FIRM PERFORMANCE IN THE CHINESE INSURANCE INDUSTRY

FIRM PERFORMANCE IN THE CHINESE INSURANCE INDUSTRY September 20, 2004 Tyler Leverty* Georgia State University Yijia Lin Georgia State University Hao Zhou Allianz General Representative Office *Corresponding Author: Tyler Leverty Department of Risk Management and Insurance Georgia State University P.O. Box 4036, Atlanta, GA 30302-4036 Tel: 404-463-2735 E-mail: insjtlx@langate.gsu.edu

ABSTRACT FIRM PERFORMANCE IN THE CHINESE INSURANCE INDUSTRY * The Chinese insurance industry has undergone tremendous change in the short period since the state monopoly was dissolved and foreign owned insurers were allowed into the market. Little empirical research has analyzed the ramifications of these changes. In this paper we utilize a unique dataset that enables us to conduct an in-depth analysis of the efficiency and productivity of the Chinese insurance industry after the changes in the market. We observe an annual average productivity growth of 15.8% over the sample period for property-casualty insurers, with the major benefits of the progress being realized by the domestic insurers. Additionally, we discover that regulatory restrictions on foreign insurer product diversity and geographical dispersion inhibit foreign propertycasualty firm s efficiency. In the life insurance market, the average annual productivity growth is 24.7% over the sample period. However, foreign and joint-venture organizations are not observed to be the principal causes of the increased productivity over our sample period. Overall, the results are consistent with there being significant increases in social welfare following the liberalization of the insurance market in China. INTRODUCTION The Chinese government set up the People's Insurance Company of China (PICC) in 1949. During its first decade of operations, PICC made steady progress establishing regional branches and offices in a majority of the provinces, autonomous regions and municipalities. By 1956, all other insurance companies in the People s Republic of China (PRC) ceased operations and foreign insurance companies were required to leave China. Thus PICC became the state-monopoly insurer and the nationalization of the industry was achieved, i.e. the State ultimately bore the risks of most hazards against which insurance would normally be sought in other countries. Political upheaval and state planning, however, hindered the development of the domestic industry. In fact, the business of insurance was openly denounced by leftist ideologues in 1958 and as a result domestic insurance was virtually suspended. China s Reform and Open policy in 1979, initiated a series of reforms with the purpose of stimulating economic growth. As a result of these policies, China slowly reinvigorated its insurance industry. PICC was separated from the People s Bank of China (PBOC) in 1982, making PICC the sole state-owned independent insurance company. The state-owned monopoly was disrupted in 1988 with the * The authors would like to thank Martin F. Grace for all his comments, suggestions, and advise. We are also grateful to the participants at the Georgia State University research seminar for helpful comments. 2

entrance of a shareholder-owned insurer, Ping An Insurance Company. AIG entered the market in 1992, officially signaling the opening of the Chinese market to foreign insurers. Another event stimulated by the late 1970 s economic reform was the establishment of The Insurance Law of the People s Republic of China in 1995. The provisions of the law required insurance companies to underwrite either property insurance or life insurance, not both at the same time. As a result, the state-sponsored PICC was restructured into a group company with three subsidiaries. In addition the reform movement established the China Insurance Regulatory Commission (CIRC) 1998. The CIRC took supervisory control of all insurance companies from the PBOC. With pressure from the international insurance community, the CIRC encouraged China s government to issue licenses to new domestic and foreign insurers. Finally, China s transformation continued with their membership in the WTO becoming official on December 11, 2001. Overall, the Chinese insurance market has undergone considerable change in a relatively short period of time. The feeling in China is the transformations in the insurance market, in particular the emergence of powerful local players and the arrival of foreign insurers, has been the principle source of the growth in the Chinese insurance industry. In fact, many researchers provide anecdotal evidence that the emergence of foreign insurers and new domestic insurers has already led to benefits to the market, such as additional capital, advanced technology, experienced management, and new approaches to distributing insurance products (Shen, 2000; Sun, 2003; D Arcy and Xia, 2003). The internationalization of an industry and the emergence of new firms are expected to bring pressure on all firms in the industry and increase operational efficiency. Specifically, reducing restrictions on foreign insurer entry may improve the domestic insurance industry directly and indirectly. Foreign insurers may directly bring new and better management techniques, skills, training procedures, and technology to the domestic market and may indirectly enhance domestic insurer efficiency by stimulating competition in the domestic market. The overall impact (direct and indirect) is a reduction in insurer overhead expenses. 3

