Sales Quality and Corporate Governance - CiteSeerX

1 downloads 0 Views 312KB Size Report
Apr 18, 2002 - questions about the service quality UK life insurance. ... proceeds received (the surrender value) may be less than the premiums .... will often form the main determinant of overall customer satisfaction (Krishnan et al, 1999).
THE UNIVERSITY OF NOTTINGHAM

Centre for Risk & Insurance Studies

Persistency in UK Long-Term Insurance: Customer Satisfaction and Service Quality

Professor Stephen Diacon Chris O’Brien 18th April 2002 CRIS Discussion Paper Series – 2002.III

Persistency in UK Long-Term Insurance: Customer Satisfaction and Service Quality

Stephen Diacon and Chris O’Brien Centre for Risk and Insurance Studies Nottingham University Business School Jubilee Campus, Wollaton Road Nottingham NG8 1BB Tel: +44 (0) 115 951 5267 Fax: +44 (0) 115 846 6667

Presented to the conference, “Global Issues in Insurance Regulation” London, 18 April 2002

1

Persistency in UK Long-Term Insurance: Customer Satisfaction and Service Quality Stephen Diacon and Chris O’Brien Nottingham University Business School

ABSTRACT

The main trust of this paper is to explore the inter-company variations in persistency experience for a sample of 99 U.K. life insurers, using data collected and published by the U.K. financial services regulator. Our first objective is to ascertain whether the undoubted differences in persistency among U.K. long-term insurers are random in nature (as opposed to having some systematic features). Our analysis indicates that there are systematic differences among insurers in withdrawal rates. These differences suggest very strongly that persistency problems do not arise from random factors, but result instead from an inability of insurers to meet the service quality expectations of a wide range of customers. Our second main objective is to determine the nature of any systematic differences in persistency according to company size, efficiency, and ownership structure. We therefore use multivariate techniques to measure the relationship between withdrawal rates and those aspects of service quality that are correlated with variables such as size, new business growth, the expense ratio, and the mutual/stock distinction. The findings suggest that mutual insurers have a better persistency record than stock insurers, offices with higher expense ratios tend to demonstrate significantly higher withdrawal rates, and persistency seems to be negatively related to insurer size.

2

1.

Introduction

There has been considerable disquiet in the U.K. at what are regarded as low rates of the persistency of long-term (life and pensions) insurance business

(Personal Investment

Authority, 2001). Many policyholders have withdrawn from long-term commitments before their contract has expired, and have consequently received a surrender value that represent poor value for money. The poor persistency rates associated with these long-term savings contracts provides tangible evidence of widespread customer dissatisfaction, and raises questions about the service quality UK life insurance. The U.K. financial services regulator provides regular statistics on the persistency of life and pensions policies, and has noted that “overall levels of persistency remain low, particularly for pensions business and policies sold by company representatives” (Personal Investment Authority, 2001, p14). For example, the figures for personal pensions incepted in the U.K. in 1996 indicate that on average, 13.6% of regular premium contracts sold by company representatives failed to persist beyond the first year, and 42.8% experienced withdrawal in the first 4 years. The corresponding withdrawal rates for regular premium personal pensions contracts sold by independent financial intermediaries (IFAs) in 1996 were 10.2% and 37.7% respectively. The PIA also commented that “persistency can be indicative of the quality of the selling, although clearly the reasons why some policies lapse cannot be foreseen either by the individual investor or their adviser at the time of sale” (p3). Persistency rates in long-term insurance contracts remain low in spite of the penalties that customers, intermediaries and product providers, incur from early withdrawal from the contract. Policyholders who withdraw may suffer a financial penalty because the policy proceeds received (the surrender value) may be less than the premiums paid – particularly if withdrawal occurs in the early years of the contract. Similarly salesmen and intermediaries will also suffer, as low persistency means lower renewal commissions. Good persistency is also of vital importance to the financial performance of life insurance companies. In a survey of U.S. life insurers, Moore & Santomero (1999) report that a focus on client retention is the highest ranking corporate objective to achieve future success. Early withdrawal often means that product providers are unable to recoup their business acquisition expenses (as premiums tend to be level-loaded rather than front-end loaded). Furthermore it 3

is increasingly recognised that insurance company profit and growth are stimulated primarily by customer loyalty and retention. This is particularly important for the new stakeholder pensions products (which have a level rate of charge that cannot exceed 1% of the fund each year) since companies depend on such policies being kept in force for several years before they can make a profit at all. The main focus of this paper is on inter-company differences in persistency. A feature of the PIA statistics is the wide variation in persistency rates among different life insurance companies operating in the U.K. market. For example, the four-year withdrawal rates reported by 37 U.K. life insurers for personal pensions contracts sold by company representatives in 1996 varied between 17.2% and 51.9%; the corresponding figures for contracts sold by 23 insurers via the IFA channel varied between 9.2% and 53.1%. The paper attempts to identify factors that separate low from high persistency companies. The remainder of this paper is organised as follows. In section 2 we review persistency literature and discuss the UK long-term insurance persistency experience. Section 3 then goes on to consider what is meant by service quality and customer satisfaction, and relates these to persistency. Section 4 describes some statistical analysis of inter-company differences in withdrawal rates. The conclusions and some recommendation are in section 5.

2.

A Review of Persistency

In this section we review some previous work on persistency. However, the precise definitions adopted differ between studies, and so care is needed in making comparisons. The persistency of long-term insurance business is generally measured as the proportion of business sold which remains in force after a specified period and, for regular (as opposed to single) premium business, any premiums due have been paid. Defined in this way, persistency incorporates the effect of death (and health-related, etc) claims, whereas persistency analysis is primarily concerned with the policyholder (voluntarily) withdrawing from the contract, where withdrawal covers a policy lapsing or being surrendered, or (in the case of a regular premium policy) being made paid-up. Persistency rates are sometimes adjusted to exclude deaths, etc. Broadly the opposite of persistency rates are withdrawal 4

