick (1997) for UK, Hwang/Gao (2005) for Ireland, Qiu/Chen (2006) for China, and Fecher/Perelman/Pestieau (1991) for France). However, the differentiation between size clusters must be considered to achieve more specific results. For example, Yuengert (1993), in his application to the US life insurance market, finds increasing returns to scale for firms with up to US$15 billion in assets and constant returns to scale for bigger firms. In contrast, Cummins/Zi (1998), for the same market, find increasing returns to scale for firms having up to US$1 billion in assets, and decreasing returns to scale for all others except for a few firms with constant returns to scale. The topic of scope economies is also of importance, since there is an increasing number of cross-industry mergers involving insurers from different lines of business (see Cummins/Weiss, 2000). In general, researchers have found evidence for the existence of economies of scope, meaning that multiproduct/-branch firms are more efficient than specialized firms (see, e.g., Meador/ Ryan/Schellhorn, 2000; Cummins/Weiss/Zi, 2003; Fuentes/Grifell-Tatjé/Perelman, 2005). Again, however, looking at the study results in more detail is revealing. For example, in Berger et al. (2000), a US application, it is shown that profit scope economies are more likely to be realized by larger firms. 14
3. NEW EMPIRICAL EVIDENCE As mentioned above, the geographic coverage of efficiency studies in the insurance industry has to date been limited to certain countries or regions. The contribution of this section is to give new insights into efficiency at the international level by analyzing a large number of countries and insurance companies. We compare different (1) methodologies, (2) countries, (3) organizational forms, (4) distribution systems, (5) lines of business, and (6) company sizes, which allows us to address many of the research questions surveyed in Section 2. In each case, the results are presented at different levels of aggregation, which enables us to identify the pure effect of methodologies, countries, organization, distribution, lines of business, and size on efficiency. Additional cross-sectional insights including a Tobit regression analysis are provided at the end of this section. In our analysis, we determine and compare efficiency for 15 countries that have not been considered in the literature to date: the Bahamas, Barbados, Bermuda, Brazil, the Czech Republic, Hong Kong, Hungary, Indonesia, Lithuania, Mexico, Norway, Poland, Russia, Singapore, and South Africa. 3.1. DATA AND METHODOLOGY Our main data source is the 2007 edition of the AM Best Non US database (Version 2007.3). It contains information on 5,031 life and non-life insurance companies from 98 countries. The database has five years of data, covering the period 2002–2006. Companies were included in our analysis if they had positive values for all the inputs and outputs described in Table 2. This reduces our sample to 4,103 companies from 90 countries. Furthermore, in order to appropriately compare the different countries we require each country to have at least a total of 30 firm years and to have data for each of the five years that we analyze. This reduces our sample to 3,710 companies from 37 countries. The remaining 393 companies from 53 countries were included in the analysis as “other” countries.1/2 1
2
These countries are : Antigua and Barbuda (1 company/3 firm years), Argentina (4/15), Bahrain (4/18), Bolivia (14/37), British Virgin Islands (3/8), Bulgaria (5/14), Cayman Islands (14/57), Chile (50/144), China (8/19), Croatia (4/12), Cyprus (5/17), the Dominican Republic (1/4), Ecuador (40/107), Egypt (6/27), El Salvador (8/17), Estonia (12/47), Greece (3/6), Guernsey (2/6), Iceland (7/21), India (12/52), the Isle of Man (3/9), Israel (10/28), Jamaica (3/12), Jordan (2/6), Kazakhstan (1/5), Kenya (4/14), Kuwait (4/17), Latvia (8/29), Lebanon (1/5), Macau (4/19), Malta (2/8), Monaco (1/1), Montserrat (1/2), Nigeria (2/9), the Northern Mariana Islands (1/4), Oman (3/8), Pakistan (4/14), Panama (3/13), Peru (9/27), Qatar (3/14), Romania (3/6), Saudi Arabia (1/5), Slovakia (10/23), Slovenia (4/15), South Korea (9/45), Tanzania (5/16), Thailand (17/45), Trinidad and Tobago (6/27), Tunisia (2/10), the Ukraine (3/9), the United Arab Emirates (3/7), Uruguay (12/34), and Venezuela (46/192). Unfortunately, for many South American countries (e.g., Venezuela), there are no data available for 2006, which is why we excluded these from the country-specific analysis even though the number of companies is relatively large. However, detailed results for all these countries are available upon request. We did not include Canadian insurers, which are covered in the database, but presented in a separate section with different data fields.
15
As discussed above, there is widespread agreement in literature with regard to the choice of inputs. We thus use labor, business services and material, debt capital, and equity capital as inputs. Due to data availability, it was necessary to simplify this scheme by combining labor and business services; only operating expenses (including commissions) are available in the AM Best data. This simplification is common in many other international comparisons (see Diacon/Starkey/O’Brien, 2002; Fenn et al., 2008) for much the same reason it is made here. Furthermore, Ennsfellner/Lewis/Anderson (2004) argue that the operating expenses should be treated as a single input in order to reduce the number of parameters that will need to be estimated. We thus use operating expenses to proxy both labor and business services and handle these as a single input in the following analysis. Cummins/Weiss (2000) showed in their analysis of operating expenses in the US insurance market that these are mostly labor related, e.g., in both life and non-life insurance, the largest expenses are employee salaries and commissions. We therefore concentrate on labor to determine the price of the operating-expenses-related input factor. The price of labor is determined using the ILO October Inquiry, a worldwide survey of wages and hours of work published by the International Labour Organization (ILO; see http://laborsta.ilo.org/) and used in a variety of efficiency applications (see, e.g., Fenn et al., 2008). The price of debt capital is determined using country-specific oneyear treasury bill rates for each year of the sample period; the price of equity capital is determined using the yearly rate of total return of the country-specific MSCI stock market indices (all data were obtained from the Datastream database; see Cummins/ Rubio-Misas (2006) for a comparable selection and a discussion on selection depending on the insurer’s capital structure and portfolio risk). To ensure that all monetary values are directly comparable, we deflate each year’s value by the consumer price index to the base year 2002 (see Weiss, 1991b; Cummins/Zi, 1998). Country-specific consumer price indices were obtained from the International Labour Organization. As is usual in most studies on efficiency in the insurance industry, we use the valueadded approach to determine the outputs. We thus distinguish between the three main services provided by insurance companies—risk-pooling/-bearing, financial services, and intermediation. According to Yuengert (1993), a good proxy for the amount of risk-pooling/-bearing and financial services is the value of real incurred losses, defined as current losses paid plus additions to reserves. As different types of services are provided by life and non-life insurance firms, we need separate output measures for each type of firm (see Choi/Weiss, 2005). We use the present value of claims plus additions to reserves as a proxy for the output volume for non-life insurance and the present value of net incurred benefits plus additions to reserves for life insurance. As done, for example, in Berger/Cummins/Weiss (1997), the price of risk-pooling/-bearing and 16
financial services is calculated as net premiums minus the volume proxy, expressed as a ratio to the volume proxy. The output variable, which proxies the intermediation function, is the real value of invested assets. To obtain present values we again deflate each year’s value using the consumer price index. The price of the intermediation output is the expected rate of return on assets (see Berger/Cummins/Weiss, 1997); we therefore again use the yearly rate of total return of the country-specific MSCI stock market indices. Panel A of Table 2 presents an overview of the inputs and outputs used in this analysis. Each input and output is calculated using a volume proxy that we obtained from AM Best. This volume proxy is then divided by a proxy for price. Panel B of Table 2 contains summary statistics on the variables employed. All numbers were converted into US dollars for comparative purposes using the exchange rates published in the AM Best database. Table 2: Inputs and Outputs Panel A: Overview Inputs Labor and business service Debt capital Equity capital Outputs Non-life losses Life benefits
Proxy for volume AM Best operating expenses AM Best total liabilities AM Best capital & surplus Proxy for volume AM Best claims + addition to reserves
AM Best net incurred benefits + addition to reserves Assets AM Best total investments Panel B: Summary statistics for variables used Variable Unit Mean AM Best operating expenses Million $ 220.20 AM Best total liabilities Million $ 6473.90 AM Best capital & surplus Million $ 517.26 AM Best losses + changes in reserves Million $ 999.47 AM Best benefits paid + changes in reserves Million $ 2802.30 AM Best total investments Million $ 5212.96 ILO consumer price index % 3.78 ILO October Inquiry wages per year % 5.11 Long-term government bonds rates % 12.97 MSCI stock market indices returns $ 29284 (Net premium income - Proxy for non-life $ 1.32 volume) / Proxy for non-life volume (Net premium income - Proxy for life volume) / $ 0.33 Proxy for life volume Labor and business service Quantity 4.15 Debt capital Quantity 1213.42 Equity capital Quantity 299.27 Non-life losses Quantity 306.34 Life benefits Quantity 1194.79 Assets Quantity 1213.42
Proxy for price ILO October Inquiry wages per year Long-term government bonds rates MSCI stock market indices returns Proxy for price (Net premium income - Proxy for volume) / Proxy for volume (Net premium income - Proxy for volume) / Proxy for volume MSCI stock market indices returns St. Dev.
Min
Max.
