Life insurance lapse behaviour: evidence from China
Lu Yu2• Jiang Cheng1•Tzuting Lin3
Received: 21 May 2017 / Accepted: 3 January 2018 The Geneva Association 2018
Abstract We investigate determinants of lapse rates in the Chinese life insurance industry using firm-province level data from 2005 to 2013. We conduct a panel study, exploring significant regional differences and dramatic changes in demo-graphic conditions during the urbanisation process in China. First, we find that the unemployment rate is positively related to lapse rates, and the driving force is migrant population rather than local urban residents. This provides evidence for the emergency fund hypothesis from a new perspective. Second, we find that an insurer’s reputation is negatively linked to lapse rates. We define this as the ‘rep-utation hypothesis’. Third, our findings are also consistent with the interest rate hypothesis. We extend the literature by decomposing life insurance products into three types and find that interest rates are positively (negatively) associated with lapse rates of investment-type (protection-type) products. Lapse rates of health products are not related to interest rates. Fourth, our empirical result suggests that high lapse rates can potentially weaken the insurers’ financial soundness and harm new business.
Keywords Lapse rate Migrants Reputation risk Financial soundness
& Jiang Cheng jiangcheng@ln.edu.hk
1 Department of Finance and Insurance, Lingnan University, Tuen Mun, Hong Kong 2
Department of Finance of Henan Province, Zhengzhou, China
3
Department of Finance, College of Management, National Taiwan University, Taipei, Taiwan https://doi.org/10.1057/s41288-018-0104-5
Introduction
This paper studies the determinants of life insurance lapse rates in China. We also investigate the consequences of lapse on insurers’ financial soundness. Lapses in life insurance have been an area of intense academic interest since the 1970s (Kiesenbauer 2012). Lapse is important and constitutes the largest submodule in terms of solvency capital requirements within the life underwriting risk module under the SolvencyII framework. Regardless of the theoretical models or empirical evidence, existing research mainly focuses on developed countries in which life insurance markets have reached maturity. Lapse rates have not been studied empirically in the Chinese life insurance market. To the best of our knowledge, this is the first paper to study lapse rates in life insurance using Chinese data. Moreover, examining insurance lapse behaviour in China could be useful for other emerging markets.
The background of this paper is that China is facing a severely high lapse problem. Over the past three decades, the huge population base and the rapid economic development provided a unique opportunity for the Chinese insurance industry to grow. Now, China accounts for 6% of the world’s premium volume and more than one third of the emerging markets’ premiums (Swiss Re2014). However, we have also witnessed a serious lapse trend in the Chinese life insurance market in recent years. In Fig.1a, b, we present the industry lapse value and lapse rates respectively from 2005 to 2013 issued by the China Insurance Regulatory
2005 2006 2007 2008 2009 2010 2011 2012 2013
Lapse Value (billion) 48.80 54.64 92.20 96.62 106.97 115.87 95.78 119.81 190.66 0.00
40.00 80.00 120.00 160.00
200.00 Lapse Value (billion)
2005 2006 2007 2008 2009 2010 2011 2012 2013 Lapse Rate㸦%㸧 4.75 3.53 4.73 3.78 3.54 3.05 2.57 2.76 3.80 2.00 3.00 4.00 5.00 6.00 Lapse Rate㸦%㸧
a
b
Fig. 1 aThe CIRC lapse value. b The CIRC lapse rates. Data source China Insurance Regulatory Commission
Commission (CIRC). In 2013, the total lapse rate was 3.8%, and the total lapse value reached a record high amount of RMB 190.66 billion, which is RMB 70.85 billion more than the amount in 2012 (RMB 119.81 billion), based on the report of the CIRC.1
The high lapse rate might heavily influence the profitability and liquidity of life insurers through acquisition costs, adverse selection, and cash surrender values (Kiesenbauer 2012). It also has a negative effect on the insurer’s reputation and financial soundness, both of which result in further policyholders lapsing, as well as doing harm to insurers’ efforts to expand new business (Eling and Kochanski2013). Further, high lapse rates might hurt the confidence of customers of the whole life insurance industry and create a potential systemic risk if a series of policyholders’ lapse behaviour leads to widespread panic among consumers similar to a bank run. Therefore, this study not only investigates the determinants of lapse, which most of the existing studies put emphasis on, but also tests the impact of lapse rates on the insurers’ financial strength and new business development.
In this study, we first revisit the traditional emergency fund hypothesis, which states that the embedded cash value in the life policies could be considered as the emergency fund (Linton1932; Outreville1990; Kuo et al. 2003; Kim2005a). We identify migration as an important factor which significantly affects lapse behaviour. This factor has never been studied in the literature, probably due to the lack of data and its insignificant effects on lapse in the developed markets. Large-scale population mobility is a major aspect of China’s industrialisation and urbanisation, which has significantly affected people’s behaviour as well as financial decisions. There were about 236 million migrants in 2012, and there are comparable numbers for other years during the past decade, suggesting one-sixth of the Chinese population was migrating.2
In addition, the significant mobility of migrants in China due to the quick changes in regional economic developments has led to a marked revolution in demand for life insurance in China. In the late 1990s, eastern and southern coastline cities witnessed a huge immigrant increase from the inland provinces, mainly rural residents, who sought work in labour-intensive industries. Over time, the increasing living costs, labour costs and environmental burdens forced coastline cities to shift from labour-intensive industries to capital- and especially knowledge-intensive industries. This significantly reduced the blue-collar job opportunities. As a result,
1 The CIRC lapse rate is calculated as the lapse amount in the current period divided by the sum of the
total accumulated reserves at the beginning of the period and net premiums written in the current period. The literature normally defines lapse rate as the number of policies surrendered divided by the number of total policies (Jiang,2010). Because policy number is not available in China, we define lapse rate as the lapse amount in the current period divided by the premiums written at the beginning of the year. In this case, the lapse rate will increase to an astonishing number, 10.1% in 2014. Please note that our analysis in this study is at the individual firm-province level. While we can obtain the data on the premiums written at the individual province level, we cannot find data of accumulated reserves at the individual firm-province level to apply to the definition of the lapse rate.
2
labour-intensive industries largely moved back to cities in inland provinces.3This led to significant inverse blue-collar labour migration in the last ten years in China. While migrants move to new developing regions which provide new job opportunities, their financial conditions change due to the payment from new jobs and the local living costs. This leads to a change in the demand for financial security and appropriate life insurance coverage. Consequently, the Chinese setting also provides an ideal motivation to analyse how changes in the demand for insurance affect migrants’ life insurance lapse behaviour. We posit that migrants are more financially vulnerable than natives, especially when they are unemployed, which forces them to lapse life insurance as the last resort. In other words, the migrants are more likely to surrender their life insurance to meet emergency financial needs. Thus, we expect to find stronger evidence to support the traditional emergency fund hypothesis on lapse for migrants.
Second, we hypothesise that lapse behaviour is similar to a bank run in that people are easily affected by negative news of insurers when making surrender decisions. More specifically, we test the reputation hypothesis by studying how negative news about insurers affects insurers’ regional lapse rates. Negative news is represented by administrative penalties issued by the regional CIRC branches on market misconduct of insurers, e.g. improper marketing and underwriting practices. Our proxy of administrative penalties is on the market conduct of insurer branches, which is under the supervision of corresponding CIRC regional branches. The penalties are not related to the insurers’ financial strength and lapse behaviour. Therefore, we are able to partially address the endogenous concern of insurers’ financial strength and their lapse rates.
