‘What could make me stay at work’: Retirement transition profiles

Aging populations pose a persistent challenge to the sustainability of public pension systems. To tackle these financial pressures, many countries strengthen the incentives to work by increasing the statutory retirement age and reducing early retirement benefits. These policy reforms make retirement a topic of utmost importance for individuals, organizations, and societies. Although retirement predictors are already a widely studied topic in the literature, there is still much to investigate about why people decide to retire when they do. In particular, the role of work-related variables in the retirement decision-making process is still not fully understood. Thus, the aim of this study was to examine how individual and work factors influence retirement timing (early, on-time, and later retirement). Forty-one interviews were conducted, and data were subjected to content analysis. The inter-relationship between the multiple categories was analyzed by a Multiple Correspondence Analysis (MCA) combined with Cluster Analysis. Results revealed three distinct profiles, which allowed us to group the participants into three clusters. The stay factors profile (e.g., high positive experiences at work, having no dependents, the spouse/partner not being retired) was associated with later retirement. These results can be important for organizations that want and need to retain the best senior talents, by acknowledging that positive experiences at work are associated with older workers’ desire of postponing retirement.


Introduction
Retirement is a phenomenon widely studied by various scientific disciplines, including psychology, sociology, economics, gerontology, and health sciences. Retirement research has received a boost of scientific attention in the last decade as a result of demographic changes, technological developments, and economic and financial crises (Sargent et al., 2013;Wang & Shi, 2014).
With the institutionalization of retirement in the midtwentieth century, public and private pension schemes became common, and the statutory age of retirement was defined in Western industrialized countries (Phillipson, 2013). Later, with the rise of mass unemployment following the 1970s oil crisis, early retirement became a social trend that "resulted from unintended policies and collective responses to the new socioeconomic pressures" (Ebbinghaus, 2001: p. 76). The efforts of governments to reduce labor supply led to improvements in public pension systems (e.g., broadened eligibility for pensions before the normal retirement age) and other welfare programs (e.g., unemployment insurance), which contributed to decreasing employment rates of older male workers (55-65 years old) (Guillemard & Rein, 1993;Hofäcker et al., 2015). Also, the economic downturn led organizations to increase productivity and lower costs by restructuring and downsizing, which had a direct effect on older workers (e.g., early retirement incentives) (Ebbinghaus, 2001). Together, these measures conveyed the idea to older workers that early retirement was less costly and more attractive than continuing to work.
At the end of the twentieth century, policy makers in industrialized countries became aware of the problems created by this policy of early retirement. In times of aging populations and threats to the sustainability of public finances, many countries adopted policies to support an extended working life, such as the substantial rise in public pension eligibility ages and increasing financial penalties for early retirement  Organisation for Economic Co-operation and Development [OECD], 2017). This change of mindset from early retirement to active aging materialized in the promotion of increasing employment of older workers, full participation in society, and healthy, independent and dignified living of older people (Duchemin et al., 2012;Hess, 2017). Hence, within a short historical period, the nature of retirement has changed in terms of both meaning and timing, and it is likely that the future will bring more changes (Henkens et al., 2018;Phillipson, 2013;Sargent et al., 2013). As a result, the concept of retirement itself has become more unpredictable and difficult to define.
Despite its many definitions, retirement can be conceptualized as a decision-making process that occurs over time and that involves a number of reflections and decisions about when and how people will enroll in retirement (Beehr, 1986;Moen, 2012). Research on retirement has been fruitful showing that, in this process, there is a combination of individual (e.g., age, gender, education, work centrality) and situational (e.g., eligibility, family commitments, organizational policies) factors that interact and shape an individual's transition to retirement (De Preter et al., 2013;Fisher et al., 2016;Flynn, 2010a;Scharn et al., 2018).
Despite these important advances, studies are often based on survey data (Kojola & Moen, 2016;Robertson, 2000). These studies, especially those with a longitudinal design and large samples (e.g., English Longitudinal Study of Ageing (ELSA), Health and Retirement Study (HRS)) and cross-national databases (e.g., Survey of Health, Ageing and Retirement in Europe (SHARE)), have provided useful insights regarding how people plan, prepare, decide about, and adjust to retirement. However, given the dynamic and complex nature of retirement, quantitative research does not capture the quality of individual experiences as these decisions do not always fit into dichotomous and finite categories, "particularly when categories are based on out of date and evolving conceptualizations of retirement" (Kojola & Moen, 2016, p. 61). Retirement decisions stem from the interaction between multiple and dynamic factors whose importance varies from person to person as a result of increasingly diverse career pathways. This variability makes this decision-making process even more complex and unpredictable.