There is an extensive literature detailing the impact of liberalization and deregulation on bank performance (e.g. Demirguc-Kunt, Levine, and Min, 1998; Claessens, Demirguc-Kunt, and Huizinga, 2001; Demirguc-Kunt, Laeven, and Levine, 2002; and Clark, Cull, Martinez-Peria, and Sanchez, 2003). 1 The general findings are that greater foreign bank presence tends to increase domestic banking system efficiency, which implies that liberalizing restrictions of foreign bank entry enhances domestic banking efficiency with positive implications for total factor productivity (TFP) growth. The liberalization and deregulation of developing insurance markets has also led researchers to investigate insurer efficiency improvements. Boonyasia, Grace and Skipper (2004) use DEA to examine the impacts of liberalization and deregulation on four life insurance markets: Korea, Philippines, Taiwan and Thailand. Their results suggest that in a restrictive regulatory environment, welfare gains will be minimal if deregulation does not closely follow liberalization. Jeng and Lai (2003) also use DEA to explore efficiency and solvency issues in the Chinese insurance industry. 2 Jeng and Lai (2003) utilize two different approaches for measuring insurance company output and each approach leads to a different conclusion. Using the production (value-added approach) for output measurement they discover that state-owned insurers are more efficient than stock insurers. However, when they utilize the flow (or financial intermediation) approach they find that state owned insurers are less efficient than stock insurers and composite insurers are more efficient than specialized insurers. 3 Overall, due to scarcity of data, little is known about the efficiency and productivity gains of the firms in the Chinese insurance market during this period. We, however, have a unique dataset that enables us to conduct a more in-depth analysis of the efficiency and productivity growth over the 1995-2002 1 Levine (2001) provides a survey of this literature. 2 There are a couple of limitations of the Jeng and Lai (2003) paper. First, they conduct the DEA analysis with property-casualty and life insurers in the same sample. Combining both PC and life insurers into the same DEA model may result in misleading conclusions because they are producing different products. Second, Jeng and Lai (2003) utilize only univariate statistics due to their limited sample size and it does not all allow them to separate out potentially confounding market effects. 3 One way to think about these two approaches is the value-added approach regards insurers as profit-maximizing producers of financial services such as risk-bearing, real insurance services, and intermediation (see Cummins and Weiss, 2001); whereas the flow approach considers insurers as pure financial intermediaries (transformers of liabilities into assets) whose principal objective is the maintenance of firm solvency for its policyholders and employees and for the regulators of insurance (see Brockett, et al., 2004). 4

period for property-casualty insurers and the 1999-2002 period for life insurers. We estimate total technical efficiency, purely technical efficiency, and scale efficiency using data envelopment analysis (DEA). In addition, we utilize the Malmquist approach, which is a DEA-based technique, to measure the evolution of productivity and efficiency of Chinese insurers over time. Our results suggest that the presence of technically efficient foreign firms in the property-casualty insurance market has led to productivity gains for all the firms in the industry. In fact, there was a 15.8 % annual productivity growth over the 1995-2002 period with the major benefits being realized by the domestic insurers. However, we discover that Chinese regulatory restrictions on foreign insurer product diversity and geographical dispersion inhibit foreign PC insurers pure technical efficiency. Furthermore, in the Chinese life insurance industry, annual productivity growth over the 1999-2002 period is 24.7%. Though, we do not discover foreign and joint-venture organizations to be the principal causes of the increased productivity in our life insurer sample. Our paper is organized as follows. In Section 2 we review the Chinese insurance industry. In Section 3 we will briefly discuss the methodologies that we will be using to estimate efficiency and productivity. Section 4 contains the explanation and the results of our empirical procedures that analyze whether the rapidly changing Chinese insurance industry (the dissolution of the state monopoly, the presence of shareholder-owned insurers, the presence of foreign insurers, the accession of China into the WTO, etc.) has resulted in efficiency gains. Section 5 concludes the paper. OVERVIEW OF THE CHINESE INSURANCE MARKET Governments worldwide continue to take steps to deregulate and/or liberalize insurance markets to enhance consumer choice and welfare. 4 In the Chinese insurance market, the last twenty-five years have brought about numerous; however, few changes in national regulation of insurance have accompanied the modest changes in market access restrictions. In fact, China s moderate liberalization efforts were not accompanied by any attempts to deregulate the insurance industry. 4 Deregulation is generally defined as the lessening of national regulation. Whereas liberalization is the reduction of government and/or other barrier to market access, especially as it relates to foreign firms. 5

The most significant liberalization acts in the Chinese insurance market were the disruption of the state-owned monopoly in 1988, the initial opening of the market to foreign insurers in 1992, and the stepup in the issuance of licenses to new domestic and foreign insurers in 1998. Despite these liberalizing acts, the property-casualty and life insurance industries are still highly concentrated and relatively little business is written by foreign insurers (see Table 1, Panels A and B). For example, in 2002, two firms wrote 83.4% of the total property-casualty business and two firms wrote 80.1% of total life business in China. While the Herfindahl index in 2002 was 5,236 for the property-casualty insurance industry and 3,900 for the life insurance industry. Table 1 Market Concentration Property-Casualty Insurance Industry Life Insurance Industry Year Number Concentration Ratios HHI a Number Concentration Ratios HHI a of Total Premiums Written (%) of Total Premiums Written (%) Firms 1-Firm 2-Firm 4-Firm 8-Firm Firms 1-Firm 2-Firm 4-Firm 8-Firm 1995 8 61.6 94.1 97.7 100.0 4,862 1996 8 58.3 93.5 97.3 100.0 4,651 1997 13 53.3 90.0 94.1 99.0 4,206 1998 14 79.4 90.4 98.2 99.5 6,482 1999 15 79.4 89.9 98.1 99.4 6,462 12 68.6 88.9 97.5 99.9 5,179 2000 17 77.0 88.3 97.0 99.0 6,128 13 65.0 87.3 97.2 99.8 4,802 2001 19 74.1 86.6 96.9 99.0 5,738 16 55.1 82.0 96.9 99.7 3,945 2002 21 70.3 83.4 95.5 98.5 5,236 20 56.6 80.1 94.6 99.3 3,900 a Herfindahl-Hershman Index In preparation for (and as a result of) accession into the WTO, China expanded the number of licenses given to insurers and increased the allowed ownership levels for foreign investors (Ng and Whalley, 2004). As a result, the number of new foreign insurance companies (or joint ventures) entering the market has been rising faster than the number of new domestic companies, especially in the life business. Specifically, the number of foreign-owned life insurers and joint ventures increased from 1 in 1992 to 11 in 2002. While only six new domestic life insurers entered the market during this period. However, despite the fact that the number of foreign participants in the insurance market has been steadily increasing, the market share of foreign insurers is still extremely limited (see Table 2). In 2002 6

