Entrepreneur: Start & Grow Your Business

Determinants of the lapse rate in life insurance operating companies.

by Mauer, Laurence^Holden, Neil

Review of Business • Fall, 2007 •

Introduction

The purpose of this study is to determine whether factors influencing the lapse rates on ordinary life insurance products can be identified and their importance statistically assessed. In this study, we refer to ordinary life insurance as including the two traditional components of term and whole life. The lapse rate is a key operating parameter that reflects both consumer behavior and insurer managerial decisions involving life policies. Projections of the lapse rate are also critical in structuring securitization arrangements.

Interest in lapse rate determinants has become more prominent in recent years as some financial service firms have applied securitization techniques to life insurance policies held by individuals. Under securitization arrangements, the insured person conveys the payment rights of his/her policy to the firm structuring the securitization product. In return, the insured person receives a one-time cash payment. The securitizing firm pools these policies, using them as the basis for asset-backed securities. Insurance companies, too, may engage in the securitization of their life insurance liabilities by transferring life policies and the assets that back them. For example, such an action was undertaken by Prudential Financial in 2001 as a component to its demutualization. In both cases, assumptions concerning lapse rates are basic to the securitization process. In this paper, we do not attempt to assess the impact of securitization, since this trend is still in its early stages. (1) Rather, we direct our efforts to expand our understanding of lapse rate determinants.

Of special interest in the area of lapse rate analysis is the role of the financial stress position of the insurer, and whether and in what ways stress considerations may influence policyholder lapsation/persistence. To our knowledge, financial stress has not been previously examined in the framework of a formal cross-sectional statistical study to identify the influence on the lapse rates for ordinary life policies.

The model presented in this paper is shown to work best in the case of life companies operating under relatively high financial stress.

We interpret this result as suggesting that a life company under financial stress operates with a reduced level of managerial discretion since company management, in these circumstances, is typically concerned about the company's credit rating position. In contrast, the ordinary life lapse rates for companies characterized by low and mid-range financial stress do not closely fit the model. These findings provide analytical direction to studies of ordinary life products by showing the important role of the financial stress dimension.

Lapsation Issues Shaping Model Development

The lapse rate on life policies has traditionally been one of the central parameters in the managerial framework for life insurers. Consumer demand is widely recognized to be sensitive to pricing for both term and whole life product lines. These ordinary life products are regarded as "commodities." That is, these products are relatively homogeneous in nature, offered by a large number of insurers, and are competitively priced. Lapsation and policy terminations may be initiated by consumers at any time by failing to pay premium billings on term and whole life policies. The non-renewal of term and whole life policies may also occur on renewal anniversaries for policies with guaranteed renewable riders.

From the company's perspective, life insurance products, and especially whole life products, typically entail large underwriting and upfront origination costs, heavily driven by sales costs and commissions. (2) This cost structure provides insurers a strong motive toward lower lapse rates. Too high a lapse rate may impair the ability of the insurer to recoup these costs, given the projected benefits payouts required under these lines of insurance. Companies that are in a stronger market position will be able to price more aggressively than companies in a weaker market position.

From the consumer's perspective, both term and whole life policy holders with health or other insurability problems tend to lapse less frequently, because their alternatives are limited and can be more expensive. The effect is to introduce a tendency toward adverse selection. Healthy individuals may lapse or fail to renew at the guaranteed-renewal dates if they can find less expensive term insurance when passing a medical exam. This adverse selection effect causes the insurer to experience a higher rate of claims on the remaining policies than would have been expected from the entire pool of original insureds. (3)

Managers in the life insurance industry incorporate assumptions about the expected level of lapsation in the design of life products. Unfortunately, information relating to the planned or anticipated lapse rates for companies is proprietary, and thus is not generally available to the public. By studying life companies for which data is available, conclusions can be drawn regarding the factors that affect life policy lapsation.

The Statistical Approach, Data Availabilities, and Variable Specification

This study uses the single equation-multiple regression statistical framework to examine the issue of lapse rate determinants. The cross-section analytical approach is selected. In the course of our study, we identified heteroskedasticity problems, and chose to deal with these using the White (1980) estimation technique.

The lapse rate and other data in this study are compiled based on the set of 162 companies rated for financial stress by the Moody's Investors Service credit rating organization. All available Moody's rated companies were used. This company set was matched to operating life insurance companies reported by the National Association of Insurance Commissioners (NAIC). The data used were taken from the Five-Year Historical Data summary tables in the NAIC "Blue Book." After accounting for data availabilities, the number of useable company observations was 139. These firms constitute 52% of the total life policies in force in the United States in 2003.

