Work After Retirement: the Experiences of Older Men in the United States

全文

(1)SOCIAL SCIENCE RESEARCH. 23, 82-107 (1994). Work after Retirement: The Experiences of Older Men in the United States MARK. D.. HAYWARD. Department of Sociology and the Population Research Institute,. Pennsylvania State University. MELISSA. A.. HARDY. Department of Sociology and the Institute on Aging, Florida State University. AND. MEI-CHUN LIU. Institute for Labour Research, National Chengchi University, Taipei, Taiwan The permeability of the work/ retirement boundary is examined by investigating the labor force reentry process among a group of male retirees. Using data uom the National Longitudinal Survey of Older Men, hazards models are estimated to identify the determinants of postretirement work. Reentry is distinguished according to part·time and full-time work to capture potentially important sources of diversity. The results show that reentry occurs quickly-typically within the first year or two after a labor force exit. Despite career inteuuption , several work career factors significantly alter the overall chances of reentry. Comparing reentry determinants of full-time versus part-time work suggests that postretirement, part time work is a disti.nct state referencing partial retirement ; it is not a middle ground on a continuum between career work and complete retirement. In addition, characteristics positively linked to an initial early retirement negatively affect reentry into full-time (although not part-time) work. The determinants governing moves into and out of the labor force, and into and out of partial retirement are quite different. Overall, the results demonstrate that the transition uom work to retirement is neither uniform nor irreversible. ~ 1994 Academic Press, Inc. This work was partially supported by grants from the National Institute on Aging (No. AG04915) and the AARP Andrus Foundation. We thank Dan Lichter, Mary Fennell and an anonymous reviewer for their comments. Address correspondence and reprint , requests to Mark D. Hayward, Department of Sociology, Pen.!} State University, University Park, PA 16802-6207 .. 82 0049-089X/94 $6.00 Copyright © 1994 by Academic Press, Inc..

(2) WORK AFTER RETIREMENT. 83. Recent life course theories posit that public systems of age grading have intensified the periodization of life stages in modern society. The result is a tripartition of the life course into segments of labor force preparation, economic activity, and retirement (Kohli, 1.986; Mayer and Schoepftin, 1989; Meyer, 1986). Within this framework, retirement represents a tran sition to a nonwork status closing the work career. It signifies the end of the period in which individuals' ambitions and achievements are centered on the labor force and the start of the period in which people turn toward leisure (Ekerdt, 1986). Public pension programs with universal coverage and formal eligibility rules based on age reinforce the perception of re tirement as predictable and routine. Moreover, the stringent earnings test for Social Security pensions penalizes paid work consequent to retirement. A growing body of empirical evidence suggests that late-life work ac tivity is more varied and complex than this description implies. At a minimum, the high degree of postretirement labor force participation demonstrates that retirement is not the end of the work career for a substantial portion of people (Beck, 1986; Hardy, 1991; Myers, 1991; Quinn, Burkhauser and Myers, 1990; Ruhm, 1989). Indeed, during the 1970s as the age of retirement declined, the incidence of reentry more than doubled among men and women aged 55 years and older (Hayward, Grady, and McLaughlin, 1988b). These empirical accounts suggest that the transition between the economic activity and retirement life course stages is often a complex series of events. Although some workers may experience retirement as an abrupt shift from work to leisure, for others the work-to-retirement transition is less clear cut. Rather than a single event, retirement is embedded in a more prolonged process involving movements both out of and into the labor force . The complexity of these work transitions makes retirement a more ambiguous event. The increas ing frequency of reentry also suggests a blurring of retirement and work statuses rather than movement toward further segmentation of these life stages (Hardy, 1991). If an increasing number of retirees are choosing to resume employment, what are the factors structuring the reentry process? The answer to this question depends, in part, on whether and how preretirement and postre tirement employment are linked. We therefore investigate the extent to which preretirement employment characteristics, as well as factors op erating during retirement, shape postretirement employment. Our theo retical interest lies in exploring how reentry transitions are embedded in the overall structure of work careers .. Retirement Transitions and Work Careers. After an initial retirement, older workers can either accept retirement as a permanent status or attempt to find alternative employment. 1 If the.

(3) 84. HAYWARD, HARDY, AND LIU. search is unsuccessful, retirement becomes permanent by default. The reentry process thereby implies two selection processes: retirees' self Se lection into employment, and the market selection of retirees from the pool of available labor. These two processes are intertwined through supply/demand market relations. Theoretically, the market can pull any one into the labor force if the terms are sufficiently enticing. What is distinct about retirees is that, for many, the alternative to work (i.e., continued retirement) carries with it some consistent level of income replacement. If income replacement is too low, presumably workers would not have chosen retirement in the first place. If retirement income is sufficient, then returning to work may hold little attraction. Retirement transitions, however, do not always occur as planned. The fit between the costs of financing retirement and the flow of retirement income can be more ragged than anticipated. Older workers also may misjudge their retirement preferences . Even if initial plans are on target, the erosion of income purchasing power can lead retirees back to the labor force. The analytic focus of this study is on older men's actual reentry behavior rather than intentions or future work plans. Persons at risk of reentry are those men observed to have exited the labor force after age 55. Reentry occurs when a person out of the labor force resumes paid employment. Note that retirees may remain retired either because they don't want to reenter or because they cannot find suitable positions. The incidence of reentry, therefore, is indicative of both desire and success; it captures retirees who, at some point, wanted to return to work and were successful in securing satisfactory positions . Workers who reestablish a connection to the work force can reenter either as full-time or part-time workers. This distinction has important implications for understanding the more general retirement process. Per sons returning to the labor force in a full-time capacity relinquish by standard definition the retirement status. In contrast, retirees who return to work part-time may simply be modifying the conditions of retirement relative to their perceived needs or modifying their work schedules relative to a desire for increased leisure. These part-time workers/part-time re tirees occupy a status generally referred to as partial retirement (Reimers and Honig, 1993). A desire for structured activity, social contacts, or additional income may prompt retirees to seek part-time jobs. Preferences for part-time or full-time work, coupled with older workers' ability to compete in the labor market, may result in distinct reentry age 55. Thus, our study conforms to prior research using labor force status changes to examine the work-to-retirement transition. tabor force exiting events are di-fferentiated in terms of respondents' self-reported ability to work. This allows us to distinguish disabled and nondisabled persons out of the labor force ..