foreign insurers only had 0.8% of the Chinese property-casualty market and 1.54% of the life insurance market (0.4% for joint-venture firms and 1.14% for foreign insurers). The principal reason for the low market share of foreign insurers is regulatory restrictions. Specifically, China imposes strict restrictions on foreign company ownership and on the regions in which foreign companies can operate. 5 In addition regulators mandate reinsurance cession for foreign insurance companies and require other business restrictions. 6 Moreover, rate and product regulation is in the Chinese insurance market extensive (Sun, 2003). Another potential barrier to foreign insurer market share is the complexity of business in China. Wu and Strange (2000) state that foreign insurance companies see the Chinese market as complex due to China s unusual organizational forms, complicated legal arrangements, different business customs, and the need for Chinese language skills to conduct business. The WTO, however, mandates that China have open markets by 2007, which may ultimately alter China s strict regulatory practices. In fact, Sun (2003) states that [R]eform of the insurance regulation system is the key to reforming China s insurance market The most important impact of WTO accession will be on the government, meaning that the government must get rid of anything inconsistent with WTO rules and reposition itself in a market economy. Specifically, WTO guidelines state that three years after the WTO accession foreign life companies will be able to hold a 51% ownership claim and full ownership will be permitted five years after accession 7. Other restrictions that will be removed as a result of China s entrance into the WTO are the regional and life business restrictions (after three years of WTO accession), the non-life business restrictions (after two years), and the mandatory purchase of reinsurance from China Re the Chinese domestic reinsurance company (four years from the accession). Therefore, the period after December 5 Prior to December 2003 foreign property-liability insurers could only underwrite the risks of foreign-invested enterprises located in the insurer s local geographical area, and foreign life insurers could only sell individual life polices and were banned from selling group life products. 6 For example, foreign insurance companies still can not sell mandatory auto third-party liability insurance, public transportation vehicle liability and commercial auto and contract mover liability insurance in China. 7 Currently, with the exception of American International Assurance (AIA), 100% foreign ownership is not allowed in the life insurance market. In 1992, Shanghai was selected by the Chinese government as an experimental zone to test the impact of foreign participation, and AIA became the first foreign insurer to be granted a license to underwrite both life and non-life insurance (Wu and Strange, 2000). After that, no new foreign fully-owned life insurance company was given license to operate in China. 7

2001 is a transition phase for Chinese insurance regulation. The process of opening the Chinese insurance market follows a step-by-step procedure and some of the current restrictions will be totally removed by the end of 2004. Table 2: Market Share in the Chinese Insurance Market Panel A: Foreign and Domestic Firms in the Chinese PC Insurance Industry Year Foreign Firms Domestic Firms Number Market Share (%) Number Market Share (%) of Firms (Total Premiums Written) of Firms (Total Premiums Written) 1995 3 1.39 5 98.61 1996 2 0.78 6 99.22 1997 4 2.32 9 97.68 1998 5 0.65 9 99.35 1999 6 0.67 9 99.33 2000 7 0.84 10 99.16 2001 9 0.72 10 99.28 2002 10 0.80 11 99.20 Panel B: Foreign, Joint Venture, and Domestic Firms in the Chinese Life Insurance Industry Year Joint Venture Firms Foreign Firms Domestic Firms Number Market Share (%) Number Market Share (%) Number Market Share (%) of Firms (Total Premiums Written) of Firms (Total Premiums Written) of Firms (Total Premiums Written) 1999 3 0.11 2 1.50 7 98.38 2000 3 0.21 3 1.70 7 98.08 2001 5 0.46 3 1.35 8 98.19 2002 7 0.40 4 1.14 9 98.46 Nevertheless, much doubt has been raised about the feasibility of implementing the WTO mandated changes in such a short amount of time. Indeed, the starting point for these policy changes seems so highly restricted that even threats of eventual retaliation from WTO partners may not be enough of an impetus to speed things along (Whalley, 2003). Furthermore, membership of the WTO is not a magic formula; it will not automatically erase those prejudicial/cautionary measures that kept China s market closed for such a long time. In spite of the strict governmental restrictions and accompanying low market share, foreign insurers do have a wide presence in the marketplace through representative offices and 8

insurance companies, 8 which may influence the potential/perceived competition in the marketplace. MODEL AND METHODOLOGY This section briefly discusses our unique database and explains the inputs and outputs used in our analysis. The section concludes with a short description of the DEA and Malmquist methodologies utilized in our analysis. The Data The data used in this study are drawn from the Chinese Insurance Yearbook. Some of the companies have subsidiaries or branches and these subsidiaries or branches report their own financial statements (e.g. AIA Shanghai, AIA Guangzhou). In this case, we treat each accounting reporting entity as a business entity. In our sample we have all the firms in the two industries for which there is adequate data. For each year, the number of firms in our property-casualty and life samples is shown in Table 2. All of financial statements in the yearbooks are reported in the local currency RMB. Outputs and Inputs Output Measurement Insurer s outputs are primarily intangible financial services which make it necessary to find suitable proxies for the volume of services. Consistent with most of the recent literature on financial institutions, we adopt a modified version of the production (or value-added) approach to identify the important outputs. The production approach employs as important outputs all asset and liability categories that have substantial value-added, as judged by operating cost allocations (Berger and Humphrey, 1992). Operating expense allocations identify three principal services that insurers provide (Cummins and Weiss, 2001): 8 By the end of 2000, foreign insurance companies from 17 countries have set up 196 representative offices in China in an attempt to obtain a business license (Wang and Lin, 2001). Representative offices are permitted to participate only in non-business activities like liaison work, market research and technical exchanges. The motivation for establishing a representative office is the requirement that a foreign insurer must maintain a representative office for at least two years prior to being eligible to apply for a license to conduct formal business operations. 9