1174.09 44765.36 2530.88 6444.17 10974.29 27563.82 5.29 5.65 23.73 23965 17.76
0.00 0.00 0.00 0.00 0.00 0.00 -3.07 0.00 -58.07 214 -1.00
26065.20 1568570.53 82010.08 270776.30 279073.87 792591.93 44.96 57.96 101.09 120872 999.00
8.49
-1.00
499.00
32.74 10647.83 1365.88 1659.82 11053.11 10647.83
0.00 0.00 0.00 0.00 0.00 0.00
867.79 372071.22 29831.63 47414.65 393291.49 372071.22
17
In the next section, we analyze two methodologies (data envelopment analysis, stochastic frontier analysis), 37 countries (see Table 3 for listing), three organizational forms (stocks, mutual, other), three distributional forms (agency based, direct writers, both), three lines of business (life, non-life, groups), and three company sizes (large, medium, small). For data envelopment analysis, we use Deap, Version 2.1 (see Coelli, 1996a) and calculate efficiency values assuming input orientation and variable returns to scale. For stochastic frontier analysis, we calculate a translog production frontier using Frontier, Version 4.1 (see Coelli, 1996b), assuming the inefficiencies to follow a half-normal distribution. Company-specific information on domiciliary country, organization type, distribution type, and lines of business is extracted from the AM Best database. Total assets is a widespread measure of insurer size (see, e.g., Cummins/Zi, 1998; Diacon/Starkey/O’Brien, 2002). For comparison of different company sizes, we thus subdivide all companies by their total assets into large (total assets larger than $816 million), medium, and small (total assets smaller than $64 million) insurers. 3.2. RESULTS Data envelopment analysis The results of the data envelopment analysis are set out in Table 3, presented at different levels of aggregation so as to focus on different aspects of efficiency. The first focus is on countries (Panel A), the second on organization (Panel B), the third on distribution (Panel C), the fourth on lines of business (Panel D), and the fifth on size (Panel E). For comparison purposes, the average values are presented in the last line of the table. Altogether our analysis covers 4,103 insurers, with a total of 15,306 firm years. The last line of Table 3 shows that across all 4,103 companies efficiency increased by 17.08% from 2002 to 2006. There is a steady growth in efficiency from 0.69 in 2002 to 0.75 in 2006. However, large differences can be found between countries (see Panel A). The country with the highest efficiency is Japan (average efficiency 0.89), followed by Denmark (0.87) and Switzerland (0.83). The lowest efficiency values are found for the Philippines (average efficiency 0.51), the Bahamas (0.57), and Ireland (0.60). Overall, it appears that developed countries in Asia and Europe achieve higher efficiency scores than do emerging market countries. The efficiency of the largest economies under evaluation is in the upper-middle field: Germany is in 13th place (average efficiency 0.77); France is in 15th place (0.76). Relatively low efficiency values were found for the United Kingdom, which is in 28th place (0.69). Considering the development of efficiency over the sample period, the highest efficiency gains were obtained in Sweden (+50.12%) and Turkey (+47.07%). Austria (–1.76%) and the Czech Republic (–3.06%) are the only countries that did not increase efficiency from 2002 to 2006. 18
Table 3: Results of the Data Envelopment Analysis Year
No. of No. of firm 2002 2003 2004 2005 2006 insurers years Panel A: Comparison of countries Australia 107 419 0.64 0.66 0.73 0.72 0.72 Austria 48 219 0.81 0.80 0.82 0.82 0.78 Bahamas 10 43 0.54 0.50 0.56 0.61 0.68 Barbados 10 41 0.65 0.70 0.71 0.67 0.77 Belgium 108 421 0.70 0.73 0.81 0.78 0.84 Bermuda 94 333 0.69 0.72 0.73 0.74 0.76 Brazil 61 200 0.69 0.72 0.75 0.78 0.79 Czech Repub. 20 84 0.79 0.71 0.82 0.81 0.78 Denmark 181 706 0.79 0.85 0.90 0.91 0.96 Finland 62 266 0.74 0.78 0.84 0.89 0.90 France 241 923 0.71 0.74 0.77 0.78 0.81 Germany 561 2323 0.66 0.77 0.79 0.81 0.85 Hong Kong 22 84 0.66 0.72 0.75 0.70 0.83 Hungary 10 35 0.80 0.78 0.83 0.89 0.92 Indonesia 18 49 0.70 0.68 0.77 0.77 0.81 Ireland 146 481 0.51 0.61 0.66 0.65 0.58 Italy 163 651 0.75 0.76 0.80 0.80 0.83 Japan 58 282 0.84 0.86 0.90 0.94 0.93 Lithuania 28 115 0.59 0.58 0.66 0.66 0.73 Luxembourg 24 96 0.77 0.81 0.82 0.83 0.84 Malaysia 43 172 0.77 0.74 0.75 0.72 0.83 Mexico 76 225 0.70 0.65 0.73 0.66 0.73 Netherlands 289 1105 0.67 0.73 0.76 0.78 0.82 New Zealand 32 124 0.62 0.61 0.68 0.62 0.70 Norway 55 209 0.70 0.79 0.85 0.83 0.87 Other 393 1309 0.66 0.69 0.71 0.71 0.78 Philippines 20 56 0.45 0.55 0.52 0.56 0.53 Poland 19 74 0.65 0.66 0.75 0.77 0.83 Portugal 31 127 0.76 0.78 0.83 0.84 0.88 Russia 37 143 0.69 0.63 0.59 0.68 0.75 Singapore 16 58 0.71 0.75 0.78 0.73 0.80 South Africa 38 139 0.67 0.61 0.71 0.75 0.80 Spain 348 1241 0.77 0.82 0.83 0.81 0.82 Sweden 118 417 0.57 0.75 0.80 0.80 0.86 Switzerland 154 493 0.82 0.81 0.83 0.85 0.83 Taiwan 14 63 0.71 0.80 0.72 0.75 0.80 Turkey 17 54 0.51 0.66 0.69 0.80 0.75 UK 431 1526 0.65 0.69 0.70 0.69 0.74 Panel B: Comparison of organizational types Stocks 3415 12657 0.74 0.81 0.83 0.82 0.87 Mutual 664 2585 0.77 0.84 0.61 0.67 0.70 Other 24 64 0.68 0.73 0.76 0.77 0.80 Panel C: Comparison of distributional systems Agency based 205 850 0.71 0.72 0.75 0.74 0.77 Direct writers 80 332 0.69 0.73 0.73 0.74 0.78 Both 3818 14124 0.69 0.74 0.77 0.78 0.81 Panel D: Comparison of lines of business Life 1212 4484 0.78 0.82 0.89 0.90 0.92 Non-life 2353 8440 0.61 0.68 0.69 0.69 0.74 Groups 538 2382 0.84 0.82 0.83 0.84 0.85 Panel E: Comparison of company size Large 1341 5102 0.80 0.83 0.85 0.86 0.86 Medium 1603 5102 0.67 0.72 0.74 0.75 0.77 Small 1654 5102 0.62 0.69 0.72 0.71 0.77 Total 4103 15306 0.69 0.74 0.77 0.78 0.81 *: Some companies changed size during the sample period, which is why the sum of insurers is larger than the total of 4,103 insurers.
Average
Growth 2002 to 2006 (%)
0.69 0.81 0.57 0.70 0.77 0.72 0.73 0.78 0.87 0.82 0.76 0.77 0.72 0.83 0.71 0.60 0.78 0.89 0.64 0.81 0.75 0.69 0.74 0.64 0.80 0.70 0.51 0.71 0.81 0.66 0.75 0.69 0.81 0.75 0.83 0.76 0.64 0.69
12.93 -3.06 24.65 18.28 19.22 9.59 14.03 -1.76 20.67 22.33 14.76 29.44 25.11 14.29 16.77 13.87 10.68 10.54 24.54 9.75 7.09 4.75 22.39 12.08 24.69 18.95 17.99 28.41 15.11 8.43 12.83 18.63 7.18 50.12 1.08 13.63 47.07 12.95
0.81 0.68 0.74
17.00 -8.83 17.34
0.74 0.73 0.75
7.64 12.70 17.95
0.85 0.67 0.83
16.93 21.64 1.59
0.84 7.38 0.72 13.91 0.69 23.06 0.75 17.08 large, medium, and small
19
The second focus of our analysis concerns different organizational forms and their effects on efficiency (see Panel B of Table 3). We find evidence for the expense preference hypothesis, as the average efficiency values of stock companies (0.81) are much higher than those of mutual insurers (0.68). This finding is in line with most research in the field, e.g., Cummins/Weiss/Zi (1999) and Erhemjamts/Leverty (2007). Other legal forms (e.g., public companies) play only a very minor role in our sample. While there is a steady growth in efficiency over time for stocks and other companies, we find large variations and a decreasing efficiency for mutuals over time, an effect that might be due to mergers, acquisitions, and demutualization occurring during the period of investigation. Panel C of Table 3 shows the results for different distribution systems. The DEA results do not confirm Ward’s (2002) finding that insurers focused on one distribution system are more efficient than those using more than one system. Indeed, in our study, the highest efficiency values are found for insurers using different distribution systems (average efficiency 0.75), followed by agency based (0.74) and direct writers (0.73). However, the differences are minor. Furthermore, note that the number of insurers that identify as pure direct writer or agency based is relatively small in our sample (but still large in absolute terms compared to other studies covering this aspect; e.g., Ward, 2002 covers 44 companies and Klumpes, 2004, 90 companies). Interesting differences can be found when comparing different lines of business (see Panel D of Table 3), a field that has not been in focus of literature to date. On a very aggregate level (detailed results on less aggregated level are available upon request), we compare life insurers, non-life insurers, and groups (i.e., companies offering both life and non-life insurance). We find that average efficiency in life insurance is much higher than in non-life insurance (0.85 vs. 0.67). This effect might be connected with the size effects and economies of scale (see next paragraph) as, based on total assets, the average life insurer is six times larger than the average non-life insurer. We find no evidence for economies of scope; the multi-product/-branch firms are not more efficient than specialized firms. In Panel E of Table 3 the total sample is subdivided by total assets into three size groups—large, medium, and small insurers. In agreement with most research on economies of scale, we find that large companies have much higher efficiency than small companies. Average efficiency for large companies is 0.84, while it is only 0.72 for medium companies, and 0.69 for small companies. However, smaller companies have the most rapid gains in efficiency: from 2002 to 2006, efficiency increased for
20