(4) WORK AFTER RETIREMENT. 85. processes . The differentiation of reentry destinations according to part time versus full-time employment allows us to determine whether the reentry process itself is sensitive to the level of labor supplied. If no differences exist, then support is indicated for the idea that the reentry dynamic is indifferent to work schedule demands. Indeed, such evidence would suggest that the relative frequency of part-time vs full-time reentry is a reflection of such positions in the market. More likely is the possibility that the processes driving reentry to full and part-time work differ either in terms of degree or in the structure of determinants (or both). Differ ences in terms of degree support the idea that partial retirement is part of a ;l abor supply continuum ranging from career work to complete re tirement (Reimers and Honig, 1993). For example, one might expect that medium inflation levels fuel reentry to part-time work while high inflation rates prompt reentry to full-time work. Differences in the structure of determinants, however, signify discontinuities in the labor supply function related to the retirement state, supporting the idea that partial retirement is a distinct status (Reimers and Honig , 1993) . Given the negative career consequences generally associated with in terrupted work activity, the view of retirement as a career ending transition is not surprising. Empirical evidence gives some credence to this position. Older retirees' skills and experience , for example, often are not trans ferable to other settings, and there are substantial declines in occupational status and wages for reentrants (Beck, 1986; Hayward, Grady, and McLaughlin , 1988a; Honig and Reimers , 1987; Palmore et al., 1985; San dell, 1985) . Career achievements for those reentering the labor force thus are not maintained. Also problematic is the fact that retirees' age and inexperience in conducting a job search contribute to relatively high rates of unsuccessful searches (Sandell , 1985). Despite such evidence, there are grounds to suspect that much of postre tirement labor force behavior is governed by at least some of the same social structural factors shaping the careers of prime-aged persons. Some researchers have argued that reentry--especially reentry to full-time em ployment-is often a planned event and a part of a person's labor force career (Myers, 1991; Honig and Reimers, 1987). Structural circumstances make it difficult to shift to partial retirement in one's career job by lowering work effort, job exits generally are required to trigger private pension receipt, and few part-time jobs may be available for postretire ment employment (Reimers and Honig, 1993). Workers unable to time the career job's end with beginning a new job thus experience "retire ment" spells. From this perspective, retirement is a pause in an on-going career rather than the point of career termination. Reentry is a reversal of the retirement transition; it is a move back into previous patterns of work activity rather than to some new organization of work/leisure time. In many respects, it may be indistinguishable from other job shift be.

(5) 86. HAYWARD, HARDY, AND LIU. haviors. Reentry, especially to full-time work, is thus likely to be linked closely to prior position in the occupational structure or to the skills and experience gained through job performance. With regard to partial retirement, there is much less to suggest a con tinuation of traditional career determinants. By its very definition, partial retirement refers not to the rejection of the retired status, but to the modification of the social or economic conditions of retirement. A sub stantial number of older workers want to work after retirement, and most prefer part-time employment (Louis Harris and Associates, 1981)-a pref erence often based on the desire to combine leisure activities with sup plemental income and social contacts (Palmore, Burchett, Fillenbaum, George, and Wallman, 1985; Parnes and Less, 1985). It follows, therefore, that conditions affecting the desired amount of activity or income in retirement--conditions such as inflation and boredom--can provide the impetus for reentry to part-time work and that the dynamic of reentry will be governed more by general work-related skills and retirement con ditions rather than the specific characteristics of the preretirement oc cupational position., Determinants of Reentry. In this analysis, we include a variety of individual and structural char acteristics referencing the preretirement job and the conditions and ex periences of retirement. Many of the factors we are considering are as sociated with older men's retirement (Hayward, Grady, Hardy, and Sommers, 1989; Palmore et aI., 1985). Only recently has research ex amined directly the determinants of postretirement work (e .g., Beck, 1986; Myers, 1991; Reimers and Honig, 1993), although problems of definition hinder explicit comparisons. Characteristics of the preretirement job allow us to make some assessment of the relationship between pre and postretirement employment. These factors are expected to influence reentry chances to the extent that preretirement jobs define retirees' range of job opportunities and work preferences. Characteristics of retirement conditions and experiences aUow us to relate the quality of the retirement status to reentry. With regard to the latter, we examine the effects of individual-level measures of financial status, and disabling and nondisabling health con ditions . These factors represent core dimensions defining retirees' quality of life and the immediate constraints on reentry decisions (Myers, 1991). They influence the decision based on an evaluation of continuing retire ment against the structure of labor market alternatives . The prevalence and attractiveness of labor market alternatives are themselves a product of conditions within the economy. For example, assuming employers treat older workers as employees-of-last-resort, reentry chances should be greater during periods of low unemployment. Similarly. persons whose.

(6) WORK AFTER RETIREMENT. 87. retirement income is threatened by inflation should be more likely to reenter to supplement pension benefits with earnings . Measures of the annual unemployment rate and the inflation rate , therefore, are included to capture the effects of labor market conditions . Our model also takes account of individuals' social and demographic characteristics. Older men's retirement has been associated with a number of ascribed and achieved characteristics , including age, race, education, and marital status (Hayward et al. , 1989; Palmore et al., 1985). Prior research has shown that nonwhites have higher disability rates (Parsons, 1980) mitigating their reentry chances. Net of occupation, we anticipate that education win be associated positively with reentry based on the positive relationship between education and the preference for work. Marital status is an indicator of household consumption levels, financial resources, and social support, suggesting that single men may be more likely to reenter since there are fewer familial constraints. Age is expected to be negatively related to reentry. Age is related to the social acceptability of retirement and is a major criterion of pension eligibility. Perceptions of life expectancy also are age-linked ; thus , age defines the time horizon for reentry. The nature of work influences retirement decisions by defining the attractiveness of continued work activity relative to nonwork alternatives. Workers with the lowest retirement rates , for example, hold jobs char acterized by autonomy, intellectual challenge, and low physical demands (Hayward et al. , 1989) . This pattern may reflect individual work pref erences, personality characteristics (e.g., job satisfaction and ideational flexibility ; Kohn and Schooler, 1982; Mortimer, Finch , and Maruyama , 1988) , structural opportunities (Hayward and Grady, 1986; Kohn and Schooler , 1982; Kaufman and Spiierman, 1982), or some combination . In any case , since workers holding these jobs are less likely to retire in the first place , retirees from these positions may be more likely-through preference and opportunity-to reject the retirement role . In addition, careers also define general and firm-specific skill development, thereby affecting the older workers' marketability across employers. Persons in jobs mandating extensive firm-specific skills (e.g., managers) have few options for mobility . In contrast, workers employed in positions where they use and develop more generalizable skills (e.g., professionals, crafts workers , and service workers) are likely to have more mobility oppor tunities. Workers able to acquire these general skills may be more likely to return to work, since their skills may increase the range of jobs for which they qualify . Given older workers' preferences for partial retire ment, general skills may increase the chances of reentry through part time work. Several work-related measures included in the model are linked indi rectly to the availability and amount of retirement income. For example,.

(7) 88. HAYWARD, HARDY, AND LIU. a number of older workers in our sample were subject to mandatory retirement regulations. These regulations were designed to routinize the movement of older' workers out of the work force, a labor management policy justified by the belief that workers' productivity declined with age. Workers who were covered by mandatory retirement policies often qual ified for relatively liberal pension benefits. Given this connection, it is not surprising that older workers responded to the joint presence of man datory retirement regulations and private pension income by exiting the labor force at higher rates (Hardy, 1982; Hayward et aI., 1989). Within the context of pension plan restrictions on subsequent labor force activity, workers who retired under these circumstances also should be unlikely to reenter the labor force. Employment stability also can be linked to private pension benefits as well as to a variety of worker traits (e.g., firm loyalty, satisfactory job performance). Job mobility and retirement are tightly bound since mobility is linked to private pension eligibility and pension wealth. Indeed, private pensions can be construed as part of an implicit contract rewarding stable employment. Over time , the acquisition of firm-specific skills increases the employee's value to their present employers, a factor often reflected in the accumulation of pension wealth relative to years of service . Because these skills are developed relative to firm-specific routines, however, they may be of little value to other employers . Whether job tenure reflects pension wealth or the acquisition of firm-specific skills, it nonetheless should be negatively associated with reentry .. ANALYTICAL MODELS. This study advances our understanding of postretirement work activity in a number of ways. We portray work after retirement as a dynamic process in order to characterize the reentry experiences of a cohort of older men as they age. This approach overcomes methodological limi tations of cross-sectional studies relying on prevalence rates and the ex position of compositional differences between those in and out of the labor force (e.g., Hardy, 1982; Quinn, 1977) or differences among the everretired population (e.g., Beck , 1985,1986; Myers, 1991). By masking the timing of the actual transition, compositional differences reflect not only the differences that existed at the time of the person's departure but also any changes ill characteristics since the transition . In addition, because persons typically are not observed prior to exiting, changes in character istics leading up to the transition also are masked. We also differentiate reentry accord~ng to full-time and part-time work to examine the linkage between the preretirement career and postretire ment employment. Current details of the incidence of reentry are almost nonexistent both for the general population of older males and for smaller subgroups such as those defined by race or occupation. Through this.

