Retirement Preparation Programs: Differentials in Opportunity and Use 1

Journal of Gerontology 1984, Vol. 39. No. 5. 596-602 Retirement Preparation Programs: Differentials in Opportunity and Use 1 Scott H. Beck, PhD : Two issues rarely addressed in the retirement planning field are (a) the proportion of older workers who participate, or have the opportunity to participate, in retirement preparation programs; and (b) socioeconomic differentials in access to such programs. Data from the National Longitudinal Surveys of older men were used to investigate these two issues. The data indicate that fewer than 4% of this sample of men aged 60 to 74 in 1981 had participated in a retirement preparation program. Logistic multiple regression analysis indicated that level of education, occupational status, government employment, and private pension coverage were positively related to the likelihood of participation as well as the likelihood of opportunity to participate. Conclusions from this analysis are (a) very few older men are ever exposed to retirement preparation programs and (b) those who would seem to benefit most from preparation programs, low status and low income workers, are the least likely to have access to these programs. Key Words: Socioeconomic factors, National Longitudinal Surveys A TUCH of the research on retirement preparation -LV1 programs has focused on the structure (type) and content (topics covered) of such programs (e.g., Kasschau, 1974; Pyron, 1969; Siegel & Rives, 1978, 1980), or their effect on attitudes and behaviors (e.g., Charles, 1971; Glamser, 1981; Hunter, 1968). Although these issues are important, especially from an applied perspective, there are other rarely addressed issues regarding retirement planning programs and their participants. Two such issues are (a) the proportion of older workers having an opportunity to participate and the proportion who actually participate in retirement preparation programs, and (b) socioeconomic differentials in participation and opportunity to participate in such programs. The purpose of this investigation was to address these two issues using data from the National Longitudinal Surveys of older men. The usual method of data collection for research on retirement preparation programs is to send relatively short questionnaires to personnel directors or other executives of hundreds of large corporations. These samples are admittedly not representative of all corporations and probably not representative of all large corporations, and the return rates of the questionnaires are usually below 50%. Under- 'This research was supported, in part, by National Institute on Aging grant AG-03565. The research and original drafts of this paper were completed while the author was an assistant research professor in the Center on Aging Studies, University of Missouri-Kansas City. 2 Andrus Gerontology Center, University of Southern California, University Park MC 0191, Los Angeles, CA 90089. standably, findings from these surveys vary widely. Results from such surveys suggest that anywhere from 13 to 50% of corporations have some type of preretirement counseling program or preparation program (O'Meara, 1977; National Council on Aging, NCOA, 1980; Siegel & Rives, 1978, 1980). Group retirement preparation programs are also sponsored by institutions other than corporations. Reich (1977) estimated that as of the mid-1970s over 100,000 people had participated in some type of group program. This number constitutes only a small proportion of the older work force. The NCOA-sponsored Harris surveys (1975, 1981) indicate that only 8% of those aged 65 and over reported having enrolled in retirement counseling or preparation courses. In these reports no breakdown by sex, occupation, or other characteristics were provided for this response. Men should be more likely to have participated because of their higher labor force participation rate, and we would expect participation to vary by other characteristics as well. Because most researchers in the retirement planning field assert that these programs are helpful to older workers in planning for and adjusting to retirement, it is important not only to obtain an estimate of the proportion of older workers who have participated, but also the proportion who had an opportunity to participate. The other focus of this investigation concerned the socioeconomic characteristics of those who do participate or have the opportunity to participate in planning programs. Atchley (1976) asserted that middle status workers (clerical, sales, and skilled 596

