PREM CHANDRA SAXENA - DHIRENDRA KUMAR

DIFFERENTIAL RISK OF MORTALITY AMONG PENSIONERS AFTER RETIREMENT IN THE STATE OF MAHARASHTRA, INDIA^*

1. INTRODUCTION

   Retirement constitutes a major problem for many older people because their occupational role plays such a central part in defining their total social position, and the loss of this role therefore creates a traumatic shift in their personal situation (Tibbitts, 1966). Work is often the most important dimension integrating an individual into society. Identity, lifestyle and general life participation patterns are all conditioned by work. The process of retirement therefore undercuts the major social supports of an individual - by removing him or her from the world of work in which these supports are rooted. There has been much concern about retirement practice - particularly if retirement is compulsory.

   Mandatory regulation of retirement is a key issue facing many developing countries because of their increasing levels of general unemployment. Rising unemployment increases societal pressure to reduce the volume of older skilled manpower from the labour force - whether or not they are physically, mentally or economically ready for retirement. Sudden and abrupt disengagement from active economic and social life may contribute to numerous socio-psychological problems among retirees (Chown, 1977). These problems may affect their personal and social life as they shift to a non-productive role with less income, possibly more social isolation, and perhaps with reduced status within the family. Such role loss and attendent feelings of worthlessness may have deleterious effects on the health of retirees in terms of tension, stress and worry (Tibbitts, 1954; Havighurst, 1963; Havighurst et al., 1968). In societies where the process of withdrawal from the labour force is gradual, these adverse socio-psychological effects may not be very great. But in cases of abrupt withdrawal from the labour force, the effects may be much more pronounced and experienced by most individuals.

   The elderly people of India constitutes a heterogeneous category, and their problems vary - for example according to residential location and socio-cultural characteristics. Social scientists have attempted to study these problems from different perspectives. Thus demographers view aging and its consequences as a by-product of the demographic transition from high fertility and mortality to low fertility and mortality due to the process of modernization. Biologists consider aging apropos the processes that lead to a gradual reduction in the capacity to adapt to stress and a gradual increase in the susceptibility to disease. Psychologists focus on the behavioral, attitudinal, emotional and other characteristics of the elderly. And sociologists deal with the gradual unfolding of the social development of individuals - including changes in their social status, roles and positions.

   Of late, researchers in western countries have shown much interest in the age-patterns of mortality of people after their withdrawal from the paid labour force (Hayward et al., 1988, and 1990). The retirement in both developed and developing countries is sometimes voluntary. The reasons for voluntary retirement are many - but they include ill health.

   The process of retirement in the formal employment sector in India, and most developing countries, is such that the age of retirement is fixed, and normally a person cannot remain in service after a given age. Several studies of the problems of the aged in general - and of pensioners in particular - have been undertaken in India. Their main thrust has been to analyze the socio-psychological problems following retirement. Unfortunately, however, little is known about the age pattern of mortality among pensioners who retire after reaching the age of superannuation i.e., normal retirement age. Accordingly, the present study attempts to estimate the relative risks of mortality, by occupation, of those who retired from the Maharashtra State Government Service after superannuation during the period 1981-90.

2. A NON-CONVENTIONAL INSTITUTIONAL DATA SOURCE

   The major problem facing researchers who wish to study the risks of mortality among pensioners is the lack of information on duration of survival after retirement. It is obviously the case that to construct life tables for retired persons and estimate risks of mortality by age, a large body of data on the mortality of pensioners is required. In principle the collection of such data is possible through large scale follow-up studies of cohorts of pensioners. But in practice such an approach would be both time consuming and expensive.

   Recently Saxena (1991) has uncovered a non-conventional data source which can provide reliable information with which to study mortality patterns and differentials among pensioners according to various socio-economic characteristics. The Government Pay and Accounts Office/Treasury Office of each district in all states of India maintains pension records after retirement (or any other sort of withdrawal from government service leading to the monthly disbursement of pensions). It should be noted that the payment of pensions to retirees or to their bonfide beneficiaries (wife/husband or non- adult children in the case of family pensions) is made only upon the receipt of a Pension Payment Order (PPO) issued by the Office of the Accountant General in each state¹.