(8) WORK AFTER RETIREMENT. 89. analysis, we can begin to elaborate how retirees respond to their work status changes. What factors determine whether retirees reject their re tirement role, whether they modify the economic and social terms of their retirements by supplementing their incomes with earnings and their ac tivities with part-time employment, or whether retirement proves to be a permanent exit from the labor force? Finally, we estimate a set of models allowing us to test some specific hypotheses regarding the linkage of retirement transitions and careers. If, for example, reentry to full-time employment represents an extension of career, then the process of reentry should be strongly governed by previous occupational position. If full-time, postretirement employment is not a continuation of a career trajectory, even loosely defined, then the reentry process should be dominated by proximate conditions. It should depend more on general skills, health, and financial status of the older worker than the structural position of last employment. Moreover, we would expect reentry rates to be sensitive to labor market conditions. The analysis is based on three basic models. The first model posits the overaU risk of reentry to be a function of variables indexing the deter minants previously outlined. Two additional models are estimated distin guishing reentry to full- or part-time work. We compare the parameter estimates from the latter two models to assess whether the dynamic of reentry differs primarily in terms of degree or in the structure of deter minants. DATA SET, VARIABLES, AND STATISTICAL APPROACH. Data Set and Variables. The National Longitudinal Survey of Older Men (NLS) is used to estimate the reentry models. The respondents in the original sample are representative of approximately 15 million men in the U.S. civilian non institutionalized population who were between the ages of 45 and 59 in 1966, the date of the first interview. Respondents were interviewed an nually or every 2 years until 1983. In the present study, we examine reentry for a subsample of older men who had exited the labor force at some point after exact age 55 and prior to the end of observation (N = 1936). This condition is imposed to ensure homogeneity with respect to the initial exiting process and avoid left-censoring problems. Respondents are followed from the age at which they experienced a labor force exit until one of three events: reentry (the respondent acquired employment), death, or loss to follow-up (the respondent left observation). Should those persons who reentered the labor force subsequently exit again, they rejoin the reentry risk set and are observed until an event of in.terest or censoring occurs. Because we are able to observe only the age interval within which an.

(9) 90. II. HAYWARD , HARDY, AND LIU. event occurs, the analysis is based on a file of " exposure intervals, " each of which contains information about the labor force experiences of a person during a single-year age interval. Exposure intervals are adjusted such that, on average, men are exact age x at the beginning of each interval. The file contains 6331 intervals, of which 6263 have complete covariate information. Deaths , attrition, and reentry transitions are all assumed to have occurred in the middle of an interval, and exposure is adjusted accordingly to derive central rates (Namboodiri and Suchindran, 1987).2 Analyzing exposure intervals as individual observations does not inflate the statistical significance of the models. The value of the likelihood function remains the same regardless of whether individuals (with multiple intervals) or person intervals are the units of analysis (Petersen , 1986:233). The total number of reentry events is 695. When we differentiate reentry according to destination, the total number of reentry events drops to 670 due to missing information on the criteria used to distinguish labor supply. Of these, there are 487 entries into full-time employment (working 35 h per week or more) as opposed to 183 entries into part-time positions. Given the relative infrequency of movement to part-time positions, the standard errors are larger and our description of the process determining these events is less detailed than would be possible if the estimates were based on a larger number of events. Nevertheless, as described below, the findings point to several interesting contrasts with reentry to full-time work. Each exposure-interval record includes age-specific statuses on all cov ariates, some of which change over time (e.g., health status). The cov ariates considered in the analysis are summarized in Table 1 and parallel the previous theoretical discussion. Although measurement of these cov ariates follows the general procedures .\-!sed in previous related research (e.g., Hayward et aI., 1989), one caveat is worth noting. We are not able to directly measure pension attributes such as the amounts of public and private pension benefits, and private pension eligibility . The data restrict our attention to three indicators of financial well-being: a measure of permanent family income , earnings in the preretirement job, and pension coverage . Permanent family income is measured as the average family income during the 3-year period prior to first retirement. Retirement income per se is both poorly and inconsistently measured in the NLS. This is compounded by the fact that most retirement spells are quite short for those persons who reenter, making it almost impossible to attribute income to specific labor force statuses surrounding an event. Our use of 2 Blacks were substantially oversampled in the NLS. We therefore weighted both exposure and events so that the resulting coefficients would apply to the population of older men who were 45- 55 in 1966. The analysis treats the weighted sample as a simple random sample. We applied the 1966 sample weights which were normalized to average 1.0..

(10) WORK AFTER RETIREMENT. -. 91. TABLE 1. Description of Variables Considered in Reentry Analysis. Age: The natu~al logarithm of age. Age is defined as chronological age beginning at 56, the initial pomt of the process. Spell: Duration of current OLF speIl. Number of prior OLF spells Inflation rate based on growth in the current price index for year t. Civilian unemployment rate for year t. Education: Number of years of completed schooling. (1 ':' high school or less; 2 some college; 3 = college degree or more.) Race: Nonwhite = 1; white = O. Region: Geographical region of residence. (0 = South; 1 = non-South) SMSA: Residence in an SMSA. (0 = No; 1 = Yes.) Marital Status: Marital status and wife's employment status. (0 = single; 1 = married, wife not working; 2 = married, wife working.) Disability: Physically incapable of working at the time of exiting the labor force (0 = no; 1 = yes.) Health: Does health limit amount or kind of work? Yes = 1; 0 = No . Permanent family income (in 1981 constant dollars) . This is an average of total family income in the 3-year period prior to first retirement. Double pension coverage vs Social Security only_ Class of worker: Class of worker in last job. (1 = private sector; 2 = government sector; 3 = seU employment.) No. of hours worked per week : Hours worked dur,ing survey week in last job . Hourly wage rate: Hourly wage in last job . (1981 dollars) Compulsory retirement age. Was there a compulsory retirement regulation in the worker's preretirement job and at what age? Job tenure: Number of years worked for last employer. Occupation of preretirement job: One-digit 1960 census occupational categories Substantive Complexity": A multi-item scale derived from a factor analysis of variables from the fourth edition of the Dictionary of Occupational Titles (DOT) . The scale reHects the degree to which work in the occupation involves working with data , the training require ments of the occupation , whether the work is abstract and creative , and the degree to which work is repetitive. Manipulative skill': This is a multi-item , DOT-based scale that represents the degree to which work in the occupation involves working with things and requires aptitudes such as motor coordination, finger dexterity , manual dexterity, and eye-hand-foot coordination. Physical demands' is a multi-item, DOT-based scale that indicates the physical demands and environmental hazards associated with work in an occupation. Social skill' is a multi-item, DOT-based scale that identifies wheth er work in the occupation involves dealing with people and requires the physical capabilities of talking, hearing, and seeing . • A more complete discussion of this measure is available in Hayward and Grady (1986).. permanent family income, albeit crude, nevertheless reflects the income. stream prior to retirement. We supplement this measure with a wage. measure and pension coverage variable. Wage is a rough approximation. . of both public and private pension benefit levels. Further, the pension.