rates, i.e. the proportion of policies in force at the beginning of a period where withdrawal takes place during the period. Some early U.S. analyses are of particular interest. Richardson & Hartwell (1951) analysed persistency data from the life insurance company Mutual Life of New York. They found that withdrawal rates were lower for high-income policyholders. Withdrawal rates were also lower for more experienced salesmen, with new agents who leave early having the highest withdrawal rates. Buck (1961) analysed the first-year lapse rates of Lincoln National and discovered a strong negative relationship between the number of years’ experience of the salesman and withdrawal rates. He also found that withdrawal rates were significantly lower if the person had a previous policy with the company. Patrick & Scobbie (1969) used data from five Scottish life offices and analysed life policies lapsing or surrendering in 1965. They determined that the rate of withdrawal reduced the longer the policy was in force, and discovered that 79.9% of policies were still in force after 4 years1. Withdrawal rates appeared to be higher for younger customers, and depended on policy type (highest for non-profit whole of life policies). The withdrawal rates for manual workers were 50% higher than for professionals and mangers. The authors suggested that the factors thought to affect withdrawal rates included: increased sales to low income groups who may wish to surrender in the event of temporary financial difficulties, mis-selling, house moves, and a change in economic conditions or tax structure. Proudfoot (1969) and Webster (1969) suggested that the withdrawal rates were higher in Australia and the U.S. respectively, in comparison with the Scottish figures. Crombie et al (1978) analysed the data of seven Scottish life offices in the mid-1970s and found that 81.3% of policies had persisted beyond 4 years. Interestingly, they found that insurers that had high withdrawal rates for one product type had high rates for other types. Furthermore high rates of withdrawal appeared to persist in some companies for policy durations of 15 years and more, suggesting that there were characteristics of high withdrawal offices that were not merely sales-related.

1

The corresponding PIA data for endowment policies sold in the U.K. by IFAs in 1996 averaged 83.1%.

5

Renshaw & Haberman (1986) analysed the Crombie et al (1978) data further, and drew attention to the dangers of using cross-section data to analyse withdrawals, (which were better considered by cohort over time). They looked at insurer-specific effects, and also age of policyholder, duration of policy in force and type of policy, and concluded “that, essentially, all offices experience a similar pattern of lapses across the different combined levels of the other factors under investigation, but to varying degrees of intensity” (p472). Increased concern about poor persistency in the 1990s motivated the U.K. regulatory authorities to collect industry-wide data for more detailed analysis. The Securities and Investment Board (1991) published data prepared by consulting actuaries AKG that showed withdrawal rates of 13.8% (17.9%) in the first year for non-linked life (pensions) business, and linked business had higher rates still. Sales through direct sales forces had higher withdrawal rates than sales through independent financial advisers. However, Laslett et al (2002) reported that commission bias in the giving of advice is more serious in the IFA sector than for company representatives (where the market research suggested there was little, if any, evidence of such bias). The Office of Fair Trading (1994) expressed concern about low surrender values on life policies and felt that indirect evidence suggested a link between low surrender values and high lapse rates. A link between economic factors and withdrawal rates in the U.K. was estimated approximately by Mehta (1992), who found that a stock market decline of 15% was associated with an increase in the withdrawal rates on regular premium products of 3% in the first policy year and 1% p.a. thereafter. The latest UK long-term insurance persistency statistics are provided in Table 1. To illustrate, of those endowment policies sold by company representatives in 1996, 93.3% persisted at least a year and 76.8% at least four years, implying a withdrawal hazard rate of 6.7% in the first year, and 17.68% over the next three years2. Policies sold by company representatives have lower persistency rates than those provided via independent financial advisers for every year between 1993 and 19993. 2

The first-year withdrawal hazard rate is obviously h(1) = 100-93.3 = 6.7%. The withdrawal rate for years 2-4, for those policies that were in force at the start of year 2, is then h(2_4) = (23.2-6.7)/0.933 = 17.68% 3 The PIA provides data for two other sales channels for endowments: policies sold through direct offer advertisements have persistency similar to IFAs; but persistency is markedly lower for sales through home service offices.

6

Some recent figures from the U.S. suggest that U.K. experience is not especially low. Moore & Santomero (1999) show that 87-95% of life policies are in force after 13 months, 60-82% after 4 years (the figures vary by product). The Society of Actuaries Direct Response Persistency and Mortality Task Force (1995) implied that only 39% of term and 52% of whole life policies sold direct would be in force after 4 years. TABLE 1:

Endowment Policies

Personal Pensions

Percentage of Policies Persisting for at least 1 and 4 Years (by Sales Channel).

Year of Inception 1993 1994 1995 1996 1997 1998 1999 1993 1994 1995 1996 1997 1998 1999

Company Independent Representatives Financial Advisers 1 year 4 years 1 year 4 years 91.7 76.7 94.4 83.8 91.8 76.9 94.6 83.6 92.2 76.4 94.8 81.9 93.3 76.8 95.1 81.2 93.2 95.8 92.1 95.5 91.7 95.4 84.1 56.7 91.5 70.5 83.7 57.1 90.9 66.9 85.4 57.8 90.2 64.7 86.4 57.2 89.8 62.3 85.6 90.2 85.2 88.3 84.7 87.2 Source: Personal Investment Authority (2001)

Further details on the U.K. persistency experience over the period 1993 to 1999 are provided in Charts 1 to 4. These show the average year-on-year withdrawal hazard rates4 plotted against persistency term (for the first four years of the policy), for the two main types of regular premium policy (endowment and personal pensions) and two types of distribution channel (company representatives CR, and independent financial advisers IFA). Charts 1-4 about here The average withdrawal hazard rates in Charts 1-4 represent the probability that a policy will be withdrawn in policy year t, given that the contract was still in force at the end of the

7

previous year. Raw data was obtained from the insurance regulator’s annual persistency survey (Personal Investment Authority, 2001), and the average hazard rates were calculated as unweighted averages of individual company withdrawal hazard rates. The number of companies in the dataset varies by year, policy type and distribution channel as follows: Endowment/CR 40-60; Pensions/CR 37-54; Endowment/IFA 15-20; Pensions/IFA 23-32. Several interesting features emerge from a comparison of the Charts: •

The average withdrawal hazard rates confirm the pattern of persistency figures in Table 1. Hazard rates are higher for the company representative channel than for IFAs, for both endowment and pensions policies.



The hazard rates are usually highest in the first year of the policy (with the exception of Pensions/IFA) and then decline as the term increases. This pattern provides circumstantial evidence of problems with the initial sales process.