large insurers by 7.38%, by 13.91% for medium insurers, and by 23.06% for small insurers. Stochastic frontier analysis Table 4 is structured like Table 3 and shows the results of the stochastic frontier analysis. Again, average efficiency is presented for different countries (Panel A), organization types (Panel B), distribution systems (Panel C), lines of business (Panel D), and company sizes (Panel E). The last line of the table shows the efficiency values of the full sample. Below, we briefly highlight the main results and the main differences compared to the DEA (Table 3). One obvious difference between the results in Tables 3 and 4 is that the efficiency values are lower and more variable under the stochastic frontier analysis. The average efficiency across all 4,103 companies is only 0.36 compared to 0.75 with the data envelopment analysis. Nevertheless, the results of the two types of analysis are very similar, leading to identical economic insights. Under the SFA, efficiency increased by 30.20% from 2002 to 2006 and there is a growth in efficiency from 0.30 in 2002 to 0.38 in 2006. Large differences between countries are also found when employing stochastic frontier analysis. The country with the highest efficiency is again Japan (average efficiency 0.62) and the country with the lowest efficiency is, again, the Philippines (0.18). We again confirm the expense preference hypothesis: the average efficiency value of stock companies (0.36) is higher than that of mutual insurers (0.33). Again, efficiency in life insurance is on average much higher than in non-life insurance (0.48 vs. 0.27). Efficiency for the multi-line groups (0.43) is lower than for the mono-line life insurers; we thus, once more, find no evidence for economies of scope. Finally, we find that efficiency for large companies (0.61 on average) is much higher than for medium-sized (0.28) and small companies (0.19). All these observations confirm the results of the data envelopment analysis. One difference between the results in Tables 4 and 3 has to do with distribution systems. The results of the SFA confirm Ward’s (2002) result that insurers focusing on one distribution system have higher efficiency than those using more than one system. Efficiency of agency-based insurers (0.51 on average) is again higher than that of direct writers (0.45), but the individual efficiency of these two groups is now much higher than the efficiency of those insurers using both (0.35). SFA thus provides more clear results than DEA in this case. The finding that agency-based insurers are more efficient than direct writers is in agreement with the findings of Brockett et al. (1998, 2004) for the United States and the findings of Klumpes (2004) for the United Kingdom. 21
Table 4: Results of the Stochastic Frontier Analysis Year
No. of No. of firm 2002 2003 2004 2005 2006 insurers years Panel A: Comparison of countries Australia 107 419 0.34 0.52 0.35 0.44 0.34 Austria 48 219 0.29 0.42 0.33 0.50 0.29 Bahamas 10 43 0.16 0.40 0.14 0.23 0.30 Barbados 10 41 0.27 0.26 0.29 0.33 0.22 Belgium 108 421 0.28 0.40 0.28 0.38 0.32 Bermuda 94 333 0.32 0.35 0.37 0.45 0.30 Brazil 61 200 0.19 0.37 0.17 0.32 0.31 Czech Repub. 20 84 0.18 0.29 0.19 0.37 0.21 Denmark 181 706 0.24 0.37 0.24 0.34 0.40 Finland 62 266 0.37 0.59 0.37 0.47 0.51 France 241 923 0.42 0.54 0.43 0.51 0.56 Germany 561 2323 0.35 0.52 0.32 0.42 0.45 Hong Kong 22 84 0.21 0.21 0.17 0.29 0.19 Hungary 10 35 0.18 0.31 0.20 0.41 0.15 Indonesia 18 49 0.23 0.19 0.20 0.39 0.14 Ireland 146 481 0.27 0.40 0.25 0.35 0.28 Italy 163 651 0.39 0.55 0.42 0.51 0.49 Japan 58 282 0.56 0.73 0.54 0.66 0.60 Lithuania 28 115 0.15 0.20 0.14 0.27 0.18 Luxembourg 24 96 0.27 0.45 0.27 0.36 0.29 Malaysia 43 172 0.24 0.33 0.20 0.31 0.22 Mexico 76 225 0.18 0.33 0.17 0.26 0.34 Netherlands 289 1105 0.32 0.41 0.30 0.38 0.34 New Zealand 32 124 0.20 0.29 0.18 0.26 0.22 Norway 55 209 0.32 0.34 0.33 0.38 0.40 Other 393 1309 0.20 0.32 0.20 0.38 0.27 Philippines 20 56 0.18 0.22 0.14 0.24 0.15 Poland 19 74 0.29 0.43 0.27 0.33 0.29 Portugal 31 127 0.27 0.45 0.28 0.38 0.28 Russia 37 143 0.18 0.25 0.18 0.33 0.15 Singapore 16 58 0.24 0.20 0.20 0.32 0.24 South Africa 38 139 0.26 0.44 0.30 0.40 0.50 Spain 348 1241 0.23 0.34 0.22 0.33 0.33 Sweden 118 417 0.31 0.39 0.31 0.41 0.41 Switzerland 154 493 0.37 0.41 0.32 0.41 0.54 Taiwan 14 63 0.36 0.40 0.33 0.43 0.32 Turkey 17 54 0.19 0.25 0.22 0.33 0.22 UK 431 1526 0.27 0.49 0.34 0.45 0.39 Panel B: Comparison of organizational types Stocks 3415 12657 0.30 0.43 0.30 0.42 0.39 Mutual 664 2585 0.29 0.40 0.27 0.36 0.38 Other 24 64 0.17 0.27 0.35 0.43 0.31 Panel C: Comparison of distributional systems Agency based 205 850 0.48 0.54 0.48 0.56 0.49 Direct writers 80 332 0.44 0.50 0.41 0.50 0.39 Both 3818 14124 0.28 0.42 0.28 0.39 0.38 Panel D: Comparison of lines of business Life 1212 4484 0.31 0.73 0.30 0.49 0.61 Non-life 2353 8440 0.28 0.25 0.27 0.32 0.26 Groups 538 2382 0.34 0.47 0.39 0.58 0.38 Panel E: Comparison of company size* Large 1341 5102 0.52 0.73 0.53 0.65 0.60 Medium 1603 5102 0.24 0.37 0.22 0.31 0.23 Small 1654 5102 0.17 0.23 0.14 0.23 0.18 Total 4103 15306 0.30 0.43 0.30 0.41 0.38 *: Some companies changed size during the sample period, which is why the sum of insurers is larger than the total of 4,103 insurers.
Average
Growth 2002 to 2006 (%)
0.40 0.37 0.24 0.27 0.33 0.36 0.26 0.25 0.31 0.46 0.48 0.41 0.21 0.26 0.22 0.31 0.47 0.62 0.19 0.33 0.27 0.24 0.35 0.23 0.35 0.26 0.18 0.33 0.34 0.22 0.24 0.36 0.28 0.36 0.38 0.37 0.23 0.39
0.15 -0.62 84.41 -21.45 12.78 -5.00 64.50 18.48 67.20 38.38 31.97 28.39 -7.85 -19.89 -40.83 4.83 25.84 6.01 17.46 9.14 -7.18 83.77 6.48 12.29 26.99 35.72 -14.53 -0.80 4.33 -17.35 -2.22 89.25 44.23 35.28 47.40 -11.79 12.14 44.40
0.36 0.33 0.34
30.58 29.79 80.14
0.51 0.45 0.35
0.82 -12.87 33.41
0.48 0.27 0.43
96.20 -5.69 11.07
0.61 14.38 0.28 -3.06 0.19 4.92 0.36 30.20 large, medium, and small
22
Further cross-sectional evidence and Tobit regression analysis In this section we provide a cross-sectional analysis of the efficiency results presented in Tables 3 and 4. Additional insights are provided by a Tobit regression analysis. The groups shown in Table 5 are the result of combining the five comparison criteria from the previous tables (countries, organization, distribution, line of business, size). We ranked these groups according to their average efficiency score. To provide a sound comparison, we display only the groups that have a minimum of ten firm years, resulting in a total of 261 groups. Table 5: Results of Cross-Sectional Analysis Rank Country Organization Distribution Line of business Panel A: Based on efficiency values from data envelopment analysis 1 Japan Mutual Both Life 2 Denmark Stock Both Group 3 Japan Mutual Agency Life 4 Denmark Stock Both Life 5 Denmark Mutual Both Life 6 Norway Stock Both Life 7 Japan Stock Agency Life 8 Brazil Stock Both Life 9 Japan Stock Both Non-Life 10 Spain Mutual Both Group ... … … … … 261 Ireland Stock Agency Non-Life Panel B: Based on efficiency values from stochastic frontier analysis 1 Japan Mutual Both Life 2 Japan Mutual Agency Life 3 Netherlands Stock Agency Group 4 Japan Stock Agency Non-Life 5 France Stock Agency Group 6 Japan Stock Agency Life 7 Italy Stock Direct Group 8 Other Stock Agency Group 9 Other Stock Both Non-Life 10 Norway Mutual Both Non-Life ... … … … … 261 Philippines Stock Both Non-Life
Size
Firm years
Efficiency
Large Medium Large Large Large Large Large Large Large Small … Medium
15 11 20 37 86 25 14 14 38 23 … 12
0.99 0.98 0.97 0.97 0.97 0.97 0.97 0.97 0.96 0.96 … 0.32
Large Large Large Large Large Large Large Large Large Large … Small
15 20 10 25 29 14
0.91 0.91 0.89 0.89 0.87 0.86 0.85 0.82 0.82 0.82 … 0.16
12 11 10 … 35
Panel A of Table 5 shows the results for the efficiency scores based on data envelopment analysis. The best companies are large, mutual Japanese life insurers that use both agents and direct writers (average efficiency score 0.99). The best ten groups are dominated by companies from Japan (4 out of the top 10), by stocks (6 of 10), by companies using both agency based and direct writers (8 of 10), companies operating in life insurance (7 of 10), and by large companies (8 of 10). The worst group consists of medium-sized Irish non-life insurers, which have an efficiency of 0.32.
23
In Panel B of Table 5, the results for the efficiency scores based on stochastic frontier analysis are presented. The best group in Panel A (large, mutual Japanese life insurers using both agents and direct writers) is also the best group in Panel B, the second best group with the DEA (large, mutual Japanese life insurers using agents) is in third place with SFA. We again find mostly Japanese companies among the best ten groups (4 of 10). In contrast to the DEA, there are no Danish companies among the best 10, but, instead, insurers from the Netherlands, France, Italy, and Norway. The worst group with SFA are small non-life insurers from the Philippines, which have an efficiency of 0.16. In summary, it appears that the most efficient companies are large life insurers from Japan, operating as either stock or mutuals. Finally, we conduct a Tobit regression analysis in order to identify the key drivers of efficiency with respect to different independent variables. As the dependent variable (i.e., the efficiency value) is limited between zero and one, standard regression coefficients will be underestimated and thus need to be corrected. This correction is provided by Tobin’s (1958) modified form of a standard regression analysis for the investigation of limited dependent variables. We determine Tobit regression parameters using the statistic software R. We use the following independent, explanatory variables for our regression on the efficiency scores (see Diacon/Starkey/O’Brien (2002) for a comparable selection): (1) organization (1 if company is stock; 0 otherwise), (2) distribution (1 if both agency based and direct writers; 0 otherwise), (3) line of business (1 if life/group; 0 if non-life), (4) company size (natural logarithm of the total assets), and (5) squared size (to capture nonlinearities between size and efficiency). Country and year dummy variables are included to pick up country and time effects. The Tobit regression results for the efficiency values based on data envelopment analysis are presented in Panel A of Table 6; the results based on stochastic frontier analysis in Panel B of Table 6. Considering Panel A, the results of the Tobit regression analysis confirm that organization and line of business are particularly and strongly associated with efficiency; these have the (absolute) highest t-statistics and lowest p-values. Size and squared size are also important for determining efficiency. The country dummies confirm the international differences previously identified. The results confirm the above findings that the most efficient companies are insurers from Japan. Considering Panel B, the results again show the importance of company size for efficiency. Surprisingly, though, and in contrast to Panel A, organization and distribution do not significantly influence efficiency. Line of business is again positively connected with efficiency.