Third, we examine the interest rate hypothesis (IRH), which emphasises the arbitrage by policyholders when the market rates are high (Schott1971; Pesando
1974; Kuo et al. 2003). We extend the literature by investigating the impact of interest rate on three life insurance products, investment-, protection-, and health-type products, in China.4 We argue that the investment-type products are more closely associated with the asset market, and the changes in interest rates directly affect investment returns. Unlike protection-type and health-type products, substi-tutes can be found easily for investment-type products. Thus, high interest rates are more likely to increase the lapse rates of investment-type products than protection-type and health-protection-type products.
Fourth, we test the impact of the lapse risk on insurers’ financial soundness and whether the new business expansion is negatively affected concurrently. High lapse
3 A typical example is Foxconn which employs more than 200,000 blue-collar labourers in China.
Foxconn moved its factory from Shenzhen, one of the most developed metropolises adjacent to Hong Kong, to Zhengzhou, an inland city of Henan province, which is famous for its emigrants to coastal provinces.
4 In China, the most authoritative data source is the annual report of the China Insurance Regulatory
Commission. According to the reports submitted to the regulator by life insurers, we classify the total of five categories of life insurance products into three types: protection, investment and health. The protection type is traditional life insurance, which includes whole life insurance and term life insurance. The investment type includes participating insurance, universal insurance and unit-linked insurance. The health type includes health insurance. Health insurance typically covers illness as well as death.
rates might negatively affect an insurer’s reputation and harm new business, although we find no empirical results in literature (Eling and Kochanski2013).
By way of preview, we find several important results. First, our empirical investigation supports the emergency fund hypothesis with a focus on the effect of labour migration on life insurance lapse rates. We show that the positive relationship between unemployment rates and lapse rates is mostly due to the decisions of migrants in China. Second, our results are consistent with the reputation hypothesis, which holds that reputation is negatively associated with life insurance lapse. Insurers receiving administrative penalties suffer a higher lapse rate than those without unfavourable news. Third, we also find evidence consistent with the IRH, which states that the lapse rate is positively associated with the increase in the interest rate. Further, by decomposing life insurance products into three types, we find that interest rates are positively (negatively) associated with lapse rates of investment-type (protection-type) products. Lapse rates of health products are not affected by interest rates. Fourth, our empirical results show that high lapse rates can weaken financial soundness and impede insurers’ efforts in underwriting new policies.
Our study contributes to the literature in the following important ways. First, to our knowledge, this is the first empirical lapse study of the Chinese life insurance market. By focusing on migrants in the largest emerging market, we provide additional empirical evidence supporting the emergency fund hypothesis. Migration is not unique to China, and our findings on migration associated with urbanisation can therefore be applied to other developing markets.
Second, our firm-province level data enables us to clearly test the reputation hypothesis. As reputational risk is difficult to quantify given the limited data available in the insurance market, the relationship between reputation and lapse is still untested in the literature. Our evidence on the reputation hypothesis also prompts insurers to enhance internal control on underwriting practices so as to improve customer service, which also contributes to the build-up of social confidence in the insurance industry.
Third, most of the literature about lapse rates uses aggregate macroeconomic data of the whole country. In China there are regional differences between the provinces regarding socio-economic and demographic factors, allowing for meaningful cross-sectional analysis within a single economy. We adopt insurance firm-province level data to study the impact of regional macroeconomic factors and firm characteristics on life insurance lapse rates. We classify life insurance products into three types, i.e., protection type, health type, and investment type, and study which types of insurance products are subject to greater lapse risk. We find evidence consistent with the IRH, mostly associated with the investment-type and protection-type products, albeit with opposite directions. Our research thus partially reconciles the conflicting findings in the previous research on the IRH.5This is the first attempt in the lapse literature, as far as we know.
5
Dar and Dodds (1989) find that there is a negative relationship between interest rates and lapses. By contrast, Kuo et al. (2003) give the opposite result that interest rates positively affect lapse rates. The inconsistent results might be due to the unique data samples studied, the specific time periods covered and the different methods used.
Fourth, we are the first to provide empirical evidence with regard to the impact of the lapse on insurers’ financial soundness as well as their efforts in underwriting new policies. High lapse rate can lead to the weakened financial strength of insurers and ultimately potential insolvency in extreme cases. Insolvency of insurers might lead to panic among policyholders and can be a source of systemic risk if not under proper control. Our empirical evidence suggests that a proper understanding of lapse dynamics is particularly important as lapse has a significant influence on insurance managers, customers and regulators (Kiesenbauer 2012). Thus, our study has important policy implications for the life insurance industry in China as well as for the global financial market.
The remainder of the paper is structured as follows: in the following section, we summarise the literature on life insurance lapse rates. In the third section we discuss changes in China’s demographic and macroeconomic factors as well as insurer characteristics that have an impact on customers’ lapse decisions, and develop the hypotheses accordingly. We also explore the consequence of high lapse rates. The fourth section describes the data, control variables and methodology employed. The fifth section provides the results and discussion, and the final section concludes.
Literature review
The Life Insurance and Market Research Association (LIMRA) (1977) gives the earliest definition of lapse, explaining that it occurs when valid life insurance policies are forced to be terminated or have lost their cash surrender value. Outreville (1990) and the Society of Actuaries use almost the exact same definition as LIMRA (1977), which states that a policy lapses if its premium is not paid by the end of a specified time. The subtle differences between lapse and surrender were not noticed before Kuo et al. (2003). They argue that policyholders could actively terminate their policies through surrendering or letting their policies lapse by not paying premiums. Early surrender usually involves cash value payment, but policy lapse might not. The reason is that when policyholders stop paying premiums some policies deduct the premiums still due from cash surrender values so that policies are kept valid until cash values are used up. Both lapse behaviour and surrender activity can erode the customer base, leading to an increase in fixed administrative costs per policy and requiring substantial marketing expenditure to offset these (Russell et al. 2013). Most empirical research broadens the meaning of lapse to account for surrender as well as lapse, referring to a lapse in the insurance contract as termination before maturity (Kuo et al.2003; Kiesenbauer 2012; Renshaw and Haberman1986; Eling and Kiesenbauer2014). The term surrender usually appears in empirical papers studying the secondary market and in theoretical modelling studies (Kim2005a,b; Consiglio and De Giovanni2010; Gatzert2009; Tsai et al.
2002; and Hilpert et al. 2014). When using Chinese data, we cannot distinguish lapse behaviour from surrender activity. Thus, following Jiang (2010) we make no distinction between them in this paper.
The extant literature analyses the economic importance of life insurance lapse from three perspectives. First, high lapse might do harm to an insurer’s profitability,
and surrender cash flows might influence insurer financial stability, which is a potential factor leading to systemic risk (Kuo et al.2003; Kiesenbauer2012; Eling and Kiesenbauer2014). Second, lapse rate is a vital parameter to determine the risk-based capital under SolvencyII (Kiesenbauer 2012; Eling and Kiesenbauer 2014; Eling and Kochanski2013). Third, for potential customers, lapse rate is an indicator of an insurer’s product and service quality (Kiesenbauer2012).