The base year for our work is 2003; this was the latest available year at the time of this study. By 2003, a slow economic expansion was widely acknowledged to be underway, following the recession that ended in November 2001. This base year is not cyclically exceptional in the sense of being neither a year of recession nor the peak of a boom.

The company-level characteristics considered in this study, and the specifications for these variables, are now considered.

Lapse Rate. As compiled by the NAIC organization, the lapse rate is reported for ordinary life insurance products. To form the lapse rate for a specific life insurer, the value of lapsed policies for ordinary life products is divided by the average total life insurance in force during the time period. This ratio is multiplied by a scaling factor of 100. Ordinary life policies include the following types of insurance plans: level term life; decreasing term; renewable term; traditional whole life; interest sensitive whole or universal life; and graded-premium whole life.

Financial Stress Ratings. Several outside agencies provide financial stress ratings, including: Moody's, A. M. Best, and Fitch. After reviewing the methodologies of these agencies and after discussions with practitioners and state-level insurance regulators, we chose the Moody's ratings because of the greater degree of delineation provided by their evaluation system. We wish to examine whether the Moody's financial stress measure is statistically supported as a determinant of the lapse rate.

Life operating companies face the decision of how to structure assets and liabilities to achieve specific target financial ratings. The lapse rate is one of the internal operating parameters that many analysts believe will be affected by a company's rating. For our purpose, the Moody's ratings range from Aaa (as 1) to B2 (as 15). A positive relationship is expected between financial stress ratings and the lapse rate.

These considerations are examined in this study for the Full Sample of life companies taken together, and also by segmenting the sample, as follows: the Low Risk panel consists of all life companies rated by Moody's as Aa1 and above; the Mid-range Risk panel, for Aa2 to A1; and the Higher Risk panel, with ratings below A1. In the segmented analysis, the effect of the financial stress rating will be transferred to the constant term within the linear regression framework. The constant intercept term is expected to be higher in the case of the Higher Risk panel, relative to the Mid-range and Low groups. The Moody's ratings variable is included in the segmented analysis to capture any possible "within panel" financial stress effects.

The Joint Pricing Variables. Ordinary life insurance pricing can be interpreted at two levels. The first interpretation is the traditional time-series oriented microeconomic perspective, where pricing serves as a mechanism for rationing consumer demand. Under the traditional micro interpretation, a higher price may contribute to increased lapsation, a positive relationship.

The second interpretation of ordinary life pricing arises from the inherent complexity of life insurance markets in a cross-sectional sense, the data dimension selected for this paper. Some companies have pursued sales and marketing strategies that result in stronger market positions. Examples of such market specializations or niches are companies that market to fraternal or religious organizations or that address specific geographic or urban-rural market segments. (4) In this context, the pricing variable may serve as an indicator of the ability of a life company to acquire a sufficient position of market strength to allow its price to deviate from pricing on average across all companies. Under this interpretation, a higher price may indicate a stronger market position, implying a lower lapse rate, a negative relationship. (5)

The approach of this study provides a natural basis for evaluating which of these two interpretations of the pricing variable is dominant in a cross-sectional context. To address this question, our study employs two pricing-related variables: one for pricing (per-unit premium revenues) directly; and the other, a compositional variable to reflect the mix between term and whole life policies. The average price per dollar of ordinary life is computed by dividing the premium income from ordinary life by the value of ordinary life in force. The compositional variable is the ratio of term life in force to whole life in force. A higher percentage of term policies (without a renewal rider) may be associated with a lower lapse rate, given that the expiration of non-renewal term policies is not regarded as lapsation; in contrast, the cancellation of a whole life policy is reported as a lapse. The inclusion of the compositional variable in the analysis has the effect of holding the term-whole life composition constant, allowing the isolation of the pricing effects. When combined together in a regression framework, these joint variables should capture the role of the pricing factor as it affects the lapse rate. (6)

Life Insurer Size. Insurer size may influence the lapse rate by serving as a proxy for the financial health of the life company in ways that are not fully captured by the Moody's financial stress variable. Under this view, a large company may achieve a higher level of diversification than a smaller firm. In addition, the larger firm may have support in the event of extreme financial stress from the "too big to fail" doctrine that has been invoked from time to time by regulatory authorities in allied financial service fields. Given these perspectives, we wish to account for possible size effects in model specification. Following earlier studies, we use the natural log of the total admitted assets of the insuring company. We expect a negative association between the size variable and the lapse rate.

