سوابق و پیامدهای ناشی از برنامه ریزی بازنشستگی و تصمیم گیری : مدل تجزیه و تحلیل متا

In this study, meta-analytic procedures were used to examine the relationships between retirement planning, retirement decision and their antecedent and consequences. Our review of the literature generated 341 independent samples obtained from 99 primary studies with 188,222 participants. A small effect size (ES) for antecedents of retirement planning (poor health, negative working conditions and positive attitudes toward retirement) was obtained (ranging from r = .05 to r = .19), whereas a medium ES was obtained for work involvement and job satisfaction (r = −.31 and r = −.34). Regarding retirement decision, lower effect sizes were obtained. Effect sizes for the relationships with consequences were medium for retirement planning and bridge employment (r = .28), for retirement decision-volunteer work (r = .26), and for retirement decision-retirement satisfaction (r = .26). Structural equation analysis using the pooled correlation matrix allowed us to test a more complex model. Potential moderator variables were examined, and it was found that they explained only a small percentage of variability of primary studies. Results are discussed, and theoretical and empirical implications are suggested.

The SHARE survey (The Survey of Health, Ageing, and Retirement in Europe) reveals that, in 2004, the number of retired people in Europe ranged from 34% of the participants in Spain to 66% in Austria ( Brugiavini, 2005), and these numbers will probably increase as the baby boomers leave the job market and begin to enjoy their retirement. The economic activities and retirement decisions of people over 50 years old have been the subject of contemporary debate about the reforms of the pension systems in the European continent. However, although retirement does not seem to be a phenomenon that is restricted to industrialized countries, in many developing countries there is continuity in older people’s participation in the labor world, which is a sign that retirement has not been completely institutionalized ( Szinovacz, 2003). Retirement is a complex phenomenon that involves procedural aspects related to preparation for retirement and, at the same time, particular aspects linked to the decision to retire here and now. Although retirement is frequently seen as an abrupt switch from being employed one minute to total ceasing of work activity in the next minute, evidence suggests that it is a more complex and progressive transition (Pinquart and Schindler, 2007 and Wang, 2007). However, there is an essential moment in the process of deciding to retire, which requires one to consider people’s freedom of action when deciding to retire, on the one hand, and, on the other, the push and pull factors that lead them to retire at a certain moment ( Szinovacz, 2003). A large number of primary studies have addressed the diverse factors related both to planning retirement as well as the decision to retire (i.e., Abel & Hayslip, 1986; Adams & Beehr, 1998; Adams & Rau, 2004; Alcover, Crego, & Martínez-Íñigo, 2007; Anderson & Weber, 1993; Austrom et al., 2003 and Beehr et al., 2000; Blekesaune & Solem, 2005; Butterworth et al., 2006; Buxton, Singleton, & Melzer, 2005; Charles, 1999; Chase, Eklund, & Pearson, 2003; Chiesa, 2007; Choi, 2001; Cron, Jackofsky, & Slocum, 1993; Debrand & Lengagne, 2007; Depolo et al., 2007 and DeWitt et al., 1998; Dorfman, 2002; Dosman et al., 2006, Draper et al., 1997 and Dunlop et al., 2003; Earl, 2004; Ekerdt & DeViney, 1993; Ekerdt et al., 2001, Ekerdt et al., 2000, Elovainio et al., 2003 and Evans et al., 1985; Feldman & Kim, 2000; Fernandez, Mutran; Reitzes, & Sudha, 1998; Fletcher & Hansson, 1991; Fretz et al., 1989 and Gall et al., 1997; Gee & Baillie, 1999; Gill et al., 2006; Gowan, 1998; Hanks, 1990; Hayward, Friedman, & Chen, 1998; Henkens, 1999; Henretta, Chan, & Rand, 1992; Hershey & Mowen, 2000; Honing, 1996; Hyde, Ferrie, Higgs, Mein, & Nazroo, 2004; Isaksson and Johansson, 2000 and Isaksson and Johannson, 2003; Johnson & Favreault, 2001; Keith, 1985; Kim & DeVaney, 2005; Kim & Feldman, 2000; Kim & Moen, 2002; Kosloski, Ekerdt, & DeVaney, 2001; Kremer & Harpaz, 1982; Luchak, 1997; Lund, Iversen, & Poulsen, 2001; Lund & Villadsen, 2005; Martinez, Lozano, Ancizu, Valdés, & Candenas, 2003; Mein, Martikainen, Hemingway, Stansfeld, & Marmot, 2006; Mein et al., 2000; Morrow-Howell & Leon, 1988; Pransky et al., 2005 and Reitzes et al., 1996a; Reitzes & Mutran, 2004; Richardson & Kilty, 1992; Shultz et al., 1998 and Siegrist et al., 2006; Singh & Verma, 2003; Stetz & Beehr, 2000, Taskila-Abrandt, Pukkala, Martikainen, Karjalainen, & Hietanen, 2005; Taylor & McFarlane, 1995; Theriault, 1994; Wallman, Burell, Kullman, & Svardsudd, 2004; Zaniboni, 2007), and up-to-date reviews of the literature have been conducted (Beehr and Adams, 2003 and Wilson and Palha, 2007). Nevertheless, no quantitative review has been carried out nor has any meta-analysis been published to date. Therefore, given the importance of retirement in people’s lives (Rosenkoetter & Garris, 1998) and of the political decisions that retirement patterns of the population require (Beehr & Adams, 2003), in this work: (a) a meta-analysis of the antecedents and consequences of the experience of retirement was carried out; (b) the meta-analytical correlations were integrated to test a more complex model of the results of many empirical studies; and (c) the situational factors (both conceptual and methodological) that may moderate the relations of the antecedents and consequences with the experience of retirement were explored. 