   The PPO gives particulars about pensioners - such as their age, sex, type of pension, length of service prior to retirement, the amount of pension fixed at the time of retirement, the date of death (if applicable) and their religion. This paper utilizes information from PPOs in order to study the mortality risks and differentials by occupation among pensioners who were in Maharashtra Government Service prior to retirement.

   The relevant information was compiled from PPOs available at the Pay and Accounts Office in Bombay. The data compilation was completed in two stages between March 1990 and January 1991. In the first stage, a separate list of pensioners was prepared from the entries of the pension papers received during the period between April 1, 1981 and March 3 1, 1990, and contained in the registers maintained by the Pay and Accounts Office. Since information on pensioners who retired before 1981 was not readily available, we have confined the present study to the period 1981-90. In the second stage, the particulars of pensioners were compiled from the available PPOs as stipulated in Table 1.

   The total number of pensioners enrolled at the Pay and Accounts Office in Bombay during the period 1981-90 was 18,201. Out of these 12,685 (69.6%) were retired after superannuation (i.e. the M category pension) followed by 2,970 (16.3%) who died while in service and a family pension was granted (i.e. the MF category pension). The MF/ADHOC category of pension accounted for 2,546 cases (14.0%) and applied when a government employee died while in service, but before October 1, 1977 - the date when the family pension scheme was introduced. Table 2 gives the distribution of pensioners who retired from the Maharashtra State Government Service during the period 1981-90 by year of registration in the Pay and Accounts Office (Bombay) and by category of pension. However, the present study is confined to the analysis of data for those pensioners who retired after reaching the normal age of retirement (i.e. the M category pension).

   TABLE 1

   Particulars of pensioners compiled from pension payment orders (PPOs)

   Age	The age of the pensioner at the reference date of the survey, i.e. March 31, 1990
   Sex	Male = 1; Female = 2
   Type of pension
   M	Pension after superannuation (i.e. normal retirement age)
   MF	Family pension paid in case of the death of a person while in service
   MF/ADHOC	Family pension paid in case of the death of a person while in service, but before introducing the Family Pension Act of 1964, i.e. October 1, 1977
   Occupation	Type of service prior to retirement
   Income	The amount of pension per month at the time of retirement (dearness allowance is not included in the pension)
   Retirement age	58 years at superannuation from government service
   Date of death	In the event of death of the pensioner
   Religion	Hindu=1; Muslim=2; Christian=3; Other=4

   2.1 Occupational categories

   Researchers have identified two major dimensions underlying occupational differences in mortality - namely exposure and life-style. By 'exposure' is meant the duration of service in a particular occupation; and 'life-style' is defined as a by-product of social status which includes health behavior, access to health care facilities, attitudes towards health care, and administrative power to take decisions (Benjamin 1965; Fox and Adelstein 1978; Moore and Hayward, 1990). Typically researchers view occupation as the most accurate single indicator of social status and life-style. For retired people, occupation often has more social importance - since they may try to derive present social status from their pre-retirement occupation (Clark, 1972). Therefore, in order to study occupational differences in mortality after the age of retirement (i.e. superannuation) we have classified the M category of pensioners into three broad groups according to the degree of similarity of occupation expressed in terms of life-style and experience before retirement. Group I consists of those pensioners who were in the Maharashtra State Government Police Service before retirement; Group II consists of those categories of occupations which involved administration and/or direct public dealing; and Group III consists of all those occupations of Government Service which had no public dealing at all. The two occupational categories of Groups I and II may seem alike in terms of their administration and public dealing activities, but a major difference lies in their life styles. Prior to their retirement, police personnel maintain a mandatory exercise regime which keeps them physically fit. However, this regime may well be relaxed after retirement. In contrast , this physical fitness factor is entirely absent in the case of pensioners in Group II.