Ordinary Life Profitability. The management of ordinary life products within companies will be influenced by the relative importance of ordinary life insurance lines within the overall profit mix of the firm. Companies for which ordinary life products are important in the product mix will be more responsive to adverse changes in the lapse rate for these products. We address this issue using a profit margins measure. This is defined as the ratio of the net operating gains from ordinary life policies (after Federal income taxes), divided by total net gains across all product lines for the company. (7)

We expect companies to adjust their marketing and sales strategies to emphasize those product lines with the highest profitability, whether sold using commission-compensated agents or a non-agency distribution system. Through sales and marketing choices, approaches can be undertaken to use information and product alternatives that can influence the lapse behavior of policy holders. In cases where ordinary life would be more profitable relative to other lines of business, we would expect companies to undertake strategies to hold down the lapse rate. In such cases, we would expect a negative association between this profitability variable and the lapse rate in the ordinary life business line. We identify this circumstance as "lapse avoidance."

It is possible, however, that life companies may wish to de-emphasize coverage in the life area. In this case we would have circumstances of "lapse preference," a positive relationship between the ordinary life policies and the lapse rate. Where a firm wishes to raise its lapse rate, such a company might pursue the offering of incentives to migrate ordinary life policy holders to more attractive business lines. It is not clear on an a priori basis whether lapse avoidance is the dominant mode or whether the preference is for higher lapse rates for the sample of firms included in this analysis.

Empirical Results and Interpretation

The model described in the previous section was tested using the least squares (LS) regression procedure. In early investigations, it was determined that the LS residuals exhibited evidence of unequal variances, identified more commonly as heteroskedasticity. To correct this problem, we adopted the White (1980) estimation procedure, which allows model estimates to be interpreted as consistent, with asymptotically efficient standard errors. The regression results presented in this study are estimated using this technique. Exhibit 1 reports descriptive statistics for the lapse rate and the independent variables.

In Exhibit 2, the Full Sample equation presents the model estimates using 139 company observations. While the equation for the full sample is statistically significant at the 5 % level, the R bar 2 for the regression is rather low, at 6%. The equation has four statistically significant variables. Of special note in this regard is the positive-signed coefficient on the Moody's rating variable, which is significant at the 1% level. Also significant are PRICE, at 10%, and its joint variable TERMWHL, at 1%. PRICE is negative-signed, consistent with the view that the relative market position of the sample companies dominates the PRICE variable. TERMWHL is also negative-signed, reflecting the lower lapse tendency of term policies, compared with whole life policies. And finally, the ordinary life profitability measure ORDPM is significant at the 1% level with a negative sign, suggesting that internal company policies on balance are "lapse avoidance" oriented.

To examine the issues related to financial stress, we divided the companies into three panels, according to the Moody's ratings, as described above in the variables specification section. The Higher Risk group, the Mid-Range Risk group, and the Low Risk group have 46, 54, and 39 company observations, respectively.

It is seen in the descriptive statistics for the three panels that the mean value for the lapse rate is highest in the case of the Higher Risk group; this group is also characterized as having a significantly higher TERMWHL, the ratio of term to whole life policies outstanding. However, regarding the PRICE, LNSIZE, and ORDPM variables, there is considerable commonality in the mean values among the three panels.

The statistical findings from the Equation 1 regression model are reported in Exhibit 2 for Panels A, B, and C, in addition to the Full Sample. In the panel regressions, we retain the Moody's variable as a test for "within group" financial stress effects and to maintain comparability with our Full Sample model results. For these results, and in contrast to the Full Sample, the Moody's variable was not statistically significant in any of the panels, reflecting its limited variability within each panel grouping.

In the Panel A case for Higher Risk companies, the model fits quite well, with an R bar 2 of 44%, a high showing in a cross-section study. For these companies, the joint pricing variables are significant; PRICE, at 10%, and TERMWHL, at 1%. Additionally, ORDPM and the LOGSIZE variable are significant at 1%. The signs of all variables in the Higher Risk group coincide with those found for the Full Sample group. Additionally, the constant term in the Panel A case is large and significant at the 1% level, as anticipated from the approach of segmenting on the financial stress dimension.

For the life companies rated as Mid-Range Risk, Panel B, the notable statistical result is that the model's R bar 2 is close to zero and does not pass conventional levels of statistical significance. In the Panel C case for the Low Risk companies, the R bar 2 is again low and is not statistically significant.

In comparing these results, it is evident that lapsation determinants between the Higher Risk Panel and the other two groups are the result of different processes. The simple model developed in this paper does well for those companies faced with high financial stress. In contrast, the model fits less well in the case of the Low Risk and Mid-range Risk companies. Our explanation for this result is that the companies operating under a higher level of financial stress are constrained by their circumstances to work under operating procedures that are captured by our model. In effect, the managerial discretion of the Higher Risk companies is concentrated to a narrower range of decision choices and strategies, influenced by their weaker financial positions.