1.1. Retirement: Process and decision It is not easy to provide a univocal definition of a multifaceted phenomenon like retirement. In fact, retirement can be seen as a process or an act (Beehr, 1986). There are a series of factors that lead people to think of retiring (planning or preference) and such preferences of plans have an impact on the decision to retire (retirement intention). As noted by Beehr & Adams (2003), “retirement is plural.” The first theoretical perspectives analyzed the issue placing emphasis on only one facet—the retirement decision (hereafter, RD)—and they focused on the specific moment of leaving work. Thus, they underscored the loss of roles suffered by retirees and their negative effects on personal well-being, at the same time as the palliative effects of substitutive tasks such as volunteer work or caring for relatives (O’Brien, 1981). This viewpoint was challenged by the one provided by Atchley, 1976 and Atchley, 1989 theory, which placed more emphasis on the continuity between the moments preceding and following retirement and accumulated favorable empirical evidence (Calasanti, 1996, Ekerdt et al., 1989 and Reitzes et al., 1996b). Later on, the second facet—retirement planning (hereafter, RP)—was attended and the studies stressed the existence of many pathways leading to retirement. The concept of pathways emphasizes that RD implies a long-term sequential process within the life span (Settersten, 2003). This more complex viewpoint of retirement uses four key concepts to understand personal well-being after retirement: the relation of the process with the personal context, interdependence between the vital spheres involved in retirement, the importance of considering the moments within the process, and the existence of diverse pathways or possible trajectories. Within Organizational Psychology, Beehr’s (1986) conceptualization of the process of retirement has made considerable impact. His model distinguishes various related facets: preference or RP and intention or RD, and lastly, the very act of retiring. Despite the fact that in his model, RP and RD are aspects that are doubtless related, there are theoretical reasons to assume that preferences or plans and decisions are not simply equivalent or exchangeable (Beehr, 1986). For example, some antecedents probably explain RP better than RD, in particular, if we take into account that RP—compared to switching jobs—is more accessible and more socially acceptable for older workers. In contrast, some antecedents probably explain RD better, as Beehr (1986) himself notes. This viewpoint is coherent with the long theoretical tradition about judgment and decision-making, which notes that planning the action and the making the decision are two different processes. When a problematic situation is presented and the person should plan a course of action, they begin by identifying certain salient traits of the problem, recuperating relevant information from the memory and creating a meaningful organization of such information by developing a representation of the problem that serves to reduce uncertainty (Berkeley and Humphreys, 1982, Hastie, 2001 and Pitz and Sachs, 1984). These cognitive operations are different from decision-making, although both of them are related. In order to make decisions, people analyze the costs and benefits, combine their desires or preferences and their expectations about the situation (Camerer, 1995, Luce, 2000, Savage, 1954 and von Neumann and Morgenstern, 1947), or even avoid such analysis and base their decision on prudential rules that imply moral considerations and concerns about self-control (Prelec & Herrnstein, 1991). The complexity of these processes does not allow a fixed description of them (Mellers, Schwartz, & Cooke, 1998), but it is clear that theory has revealed the distinction between planning a course of action and making a specific decision about such a course of action. Although Beehr’s model states that it is reasonable to expect that the influences on RP will be different from those on RD, when exploring the empirical research that this model generated, it is often noted that RP and RD have, to some extent, been dealt with as interchangeable concepts. This can be seen in the fact that, when the studies have operationalized retirement, they have sometimes assessed positive or negative feelings towards continuing to work, and at other times, plans to continue working. Likewise, they have measured attitudes towards retirement without establishing the appropriate distinctions (Barnes-Farrell, 2003). In the present review, in contrast, we followed Beehr’s (1986) conceptual distinction and examined RP and RD separately. A similar procedure was adopted in the only meta-analytical review, not published to date, about the antecedents of retirement (Shultz & Taylor, 2001) that uses planned retirement age for and retirement decision as criteria. 