RETIREMENT PREPARATION PROGRAMS 597 blue collar) are most likely to participate in preparation programs because they are favorably disposed toward retirement and tend to work in organizations providing such programs. Most research on the effects of planning programs involves one company (Fitzpatrick, 1979) or a few firms in one industry (Glamser & DeJong, 1975). There does not appear to be any existing research that compares demographic, socioeconomic, or other characteristics of participants in preparation programs with those of the larger population of older workers. One study (Morrow, 1981) did use a stratified random sample of older employees at a state university to compare characteristics of participants in a retirement planning course with those not participating. Morrow found no significant differences between participants and nonparticipants in level of education, occupation, or income. As Morrow noted, however, this sample cannot be construed to be representative of older workers in general. Company-sponsored retirement preparation programs may be viewed as a fringe benefit (Hall, 1980), and such programs in general may be considered a valued resource for preretirees. Like other retirement benefits (e.g., pensions), opportunity to participate and actual participation in programs are distributed differentially among workers based on certain social and economic attributes. Those socioeconomic factors found to be important in research on other aspects of retirement education, occupation, industry, and class of worker (self-employment) were used in this investigation, as well as race and pension coverage. METHOD Sample. The data used for this analysis were from the National Longitudinal Surveys (NLS) of older men. The NLS is based on a multistage probability sample of the noninstitutionalized population of men aged 45 to 59 in 1966 (Center for Human Resource Research, 1982). The sample was drawn by the Census Bureau from sampling units that originally had been selected for Monthly Labor Surveys in the mid-1960s. The sample size for this cohort of men for the first survey in 1966 was 5,020. Blacks were oversampled by a ratio of 3:1 in order to allow for separate analyses by race. Interviews have been conducted either every year or 2 years, the latest interview having taken place in 1983. Information contained in the 1981 survey as well as information from prior waves was used in this analysis. In the 1981 NLS survey there were 2,832 respondents who at the time were between age 60 and 74. The primary cause of attrition was death (1,307 respondents). This natural attrition does not affect the representativeness of the sample because we wish to generalize to the cohort of men 60 to 74 in 1981. Between the 1966 and 1981 surveys 881 respondents (17.5% of the original sample) dropped out of the study for reasons other than death. This attrition slightly affects the composition of the sample. Comparing the characteristics of those who dropped out with those remaining in the sample as of 1981, we find that those who dropped out have slightly higher mean occupationl status (measured in 1966) and more education. Otherwise the two groups are very similar. One purpose of this study was to obtain estimates of the proportion of older men who have participated in, and those who have had the opportunity to participate in, retirement preparation programs. Because blacks were oversampled in the NLS (in 1981 blacks comprised 28% of the sample) and because attrition due to causes other than death introduces some degree of bias, generalization to the population of men aged 60 to 74 in 1981 is problematic. Weights assigned to the respondents still in the sample in 1981, however, adjusts for these two factors. Thus, weighted frequencies were used to obtain the estimates of participation and opportunity to participate. The second part of this analysis involved estimating the effects of certain socioeconomic factors on the probabilities of participating and having an opportunity to participate. The small amount of attrition bias should not have much of an effect on the estimation of parameters. Variables and method of analysis. In the 1981 interview several items were included pertaining to retirement preparation programs. Two questions form the basis of this analysis. The first question, asked of all respondents, was "Have you ever taken a course, either at work or anywhere else that attempted to help you prepare for retirement?" Those who responded no were asked, "Did you ever have an opportunity to take such a course?" These questions appear to exclude personal briefings or counseling of workers by employers' personnel staff or other one-to-one meetings. The questions obviously do include courses and programs other than those offerred by corporations. The items are very general and demand recall of something that may have occurred many years in the past. Participation in a structured course on retirement planning, however, is an uncommon