24
Table 6: Results of the Tobit Regression Analysis Panel A: Panel B: Based on efficiency values from Based on efficiency values from data envelopment analysis (Adjusted R2=0.24) stochastic frontier analysis (Adjusted R2=0.24) Coefficient St. Error t-Statistic p-Value Coefficient St. Error t-Statistic p-Value Intercept 0.86 0.03 33.61 0.00 0.79 0.04 20.65 0.00 Organization -0.05 0.00 -13.69 0.00 0.00 0.01 0.32 0.75 Distribution 0.03 0.01 4.88 0.00 0.00 0.01 -0.22 0.83 Line of business 0.15 0.00 48.33 0.00 0.02 0.01 5.27 0.00 Size -0.04 0.00 -10.32 0.00 -0.13 0.01 -22.27 0.00 Squared size 0.00 0.00 13.09 0.00 0.01 0.00 34.25 0.00 Australia -0.03 0.01 -3.38 0.00 0.01 0.01 0.67 0.50 Austria 0.01 0.01 0.56 0.58 -0.08 0.02 -4.43 0.00 Bahamas -0.15 0.03 -5.82 0.00 0.01 0.04 0.35 0.73 Barbados 0.05 0.03 1.83 0.07 0.00 0.04 0.07 0.94 Belgium 0.02 0.01 1.82 0.07 -0.03 0.01 -2.10 0.04 Bermuda 0.03 0.01 3.46 0.00 -0.01 0.02 -0.61 0.54 Brazil 0.02 0.01 1.62 0.11 0.01 0.02 0.29 0.77 Czech Repub. 0.02 0.02 1.10 0.27 -0.05 0.03 -1.91 0.06 Denmark 0.12 0.01 16.07 0.00 -0.03 0.01 -2.81 0.01 Finland 0.03 0.01 2.57 0.01 0.01 0.02 0.62 0.54 France -0.01 0.01 -1.51 0.13 0.00 0.01 -0.10 0.92 Hong Kong 0.05 0.02 2.58 0.01 -0.06 0.03 -2.09 0.04 Hungary 0.04 0.03 1.30 0.20 -0.05 0.04 -1.23 0.22 Indonesia 0.06 0.02 2.33 0.02 -0.02 0.04 -0.60 0.55 Ireland -0.11 0.01 -12.62 0.00 -0.02 0.01 -1.29 0.20 Italy 0.01 0.01 1.98 0.05 0.00 0.01 0.03 0.98 Japan 0.08 0.01 7.96 0.00 0.00 0.02 -0.06 0.95 Lithuania -0.07 0.02 -4.32 0.00 -0.01 0.02 -0.26 0.80 Luxembourg 0.07 0.02 3.77 0.00 -0.04 0.03 -1.52 0.13 Malaysia 0.05 0.01 3.87 0.00 -0.05 0.02 -2.39 0.02 Mexico -0.04 0.01 -3.54 0.00 -0.02 0.02 -1.16 0.25 Netherlands 0.01 0.01 1.84 0.07 0.00 0.01 -0.05 0.96 New Zealand -0.04 0.02 -2.41 0.02 -0.01 0.02 -0.48 0.63 Norway 0.09 0.01 7.42 0.00 0.00 0.02 -0.27 0.79 Other -0.15 0.02 -6.59 0.00 -0.03 0.03 -0.85 0.39 Philippines 0.00 0.02 -0.14 0.89 0.01 0.03 0.51 0.61 Poland 0.07 0.02 4.45 0.00 -0.02 0.02 -1.07 0.28 Portugal -0.08 0.01 -5.50 0.00 -0.03 0.02 -1.17 0.24 Russia 0.07 0.02 3.08 0.00 -0.07 0.03 -2.26 0.02 Singapore -0.04 0.02 -2.76 0.01 0.01 0.02 0.53 0.60 South Africa 0.07 0.01 12.00 0.00 -0.02 0.01 -2.29 0.02 Spain 0.01 0.01 1.38 0.17 -0.03 0.01 -2.18 0.03 Sweden 0.12 0.01 14.05 0.00 -0.01 0.01 -0.53 0.59 Switzerland 0.04 0.02 1.79 0.07 -0.05 0.03 -1.73 0.08 Taiwan -0.05 0.02 -2.26 0.02 -0.03 0.03 -1.01 0.31 Turkey -0.05 0.01 -9.16 0.00 -0.03 0.01 -3.93 0.00 Un. Kingdom -0.03 0.01 -4.61 0.00 -0.01 0.01 -1.06 0.29 2006 0.07 0.01 13.36 0.00 -0.12 0.01 -15.84 0.00 2005 0.03 0.00 8.19 0.00 -0.05 0.01 -7.42 0.00 2004 0.03 0.00 7.23 0.00 -0.15 0.01 -24.99 0.00 2002 -0.04 0.00 -11.24 0.00 -0.12 0.01 -21.10 0.00 Note: The country Germany and the year 2003 were excluded from the analysis due to collinearity.
25
4. CONCLUSION The purpose of this paper was to review recent studies on efficiency in the insurance industry and to provide new empirical evidence on this widely discussed topic. Our overview extends two earlier surveys, one by Berger/Humphrey (1997) and the other by Cummins/Weiss (2000), and analyzes 83 studies on efficiency measurement in the insurance industry. Our results show that during the last several years, methodologies have been refined, new topics have been addressed, and geographic coverage has been extended beyond a US focused view to a broad set of countries. The large number of studies illustrates the increasing interest in the international competitiveness and efficiency of insurance companies. In the second part of the paper, we extended existing cross-country comparisons of efficiency in the insurance industry by analyzing a broad international dataset that has not yet been the subject of an efficiency study. The cross-country analysis covers data on 3,710 insurance companies from 37 countries, which makes our dataset, at least to our knowledge, the largest ever analyzed in insurance-related efficiency literature. A total of 15,306 firm years were analyzed using both data envelopment analysis (DEA) and stochastic frontier analysis (SFA), which allows us to provide a broad range of new insights into the efficiency of the international insurance industry: During the sample period from 2002 to 2006, we find a steady growth in efficiency in the international insurance markets, although there are large differences between countries. In our cross-country comparison, Japan is the most efficient country, while insurers from the Philippines have the lowest efficiency values. We are the first to determine efficiency for the following 15 countries: the Bahamas (0.57 average efficiency under DEA/0.24 average efficiency under SFA), Barbados (0.70/0.27), Bermuda (0.72/0.36), Brazil (0.73/0.26), the Czech Republic (0.78/0.25), Hong Kong (0.72/0.21), Hungary (0.83/0.26), Indonesia (0.71/0.22), Lithuania (0.64/0.19), Mexico (0.69/0.24), Norway (0.80/0.35), Poland (0.71/0.33), Russia (0.66/0.22), Singapore (0.75/0.24), and South Africa (0.69/0.36). The results illustrate that SFA provides lower efficiency values than DEA. However, in general, the results of the two approaches and the economic insights that can be derived from them are actually very similar. One exception to this general similarity has to do with different distribution types (agency-based vs. direct writers): the results of the SFA are in accordance with existing literature, which shows that agency-based insurers are more efficient than direct writers; however, no difference in efficiency based on distribution type is found with DEA. In accordance with most other research on the topic, we find evidence for the expense preference hypothesis, which states that the average efficiency value of stock 26
companies is higher than that of mutual insurers. Again in line with most of the literature, we find that the efficiency of large insurers is higher than the efficiency of small insurers, which is evidence for economies of scale. However, in contrast to most other efficiency studies (see, e.g., Meador/Ryan/Schellhorn, 2000; Cummins/Weiss/Zi, 2003; Fuentes/Grifell-Tatjé/Perelman, 2005), we find no evidence for economies of scope in the international insurance industry. All results are confirmed by both DEA and SFA. Further cross sectional analysis including a Tobit regression analysis reveal that organization type and line of business are strongly associated with efficiency. The most efficient insurers are large companies selling life insurance in Japan. All these results provide valuable insights into the competitiveness of insurers in different countries. Frontier efficiency methodologies can be used to determine improvement potential for different economies by measuring industry performance relative to the best practice industries. At the individual-company level, the results can be used to compare performance with other firms in the industry and to identify areas and operations that need improvement. A number of implications for different managerial decisions can be drawn, including, e.g., decisions on mergers and acquisitions, organization, and distribution. Further research should be undertaken to validate the results presented in this first draft of the paper. Additional frontier efficiency methodologies (e.g., the distributionfree approach presented by Meador/Ryan/Schellhorn (2000)), further cross-sectional tests, and different types of efficiency (technical, scale, cost, revenue, and profit efficiency) might be addressed. Furthermore, in order to deepen the integration of the results with existing literature, the results should be analyzed at a more disaggregated level, for example, the results might be broken down by different countries and compared to the results found for different countries in existing literature. Another idea worth pursuit is to analyze the different lines of business at a more disaggregated level, especially for the non-life insurance industry. We would be delighted to present our paper, along with some of this further research, at the conference on “Performance Measurement in the Financial Services Sector” and to contribute to a special issue of the Journal of Banking and Finance.
27
REFERENCES Badunenko, O., Grechanyuk, B., Talavera, O., 2006. Development Under Regulation: The Way of the Ukrainian Insurance Market. Discussion Papers of DIW Berlin 644, DIW Berlin, German Institute for Economic Research. Banker, R. D., 1993. Maximum Likelihood, Consistency and Data Envelopment Analysis: A Statistical Foundation. Management Science 39 (10), 1265–1273. Banker, R. D., Charnes, A., Cooper, W. W., Swarts, J., Thomas, D. A., 1989. An Introduction to Data Envelopment Analysis with Some of its Models and Their Uses. In: Chan, J. L., Patton, J. M., eds., Research in Governmental and Non-Profit Accounting, Vol. 5, JAI Press, Greenwich, 125–164. Barros, P., Barroso, N., Borges, M. R., 2005. Evaluating the Efficiency and Productivity of Insurance Companies with a Malmquist Index: A Case Study for Portugal. Geneva Papers on Risk and Insurance 30 (2), 244–267. Barros, P., Obijiaku, E. L., 2007. Technical Efficiency of Nigerian Insurance Companies. Department of Economics, Institute for Economics and Business Administration (ISEG), Technical University of Lisbon. Working Papers No. 18. Berger, A. N., Cummins, J. D., Weiss, M. A., 1997. The Coexistence of Multiple Distribution Systems for Financial Services: The Case of Property-Liability Insurance. Journal of Business 70 (4), 515–546. Berger, A. N., Cummins, J. D., Weiss, M. A., Zi, H., 2000. Conglomeration Versus Strategic Focus: Evidence from the Insurance Industry. Journal of Financial Intermediation 9 (4), 323–362. Berger, A. N., Humphrey, D. B., 1992. Measurement and Efficiency Issues in Commercial Banking. In: Griliches, Z., ed., Output Measurement in the Service Sectors, Vol. 56, National Bureau of Economic Research, Studies in Income and Wealth, University of Chicago Press (Chicago, IL), 245–279. Berger, A. N., Humphrey, D. B., 1997. Efficiency of Financial Institutions: International Survey and Directions for Future Research. European Journal of Operational Research 98 (2), 175–212. Bernstein, J. I., 1999. Total Factor Productivity Growth in the Canadian Life Insurance Industry: 1979–1989. Canadian Journal of Economics 32 (2), 500–517. Bikker, J. A., van Leuvensteijn, M., 2008. Competition and Efficiency in the Dutch Life Insurance Industry. Applied Economics (forthcoming). Boonyasai, T., Grace, M. F., Skipper, Jr., H. D., 2002. The Effect of Liberalization and Deregulation on Life Insurer Efficiency. Working Paper No. 02-2, Center for Risk Management and Insurance Research, Georgia State University, Atlanta. Brockett, P. L., Cooper, W. W., Golden, L. L., Rousseau, J. J., Wang, Y., 1998. DEA Evaluations of the Efficiency of Organizational Forms and Distribution Systems in the US Property and Liability Insurance Industry. International Journal of Systems Science 29 (11), 1235–1247. 28