Two main hypotheses about lapse behaviour have received the most attention in existing literature. The first is the emergency fund hypothesis (EFH), which states that policyholders consider life policies as emergency funds when faced with financial hardship. In addition, because of shortage of money to pay for the contracts, policyholders have to choose to let their policies lapse (Linton 1932; Outreville1990; Kuo et al. 2003; Kim2005a).
The second is the IRH, which emphasises the arbitrage by policyholders when the market rates are high (Schott1971; Pesando1974; Kuo et al.2003). In order to take advantage of returns offered by alternative investment vehicles, policyholders may be willing to surrender their policies.
Alternatively, the policy replacement hypothesis (PRH) is derived from Outreville’s replacement theory, which argues that consumers make decisions about lapse with the intent of replacing the original policy with one that is more suitable for them (Outreville1990; Russell et al.2013). Outreville (1990) uses the ratio of new ordinary life insurance business to in-force ordinary life insurance as a proxy for replacements, and finds a positive and significant relationship directly related to early lapse. Russell et al. (2013) employ this measure of replacement activity as well. However, unlike the two classic hypotheses above, the PRH referred to the specific situation of the policyholder that can be better tested by confidential individual data, e.g. Fier and Liebenberg (2013). Also, on the basis of a crowding-out effect between policy loans and policy surrenders, Jiang (2010) posits a crowding-out hypothesis that stresses competition for the withdrawal of cash values. In this paper, macroeconomic data allow us to test IRH and EFH. We also try to use the connotation of the PRH to analyse the motivation of customers’ lapse behaviour.
When it comes to empirical methodologies of life insurance lapse study, generally they can be classified into two main categories based on the explanatory variables. The quantity of studies in the first set of the literature is limited by data that is not always publicly available. These papers apply microdata on individuals or households to study policyholder features and product characteristics (Renshaw and Haberman1986; Kagraoka 2005; Cerchiara et al. 2008; Milhaud et al.2011; Fier and Liebenberg 2013). The second set of research uses macrodata focusing on environmental variables which include macroeconomic and company factors. Initially, Dar and Dodds (1989) and Kuo et al. (2003) only consider variables on interest rate and unemployment. Outreville (1990) adds other variables, such as real transitory income, price of insurance and the inflation rate. This work is extended by Kim (2005a,b) and Cox and Lin (2006), which add policy age, GDP growth, and surrender charges as additional variables. Kiesenbauer (2012) applies more macroeconomic factors (i.e., buyer confidence) and company-specific explanatory variables (i.e., company age, distribution channels, legal form, firm size and
participation rate).6Our study generally follows Kiesenbauer (2012) and analyses life insurance lapses by using unique Chinese data to study a few untouched determinants, notably labour migration and insurers’ reputation.
Hypotheses
Determinants of lapse
Labour migration
According to the data by National Bureau of Statistics of China, there were approximately 236 million migrants moving into urban cities from rural areas in 2012, and the number is expected to continue growing during the next decade due to the new urbanisation occurring in China. It is the largest migration of its kind in human history. To the best of our knowledge, there is no study that has explicitly investigated the relationship between the lapses of life insurance and labour migration. In this paper, we postulate that the prevalence of labour migration has a significant positive impact on lapses of life insurance because of relatively reduced financial resources of migrants compared with local urban residents and the changing needs for life insurance. This leads to our first hypothesis:
Hypothesis 1 Labour migration is positively associated with life insurance lapse. As elsewhere in the world, the remarkable Chinese labour migration centres on mobility, which means some people move to urban areas from rural areas and some people move out of urban areas to rural areas or to another city. It is from this perspective that we analyse the relationship between migrant population and life insurance lapse.
First, the kind of people who move to urban areas are more easily faced with economic hardships. According to the 2013 Report on China’s Migrant Population Development, labour migration is experiencing a generational shift, which means that the young generation makes up a dominant proportion of the migration population. Compared with the aged population, the young are more keen to live in the major cities for the sake of better education and health care opportunities.7 However, high housing prices and living costs cause the immigrants to live a stressful life. The situation is expected to be exacerbated when the unemployment rate in an area increases due to reduced job opportunities and increased competition from local residents and other immigrants. Thus, we postulate that:
H1a Increase in the unemployment rate leads to a higher lapse rate when the migrant ratio is relatively high.
More specifically, the interaction of unemployment and migration is expected to have a positive relationship with life insurance lapse. According to the emergency
6
Please see Eling and Kochanski (2013) for an excellent review on lapse in life insurance.
7 A nationwide survey conducted by the National Population and Family Planning Commission revealed
that in 2012 the average age of the migrant population was 28, and more than 70% expressed hopes of settling in big cities.
fund hypothesis (EFH), policyholders consider their life insurance policies as a source of emergency funding and they surrender their policies in times of need. When unemployment increases, the migrant population is more motivated to surrender their insurance contracts. Furthermore, new migrants are moving away from their native homes, making it harder to access emergency funding resources since they are now living independently away from relatives and close friends. Zhang and Song (2003) suggest that labour migration reduces the traditional mutual dependency of family members. When migrants are faced with economic distress, they are less likely to receive enough financial support in contrast to the local residents who find it easier to get support from relatives and friends.
Second, people who tend to move out of their original urban areas might let their life policies lapse and search for an alternative contract to better fit their new jobs in new cities. A significant number of Chinese workers have moved with their firms from relatively developed regions to less developed provinces in the past decade. This is due to the adjustment in industrial structure that causes companies to relocate to areas with lower labour and environmental costs. The Chinese government has encouraged the upgrade of the industrial structure in developed regions, and many labour-intensive enterprises have moved from the southern or eastern coastal cities to the middle or western inland areas of China.
Further, ‘‘return to the roots’’ is the tradition in Chinese culture. In contrast to people moving to urban areas, the aged and less educated migrant population is the main group of people who move out of developed cities. Because of the differences between regions and a wide gap between urban and rural regions, the aged population usually finds that they cannot completely become acclimatised to the city they work in. For this reason, many of the aged are inclined to go back to their rural home towns. Overall, the change in working and living conditions might change the life insurance demands of migrants and the lapse behaviour.
Reputation
The flagging reputation due to market misconduct and deteriorating financial strength are two main causes of insurers’ lapse. While it is not that straightforward for consumers to compare the financial strength of insurers, consumers might rely more on insurers’ reputation to make a purchase or lapse decision in China.