598 BECK event and would most likely be remembered. This may be less true in the case of opportunities to participate. From these questions two dependent variables were created: actual participation in a preparation course, based on the first question, and opportunity to participate, combining responses from both questions. A total of six independent variables were used in the analysis of the dependent variables. Although Morrow (1981) did not find any differences in education level between participants and nonparticipants in her study, it was expected that men with more years of formal schooling would be more likely to participate and to report having an opportunity to participate. Men with little formal education may be reluctant to enroll in courses demanding some verbal and written skills. Furthermore, because the questions upon which the dependent variables are based include programs offered in the community, men with higher education may have more knowledge of such offerings. Based on Atchley's (1976) assertion that exposure to such programs differs by occupation, it was hypothesized that occupational status (measured by the Duncan Socioeconomic Index) would be positively related to both participation and opportunity to participate. It is well known that the large majority of company-sponsored retirement preparation programs are offered by large corporations with extensive economic resources (NCOA, 1980; O'Meara, 1977). The NLS does not contain information on size or other characteristics of the employer, so it is not possible to directly measure this factor. These types of corporations, however, tend to be located in certain industries (O'Rand & Henretta, 1982; Tolbert et al., 1980), usually termed the "core." Tolbert et al. (1980) schema was used to classify industries into core and periphery sectors. Because this dichotomy is very general, I separated from the core the goverment sector, which closely resembles large corporations in size, economic resources, and provision of pensions. It was hypothesized that employment in core industries, and government employment especially, would result in greater likelihoods of participation and opportunity to participate. There is no information on the propensity of self-employed workers to enroll in planning programs because virtually all research on such programs is based on surveys of large corporations. It seems likely that the self-employed would have lower probabilities of participation and opportunity to participate, unless they are more likely than wage and salary workers to utilize noncorporate programs. Atchley (1980) noted that almost all companysponsored retirement preparation programs exist alongside pension plans. Thus, private pension coverage was hypothesized to increase the chances of participation and opportunity to participate, even if noncorporate programs are included in the definition of the dependent variables. Pension coverage, which is a dichotomy (yes/no) based on a question in the 1966 survey, was viewed as an intervening variable between the four independent variables just discussed and the two dependent variables. The sixth predictor, race, was included as a control variable. At the bivariate level, whites may be more likely to participate and to have an opportunity to participate in preparation programs. The net effect of race with the five other predictors in the models, however, should be trivial and nonsignificant. Because both of the dependent variables are dichotomous, logistic multiple regression (Hanushek & Jackson, 1977; Nerlove & Press, 1973) was used. This maximum likelihood method allows for the inclusion of both categorical and interval-level variables. This is crucial in this analysis due to the occupational status and education variables. RESULTS Descriptive statistics. Shown in Table 1 are the actual frequencies as well as weighted proportions of NLS respondents who participated in prep- Table 1. Descriptive Statistics on Participation and Opportunity to Participate in Retirement Preparation Programs Variables Total NLS sample (n = 2,832) Actual Weighted 1 " Actual Group of ever-retired NLS respondents (n = 2,012)' Weighted Opportunity to 305 12.1% 207 11.7% participate (10.8%) (10.9, 13.3) (10.3%) (10.3, 13.1) Participants 78 3.4% 62 3.8% (2.8%) (2.7, 4.1) (3.1%) (3.0, 4.6) Nonparticipants 227 (8.0%) 8.7% (7.7, 9.7) 145 (7.2%) 7.9% (6.7, 9.1) "Ever-retired respondents were those men who reported themselves as retired in one or more surveys over the 1966 to 1981 period. Almost 90% of these respondents were retired in the 1981 survey. 'The percentages in this column are based on weighted frequencies. The numbers in parentheses below the weighted frequencies represent the lower and upper bounds of a 95% confidence interval.