After an initial fall, the hazard rates often begin to rise in years 3 and 4; again the Pensions/IFA case is an exception. This pattern suggests that other factors, in addition to sales quality, may then come into play to affect persistency.



The case of Pensions/IFA is quite different: hazard rates appear to rise from an initial low, and peak around year 3. This pattern suggests that pensions policyholders using the IFA channel take rather longer to realise that policy may not be appropriate for them.

3.

Customer Satisfaction and Service Quality

The Distinction Between Customer Satisfaction and Service Quality Customer satisfaction refers to the degree to which customers derive value from the quality of products and services provided by a firm. In the case of financial services, where the products are intangible and are sampled only rarely, the services accompanying the product will often form the main determinant of overall customer satisfaction (Krishnan et al, 1999). Service quality, on the other hand, refers to the performance of the provider, in terms of the technical quality of what is provided and the functional quality of how it is provided (Lassar 4

The raw data shows that proportion of policies that are still in force at the end of policy term t, denoted by

8

et al, 2000). However the intangibility of services (as opposed to products) means that a customer’s perceptions of both service quality and customer satisfaction are necessarily subjective, and may be intertwined. In an attempt to measure customer satisfaction in financial services and investigate the link with service quality, Krishnan et al (1999) design a questionnaire instrument to investigate four distinct factors relating to the customer experience of a firm’s offerings - in terms of personal contact, usage of telephone and IT systems, product performance, and periodic financial statements. Lassar et al (2000) attempt a similar exercise, but provide separate measures of customer satisfaction of the firm’s technical offerings (in terms of product and systems performance) and functional offerings (in terms of the interface with front-office staff). The results confirm the findings of previous research that customer satisfaction with a company’s services is determined to a large degree by the quality of service the customer receives (Parasuraman et al, 1985; Cronin and Taylor, 1992). Considerable efforts have been expended since the pioneering work of Parasuraman et al (1985) to measure service quality, and then to explore the impact on customer satisfaction and firm performance5. The principal measure of service quality attempts to measure the gap between customer expectations and perceived actual service performance, in terms of five dimensions relating to product and process (namely reliability, responsiveness, assurance, empathy and tangibility). Of these dimensions, the first four relate to the element of human interaction/intervention in the service delivery: their inclusion has been validated by numerous studies which highlight the importance of the customer interface in determining service quality (eg Lewis, 1995; Roth and Jackson, 1995; Krishnan et al, 1999). Little formal research appears to have been conducted which attempts to measure service quality in insurance6. However, there appears to be a widespread view that service quality is driven primarily by the technical and functional performance during the sales process particularly for long-term life and pensions policies. Thus Diacon and Watkins (1995, p252) and McCabe et al (1997, p62) suggest there are strong links between an insurer’s service S(t). The year-on-year withdrawal hazard rate for term t was then calculated as (S(t)-S(t-1)) / (1-S(t-1)). 5 There is now a considerable literature that charts the beneficial impact of quality improvements on firm performance (see Rust et al (2000) for a review). 6

Wells and Stafford (1995) is an exception, although it concentrated on motor insurance in the U.S..

9

quality and the quality and professionalism of advice provided by sales staff7. Furthermore, surveys of financial services consumers in the U.K. often indicate concern about the performance of the sales process8 (Diacon and Ennew, 2001). Public concern about the service quality of long-term insurers has been hightened by a series of mis-selling problems and official criticism. In the U.K., Gower (1984) was critical of the lack of training of life insurance salesmen and the conflicts of interest that can arise from commission payments. This led to the marketing and sale of most life assurance policies being regulated by the Financial Services Act 1986. Under the Act, salesmen had to comply with a code of conduct that required them to consider the needs and circumstances of their customers and give “best advice”. This code did not prevent the widespread mis-selling of personal pension business in 1988-94, where the regulators took action following the report of KPMG Peat Marwick (1993)9. However, there is also evidence that service quality in the U.K. life and pensions industry is far from uniformly low. A survey undertaken by the insurance regulator (the Financial Services Authority, 2000) indicated that, of people who had taken out a policy within the last 12 months, 66% were very and 28% fairly satisfied with their adviser. Furthermore Laslett et al (2002) found that the advice market was “not riddled with [commission] bias” (page vi) and indeed there was no detectable bias on regular premium products. Persistency, Service Quality and Customer Satisfaction Krishnan et al (1999) commented that “the basic argument is that satisfied customers of a firm decide to stay with the firm for future business” (p1194). It is natural therefore to suggest that the persistency of business can be regarded as an indicator of customer satisfaction. Furthermore the pattern exhibited in Charts 1-4 indicate that this is likely to be determined by service quality in terms of the quality of the sales process.

7

The industry has been conscious of criticisms of its sales practices, and has introduced steps intended to improve quality. In the U.S., the Quality Insurance Congress was founded to facilitate change that would improve customer satisfaction (in 1993). In the U.K., the Association of British Insurers (2000) has formed a “Raising Standards” scheme which has been adopted by a number of companies. McCabe (2000) charts the difficulties that companies have experienced in implementing quality initiatives. 8 In terms of the possibility of experiencing unacceptable sales pressure, receiving unsound or biased advice, or being levied high but unobservable charges.

10

The casual link between service/sales quality and customer satisfaction (in terms of persistency) has been made by several commentators. Pullan (1995) noted that “persistency is probably as good a measure as any of the quality of life assurance selling”. Similarly a report prepared by a leading life and health reinsurance company also made this link explicit. After citing the example of a Canadian company whose withdrawal rates had dropped by a quarter following a switch to paying level commission, Mercantile & General Reinsurance (1993) concluded that “the quality of life assurance sales is dependent on the quality of the people selling, the training they receive, the commission structure by which they are remunerated and the cultural environment in which they work” (p1). Finally a consumer survey conducted by the U.K. insurance regulator reported that the policy type with the highest withdrawal rate (mortgage endowments) also had the highest proportion of policyholders saying they regretted taking out the product and a high rate of complaints (Collard, 2001). Of course, poor persistency may be due to ‘churning’, where the salesperson persuades the policyholder to withdraw from an existing contract in order to switch insurers. In such cases, persistency problems could not necessarily be attributed to the poor sales quality of the original insurer. However the U.K. evidence suggests that most withdrawing policyholders do not switch to another life office. The Financial Services Authority (2000) reported that only 7% of withdrawals from life policies (and around 15% from personal pensions) were followed by a fund transfer to another company. However, it is also clear that the chain: poor sales quality → low customer satisfaction → low persistency may be too simplistic. In a survey of U.K. life insurance policyholders who had withdrawn from their policies, 80% said they were satisfied with the overall service provided by the company, and 88% reported that the main reason for withdrawing was a change in personal circumstances (Survey Research Associates, 1992). Similarly, in a more recent investigation by the U.K. insurance regulator, less than one-quarter of the respondents interviewed in depth were dissatisfied with the process of taking out and subsequently having to lapse a regular premium product (Financial Services Authority, 2000).