1.2. Antecedents of RP and RD The diverse theories that explored the experience of retirement have progressively extended the range of antecedents. Specifically, pioneer theories indicate that situations of disruption and loss of roles are associated with deterioration of well-being, and this deterioration is more pronounced when the roles are very relevant for personal identity. So, they propose that people who are strongly involved and satisfied with their jobs will adjust worse after retirement. The theory of continuity underlines that retirement is a process that starts before the act of retiring and that prior attitudes will have an impact on subsequent outcomes (Atchley, 1976 and Pinquart and Schindler, 2007). Despite the existence of theoretical agreement about the predictive value of such antecedents, the empirical findings are not always consistent. Adams & Beehr (1998) mention a relation of r = −.08 between work satisfaction and retirement intention, whereas Bidewell, Griffin, & Hesketh (2006) obtain a relation of r = −.20. The relation between job involvement and RD is r = −.27 according to Adams & Beehr (1998), whereas it is r = −13 according to Adams, Prescher, Beehr, & Lepisto (2002). These inconsistencies suggest that this field of investigation would benefit from the administration of meta-analytical techniques. The more complex perspective from the theoretical viewpoint—the life span approach—although it emphasizes the importance of personal circumstances and the interdependence of the spheres, does not propose specific operative hypotheses. However, other theories, such as those of control or of stress, could contribute antecedents. Specifically, the theory of control notes that situations of loss of control, such as poor health, would imply a decrease in personal well-being. Likewise, the theory of stress suggests that when one leaves a stressful role, such as adverse work conditions, one will experience relief associated with the change in the situation. Lastly, Beehr’s (1986) model indicated operatively some antecedents that could affect both the key aspects of his model, that is, RP and RD. Although it seems obvious that to have poor health will induce one to plan and decide to retire, the results are not unanimous here either. Bahrami (2001) found a weight of β = .116 of the health status on RD, whereas Negrini (2007) found a relation of r = −.23 between physical health symptoms and RP. With regard to negative work conditions, the results are also quite variable, as Bidewell and colleagues (2006) refer to a relation of r = .37 between negative conditions and RD, but Lim & Feldman (2003) only find a value of r = −.11. Again, the inconsistencies indicate the appropriateness of applying meta-analytical techniques to these primary studies. In an attempt to classify these predictors, in the present review, we selected three categories of antecedents of RP and RD: poor health, psychosocial factors (positive attitudes towards retirement, job satisfaction, and work involvement) and organizational factors (negative work conditions), partially following the suggestions of Shultz & Taylor (2001). Demographic factors, such as age or gender, were considered potential moderators that could explain the variability of the results. Based on the literature reviewed, we proposed a first set of hypotheses for this meta-analysis. In general, regarding antecedents, we hypothesized that: (a) Poor health, negative work conditions, and positive attitudes towards retirement are positively related both to RP and RD (Hypothesis 1). (b) Job satisfaction and work involvement are negatively related both to RP and RD (Hypothesis 2). 1.3. Consequences of retirement People’s adjustment to retirement has been a concern for social investigators. Although some theoretical models (i.e., Atchley, 1974 and Atchley, 1976) were formulated, empirical research is highly dispersed, and there are many partial studies focusing on the analysis of one or a few variables, which hinders generalization of their results. Despite this, there are many empirical studies of retirees’ well-being. The theoretical approaches that refer to role change associated with retirement also resort to these roles to explain subsequent adjustment. According to the continuity theory, the accessibility of alternative roles could be a desirable situation in retirement because it would reduce the negative impact on well-being (Atchley, 1989). So, RD may often be associated with searching for bridge employment or with the desire to become involved in volunteer activities. Such activities, to a certain extent a substitution of the work role, allow one to avoid the loss of social contacts and enhance maintaining personal well-being. Despite clear theoretical justifications, there are noticeable discrepancies in the empirical results. Whereas Davis (2003) reports a relation of r = −.18 between RD and participation in bridge employment, Kim & Feldman (1998) find a relation of r = .18 between the same variables. The most direct indicator in research of well-being has been health after retirement and here, too, the results are not unanimous. Herzog, House, & Morgan (1991) find a weight of β = −.28 of RD on