Brockett, P. L., Cooper, W. W., Golden, L. L., Rousseau, J. J., Wang, Y., 2004. Evaluating Solvency Versus Efficiency Performance and Different Forms of Organization and Marketing in US Property-Liability Insurance Companies. European Journal of Operational Research 154 (2), 492–514. Brockett, P. L., Cooper, W. W., Golden, L. L., Rousseau, J. J., Wang, Y., 2005. Financial Intermediary Versus Production Approach to Efficiency of Marketing Distribution Systems and Organizational Structure of Insurance Companies. Journal of Risk and Insurance 72 (3), 393–412. Carr, R. M., Cummins, J. D., Regan, L., 1999. Efficiency and Competitiveness in the U.S. Life Insurance Industry: Corporate, Product, and Distribution Strategies. In: Cummins, J. D., Santomero, A. M., eds., Changes in the Life Insurance Industry: Efficiency, Technology and Risk Management. Kluwer Academic Publishers, Boston, MA. Chaffai, M. E., Ouertani, M. N., 2002. Technical Efficiency in the Tunisian Insurance Industry: A Comparison of Parametric and Non Parametric Time Variant Models. Working Paper, Research Unit on Production Econometrics, Sfax University, Sfax. Charnes, A., Cooper, W. W., Lewin, A., Seiford, L., 1994. Data Envelopment Analysis: Theory, Methodology and Applications. Kluwer Academic Publishers, Boston, MA. Charnes, A., Cooper, W. W., Rhodes, E., 1978. Measuring the Efficiency of Decision Making Units. European Journal of Operational Research 2 (6), 429–444. Choi, P. B., Weiss, M. A., 2005. An Empirical Investigation of Market Structure, Efficiency, and Performance in Property-Liability Insurance. Journal of Risk and Insurance 72 (4), 635–673. Coelli, T. J., 1996a. A Guide to Deap Version 2.1: A Data Envelopment Analysis (Computer) Program, Working Paper No. 8/96, Department of Econometrics, University of New England. Coelli, T. J., 1996b. A Guide to Frontier 4.1: A Computer Program for Stochastic Frontier Production and Cost Function Estimation, Working Paper No. 7/96, Department of Econometrics, University of New England. Cooper, W. W., Seiford, L. M., Tone, K., 2006. Introduction to Data Envelopment Analysis and its Uses. Springer, Berlin/Heidelberg/New York. Cummins, J. D., 1999. Efficiency in the U.S. Life Insurance Industry: Are Insurers Minimizing Costs and Maximizing Revenues? In: Cummins, J. D., Santomero, A. M., eds., Changes in the Life Insurance Industry: Efficiency, Technology and Risk Management. Kluwer Academic Publishers, Boston, MA. Cummins, J. D., Dionne, G., Gagné, R., Nouira, A., 2006. Efficiency of Insurance Firms with Endogenous Risk Management and Financial Intermediation Activities. Working Paper. Cummins, J. D., Nini, G. P., 2002. Optimal Capital Utilization by Financial Firms: Evidence from the Property-Liability Insurance Industry. Journal of Financial Services Research 21 (1–2), 15–53. 29
Cummins, J. D., Rubio-Misas, M., 2006. Deregulation, Consolidation, and Efficiency: Evidence from the Spanish Insurance Industry. Journal of Money, Credit, and Banking 38 (2), 323–355. Cummins, J. D., Rubio-Misas, M., Zi, H., 2004. The Effect of Organizational Structure on Efficiency: Evidence from the Spanish Insurance Industry. Journal of Banking and Finance 28 (12), 3113–3150. Cummins, J. D., Tennyson, S., Weiss, M. A., 1999. Consolidation and Efficiency in the US Life Insurance Industry. Journal of Banking and Finance 23 (2–4), 325– 357. Cummins, J. D., Turchetti, G., Weiss, M. A., 1996. Productivity and Technical Efficiency Italian Insurance Industry. Working Paper, Wharton Financial Institutions Center, University of Pennsylvania, PA. Cummins, J. D., Weiss, M. A., 1993. Measuring Cost Efficiency in the PropertyLiability Insurance Industry. Journal of Banking and Finance 17 (2–3), 463–481. Cummins, J. D., Weiss, M. A., 2000. Analyzing Firm Performance in the Insurance Industry Using Frontier Efficiency Methods. In: Dionne, G., ed., Handbook of Insurance Economics. Kluwer Academic Publishers, Boston, MA. Cummins, J. D., Weiss, M. A., Zi, H., 1999. Organizational Form and Efficiency: The Coexistence of Stock and Mutual Property-Liability Insurers. Management Science 45 (9), 1254–1269. Cummins, J. D., Weiss, M. A., Zi, H., 2003. Economics of Scope in Financial Services: A DEA Bootstrapping Analysis of the US Insurance Industry. Working Paper, The Wharton School, Philadelphia, PA. Cummins, J. D., Xie, X., 2008. Mergers and Acquisitions in the US Property-Liability Insurance Industry: Productivity and Efficiency Effects. Journal of Banking and Finance 32 (1), 30–55. Cummins, J. D., Zi, H., 1998. Comparison of Frontier Efficiency Methods: An Application to the U.S. Life Insurance Industry. Journal of Productivity Analysis 10 (2), 131–152. Delhausse, B., Fecher, F., Pestieau, P., 1995. Measuring Productive Performance in the Non-Life Insurance Industry: The Case of French and Belgian Markets. Tijdschrift voor Economie en Management 40 (1), 47–69. Diacon, S. R., 2001. The Efficiency of UK General Insurance Companies. Working Paper, Centre for Risk & Insurance Studies, University of Nottingham. Diacon, S. R., Starkey, K., O’Brien, C., 2002. Size and Efficiency in European LongTerm Insurance Companies: An International Comparison. Geneva Papers on Risk and Insurance 27 (3), 444–466. Diboky, F., Ubl, E., 2007. Ownership and Efficiency in the German Life Insurance Market: A DEA Bootstrap Approach. Working Paper. Donni, O., Fecher, F., 1997. Efficiency and Productivity of the Insurance Industry in the OECD Countries. Geneva Papers on Risk and Insurance 22 (84), 523–535. 30
Donni, O., Hamende, V., 1993. Performance des Sociétés Belges D’AssuranceComparaison des formes institutionelles. Annals of Public and Cooperative Economics, Université de Liège. Ennsfellner, K. C., Lewis, D., Anderson, R. I., 2004. Production Efficiency in the Austrian Insurance Industry: A Bayesian Examination. Journal of Risk and Insurance 71 (1), 135–159. Erhemjamts, O., Leverty, J. T., 2007. The Demise of the Mutual Organizational Form: An Investigation of the Life Insurance Industry. Working Paper. Farrell, M. J., 1957. The Measurement of Productive Efficiency. Journal of the Royal Statistical Society A 120 (3), 253–281. Fecher, F., Kessler, D., Perelman, S., Pestieau, P., 1993. Productive Performance in the French Insurance Industry. Journal of Productivity Analysis 4 (1–2), 77–93. Fecher, F., Perelman, S., Pestieau, P., 1991. Scale Economics and Performance in the French Insurance Industry. Geneva Papers on Risk and Insurance 16 (60), 315– 326. Fenn, P., Vencappa, D., Diacon, S., Klumpes, P., O’Brien, C., 2008. Market Structure and the Efficiency of European Insurance Companies: A Stochastic Frontier Analysis. Journal of Banking and Finance 32 (1), 86–100. Fuentes, H., Grifell-Tatjé, E., Perelman, S., 2001. A Parametric Distance Function Approach for Malmquist Productivity Index Estimation. Journal of Productivity Analysis 15 (2), 79–94. Fuentes, H., Grifell-Tatjé, E., Perelman, S., 2005. Product Specialization, Efficiency and Productivity Change in the Spanish Insurance Industry. Working Paper, Université de Liège. Fukuyama, H., 1997. Investigating Productive Efficiency and Productivity Changes of Japanese Life Insurance Companies. Pacific-Basin Finance Journal 5 (4), 481– 509. Fukuyama, H., Weber, W. L., 2001. Efficiency and Productivity Change of Non-Life Insurance Companies in Japan. Pacific Economic Review 6 (1), 129–146. Gardner, L. A., Grace, M. F., 1993. X-Efficiency in the US Life Insurance Industry. Journal of Banking and Finance 17 (2–3), 497–510. Greene, W. H., 1993. The Econometric Approach to Efficiency Analysis. In: Fried, H. O., Lovell, C. A. K., Schmidt, S. S., eds., The Measurement of Productive Efficiency: Techniques and Applications. Oxford University Press, Oxford, 68–119. Greene, W. H., Segal, D., 2004. Profitability and Efficiency in the U.S. Life Insurance Industry. Journal of Productivity Analysis 21 (3), 229–247. Grosskopf, S., 1993. Efficiency and Productivity. In: Fried, H. O., Lovell, C. A. K., Schmidt, S. S., eds., The Measurement of Productive Efficiency: Techniques and Applications. Oxford University Press, Oxford. Hancock, D., 1985. The Financial Firm: Production with Monetary and Non-Monetary Goods. Journal of Political Economy 93 (5), 859–880. 31
Hao, J. C. J., 2007. Efficiency Test on Taiwan’s Life Insurance Industry-Using XEfficiency Approach. Information and Management Sciences 18 (1), 37–48. Hao, J. C. J., Chou, L. Y., 2005. The Estimation of Efficiency for Life Insurance Industry: The Case in Taiwan. Journal of Asian Economics 16 (5), 847–860. Hardwick, P., 1997. Measuring Cost Inefficiency in the UK Life Insurance Industry. Applied Financial Economics 7 (1), 37–44. Hardwick, P., Adams, M., Zou, H., 2004. Corporate Governance and Cost Efficiency in the United Kingdom Life Insurance Industry. Working Paper. Hirao, Y., Inoue, T., 2004. On the Cost Structure of the Japanese Property-Casualty Insurance Industry. Journal of Risk and Insurance 71 (3), 501–530. Huang, W., 2007. Efficiency in the China Insurance Industry: 1999–2004. Working Paper. Hussels, S., Ward, D. R., 2006. The Impact of Deregulation on the German and UK Life Insurance Markets: An Analysis of Efficiency and Productivity between 1991–2002. Working Paper, Cranfield Research Paper Series (4). Hwang, S. N., Kao, C., 2008. Efficiency Decomposition in Two-Stage Data Envelopment Analysis: An Application to Non-Life Insurance Companies in Taiwan. European Journal of Operational Research 185 (1), 418–429. Hwang, T., Gao, S. S., 2005. An Empirical Study of Cost Efficiency in the Irish Life Insurance Industry. International Journal of Accounting, Auditing and Performance Evaluation 2 (3), 264–280. Jeng, V., Lai, G. C., 2005. Ownership Structure, Agency Costs, Specialization, and Efficiency: Analysis of Keiretsu and Independent Insurers in the Japanese Nonlife Insurance Industry. Journal of Risk and Insurance 72 (1), 105–158. Jeng, V., Lai, G. C., McNamara, M. J., 2007. Efficiency and Demutualization: Evidence from the U.S. Life Insurance Industry in the 1980s and 1990s. Journal of Risk and Insurance 74 (3), 683–711. Kessner, K., 2001a. Ein Effizienzvergleich deutscher und britischer Lebensversicherungen. In: Markttransparenz und Produktionseffizienz in der deutschen Lebensversicherung. Dissertation, Ludwig-Maximilians-Universität München. Kessner, K., 2001b. Skaleneffizienz und Produktivitätswachstum in der deutschen Lebensversicherung. In: Markttransparenz und Produktionseffizienz in der deutschen Lebensversicherung. Dissertation, Ludwig-Maximilians-Universität München. Kessner, K., Polborn, M., 1999. Eine Effizienzanalyse der deutschen Lebensversicherer—die Best Practice Methode. Zeitschrift für die gesamte Versicherungswissenschaft 88, 469–488 Kim, H., Grace, M. F., 1995. Potential Ex Post Efficiency Gains of Insurance Company Mergers. Working Paper, Center for RMI Research 95-4, College of Business Administration, Georgia State University, Atlanta. 32