Thus, previous work has not been able to fully capture the heterogeneity in retirement transitions by not considering the interaction between individual and situational predictors, as well as their importance for different people. The purpose of this study is therefore to examine how individual and work factors influence people's retirement timing in a more comprehensive way. For that, semi-structured interviews were conducted with older workers (55 years old or over) and retirees who were retired for fewer than five years. We further explore how individual aspects (e.g., health, financial satisfaction) are associated with work-related variables (e.g., relationships at work, job satisfaction) in the definition of retirement transition profiles. Moreover, we investigate how these profiles are associated with retirement timing (i.e., early, on-time, and later retirement).
The contribution of this paper is threefold. First, we examine the interdependence of individual and work-related factors to shape the retirement decision-making process. By applying Multiple Correspondence Analysis (MCA) to the complex and multidimensional process that is retirement, we can better understand the many relationships that exist among a high number of (categorical) indicators that were included into retirement transition profiles. These profiles stress the multidimensionality and complexity of retirement. Second, we performed Cluster Analysis to define groups of older individuals in terms of their retirement transition profiles, showing the heterogeneity in retirement. These clusters were associated with different retirement timings, which demonstrated the utility of organizations to implement different retention strategies tailored to different groups. To the best of our knowledge, the combination of these statistical multivariate methods has not been used in studies concerning retirement timing.
Third, we consider the Portuguese context specifically since, with a few exceptions (e.g., Fonseca & Paúl, 2004), there is a lack of research about the retirement transition of Portuguese workers. In 2007, the Portuguese government introduced the sustainability factor in the calculation of the statutory age for receiving an old-age pension, linking it to changes in the life expectancy (Decree-Law no. 187/2007;Decree-Law no. 167-E/2013). These reforms represent a change in older workers' expectations about the length of their careers as they will have to work considerably longer than they had earlier planned or anticipated. Thus, considering the specific legal aspects of retirement in Portugal, this study provides a deeper understanding of why individuals prefer to exit the labor force before, at, or after the statutory retirement age.

Theoretical Background
The transition from work to retirement has attracted considerable interest for many years and empirical research has identified numerous antecedents of this transition. Research has shown that retirement decisions depend upon a wide range of individual, organizational, and social aspects: demographic characteristics, financial status, health, family context, workrelated variables, and the socioeconomic and political context (Browne et al., 2019;Flynn, 2010aFlynn, , 2010bMadero-Cabib et al., 2016;von Bonsdorff et al., 2010;Wang & Shultz, 2010).
Further, retirement is a dynamic process that occurs over time, rather than a discrete event (Beehr, 1986). The day people actually retire is preceded by planning, preparation, and a decision about the best timing to retire, considering the individual's specific characteristics or circumstances. Although some individuals' lives and careers seem similar, the life course perspective argues that there are events and situations in the past that influence workers' unique intentions and future decisions about retirement (Beehr, 2014;Moen, 1996). Particularly, researchers have demonstrated the need to include both individual and work-related variables in the analysis of this decision-making process (Fisher et al., 2016;Furunes et al., 2015;Wang & Shultz, 2010).

Individual Factors
At the individual level, finances and health are the most studied predictors of retirement intentions and decisions, and its consequences (Barbosa et al., 2016;Beehr, 2014;van Rijn et al., 2014). The well-established influence of the economic aspects in retirement can be explained by the rationaleconomic approach: Individuals estimate what will be their monthly income as a retiree, according to their pensions and savings, evaluate if this income covers the expenses, and if so they decide to retire (Feldman & Beehr, 2011;Topa et al., 2011). Thus, the assumption is that workers cannot retire until they can afford to do so financially, regardless of how they feel about work or retirement. For instance, an important reason for working beyond the normal retirement age is the individual's financial situation (Burkert & Hochfellner, 2017;Fasbender et al., 2016;Sewdas et al., 2017). In the current socioeconomic context after the recession, which was characterized by cuts in benefits, tax increases, and higher cost-ofliving (Baker & Rosnick, 2011;Szinovacz et al., 2014), it is crucial to understand how financial aspects influence retirement expectations.