Eccles et al. (2007) hold the view that firms with a better reputation among stakeholders tend to be more valuable. This conjecture is supported by Wang and Smith (2008), stating that on average a firm with a better reputation increases market value by about USD 1.3 billion. Reputation is an intangible asset of a firm, particularly for insurers. Companies with a relatively good reputation might sell products and services to customers at a higher price and/or at a lower cost than their competitors. In contrast, loss of reputation may result in financial distress through weakened investor confidence and loss of customers. A survey conducted by Swiss Re (2011) states that 26% of people who originally want to purchase insurance, decide not to because of the poor reputation of the company. It follows that insurer reputation is one of the main factors that is used by customers to assess the product and service quality of life insurance companies. Thus, we postulate that
Hypothesis 2 Reputation is negatively associated with life insurance lapse. More specifically, we apply a negative news item to proxy for firm reputation and predict that it is positively related to the lapse rate. The negative news item has two advantages when used to proxy for firm reputation. First, negative news items are consistent with the core concept of reputational risk to insurers. Following the Federal Reserve System’s regulatory definition, a negative news item can damage corporate reputation. ‘‘Reputational risk is the potential that negative publicity regarding an institution’s business practice, whether true or not, will cause a decline in the customer base, costly litigation or revenue reduction.’’ Second, customers usually lack professional knowledge when purchasing insurance coverage or deciding to surrender the life insurance contract.8They are more likely to depend on the impression from news items reported through print, electronic and social media outlets. Once a negative news item about an insurer appears on social media or television, it will rapidly circulate to existing or potential customers. Thus, we use a negative news item as a proxy variable for insurer reputation and conjecture that it has a significant positive impact on life insurance lapse decisions.
Interest rate on products
The IRH has been investigated frequently, albeit with inconsistent results in the literature. The inconsistent results might be due to the unique data samples studied, the specific time periods covered, and the different methods used. According to Kuo et al. (2003), policyholders lapse their policies in order to take advantage of higher interest rates and/or lower premiums when market interest rates rise. Therefore, we include interest rates proxied by the one-year bank deposits rate set by the central bank to test the IRH, and expect a positive sign for this coefficient.
We extend the literature of IRH in this study by studying the impact of interest rates on three different life insurance products, namely protection-, investment-, and health-type products. In other words, we test what kind of products are more sensitive to the changes in interest rates and are subject to a higher lapse rate.
There are various asset items in households’ portfolios, and life insurance can be a substitute for financial assets such as equities and other lower risk assets (Fortune
1973). With the development of investment-type products, life insurance was considered mainly as an investment channel in China in the past rather than a protective mechanism.4The changes in interest rates directly affect the investment return on the assets, and encourage customers to reallocate personal portfolios when the alternative asset market is more profitable. On the basis of this point, we conjecture that interest rates will increase their effect on lapses through investment-type products, and postulate that:
Hypothesis 3 Interest rates positively affect the lapse of investment-type life insurance products.
8 If customers purchase multiple policies from different insurers, the reputation matters a lot. They
usually depend on the reputation to decide when they want to surrender one or a few but not all the policies, and which insurer’s policies they want to surrender and which ones they want to keep.
For the protection- and health-type products, their essence is to resist the potential risk, which makes life insurance stand out from other financial assets. The potential interruptions of protection against risks are expected to deter the consumers’ lapse behaviour. In other words, if the policyholders are rational enough, they should hesitate to make a lapse decision on protection- and health-type products for only marginal returns due to the increase in interest rates. Thus, we do not expect a significant impact of interest rates on lapses of protection- and health-type products.
Impacts of lapse
Risks arising from lapse are of great importance to the insurer’s financial soundness. Unexpected high lapse directly challenges an insurer’s asset and liability management. A massive lapse event could even endanger the current liquidity of the insurer (Kiesenbauer 2012). Furthermore, high lapse rates could further negatively affect the insurer’s cash flow through impeding the effort of writing new business due to the firm’s worsening financial conditions (Eling and Kochanski
2013). Potential customers might avoid purchasing new insurance policies from insurers with high lapse rates and flagging financial conditions. Deteriorating financial soundness of insurers can have a negative impact on the financial system and could lead to systemic risk similar to a bank run. We use insurer leverage to proxy for financial strength and postulate that:
Hypothesis 4 The high lapse rate will threaten the insurer’s financial soundness. H4a The lapse rate positively affects the leverage of the insurer.
H4b The lapse rate negatively affects the insurer’s conducting of new business.
Data and methodology
Data and variables
Lapse rate
Lapse rate as a direct indicator of the lapse experience of insurers has been used in previous studies. According to the Life Insurance Marketing and Research Association (LIMRA 1977, p. 11), the life insurance industry calculates the annualised lapse rate as follows:
Annualised Policy Lapse Rate
¼ 100 Number of policies lapsed during the year Number of policies exposed to lapse during the year However, in the academic literature, researchers have done some revisions to the above formula. For example, Jiang (2010) uses the ratio of the number of
surrendered policies to the mean number of policies in force, and Eling and Kiesenbauer (2014) consider lapse rates in terms of the number of contracts and regular premiums. In the Chinese life insurance market, the number of policies lapsed is not publicly available. We define the lapse rate as the ratio of total amount of lapse to the total premiums written.
Macroeconomic explanatory variables
Unemployment rate Unemployment is used in nearly all lapse research because it is the most typical indicator when testing the emergency fund hypothesis. Previous literature provides inconsistent evidence. While Dar and Dodds (1989), Outreville (1990) and Kim (2005a) find unemployment is positively related to the lapse rate, other studies find no evidence supporting the emergency fund hypothesis, i.e. Kuo et al. (2003) We follow most studies and apply the yearly unemployment rate (Outreville1990; Kuo et al.2003; Kim2005a; Kiesenbauer2012) in our study. In accordance with the emergency fund hypothesis, we predict that the unemployment rate will be positively related to lapse.
GDP growth Kim (2005a,b), Cox and Lin (2006), and Kiesenbauer (2012) argue that GDP growth reflects not only the overall development of the economy but also the disposable income of people. Households that experienced income growth are expected to be less likely to lapse a policy contract. We thus expect a negative sign for the coefficient for the GDP growth.
Stocks index growth Even though Dar and Dodds (1989) explicitly differentiate between internal and external rates of return in the context of the IRH, they only consider risk-free alternative assets. However, as a risky alternative to life insurance products, stock investment provides another attractive financial asset for household portfolios. Thus, stock performance might provide a starting point for explaining the lapse behaviour of policyholders, especially in the case of investment-type insurance, such as unit-linked products (Kiesenbauer 2012). Furthermore, the stock market has received the most public attention in the past decade in China. Hence, it constitutes easily accessible information for customers. The annual stock market price index is used in our analysis, and a positive relationship is expected if stocks and investment-type life insurance products are substitutes. However, an increase in the stock market index might suggest an increase in household wealth and stimulate consumers’ purchasing and holding of life insurance products. Thus, we do not have predictions on the signs of the stock market index growth.
Education The education level is correlated with human capital investment and earning ability over the long term. It is also associated with wealth, financial vulnerability, and risk attitude. Fier and Liebenberg (2013) find that the level of the household’s education affects lapse positively using micro survey data, although no
explanation for the result is given. Surprisingly, we find no literature applying education as an explanatory variable in lapse studies at the macro level.9In this paper, we study the impact of education on lapse from the macro perspective. Education is defined as the proportion of the regional population with a college degree or above. We do not have a prior on the result.