Klumpes, P. J. M., 2004. Performance Benchmarking in Financial Services: Evidence from the UK Life Insurance Industry. Journal of Business 77 (2), 257–274. Klumpes, P. J. M., 2007. Consolidation and Efficiency in the Major European Insurance Markets. Working Paper, Imperial College, London. Leverty, T., Lin, Y., Zhou, H., 2004. Firm Performance in the Chinese Insurance Industry. Working Paper. Mahlberg, B., 1999. Effizienzmessung österreichischer und deutscher Versicherungen—Ein Vergleich. Wirtschaftspolitische Blätter 46 (4), 400–406. Mahlberg, B., 2000. Technischer Fortschritt und Produktivitätsveränderungen in der deutschen Versicherungswirtschaft. Jahrbücher für Nationalökonomik und Statistik 220 (5). Mahlberg, B., Url, T., 2000. The Transition to the Single Market in the German Insurance Industry. Working Paper, Austrian Institute of Economic Research. Mahlberg, B., Url, T., 2003. Effects of the Single Market on the Austrian Insurance Industry. Empirical Economics 28, 813–838. Mansor, S. A., Radam, A., 2000. Productivity and Efficiency Performance of the Malaysian Life Insurance Industry. Jurnal Ekonomi Malaysia 34, 93–105. Mayers, D., Smith, C. W., 1988. Ownership Structure Across Lines of PropertyCasualty Insurance. Journal of Law and Economics 31, 351–378. Meador, J. W., Ryan, Jr., H. E., Schellhorn, C. D., 2000. Product Focus Versus Diversification: Estimates of X-Efficiency for the US Life Insurance Industry. Working Paper, Wharton Financial Institutions Center University of Pennsylvania. Mester, L. J., 1991. Agency Costs Among Savings and Loans. Journal of Financial Intermediation 1, 257–278. Noulas, A. G., Hatzigayios, T., Lazaridis, J., Lyroudi, K., 2001. Non-Parametric Production Frontier Approach to the Study of Efficiency of Non-Life Insurance Companies in Greece. Journal of Financial Management and Analysis 14 (1), 19–26. Qiu, S., Chen, B., 2006. Efficiencies of Life Insurers in China—An Application of Data Envelopment Analysis. Working Paper. Rai, A., 1996. Cost Efficiency of International Insurance Firms. Journal of Financial Services Research 10 (3), 213–233. Rees, R., Kessner, E., Klemperer, P., Matutes, C., 1999. Regulation and Efficiency in European Insurance Markets. Economic Policy 14 (29), 363–397. Ryan, Jr., H. E., Schellhorn, C. D., 2000. Life Insurer Cost Efficiency Before and After Implementation of the NAIC Risk-Based Capital Standards. Journal of Insurance Regulation 18 (3), 362–384. Simar, L., Wilson, P. W., 1998. Sensitivity Analysis of Efficiency Scores: How to Bootstrap in Nonparametric Frontier Models. Management Science 44 (11), 49–61. Tobin, J., 1958. Estimation of Relationships for Limited Dependent Variables. Econometrica 26 (1), 24–36. Toivanen, O., 1997. Economies of Scale and Scope in the Finnish Non-Life Insurance Industry. Journal of Banking and Finance 21 (6), 759–779. 33
Tone, K., Sahoo, B. K., 2005. Evaluating Cost Efficiency and Returns to Scale in the Life Insurance Corporation of India Using Data Envelopment Analysis. SocioEconomics Planning Sciences 39 (4), 261–285. Ward, D., 2002. The Costs of Distribution in the UK Life Insurance Market. Applied Economics 34 (15), 1959–1968. Weiss, M. A., 1986. Analysis of Productivity at the Firm Level: An Application to Life Insurers. Journal of Risk and Insurance 53 (1), 49–84. Weiss, M. A., 1991a. Efficiency in the Property-Liability Insurance Industry. Journal of Risk and Insurance 58 (3), 452–479. Weiss, M. A., 1991b. International P/L Insurance Output, Input, and Productivity Comparisons. Geneva Papers on Risk and Insurance Theory 16 (2), 179–200. Worthington, A. C., Hurley, E. V., 2002. Cost Efficiency in Australian General Insurers: A Non-Parametric Approach. The British Accounting Review 34 (2), 89–108. Wu, D., Yang, Z., Vela, S., Liang, L., 2007. Simultaneous Analysis of Production and Investment Performance of Canadian Life and Health Insurance Companies Using Data Envelopment Analysis. Computers & Operations Research 34 (1), 180–198. Yang, Z., 2006. A Two-Stage DEA Model to Evaluate the Overall Performance of Canadian Life and Health Insurance Companies and Computer. Mathematical and Computer Modelling 43 (7–8), 910–919. Yao, S., Han, Z., Feng, G., 2007. On the Technical Efficiency of China’s Insurance Industry After WTO Accession. China Economic Review 18 (1), 66–86. Yuengert, A. M., 1993. The Measurement of Efficiency in Life Insurance: Estimates of a Mixed Normal-Gamma Error Model. Journal of Banking and Finance 17 (2– 3), 483–496.
34
APPENDIX: OVERVIEW OF 83 STUDIES ON EFFICIENCY IN THE INSURANCE INDUSTRY Authors
Countries
No. Sample Lines of insurers period business 163 2003Life, non-life 2005
Meth Input type od DEA Fixed assets, current assets, liabilities, equity
27
19952001
Life, non-life
DEA Wages, capital, total investment income, premiums issued
Barros/Obijiaku (2007) Nigeria
10
20012005
Life, non-life
DEA Capital, operative costs, number of employees, total investments
Berger et al. (2000)
684
19881992
Life, property- SFA liability
472
19811990 19791989
Propertyliability Life
Labor, business services, debt capital, equity capital Index Labor, buildings capital, machinery capital, materials
Bikker/van Leuvensteijn Netherlands 84-105 (2008)
19952003
Life
SFA
Boonyasai/Grace/Skipper (2002)
19781997
Life
DEA Labor, capital, materials
Brockett et al. (1998)
Korea, 49-110 Philippines, Taiwan, Thailand US 1524
1989
Propertyliability
Brockett et al. (2004)
US
1524
1989
Propertyliability
Brockett et al. (2005)
US
1524
1989
Propertyliability
Carr/Cummins/Regan (1999)
US
66
n/a
Life
13
19902000
Badunenko/ Ukraine Grechanyuk/ Talavera (2006) Barros/Barroso/Borges Portugal (2005)
US
Berger/Cummins/Weiss US (1997) Bernstein (1999) Canada
Chaffai/Ouertani (2002) Tunisia
12
Choi/Weiss (2005)
US
n/a
Cummins (1999)
US
750
Cummins et al. (2006)
US
1636
Labor, business services, reserves, equity capital
DFA
Output type
Output approach Value added
Main types of efficiencies analyzed Technical, scale
Claims paid, profits
Value added
Technical, pure technical, scale
Profits, net premiums, settled claims, outstanding claims, investment income
Value added
Technical, pure technical, scale
Premiums
Invested assets, present value of Value real losses incurred (p/l), incurred added benefits (life) Total real invested assets, real Value present value of expected losses added Number of policies Physical
Application category Selected findings Deregulation, regulation change
Increased capitalization requirements have positively influenced Ukrainian insurance market and helped improve both technical and scale efficiency
General level of efficiency and productivity, evolution over time General level of efficiency and productivity, evolution over time Scale and scope economies
Improvement of technical efficiency over time, but deterioration in terms of technological change
Cost, profit
Distribution systems
Independent agents less cost efficient but equally profit efficient as direct writers
Technical, scale
General level of Average annual productivity growth of 1% efficiency and productivity, evolution over time General level of Cost inefficiency of 25% on average efficiency and productivity, evolution over time
Cost, revenue, profit
Most companies are VRS efficient
Conglomeration hypothesis holds for some types while strategic focus hypothesis dominates for others
Physical Premium income, number of outstanding policies, sum total of insured capital, sum total of insured annuities, unit-linked fund policies Premium income, net investment Value income added
Cost
Technical, pure technical, scale
Deregulation, regulation change
Increasing productivity in Korea and Philippines due to deregulation and liberalization; little effect of liberalization on productivity in Taiwan and Thailand
DEA Surplus previous year, change in capital and surplus, underwriting and investment expense, policyholdersupplied debt capital DEA Surplus previous year, change in capital and surplus, underwriting and investment expense, policyholdersupplied debt capital DEA Surplus previous year, change in capital and surplus, underwriting and investment expense, policyholdersupplied debt capital DEA Labor (admin., agents), business services, financial capital
ROI, liquid assets to liability, solvency scores
Financial intermediary
n/a
Distribution systems
Stock companies more efficient than mutuals; agency more efficient marketing system than direct
ROI, liquid assets to liability, solvency scores
Financial intermediary
n/a
Financial and risk management, capital utilization
Solvency scores as output only with limited impact on efficiency scores
ROI, liquid assets to liability, solvency scores
Financial intermediary
n/a
Organizational form, Stock firms with higher inefficiency in the input dimension while mutuals with corporate governance higher shortfalls in all areas of output; direct systems with more inefficiencies issues than agency
Incurred benefits, additions to reserves
Value added
Cost, revenue
Distribution systems
Life, non-life
DFA, Labor, physical capital, financial SFA capital
Total premiums earned
Value added
Technical
19921998 19881995
Propertyliability Life
SFA Labor (agent, nonagent), materials, equity capital DEA Labor (admin., agents), business services, financial capital
Present value of losses incurred, total invested assets Incurred benefits, additions to reserves
Value added Value added
19952003
Propertyliability
SFA
Acquisition cost, other cost (management cost, salaries, depreciation on capital equipment, etc.)