physical illness, the studies of Gallo and collaborators find significant differences in mental health, physical functioning, alcohol consumption, and depression between retirees and workers ( Gallo et al., 2000, Gallo et al., 2001 and Gallo et al., 2006). When considering the broader results of retirement, most authors resort to life satisfaction or satisfaction with specific aspects of retirement (Floyd et al., 1992). Empirical research on the association between retirement and life satisfaction is contradictory (Pinquart & Schindler, 2007). Some studies (Herzog et al., 1991) mention significant regression coefficients of retirement on life satisfaction (B = .26, p < .01) whereas other studies ( Schmitt, Coyle, Rauschenberger, & White, 1979) found no differences in retirees compared to people who continue to work (M = 22.5 and M = 22.13, for retirees and non-retirees, respectively), or others ( Bell, 1978) even mention a significant decline in life satisfaction after retirement (t = −3.06, p < .01). More precise measures, such as satisfaction with retirement, do not provide more consistent results, because some studies ( Desmette, Gaillard, & Lienard, 2004) find high levels of relationship between RD and subsequent satisfaction (r = .33), whereas other results (Topa, Moriano, Depolo, & Morales, 2007) are not so promising (r = .06). In addition to the relations between antecedents and consequences through RP and RD, we must also be aware of direct relations among factors such as poor health experienced whilst working and illness or satisfaction after retirement, as has been repeatedly suggested (Beehr, 1986 and Taylor and Doverspike, 2003). The pattern of relations revealed in the literature is complex and there are many and sometimes inconsistent findings, so we consider the use of meta-analytical techniques promising for the review of this field of research. Based on the literature reviewed, we proposed a second set of hypotheses for this meta-analysis. In general, regarding retirement consequences, we proposed that: (a) RP is positively related to volunteer work, bridge employment, life satisfaction, and retirement satisfaction (Hypothesis 3). (b) RP is negatively related to mental/physical illness (Hypothesis 4). (c) RD is positively related to volunteer work, bridge employment, life satisfaction, and retirement satisfaction (Hypothesis 5). (d) RD is negatively related to mental/physical illness (Hypothesis 6). In an attempt to clarify this complex pattern of relationships between antecedents and consequences of RP and RD, in the first stage, we develop a theoretical model and, in the second stage, we perform a structural equation model (SEM) analysis to test all the relationships simultaneously using the pooled correlation matrix as input. However, we acknowledge that SEM cannot completely explain the complex causal relations between concepts and, therefore, longitudinal studies are needed to help clarify these causal relations and issues. The hypothesized model of relations of the antecedents and consequences is displayed in Fig. 1.1.4. Potential moderator variables The most recent theoretical reviews note that there are many indications that the relations of retirement with its antecedents and consequences vary among social groups (Szinovacz, 2003). In his model, Beehr (1986) suggested diverse moderators, such as planning, retirement expectations, economic status, health, type A behavior, and type of job. Despite the fact that this wide range of variables would be optimal, unfortunately, the lack of information provided by primary studies does not allow us to include them in our moderator analyses. As a result, we followed Lipsey & Wilson’s (2001) suggestions for moderator variables. We first considered the substantive aspects of the studies. In this sense, we selected the characteristics of the people in the sample (workers or retirees), and age and gender as variables involved in the retirement phenomenon. Then, we took into account source descriptors, including the factors involved in the general study context, such as origin of the sample and year of publication. The rationale of this decision was that these variables are often proxies for substantive or methodological variables that might not otherwise be reported and included in the moderator analyses (Lipsey & Wilson, 2001, p. 83). Based on these recommendations, we considered a set of mediator variables that is listed below. Firstly, it must be taken into account that retirement involves transition, which means that understanding the temporal process is crucial (Zickar & Gibby, 2003). This implies a series of methodological difficulties in the primary studies, which should be considered in the quantitative reviews. Many cross-sectional studies are based on the comparison of groups of older workers, near retirement, with other workers who have already retired. In this case, retirement for those who are still working is an expectation, a construct related to the future, whereas for the retirees, their job satisfaction or their previous poor health are retrospective. Therefore, it is reasonable to expect differences among the studies depending on the type of participants from whom the data was obtained (Pinquart & Schindler, 