RETIREMENT PREPARATION PROGRAMS 599 aration programs, had the opportunity to participate but did not, and the combination of these two groups. Numbers in parentheses below the frequencies are sample proportions, and numbers in parentheses below the weighted proportions represent the lower and upper bounds of a 95% confidence interval. These descriptive statistics are shown for two groups: the entire sample of older men in the 1981 survey and the group of respondents who reported themselves as retired at some time during the 15-year period of the study (hereafter referred to as the ever-retired). The estimates in Table 1 clearly show that a very small proportion of the respondents ever participated in a retirement preparation program. This is true whether or not the respondent has ever retired. Even choosing the upper bound of the confidence interval for ever-retired respondents (4.6%) as an estimate of the proportion of men aged 60 to 74 in 1981 who had been exposed to comprehensive retirement planning programs, fewer than 1 in 20 ever participate in such programs. The estimates in Table 1 also show that fewer than 10% of those men who had not taken a preparation course reported ever having an opportunity to participate in such a program. It should be noted that opportunity to take a retirement preparation course is somewhat subjective. Many adult education programs, colleges, and other organizations offer such courses to the public. The opportunity to take such a course is probably greatly affected by knowledge of such opportunities and, in turn, by degree of interest in retirement planning. Combining those respondents who did participate with those who had an opportunity to participate but did not, approximately 1 out of 8 had an opportunity to take a retirement preparation course. These estimates indicate that despite the reported proliferation of preparation programs in this country, relatively few older men have had an opportunity to participate in them. As noted before, respondents who participated or had an opportunity to participate were asked other questions concerning retirement preparation programs. One of the questions pertained to the place where the program was offered. In Table 2 frequencies and unweighted proportions (the weighted proportions do not differ much from the unweighted) are shown for the four categories of location of program. The information in Table 2 indicates that slightly more than half of the participants took the course at work, whereas one fourth went to some type of school, and one fifth attended' 'elsewhere.'' Of those reporting an opportunity to participate, Table 2. Location of Retirement Preparation Program Subsample of NLS respondents Opportunity to participate (n = 301) Participants (II = 78) Nonparticipants (n = 223) Work 125 (41.5%) 42 (53.8%) 83 (37.2%) Location of program School 95 (31.6%) 19 (24.4%) 76 (34.1%) Union 18 (6.0%) 1 (1.3%) 17 (7.6%) Elsewhere 63 (20.9%) 16 (20.5%) 47 (21.1%) "Of the 227 respondents who reported having an opportunity to take a retirement preparation course, 4 did not indicate where the opportunity was given. substantially fewer than half indicated that this opportunity was given at work. When considering the entire sample of NLS respondents, the 42 participants who took the course at work represent only 1.5%. Even combining participants with those reporting an opportunity, the 125 respondents who identified work as the location constitutes but 4.4% of the sample. Effects of socioeconomic factors. First, it should be noted that because of the extreme skew of the dependent variables, especially that of the probability of participation, the possible size of the covariance between any independent variable and the dependent variable is reduced. Thus, the independent variables will not have large effects on the dependent variables, and even with a large sample the logit coefficients may not be statistically significant. coefficients represent the change in the log-odds of the dependent variable for a one-unit change in the independent variable (Hanushek & Jackson, 1977). The anti-log of this coefficient indicates the degree to which a one-unit change in the independent variable multiplies the odds of being 1 as opposed to 0 on the dependent variable. The results of the logistic multiple regression analyses of both dependent variables are presented in Table 3. Models IA and IIA show bivariate relationships between each of the predictors and the dependent variables, Models IB and IIB present net effects of the first five predictors (pension coverage excluded), and Models IC and IIC show net effects when all six independent variables are in the equations. In both of the bivariate models, whites were much more likely to have participated and to have had an opportunity to participate compared with blacks. When other predictors are added in the equation, the significant relationship disappeared in the case of opportunity to participate but, surpris-