Social security expenditures and Herfindahl–Hirschman Index (HHI) The emer-gency fund hypothesis conjectures that when faced with economic hardship, policyholders choose to lapse their contracts for the cash surrender value. Social security expenditures can be considered a source of household assets and disposable income (Bernheim 1991; Browne and Kim 1993), which could help to alleviate consumers’ personal financial distress. However, there is no empirical research studying how social security expenditures affect life insurance lapse rates. In this paper, we measure social security expenditures as social security spending per capita. Finally, like Outreville (1996) we calculate the Herfindahl–Hirschman Index (HHI) of the premiums written for each province each year to capture the regional market competition structure.10We expect market competition to affect customers’ lapse decisions when they plan to replace original policies with other ones.
Firm-specific explanatory variables
Following the literature (e.g. Kiesenbauer2012), the first firm characteristic variable which we control for is firm size. Large firms are expected to have more financial resources to honour their promises; thus, firm size should have a negative impact on life insurance lapse. Firm size is proxied by the natural logarithm of total assets.
The second control variable is leverage. An insurer’s financial strength is expected to negatively affect a consumer’s lapse decision. We use leverage, which is the ratio of liability to assets, to reflect the insurer’s financial condition and predict that it is positively related to the lapse rate.
The third control variable is business age, following Kiesenbauer (2012). A firm that has been in the market for a long time has greater visibility—implying better reputation. Thus a negative relationship between the business age and lapse rate is expected.
Next we control for business concentration. With high business concentration, insurers are more likely focus on a few (a limited number of) product lines and are more tempted to provide attractive policies (i.e. better rates or services) for specific groups of customers. For this reason, the business concentration variable is hypothesised to be negatively related to insurance lapses.
The fifth variable we include is a dummy variable to capture firm ownership structure. We categorise insurance companies operating in China into two broad types based on their ownership structure: insurers with and without certain degrees
9
A positive relation between education and demand for life insurance is found in the literature (Truett and Truett,1990; Browne and Kim,1993; Hwang and Gao,2003; and Li et al.,2007).
10Outreville (1996) proxies for market structure by introducing two dummy variables indicating whether
of foreign shares, following the categorisation of the CIRC. Among 62 companies with data available in our sample, 37 are domestic companies without foreign ownership, and 25 firms have varying degrees of foreign insurers’ ownership. The dummy variable is equal to one if the firm has shares owned by foreign insurers, and zero otherwise. We postulate that foreign ownership structure is negatively related to life insurance lapse rates.
As an explanatory variable, product type is not rare in the existing literature on lapse. However, most studies usually conduct empirical investigation targeting one specific product type, e.g., Dar and Dodds (1989) and Kuo et al. (2003), or using individual data, e.g., Renshaw and Haberman (1986), Cerchiara et al. (2008), Milhaud et al. (2011) and Eling and Kiesenbauer (2014). Kiesenbauer (2012) suggests that life insurers treat lapse data as highly confidential. The data of total lapse amount for each company per year in China is available but not by product category. Thus, we study the impact of life insurance product lines by including the proportion of investment-and health-type premiums written in total premiums written, with the proportion of protection-type premiums omitted to avoid collinearity. We predict that investment-type products are more likely to be lapsed than health- and protection-investment-type products.
Data source
We manually collected the data from the CIRC annual reports (Yearbook of China Insurance) for the period 2005–2013. Data for negative news items, migrant population, unemployment, social security expenditures, economic development and education were collected from several sources. Insurers’ reputation proxied by negative news items was obtained from the official CIRC website.11The migrant population ratio was obtained from the China Statistical Yearbook, and the unemployment ratio from the CEInet Statistics Database. Social security expenditure was obtained from National Bureau of Statistics of China. Finally, data for education, stock index growth, GDP growth and interest rates were obtained from the CSMAR (China Stock Market & Accounting Research) and the Wind databases for various years.
In the final sample, at the individual firm-province level, we have 62 firms and 4292 observations after eliminating observations with missing information for the regression analysis. The sample firms represent 85% of total life insurance industry premiums written in 2010 and represent comparable percentages for other years.
Methodology
In this study, we take regional variability into consideration in a country for the first time in lapse research literature. Following Kiesenbauer (2012), we apply a fixed firm effect panel model to investigate the determinants of lapse rates.12The specific model is as follows,
11Please seehttp://www.circ.gov.cn/web/site0/for more details. 12
Ordinary least-square (OLS) model only treats each observation equally, but does not take individual and time effects into account. Kiesenbauer (2012) thus applies the fixed firm effects model based on the data design. By conducting Hausman specification tests, we find that the fixed firm effects (FE) model is preferred to the random firm effects (RE) model.
LRi;t¼ aiþ bk Xk;i;tþ li;t
where LR is short for lapse rate, i indicates the respective province-level firm in China (i = 1, 2,…,1922), t denotes the year (t = 2005, 2006,…,2013), k signifies different explanatory variables and aiis constant.13
First, we use non-local resident ratio as the proxy for migration. According to the household registration system of China, the threshold to decide whether or not residents are local depends on the resident’s place of registered permanent residence. If the living place is the same as the registered permanent residence, the resident is considered a local. Otherwise they are considered a migrant.
Second, as mentioned above, we apply a negative news item to proxy for firm reputation. To capture the effect of a negative news items, we adopt a dummy variable denoting whether there is negative news related to the insurer during a whole calendar year. More specifically, we adopt the administrative penalties issued by the CIRC provincial branches for two reasons: 1) Every administrative penalty is posted on the website of the CIRC and therefore is publicly available. Compared with alternative sources, messages from an authoritative source tend to have a greater impact on customers. 2) The CIRC has 36 provincial level branches. The administrative penalty announced by these provincial branches allows us to merge these data with the regional macroeconomic data to exploit the regional differences in China. Further, the administrative penalties are mainly about market misconduct of insurer branches which is not related to financial conditions of insurers. This allows us to disentangle the effect of market news on customers’ lapse decisions from firm financial strength. The dummy variable of negative news is equal to one if the insurer received any administrative penalty during the calendar year, and zero otherwise.
To disentangle the impact of interest rate on different products, we multiply the market interest rate with three product-type ratios. Comparing coefficients of three interaction terms allows us to test Hypothesis 3.
We use the firm-level data to test H4a and H4b, because the financial variables are available at firm level instead of firm-province level. We consider the leverage as the indicator of an insurer’s financial strength and apply the new business growth rate to proxy for the impact of lapse on a firm’s ability to write new business. It is defined as the annual growth rate of premiums written in new business.
Because of the skewness of some explanatory variables (e.g. social security expenditures), following Li et al. (2007) we use a logarithmic transformation.
13Kiesenbauer (2012) gives two reasons supporting the non-consideration of time effects. First, using
time effects would partially offset the impact of economic explanatory variables. Second, when the time effects are included, the impact of some explanatory variables cannot be estimated due to multicollinearity. Nevertheless, we conduct robustness tests by adding year dummies. Our main results remain quantitatively similar, except that results with regard to interest rates hypothesis are distorted to some extent as expected. Results are available from the authors on request.
Variables that are not expressed as logarithms are already expressed in percentage terms or dummy variables (e.g. interest rate, the reputation dummy). We omit the ratio of protection-type products in our regression to avoid perfect multicollinearity. Finally, we conduct several robustness tests, including replacing the non-local resident ratio with urbanisation ratio and using one-year lagged firm reputation proxy. We acknowledge the fact that it might take time for policyholders to react to the negative news. We also replace the new business growth rate with the absolute level of new business premiums written and the ratio of new business to total premiums written.