2007). Secondly, there is a strong influence of social and normative factors on the phenomenon of retirement, and on its antecedents and consequences. Beehr & Adams (2003) report that whereas retirement is firmly established in industrialized countries, retirement patterns in rural areas or developing countries are very diverse. Even in developed countries, there is great diversity in the regulation of the complex phenomenon of retirement. For example, whereas in the United States of America, rules to have right to a disability pension are very demanding, but some European countries offer generous incentives for people with reduced capacity and specific programs for workers’ gradual retirement. Whereas in some regions of the European Union, continuity in part-time employment is seen as an indicator of well-being, in Asian countries, it is discouraged, although it may be economically necessary, because the social norms would blame the family that did not take proper care of the older person. This may be why empirical research has detected notable discrepancies among the results of studies from diverse countries, so the origin of the participants can be expected to be a potential moderator of variability among the studies (Hershey, Henkens, & Van Dalen, 2007). Thirdly, age is always considered a good predictor of RD, and, at the same time, it is a demographic factor that interacts with the rest of the antecedents and consequences (Pinquart & Schindler, 2007). Despite unanimousness at the theoretical level, the results show that sometimes retirement before the age of 65 is associated with less satisfaction (Palmore, Fillenbaum, & George, 1984) and, in some cases, retirees who are quite elderly are less satisfied (McGoldrick & Cooper, 1994). Fourthly, the theories also admit that differences between men and women in the pattern of antecedents and consequences of retirement are to be expected, especially because of the impact of gender roles on couples’ retirement patterns (Szinovacz, 2003). If the husbands are still considered the main providers of family income, the patterns of retirement will be affected by this image, so that the women will either prefer to retire before their husbands, or else both partners will retire at the same time. But as the decision is also conditioned by economical factors, this can lead to retirement at different moments, which seems to wind up affecting well-being. However, empirical research on gender and retirement does not always reveal consistent results but instead diverse models have been verified of the relations between predictors and results as a function of gender (Calasanti, 1996 and Quick and Moen, 1998). Fifthly, some reviews have indicated that the time gone by since the empirical study was carried out could influence the explanation of the results. This could be due to diverse factors, some of a methodological nature, because the quality of empirical studies and the rigor in the presentation of the results may have increased over time. It could also be due to sociological factors, because retirement is a practice that has become consolidated in the last few decades in industrialized countries, so that the oldest studies might reflect results that contradict the more recent ones, at least, with regard to participants’ expectations. In this sense, the publication date of the study is proposed as a potential moderating variable of the results. Based on literature review, we proposed a third set of hypotheses regarding potential moderator variables. (a) RP in a sample of workers has a stronger relationship both with antecedents and consequences than RP in a sample of retirees (Hypothesis 7). (b) RD in a sample of retirees has a stronger relationship with both antecedents and consequences than RD in a sample of workers (Hypothesis 8). (c) RD has a stronger relationship both with antecedents and consequences among American than among European samples (Hypothesis 9). (d) RD and RP relationships with antecedents and consequences are affected by participants’ age and gender, and year of publication of the study (Hypothesis 10).

The main goal of this review was to synthesize the results of the empirical studies of antecedents and consequences of retirement and to test a more complex theoretical model from the pooled correlation matrix. In order to do this, firstly, we performed several separate meta-analyses of RP and RD on the one hand, and their antecedents and consequences, on the other. Secondly, we constructed a meta-analytical correlation matrix used as input to test the proposed SEM. Lastly, we attempted to provide some explanation of the variability of the studies from potential moderating variables. Taking the work of Shultz & Taylor (2001) as a starting point, we made a larger scale meta-analytic research on the topic. Despite the limitations of this work, as we shall comment on below, we believe that the present review makes several contributions to research of retirement and its effects on people. Firstly, with regard to the antecedents of retirement, in the meta-analysis, we confirmed the pattern of relations predicted. So, the antecedents predict RP more precisely than RD. This confirms Beehr’s (1986) suggestion, according to which, these phenomena