Also, one of the leading factors predicting retirement timing is health (Wang & Shi, 2014). Health and retirement have a complex relationship since previous research demonstrated that health can be either a push or a pull factor (Oksanen & Virtanen, 2012;Shultz et al., 1998;Topa et al., 2009). On the one hand, deterioration of health pushes individuals to retirement because they perceive that their work ability is impaired, they are afraid of further erosion of health, and retirement is an occasion to rest and recover (de Wind et al., 2013;Oksanen & Virtanen, 2012). On the other hand, healthy individuals may feel that retirement is an opportunity to enjoy life outside the work context, which pulls them toward retirement (de Wind et al., 2013;Shultz et al., 1998). Along with health, work ability has been consistently identified as an important antecedent of retirement (Boissonneault & de Beer, 2018;Cadiz et al., 2019;de Wind et al., 2015). Work ability can be conceptualized as an individual's perception of his or her ability (e.g., physical, mental resources) to meet the job demands (e.g., emotional, cognitive) (Cadiz et al., 2019;McGonagle et al., 2015). Individuals who report low work ability are more likely to exit from paid employment through early retirement as they feel that their resources are not enough to respond to the demands posed by work (Boissonneault & de Beer, 2018;Sell, 2009).
However, individuals who anticipate financial security in retirement and enjoy good health do not always choose to retire at the statutory retirement age. In fact, even when individuals meet all the requirements to receive a pension, some people decide to continue working because there are other individual variables (e.g., family context, meaning of work) and work-related characteristics (e.g., social support, challenge at work) that motivate them to do so. For instance, individuals who experience personal meaning at work (i.e., have meaningful and satisfying tasks) are more likely to engage in post-retirement employment (Fasbender et al., 2016). Also, people who experience social loss in retirement (e.g., loss of social contacts, feelings of being less needed and less respected) see work as an opportunity to fulfill their social needs, leading them to engage in post-retirement employment (Fasbender et al., 2014). Thus, the option for continuing to work beyond the moment retirement becomes possible is based on different types of motivations.
Family situation is one of the aspects of a worker's personal life that influence retirement timing. Previous research on this topic considers the caregiver role and the employment status of the spouse or partner as important predictors of an individual's intention of anticipating or postponing retirement (Kubicek et al., 2010;Szinovacz et al., 2001). If a spouse, partner, or family member needs to be taken care of, individuals might decide to retire earlier to have more flexibility to assume this role (Kubicek et al., 2010). Also, many couples plan their careers to retire at about the same time, even if they differ in age (Gustafson, 2018). Moreover, people might face retirement as a period to enjoy more time with family if relatives do not have work or school commitments.

Work-Related Factors
Research about the potential influence of the work situation on workers' decisions to retire has received growing attention since the study of Beehr et al. (2000). The modifiable nature of some aspects of the work context brought to light the role of organizations in promoting extended working lives (Browne et al., 2019;Carr et al., 2016). For example, past research found that increasing mental and physical demands, adverse working conditions, low recognition, and low support were associated with early retirement (Böckerman & Ilmakunnas, 2019;Carr et al., 2016;Dal Bianco et al., 2015). Also, leaders can be drivers for extending careers by providing support to their teams (Furunes et al., 2015). For instance, Wöhrmann et al. (2017) found that respectful leadership is positively related to older workers' desired retirement age via subjective health and work-to-private life conflict.
Work characteristics can be viewed as stay or retention factors when they attract individuals to work, or as push factors when they are perceived as negative aspects and lead older workers to retire (De Preter et al., 2013;Hofäcker & Unt, 2013;Hofäcker, 2015). Hofäcker and Unt (2013) proposed as stay factors for the institutional context the active aging policies that are implemented by Governments to improve older workers' employability and promote their maintenance in the workforce. Similarly, organizations can design and implement strategies to encourage older workers to stay active and productive at work (Armstrong-Stassen, 2008;Bal et al., 2012;Shacklock & Brunetto, 2011). As past research has shown, these strategies to prompt later retirement should include higher levels of support, more autonomy, greater decision authority, more recognition, career advancement prospects and opportunities to develop skills, active aging practices, and an age-friendly climate (Browne et al., 2019;Carr et al., 2016;Fasbender et al., 2019;Sousa et al., 2019;van Solinge & Henkens, 2014). Work experiences are, therefore, determinants of workers' satisfaction, and greater job satisfaction might lead workers to postpone retirement (Böckerman & Ilmakunnas, 2019;Oakman & Wells, 2013).
