On Voluntary Terminations of Life Insurance: Differentiating Surrender Propensity From Lapse Propensity Across Product Types

Abstract

Understanding the causes of voluntary terminations is important to the service quality, profitability, and risk management of the life insurer. This article extends the literature on the determinants of the termination propensities in four aspects. First, we decompose voluntary terminations into surrender and lapse and build models accordingly. This decomposition is important because the motives, causes, and consequences of lapse and surrender are distinct. Second, we construct models for the surrender and lapse propensities by product type. Without such construction, one would overlook the diverse motivations of buying different types of product and distinct behaviors of terminating product holding. Third, we introduce new explanatory variables (commission ratio and occupation of the insured) in modeling the propensities and these variables are found to be significant. Fourth, this is the first article on the determinants of voluntary terminations for the market of Taiwan with implications for other regions having significant Chinese people presence.

ACKNOWLEDGMENTS

The authors are grateful to the two anonymous referees and co-editor Yijia Lin for their constructive and detailed comments. The resultant revisions enhanced the quality of the article considerably.

FUNDING

The authors are grateful to the Ministry of Science and Technology of Taiwan for its financial support (project number NSC 102-2410-H-004 -027-MY3, MOST 105-2410-H-004-070-MY3, and MOST 108-2410-H-004-081-MY3), as well as to California State University Northridge (CSUN) and California State University Fullerton (CSUF) for the kind support during Jason’s visit. A significant portion of the research was completed while Jason was visiting CSUN and CSUF.

Discussions on this article can be submitted until January 1, 2023. The authors reserve the right to reply to any discussion. Please see the Instructions for Authors found online at http://www.tandfonline.com/uaaj for submission instructions.

Notes

1 Whether the policyholder of a lapsed policy receives benefits depends on whether the policy contains a nonforfeiture value clause. The nonforfeiture value originates from the policy reserve. Policy reserves are also the origins of the cash values that policyholders can access when they surrender their policies. If a policy has cash value available for surrender, lapsing this policy means the holder is entitled to a benefit no smaller than the cash value. Few policies have no nonforfeiture value. An example is yearly renewable life insurance. Voluntary terminations of this kind of policies have little impact on insurers and are not considered in this article.

2 Increases in surrender activities would probably be accompanied with decreases in new businesses, which exacerbate insurers’ liquidities (Russell et al. 2013).

3 The empirical data used in the literature do not differentiate surrender from lapse, and the numerical analyses focus on how surrenders affect insurers.

4 On the other hand, Tsai, Kuo, and Chen (2002) demonstrate that providing surrender options to policyholders might be beneficial to life insurers if policyholders do not exercise these options when the options have positive values to them (e.g., at times when interest rates are significantly higher than the crediting/pricing rates of the policies) but exercise the options when the option values are negative (e.g., when interest rates are significantly lower than the crediting rates). This seemingly contradictory finding indeed highlights the importance of appropriately modeling voluntary termination behaviors.

5 Surrender would be a sound action for policyholders, although Tsai et al. (2002) questions such rationality. Insurability after surrendering could be an issue, though.

6 The term “lapse” used in QIS5 is equivalent to our “terminations” that include the behaviors of both surrender and lapse.

7 There are some micro-oriented papers built upon theoretical modeling about how individual policyholders decide whether to terminate their policies or not. They impose specific assumptions on the transition probabilities to terminations (Buchardt, Moller, and Schmidt 2015), terminations’ intensity (Barsotti, Milhaud, and Salhi 2016), or termination rates (Loisel and Milhaud 2011; Milhaud 2013; Carson et al. 2020). Fang and Kung (2012) present and empirically implement a dynamic discrete choice model of life insurance decisions to explain terminations.

8 Note that Cole and Fier (2020) consider the substitutions between surrenders and policy loans. Lapses do not generate cash inflows to meet emergency needs of households.

9 Other hypotheses about the behaviors of voluntary terminations include the arbitrage hypothesis (Clancy 1985; Carson and Hoyt 1992), alternative funds hypothesis (Dar and Dodds 1989; Hoyt 1994), inflation hypothesis (Russell 1997), and policy replacement hypothesis (Outreville 1990; Russell et al. 2013).