are not exchangeable, despite being related to each other. It also confirms his prediction of the higher explanatory power of antecedents for RP than for RD. These results reinforce the fact that people in general perceive little control over RD, or, in other words, RD is, to a good extent, “forced.” A more detailed examination shows that all the antecedents have closer relations with planning than with the final decision to retire, except for poor initial health. This finding confirms the data of the older qualitative reviews (Kasl, 1980 and Minkler, 1981) and is coherent with the fact that deterioration of health before retirement can be a factor that reduces the possibilities of RP and, at the same time, precipitates RD. Certainly, as Szinovacz stated (2003), the relationship between health and retirement is complex. On the one hand, health problems can cause premature retirement as a consequence of long-term illness and can undermine retirees’ well-being. On the other hand, retirement from unhealthy jobs may promote early retirement and increased well-being for retirees. Moreover, a methodological problem could affect the meta-analytical results due to the fact that health problems can be viewed as consisting of three different levels, ranging from major physical illness to relatively mild psychosomatic illnesses, as Feldman (1994) stated. But, only overall measures of health are predominant in the studies included herein. As a result, we could not make a clear distinction of Feldman’s levels, and a clear understanding of the health-retirement relationship has not been provided. Job satisfaction is the best predictor of RP (in a negative sense), confirming the results of a previous quantitative review (Shultz & Taylor, 2001). At the same time, job satisfaction was the worst predictor of RD, and this fact is noteworthy. This result may be affected by the compulsory nature of retirement age in many western societies, where RD could be due to a legal command rather than a real personal decision. Secondly, with regard to the consequences, we could not test all the relations due to the lack of sufficient primary studies in some categories, but we found significant relations for RP-bridge employment, and RP-retirement satisfaction. In this sense, Taylor & Doverspike (2003) stated that planning eases the transition into retirement through two processes. First, RP allows the person to develop realistic expectations about retirement, and these expectations are more likely to be confirmed. Second, RP may facilitate goal setting, such as engaging in bridge employment. Moreover, clarity of goals may improve specific financial planning in the long term which, in turn, may affect satisfaction, at least about the financial aspect of retired life. Clearly, the most relevant consequences of RD were retirement satisfaction, volunteerism, and bridge employment. The other two relations that were meta-analyzed with RD, on the one hand, and with life satisfaction and illness, on the other, did not show relevant ESs. When the results are analyzed in detail, the strongest patterns of relations are verified between RP and bridge employment, and between RD and retirement satisfaction. Undoubtedly, the predisposition to think about and anticipate retirement is likely to lead older persons to engage in bridge employment more than if there is no reflection and no resources, and RD is produced abruptly or uncontrollably, for example, because of one’s own illness or that of one’s partner. In this sense, our results confirmed the findings of some excluded studies, based on large samples, which showed a strong pattern of the relations between quality of bridge employment and adjustment to retirement (Topa et al., 2009 and Topa et al., submitted for publication). In contrast, precisely because retirement is a process, the influence of RP on subsequent well-being also decreases over time and therefore, one could expect RD to have a stronger relation with retirement satisfaction than mere RP (Taylor & Doverspike, 2003). This is also verified by the results of the meta-analysis through the global ES of both aspects of retirement on satisfaction after retirement. Thirdly, in order to test all the relations at once, in this study, we fit a SEM using the pooled correlation matrix as input. Despite the fact that some relations were based on a limited number of studies, this procedure allowed us to reach more essential conclusions than if it were based on the findings on a single empirical study. Some concerns have been expressed about their meaning, specifically in the sense that these models seemed to bear little resemblance to what was suggested by prior theory and instead may represent statistically derived models. In contrast, this model can be seen as a research tool, which allows us to discuss overall relationships between constructs and to achieve a better understanding of the retirement process. Therefore, we offer a briefly comment on its meaning. First, we acknowledge that data fit has been mainly obtained by dropping and adding paths from the original model. Although final models should be always consistent primarily with theory, it is not uncommon that in meta-analysis, data do not allow a perfect fit. Taking