These results suggest that there are several workplace characteristics that can be changed by organizations to delay retirement timing, but it is still necessary to understand the universality of these interventions according to individual factors. Some workers are expected to benefit more from workplace modifications than others. Individuals who experience poor health, for instance, might benefit from a reduction in the workload, but probably they will continue to desire an early retirement, as they do not feel able to perform their job.
The qualitative approach used in the data collection of this study considers and accounts for multiple factors that might describe the idiosyncratic trajectories toward retirement. The combination of qualitative (i.e., content analysis) and quantitative (i.e., MCA and Cluster Analysis) data analysis can be especially useful to obtain a deeper understanding of the interindividual variability observed in the transition to retirement. We therefore explored the co-existence of individual (i.e., financial satisfaction, health, pension eligibility conditions, family situation, meaning of work) and work-related (i.e., relationship with colleagues, work experiences, job satisfaction) factors in characterizing the retirement decision-making process. Furthermore, we used a person-centered approach to group individuals who were similar in terms of the configuration of these factors, with the purpose of understanding the preference for early, on-time, and later retirement.

Participants and Procedure
Forty-one Portuguese people participated in this study, aged between 55 and 70 years old (M = 61.3, SD = 3.77). More than half of the participants were male (56.1%), and 36.6% completed higher education. Of these 41 participants, 26 (63.4%) were still working. The average age of workers was 60.8 years (SD = 3.48), and of retirees was 62.2 years (SD = 4.20). Participants' occupations ranged from banking and finance professions to technical staff including, among others, bank clerk, maintenance technician, administrative assistant, and nurse. The demographic characteristics of the participants are described in Table 1.
The data were collected through in-person semi-structured interviews until saturation was reached (Corbin & Strauss, 2008), which happened in June 2018. Each interview was performed individually, and participants were recruited through snowball sampling. The research team initiated the snowball sampling by contacting individuals in their personal networks who met the inclusion criteria. To qualify for the study, workers had to have 55 years old or over, and retirees could not be retired for more than five years. The selection of the participants intended to maximize diversity in terms of age, gender, occupation, and career pathways. Thus, participants were asked to refer friends or acquaintances who met the criteria but were dissimilar to them in terms of gender or job. Participants were informed about the study's objectives and asked for their permission to record the interview. This study complies with the Ethical Principles of Psychologists and Code of Conduct of the American Psychological Association.

Instrument
A semi-structured interview script was developed, anchored in relevant scientific literature, which focused on the specific theme of retirement. A pre-test with four participants was conducted to assess the rigor and relevance of the script, to evaluate if questions were comprehensible to individuals, and to estimate the duration of the interview.
The first part of the interview script was designed to collect demographic information about the participants (e.g., age, gender, level of education). The interviews started with a broad open question ("When did you begin to think about your retirement?"). We also developed support questions focused on the different factors that influence retirement intentions for participants who were working and retirement decision for participants who were retired. The individual factors included questions about participants' financial situation (e.g., "How was your financial situation when you began to think about retirement?"), perceived health and work ability (e.g., "How do you evaluate your health? Do you feel you will be able to work until retirement?"), and social support outside the work context (e.g., "Tell me about your family. Did your household changed from the day you began to think about retirement until today?"). The work factors comprised questions about job characteristics, experiences at work, and organizational practices, and how these aspects can encourage workers to stay active (e.g., "Which aspects of your work motivate you to continue working? What aspects make you want to retire?"). From a biographical perspective, this interview script allows individuals to narrate their story identifying significant and meaningful experiences to them based on its pertinence to the retirement decision (Poirier et al., 1983;Ramos, 2010). The temporal logic of the biographical perspective is important for understanding the dynamics of the retirement decision-making process.