(11) 92. ,I I;. HAYWARD, HARDY, AND LIU. coverage variable identifies the combination of pension benefits a person coU'ld receive either currently or at some future date. Specific comparisons are made between double pension coverage, employer-provided pension coverage only, and Social Security coverage only. Characteristics of the preretirement job include scales representing oc cupational work activities (substantive complexity, manipulative skill, physical demands, and social skill; Hayward and Grady, 1986), major occupational group (defined in terms of one-digit census occupational codes), compulsory retirement regulations (was there a compulsory re tirement regulation in the worker's preretirement job?), job tenure, class of worker (whether the person worked in a wage or salaried position Or was self-employed), the amount of hours worked during the survey week, and wage. 3 General economic conditions defining the tightness of the labor market and reductions in the purchasing power of fixed incomes are measured by the annual civilian unemployment rate for adult men and the annual rate of inflation. Individual characteristics defining personal circumstances are measured in terms of two age-dependent, life cycle attributes--current health status and financial status. Indicators of these concepts were defined above . Finally , the sociodemographic characteristics considered in the analysis are (logged) age, race, education, region, SMSA, and marital status. All measures are defined in a traditional fashion with the possible exception of marital status which is coded in combination with the wife's labor force status. Duration of the current spell out of the labor force defines the primary "time axis" along which the process takes place. Duration is measured in terms of years since exiting the labor force. Also included is a measure of the number of prior OLF (out of the labor force) spells to control for the effects of prior experience with the process. Statistical Approach The reentry models are estimated using a discrete-time hazards mod eling approach, since events are observed in I-year intervals [ao,a\), [al>a z), . . . , [ale-bale) ' Competing risk models are estimated to detect whether the structure of determinants varies according to reentry desti nation. Within this framework, reentry events are characterized as tran sition intensities (i.e. , instantaneous rates) governing individuals' expe riences. The reentry rate to a specific destination j (e.g., part-time work) 3 For persons who were self-employed, no direct measure of wage is available. We estimate wage by dividing annual earnings in the previous year by the number of weeks worked in that year and the number of hours worked during the survey week. Wage estimates for persons who were unemployed prior to exiting the labor force are obtained from the last job held..

(12) WORK AFTER RETIREMENT. 93. for the kth time interval (e.g., within 1 year subsequent to leaving the labor force) is called the conditional hazard rate (h ij) and is defined as. hij(tIX(t)). =. lim Pij(t,t + LltIX(t))/Llt.. (1). "'dO. Pij is the probability of reentry in the interval (t,t + Llt), given that the person has remained out of the labor force to time t, and conditional on a set of random variables, X( t). In the current study, X( t) is a set of explanatory variables measuring relevant features of the preretirement job and career, current circumstances, and traditional sociodemographic characteristics, some of which may change over time. Note that the time paths of the explanatory variables are assumed to follow a step function corresponding to the time points of survey observation. A semiparametric form of the baseline hazard function was chosen to minimize bias in the effects of the explanatory variables . (Covariate effects were observed to be sensitive to the imposition of a wide variety of pseudo continuous-time parametric forms of the baseline hazard function.) Du ration dependence is captured via a set of dummy vaIiables corresponding to the time intervals. Within each interval, duration dependence is as sumed to remain constant (i .e., an exponential distribution applies), but the rates are allowed to vary at will across intervals. This piece-wise exponential model is roughly equivalent to Cox regression in a continuous time framework. Specifying duration dependence in this manner is ad vantageous, since very little variation typically is left to be attributed to . unobserved heterogeneity (GaUer and Poetter, 1990) . The general reentry model estimated in the analysis thus looks like In(hjj(tIX(t))). = f3 X(t) +. alt l + a2t2 + ... + ak-Itk - I ,. (2). where f3 is a vector of parameters corresponding to the covariates in X(t), which includes a constant, and ak indicates the effect of kth time intervaL RESULTS. Descriptive Results To set the stage for the multivariate hazard model estimates, traditional life tables are calculated showing the transition rates to each type of destination and the cumulative probabilities of reentry. Duration of the OLF spell defines the time axis of the life table. Figure 1 shows the cumulative probability of reentry, where spells out of the labor force are not differentiated by prior experience. This is analogous to the total volume of reentry per unit cohort. The figure shows that the volume of reentry is approximately .33 events per person, a level achieved fairly quickly after exiting the labor force. This estimate corresponds to prior findings based on CPS data centering on the 1972-1980 period (Hayward.

(13) 94. HAYWARD, HARDY, AND LIU Cumulatlve Probability. 0.35"-------------------------------------------, 0.3 0.25 0.2 0.15 0.1. 2. 3. 5. 4. 6. 7. 8. 9. 10. Duration OLF (Years) FIG.. II. 1.. Cumulative probability of reentry among older men.. et aI., 1988a; Smith, 1982). Reentry at older ages clearly is not a rare and aberrant occurrence.. Figure 2 shows the transition rates to full-time and part-time employ. ment. The reentry rate is highest during the initial portion of the OU spell, dropping sharply during the first 2 years after leaving the labol force, although the steepness of the descent is somewhat sharper for full· time positions. Starting with the third year of the spell, the hazard rate~. Hazard Rata (h). 0.3r l ---------------------------------------------,. 0.15 0.1. 2. 3. 5. 4. 6. 7. 8. 9. 10. Duration OLF (Yaars). Reentry Destination. - Part-TIme. FIG. 2.. +. Full·TIme. *" Total. Competing risks of reentry to part-time vs full-time positions..

(14) WORK AFfER RETIREMENT. 95. TABLE 2 Mobility between Part-Time and Full-Time Work Surrounding Retirement Preretirement job Reentry job Part-time Full-time Total. ParHime. FUll-time. Total. 107 22 129. 81 468 549. 188 490 678. remain relatively unchanged at higher durations. Succinctly stated, reentry occurs quickly or not at all. Quite striking is the predominance of full-time employment as the pri mary destination. Movement to this destination exceeds movement to the alternative destinations by more than twofold during the first 2 to 3 years, at which point there is little difference due to the low overall rate of reentry. However, the overall proportion of part-time workers increases as we move from preretirement to postretirement employment distribu tions: 17% of preretirement employment was in part-time jobs compared to 33% of postretirement employment. The change in pre and postre tirement part-time employment reflects the influence of selection pro cesses, i.e., the self-selection of part-time workers for postretirement em ployment, the structure of "vacancies" available to older workers, and the movement of previously full-time workers to part-time positions. The actual level of mobility between part-time an.d full-time work sur rounding a labor force exit is shown in Table 2. The dominant tendency is that older workers reenter those types jobs held prior to exiting. Among those who held part-time postretirement jobs, 57% (107/188) worked part-time prior to exiting the labor force compared to 43% who worked full-time before exiting the labor force (81/188). Conversely, 96% of persons holding full-time postretirement jobs were employed full-time in their preretirement jobs (468/490). The pattern of part-time postretire ment employment is equally one of persistence (part-time workers reen tering through part-time employment) and labor supply reduction (full time workers reentering through part-time employment). However, the pattern of full-t,ime reentry is almost exclusively one of persistence. These patterns suggest that, once older workers accept part-time schedules, they are unlikely to move back into full-time work either because they do not desire an increase in labor supply (preference) or because they cannot secure full-time employment (underemployment) .. Multivariate Hazard Models. Table 3 reports the multivariate hazard model estimates for the general reentry process (i.e., destination is not taken into account). Model 2.