into account that statistical fit can never be a substitute for theoretical soundness, our final model seems to be quite consistent with theory. Second, the data show that the proposed antecedents really are related to the consequences, according to the hypothesized conceptual model. The data show, however, that some “specialized” paths emerged. On the one hand, RP has primarily been predicted by specific job-related attitudes (e.g., job satisfaction and work involvement), whereas RD shows additional influence of poor health and negative working conditions, which, however, does not affect RP. In this sense, it would be interesting to consider that, even in the absence of previous RP, the factors that one could call hard (health, working conditions, wealth) could precipitate RD. Meanwhile, a medium-level covariance has been observed between RP and RD, which would suggest that retirement has been included as a step in a self-managed career. As Sterns & Kaplan (2003, p. 188) acknowledged, “Similar to work, retirement has moved into the realm of self-management.” Our contention is that, if the individual is in charge of his or her career, RD does not actually exist without some form of RP. On the other hand, a similar “specialization” appears for consequences. RP appears to predict all the consequences except retirement satisfaction. In turn, RD shows the sole significant relationship with retirement satisfaction (besides the one with mental/physical illness). Moreover, when it was tested with SEM, RP showed a greater impact on consequences than did RD. This suggests that the increased focus on self-determination may lead individuals to perceive RP as a personal decision, whereas RD is seen as an imposition. Beyond this, up till now, there has been much debate about the pernicious influence of retirement on physical and mental health, which individual studies did not clarify. In our SEM, it seems clear that there is no close relationship between RD and subsequent illness, but instead, the lack of RP is a consistent and negative predictor of subsequent illness. To summarize, the final model: (a) confirms the overall pattern between antecedents and consequences, showing some “specialized” role for RP and RD; (b) suggests the multifaceted nature of the RD process, which is influenced by a large number of variables, mainly related to the work context; (c) also suggests that consequences are more directly affected by RP than by RD. Although such a pattern should be tested by means of a longitudinal study (Beehr and Adams, 2003 and Zickar and Gibby, 2003), it seems to show that planning and decision activities are far removed from being a structured and rationale process. Our data are not robust enough to allow a definitive model to be proposed. However, given the unavoidable limitations that are typical of meta-analysis, we believe that such a model can be used to confirm the centrality both of the planning and the decision processes in retirement, because RP and RD have been proven capable of describing and explaining a large part of the interwoven links between the antecedents and consequences of retirement. Fourthly, the influence of the moderators was analyzed, but this analysis yields rather unclear results. On the one hand, although differences between the types of participants have been verified, the ESs found in each one of the categories do not follow the predicted pattern. This may be due to methodological reasons, because in diverse studies included in the review, the description of the participant sample is insufficient and, as a consequence, so is the categorization for the meta-analysis. In this sense, we suggest that more methodological precision is needed when operationalizing the characteristics of the samples in the primary studies, because this contributes precision to the conclusions. As Barnell-Farrell (2003, p. 182) suggested, “retirement is a personal decision that has quite different ramifications for people from different generations, different gender role sets, different cultures, and different socioeconomic contexts.” Despite our efforts, more research in this direction is needed. This result also seems to confirm the warning of Pinquart & Schindler (2007) about the possible interference of third variables in the relations between RD and subsequent results. Moreover, it should be remembered that retired people are not a homogeneous group and some of the characteristics that differentiate them, such as how long they have been retired, should be taken into account when planning group comparisons. This might explain the inconsistency of the patterns of differences between retirees and workers that are found in the analysis of moderators in this review. Despite the methodological issue of the sample comparisons, it should also be taken into account that retirement is not the same when it is proposed to people who are still working as to those who are already retired. In the former case, expectations and perceptions of the future are being investigated, whereas in the latter, we are referring to memories and retrospective perceptions, on which subsequent experiences have had an impact, which can bias such memories negatively or positively. Another explanation for this methodological