Data Analysis
The interviews, with an average duration of 32 min, were audio-recorded and transcribed verbatim. Initially, the transcribed interviews were read through several times to obtain an overview of the retirement intentions or decision (Graneheim & Lundman, 2004). Then, data management and content analysis were conducted using MAXQDA18 (Kuckartz & Rädiker, 2019). A deductive approach was used to develop conceptual categories and some subcategories according to the literature (e.g., financial satisfaction, health, work-related factors), whereas other subcategories emerged from the data through an inductive process (e.g., meaning of work) (Patton, 2015). The coding units were sentences or paragraphs (excerpts relatively large) that were then assigned to the mutually exclusive subcategories (Bardin, 1996). Finally, a category system was developed with the name and definition of each conceptual category, along with examples of the units of analysis (Bauer & Gaskell, 2000;Patton, 2015). Most of the variables had three categories, representing the absence of answer, and the positive and negative valence of the dimension. For instance, the variable Financial satisfaction was categorized by 1 -Not mentioned, 2 -Financial dissatisfaction, and 3 -Financial satisfaction. Other variables had three categories that represent the intensity of the dimension, as is the case with the variable Positive experiences (1 -Not mentioned, 2 -Positive experiences, 3 -High positive experiences). Only the variable Meaning of work had more than three categories (i.e., five categories) to describe the different meanings that work can represent to individuals (e.g., instrumental value, sense of purpose, self-realization).
The reliability of the category system was assessed through Cohen's kappa, which evaluates the degree of agreement between two raters in the coding process (Cohen, 1960;Hsu & Field, 2003). A total of 152 randomly selected units of analysis (20% of the total units) were coded by a second independent researcher who had access to the category dictionary. Cohen's kappa showed a substantial inter-rater consistency, κ = .662, p < .001 (Landis & Koch, 1977).
An MCA was applied to find retirement transition profiles. This multivariate method is particularly suitable to analyze the associations between multiple categorical variables. Like Principal Component Analysis, MCA allows the definition of a new system of orthogonal axes (dimensions), which corresponds to the latent constructs that structure the retirement transition profiles. Each dimension is composed by all the input variables that contribute with a different discrimination measure. The most relevant variables for each dimension are the ones that have the highest discrimination/contribution values. These dimensions (orthogonal axes) are used to provide low-dimensional representations (Gifi, 1996) of the relationships among categorical variables, usually twodimensional graph (Carvalho, 2017;Le Roux & Rouanet, 2010). Using an optimal scaling procedure (Blasius & Greencare, 2006), the MCA algorithm defines optimal quantifications (coordinates) for all the categories of the variables under analysis. Those coordinates are used to represent all categories in a factorial plane. Focusing on categories, their associations are emphasized by the geometric proximity of their coordinates in the factorial plane, thereby making the structure of the interrelationships between the variables visible. In this case, MCA provided the structure of the associations between the factors that influenced the retirement decision-making process, which allowed us to assess the multivariate configuration of the retirement transition profiles. Based on new optimal quantifications obtained from the MCA algorithm for the categories of each variable, a score was also calculated for each object, which represents the average of the category quantifications of its response profile (Gifi, 1996). From the MCA standardized object scores of each selected dimension, two new composite and quantitative variables were obtained, and then entered in a Cluster Analysis to group participants according to their retirement transition profiles. First, a hierarchical cluster analysis was conducted through two different agglomerative methods: Ward and farthest-neighbor (Hair et al., 2019). The convergence of the cluster solution proposed by each agglomerative method, joint with the MCA solution, sustained the accuracy of the MCA results. Then an optimization algorithm (kmeans) was implemented to optimize the final partition of the participants across their retirement transition profiles. As the profiles had a multidimensional configuration, the clustering solution made it easier to perform a posteriori analysis. Data analysis was performed using SPSS 26.0.

Results
The results of the MCA suggested a three-dimension model to characterize the retirement decision-making process. The discrimination measures and the contributions of the active variables for the three dimensions are in Table 2. The discrimination measures in bold represent the variables for which the values were higher (or near) the inertia of each dimension. Regarding the definition of the profiles, dimensions 1 and 3 were used. Although dimensions 2 and 3 showed similar inertia (variance accounted for each dimension), dimension 3 was preferred, as the amount of contribution of the selected variables was greater (Table 2). From the joint plot of category points for dimensions 1 and 3, three profiles that represent different configurations of the associations among the categories were identified. Figure 1 shows the association between the individual and work indicators that characterize the retirement transition. The solution provided by the two agglomerative methods in the hierarchical cluster analysis converged to three clusters, which was consistent with the three profiles suggested by MCA.