Results
Descriptive statistics
Table1 displays the descriptive statistics for the dependent and explanatory variables. The mean, median and related statistics are shown at the firm-province level from 2005 to 2013. The table indicates that lapse rates vary significantly among firms and provincial branches with median and mean at 7.1 and 10.5%, respectively. The lowest lapse rate is zero while the highest lapse rate reaches 40.2%. There is also significant variance in socio-economic and demographic variables. The mean of the unemployment rate and migrant population (non-local residents) ratio is 3.5 and 19.5%, respectively. The average proportion of the population with a college degree or above is 11%. The government expenditures on social welfare have increased over time, and the mean of social security expenditures is RMB 689.5 (value before taking logarithm) per capita. The interest rate is relatively stable over our sample period with a mean of 2.9%. The stock performance fluctuates drastically during the sample years and the stock index growth ranges from a peak of 130.4–65.4%. There are also big differences among regional insurance market competition structures as the HHI ranges from 0.12 to 0.57.
The results in Table1contain the means for firm-specific explanatory variables as well. Approximately 22.2% of the sample consists of insurers with foreign-ownership structure. The investment-type products enjoy about 74.8% of the life insurance market, followed by protection-type products (12.4%) and health-type products (10.8%).
Regression results
Regression results to test the Hypotheses 1–3 are presented in Table2. Columns 1–4 report the results applying the firm-fixed-effect estimation method. Hypothesis 1, which considers migrants, and Hypothesis 2, which considers firm reputation, are generally borne out in the results in Column 1. Columns 2–4 contain the results in
Table 1 Descriptive statistics
Summary statistics for the dependent variables Mean Median Std Min Max
Lapse rate 10.518 7.130 10.840 0.000 40.190
Macroeconomic variable statistics
Non-local resident rate (%) 19.530 16.440 11.876 3.590 39.890
Negative news (dummy) 0.191 0.000 0.393 0.000 1.000
Unemployment rate (%) 3.480 3.540 0.569 2.430 4.350
Education 0.110 0.090 0.058 0.030 0.230
Interest rate (%) 2.938 3.000 0.538 2.250 3.940
Stock index growth (%) 12.575 - 6.750 53.915 –65.390 130.430
GDP growth (%) 13.938 13.070 5.646 5.830 25.530
Social security expenditure 6.453 6.555 0.757 4.286 7.504
Herfindahl–Hirschman Index (region) 0.228 0.210 0.093 0.120 0.570
Firm-characteristic variable statistics
Insurer size 10.374 10.178 1.861 5.765 12.865
Insurer leverage 0.876 0.918 0.127 0.096 0.970
Insurer business age 7.155 7.000 3.526 0.000 14.000
Business concentration 0.696 0.706 0.195 0.362 0.975
Insurer with foreign ownership (dummy) 0.222 0.000 0.416 0.000 1.000
Protection-type ratio 0.124 0.053 0.151 0.003 0.533
Health-type ratio 0.108 0.054 0.139 0.003 0.546
Investment-type ratio 0.748 0.852 0.278 0.000 0.989
This table reports summary statistics for the years 2005–2013. The firm-province level data contains 4292 observations. Negative news is a dummy variable equal to one if the firm receives administrative penalties from CIRC, and zero otherwise. Non-local resident rate is the ratio of the total number of non-local residents to the total number of residents in the same year. Unemployment rate is the ratio of yearly unemployed population to the regional total population. Education is the proportion of regional popu-lation with a college degree or above. Interest rate is the 1–year bank deposit interest rate. Stock index growth is the annual growth rate of the Shanghai stock exchange composite index measured in closing prices. GDP growth is the regional annual growth rate of gross domestic product (GDP). Social security expenditure is the logarithm of the regional amount of expenditure for social security and employment (in RMB) per capita. Herfindahl–Hirschman Index (region) is Herfindahl indexes of premiums written by province. Insurer size is the logarithm of total assets. Insurer leverage is the ratio of liability to assets. Insurer business age is the total years in business. Business concentration is Herfindahl indexes of premiums written by product categories. Insurer with foreign ownership dummy is a dummy variable equal to one if the firm has shares owned by foreign insurers, and zero otherwise. Protection-type ratio is the proportion of premiums written for protection. Health-type ratio is the proportion of premiums written for health type. Investment-type ratio is the proportion of premiums written for investment products. All continuous variables are winsorized at the 5 and 95 percentile to remove the excess effects of outliers
Table 2 Determinants of life insurance lapse
Variables Column 1 Column 2 Column 3 Column 4
Hypothesis 1
Non-local resident rate 0.057** - 0.140 0.055** - 0.134
(0.024) (0.101) (0.024) (0.099)
Unemployment rate 0.850*** - 0.634 0.813*** - 0.614
(0.301) (0.840) (0.301) (0.826)
Hypothesis 1a
Interaction (non-local and unemployment) 0.059* 0.057* (0.030) (0.030) Hypothesis 2 Negative news 1.346*** 1.304*** 1.312*** 1.272*** (0.461) (0.457) (0.460) (0.457) Hypothesis 3 Interest rate 1.524** 1.578** (0.668) (0.670)
Interaction (interest and investment type) 2.482*** 2.528***
(0.817) (0.817)
Interaction (interest and protection type) - 3.259* - 3.226*
(1.714) (1.698)
Interaction (interest and health type) - 3.008 - 2.882
(2.626) (2.624)
Investment-type ratio 6.822*** 6.872*** - 6.554 - 6.576
(2.157) (2.141) (4.828) (4.755)
Health-type ratio - 0.179 0.163 0.608 0.646
(4.607) (4.597) (11.235) (11.259)
Macroeconomic explanatory variable
GDP growth - 0.259*** - 0.264*** - 0.252*** - 0.256***
(0.045) (0.045) (0.047) (0.046)
Social security expenditure - 2.001*** - 1.938*** - 1.919*** - 1.859**
(0.721) (0.725) (0.695) (0.700)
Stock index growth 0.014*** 0.014*** 0.014*** 0.014***
(0.005) (0.005) (0.005) (0.005)
Education 9.159 8.025 8.464 7.375
(7.290) (7.363) (6.996) (7.080)
Herfindahl–Hirschman Index (region) – 20.679*** – 21.414*** – 20.858*** – 21.565***
(3.313) (3.408) (3.230) (3.319)
Firm-characteristic explanatory variable
Insurer size 3.020*** 3.028*** 3.028*** 3.035***
(0.879) (0.880) (0.880) (0.881)
Insurer leverage 6.404* 6.505* 6.629* 6.728*
which relevant interactions are added. More specifically, the interaction term of non-local resident rate and unemployment rate added in Column 2 is used to test H1a, and the interactions term of interest rate and product type ratios in Column 3 is used to analyse the sensitivity of lapse rate to interest rates by products. In Column 4, we conduct the regression adding all interactions in the model.