9

Mis-selling of life and pensions policies has not been restricted to the U.K. Keech (2002) refers to widespread mis-selling in Australia in the 1990s; while in the U.S., one company was recently fined $3bn for deceptive sales practices (Abromovitz & Abromovitz, 1998).

11

4.

Inter-Company Variations in Persistency

Hypotheses and Data Sources The main trust of this paper is to explore inter-company variations in persistency experience. Our first objective is to ascertain whether the undoubted differences in persistency among U.K. long-term insurers are random in nature (as opposed to having some systematic features). A failure to find systematic differences among insurers would suggest very strongly that persistency problems arise from random factors (such as macroeconomic fluctuations and changes in the policyholders’ personal circumstances) rather than from any dissatisfaction arising from poor service quality of insurers. Our second main objective is to determine the nature of any systematic differences in persistency according to company size, efficiency, and ownership structure. The essential idea is that persistency is an indicator of customer satisfaction, while company financial performance can, to a certain extent, proxy for service quality. Thus we attempt to measure the relationship between customer satisfaction /persistency and service quality (which is, in turn, affected by variables such as size, new business growth, the expense ratio, and the mutual/stock distinction). Inter-relationships between the various proxies for service quality necessitate a multivariate approach. Our analysis of inter-company differences uses the data obtained from a survey conducted by a division of the U.K. insurance regulator (Personal Investment Authority, 2001). Companies are only included in the survey if they sold over 1000 policies in total in the relevant year, and are only included in the individual persistency tables (by policy type and distribution channel) if they sold more than 500 policies in that category. We further restricted the sample by omitting a number of very small friendly societies that did not file ‘regulatory returns’ to the U.K. Treasury (as required by the Insurance Companies Act 1982), and this limited the sample to a maximum of 99 insurers10; none-the-less, the included companies constitute the major part of the U.K. life insurance sector. The analysis is restricted to

10

Although not all insurers provided persistency data for each year, policy type, or distribution channel.

12

Ordinary Branch regular premium endowment and personal pension policies, sold through either company representatives (CRs) or independent financial advisers (IFAs).

Table 2: Average Withdrawal Hazard Rate Correlations, by Product Type Persistency Term 1

2

3

4

Overall Average

Endowment/CR

0.165

0.225

0.205

0.178

0.193

0.525

0.427

0.421

0.403

0.444

0.345

0.326

0.313

0.291

0.319

Pensions/CR Endowment/IFA Pensions/IFA Average

Table 3: Average Withdrawal Hazard Rate Correlations, by Distribution Channel Persistency Term 1

2

3

4

Overall Average

Endowment/CR

-0.001

0.612

0.370

0.518

0.375

0.633

0.423

0.491

0.280

0.457

0.316

0.518

0.430

0.399

0.416

Endowment/IFA Pensions/CR Pensions/IFA Average

Table 4: Average Withdrawal Hazard Rate Correlations, by Persistency Term Year of Inception 1993

1994

1995

1996

Overall Average

Endowment/CR

0.634

0.434

0.463

0.451

0.496

Pensions/CR

0.562

0.088

0.228

0.139

0.254

13

Endowment/IFA

-0.070

0.405

0.506

0.675

0.379

Pensions/IFA

0.476

0.499

0.241

0.393

0.402

Average

0.401

0.357

0.360

0.415

0.383

Testing for Systematic Differences in Persistency If there are indeed inter-company differences, we expect to observe inter-correlations in withdrawal hazard rates across products (for companies reporting rates for more than one product type), by distribution channel (for companies reporting rates for more than one channel), and by persistency term or duration. The average correlation coefficients for are reported in Tables 2 to 4. Table 2 shows the average11 bivariate correlation coefficients between the withdrawal hazard rates for endowment and personal pensions policies for those insurers that provide persistency data for both product types. The table shows that the withdrawal hazard rates for different policy types tend to be correlated (with an overall average value of 0.319) and that this relationship seems fairly consistent across all persistency terms. However the relationship between the endowment and pensions hazard rates is marked stronger for insurers relying on the IFA distribution channel. Table 3 shows the average correlation coefficients for company representative and IFA sales channels for those insurers that provide persistency data for both channels: the overall picture is very similar to Table 2 (with an overall average correlation of 0.416). Table 4 shows the average inter-correlations between hazard rates of different persistency terms (for contracts incepted in the same calendar year). The figures show that the withdrawal hazard rates for persistency years 1 to 4 tend to be correlated (with an overall average coefficient of 0.383). The inevitable conclusion to be drawn from Tables 2, 3 and 4 is that withdrawal rates do not very randomly among U.K. life insurers; instead the differences are systematic. Thus insurers tend to have positively-correlated hazard rates for the two main product types for the same distribution channel (Table 2); the hazard rates for the two main distribution channels tend to

11

Averaged over the inception years 1993 to 1999 where possible.

14

be positively correlated for the same product type (Table 3); and the hazard rates at different durations also seem to be positively related (Table 4).