Drawing on the work of Shultz et al. (1998) andHofäcker (2015), these three profiles reflect distinct combinations of pull, push, and stay factors. The first profile, called push factors (Cluster 1, 22.0% of the participants), was characterized by the association between negative relationships at work, high negative experiences at work, poor health, poor work ability, financial dissatisfaction, and not being eligible in terms of Social Security contributions. In this profile there was a predominance of negative aspects at both individual and work level that induce workers to retire, making these individuals "the resistant". The second profile was characterized by an association between negative experiences at work, job dissatisfaction, the work meaning self-realization and sense of purpose, financial satisfaction, being eligible in terms of Social Security contributions, having dependents, and having a retired spouse or partner. This profile was called push and pull factors (Cluster 2, 34.1% of the participants) as there were negative aspects pushing people to retirement (e.g., job dissatisfaction) and positive factors that attracted workers toward retirement (e.g., retired spouse or partner). These older workers were "ready to go". Finally, the third profile, known as stay factors (Cluster 3, 43.9% of the participants), was characterized by the association between high positive experiences at work, perceiving work as an occupation, having no dependents, and the spouse/partner not being retired. In this profile the factors that make employment seem more attractive than retirement prevailed, which makes these older workers feel unstoppable at work.
Moreover, Fig. 2 shows the association between age, education, and gender (demographic variables), clusters (new variable obtained after the clustering), and retirement timing. Cluster 1 was associated with on-time retirement. The demographic characteristics did not differentiate the participants in this cluster. Cluster 2 was associated with early retirement and was characterized by the youngest (55-59 years old) and most qualified (higher education) participants in this sample. Finally, cluster 3 was associated with later retirement and characterized by the oldest (65-70 years old) and least qualified (9th grade or less) individuals in this sample. Gender did not differentiate the participants across clusters.

Discussion
The present study sought to examine the individual and work indicators that characterize the retirement transition profiles, as well as to group individuals according to these profiles. The results suggest the existence of three different profiles of retirement transition that were associated with different retirement timings (i.e., early, on-time, and later retirement).
Individuals in cluster 1 desired to leave the workforce as soon as they satisfied the eligibility conditions to receive a pension (i.e., on-time retirement). These workers can be seen as "the resistant" as they endured working while they needed it since they experienced several negative aspects in their work (e.g., working long hours, physically and mentally demanding tasks, highly competitive work environment), and perceived poor health and poor work ability. Only financial  200 .193 Values in bold are above inertia for each dimension Fig. 1 Topological configuration of the retirement transition profiles dissatisfaction and having a career with insufficient contributions to receive an old-age pension were motives to stay at work and retire on-time. Regardless of their age or education, individuals were pushed to retirement to avoid the negative experiences at work and to search for new ways of enjoying their time (e.g., to rest, pursue new hobbies). These results are consistent with previous research on the influence of low job satisfaction, continued economic pressure, and poor health and work ability in retirement timing (e.g., Fisher et al., 2016;Topa et al., 2009). Individuals in cluster 2 were "ready to go" as they desired to leave before the statutory retirement age (i.e., early retirement). These older individuals did not need to continue working since they were eligible for receiving an old-age pension in terms of Social Security contributions, and they were financially satisfied. Despite these being the youngest individuals in the sample, they were the most qualified, which probably contributed to a career with higher salaries and more benefits (e.g., health insurance, company car) (Münich & Psacharopoulos, 2018), and also to their satisfactory financial situation in later life, as explained by the Cumulative Advantage model (Crystal & Shea, 1990;Crystal et al., 2017). Furthermore, the family situation pulled these workers to retirement, which is also consistent with previous evidence. Individuals who have a retired spouse, and children or grandchildren of young ages want to spend more time and energy on private life (Gustafson, 2018;Kubicek et al., 2010). These older individuals were also not satisfied with their work and had negative experiences, pushing them to early retirement. Only the meaning of work to these individuals can act as a retention factor. Work provided them with a sense of self-realization (e.g., upgrade skills and knowledge, challenge one's self) and purpose (e.g., add value to society, feel useful to others), feeling that their lives have meaning and for that reason they should stay at work. This idea is supported by previous studies showing that it is especially important for older individuals to perform meaningful work and to find meaning in life (Noonan, 2005;Steger & Dik, 2009). It is likely that when individuals find other activities that meet their sense of fulfillment and purpose (e.g., volunteering), they will leave the workforce to invest their