   TABLE 2

   Pensioners retired during the period 1981-90 by year of registartion and category of pension

   Year of registration	Category of pension
   	M	MF	MF/ ADHOC	Total
   1981-82	799	191	61	1051
   1981-82	1078	260	215	1553
   1983-84	1354	274	183	1811
   1984-85	1244	323	304	1871
   1985-86	1566	344	516	2426
   1986-87	1618	402	398	2418
   1987-88	1867	430	383	2680
   1988-89	1626	413	311	2350
   1989-90	1533	333	175	2041
   Total	12685	2970	2546	18201
   	(69.7)	(16.3)	(14.0)	(100.0)

   The distribution of deaths of pensioners (in the M category) by their occupational categories prior to retirement revealed that out of a total of 1473 deaths recorded during the period 1981-90, 731(49.6%), 356(24.2%) and 386(26.2%) respectively were from Groups I, II and III. For the present analysis the duration of survival for those who died was computed by category of occupation prior to retirement.

3. METHODOLOGY

   In order to estimate the probability of death of pensioners in the years following retirement and determine mortality differentials by occupation in terms of relative risk, synthetic life-tables incorporating censored information have been constructed and a hazards analysis has been carried out. In recent years, the methodology of hazard analysis has been increasingly applied to marital dissolution (Menken et al., 1981), mortality (Trussell and Hammerslough, 1983), and birth intervals (Rodriguez et al., 1984). The concept of a proportional hazards model was first proposed by Cox (1972) and has been developed by Kalbfleisch & Prentice (1973) and Holford (1976). For details the reader is referred to Namboodiri and Suchindran (1987). In this paper the survival functions and hazard rates for retirees have been estimated incorporating censored information.

   3.1 Construction of life tables incorporating censored information

   The following assumptions have been made in order to constructsynthetic life tables for pensioners who retired during the period of 1981-90.

   Assumptions:

   1. The population is homogeneous; i.e., the risk of mortality for pensioners is constant for all individuals under study during the time period (0,T).
   2. The failure and censoring mechanisms are independent, i.e., the experience of a number of persons with differing entry dates has been consolidated so as to yield a representation of the experience of a synthetic cohort with a common entry date.

   Description of life table columns:

   Column (1)	Time interval used for grouping the observations, all intervals being of length one year.
   Column (2)	ni - Number of pensioners as of the beginning of the time interval and ni+1 = ni - ci, i = 0,1,2,...,8, where ci is defined as below.
   Column (3)	di - Number of deaths occurring during the time interval.
   Column (4)	ci - Number of pensioners who did not die during the study period and who were alive at the end of the observational period (i.e., the reference date of study March 31, 1990), i.e. the number of censored cases during that time interval.
   Column (5)	ni' - The number of individuals at risk of mortality at the beginning of the time interval. ni = ni - ci/2, i = 0,1,...,8.
   Column (6)	qi - Estimated probability of death during the interval, given that the pensioner was alive at the beginning of the interval. qi = di/ni = 1,2,...,8.
   Column (7)	pi - Probability of survival of pensioners during the interval. pi = 1-qi, i = 0,1,2,...,8.

   3.2 Relative risk of mortality

   In order to study mortality differentials by occupation, the ratio of hazard rates for each category of pensioners considered by single years after retirement has been calculated, and is denoted here as R_jk(t).

   R_jk(t) is defined as the relative risk of mortality of pensioners of the j-th occupational group with respect to pensioners of the k-th occupational group in a given time interval. Thus, R_jk(t) is defined as follows:

   Rjk(t)=	Hazard rate h(t) for pensioners of j-th occupational group during the given interval of time
   	Hazard rate h(t) for pensioners of k-th occupational group during the same interval of time

   If R_jk(t) >1, it implies that the risk of mortality among pensioners of the j-th occupational group is higher than that of the k-th occupational group for the given time interval.