A Multivariate Analysis of Differences in Withdrawal Hazard Now that it has been established that there are indeed systematic differences among insurers in their withdrawal hazard rates, it remains to try and discover something about the source of such differences. We are particularly concerned to determine whether the variation in persistency can be explained by company size, efficiency, and ownership structure. The review of Section 2 identified a number of factors that might cause an insurer’s withdrawal rates to lie above the industry average. In particular, persistency is likely to be worse in those life offices that focus on low-income, or younger customers, have a high proportion of index- or unit-linked business, or have a less experienced salesforce12. Unfortunately many of these factors are either not observable or cannot be measured with any degree of accuracy at the corporate level. We therefore focus our attention on whether inter-company variations in withdrawal rates are due to financial and ownership characteristics (summary statistics for Ordinary Branch business13, by year of inception, are provided in Table 5) as follows: Expense Ratio: The impact of an insurer’s expense ratio on persistency may differ depending on the persistency duration. In the early stages of the policy, an insurer that expends resources on servicing customers may gain the benefit of lower initial withdrawal rates. However, the opposite may occur as the duration of the policy increases as policyholders observe that high expenses produce a lower accumulated fund. The regression variable is lnEXPt being the natural logarithm of total long-term operating expenses as a % of total long-term net earned premium income. 12

Of course, the fact that their customers may be younger with low incomes does not necessarily absolve insurers from responsibility for the ensuing poor persistency, particularly if the higher withdrawal rates arise from a failure to meet these customers’ service quality expectations. 13 Data on financial performance is obtained from the supervisory returns submitted to the U.K. Treasury, and therefore relates primarily to business transacted in the U.K.. Company groups with more than one ‘Treasury Return’ have been aggregated to obtain a U.K. group total, taking account of changes in ownership resulting from mergers and acquisitions over the period 1994-1996.

15

Index-Linking: Index- or unit-linked contracts are generally designed to be more flexible and transparent than their with-profit counterparts, and usually have lower withdrawal penalties. Thus withdrawal from such contracts is more likely to be observed. The regression variable is INDEXt being index- and other linked new business regular premiums as % of total new business regular premiums. Mutuality: Mutual life insurance companies are owned by a subset of their customers, whereas stock insurers are owned by shareholders (or are subsidiaries of stock companies). A number of different arguments have been advanced about the behaviour of mutuals and stocks that may have a bearing on their persistency performance. The so-called ‘expense preference’ hypothesis suggests that customer satisfaction (and therefore persistency) may be lower for mutual insurers as a result of the self-serving behaviour of company management14. On the other hand, the ‘efficient sorting’ hypothesis implies that the customers of mutual insurers may be more satisfied (than those of stocks) because they have less need to be concerned about conflicts of interest between customers and (profiteering) owners. The regression variable is MUTUALt = 1 (mutual or owned by mutual parent company) or = 0 (stock or owned by mutual stock company). New Business: A high rate of new business growth is likely to be associated with high sales pressure and poor sales quality, as well as higher new business expenses. The regression variable is lnNBt = natural logarithm of (new business regular premiums + 0.1single premiums) as a % of total long-term net earned premium income. Size: The impact of company size on withdrawal rates is ambiguous. On the one hand, larger companies should be able to gain economies of scale in the delivering service quality15. On the other hand, large insurers will have a more diverse customer base, and may have a higher proportion of customers with a greater propensity for poor persistence. The regression variable is SIZEt = natural logarithm of total admissible assets. 14

Cummins and Zi (1998, p145) say that “The expense preference hypothesis predicts that mutuals will have higher costs than stocks because the mutual form of ownership affords owners less effective mechanisms for controlling and disciplining managers than the stock ownership form”. Their empirical results are not consistent with the predictions of the hypothesis. 15 Frei et al (1999) discovered that larger banks were associated with better service quality.

16

Table 5 about here Two aspects of the withdrawal hazard are analysed: the first-year hazard rate h(1), and the withdrawal rate for years two to four h(2_4), which represents the probability of withdrawal in years 2, 3 or 4 for those policies that were in force at the start of the second policy year16. The withdrawal rates are analysed separately for endowment and personal pensions contracts sold via the company representative and IFA distribution channels (however not all 99 companies in the sample provided data for the two policy types and distribution channels). Descriptive statistics of the withdrawal rates are also provided in Table 5. Multivariate Tobit regressions are conducted on withdrawal rates for policies incepted in the years 1994, 1995 and 1996, with explanatory variables which pick-up company-specific performance (in the year of inception) plus dummy variables for the years 1994 and 199517; the results are reported in Table 6. Table 6 about here Table 6 provides confirmation of the existence of systematic differences between insurers in their first-year h(1) and years 2-4 hazard rates h(2_4): with the exception of personal pensions sold via independent financial advisers (PI), the models show adjusted R2 figures which are generally significant at the 1% level at least. The results confirm the following features of about inter-company differences in persistency: 1. Mutuality appears to have a significantly negative impact on withdrawal hazard rates in the case of endowments sold by company representatives (first-year), and on all years 2-4 rates (except for PI). No model shows a significantly positive coefficient for the Mutual dummy variable. Thus it would appear that mutuality status is good for persistency, and that mutual policyholders are less likely to withdraw from their contracts than their stock policyholder counterparts (particularly in years 2 to 4).

16

The first-year withdrawal hazard rate is h(1) = 1-S(1) where S(1) is the proportion of policies that persisted to the end of the first policy year. The withdrawal rate for years 2-4, for those policies that were in force at the start of year 2, is given by h(2_4) = (S(1)-S(4))/S(1) where S(4) represents the proportion of incepted policies which persisted to the end of the fourth policy year. 17 A Tobit specification is utilised in recognition that h ~ [0,1] .The panel nature of the data might suggest the use of a fixed (FEM) or random (REM) effects model. However the inclusion of a largely time-invariant mutual dummy variable rules out FEM, and the REM frequently failed to converge because of the relatively small sample size.

17

These finding support the ‘efficient sorting’ hypothesis as opposed to the ‘expense preference’ hypothesis of mutual insurance company behaviour (Cummins and Zi, 1998). 2. The expense ratio (lnEXP) has a consistently positive coefficient that is significant in five out of the eight cases (and no model shows a negative coefficient for lnEXP). Thus offices with higher expense ratios tend to demonstrate significantly higher withdrawal rates in the early years of the policy term. One possible explanation is that offices which incur high acquisition costs suffer higher subsequent withdrawal rates as a result. 3. The growth in new business variable (lnNB) has a significantly positive impact on the first-year withdrawal rates of endowment policies sold by company representatives, but otherwise does not seem to impact on persistency. 4. Insurance group size (in terms of the logarithm of total long-term assets) has a significantly positive impact on withdrawal rates in four out of the eight cases, and only one model shows a negative (but insignificant) SIZE coefficient. There is therefore some evidence to suggest that large life insurers suffer lower persistency figures, either because they provide a poorer level of service quality or because they target customers who, by their personal circumstances, have service expectations which the insurers cannot meet. 5. Finally although most models show some degree of overall significant, it is fair to say that the overall success of the models (in terms of explaining the variation in withdrawal rates) is still rather low. There is therefore quite a lot about inter-company variations in persistency that still remains unexplained.