600 BECK Table 3. Logistic Regression of Participation and Opportunity to Participation in Retirement Preparation Program Independent variables Race (white = 1) Education Occupational status Core industry Government Self-employed Pension coverage Intercept Reduction in x 1 " Model IA 1.40**.20**.03**.37 1.52** -1.13 1.52** 4.05 1.22 1.03 4.55 4.56 Participation Model IB.85* 2.34.06.03** 1.03.01 1.04** 2.83-1.06-6.20 90.95 Model IC.87* 2.38.04.02** 1.02 -.12.88** 2.41 -.81.85** 2.34-6.47 98.98 Model HA.70**.16**.02**.14.91** -.12 1.00** 2.02 1.17 1.02 2.48 2.72 Opportunity to participate Model IIB.22.09** 1.09.01** 1.01 -.01.67** 1.95 -.11-3.83 137.37 Model IIC.23.08** 1.08.01** 1.01 -.10.54** 1.72.06.61** 1.84-3.96 155.76 This value represents the difference in x 2 between the model with only the intercept and the model under which the value is located. There were 6 degrees of freedom in Models IB and IIB, 7 in Models IC and IIC. *p <.05; **p <.01 ingly, not in the case of participation. Years of education had the expected positive effect on the dependent variables, but this effect is nonsignificant in Models IB and IC. The large reduction in the effect of education on participation from the bivariate model to models with other predictors was due almost entirely to the high covariance between it and occupational status, which, in this case, is related more strongly to the dependent variable. Also, as hypothesized, occupational status had significant positive effects on the dependent variables throughout the various models. It should be noted that the smaller absolute size of the coefficients for education and occupational status in comparison with other predictors was due entirely to the fact that they are interval-level variables. In terms of amount of reduction in chi square attributed to the predictors, occupational status and education had net effects as large as the other significant predictors (except in those cases where education was not significant). A surprising finding was the lack of a significant contrast between core and periphery industries in any of the models. This finding cannot be attributed to the skewed distributions of the dependent variables but may, in part, reflect the inclusion of noncompany sponsored programs in the definition of the dependent variables. Government employees were much more likely to participate in retirement preparation programs and to have had an opportunity to participate. The relatively large negative coefficients for the self-employed versus wage and salary contrast in the models predicting participation did not attain statistical significance. In this case nonsignificance may be due largely to the combination of a highly skewed dependent variable and a skewed independent variable (10% of the sample are or were selfemployed) that produced large standard errors. The very small coefficients for this contrast in models predicting opportunity to participate strongly indicated that, as with the core versus periphery contrast, the inclusion of noncompany sponsored programs in the definition of the dependent variables reduced differences in opportunity to participate. Pension coverage had the expected positive effect on both dependent variables and was always significant. The inclusion of this mediating variable did not affect the significance of other predictors in Models IC and IIC but did reduce the size of the net effects of other significant predictors, especially government employment. In general, the logistic multiple regression analyses of the probabilities of participation and opportunity to participate produced very similar findings. The positive effect of education was somewhat greater in models predicting opportunity to participate as opposed to actual participation. This may reflect more knowledge of noncompany sponsored programs by men with more years of formal schooling. DISCUSSION Before discussing the results, a note of caution is in order. The items included in the NLS on retirement preparation programs are general and somewhat ambiguous. It is possible that some respondents who participated in group programs at work were confused by the term "course" and did not answer in the affirmative (the same would hold true

RETIREMENT PREPARATION PROGRAMS 601 for the opportunity item). Second, the question on opportunity is certainly subject to a degree of recall error, especially by those respondents who were never interested in such programs. Because the NLS was not constructed for investigating exposure to retirement planning programs, there is no information on structure or content of the programs. (There are questions on length of program however.) Thus, we have no idea of the distribution of different programs among this sample of older men. Finally, in regard to the analysis of differentials in opportunity and use, the effect of one important factor could not be investigated attitude toward retirement. Items on attitude toward retirement were not included in the NLS until the 1978 survey, and a large proportion of the respondents had already retired. These drawbacks, which are common in national surveys, are not considered to constitute a serious limitation to the use of information in the NLS. In regard to the first issue addressed in this study, the NLS data indicated that very few men in the present 60-to-74 cohort were ever exposed to programs or courses designed to help them plan for retirement. Given that less than 2% of the NLS sample of older men reported taking a retirement preparation course at work, it becomes evident that the "proliferation" of such company-sponsored programs is restricted to a small segment of the work force. As noted previously, most researchers and practitioners in the retirement planning field have focused their efforts on describing and assessing the content of company-sponsored programs and/or evaluating the impact of such programs on the individuals involved. Although the empirical evidence is equivocal, most writers in this area claim that preparation programs are helpful