Labor (admin., agents, risk manage- Present value of losses incurred, Value ment), material and business service, invested assets, dollar duration of added debt capital, equity capital surplus
Exclusive dealing insurers less efficient than nonexclusive dealing or direct writers; nonexclusive dealing insurers should focus on fewer product lines; firms adopting one of Porter’s 3 generic strategies are more efficient than rivals Significant potential for increase of efficiency
General level of efficiency and productivity, evolution over time Cost, revenue Market structure Cost-efficient firms charge lower prices and earn higher profits; prices and profits higher in revenue-efficient firms Pure technical, scale, General level of Efficiency scores in insurance relatively low compared to other financial serallocative, cost, efficiency and produc- vices industries; also widely dispersed; small insurers operate with IRS; big revenue tivity, evolution over insurers with DRS; brokerage system most efficient time Cost Financial and risk Risk management and financial intermediation increase efficiency management, capital utilization
Cummins/Nini (2002)
US
770
19931998
Propertyliability
Cummins/Rubio-Misas Spain (2006)
508
19891998
Life, non-life
Cummins/RubioMisas/Zi (2004)
Spain
347
19891997
Life, non-life
DEA Labor, business services, debt capital, equity capital
Cummins/Tennyson/Weiss (1999)
US
750
19891995
Life
DEA Home-office labor, agent labor, Incurred benefits, additions to business services (including physical reserves capital), financial capital
Cummins/Turchetti/Weiss (1996)
Italy
94
19851993
Life, non-life
DEA Labor (acquisition, admin.), fixed capital, equity capital.
Cummins/Weiss (1993) US
270
19801988
Propertyliability
Cummins/Weiss/Zi (1999)
US
417
19811990
Propertyliability
Cummins/Weiss/Zi (2003)
US
817
19931997
DEA Labor (office, agent), materials and business service, financial equity capital
Technical, cost, revenue
US
1550
19942003
Health/life/accident: Incurred benefits plus changes in reserves; Non-life: Losses incurred, real invested assets DEA Labor (admin., agent), materials and Present value of losses incurred, business services, financial equity real invested assets capital
Value added
Cummins/Xie (2008)
Life incl. health; property-liability, diversified firms Propertyliability
Value added
Cummins/Zi (1998)
US
445
19881992
Life
Benefit payments, additions to reserves
Value added
DelBelgium, 434 hausse/Fecher/Pestieau France (1995) Diacon (2001) 6 European 431 countries
19841988
Non-life
Premiums
Value added
Technical, scale
1999
General insurance
Technical
19961999
Diboky/Ubl (2007)
90
20022005
Life incl. pension, and health Life
Net earned premiums (general, long-term), total investment income Net earned premiums (general, long-term), total investment income Gross premium, net income
Value added
Diacon/Starkey/O’Brien 15 European 454 (2002) countries
DEA, Labor, financial capital, materials DFA, FDH, SFA DEA, Labor costs, other outlays (capital SFA consumption, purchase of equipment and supplies, etc.) DEA Total operating expenses, total capital, total technical reserves, total borrowings from creditors DEA Total operating expenses, total capital, total technical reserves, total borrowings from creditors DEA Labor, business services, financial debt capital, equity capital
Cost, technical, allocative, pure technical, scale, revenue Cost, technical, allocative
n/a
19831991 19821988
Germany
Donni/Fecher (1997)
15 OECD countries Donni/Hamende (1993) Belgium
Ennsfellner/Lewis/Anderson (2004)
Austria
300
97
19941999
DEA Labor (office, sales), materials and business service, financial equity capital DEA Labor, business services, debt capital, equity capital
Present value of losses incurred, Value total invested assets added
Technical, allocative, cost, revenue
Non-life losses incurred, life Value losses incurred, reinsurance added reserves, nonreinsurance reserves, invested assets Life and non-life insurance losses Value incurred added
Cost, pure technical, allocative, scale
Value added
Life insurance: sum of life Value insurance benefits, changes in added reserves, invested assets Non-life insurance: Losses incurred, invested assets SFA Labor, capital, intermediate materials Discounted incurred losses, loss Value added reserves, loss adjustment expense reserve, unearned premium reserve, loss adjustment expenses paid DEA Labor, business services, debt Real additions to reserves, real Value capital, equity capital incurred benefits added
Life, non-life
DEA Labor
Net premiums
Life, non-life
FDH Labor cost, other cost
Premiums; alternatively, losses incurred
Life, health, non-life
SFA
Health/life: Incurred benefits, changes in reserves, total invested assets Non-life: Claims incurred, total invested assets
Net operating expenses, equity capital, technical provisions
Technical, allocative, cost, revenue
Cost, technical, allocative, pure technical, scale, revenue Technical
Financial and risk management, capital utilization Deregulation, regulation change
On weighted industry average, firms could reduce labor by 62%, materials by 36%, and capital by 46%; capital is used suboptimally Consolidation leads to growth in TFP and increases number of firms operating with decreasing returns to scale
Organizational form, In cost and revenue efficiency, stocks of all sizes dominate mutuals in the corporate governance production of stock output vectors, and smaller mutuals dominate stocks in the issues production of mutual output vectors; larger mutuals are neither dominated by nor dominant over stocks Mergers M&A beneficial for efficiency; target life insurers achieve greater efficiency gains than firms that have not been involved in M&As
General level of Stable efficiency over time (70–78 for the industry) with sharp decline (25% efficiency and produc- cumulative) in productivity due to technological regress tivity, evolution over time
Cost
General level of Large insurers at 90% relative to their cost frontier; medium and small insurers efficiency and produc- at 80% and 88%, respectively; scale economies with small and medium-sized tivity, evolution over insurers time
Technical, cost
Organizational form, Stock cost frontier dominates mutual cost frontier corporate governance issues Scale and scope Weak evidence for existence of economies of scope economies
Mergers
M&As in property-liability insurance are value enhancing; acquiring firms achieved more revenue efficiency gains than nonacquiring firms, and target firms experienced greater cost and allocative efficiency growth than nontargets
Methodology issues, comparing different techniques or assumptions Intercountry comparisons
Choice of estimation method can have a significant effect on the conclusions of an efficiency study; efficiency rankings are well preserved among the econometric methods; but the rankings are less consistent between the econometric and mathematical programming methods French companies on average more efficient than Belgian ones; overall low efficiency levels; high correlation between results of both approaches
Intercountry comparisons
Average efficiencies: UK (77%), France (67%), Germany (70%), Italy (56%), Netherlands (69%), Switzerland (66%)
Value added
Pure technical, scale, Intercountry comparimix sons
Striking international differences and decreasing levels of average technical efficiency over sample period
Value added
Technical, cost and allocative efficiency
Stock ownership is superior to mutual and public structure
Value added Value added
Technical
Value added
Technical
Technical
Organizational form, corporate governance issues Intercountry comparisons Organizational form, corporate governance issues Market structure
Average efficiency levels rather high and dispersed; growth in productivity observed in all countries and due to improvements in technical progress Superior efficiency of nonprofit insurance companies
Deregulation had positive effects on production efficiency
36
Erhemjamts/Leverty (2007)
US
697
19952004
Life
DEA Labor, business services, equity capital, policyholder-supplied debt capital DEA, Labor cost, other outlays SFA
Incurred benefits, additions to reserves
Value added
Technical
Fecher et al. (1993)
France
327
19841989
Life, non-life
Gross premiums
Value added
Technical, cost
Fecher/Perelman/Pestie France 327 au (1991) Fenn et al. (2008) 14 European n/a countries
19841989 19952001
Life, non-life
SFA
Labor cost, other outlays
Gross premiums
Cost
SFA
Capital, technical provisions, labor, debt capital
Value added Value added
Life, non-life, composite
Cost
Health, life, non-life
SFA
Labor costs, composite input
Net incurred claims (= gross claims paid – claims received from reinsurers + increase in loss reserves + bonuses and rebates) Annual premiums Value added
Fuentes/GrifellTatjé/Perelman (2001)
Spain
55-70
19871994
Fuentes/GrifellTatjé/Perelman (2005)
Spain
n/a
19871997
Health, life, propertyliability Life
SFA
Labor costs, composite input
Annual premiums
Value added
Technical
Fukuyama (1997)
Japan
25
19881993
DEA Labor (office, sales), capital
Insurance reserves, loans
Technical, scale
19831994
Non-life
DEA Labor, capital
Reserves, loans, investments
Financial intermediary Financial intermediary
Fukuyama/Weber (2001)
Japan
17
Gardner/Grace (1993)
US
561
19851990
Life
DFA
Labor, physical capital, misc. items
Premiums, securities investments Value added
Greene/Segal (2004)
US
136
19951998
Life
SFA
Labor, capital, materials
Premiums, investments
Value added
Hao (2007)
Taiwan
26
19812003
Life
DFA
Labor, physical capital, claims
Premiums, investments
Value added
Hao/Chou (2005)
Taiwan
26
19771999
Life
DFA
Labor, physical capital, claims
Premiums, investments
Value added
Hardwick (1997)
UK
54
19891993
Life incl. pension, and health
SFA
Labor, capital
Premiums
Value added
Hardwick/Adams/Zou (2004)
UK
50
19942001
Life
DEA Labor, capital
Incurred benefits, additions to reserves
Value added
Hirao/Inoue (2004)
Japan
33
19801995
Propertyliability
SFA
Value added
Huang (2007)
China
n/a
19992004
Life, property- SFA liability
Real incurred losses (net claims paid and changes in loss reserves) Premiums earned, incurred benefits and additions to reserves, total invested assets
Hussels/Ward (2006)
Germany and UK
47 (UK) 31 (GE)
19912002
Life
DEA, Labor, capital DFA
Hwang/Gao (2005)
Ireland
11
19912000
Life
DFA
Labor, agencies, materials
Labor, capital, business services
Labor (admin., agent), financial capital
Technical
Organizational form, corporate governance issues General level of efficiency and productivity, evolution over time Scale and scope economies Market structure
Stock production technology dominates mutual technology; mutuals that are further away from mutual efficient frontier more likely to demutualization; access to capital important reason for demutualization High correlation between parametric and nonparametric results; wide dispersion in the rates of inefficiency across companies
Methodology issues, comparing different techniques or assumptions Scale and scope economies
Malmquist index of productivity growth can also be estimated on basis of parametric frontier approach
Increasing returns to scale Most European insurers operating under IRS; size and domestic market share lead to higher levels of cost inefficiency
Overall low productivity growth over time (less than 2% per year), multi-branch companies perform better than specialized firms
Organizational form, corporate governance issues Technical Methodology issues, comparing different techniques or assumptions Cost General level of efficiency and productivity, evolution over time Cost Organizational form, corporate governance issues Cost General level of efficiency and productivity, evolution over time Cost General level of efficiency and productivity, evolution over time Economic, scale, total General level of inefficiency efficiency and productivity, evolution over time Cost, technical, Organizational form, allocative corporate governance issues Cost, scale Scale and scope economies
Mutual and stock companies possess identical technologies; productive efficiency and productivity performances differ across 2 ownership types and different economic conditions Productivity and technological progess over time in Japan
Value added
Cost, profit
Non-state-owned companies and foreign companies are superior in terms of cost efficiency to the property insurance industry, state-owned companies, and domestic companies
Net written premiums, additions to reserves
Value added
Cost, technical, allocative, scale
Insurance benefits, investable funds
Value added
Cost