(15) 96. HAYWARD, HARDY, AND LIU TABLE 3 Hazard Model Estimates for Labor Force Reentry (N of Events = 695). Variable Intercept Age Spell 1 2 3 4. 5 6. II. I". Number of prior OLF spells Inflation Civilian unemployment rate Education • 12 years or less 13-15 years 16 years and more Race Region SMSA Marital status Married, spouse OLF Married , spouse ILF Single Disability Health Permanent family income Double pension coverage Employer-provided coverage Class of worker of preretirement job Private sector Government sector Self-employed Compulsory retirement age No compulsory retirement Compulsory at 50-64 Compulsory at 65 Compulsory at 65 + No. of hours worked per week Hourly wage rate Job tenure Occupation Professional Manager Clerical Sales Craftsmen Operative Service f~l-l'I\eI'S. (1). (2). 27.3467 (3.8814)*' -7.1930 (.9511)" 2.7138 2.2884 1.5060 .6522 .7972. (.3882)" (.3902)" (.4068)" (.4641) (.4841). Ref. 27.8031 (3.8892)*' -7.2068 (.9518)" 2.7069 2.2772 1.4932 .6374 .7825. (.3881)" (.3901)" (.4068)"" (.4641) (.4841). Ref. .0879 (.0821) .0361 (.0143)' - .1816 (.0390)". .0965 (.0824) .0342 (.0143)' - .1828 (.0388)". - .0191 (.1629). .0782 (.1973). - .1413 (.1531) .0226 (.1891). Ref. Ref. .0415 (.1486) .0094 (.0895) - .2475 (.0870)". .0585 (.1470) .0042 (.0893) - .2512 (.0867)". -.3641 (.1160)" - .1212 (.1196). -.4040 (.1154)" -.1745 (.1188). Ref. Ref. -1.0459 (.1473)': - .2841 (.0811)". -1.0661 (.1473)" - .2769 (.0809)". .DODO ( .0000). .0000 (.0000). - .1250 (.0943). - .2554 (.2772). - .1387 (.0935) - .2690 (.2759). -.62n (.1104)" - .6066 (.1503)". - .5428 (.1038)" - .5619 (.1402)". Ref. Ref. .2040 (.1363) .1452 (.3815) .0706 (.1475). .2213 (.1358) .0862 (.3796) .0716 (.1469). Ref .D019 (.D027) - .0106 (.D081) - .0116 (.D033)". .DO19 ( .D(27) - .0117 (.D082) - .0123 (.D033)··. .5655 .2870 .1341 .4892 .2115 .1574 .1653 ,,, ') <. (.2452)' (.2196) (.2601) (.2467)' (.2002) (.2066) (.2354). i 2440'. Ref.

(16) 97. WORK AFTER RETIREMENT TABLE 3-Continued Variable Farm laborers Laborers substantive complexity Manipulative skill Physical demands Social skill Log likelihood. (1). (2). .7337 (.2620)*' Ref. .0904 .0468 .0279 .0939 - 2193.3650. (.0632) (.0527) (.0479) (.0542). -2198.1840. Note . Values in parentheses a re the standard errors_ • p .;: .05 . ,. P .;: .01.. corresponds to model 1 except that four measures of occupational char acteristics are introduced in place of the standard occupational titles. Note that the occupational titles and the four occupational characteristics are related definition ally and thus cannot be included simultaneously in the same model. We maintained the detail provided by the preretirement three-digit occupations in assigning values of the occupational character istics , thereby overcoming limitations imposed by our admittedly heter ogeneous occupational groupings. Because the DOT characteristics are measured as continuous variables, their effects can be interpreted directly rather than in comparison with a reference category. Not surprising given the prior life table results, the reentry process is clearly duration dependent. In addition to estimating the risk of reentry relative to retirement duration , the coefficient for age indicates that reen try is more likely among younger retirees. All else being equal, older retirees are more likely to remain in retirement. The strength and stability of previous labor force attachment is linked to the reentry process through the negative effect of job tenure . Older workers who had remained with the same employer for a relatively long time before their retirements were less likely to reenter than workers with a more mobile work history. However, neither labor supply (i.e., hours worked in the preretirement job) nor previous spells out of the labor force were significantly related to reentry . The former result may be a function of the reemployment patterns previously noted in Table 2. The latter result suggests that, net of other factors, reentry behavior is not predominantly an extension of a previously established pattern of exits and entries. Although the effect of job tenure is net of pension coverage, we suspect it does reflect the impact of pension wealth (which is positiVely related to job tenure). Here, job stability is most likely operating as a mechanism of self selection, since workers with more stable careers have an advantage in securing claims to adequate retirement income..

(17) 98. HAYWARD, HARDY, AND LIU. None of our indicators of financial resources were related significantly to reentry; neither sources of retirement income nor level of preretirement income modified the risk of reentry. The only financial factor that did affect the process was inflation. It appears that it is not so much the sources or level of retirement income determining reentry behavior once other factors are controlled, but unanticipated changes in that income (e.g., erosion of the purchasing power). Clearly, reentry is not uniform across categories of preretirement em ployment. Reentry is more likely for previously self-employed workers compared to private and government sector workers, and among previ ously employed professionals, sales workers, and farm laborers. Self employed workers, professionals, and sales workers also display patterns of delayed retirement relative to other occupational categories (Hayward et at., 1989). Career longevity among these workers is thus due both to employment continuity in career jobs and to workers' desire and ability to reenter employment after an initial retirement (i.e., the flexibility to move from one job to another even at older ages). It is often assumed that self-employed workers are able to exert more control over their work schedules and activities (i .e ., they create their own jobs), thereby allowing a greater opportunity for them to act on their preferences for work (Fuchs, 1982; Quinn, 1980). Also pertinent to older self-employed workers' labor force decisions is the fact that this type of work brings a different structure of income constraints (e.g., a general absence of conventional second pensions and a greater reliance on asset income in old age) . Like the self employed, professionals can operate outside large bureaucratic firms and organizations . The more common career attainment process tied to the acquisition of firm-specific skills is bypassed. Instead, the connection of more generalizable skills with these tyP.~s of occupations allows career enhancement ,through interfirm mobility. In contrast to these indicators of occupational structure, our occupa tional skin measures do not affect reentry chances. The absence of effects for the DOT-derived measures of job tasks is somewhat surprising es pecially given the demonstrated relationship of substantive complexity and physical and environmental demands to the initial retirement tran sition. High levels of substantive complexity and low physical demands on the job significantly delay retirement. Apparently, the importance of these factors does not extend to work/retirement behavior beyond the initial exit. Not surprisingly, health status significantly affects reentry, with men in poor health being the least likely to return to work. This general rela tionship is reflected in the net effects of both disability and self-reported health status. These measures define a gradient of health status where disability status anchors the most severe end of the continuum; reentry becomes less and less likely as health problems mount. Married retirees.