   The basic data needed to construct the synthetic life tables are:

   1. Date of entry i.e., the date of retirement of the pensioners from service.
   2. Date of failure i.e., the date of death of pensioners.
   3. Date of exit from the study i.e., the study's reference date.

   Theoretically, the date of entry marks the start of exposure of the individual under study to the risk of mortality after retirement. The date of exit denotes the date of termination of the observational period. In the present study pensioners were observed up to the date of exit, i.e. until the reference date (March 31, 1990). Hence, if an individual did not die before the date of exit (i.e., the reference date) his or her record has been treated as censored.

4. RESULTS

   4.1 Mortality risk among pensioners after retirement

   The estimates of the survival, probability density, and hazard functions for pensioners retired following superannuation during the period 1981-90 are given in Tables 3a and 3b. Column 1 of Table 3a shows the single year interval of life span of pensioners after superannuation. Column 2 gives the total number of pensioners by single year of retirement at the beginning of the study and who were still under observation at the reference date (these pensioners were therefore exposed to the risk of mortality during the observational period). Column 3 gives the number of deaths of pensioners in the corresponding intervals, and column 4 shows the number of pensioners who did not die during the interval and who were alive at the end of the period of observation (these last constituted the censored cases). As is evident from the table, 12,685 pensioners were alive at the beginning of the study period and were exposed to the risk of mortality. Out of this total of 12,685 pensioners, 276 died before completing their first year of retirement and 1,677 pensioners were censored in the first year of exposure.

   Further, only 10,732 retirees were exposed to the risk of mortality during the interval [1,2); 257 died between the period of I and 2 years after retirement; and 1, 13 2 pensioners had only two years of exposure and did not die before the end of the study period (and therefore their information too is treated as censored). A similar interpretation can be made for the other intervals.

   Column 6 of table 3a gives the conditional probabilities of dying in the corresponding intervals. This column reveals the pattern of mortality of pensioners by single years of life after retirement. Columns 2, 3 and 4 of Table 3b give respectively the estimated survival function, probability density function and hazard function of the pensioners for the observational period 1981-90. As expected the hazard rate increases with duration (see column 4). It may, however be noted that the pace of increase in the hazard rate h(t) is not uniform. The minimum and maximum increases in h(t) are found during the 2^nd and 7^th years of retirement.

   TABLE 3a

   A synthetic table for pensionners retired during the period 1981-90

   Survival time	ni	di	ci	n'i	qi	pi
   (Year i)
   (1)	(2)	(3)	(4)	(5)	(6)	(7)
   [0 1)	12685	276	1677	11846.5	0.02329	0.97671
   [1 2)	10732	257	1132	10166.0	0.02528	0.97472
   [2 3)	9343	232	1412	8637.0	0.02686	0.97314
   [3 4)	7699	207	1281	7058.5	0.02933	0.97067
   [4 5)	6211	179	1162	5630.0	0.03179	0.96821
   [5 6)	4870	142	1351	4194.5	0.03385	0.96615
   [6 7)	3377	102	1173	2790.5	0.03655	0.96345
   [7 8)	2102	61	1249	1477.5	0.04129	0.95871
   [8 9)	792	17	775	404.5	0.04203	0.95797
   Note: The columns of the table are defined in section 3.1

   TABLE 3b

   Estimates of survival S(i) , probability density f(t), and hazard function h(t) of pensioners based on the data given in table 3a

   Survival time	S(i)	f(t)	h(t)
   (Year i)
   (1)	(2)	(3)	(4)
   0	1.00000	0.02329	0.02356
   1	0.97671	0.02469	0.02560
   2	0.95202	0.02557	0.02722
   3	0.92645	0.02718	0.02977
   4	0.89927	0.02858	0.03229
   5	0.87069	0.02948	0.03444
   6	0.84121	0.03074	0.03723
   7	0.81047	0.03447	0.04217
   8	0.77700	-	-
   Note: S(i)=p0xp1x ... xpi-1xpi i=0,1,....,8. f(t) = Si - Si+1 i=0,1,....,8. and, h(t) = f(t)/S(t) where S(t) = (Si + Si+1)/2 i=0,1,....,8.