General level of efficiency and productivity, evolution over time Deregulation, regulation change Scale and scope economies
Persistent inefficiency
Inefficiency negatively associated with profitability; stock companies as efficient and profitable as mutual companies Firms with large market share tend to be cost efficient
Firms with larger market share are more profitable; product diversification does not improve efficiency
High level of inefficiency; increasing returns to scale
Cost efficiency positively related to size of corporate board of directors
Statistically significant economies of scale and scope
Comparability of results from DEA and DFA; UK efficiency frontier less efficient than German frontier; no clear evidence for link between deregulation and efficiency levels Increasing returns to scale; magnitude of cost economies varies with firm size
37
Hwang/Kao (2008)
Taiwan
24
20012002
Non-life
DEA Operation expenses, insurance expenses
Jeng/Lai (2005)
Japan
21
19831994
Non-life
DEA VA: Labor, business services, capital (debt + equity) FI: ROA, financial condition of insurer (proxied)
Jeng/Lai/McNamara (2007)
US
11
19801995
Life
Kessner (2001a)
Germany and UK
87 (UK) 78 (GE)
19941999
Life
Kessner (2001b)
Germany
75
19891994
Life
Kessner/Polborn (1999) Germany
110
19901993
Life
DEA VA: Labor, business services, equity capital FI: ROA, financial condition of insurer (proxied) DEA New business cost, administration cost, cost for capital management, reinsurance contributions DEA New business cost, administration cost, cost for capital management, reinsurance contributions DEA New business cost, administration cost
Kim/Grace (1995)
US
248
19881992
Life
DFA
Klumpes (2004)
UK
40
Life
SFA
Klumpes (2007)
8 EU countries
1183
19941999 19972001
Life, general insurance
19952002
Life, property- DEA Business expenses, financial equity casualty capital, debt capital
Value added
Value added
Leverty/Lin/Zhou (2004) China
20-41
Austria and n/a Germany
19921996
Life, health, propertyliability
DEA
Mahlberg (2000)
Germany
348
19921996
Life, health, propertyliability
DEA
Mahlberg/Url (2000)
Germany
464-533 19921996
DEA
Mahlberg/Url (2003)
Austria
70
19921999
Life, health, propertyliability Life, health, propertyliability
Mansor/Radam (2000) Malaysia
12
19871997
Life
DEA
Meador/Ryan/ Schellhorn (2000) Noulas et al. (2001)
US
358
Life
DFA
Greece
16
19901995 19911996
DEA
Methodology issues, New relational model is more reliable in measuring efficiencies than independcomparing different ent models techniques or assumptions
Value Technical, cost added/finan cial intermediation
Organizational form, Keiretsu firms more cost efficient than NSIFs; otherwise not possible to reject corporate governance null hypothesis that all equally efficient; deteriorating efficiency for all company issues types; FI and VA approaches with different, but complementary, results
Value Cost, technical, added/finan allocative cial intermediation Value Technical added
Organizational form, For both approaches, no efficiency improvement after demutualization; excepcorporate governance tion: improvement for mutual control insurers under FI approach issues Intercountry comparisons
British insurers more efficient than German insurers; increasing efficiency in both markets Small companies with increasing returns to scale; big companies with decreasing returns to scale
Value added
Technical, scale
Scale and scope economies
Value added
Technical
Claims, changes in reserves, investment expenses
Value added
Cost
Value added Value added
Cost, profit
General level of High level of inefficiency efficiency and productivity, evolution over time Mergers Smaller firms with larger cost savings from mergers than large firms; no cost savings in mergers of mutuals; mergers of efficient with less efficient companies increase combined firm efficiency Distribution systems IFA-based firms less cost and profit efficient than AR/CR firms
Life: Net premiums written, real invested assets P&C: Losses incurred, real invested assets Administration and distribution cost (1 P&L: Claims, change in reserves, input) refund of premium Life, health, accident: Claims, change in reserves, refund of premium Administration and distribution cost (1 P&L: Claims, change in reserves input) Life, health, accident: Claims, change in reserves (DeckRST), refund of premium Expenditures on labor, material, Claims, net change in provisions, energy, depreciation, marketing, allocated investment returns, commissions (1 input) bonuses and returned premia Expenditures on labor, material, Claims, net change in provisions, energy, depreciation, marketing, allocated investment returns, commissions (1 input); capital mgt. bonuses and returned premia cost (1 input) Claims, commission, salaries, New policy issued, premium, expenses, other cost policy in force
Labor, physical capital, misc. items
New 2n/a stage production process
Sum insured (new and existing business), net returns on capital investments Sum insured of new and in-force business
Labor (home office, agent), business Claims, real invested assets services, financial capital DEA Labor, business services, debt Premiums, investment income capital, equity capital
Mahlberg (1999)
Non-life
Labor (agent, nonagent), capital, materials
Intermediate products: direct written premiums, reinsurance premiums; Final outputs: underwriting profit, investment profit VA: Number of policies, total invested assets FI: Equity capital, underwriting and investment expenses, debt capital supplied by policyholders VA: Benefits FI: Equity capital, underwriting and investment expenses, debt capital supplied by policyholders Gross and net written premiums, interest on capital
Cost, technical, Mergers allocative, pure technical, scale, revenue Technical, scale, pure General level of technical efficiency and productivity, evolution over time Technical Intercountry comparisons
Acquiring firms achieve greater efficiency gains than either target firms or firms not involved in mergers; no beneficial effect of mergers on target firms; M&A driven mostly by solvency objectives
Value added
Technical
Deregulation, regulation change
Decreasing efficiency; increasing productivity
Value added
Technical
Deregulation, regulation change
Still cost-saving potential; increasing divergence between fully efficient firms and efficiency laggards; low cost-savings potential from further mergers
Value added
Technical
Deregulation, regulation change
Still considerable inefficiency; increased productivity
Value added
Technical
General level of Productivity growth; but low compared to real growth of economy efficiency and productivity, evolution over time Scale and scope Multi-product firms more efficient than focused firms economies General level of Industry highly inefficient, with notable differences between different companies efficiency and productivity, evolution over time
Premiums, securities investments Value added DEA Salaries and expenses (1 input) and Premium income, revenue from Value payment to insurers and expenses investment activities added incurred in the production of services (1 input)
Cost Technical
Productivity growth; in P&C due to presence of technically efficient foreign firms
Inefficiencies in both markets; Austrian insurers more efficient than German insurers
38
Qiu/Chen (2006)
China
Rai (1996)
11 countries 106 internationally Germany n/a and UK
Rees et al. (1999)
14-32
20002003
Life
19881992
DEA Labor, equity capital, other
Benefit payments, additions to reserve, yield of investment
Value added
Technical, pure technical, scale
Life incl.health, DFA, Labor, capital, benefits and claims non-life SFA
Premiums (life and non-life)
Value added
Cost
19921994
Life
19901995 19891991 19822001
Life
DFA
Labor, financial capital, materials
Non-life
SFA
Labor
Total premium income and Value change in total premium income added (UK), aggregate sum insured and change in aggregate sum insured (GE) Benefit payments, additions to Value reserves added Number of units produced Physical
Life
DEA Labor, business services, debt capital, equity capital
Present value of real losses Value incurred, ratio of liquid assets to added liabilities
Technical, allocative, cost, scale
19901997 19761980
Life
SFA
Claims, additions to reserves
Cost, revenue, profit
Life
Index Labor (supervisor, agent, other); materials; capital (home office, field)
DEA Distribution cost, administration cost
Ryan/Schellhorn (2000) US
321
Toivanen (1997)
Finland
21
Tone/Sahoo (2005)
India
1
Ward (2002)
UK
44
Weiss (1986)
US
2
Weiss (1991a)
US
100
19801984
Propertyliability
SFA
Weiss (1991b)
CH, F, GE, n/a J, US Australia 46
19751987 1998
Propertyliability General insurance
Index
Wu et al. (2007)
Canada
71-78
19961998
Life incl. health DEA
Yang (2006)
Canada
72
1998
Life incl. health DEA
Yao/Han/Feng (2007)
China
22
19992004
Life, non-life
Yuengert (1993)
US
765
1989
Worthington/Hurley (2002)
DEA
DEA
Labor, capital
Number of policies, constant dollar insurance in force, real premium
Value added Physical
Technical
Cost Cost
General level of Average technical efficiency declining over time; increasing returns to scale efficiency and productivity, evolution over time Intercountry compari- Firms in Finland and France with lowest inefficiency; firms in UK with highest; sons small firms more cost efficient than large firms; specialized firms more cost efficient than combined firms Deregulation, regula- Looser regulation and increased competition increase efficiency tion change
Deregulation, regulation change Scale and scope economies General level of efficiency and productivity, evolution over time Distribution systems
n/a
Methodology issues, comparing different techniques or assumptions Labor (agent, supervisory, nonsuper- Incurred losses, reserves Value Technical, allocative, General level of visory), material, capital added scale efficiency and productivity, evolution over time Labor, capital Incurred losses, reserves Value n/a Intercountry compariadded sons Labor, information technology, Net premium revenues, invested Value Pure technical, scale, General level of physical capital, financial capital assets added allocative, cost efficiency and productivity, evolution over time Prod: Labor expenses, general Prod: Net premiums written, net Value Systematic, technical Methodology issues, operating expenses, capital equity, income added/finan (production), investcomparing different claims incurred Inv: Investment gains in bonds cial interment techniques or assumpInv: Net actuarial reserves, investand mortgages, investment gains mediation tions ment expenses, total investments, in equities and real estate total segregated funds Prod: Labor expenses, general Prod: Net premiums written, net Value Systematic, technical Methodology issues, operating expenses, capital equity, income added/finan (production), investcomparing different claims incurred Inv: Investment gains in bonds cial interment techniques or assumpInv: Net actuarial reserves, investand mortgages, investment gains mediation tions ment expenses, total investments, in equities and real estate total segregated funds Labor, capital, payment and benefits Premiums, investment income Value Technical General level of added efficiency and productivity, evolution over time Labor, physical capital Reserves, additions to reserves Value Cost, scale Scale and scope added economies
Life incl. SFA, accident and TFA health Abbreviations: DEA: data envelopment analysis; DFA: distribution-free approach; FDH: free-disposal hull; Index: index numbers; SFA: stochastic frontier approach; TFA: thick frontier approach.
Unchanged efficiency levels after RBC became effective Diseconomies of scale at firm and economies of scale at branch level; economies of scope in production Increasing allocative inefficiencies after 1994; increase in cost efficiency in 2000
Cost benefits for firms focusing on one mode of distribution Theoretically sound method for measuring productivity in life insurance industry has been introduced
Estimated inefficiency costs of 12–33% of premiums
Japan with weakest productivity growth; US and Germany with overall high productivity Low average level of efficiency; mostly due to allocative inefficiency
New model allows integration of production performance and investment performance; Canadian companies operated very efficiently
New model allows integration of production performance and investment performance; Canadian companies operated fairly efficiently
Average efficiency of 0.77 for non-life and 0.70 for life companies
Economies of scale, but not for whole sample; x-inefficiency 35–50%
39