(18) WORK AFfER RETIREMENT. 99. whose wives are out of the labor force are also unlikely to reenter em ployment, a result supportive of the view that couples attempt to coor dinate their work/retirement activities. In addition to individual traits and the classification of preretirement employment, labor market characteristics also influence the reentry pro cess. Retirees in metropolitan areas are less likely to reclaim employment, and this effect obtains even when the dampening effect of the unem ployment rate is controlled. Whether these factors operate by restricting the self-selection of retirees who seek to reenter employment, through the employers' selection of available workers or through some combination of the two, cannot be answered by this analysis. However, since retirees are more likely to reenter the labor force when competition for jobs is low, the future labor force activity rates of older workers also may be responsive to changes in the age composition of the labor force brought on by population aging.. Reentry to Part-Time vs Full-Time Positions Table 4 reports results from reentry models specifying returns to full time work (columns 1 and 2) and part-time work (columns 3 and 4).4 Estimates of the effects of previous work activity reinforce what we learned from the general model. Reentry into both part-time and full-time jobs primarily occurs within the first 2 years following retirement and is less likely for older compared to younger retirees. Workers with relatively stable preretirement employment experiences continue to be less likely to return to employment of any sort. The relationship between prere e tirement labor supply and hours worked by reentrants noted in Table 2 persists: working longer hours in a preretirement job increases the rate of reentry into full-time positions, but decreases the rate of reentry into part-time positions. Results for the indicators of financial resources do provide some ad ditional information. Retirees who are not covered by private pensions have the highest rate of reentry to full-time employment and relatively low rates of reentry to part-time employment. In contrast, retirees who are receiving only private pensions are less likely to reenter full-time employment. Once content characteristics of the preretirement job are • Although parameter estimate differences for the full- and part-time reentry models are not tested directly due to software limitations, we have attempted to assess these differences indirectly using logistic regression. This approach is based on describing the reentry process as a two stage jump process where one set of parameters determines the timing of a move and another set determines the destination given that a move occurred. Thus, our logistic regression model tells us whether a variable affects the reentry destination among persons Who reentered. Significant variables (p < .05) in these models were inflation , marital status, education, hours worked in the preretirement job, health status, pension coverage, and the substantive complexity of the preretirement job..

(19) .. r---. - - - _.. ~- --- ~ - . ........ TABLE 4 Hazard Model Estimates for Labor Force Reentry into Full-Time and Part-Time Positions Full-time. Intercept Age Spell. (2). 33.4903 (4.7371)**" - 8.8165 (1.1579)" 2.9628 2.6015 1.7228 .7641 1.0206. 1 2. 3 4 5 6 Number of prior OLF spefls Inflation Unemployment rate Education 12 years or less 13-15 years 16 years and more Race Region SMSA Marital status Married, spouse OLF Married, spouse ILF Single Disability Health Permanent family income. Part-time. (1). Variable. ....... 8. (3). 33.2001 (4 .7446)" - 8.6600 (1.1583)". (.5876)" (.5891)" (.6088)" (.6914) (.7077). 2.9773 2.6101 1.7263 .7615 1.0142. Ref. (.5874)" (.5890)" (.6087)" (.6913) (.7077). 22.5808 (7 .7607)" - 6.9686 (1.8913)" 3.5453 2.9765 2.1936 1.7677 1.6349. (1.0148)" (1.0178)" (1.0407)' (1.0710) (1.1193). Ref. Ref. (4). 25.9342 (7.7331)" -7.5630 (1.8856)" 3.4359 2.8778 2.1223 1.7229 1.6054. (1.0147)" (1.0179)" (1.0407)' (1.0711) (1.1193). Ref. - .0011 (.1145) .0042 (.0171) - .2549 (.9461)". .0026 (.1145) .0031 (.0171) - .2514 (.0460)". .1396 (.1357) .1390 (.0296)" - .0422 (.0794). .1 860 (.1370) .1315 (.0295)" - .0612 (.0786). - .2082 (.1871) .0352 (.2208). - .3884 (.1752)' - .0610 (.2106). .8737 (.4108)' .0396 (.5620). .9058 (.3820)' .0593 (.5467). Ref. Ref. Ref. Ref. .0909 (.1791) - .0671 (.1073) -.2313 (.1045)'. .0789 (.1776) .0679 (.1073) - .2105 (.1039)'. .0653 (.2748) .0952 (.1792) -.2423 (.1735). .0718 (.2708) .0290 (.1761) - .3281 (.1696). - .5744 (.1377)' , -.2324 (.1394). - .5782 (.1374)" - .2557 (.1388). .1986 (.2461) .2407 (.2647). .0400 (.2397) .0955 (.2596). Ref. Ref. Ref. -1.0132 (.1721)" - .4525 (.0985)" .0000 (.0000). -1.0169 (.1717)" - .4497 (.0984)" .0000 (.0000). Double pension coverage. - .1680. (.1141). - .1838. ( .1132). Ernolover-Drovided covc.raalt!. -. (.3508). -. (.3494). . 7373. . 7742. -1.1511 (.3116)' .2061 (.1661) - .0000 (.0000) -.0448. (.1811). _BH4R._ {.49(2). Ref. -1.2260 (.3119) .2245 (.1654) - .0000 (.0000) -.0595 (.1784) .9.52H~.-!.. ::r:. ;». >-<. ~. § ::r:. ;». ::0 tI. >-<. ;» Z. tI r'. ;::.

(20) Class o f worlc:cr. Private sector Government sector Self-employed Compulsory retirement age No compulsory retirement Compulsory at 50-64 Compulsory at 65 Compulsory at 65 + No. of hours worked per week Hourly wage rate Job tenure Occupation Professional Manager Clerical Sales Craftsmen Operative Service Farmers Farm laborers Laborers Substantive complexity Manipulative skill Physical demands Social skill Log likelihood. -0.4419 (.1368)". - .3955. -0.3511 ( .1809) Ref. - .3119 (.1694) Ref. - .6438 (.2117)·> - .8674 (.3208»> Ref. .1504 ( .1624) - .0285 (.4416) .0764 (.1729) Ref .0211 (.0030)'" - .0096 ( .0091) - .0083 (.0039)'. .6397 (.3107)"· .6485 (.7765) .5125 (.3343) Ref - .0504 (.0055)" - .0131 (.0167) - .0210 ( .0067)". .1415 ( .1629) .0442 (.4429) .0713 (.1735) Ref .0215 (.0031)"' - .0089 ( .0091). - .0079 (.0040)' .5855 (.2986)"· .1382 (.2734) .1856 (.3100) .5739 (.3025) .2383 ( .2485) .1934 (.2537) .1206 ( .2887) .1242 (.3055) .4884 (.3365) Ref. -1620.3476. • p "';. .05 .. - .43~~ ( .19~3)· - .7535 (.300'7» Ref. .6992 (.3006)"· .6285 (.7731) .5363 ( .3337) Ref - .0527 (.0056)" - .0142 (.0167) - .0221 (.0067)". .8516 (.4765)". .5233 (.3858). .2120 ( .5297). .5619 ( .4430). .4255 (.3462). .2635 ( .3676). .0521 ( .4773). .1655 (.4203). 1.0201 (.4425)'. Ref. .0308 .0705 .0311 .0399. • Values in parentheses are the standard errors . b The variable as a whole is not statistically significant •• p ",; .01.. (.1298)'·. -1624.4888. (.0767) ( .0645) (.0570) (.0648). .2859 .1484 - .0644 .1554 -748 .2407. -747.9740. (.1215)' (.1002) (.0960) (.1064).