10 We are able to do this because the data-providing insurer kept detailed records on changes in the statuses of policies, including policyholders’ actions about paying premiums. Refer to Section 3.1.

11 Statistics about the life insurance markets in the United States (see 2020 Life Insurers Fact Book by ACLI) and Taiwan (see http://www.lia-roc.org.tw) show no downward trend of purchasing conventional life insurance products in the United States and Taiwan in the low interest rate environment.

12 Insurance penetration is defined as the total premiums of the market divided by the gross domestic product (GDP) of a country or region; insurance density means premiums per capita. The life insurance market of Taiwan was ranked eighth in the world in 2019 in terms of annual total premiums. Total insurance penetration remains number one and life insurance density increased to number four in 2019.

13 Yu, Cheng, and Lin (2019) investigate determinants of lapse rates of the life insurance industry in China. What they use is firm province-level data instead of policy-level data.

14 The behaviors of insurance holders depend on regional insurance markets and are affected by sex, social status, education, income, religion, and ethnicity, among others (e.g., Chui and Kwok 2008, 2009). A Chinese person in Hong Kong behaves somewhat differently from an English person in Hong Kong in relation to food consumption, retail habits, eating, drinking, shopping, technology adoption, value systems, identities, investments, and insurance purchases, among others (e.g., De Mooij and Hofstede 2011; Lee, Trimi, and Kim 2013). Therefore, the voluntary termination behaviors of Chinese people depend on regional insurance markets and the complex factor of “Chinese identity.” Chinese people in Taiwan may behave differently in the purchasing and terminating of life insurance than those in China or Hong Kong do, but the particularities of these differences can be addressed in future research; our article marks the first step toward identifying the factors affecting the life insurance termination behaviors of Chinese people.

15 Some studies employ dummy variables to take into account the possible differences in the propensities toward voluntary terminations across product types. Using dummy variables would not fully capture the differences, however.

16 We suspect that what some of these papers measure is indeed the insureds’ ages rather than policyholders’ ages. For instance, Renshaw and Haberman (1986) mention “entry age” and their age group starts from age 15 years. Milhaud, Loisel, and Maume-Deschamps (2011) calculate “underwriting age” and some are below 20 years. Pinquet, Guillén, and Ayuso (2011) design a bonus-malus coefficient to measure a “policyholder’s” health status. Eling and Kiesenbauer (2014) mention “the youngest policyholders’ … policies might initially be purchased by the policyholder’s parents.”

17 Note that most agents are exclusive agents in Taiwan.

18 Trigo-Gamarra (2008) and Eckardt and Räthke-Döppner (2010) find that independent agents provide better services than exclusive agents in the German market. Eling and Kiesenbauer (2014) further find that the channel of independent agents results in a lower early termination rate.

19 Brokers remain an immaterial distribution channel in Taiwan. Independent agents are more popular than brokers but their market share is minor as well, in particular with the data-providing insurer during the sampling period.

20 Ideally, this variable should indicate the status change due to surrender and lapse only. We attempted to use this definition, but encountered the non-identification problem. The year dummies indicating surrender or lapse as a whole (there are 16 – 1 = 15 year dummies) are almost the same as the dependent variable, with the difference being merely one dummy year.

21 We do not establish hypotheses to explain how a factor would affect merely one type of behavior. Instead, we speculate that both types of behaviors are affected by the same set of factors to different extents or sometimes with opposite effects. Our speculations are supported by empirical results presented later in Tables 7 and 8.

22 This insurer was ranked number 9 in Taiwan and would be equivalent to the 52nd largest life insurer in the United States by total assets at the end of 2013.

23 Examples of nonconventional life insurance are variable life, universal life, and interest-linked life insurance products. The sample size is adequate for statistical analysis but inadequate for representing the market, which is a limitation of the current study.