5.

Conclusions

There has been considerable disquiet in the U.K. at what are regarded as low rates of the persistency of long-term insurance business. Many policyholders have withdrawn from longterm commitments before their contract has expired, and the high initial withdrawal rates 18

associated with these long-term savings contracts provides tangible evidence of widespread customer dissatisfaction, and raises questions about the service quality UK life insurance. Figures provided by the U.K. financial services regulator show that, for regular premium personal pensions policies incepted in the U.K. in 1996, 42.8% of contracts sold by company representatives and 37.7% of those sold by IFAs experienced withdrawal in the first 4 years. The corresponding withdrawal rates for regular premium endowment contracts were 23.2% and 18.8% respectively. Although there may be reasons why some policies which lapse cannot be foreseen either by the individual investor or their adviser at the time of sale, the high withdrawal rates are thought be many commentators to be indicative of poor service quality – particularly in the sales process. The main trust of this paper is to explore inter-company variations in persistency experience. Our first objective was to ascertain whether the undoubted differences in persistency among U.K. long-term insurers are random in nature (as opposed to having some systematic features). The analysis in Section 4 indicates that insurers tend to have positively-correlated hazard rates for the two main product types for the same distribution channel, the hazard rates for the two main distribution channels tend to be positively correlated for the same product type, and the hazard rates at different durations also seem to be positively related. These systematic differences among insurers suggest very strongly that persistency problems do not arise from random factors, but result instead from an inability of insurers to meet the service quality expectations of a wide range of customers. Our second main objective was to determine the nature of any systematic differences in persistency according to company size, efficiency, and ownership structure. We therefore use a Tobit regression model to measure the relationship between withdrawal rates and those aspects of service quality that are correlated with variables such as size, new business growth, the expense ratio, and the mutual/stock distinction. The multivariate regression findings suggest that, after allowing for size and growth, mutual insurers have a better persistency record than stock insurers; thus mutual insurers seem more able than stock companies to meet the service quality expectations of their respective customers. Further impetus to the argument that insurers may be failing to meet quality expectations is provided by the finding that offices with higher expense ratios tend to demonstrate significantly higher 19

withdrawal rates. We are unable to explain why persistency seems to be negatively related to insurer size – but this may be because large offices are unable to meet the diversity of service expectations resulting from their wide range of customers. Finally, although the analysis has provided an explanation of some of the inter-company differences in withdrawal rates, there is still a lot that remains unexplained. Other authors have noted that the causes of poor persistency tend to be complex and pervade many aspects of an insurer’s back-office operations and not just the sales process. The financial services regulator will continue to experience difficulty in analysing the sources of these retention problems without improving the nature of the persistency data that it collects. In particular, we believe that further insights into inter-company variations in persistency can only be achieved if the following improvements to the publicly available persistency data are implemented: •

Survival or retention rates should be collected for longer persistency periods than the current four years. This data should be back-dated wherever possible.



The cut-off sales level18 for returning persistency data (by contract type and distribution channel) should be reduced in order to increase the number of included companies.



The persistency figures should be split between with-profit and index- or unit-linked contracts



Representative information on sales performance (which matches up with the reported persistency data) should also be provided. This should include the number, average premium income, average sums insured and average maturity term of policies sold in the inception year.



The status of persistency data from companies that are involved in mergers should also be clarified. At the moment, it is unclear why insurance groups continue to report separate retention data from companies that have merged.

18

Currently of 500 contracts per annum

20

REFERENCES Abromovitz, H. and Abromovitz, L. (1998) Insuring Quality, Boca Raton, Florida: St. Lucie Press. Association of British Insurers (2000) Raising Standards Manual, London Buck, N.F. (1961) First Year Lapse and Default Rates, Transactions of the Society of Actuaries, 12, 258-293. Collard, S. (2001) Consumers in the Financial Market: Financial Services Consumer Panel Annual Survey of Consumers 2000 , Financial Services Consumer Panel, London Crombie, J.G.R., Forman, K.G., Gibbens, P.R., Mason, D.C., Paterson, M.D., Shaw, P.C., Smart, J.M.G., Smith, H., Thomson, C.G. and Thomson, R.G. (1978) An Investigation into the Withdrawal Experience of Ordinary Life Business, Transactions of the Faculty of Actuaries, 36, 263-295. Cronin, J.J. and Taylor, S.A. (1992) Measuring Service Quality: A Re-Examination and Extension, Journal of Marketing, 56, 55-68 Crosby, L.A. and Stephens, N. (1987) Effects of Relationship Marketing on Satisfaction, Retention, and Prices in the Life Insurance Industry, Journal of Marketing Research, XXIV, 404-411. Cummins, J.D., and Zi, H. (1998) Comparison of Frontier Efficiency Methods: An Application to the U.S. Life Insurance Industry, Journal of Productivity Analysis, 10, 131-152 Diacon, S.R. and Ennew, C.T. (2001) Consumer Perceptions of Financial Risk, Geneva Papers on Risk and Insurance, 26, 3, 389-409 Diacon, S.R. and Watkins, T. (1995) ‘Insurance Marketing’, in Ennew, Watkins and Wright (1995), ch 11 Ennew, C.T., Watkins, T. and Wright, M. (1995) Marketing Financial Services, Butterworth Heinemann, Oxford, 2nd ed. Financial Services Authority (2000) Persisting: Why Consumers Stop Paying into Policies, Financial Services Authority, London. Financial Services Authority (2001) Treating Customers Fairly after the Point of Sale, Financial Services Authority, London. Forty, J. and Ferguson, J. (1999) Why are Persistency Improvements so Elusive? Emphasis on Management in Financial Services, 6, 16-18, Tillinghast Towers Perrin. Frei, F.X., Kalalota, R., Leone, A.J. and Marx, L.M. (1999) Process Variation as a Determinant of Bank Performance: Evidence from the Retail Banking Study, Management Science, 45, 9, 1210-1220 Gower, L.C.B. (1984) Review of Investor Protection, Report: Part I, Cmnd 9125, HMSO, London). Jenkins, J.A., Bolton, M., Green, M.D., Heron, S.P., Hylands, J. & McKee (1999) Report of the [Faculty and Institute of Actuaries] Endowment Mortgage Working Party, Institute of Actuaries, London