problem is that people for whom retirement is further away view it as a response to job conditions. On the contrary, people who are close to retirement “envision retirement primarily in terms of desirability/undesirability of the new role,” as Barnell-Farrell (2003, p. 173) noted. In any case, the utility of making simple comparisons between employees and retirees is questionable. In contrast, there is no consistent pattern of results for the origin of the sample. Once again, the differences are statistically significant form some analyses but the ESs do not follow the pattern predicted. This may be due, on the one hand, to the existence of diverse factors observed in the same society at different temporal moments, such as the normative variations regarding the expected retirement age. For example, at a time of economic bonanza, older people may be encouraged to remain employed, whereas in times of high unemployment, they may be considered to be depriving the youth of work. Thus, within the same category, the variability of the results could even be due to the time at which the study was carried out. However, the categories are very heterogeneous because the social and pension systems are not uniform, either in the formal or in the informal aspects, not even in the European Union. There are even differences between professional groups in the same country or between the public and the private sectors, which adds still more heterogeneity. This could also be considered an example of social factors indirectly affecting subsequent well-being, through beliefs about the appropriate retirement age and the discrimination of older employees, which affects the moment of RD and subsequent well-being. Finally, some of the conflicting findings regarding this moderator analysis may be a function of the inadequate attention that has been paid to a valid cross-cultural definition of retirement, specifically when the purpose is to specify its occurrence in non-Western, non-industrial societies (Luborsky & LeBlanc, 2003). As a result, the studies that included participants from non-Western societies should be not compared to those of Western societies without considering these methodological issues. In the fifth place, we tested a regression model in which three continuous variables, age, sex, and the year of publication or of conducting the study were used as moderators. In this case, the results yielded a pattern that follows the prediction, as sex, taken alone or with other variables, explained an important percentage of the variability in the studies of various cases. In some analyses, in contrast, the most powerful predictor of variability was the year of publication. These results provide meta-analytical support to the statements from diverse models, both psychosocial and economical, which refer to many reasons for sex to have an impact on subsequent adjustment to retirement. The differentiated social roles of men and women lead the latter to a discontinuous work trajectory, where factors such as the need to care for elderly relatives, their partner, or grandchildren affect their decision to retire. In turn, the impact of gender on subsequent adjustment may be mediated by women’s ease to develop work activities (bridge employment or volunteering) simultaneously with domestic responsibilities, taking into account their greater experience in the management of simultaneous work and domestic responsibilities. Moreover, the impact of discontinuous and shorter working careers may be noticeable in a specific feature of retirement satisfaction, economic income, because it is a proven fact that women’s pensions are lower than men’s. This outcome is consistent with the statement that gender differences are revealed “in the extent to which retirement decisions are primarily driven by the pull influence of opportunities to carry out the family role responsibilities versus the push influences of negative affect toward job and the organization” ( Barnes-Farrell, 2003, p. 173). Lastly, the explanatory weight of the year of publication of the study may be concealing other methodological variables, such as the quality of the investigation reported, or the percentage of working women in the sample. It is reasonable to assume that, if the study is older, the percentage of female participants who had been in the labor market for a long time would be lower. It is also to be expected that, due the progressive implementation of the pension systems, especially in developing countries, participants’ expectations and plans as reported in the studies carried out during the decade of the 1970s will be different from the contemporary ones. As the results are not completely clear, it is advisable to continue to explore in this direction. It would also be appropriate to add to this regression equation other variables we could not categorize from the primary studies, such as previous occupational level, the kind of work carried out, or job tenure in the organization. As in the previous unpublished meta-analysis of Shultz & Taylor (2001), the hypotheses about these moderators could not be tested. Once again, we recommend authors of the primary studies to provide exhaustive information about these variables in their original works.