(21) 102. 1. 1··1. I,. 1. \1. HAYWARD, HARDY, AND LIU. controlled, however, they are more likely than their counterparts to reen. ter part-time employment. Further, although changes in inflation are ir. relevant to reentry. to fulHime work, higher inflation rates speed reentry into part-time jobs. Retirees whose income is limited to private pension benefits tend to be younger than retirees drawing Social Security (either alone or in combination with pension benefits). Unlike Social Security, private pension benefits generally are not indexed for inflation. It therefore appears that much of post-retirement employment has been fueled by the increase in early retirement incentive programs allowing workers to accept' their pensions before they are entitled to Social Security benefits. When high inflation rates threaten the purchasing power of these pensions, early retirees supplement their incomes with earnings from part-time but not full-time employment. The destination-specific reentry models also reveal some interesting relationships between reentry, skills, and health status. Although not significant in the general model, retirees from positions characterized by a high degree of substantive complexity show elevated rates of reentry into part-time positions, but are neither more nor less likely to reenter full-time employment. Experience in performing these types of job tasks may affect retirees' preferences for part-time, postretirement work (Louis Harris and Associates, 1981), as well as employers' willingness to hire them. Education also affects reentry to part-time employment; less ed· ucated retirees (those with, at most, a high school diploma) are more likely to move into part-time jobs after an initial retirement. This differ ence again may reflect a se.lf-selection process: less educated retirees typ ically spent their work careers in jobs that provided lower levels of com pensation and a more modest level of retirement income. In addition, these retirees may be more likely to view jobs allowing part-time work as satisfactory. The extent to which reentry is impeded by health problems also provides an interesting contrast between the two destinations. As we would expect, retirements prompted by disability are unlikely to lead to either fuJII-time or part-time reentry. However, although rates of full-time reentry are sharply lower for retirees who report a work-limiting health condition, the presence of a health limitation does not affect rates of reentry into part-time employment. In fact, although not significant, the coefficient is actually positive. FinaUy, labor market characteristics are more relevant to reentry to full-time rather than part-time employment. Retirees in metropolitan areas have lower rates of reentry into ful:!-time employment than retirees who reside in less urban areas. In addition, reentry into full-time employment is slowed by high unemployment rates..

(22) WORK AFTER RETIREMENT. 103. DISCUSSION. The idea that the life course is increasingly being partitioned into three distinct stages suggests a fairly orderly progression from a period of train 'ng to a period of employment to a period of leisure. Further, individuals ~resumablY move from one stage to the next, while the door to the previous stage closes behind them. We therefore would expect American men to move in a rather orderly way from work to retirement. However, as relatively healthy workers have become retirees at increasingly younger ages, the actual transition from work to retirement has become more ambiguous. It is not tenable to assume that someone who retires from the labor force will never return. The evidence from this study and others is that retirement is not an absorbing state. In fact, for a growing minority of workers, there appears to be a revolving door between work and retirement (Hayward et aI., 1988a, b). Although the finding that almost one-third of retirees reenter the labor force was not "new" information, this research provides considerable detail as to how the process is structured relative to the characteristics of retirees, their preretirement jobs, and how the rate at which this behavior occurs is linked to general economic trends in inflation and unemployment and other, more individual conditions of the retirement status. What is clear from this analysis is that the flow of retirees back into the labor force varies relative to all these dimensions. In addition, our analysis underscores the importance of distinguishing among types of reentry po sitions, since the structure of the process generally differs by destination. Somewhat surprisingly, over two-thirds of retirees returning to the labor force do so via full-time work. Partial retirement therefore seems more of a dream than reality for many older workers. The longer the retirement status is occupied, the less likely the retiree will move back into employment. In fact, the vast majority of men reen tered within 1 year's time. We also observed a pattern of age dependency. At older ages, retirees are unlikely to move into positions of full-time or part-time employment. This pattern persists whenwe control for the health status of the retirees. Even among relatively healthy retirees, reentry at. older ages is distinctly less likely than reentry at younger ages. In fact, among retirees who reentered part-time positions, age was more important than health status among the nondisabled, since retirees with less profound health limitations were as likely as healthy retirees to move back into part-time employment. Although persons moving back into full-time employment essentially relinquish the retirement status, the pattern of their employment transi tions is nominally distinct from younger workers' patterns. Older workers spend the interim between jobs occupying the nonwork status of retire ment T:lthp.T th:ln l1np.rnnlnvrnpnt nr hp.incr tprnnnr:lrilv nllt nf thp hhnr.

(23) 104. I. \·t. HAYWARD, HARDY, AND LIU. force. Whether these interim statuses differ in more than name remains to be seen. Mixed support is shown for the idea that postretirement labor force behavior is linked to persons' prior positions in the labor force, and these effects appear to be somewhat different in their impact on reentry to fuB. time rather than part·time positions. The few effects of structural position observed were linked to workers' positions outside bureaucratic, firm. based labor markets. Self-employed workers, professionals, sales workers, and farm laborers aU exhibited higher rates of overall reentry compared to workers whose careers developed within bureaucratic firm-based labor markets. And, although the evidence is somewhat sketchy, professionals and sales workers appear more likely to reenter full-time, whereas self. employed and farm laborers were more oriented toward part-time em. ployment. Reentry for workers in occupationally based labor markets thus is akin to a job shift . The type of occupational skill per se has no significant effect on reentry to full· time work. Despite the effects of these factors on delaying retirement, they have no reciprocal effect on extending the primary career once interrupted by a labor force exit. The exception is substantive complexity's positive effect on reentry rates to part-time work, a finding consistent with the idea that such workers are able to capitalize on their preferences for partial retirement. When we compared the overall pattern of results for reentry to fuB· and part-time work, we observed support for the position that partial retirement is a distinct state (Reimers and Honig, 1993). The effects of marital status, pension coverage, health status, and trends in the general economy all support this general idea. Studies repeatedly have reported the linkage between early retirement behavior, health limitations, and private pensions. Here, we have another perspective: men whose early retirement is prompted by private pensions, for example, also are prone to reenter employment through part-time positions, and this tendency increases when the real value of their retirement income is jeopardized. Further, the view of retirement transitions is somewhat different from that often portrayed in the literature. Instead of persons gradually ad justing their labor supply based on a common set of factors, a more accurate characterization of the retirement process is one in which indi viduals move between discrete statuses. That is, there are discontinuities in the labor supply function of older workers. The factors determining moves between statuses are shaped by the statuses in which individuals reside, the skills they have acquired, and the characteristics of the alter native statuses. Decisions often are made sequentially, and, as events unfold, a variety of new factors come into play in determining subsequent even.ts. Thus, while exiting and reentry behaviors combine to makeup the retirement process, the structures of determinants governing moves into.