to workers and that comprehensive programs are the most successful. Very few analysts in the field address the issue of whether such programs should be available to more older workers and how this might be accomplished. Monk (1972) did propose federal government involvement in offering counseling and/or structured programs on retirement planning to the general public. Other possibilities exist, such as expanding community-based programs already in existence. More attention might be given by those involved in retirement planning to ways in which programs and counseling services for older workers could be expanded to encompass a significant proportion of this population. The results of the analysis of socioeconomic differentials in opportunity and use of preparation programs indicated that older men with more years of education, higher occupational status, and private pensions have greater access to these types of programs, as do government employees. One interesting finding was that the self-employed and those in periphery industries were just as likely to have opportunities to participate and to actually participate in preparation programs as wage and salary workers and those whose jobs are located in core industries. This appears to be due to the availability of such programs outside the workplace. The finding that government employment significantly increases the likelihood of participation and opportunity to participate, however, implies that workers in large bureaucratic organizations are more likely to receive this type of fringe benefit. These findings reveal that it is the socially and economically advantaged worker who is most likely to have an opportunity to participate in retirement preparation programs. Perhaps this is what Walker and Lazer meant when they asserted, "Often, those who attend are those who need the assistance the least" (1978, p. 194). It appears that the group of older workers who would benefit most from retirement preparation programs those with less education, lower occupational status, no pension coverage, and consequently lower retirement income are the least likely to have access to such programs and services. REFERENCES Atchley, R. C. The sociology of retirement. Schenkman Publishing, Cambridge, MA, 1976. Atchley, R. C. The social forces in later life. (3rd ed.). Wadsworth Publishing, Belmont, CA, 1980. Center for Human Resource Research. The National Longitudinal Surveys Handbook. College of Administrative Science, Ohio State University, Columbus, 1982. Charles, D. C. Effect of participation in a preretirement program. The Gerontologist, 1971, //, 24-28. Fitzpatrick, E. W. Evaluating a new retirement planning program: Results with hourly workers. Aging and Work, 1979, 2. 87-94. Glamser, F. D. The impact of preretirement programs on the retirement experience. Journal of Gerontology, 1981, 36, 244-250. Glamser, F. D., & DeJong, G. F. The efficacy of preretirement preparation programs for industrial workers. Journal of Gerontology, 1975, 30, 595-600. Hall, G. Retirement planning: Suggestions for management. Aging and Work, 1980, 3, 203-209. Hanushek, E., & Jackson, J. Statistical methods for social scientists. Scientific Press, New York, 1977. Harris, L., & Associates. The myth and reality of aging in America. National Council on Aging, Washington, DC, 1975. Harris, L., & Associates. Aging in the eighties: America in transition. National Council on Aging, Washingtor, DC, 1981.

602 BECK Hunter, W. W. Preretirement education for hourly-rated employees. Division of Gerontology, University of Michigan, Ann Arbor, 1968. Kasschau, P. L. Reevaluating the need for retirement preparation programs. Industrial Gerontology, 1974, /, 42-59. Monk, A. A social policy framework for pre-retirement planning. Industrial Gerontology, 1972, 15, 63-70. Morrow, P. C. Retirement planning programs: Assessing their attendance and efficacy. Aging and Work, 1981,4, 244-252. National Council on the Aging. Retirement preparation: Growing corporate involvement. Aging and Work, 1980,-J, 1-13. Nerlove, M., & Press, S. J. Univariate and multivariate loglinear and logistic models. The Rand Corporation, Santa Monica, CA, 1973. O'Meara, J. R. Retirement: Reward or rejection. The Conference Board, New York, 1977. O'Rand, A. M., & Henretta, J. C. Delayed career entry, industrial pension structure, and early retirement in a cohort of unmarried women. American Sociological Review, 1982, 47, 365-373. Pryon, H. C. Preparing employees for retirement. Personnel Journal, 1969, 48, 722-727. Reich, M. H. Group preretirement education programs: Whither the proliferation? Industrial Gerontology, 1977, 5, 29-43. Siegel, S. R., & Rives, J. M. Characteristics of existing and planned preretirement programs. Aging and Work, 1978, /, 93-99. Siegel, S. R., & Rives, J. M. Preretirement programs within service firms: Existing and planned programs. Aging and Work, 1980, J, 183-191. Tolbert, C. M., Horan, P. C, & Beck, E. M. The structure of economic segmentation: A dual economy approach. American Journal of Sociology, 1980,85, 1095-1116. Walker, J. W.,&Lazer, H. L. The end of mandatory retirement. John Wiley & Sons, New York, 1978. 37th Annual Scientific Meeting of The Gerontological Society of America SanAntonk) One of Americas four unique cities. November 16-20, 1984 San Antonio Convention Center San Antonio with its rich ethnic and historical heritage, provides the perfect environment for the theme of the meeting, "Gender and Ethnicity as Variables in Aging."