21

Keech, C. (2002) The Evolution of a Multi-tied Marketplace: the Australian Experience, Emphasis on Management in Financial Services, 9, 4-9. KPMG Peat Marwick (1993) Pension Transfers, Report to SIB, Securities and Investments Board, London. Krishnan, M., Ramaswamy, V., Meyer, M. and Damien, P. (1999) Customer Satisfaction for Financial Services: The Role of Products, Services and Information, Management Science, 45, 9, 1194-1209 Laslett, R., Wilsdon, T. and Malcolm, K. (2002) Polarisation: Research into the Effect of Commission-based Remuneration on Advice, Financial Services Authority, London. Lassar, W.M., Manolis, C. and Winsor, R.D. (2000) Service Quality Perspectives and Satisfaction in Private Banking, Journal of Service Marketing, 14, 2/3, 244-272 Lewis, B.R. (1995) ‘Customer Care and Service Quality’, in Ennew, Watkins and Wright (1995), ch 9 Mehta, S. (1992) Allowing for Asset, Liability and Business Risk in the Valuation of a Life Office, Journal of the Institute of Actuaries, 119, III, 385-440. Mercantile & General Reinsurance (1993) Business Retention – Solving a Persistent Problem¸ Mercantile & General, London McCabe, D. (2000) The Swings and Roundabouts of Innovating for Quality in UK Financial Services, The Service Industries Journal, 20, 4, 1-20 McCabe, D., Knights, D., and Wilkinson, A. (1997) Financial Services – Every Which Way But Quality?, Journal of General Management, 22, 3, 53-73 Moore, J.F. & Santomero, A.M. (1999) ‘The Industry Speaks: Results of the WFIC Insurance Survey’, in Cummins, J.D. and Santomero, A.M., eds (1999) Changes in the Life Insurance Industry: Efficiency, Technology, and Risk Management , Kluwer, Norwell. Office of Fair Trading (1994) Surrender Values of Life Insurance Policies, Office of Fair Trading, London Parasuraman, A., Zeithaml, V.A. and Berry, L.L. (1985) A Conceptual Model of Service Quality and its Implications for Future Research, Journal of Marketing, 49, 41-50 Patrick, F.D. and Scobbie, A. (1969) Some Aspects of Withdrawals in Ordinary Life Business, Transactions of the Faculty of Actuaries, 31, 3, 53-93. Personal Investment Authority (2001) Seventh Survey of the Persistency of Life and Pensions Policies, Personal Investment Authority, London. Proudfoot, W. (1969) Comment, pages 99-102 in: Discussion on Crombie, J.G.R., Forman, K.G., Gibbens, P.R., Mason, D.C., Paterson, M.D., Shaw, P.C., Smart, J.M.G., Smith, H., Thomson, C.G. and Thomson, R.G. (1969) An Investigation into the Withdrawal Experience of Ordinary Life Business, Transactions of the Faculty of Actuaries, 36, 263-295 Pullan, C. (1995) Collecting Persistency Rates, Improving Business Persistency, LIMRA International (conference).

22

Renshaw, A.E. & Haberman, S. (1986) Statistical Analysis of Life Assurance Lapses, Journal of the Institute of Actuaries, 113, III, 459-498. Richardson, C.F.B. (1961) Comment, pages 296-302 in Discussion on Buck, N.F. (1969) First Year Lapse and Default Rates, Transactions of the Society of Actuaries, 12, 258-293. Richardson, C.F.B. & Hartwell, J.M. (1951) Lapse Rates, Transactions of the Society of Actuaries, 3, 338-374. Roth, A.V. and Jackson, W.E, (1995) Strategic Determinants of Service Quality and Performance: Evidence from the Banking Industry, Management Science, 41, 11, 1720-1733 Rust, R.T., Danaher, P.J. and Varki, S. (2000) Using Service Quality Data for Competitive Marketing Decisions, International Journal of Service Industry Management, 11, 5, 438-469 Schlesinger, H. & Graf von der Schulenberg, J.-M. (1993) Consumer Information and Decisions to Switch Insurers, Journal of Risk and Insurance, 60, 591-615. Securities and Investments Board (1991) Life Assurance – Policy Termination Rates, Securities and Investments Board, London. Society of Actuaries Direct Response Persistency and Mortality Task Force (1995) Final Report, Society of Actuaries, Schaumburg. Survey Research Associates (1992) A Presentation of the Main Findings of Persistency Research, Survey Research Associates, London. Treagus, R. (1995) Persistency: Why, What and How, Improving Business Persistency, LIMRA International, (conference). Webster, A.C. (1969) Comment, pages 106-108 in: Discussion on Crombie, J.G.R., Forman, K.G., Gibbens, P.R., Mason, D.C., Paterson, M.D., Shaw, P.C., Smart, J.M.G., Smith, H., Thomson, C.G. and Thomson, R.G. (1969) An Investigation into the Withdrawal Experience of Ordinary Life Business, Transactions of the Faculty of Actuaries, 36, 263-295. Wells, B.P. and Stafford, M.R. (1995) Service Quality in the Insurance Industry, Journal of Insurance Regulation, 13, 4, 462-477

23

Chart 1: Withdrawal Hazard Rates Endowment / CR 0.09 0.085 0.08

Rate

0.075 0.07 EC_1993 EC_1994 EC_1995 EC_1996 EC_1997 EC_1998 EC_1999

0.065 0.06 0.055 0.05 1

2

3 Term

24

4

Chart 2: Withdrawal Hazard Rates Pensions / CR 0.15

0.14

PC_1993 PC_1994 PC_1995 PC_1996 PC_1997 PC_1998 PC_1999

Rate

0.13

0.12

0.11

0.1

0.09 1

2

3 Term

25

4

Chart 3: Withdrawal Hazard Rates Endowment / IFA 0.06

0.055

EI_1993 EI_1994 EI_1995 EI_1996 EI_1997 EI_1998 EI_1999

Rate

0.05

0.045

0.04

0.035

0.03 1

2

3 Term

26

4

Chart 4: Withdrawal Hazard Rate Pensions / IFA 0.15 0.14 0.13

PI_1993 PI_1994 PI_1995 PI_1996 PI_1997 PI_1998 PI_1999

Rate

0.12 0.11 0.1 0.09 0.08 1

2

3 Term

27

4

Table 5 Descriptive Statistics of Dependent and Explanatory Variables U.K. Life Insurance Companies, 1994 to 1996 N