The Low Risk and Mid-Range Risk companies are less bound by these constraints. These companies pursue a wide range of operating strategies, directed at objectives that are more complex in terms of product development, new markets addressed, and management restructurings. (8) Such strategies are typically multi-year in nature. This complexity is simply not captured in the statistical model used in this study.

Summary and Implications for Further Work

The lapse rate is a key operating ratio for insurance companies providing ordinary life products. Companies offering these insurance products can take actions to influence their lapse rates. Further, understanding lapse rate determinants is also critical for structuring securitization offerings of life insurance-based assets and liabilities.

This study presents a model of lapse behavior for 139 companies offering ordinary life, based on the following factors: Moody's rating of financial stress; two pricing-related variables; insurer size; and the relative profitability of ordinary life. We find this model is most representative when applied to ordinary life insurers in the higher financial stress category (ratings below A1). We believe that this result occurs because companies in this risk category are forced by their financial stress circumstances to operate their ordinary life business in the basic ways that are captured by this model.

These findings suggest that work focused on the lapse rate can provide further insight into company behavior in the area of ordinary life insurance products. Future research may follow more complex methodologies, such as simulations of the behavior of consumers and insurers. Additionally, obtaining proprietary lapse data, perhaps through using a case methods approach, may yield further insights into lapse behavior. For instance, data drawn from companies on a case-by-case basis can allow the study focus to shift to more detailed customer groups and product lines. As this research proceeds, guidance will be provided to consumers as they make their ordinary life insurance choices.

References

1. Carson, J. and R. Dumm. 1999. "Insurance company-level determinants of life insurance product performance." Journal of Insurance Regulation 18 (2) (Winter): 195-206.

2. Carson, J. and R. Dumm. 2000. "The relationship of insurer characteristics and life insurance surrender values." Journal of Financial Service Professionals: 54 (5) (September): 86-90.

3. Cummins, J. D. and M. A. Weiss. 2004. "Securitization of life insurance assets and liabilities." Working Paper 04-03, Wharton Financial Institutions Center, University of Pennsylvania (January).

4. Matthewson, G. F. 1983. "Information, search and price variability of individual life insurance contracts." Journal of Industrial Economics (32): 131-48.

5. Moody's Investors Service. 2003. "Credit issues and trends for U.S. life insurance, special comment." Global Credit Research, Appendix II, (May), 15-20.

6. Praetz, P. (1980). "Returns to scale in the U.S. life insurance industry." Journal of Risk and Insurance 47 (3): 525-533.

7. Purushotham, M. 2006. U.S. Individual Life Persistency Report, LIMRA International and the Society of Actuaries.

8. Schlesinger, H. and M.G. & von der Schulenburg. 1991. "Search costs, switching costs and product heterogeneity in an insurance market." Journal of Risk and Insurance 58 (1) (March): 109-119.

9. White, H. 1980. "A Heteroskedasticity-consistent covariance matrix estimator and direct test for heteroskedasticity." Econometrica 48: 817-838.

Endnotes

(1) For a more complete description of these trends in securitization, see Cummins, J. D. & Weiss, M.A. (2004). Securitization of life insurance assets and liabilities. Working Paper # 04-03, Wharton Financial Institutions Center (January).

(2) This is particularly true in whole life because the first-year commission is typically quite high, often 90% or more of the first-year premium.

(3) This phenomenon is more pronounced in the older insured age groups.

(4) Additionally, in recent years, life companies have undertaken cross-selling with products of non-insurance financial services firms, as a further way to differentiate their services.

(5) Moody's indicates that the lapse rate is one of the considerations that is used in rating insurance companies. However, in its methodology description, Moody's does not indicate how much weight is placed on the lapsation factor. It is possible that the causation may run from the lapse rate to the ratings, rather than the direction indicated in this study. If the lapse rate plays a major role in Moody's rating methodology, we would expect to identify this when regressing the lapse rate on the ratings variable alone; that is, not in the context of the model presented in this study. We test this interpretation for the sample as a whole, using the White estimation technique, and find there is no statistically significant relationship between the lapse rate and ratings variable used alone.

(6) To guard against the possibility of a reverse causality interpretation, the pricing variable is lagged one year, to 2002.

(7) To guard against the possibility of a reverse causality interpretation, the operating income ratio is lagged one year, to 2002.

(8) Additional examples include some highly rated companies that emphasize structured settlement annuities also sell large term insurance policies to affluent customers, who typically have very low lapse rates. Also and in a different direction, some traditional mutuals still carry on the career-agency system that emphasizes whole life; these mutuals have high attrition among newly-hired agents, which tends to increase lapsation as attrition severs the company's connection to customers.

Laurence Mauer, The Peter J. Tobin College of Business, St. John's University

Neil Holden, College of Business, Ohio University