(24) WORK AfTER RETIREMENT. 105. and out of the labor force and into and out of partial retirement are quite different. Finally, the persistent effects of job tenure indicate that the reentry te to aU destinations was elevated for persons who had spent a com :arativelY short time in their previous positions. Therefore, the idea of a long and stable career with a single employer followed by full and permanent retirement is not necessarily debunked as an alternative. How ever, workers with less stable career tracks are more likely to reenter after retiring. Among the "permanent" retirees, there is considerable heterogeneity with regard to age, health, prior occupation, and financial status. It should not be forgotten that these persons constitute the majority of retirees. In this analysis, for every retiree who reentered, two did not. What this analysis cannot answer is the extent to which this statistically normative pattern of behavior is primarily due to the dynamics of labor supply (i.e., the fact that the majority of retirees are simply unavailable labor either because their health prevents it or because they're simply not interested in further employment) versus the demand for the labor of older workers. While the analysis does suggest that demand plays a ro1e in reentry to full-time work, the question of availability remains somewhat ambiguous. Our analysis has tracked successes; we know nothing of re- \ tirees who might have attempted to return but failed to find positions. The fact that successful reentrants occupy a position of retirement between employment episodes rather than a position of unemployment suggests that the status of unemployment-that status generally used to designate available labor that is not presently being utHized--does not measure availability among retirees. If the demographic projections are correct, labor force opportunities for olderwoJ:kers will increase. If, in fact, the number of retirees interested in (at least part-time) employment also is growing, the boundaries between employment, availability, and retirement will become increasingly ambiguous, since it will be more and more dif ficult to assume that this retirement is the last... REFERENCES Beck, S. H. (1985). "Determinants of labor force activity among retired men," Research on Aging 7,251-280. Beck, S. H. (1986). "Mobility from preretirement to postretirement job," Sociological Quarterly 27, 515-532. Ekerdt, D . 1. (1986). "The busy ethic: Moral continuity between work and retirement," The Gerontologist 26, 239-244. Fuchs, V. R. (1982). "Self employment and labor force participation of older males," Journal of Human Resources 17, 339-357. GaUer, H . P., and Poetter, U. (1990) . "Unobserved heterogeneity in models of unem ployment duration," in Event History Analysis in Life Course Research, (K. U . Mayer and N. B. Tuma, Eds.), pp. 226-240, The University of Wisconsin Press, Madison,. WI..

(25) 106. HAYWARD, HARDY, AND LIU. Garfinkel, S. H. (1977). "The outcome of spells of unemployment ," Monthly Labor Review 100, 54-57. Hardy, M. A . (1982) .."Social policy and determinants of retirement: A longitudinal analysis of older white males, 1969-1975," Social Forces 60, 1103-1122. Hardy, M. A. (1991). "Employment after retirement: Who gets back in? Research on Aging 13, 267-288. Harris, L., and Associates (1981), Aging in the Eighties: America in Transition, National . Council on Aging, Washington, DC. Hayward, M. D., and Grady, W. R. (1986). "The occupational retention and recruitment of older men: The influence of structural characteristics of work," Social Forces 64, 644-666. Hayward, M. D., Grady, W. R., Hardy, M. A., and Sommers, D. (1989). "Occupational influences on retirement, disability and death," Demography ~6, 393-409. Hayward, M. D., Grady, W. R . , and McLaughlin, S. D. (1988a). "Changes in the retirement process among older men in the United States: 1972-1980," Demography 25, 371-386. Hayward, M. D., Grady, W . R., and McLaughlin, S. D. (1988b). "Recent changes in mortality and labor force behavior among older Americans: Consequences for non working life expectancy," Journal of Gerontology: Social Sciences 43, S194-S199 . Honig, M., and Reimers, C. (1987). The labor market mobility of older persons , paper presented at the Annual Meeting of the Eastern Economic Association, Washington, DC. Kaufman, R . L., and Spilerman, S. (1982). "The age structures of occupations and jobs," American Journal of Sociology 87, 827-85l. Kohli, M . (1986). "Social organization and subjective construction of the life course," in Human Development and the Life Course: Multidisciplinary Perspectives, (A. B. S~ren sen, F. E . Weinert, and L. R. Sherrod, Eds.), pp. 271-292. Lawrence Erlbaum As sociates, Hillsdale , N.1. Kohn, M . L., and Schooler, C. (1982). "Job conditions and personality: A longitudinal assessment of their reciprocal effects," American Journal of Sociology 87, 1257-1286. Mayer, K. U. , and Schoepflin, U. (1989). "The state and the life course," Annual Review of Sociology 15, 187-209. Meyer, J. W. (1986) . "The self and the life course: Institutionalization and its effects," in Human Development and the Life Course: Multidisciplinary Perspectives, (A. B. S~ren sen, F. E. Weinert, and L. R. Sherrod, Eds.)"pp., 199-216, Lawrence Erlbaum As sociates, Hillsdale, N.I. Mortimer, J. T ., Finch, M. D., and Maruyama, G. (i988). "Work experience and job satisfaction: Variation by age and gender," in Work Experience and Psychological Development Through the Life Span, (J. T. Mortimer, and K. M. Borman, Eds.), pp. 109-155, Westview Press, Boulder, CO. Myers, D. A. (1991). "Work after cessation of career job," Journal of Gerontology: Social Sciences 46, S93-S102 . Namboodiri, K., and Suchindran, C. M. (1987). Life Table Techniques and Their Appli cations, Academic Press , New York. Palmore, E . B., Burchett, B. M. Fillenbaum, G. G., George, L. K., and Wallman, L. M. (1985). Retirement Causes and Consequences, Springer, New York. Parnes, H . S., and Less, L. J. (1985). "Economic well-being in retirement ," in Retirement Among American Men, (H. S. Parnes, J . E . Crowley, R. J. Haurin, L. J. Less, W. R . Morgan, F. L. Mott, and G . Nestel, Eds.), pp. 91-118, D.C. Heath, Lexington , MA. Parsons, D. (1980). "The decline in male labor force participation," Journal of Political Economy 88, 117-134. Peterson, T. (1986). "Estimating fully parametric hazard rate models with time-dependent covariates," Sociological Methods and Research 14, 219-246..

(26) WORK AFfER RETIREMENT. 107. Quinn, J . F. (1977) . "Microeconomic determi nants of early retirement : A cross-sectional view of white married men," Journal of Human Resources 12, 329-346. Quinn, J. F. (1980). " Labor-force participation patterns of older self-employed workers," Social Security Bulletin 43, 17-28. Quinn, J. F., Burkhauser, R. Y., and Myers, D. A. (1990) . Passing the Torch: The Influence of Economic Incentives on Work and Retirement, W. E . Upjohn, Institute for Em ployment Research, Kalamazoo, ML Reimers, c., and Honig, M. (1993). "The perceived budget constraint under social security: Evidence from re-entry behavior," Journal of Labor Economics 11, 184-204. Ruhm, C. J. (1989). "Career jobs, bridge employment, and retirement," in Bridges to Retirement: Older Workers in a Changing Labor Market, (P. B. Doeringer, Ed.), pp. 92-107, ILR Press, Ithaca, NY. Sandell, S. H. (1985). "Employment policy and older Americans: A framework and analysis," in Current Perspectives on Aging and the Life Course, (Z. Blau, Ed.), pp. 335-354, JAI Press, Greenwich, cr. Smith, S. J. (1982). Tables of Working Life: The Increment-Decrement Model, Bulletin 2135 , u.S. Bureau of Labor Statistics..

(27)