Risk-pooling and risk-bearing: The main function of insurance is to resolve risk and uncertainty. Insurance provides a mechanism through which consumers and businesses exposed to losses can engage in risk reduction through the diversification effect of pooling. Pooling is the collection of premiums in advance from customers and redistributing most of these funds to the policyholders that sustain losses. The actuarial, underwriting, and related expenses incurred in operating the risk pool are a principal component of value added in the insurance industry. Furthermore, the equity capital that insurers hold also creates value-added by increasing economic security as a result of the cushion it provides against unexpected losses and investment shocks. Real financial services relating to insured losses: Insurers provide a variety of real services for policyholders such as the design of risk management programs (i.e. risk surveys and recommendations regarding coverage, deductibles, and policy limits), loss prevention, financial planning, the provision of legal defense in liability disputes, and administration of group life, annuity and health insurance plans. By contracting with insurers to provide these services, policyholders can take advantage of insurers expertise to reduce the costs of managing risk. Financial intermediation: Insurers issue insurance policies, a type of debt contract, and invest the funds in financial assets until they are needed to pay claims or fund withdrawals. In return, policyholders receive a discount in the premiums they pay to compensate for the opportunity costs of the funds held by the insurer. For life insurers, financial intermediation is a principal function, accomplished through the sale of asset accumulation products such as annuities. For property-casualty insurers, financial intermediation is a somewhat incidental function resulting from the collection of premiums in advance of claims payment to minimize contract enforcement costs. Insurers value-added from intermediation is represented by the net interest margin between the rate of return earned on invested assets and the rate credited to policyholders. In defining measures for insurance output, we are searching for proxies for the quantity of insurance services provided. Accordingly, the output variables should be highly correlated with the quantity of financial services provided. Since the products offered and the data reported by life and property-casualty insurers differ significantly, different sets of output definitions are used for each. For life insurance outputs, the most recent insurance efficiency research uses incurred benefits plus additions to reserves (Yuengert, 1993; Cummins, Tennyson, and Weiss, 1999; and Berger, et al, 2000). Incurred benefits are payments received by policyholders in the current year. Incurred benefits are useful proxies for the risk-pooling and riskbearing functions since they account for the amount of funds pooled by insurers and redistributed to policyholders as compensation for insured events. Most life insurance products entail the accumulation of assets to pay future death benefits. The funds received that are not needed for benefit payments and expenses are added to policyholder reserves. Thus, additions to reserve should be highly correlated with 10

the intermediation output. Both incurred benefits and additions to reserves are correlated with real services provided by insurers, such as benefit administration and financial planning. Premium income has also been used as a proxy for the risk-bearing and real insurance services output in insurance efficiency studies (Houston and Simon, 1970; Fecher et al, 1993; Gardner and Grace, 1993; Grace and Timme, 1992; Rai, 1996; Donni and Fecher, 1997; Hardwick, 1997; Kim, 2002; and Boonyasia, Grace, and Skipper, 2004). Premiums are viewed as including the flow of services to insureds. Although premiums capture the flow of services to policyholders, they also include a component for expenses and profits. In fact, premium income is really a form of revenue (price times quantity), not the quantity of output (Yuengert, 1993). As such, systematic differences in price across insurers may lead to misleading inferences. Furthermore, Doherty (1981) critiqued the use of premiums because it results in simultaneous equation bias. Constraints imposed by the data for Chinese insurers require us to utilize net premiums written as the output. Although the utilization of premium income results in simultaneity bias, we believe that premium income, as an output measure, is appropriate under the assumption that life insurance is a homogeneous product and that all insurers charge the same price. This assumption is not very restrictive since life insurance is in fact a fairly homogenous product and because the Chinese government strictly regulates insurance prices. Consequently, net premiums written, representing risk-bearing and realinsurance services is not inappropriate given our data limitations. Because the products offered by life insurers differ in the risk-pooling and real service components of output, we categorize the life insurance product into group and individual life products. For property-casualty insurers, the most common proxy for the quantity of risk-pooling and real insurance services is the present value of real losses incurred (Berger, Cummins, and Weiss, 1997; Cummins, Weiss, and Zi, 1999; and Cummins and Weiss, 2001). Losses incurred are defined as the losses that are expected to be paid as a result of providing insurance coverage during a particular period of time. Because the objective of risk-pooling is to collect funds from the policyholder pool and redistribute 11