time in these new roles, where they feel that their skills are used and valued (Higginbottom et al., 1993). Individuals in cluster 3 showed a preference for later retirement, which conveys the message that "nothing can stop me now". First of all, work is a source of pleasure to these older individuals, as they had high positive experiences such as autonomy, challenging cognitive tasks, or opportunities to learn. As shown by past research, these aspects can be crucial to retain older workers and encourage them to retire later (e.g., Browne et al., 2019;Carr et al., 2016). Second, work meant being active and occupied, and having a routine, since the life outside the work context did not attract the individual to retirement (i.e., having no dependents, having a spouse/partner who is still working). These individuals were the oldest and the least educated in the sample, a combination of factors that is associated with lower income throughout careers, which could, in turn, influence the need for working beyond traditional retirement ages (Crystal et al., 2017). However, the economic status of these participants did not emerge as a Fig. 2 Associations between clusters (according to retirement transition profile), retirement timing, and demographic characteristics factor crucial to their desire to postpone retirement. In fact, positive experiences at work seem to be one of the key factors in retaining these older workers. It is likely that these older individuals will continue to participate in the workforce until their family and/or work situations change.
The findings of our research corroborate the idea that retirement is a multidimensional and complex process in which different factors have varying degrees of relevance to older individuals and relate to each other in different ways. This multidimensionality and complexity were well captured by MCA and Cluster Analysis, two methods that allowed the identification of different retirement transition profiles taken by different groups of older individuals. Hence, the transition to retirement is a life-altering decision that has different meanings and results from a wide variety of answers to the question of why people retire.

Theoretical and Practical Implications
The changing landscape of retirement contributed to a growing interest of researchers and organizations in the topic, particularly in the identification of the factors that can contribute to extended working lives. This study has important theoretical implications for this line of research. Whereas previous research shows the singly importance of few retirement predictors (e.g., Carr et al., 2016;Fisher et al., 2016), this study provides evidence of how the combination of multiple variables can define idiosyncratic transitions to retirement. The present investigation thus extends recent work on the role of individual and work-related factors in retirement timing, suggesting that the association between these variables is likely more complex than simply a linear relationship.
Further, our results extend prior research on late-career attitudes and preferences by demonstrating the heterogeneity of older individuals in the motives that influence their desire for early, on-time, or later retirement. This heterogeneity also questions the value and appropriateness of the concepts of voluntary and involuntary retirement, by showing that this decision may not be so straightforward. These concepts present a dichotomous and oversimplified perspective of the retirement decision-making process, and according to Kerr and Dacyshyn (2000), their distinction may be artificial in some situations.
This study highlights the relevance of considering the uniqueness of individual trajectories throughout life in organizational psychology. There is a growing amount of research demonstrating the importance of hiring and retaining older workers for organizational success (e.g., Ng & Feldman, 2010;Peeters & van Emmerik, 2008;Peterson & Spiker, 2005). However, in practice most organizations remain passive in implementing practices that keep older workers motivated, healthy, active, and productive at work (Conen et al., 2012;Van Dalen et al., 2009). The findings of the present study can help organizations to develop and implement practices tailored to each of the three groups of older individuals. Such practices are not only important to retain older workers' knowledge and experience, but also to attract those older individuals who choose bridge employment (i.e., other forms of participation in the labor market between retirement from the main career job and complete workforce withdrawal; Topa et al., 2014).
Workers who feel that they need to stay at work, despite the lack of motives to do so, might benefit from accommodative practices, such as additional leave or reduced workload, and maintenance practices, such as ergonomic adjustments (Bal et al., 2013;Kooij et al., 2014). These practices can reduce work demands, improve working conditions, and prevent health deterioration, contributing to a late-career phase with quality of life. Nevertheless, these individuals who experience losses in their capabilities should also receive special attention from public policies. For cases in which people's health does not allow them to work, disability pensions can be the only option even if it represents a significant loss of future income (OECD, 2003). Increasing the statutory retirement age without implementing policies to protect workers with health problems can threaten the retirement security of many older workers. At a societal level, a proactive strategy is needed to monitor the health and aging of potential at-risk workers (e.g., chronic patients, individuals in highly demanding jobs), preventing their premature departure from the workforce.