First, the positive and significant coefficient of the migrant population variable in Column 1 indicates that labour migration is positively associated with life insurance lapse. The provinces with a high non-local resident rate are more likely to witness a high lapse rate, ceteris paribus. Due to the lack of access to financial support from relatives and friends when in financial trouble, the migrant population is more likely than the local residents to resort to lapsing their life insurance for emergency funds. Note that the connotation of policy replacement hypothesis provides an alternative explanation. Migrants purchase life insurance in the city in which they live, partially seeking a sense of belonging. When they change the place where they live for alternative job opportunities or when they become homesick, they tend to lapse their policies in the place where they previously lived and search for an alternative one in the new location they migrate to. This also leads to a high lapse rate for migrants. We also find empirical evidence supporting Hypothesis 2. The coefficient of negative news is positive and significant at the 1% level in Column 1, consistent with Hypothesis 2 on the impact of negative news on insurers. For customers, negative news acts as a signal for insurers’ reputation. Consumers do not possess professional knowledge about insurance, and they tend to lapse their policies immediately when they get negative news about their insurers. Regulators and
Table 2 continued
Variables Column 1 Column 2 Column 3 Column 4
Insurer business age - 0.559* - 0.639* - 0.489 - 0.566*
(0.327) (0.336) (0.327) (0.337)
Business concentration – 11.386*** – 11.455*** – 13.522*** – 13.568***
(2.615) (2.574) (2.596) (2.568)
Insurer with foreign ownership - 1.021 - 1.140 - 0.765 - 0.882
(2.323) (2.359) (2.398) (2.433)
Constant - 8.090 - 2.670 3.247 8.537
(8.339) (9.317) (8.613) (9.567)
Number of observations 4292 4292 4292 4292
R-squared 0.123 0.124 0.127 0.128
This table reports results testing the determinants of life insurance lapse in China using aggregate firm-province level data for the sample period 2005–2013. Column 1 reports results using macroeconomic variables and firm-specific variables. Column 2 reports results adding the interaction of non-local resident rate and unemployment rate. Column 3 reports results adding the interactions of interest rate and product-type ratios. Column 4 reports results adding all interactions. We use the firm-fixed-effect method in Columns 1–4. All other variables are defined in Table1. Robust standard errors are reported in paren-theses below each coefficient
insurers should be aware that nowadays information spreads quickly through new media, including social networks, and may contribute to a potential bank-run risk for insurers.
In Column 1, the coefficient of the unemployment ratio is significantly positive, supporting the EFH stating that consumers lapse policies because of shortage of money for emergency situations. The coefficient of interest rate is also positive and significant, consistent with the IRH.
The GDP growth and the social security expenditure are negatively associated with the life insurance lapse rate, which is consistent with the argument that the growth in GDP and increase in social security expenditure leave people with more money in their pockets. This reduces the need to lapse life insurance to survive economic hardships. On the contrary, the coefficient of stock index growth is significantly positive, suggesting that stock is an alternative asset to life insurance investments. From the perspective of asset portfolios, once customers cash out their life insurance to reallocate to equities when the stock market is more profitable, we would witness a positive relationship between stock index growth and lapse rate. The coefficient of education is not significant in Column 1. The coefficient of the HHI is significant and negative in Column 1, suggesting that a higher degree of competition (associated with a low HHI) might provide customers with more attractive contracts with better terms and rates and thus increase the lapse rate.
Column 1 reports the results of firm-specific variables as well. We expect that firm size has a negative impact on the life insurance lapse rate. However, the coefficient of firm size is positively significant. Perhaps this is because large firms have significant disparities in employee abilities and service quality over regions. While smaller insurers normally operate only in large cities, large insurers have subsidiaries around the country, and employee and relationship development are quite divergent.14
The positive coefficient of leverage in Column 1 indicates that a higher leverage leads to increased lapse rates, suggesting insurer financial strength is important to policyholders when they make lapse decisions. The coefficient of the business age is negatively significant, which is consistent with our expectation that insurers build up their reputation over the years.
A higher business concentration might lead insurers to focus on several specific areas. This allows insurers to provide policies catering more specifically to the demand of a small group of target customers if their customers are similar, e.g. clustered in one city rather than in the whole country, which leads to the lower lapse rate. To our surprise, insurers with a foreign ownership structure do not have an advantage over the domestic firms. This might be because the domestic firms have a long history of operating in local markets, and the largest ownerships are usually controlled by the government, which is expected to have more financial resources to honour their promises.
14According to the report of the China Insurance Regulatory Commission, in 2013 the total lapse amount
for the whole industry reached RMB 190 billion. The lapse amount of the four largest insurers publicly traded in A-shares stock market (China Life, Ping An, New China Life, China Pacific Insurance) is RMB 121 billion, accounting for 63.68% of the whole volume. Correspondingly, the premiums written of these four largest insurers are RMB 326.72 billion, 146.09 billion, 103.64 billion, and 95.10 billion, representing 62.52% of the market share.
The results in Column 1 also indicate that compared with insurers writing more protection-type business, insurers writing more investment-type business are subject to a significantly higher lapse rate. Health-type business seems to have a lapse rate comparative to type products. A plausible explanation is that protection-type and health-protection-type insurance products are necessities for policyholders, and it is not easy to substitute these products for other financial products.15 By contrast, customers are more likely to surrender investment-type products if the alternative product substitute in the financial market provides a higher return.
We add the interaction of the unemployment rate and the non-local resident rate in Column 2 to test Hypothesis 1a, which states that an increase in the unemployment rate leads to a higher lapse rate when the migrant ratio is relatively high. The coefficient is significantly positive, as expected. More specifically, a one standard deviation (0.569) increase in the unemployment rate for non-local residents would lead to an estimated 3.4% (0.569 9 0.059) increase in the lapse rate, according to the result in Column 2, Table2. That is about a 32% increase in the lapse rate at the mean (0.034/0.10518).16 Thus, unemployment has a high economically significant impact on the lapse rate for non-local residents, i.e. when the migrant ratio is relatively high, unemployment has a larger impact on the lapse rate. This result suggests that once the migrant population is unemployed, they are more likely to lapse their life insurance policies for the surrender cash value than local residents who lose their jobs.
The insignificant coefficient of the unemployment rate suggests that the evidence consistent with the EFH hypothesis in Column 1 is mainly driven by the migrant population. A plausible explanation is that due to the traditional Chinese Confucian culture it is relatively easy for local residents to get financial support from their local connections for emergency fund purposes. However, this support is unavailable to migrants who can only count on themselves in an emergency when they are far away from their home town.17In Column 2, the other results are consistent with the results of Column 1.
We add the interactions of interest rate and product type ratios in Column 3 to test Hypothesis 3. The results suggest that the effects of interest rate on products are different. The coefficient of the investment-type product interaction term with the 15
Note that the fluctuation of interest rates was not that dramatic during our sample period. Also, our data allow us to decompose total premiums written into three types of products, protection-, investment-, and health-type products. However, we do not have a lapse amount for each category. Thus, our results with regard to the lapse possibility associated with product category should be interpreted with caution.
16
Similarly, the result in Column 4 Table2suggests an increase of 31% of the lapse rate at the mean (0.569 9 0.057/0.10518).