them to those who incur losses, proxying output by the amount of losses incurred is appropriate. In addition, the use of losses incurred is consistent with the economic theory of insurance risk-averse agents subject to random shocks to wealth are willing to pay more than the expected value of loss in exchange for transferring risk to the insurer. Losses are also an excellent proxy for the quantity of real services provided, since the amount of claims settlement and risk management services are also highly correlated with loss aggregates. To capture the different types of services provided by the main types of property-liability insurance, we use as separate output measures personal lines short-tail losses, personal lines long-tail losses, commercial lines short-tail losses, and commercial lines long-tail losses. 9 Since the payout characteristics vary amongst the principal types of insurance, the use of present values is typically used to recognize differences in payout tails by line of insurance (Berger, Cummins, and Weiss 1997; Cummins, Weiss, and Zi, 1999; and Cummins and Weiss, 2001). Due to the constraints of our database, we are not able to estimate cash flow patterns; however, since the vast majority of insurance written in China is of the short-tail variety, 10 the inability to estimate cash flow patterns is not a serious limitation. In addition to the risk-bearing and real insurance services, we also account for the intermediation function of borrowing from policyholders and investing the funds in marketable securities. Consistent with recent insurance efficiency studies (e.g. Berger, Cummins, and Weiss 1997; Cummins, Weiss, and Zi, 1999; and Cummins and Weiss, 2001), we utilize total invested assets for each year as our proxy measure. All outputs are expressed in real terms by deflating to 1995 using China s Consumer Price Index (CPI). 11 9 The tail length refers to the length of the loss cash flow stream. The lines of business definitions are described in Phillips, Cummins, and Allen (1998) and in the line classification in Schedule P of the U.S. National Association of Insurance Commissioners (NAIC) regulatory annual statement for property-liability companies. 10 Over the 1995-2002 time period short-tail personal lines premiums written comprised 83.6% of total premiums written, 13.8% of total premiums written consisted of short-tail commercial lines, and the remaining 2.6% of total premiums written was split between the two long-tail line groupings 11 CPI data comes from the "Annual Statistics Communique of National Economy and Social Development ", National Bureau of Statistics of China, http://www.stats.gov.cn/. 12

Input Measurement Inputs are usually easier to identify and measure relative to outputs since the units of measurement are more tangible and directly observable. Additionally, insurance inputs, unlike outputs, tend to be similar for life and property-casualty insurers, so the same input definitions are used for both types of insurers. Insurer inputs are most commonly classified into three broad groups: labor, business services and materials (including physical capital), and capital (Cummins and Weiss, 1993; Gardner and Grace, 1993; Cummins, Tennyson, and Weiss, 1999; and Berger, et al, 2000; Cummins and Weiss, 2001). Our database precludes us from further classifying the three broad input categories into more exact categories. In fact, we combine administrative labor, agent labor, and business services and materials into a single category, business expenses. We are able to further group capital into financial equity capital and debt capital. Financial equity capital is considered an important input in the theory of the firm and financial institutions studies (McAllister and McManus, 1993; Berger, Cummins and Weiss, 1997; Hughes and Mester, 1998; and Hughes, Mester and Moon, 2001). Besides satisfying regulatory requirements, the inclusion of financial equity capital is warranted under the modern theory of the firm where a firm s technology includes all the contractual relationships which encompass the firm. In addition, the financial theory of insurance pricing, views insurance as risky debt in which the financial equity of the insurance company plays a critical role in reducing firm s insolvency risks (Cummins and Danzon, 1997). Accordingly, better capitalized insurers should obtain higher prices for their products than riskier firms, ceteris paribus, since more capital implies a higher probability that losses will be paid if losses are higher than expected. In sum, capital levels ultimately affect the revenue and profit of an insurer. The quantity of financial equity capital is defined as the sum of capital and policyholders surplus. Debt capital for insurers is mainly comprised of funds borrowed from policyholders. The rationale for the segmentation of capital into debt capital is that insurers raise debt capital by issuing 13

policies and then transform this capital into invested assets. For life insurers, these funds consist of the aggregate reserve for life policies and contracts, the liability for premium and other deposit funds, and other reserve items. For property-casualty insurers, debt capital includes the sum of loss reserves and unearned premiums reserves. To summarize, we utilize three outputs for life insurers: net premiums written for group and personal lines and real invested assets. Five outputs are used for property-casualty insurers: losses incurred for short-tail personal, long-tail personal, short-tail commercial, and long-tail commercial lines and real invested assets. The same inputs are used for both life insurers and property-casualty insurers-- business expenses, financial equity capital, and debt capital. All variables are deflated to real 1995 terms via China s Consumer Price Index. ESTIMATION METHODOLGY The methodology to estimate efficiency is data envelopment analysis (DEA). DEA is a non-parametric method that compares each firm in the industry to a best-practice efficient frontier formed by as a convex combination of the most efficient firms in the sample. DEA is appropriately named since it truly envelops the entire data set making no accommodation for random noise outside the control of DMU s. In essence, DEA uses a standard linear programming technique to pinpoint peer groups of efficient firms for each firm or decision-making unit (DMU) being evaluated. A firm is fully efficient (efficiency of 1.0) if it lies on the frontier and inefficient (efficiency < 1) if it is not on the frontier, which means that its outputs could be produced more efficiently by another firm or firms. DEA has been widely used to measure efficiency for financial institutions (see Berger and Humphrey, 1997). We adopt DEA for this study for four principal reasons: (1) it is a non-parametric method and thereby it is not necessary to identify a functional form or make distributional assumptions; (2) it is able to handle relatively small sample sizes, which is ideal for analyzing the Chinese insurance market; (3) it allows for convenient decomposition of total technical efficiency (TE) into pure technical 14