Individuals who desire to retire before the statutory retirement age, due to the forces that pull and push them to retirement, could possibly be encouraged to remain active if they were to find positive experiences at work. Also, as work gives these individuals a sense of self-realization or purpose, both developmental and utilization practices can influence workers' decisions to postpone retirement (Kooij et al., 2014;Kuvaas, 2008). If people are not satisfied with their job, organizations can redesign jobs to enrich the tasks or create mentoring programs, giving workers autonomy and the opportunity to share their knowledge and contribute to the development of mentees' skills. Career management practices such as career counselling or career planning workshops can increase the individual competencies required for a successful performance, as well as provide important information to workers about career paths and promotions (Bagdadli & Gianecchini, 2019). These practices can at the same time enhance individuals' sense of self-realization and purpose, creating a meaningful work experience.
Finally, for those individuals who have already decided to delay retirement, a wide variety of organizational practices can still be essential to promote positive experiences at work. For instance, development practices such as regular training can provide older workers with increased knowledge and skills to take more responsibility in their job, allowing the organization to take advantage of their full potential. It is possible that in the absence of these practices, individuals will no longer have positive experiences, which may lead them to retire earlier than they truly wish. Besides the importance of organizational practices in this late-career phase, such practices were probably important throughout the career of these individuals. A performance evaluation system that provides feedback on current performance (maintenance practice) and on the performance required to achieve a higher job position (development practice), for example, may have provided workers with opportunities to learn and to grow and challenge themselves, contributing to an overall positive experience at work.
The growing diversity of retirement transition profiles requires that organizations develop differentiated practices to meet workers' needs and preferences over time. The ability of organizations to create and implement age-sensitive practices that support workers throughout their careers, especially in the late phase, can influence their intentions and actual retirement timing, fostering longer and healthier working lives.

Limitations and Future Research
This research may have two limitations. The first concerns the form of retirement analyzed in this study. All interviewees were considering or had already made a transition from fulltime employment to full-time retirement (i.e., full-time work one day, retired the next), since this is still the most common form of retirement in Portugal. However, instead of the abrupt retirement, individuals may retire gradually over time by reducing their work hours in a current job, or moving to a less demanding job with a different employer. The different forms of retirement are a vital issue for future research on retirement patterns.
A second limitation is related to the context of the study. Despite the contribution of the findings to better understand the retirement transition in the Portuguese context, we recognize that the categorization process depends on the characteristics of the pension systems across countries (e.g., mandatory retirement, different retirement ages for men and women). For instance, if there is a mandatory retirement age, the relevance of work-related factors in motivating older workers to continue working sharply decreases. Thus, this study should be replicated in different contexts (e.g., countries, jobs) to understand the relative importance of the retirement regulation. Future work could also explore the role of retirement expectations in these retirement transition profiles.
Finally, longitudinal studies would also be an important contribution to understand how the roles of individual and work-related factors change over time for different groups of individuals. Participants could be interviewed, for instance, at three distinct times: about five years before the statutory retirement age, near the statutory retirement age, and five years after the moment of retirement. These three moments would allow researchers to explore future expectations, in the long and short term, as well as a retrospective analysis of the retirement decision-making process.

Conclusion
This research has demonstrated the multidimensionality and complexity of the retirement decision-making process by uncovering three retirement transition profiles that were associated with different timings of retirement. Older individuals who have been resisting and choose on-time retirement despite the factors that push them into retirement can be found in the first profile. People in the second profile are ready to go since they are being influenced by push and pull factors to take on early retirement. Finally, individuals in the third profile feel unstoppable at work due to the predominance of stay factors, and prefer later retirement.
These heterogeneous profiles draw attention to the need of developing and implementing measures tailored to each group of older individuals to prolong working lives. Governments and organizations are called to create strategies that protect and promote workers' health, and foster positive experiences at work, such as job autonomy and opportunities to learn and transfer knowledge.