17
The traditional Chinese Confucian culture is associated with a strong obligation for providing mutual support to the members of an extended family (Kotlikoff and Summers,1981). The extended family relationship can be considered an effective instrument to provide emergency funds and is a substitute for life insurance lapse. However, the significant mobility of migrants in China over the past decades widens the interpersonal relationships and reduces the traditional mutual dependency of family members, especially for migrants (Zhang and Song,2003). In contrast, local residents normally maintain strong connections with relatives, friends or colleagues in the same city. Moreover, the financial conditions in migrants’ home towns are normally inferior to those of their current cities of residence. This further limits migrants’ possibilities of getting emergency assistance in the form of financial support from their social connections and increases their incentive to lapse life insurance.
Table 3 Impacts of life insurance lapse
Variables Column 1 Column 2
Panel A: Dependent variable: insurer leverage
LAPSE RATE 0.003**
(0.001)
LAPSE RATE lag one year 0.001
(0.001)
Insurer size 0.122*** 0.076***
(0.023) (0.022)
Insurer with foreign ownership 0.129** 0.098*
(0.056) (0.052)
Insurer business age - 0.009 - 0.003
(0.010) (0.008) Business concentration - 0.129* - 0.119* (0.068) (0.069) Constant - 0.249* 0.172 (0.143) (0.146) Observations 448 416 R-squared 0.402 0.249
Panel B: Dependent variable: new business growth rate
LAPSE RATE - 6.127**
(2.603)
LAPSE RATE lag one year - 4.701*
(2.454)
Insurer size – 197.631*** – 201.870***
(43.199) (44.134)
Insurer leverage – 361.417 – 423.083
(271.000) (259.395)
Insurer with foreign ownership – 131.668 – 102.701
(120.474) (116.080)
Insurer business age 11.718 14.143
(13.760) (13.763) Business concentration 309.710* 309.712* (177.479) (182.697) Constant 2094.228*** 2126.176*** (364.159) (357.501) Observations 407 407 R-squared 0.293 0.285
This table reports results testing the impacts of life insurance lapse in China using aggregate firm-level data for the sample period 2005–2013. This table contains two parts: Panel A and B apply the insurer’s leverage and the insurer’s new business growth rate as the dependent variable, respectively. There are two columns in each panel. Columns 1 and 2 report the results using the lapse rate and the lagged one-period lapse rate as independent variables, respectively. We use the firm-fixed-effect method in all models. Robust standard errors are reported in parentheses below each coefficient
interest rate is positively significant, which shows that investment-type products are more sensitive to the increase in interest rates and subject to a higher lapse rate. In contrast, a significantly negative coefficient is found for the protection-type product interaction term with the interest rate. It indicates that the marginal returns due to the increase of interest rates are not large enough for rational policyholders to make a lapse decision on protection-type products, given the potentially high lapse costs. There are no notable changes in other results in Column 3. Finally, in Column 4, we conduct the regression with all interactions, and the results remain the same.
In Table3, we present results with regard to Hypothesis 4. In Panel A, the coefficient of the lapse rate is positively significant in Column 1 but not significant in Column 2. This suggests that the leverage is affected by the lapse rate of the same year but not by the lapse rate of the previous year and is consistent with H4a. A plausible explanation is that insurers’ leverages are closely monitored by regulators, and they take immediate actions to adjust their leverage back to the firm target. In Panel B, the results show that there is a negative relationship between the new business growth rate and the lapse rate. The new business growth rate is not only affected by the lapse rate of the same year but also by the previous year. It supports the H4b that the high lapse rate gets in the way of insurers conducting new business. Moreover, it impedes the business development over a relatively long time.
In summary, we find results supporting our Hypotheses 1–3 in Table2. Empirical results also support our expectations about the coefficients of most variables in the table. In Table3, Hypothesis 4 also gets support. Now we turn to the results of the robustness tests that are available from the authors on request. We replace the non-local resident rate with urbanisation rate, which could also reflect the labour migration conditions. Considering the fact that policyholders might react to the negative news only after a certain time lapse, we replace the proxy for reputation with a lagged variable. The main results remain unchanged, which lends confidence to the results with regard to our hypotheses. We also replace the new business growth rate with the absolute level of new business premiums written and the ratio of new business to total premiums written. The results remain similar.
Conclusion
This article investigates lapse behaviour in the Chinese life insurance market. The sample period is from 2005 to 2013, a period in which the Chinese insurance industry witnessed a rapid development. The province-firm level data allows us not only to capture the impact of firm characteristics on lapses, but also to test how regional variabilities, especially labour migration, affect lapse decisions within a single economy.
Empirical results indicate that labour migration is positively associated with life insurance lapses, which is consistent with our hypothesis. Because of the features of the migrant population, they are more likely to make lapse decisions to acquire the cash value of their policies or to find alternative policies with a better fit for their new environment, lending support to the emergency fund hypothesis and providing a new perspective for understanding the policy replacement hypothesis.
The empirical results also suggest that the similarity of the lapse rate in China compared to other countries noted in the literature might be masked by two contradictory forces: culture and migration (Kuo et al.2003). While the traditional Chinese Confucian culture of interpersonal relationships has a negative impact on the lapse rate, this effect is significantly weakened by the immigrant factor. Thus, our study suggests that the effect of the emergency fund hypothesis merits further investigation in future studies based on countries’ specific characteristics.
We also test the effect of insurer reputation risk on consumers’ lapse decisions. The insurer’s reputation hypothesis is particularly relevant in China because of its rapid increase in the volume of premiums written accompanied by the regulatory lag and the lack of qualified labour supply. We proxy insurer reputation risk with negative news issued publicly by the regulator on the specific insurers. We find reputation risk has a positive relationship with the lapse rate, evidenced by the significant and positive coefficient of the negative news variable.
In addition, the IRH is supported by the emerging market data. Note that there is an extreme imbalance in the product mix in China. The investment-type products, including participating insurance, universal insurance and unit-link insurance, make up a dominant proportion of all products, as shown in Table1. Compared with consumers of protection- and health-type products, consumers of investment-type products are more sensitive to the impact of change in financial markets on their personal financial portfolios. We find that the increases in interest rates have significant albeit opposite impact on the lapse rate of investment- and protection-type products, indicating that future study of the impact of interest rate on lapse should differentiate between insurance products.
We also find that insurer business age and business concentration are negatively related to lapse rates. Firm size and leverage are positively associated with lapse rates. Insurers with a higher proportion of business in investment-type products witness a relatively higher lapse rate than insurers writing more business in the health- and protection-type life insurance products.
When it comes to the impacts of lapse, we provide initial empirical results that high lapse rate weakens an insurer’s financial strength, evidenced by the increase in leverage and decrease in new business written.
Our research is the first study on life insurance lapse rates using regional data in China. Empirical results suggest that life insurance companies should properly manage the lapse risk, and regulators should strengthen the monitoring of lapse. Our findings on migration are not unique to China and can also be applied to other developing markets. Additional research on insurers’ reputation would also be valuable. Last but not the least, while the proportion of life insurance products varies between China and other countries, our research sheds light on the importance of considering the product mix in the lapse study. Similar studies in other countries would help to establish the potential underlying driving force of life insurance lapse.
AcknowledgementsThe authors benefited from discussions and comments by Mary A. Weiss, Hong Mao (discussant) and seminar participants at the second Shanghai Risk Form, Zhengzhou University, Capital University of Economics and Business, and Shanghai University of Finance and Economics.