efficiency (PTE) and scale efficiency (SE); and (4) the Malmquist technique, which is the standard approach for measuring the evolution of productivity and efficiency over time, is DEA-based. In-depth descriptions of the DEA methodology are provided in Lovell (1993), Charnes, Cooper, Lewin, and Seiford (1994) and Zhu (2003). The DEA methodology is widely utilized and it is also extensively outlined in insurance studies (e.g. Cummins and Zi, 1998; Cummins, Weiss, and Zi, 1999; Cummins, Tennyson, and Weiss, 1999; Cummins and Weiss, 2001; Cummins and Nini, 2002; Boonyasia, Grace, and Skipper, 2004). For each year, technical efficiency is estimated separately for each firm in the sample by solving linear programming problems. Technical efficiency (TE) refers to the ability to avoid waste by producing as much output as input usage allows, or by using as little input as output production allows. TE is measured relative to a constant returns to scale (CRS) frontier. A firm can achieve TE by moving to the CRS frontier. Technical efficiency can be decomposed into pure technical efficiency (PTE) and scale efficiency (SE), where TE=PTE*SE, by solving additional linear programming problems. PTE is measured relative to a variable returns to scale (VRS) frontier, which may have segments where best practice firms operate with increasing returns to scale (IRS), CRS, or decreasing returns to scale (DRS). Therefore, a firm can achieve PTE by moving to the VRS frontier. If the firm is operating in an IRS or DRS region of the VRS frontier, it could further improve its TE by operating with CRS. Firms with PTE=TE are operating with CRS and are thereby scale efficient, SE=1. To distinguish between DRS and IRS, an additional linear programming problem is solved in which firm efficiency is measured relative to a non-increasing returns to scale (NIRS) frontier. If TE does not equal PTE, and PTE equals the NIRS efficiency measure then the firm is operating with DRS. However, if TE does not equal PTE, and PTE does not equal the NIRS efficiency measure then the firm is operating with IRS (Aly et al., 1990). 15

The Malmquist index approach is employed to analyze changes in efficiency and productivity over time. 12 This analysis permits us to examine shifts in the best practice technical frontier over time. If the Malmquist index is greater than one, then there has been total factor productivity progress, and if the Malmquist index is less than one, then there has been a decrease in total factor productivity. The Malmquist approach allows us to part shifts in the frontier, technological change (TC), from the improvements in efficiency relative to the frontier, technical efficiency change (TEC). The product of TC and TEC is total factor productivity change (TFP), which measured by the Malmquist index (Grosskopf, 1993). Similar to the Malmquist index, TEC has a value of one when there is no change in technical efficiency, and has a value greater than or less than one when technical efficiency has improved or declined. Thus if the firm is closer to the frontier in period t+1 than in period t, the ratio will be greater than 1 and if the firm is further from the frontier in period t+1 than in period t, TEC will be less than 1. Like TFP and TEC, TC has a value of one when technical change has not occurred, and has a value greater than or less than one as technical change has been progressive or regressive. In sum, we employ these measures to examine whether the recent rapid changes in the Chinese insurance industry (the dissolution of the state monopoly, the presence of shareholder-owned insurers, the presence of foreign insurers, the accession of China into the WTO, etc.) has resulted in (1) gains in total efficiencies (e.g., technical efficiency, purely technical efficiency, and scale efficiency), (2) growth in productivity changes (total factor productivity, technological change, technical efficiency change), and (3) the realization of a change in productivity. EMPIRICAL RESULTS This section presents the results of our efficiency and productivity of the Chinese property-casualty and life insurance industries. Due to data availability constraints, we focus of property-casualty insurers during the period 1995-2002 and life insurers for the period of 1999-2002. We first present summary 12 Detailed descriptions of the Malmquist index are presented in Grosskopf (1993), Lovell (1993), Charnes, Cooper, Lewin, and Seiford (1994), and Zhu (2003). Furthermore, this technique is commonly used in the insurance literature (e.g. Cummins, Tennyson, and Weiss, 1999; Cummins and Weiss, 2001; Cummins and Rubio-Misas, 2002; and Boonyasia, Grace, and Skipper, 2004) 16

statistics on the results of the DEA and Malmquist analyses. We next conduct a series of regressions to analyze whether the rapid transformation in the Chinese insurance market impacted firm efficiency and productivity. Summary and Univariate Statistics The summary statistics of efficiency for property-casualty insurers are shown in Tables 3. As depicted, the average technical efficiency (TE) for property-casualty insurers is 0.866. This means that the average firm is 86.6% as efficient as the most efficient firm. Domestic property-casualty firms have a slightly greater TE than foreign firms. For pure technical efficiency (PTE), the mean for PC insurers is 0.902, and domestic insurers are marginally more PTE. While scale efficiency (SE) for domestic insurers is slightly lower than scale efficiency for foreign firms. However, none of the differences between foreign and domestic efficiency (TE, PTE, or SE) are statistically significant. The TE Frontier and PTE Frontier variables are counts of the number of times a particular firm is represented in the best-practice reference set for the CRS and VRS models, respectively. If a firm is fully efficient (efficiency=1) then it has only one peer group firm, itself. On the other hand, if the firm is inefficient (efficiency<1) then that firm s efficient frontier is formed by a convex combination of the most efficient firms in the industry. As shown in Panel A in Table 3, the mean number of times a firm is represented in the best-practice frontier is 1.729 and 2.000 for the TE (CRS frontier) and PTE (VRS frontier), respectively. Interestingly, we observe that domestic insurers are significantly (at the 10% level) more likely to be PTE reference firms than foreign insurers. However, a univariate setting may be misleading since the number of domestic insurers outweighs the number of foreign firms. Another key variable in Table 3 is the # of Firms in Year variable, which represents the number of firms in the Chinese insurance market. This variable will be used the regression analysis to control for the possibility that the number of firms in the market influences the efficiency measure. Under DEA, the piecewise-linear convex isoquant constructs the best-practice frontier from the sample. Therefore, the sample size affects the efficiency scores. With a small number of sample firms, efficiency is likely to be 17