   4.2 Differentials in the risk of mortality

   To study occupational differentials in mortality after retirement, the analysis carried out in the previous section has been replicated for pensioners in each of the three categories of occupations considered here. Table 4 gives the estimated hazard rates and relative risks of mortality for the three occupational groups. It indicates that the risk of mortality of pensioners belonging to Group I was higher than that of those belonging to Group II throughout the entire observational period. Although the risk of mortality of occupational Group III was lower compared to the pensioners of Groups I and II, the risk was higher after the 5th year of retirement. These differences are found statistically significant. Similar findings have been observed from analysis of relative risk of death, where R₁₃(t) and R₂₃<1 for all t > 6.

   Thus, the broad findings from the present analysis are that the first six years of retirement of pensioners from the police service were more hazardous compared to those of pensioners in the other two occupational groups i.e., (i) those retired from administrative service and occupations which involved public dealing and (ii) those where employees had little or no contact with the public. Similarly, the latter had lower risks of mortality than those retired from occupations which required administration and public dealing. Figure 1 shows the hazard rates for pensioners of the three occupational categories as detailed in Table 4.

   TABLE 4

   Estimates of hazard and relative risk* of mortality of pensioners of the three occupational groups^**

   	Hazard rate h(t)			Relative risk Rjk(t)
   Survival time
   (Year i)	Group I	Group II	Group III	R12(t)	R13(t)	R23(t)
   0	0.02447	0.02379	0.02250	1.029	1.088	1.057
   1	0.02717	0.02552	0.02439	1.065	1.114	1.046
   2	0.02892	0.02767	0.02626	1.045	1.101	1.054
   3	0.03097	0.02950	0.02830	1.050	1.094	1.042
   4	0.03379	0.03294	0.03031	1.026	1.115	1.087
   5	0.03728	0.03476	0.03474	1.072	1.073	1.001
   6	0.03916	0.03671	0.03924	1.067	0.998	0.936
   7	0.04272	0.04165	0.04320	1.026	0.989	0.964
   Notes: * As defined in section 3.2 ** Group I: Pensioners who were in the police service before retirement. Group II: Pensioners who were in occupations which involved administration and/or direct public dealing before retirement. Group III: Pensioners who were in occupations which did no involve administration and/or direct public dealing before retirement.

   FIGURE 1

   Mortality risk by occupational group

   Source: Table 4

5. DISCUSSION

   Retirement is a major turning point in life. According to Holmes and Rahe's Social Readjustment Rating Scale (1976), it comes tenth in a list of forty-three life crises which includes such events as the death of a spouse, divorce, a jail term, the death of a close family member, and personal injury or illness. The implications of retirement probably deepen in societies where it is mandatory and where people cannot anticipate what their lives will be like when they are no longer in employment. In developing countries few people look forward to retirement, while the majority dread it. However, the way pensioners react depends mainly upon their social liability and their unmet needs at the time of retirement.

   Studying the mortality patterns of men working in the rubber industry, Haynes et al. (1974) detected higher mortality rates among 60-year-olds and 64-year-olds. The researchers could not explain the causes of the higher death rate at 60 years of age. But they attributed the higher rate among those aged 64 to stress connected with their being only one year away from retirement age (i.e.,65 years) (Papalia and Old, 1981). The stress theory is equally applicable in the post-retirement period, particularly during the first few years. Due to increased stress the incidence of morbidity may be raised - which can enhance the risk of mortality during the early years of retirement.