Minimum

Maximum

Mean

Std. Deviation

1994 h(1) EC h(1) PC h(1) EI h(1) PI h(2_4) EC h(2_4) PC h(2_4) EI h(2_4) PI Expense Ratio % Index-linked % Mutual=1 New Business % ln(Assets £000)

51 46 17 27 44 40 17 25 72 72 72 72 72

0.009 0.014 0.000 0.037 0.028 0.107 0.017 0.106 6.929 0.000 0.000 1.208 6.777

0.241 0.267 0.085 0.153 0.282 0.379 0.156 0.333 639.336 100.000 1.000 71.872 17.550

0.083 0.142 0.051 0.096 0.152 0.297 0.122 0.257 35.065 57.364 0.500 15.590 14.517

0.050 0.050 0.021 0.030 0.058 0.053 0.035 0.060 76.207 33.000 0.504 10.676 1.904

1995 h(1) EC h(1) PC h(1) EI h(1) PI h(2_4) EC h(2_4) PC h(2_4) EI h(2_4) PI Expense Ratio % Index-linked % Mutual=1 New Business % ln(Assets £000)

52 45 18 29 43 38 18 26 72 72 72 72 72

0.002 0.005 0.023 0.002 0.017 0.112 0.039 0.057 8.897 0.000 0.000 0.019 6.593

0.167 0.235 0.096 0.197 0.254 0.393 0.193 0.480 207.746 100.000 1.000 126.549 17.613

0.078 0.135 0.055 0.093 0.161 0.299 0.132 0.290 26.328 58.803 0.500 14.294 14.700

0.039 0.050 0.022 0.041 0.054 0.070 0.047 0.084 25.586 32.730 0.504 17.175 1.875

1996 h(1) EC h(1) PC h(1) EI h(1) PI h(2_4) EC h(2_4) PC h(2_4) EI h(2_4) PI Expense Ratio% Index-linked % Mutual=1 New Business % ln(Assets £000)

42 41 17 27 38 36 15 23 69 69 69 69 69

0.002 0.000 0.021 0.005 0.036 0.067 0.047 0.074 8.434 0.000 0.000 0.633 6.571

0.174 0.218 0.075 0.183 0.253 0.400 0.206 0.426 99.539 100.000 1.000 136.221 17.738

0.068 0.123 0.047 0.098 0.170 0.296 0.130 0.295 22.099 52.345 0.478 13.790 15.006

0.039 0.046 0.019 0.042 0.059 0.071 0.054 0.085 14.672 30.709 0.503 15.461 1.878

EC=Endowment/Company Representative; PC=Pensions/ Company Representative EI=Endowment/IFA; PI=Pensions/IFA

Table 6: Tobit Regression Results- Withdrawal Hazard Rates,

28

U.K. Life Insurance Companies, 1994 - 1996 First-Year Hazard Rates h(1) EC

Constant INDEX lnEXP lnNB MUTUAL SIZE 1994 Dummy 1995 Dummy Sigma Adjusted R2 N

h(1) PC

h(1) EI

h(1) PI

Coeff

Prob

Coeff

Prob

Coeff

Prob

Coeff

Prob

-0.10169 -0.00017 0.01497 0.01115 0.00203 0.00708 0.01518 0.01369

0.0302 0.1059 0.0187 0.0070 0.7802 0.0005 0.0671 0.0980

0.09375 0.00006 0.02426 -0.00807 -0.03159 -0.00078 0.01432 0.00741

0.2311 0.7029 0.0255 0.4064 0.0003 0.8105 0.1536 0.4590

-0.12714 -0.00004 0.02525 -0.01047 -0.00234 0.00819 0.00550 0.00662

0.0143 0.6547 0.0054 0.1107 0.6519 0.0000 0.3301 0.2379

0.07629 -0.00023 0.01433 -0.00954 0.00183 0.00109 -0.00129 -0.00503

0.4338 0.1841 0.3346 0.3135 0.8314 0.8263 0.8993 0.6184

0.03949 0.0000 0.10440 0.00223 145

0.04508 0.0000 0.10426 0.00397 132

0.01602 0.0000 0.29684 0.00159 52

0.03612 0 -0.01286 0.54892 83

Hazard Rates for Years 2 to 4 h(2_4) EC

Constant INDEX lnEXP lnNB MUTUAL SIZE 1994 Dummy 1995 Dummy Sigma Adjusted R2 N

h(2_4) PC

h(2_4) EI

h(2_4) PI

Coeff

Prob

Coeff

Prob

Coeff

Prob

Coeff

Prob

-0.01380 -0.00016 0.02449 0.00636 -0.02735 0.00766 -0.01686 -0.00533

0.8282 0.2883 0.0042 0.2444 0.0083 0.0056 0.1430 0.6452

0.10945 0.00025 0.02700 0.02109 -0.03600 0.00389 -0.01124 -0.00344

0.3782 0.2437 0.0775 0.2243 0.0032 0.4312 0.4287 0.8087

-0.13461 0.00000 0.00954 -0.00325 -0.01908 0.01645 -0.00770 -0.00113

0.2372 0.9807 0.6305 0.8193 0.0913 0 0.541 0.9279

-0.01798 0.00093 0.03433 -0.01374 0.00330 0.01253 -0.04723 -0.01950

0.9425 0.0092 0.3031 0.6713 0.8543 0.2427 0.0308 0.3642

0.05141 0.0000 0.12379 0.0019 125

0.05939 0 0.08575 0.01974 114

0.03413 0 0.31380 0.00137 50

0.07188 0 0.05075 0.16379 74

Notes •

EC=Endowment/Company Representative; PC=Pensions/ Company Representative; EI=Endowment/IFA; PI=Pensions/IFA



h(1) is the first-year hazard rate h(1). h(2_4) is the withdrawal rate for years two to four (ie the probability of withdrawal in years 2, 3 or 4 for those policies that were in force at the start of the second policy year)

29