inflated, as these firms are likely to define the frontier. As the number of firms in the sample grows, efficiency is likely to decrease because the best-practice firms will lie on the frontier and envelop the rest of the firms. In sum, as the number of firms in a particular year increases, the average efficiency score of the Chinese property-casualty insurance industry is prone to decline. Table 3 Panel A: Summary Statistics of Efficiency Property-Casualty Insurer Sample Variable ALL FIRMS (N=85) DOMESTIC FIRMS (N=44) FOREIGN FIRMS (N=41) Mean StdDev Min Max Mean StdDev Min Max Mean StdDev Min Max TE 0.866 0.190 0.340 1.000 0.876 0.194 0.340 1.000 0.856 0.187 0.413 1.000 PTE 0.902 0.159 0.349 1.000 0.917 0.168 0.349 1.000 0.886 0.148 0.522 1.000 SE 0.957 0.093 0.500 1.000 0.956 0.112 0.500 1.000 0.958 0.069 0.727 1.000 TE Frontier 1.729 2.762 0.000 11.000 1.977 2.905 0.000 11.000 1.463 2.609 0.000 11.000 PTE Frontier 2.000 2.849 0.000 10.000 2.523 3.166 0.000 10.000 1.439 2.377 0.000 10.000 Log Tot Prem Written 4.732 2.143 1.147 10.914 5.710 2.440 1.311 10.914 3.682 1.030 1.147 5.199 # of Firms in Year 13.235 5.075 5.000 19.000 5.500 2.563 2.000 9.000 5.125 3.357 1.000 10.000 Panel B: T-test of Significant Difference between Property-Casualty Foreign and Domestic Companies Variable H0: Domestic > Foreign Difference T-test Statistic TE 0.020 0.470 PTE 0.031 0.920 SE -0.002-0.100 TE Frontier 0.514 0.860 PTE Frontier 1.084 * 1.790 Log Tot Prem Written 2.028 *** 5.050 # of Firms in Year 0.375 0.576 Note: TE is total technical efficiency; PTE is pure technical efficiency; and SE is scale efficiency. TE frontier is the number of times a particular firm is represented in the CRS frontier. PTE frontier is the number of times a particular firm is represented in the VRS frontier. *** Significant at 1% level; ** Significant at 5% level; * Significant at 10% level. In addition to number of firms, the logarithm of total premium written (Log Tot Prem Written) is included to account for the size differences amongst the firms. From the Panel B in Table 3, we observe that domestic firms are significantly larger than foreign insurers (at the 1% level), which is not surprising considering the drastic market share differences that we discussed earlier. Table 4 displays the returns to scale exhibited in the Chinese property-casualty industry. Forty percent of the firms demonstrate increasing returns to scale (IRS), fifty-six percent constant returns to 18

scale (CRS), and only four percent decreasing returns to scale (DRS). Due to the additional legal regulatory constraints put on foreign firms, 13 foreign firms are more likely to have IRS and less likely to have constant and decreasing returns to scale. However, the differences in the returns to scale between domestic and foreign companies are not significant. Table 4 Returns to Scale Property-Casualty Insurer Sample (1995-2002) Variable All Firms Domestic Foreign H 0 : Domestic > Foreign # of Firms % of Firms # of Firms % of Firms # of Firms% of Firms Difference T-test Statistic IRS 34 40% 15 34% 19 46% -0.1225-1.15 CRS 48 56% 27 61% 21 51% 0.1014 0.94 DRS 3 4% 2 5% 1 2% 0.0211 0.53 Note: IRS is Increasing Returns to Scale; CRS is Constant Returns to Scale; and DRS is Decreasing Returns to Scale; *** Significant at 1% level; ** Significant at 5% level; * Significant at 10% level. The property-casualty insurers summary statistics for the Malmquist analysis of productivity are located in Panel A in Table 5. Total factor productivity (TFP) represents the average annual productivity growth in a period and as such it can be viewed as an indirect measure of the social welfare gains in the industry over the sample periods. Therefore, the mean TFP of 1.158 over the 1995-2002 period demonstrates that the average annual productivity growth (or welfare gain) in the property-casualty insurance industry was 15.8%. The considerable productivity growth over the period is due, in almost equal parts, to technological progress (increases in the production frontier TC) and improvements in technical efficiency (better use of inputs TEC). The mean technological change (TC) is 7.6% and the average technical efficiency change (TEC) is 10.7%. Overall, the changes in the property-casualty insurance industry in China over the period such as increased participation in the market by 13 The restrictions on the foreign insurance companies before China entered the WTO on December 11, 2001 include the requirement that the foreign insurer must establish a representative office in China for two years prior to writing business, a ban on wholly foreign-owned subsidiaries, and restriction on the locations of where they can operate the business. Furthermore, foreign property-casualty companies are currently only allowed to underwrite property coverage, limited lines of liability coverage, and credit insurance for subsidiaries of foreign companies operating in China. 19