   Furthermore, it is generally the case that workers at higher levels of education and employment status are less eager to retire. Obviously, such workers experience greater stress in the initial years after mandatory retirement because of their inability to fully cope with the changed phase of life. On the other hand, workers with low-level jobs requiring hard physical labour often willingly accept retirement; and when they stop working, their health is likely to improve (Butler and Lewis, 1977; Shanas, et al., 1968). Clearly, due to the ready acceptance of their changed life style following retirement, such people are less likely to suffer stress and may experience a comparatively low mortality risk during the initial years after retirement.

   The observed differentials in the risk of mortality between pensioners in the three categories of occupation support the stress theory and the adverse effects of a changed life style in the years immediately after retirement. It will be recalled that the criterion adopted for defining the different categories of occupation was primarily the similarity of life style within a group of retirees prior to their superannuation. However, the three groups were expected to differ in the extent of the change in their life style following retirement. The elevated risk of mortality during the first few years after superannuation among the pensioners who retired from the police service, as compared to the other two occupational categories, can be explained in terms of the magnitude of change experienced in their life style. In fact, during the service period, both police personnel and those in jobs which involved administration and public dealing (i.e. Group II) enjoyed authority and power. Soon after retirement, the sudden loss of status (including loss of power, income, respect and identity) often results in emotional shock and psychological stress. These considerations may contribute to declining health and early aging, thus increasing the risk of mortality. The influence of these factors may be greater during the initial years of retirement.

   It should be recalled that the pensioners who retired from the police service followed a mandatory regime of physical fitness during their years of active service. This fitness regime comes to an abrupt end after their superannuation. Thus they experience a major change in their life style ascompared to those people who retire from occupations which simply involve administration and public dealing (Group II). The contrast in life styles before and after superannuation in the case of police personnel may therefore involve a faster rate of aging and earlier death during the first few years after retirement.

   The pensioners of Group III - who retired from occupations where there was little or no change in life style - had relatively small risk of mortality during the early years of retirement as compared to those who either retired from the police service (Group I) or from occupations which involved administration and public dealing (Group II). It should be noted that the majority of Group III retirees experienced little difference in their life styles during their pre- and post- retirement periods. Thus, because they experienced little or no change in life style they may not lose their sense of continuity of service after retirement. And their risk of mortality during the first few years of retirement was not as high as for occupational categories I and 11. However, in later years and perhaps due to lower income, a poorer diet, and less health care, its category of pensioners experienced higher risks of mortality as compared to pensioners retired from the police service and from occupations involving administration and public dealing.

6. CONCLUSION

   Using a new data set, this paper has examined the risks of mortality by occupation for those who retired from the Maharashtra State Government Service during the period 1981-90. Three occupational groups were considered. These were: (i) pensioners who were in the police service before retirement (Group I), (ii) pensioners who were in occupations which involved administration and/or direct public dealing before retirement (Group II), and (iii) pensioners who were in occupations which did not involve administration and/or public dealing before retirement (Group III). The study focused on those people who retired from active service at the normal retirement age under a formal retirement system, i.e., following superannuation. Post retirement risks of mortality have been examined. The findings have established that the risk of mortality was higher up to the 5th year after retirement for retirees of occupational Group I, followed by those of Group II, as compared to the risk for those who retired from occupations which did not involve administration and/or public dealing (Group III).

   In India the majority of government employees do not want to retire at the age of fifty-eight (normal retirement age), and the proportion of workers affected by mandatory retirement policies is very large. Thus, there is a pressing need to formulate comprehensive retirement policies which include pre-retirement programmes for government employees so as to help them appreciate how they can best control their future, plan to overcome social liabilities, and be emotionally and psychologically prepared for their forthcoming years of retirement (Macionis, 1987). Such programmes, if planned and properly implemented, may go a long way towards helping prospective retirees ease their transitions into retirement and minimize the risks of mortality during the initial years.

¹ A detailed description of the process involved in the preparation of pension papers and PPOs, and the rules and regulations governing the maintenance of pension records and the disbursement of pensions by the Pay and Accounts Office/Treasury Office can be found in Kumar (1992). Back