24 Descriptions of a representative policy for each type of products are in Appendix B.

25 We define the savings component as the present value of expected premiums minus the present value of death benefit (i.e., the protection component). For the sampled policies in Appendix B, term-life insurance has no savings component while the savings component of whole-life insurance is minimal (0.001%). The importance of the savings component jumps to 90% for E and increases to 96% for LB. Note that the importance of the savings component decreases with age in general. Further note that no savings component does not mean no surrender value. Surrender value derives from policy reserves that reflect the differences between accumulated paid premiums and accumulated costs of insurance from the retrospective point of view. Indeed, levels premiums are the root cause of the accumulation of policy reserves. Therefore, even term life insurance may have policy reserves and surrender values.

26 More specifically, we use the explanatory variables to see whether a policyholder will terminate his or her policy during the period from the inception of individual policies to the end of policy lives or 2013, whichever is earlier.

27 Information about policyholders is relevant as policyholders are by definition the decision makers in cases of early terminations. Such information is unavailable, as in Cox et al. (1992) and some papers listed in Table 1, because underwriting focuses on the information about the insured. Furthermore, the insurer may not be in a legitimate position to secure personal information on policyholders unless there are moral hazard issues or other underwriting concerns. More specifically, the Application for Life Insurance asks personal information about the insured only, unless the insured and the policyholder are different persons. Even in these cases, the application focuses on the relationship between the insured and the policyholder. We speculate that the insured and the policyholder are the same person for most policies, since all insureds are adults. Even when the policyholder is not the insured (e.g., a wife for a husband or vice versa), the early termination decision is probably jointly made by the insured and the policyholder. On the other hand, information about the insured is relevant since such information affects the fundamental demand for insurance. Our results thus may complement some previous papers using the information about policyholders and enhance the understanding of the determinants of early terminations.

28 The data-providing insurer did not sell traditional life insurance through brokers or independent agents.

29 Paying premiums by automatic transfers from bank accounts or through recurring credit card payments is essentially the same to policyholders. We thus regard these two automatic/recurring payment methods as one.

30 We do not include the insureds with ages below 25 years because they are regarded (by demography) as either Childhood [0, 14) or Teenage/Young Adult [14, 25) and probably are not involved in early termination decisions.

31 The exchange rate used in the article is 32 NTD/1 USD.

32 There are few policies with death benefits of 10,000 NTD (a little over 300 USD). We examined and confirmed the consistency of these policies with their specifications (e.g., the consistency between premiums and death benefit). We did robustness checks in which the 1% tails (by policy size) were taken out of the samples and confirmed that the empirical findings remained intact except for making the policy size variable insignificant. Careful readers may find that the largest face amount of our sampled policies (a whole-life policy) is smaller than the largest policy size. This is because the latter policy provides periodic living benefits.

33 With the coefficients of both variables Age and Age² being negative, the propensity toward early termination has a theoretical maximum in the second quadrant on the age–logit(p_i) plane and exhibits a nonlinear decreasing curve in the first quadrant.

34 The findings of Pinquet, Guillén, and Ayuso (2011) and Milhaud, Loisel, and Maume-Deschamps (2011) regarding the Spanish market are contradictory.

35 Endowment (E) and Life Insurance with Periodic Living Benefits (LB) are more savings oriented than Whole-Life (W), as footnote 24 and Appendix B indicate.

36 With 0.0559 and –1.4153 being the coefficients of the variable Policy Age² and Policy Age respectively, the propensities toward early terminations exhibit a nonlinear decreasing curve in the first quadrant up to the 13th policy year.

37 The insignificance of the age variable itself does not change the previous finding that early termination propensities in Taiwan decrease nonlinearly and monotonically with age. The insignificance simply implies that the vertex of the parabola is on the Y-axis.

38 We provide the results of disaggregating riskier tiers in Appendix C. They provide consistent but more detailed results compared to those in Table 7.

Additional information

Funding

The authors are grateful to the Ministry of Science and Technology of Taiwan for its financial support (project number NSC 102-2410-H-004 -027-MY3, MOST 105-2410-H-004-070-MY3, and MOST 108-2410-H-004-081-MY3), as well as to California State University Northridge (CSUN) and California State University Fullerton (CSUF) for the kind support during Jason’s visit. A significant portion of the research was completed while Jason was visiting CSUN and CSUF.