new domestic firms and foreign firms and China s recent introduction into the WTO led to considerable progress in the industry. Table 5 Panel A: Summary Statistics of Productivity Property-Casualty Insurer Sample (1995-2002) Variable ALL FIRMS (N=61) DOMESTIC FIRMS (N=32) FOREIGN FIRMS (N=29) Mean StdDev Min Max Mean StdDev Min Max Mean StdDev Min Max TC 1.076 0.490 0.318 3.156 1.062 0.453 0.318 2.151 1.091 0.536 0.352 3.156 TEC 1.107 0.418 0.413 2.944 1.127 0.447 0.539 2.944 1.085 0.392 0.413 2.423 TFP 1.158 0.601 0.313 3.376 1.144 0.512 0.318 2.365 1.175 0.696 0.313 3.376 Log Average TPW 4.852 2.100 1.407 10.875 5.754 2.453 1.407 10.875 3.858 0.917 1.476 5.108 # of Firms in Period 11.984 4.522 3.000 17.000 4.571 2.820 0.000 8.000 4.143 3.185 1.000 9.000 Panel B: T-test of Significant Difference between Property-Casualty Foreign and Domestic Companies Variable H0: Domestic > Foreign Difference T-test Statistic TC -0.029-0.230 TEC 0.042 0.390 TFP -0.031-0.200 Log Average TPW 1.896 *** 4.070 # of Firms in Period 0.429 0.554 Note: TC is Technological Change; TEC is Technical Efficiency Change; and TFP of the Malmquist Index of Total Factor Productivity; *** Significant at 1% level; ** Significant at 5% level; * Significant at 10% level. There are, however, no significant differences between domestic and foreign insurers with respect to TC, TEC, or TFP; however, domestic firms are significantly larger than foreign firms, as measured by the log of the average total premiums written between years t and t+1. The summary statistics of efficiency for life insurers are shown in Panel A in Table 6. The average TE, PTE, and SE for the life insurer sample over the 1999 to 2002 are 0.976, 0.989, and 0.987, respectively. Domestic and foreign insurers have significantly greater PTE than joint venture firms (at the 5% level), which suggests that the joint-venture relationship between a foreign insurer and a domestic insurer is not as purely technically efficient as a solely owned and operated domestic or foreign insurer. However, the foreign insurer efficiency results may not be representative of all foreign insurers, since 20

AIG is the only foreign insurer allowed to operate in China s life insurance. 14 Domestic firms have significantly greater TE than joint-venture firms (Panel B). The average number of times a firm is represented in the best-practice frontier is 1.627 for TE Frontier and 1.431 for PTE Frontier. Interestingly even though the average number of foreign firms over the period is less than that of the joint-venture firms, the foreign firms are still significantly more likely to be represented in the PTE frontier (at the 10% significance level). While domestic insurers are significantly more likely to be represented in the TE frontier than joint-venture firms, this result may be misleading since the number of domestic firms is greater than the number of joint-venture firms (although not significantly greater). Furthermore, similar to the results property-casualty insurers, domestic life insurers are significantly larger than domestic firms (at the 5% level) and joint-venture firms (at the 1% level). Additionally, foreign insurers are significantly larger than joint-venture firms (at the 1% significance level). Panel A in Table 7 displays the returns to scale summary statistics in the life insurance industry. Eighteen percent of the life insurers in the sample exhibit IRS, seventy-five percent demonstrate CRS, and only eight percent reveal DRS. Interestingly no domestic firms exhibit IRS, while six joint ventures and one foreign firm do. Similar to property-casualty insurers, foreign life insurers (and joint-venture firms) face regulatory restrictions on the types of business they can write, which contributes to there inability to realize efficient returns to scale, i.e. CRS. 15 Furthermore, the differences in IRS between domestic and joint-venture insurers and domestic and foreign firms are significant at the 1% and 5% levels, respectively (Panel B). In addition, domestic firms are significantly more likely to have CRS than joint-venture firms (at the 5% level), and foreign firms are also more likely to exhibit CRS than jointventure firms (at the 1% level). Finally, domestic firms exhibit DRS significantly more than foreign firms (at the 10% level). 14 See footnote 7. 15 Foreign life companies are only allowed to provide individual (nongroup) life insurance to both Chinese and foreign citizens. 21

The life insurance summary statistics for the Malmquist productivity index are shown in Panel A in Table 8. The mean TFP of 1.247 illustrates that extensive total factor productivity progress (welfare gain) has been made over the period 24.7% average annual productivity growth over the 1999 to 2002 period. The productivity growth is due primarily to improvements in technological progress (TC), i.e. increases in the production frontier, rather than gains in technical efficiency (TEC). Specifically, TEC observed a modest 2.7% average annual growth rate while TC noticed a 21.8% average annual growth rate. There are no significant differences between domestic and foreign life insurers with respect to TC, TEC, or TFP (Panel B); however, domestic firms are significantly larger than foreign firms, as measured by the log of the average total premiums written between years t and t+1. Furthermore, foreign firms have experienced significantly greater TC and TFP over the period in comparison to joint-venture firms (both at the 5% significance level). Additionally, domestic firms are significantly larger than jointventure firms (1% significance level) and foreign firms have significantly greater size than joint-venture firms (5% significance level). 22