Matching the Pieces: The Presence of Idiosyncratic Deals and Their Impact on Retirement Preferences Among Older Workers

Despite working life prolongation having been at the center of the policy agenda in Europe for the last two decades, organizations’ engagement in formal age-management activities intended to strengthen older workers’ motivation and work ability appears limited. Given policies to extend working lives, negotiated individualized work arrange-ments—often called idiosyncratic deals (I-deals)—can be an informal and complementary approach to formalized age-management practices, improving the person–job fit and helping older workers extend their working lives. Nevertheless, research on I-deals and retirement preferences remains scarce in the Nordic context, where collective agreements regulate conditions of employment and the employer–employee relationship. Using confirmatory factor analysis and structural equation modeling, this study examines five areas of I-deals (i.e., Task and Work Responsibilities , Workload Reduction , Schedule Flexibility , Location Flexibility , and Financial Incentives ) and their relationships with re-tirement preferences among Swedish public-sector employees aged 55 years or older ( n = 4,499). Findings show that I-deals are generally less prevalent

Population aging and legislative actions to extend working life are rapidly increasing the share of older workers in many European labor markets.For example, from 2002 to 2018, the employment rate of older workers aged 55-64 years increased from 38% to 58% (Eurostat, 2018).However, there are still some missing elements necessary for the frictionless extension of working lives, elements that bridge the gap between state legislation and social security systems, organizational readiness and measures, and older workers' motivation and ability to continue working.To meet the needs of a more diverse and aging workforce, scholars have called for action at the organizational leveloften referred to as age-management strategies-to promote extended working lives (Hasselhorn & Apt, 2015;Ilmarinen, 2006;Naegele & Walker, 2006, Truxillo, Cadiz, & Hammer, 2015).Present labor and skill shortages, not least in the public sector, further emphasize this argument.However, even though age-management strategies are continuously advocated in the management literature and have been part of the European Union's (EU's) policy agenda for over two decades (Foster & Walker, 2015;Walker, 2005), past research indicates that organizations often take a rather "passive" approach in implementing measures to retain older employees (Conen, Henkens, & Schippers, 2012;Fuertes, Egdell, & McQuaid, 2013;Jonsson, Lindegård, Björk, & Nilsson, 2020;Vickerstaff, Cox, & Keen, 2003).
Considering the gap between age-management theories and employers' allegedly passive approach in relation to them, this study Work, Aging and Retirement, 2021, Vol. 7, No. 3, pp. 240-255 doi:10.1093/workar/waab003Advance Access publication 1 April 2021 Empirical Article focuses on individually negotiated work arrangements, so-called idiosyncratic deals (I-deals; Rousseau, 2001).In contrast to more formal age-management strategies imposed by employers, I-deals are informal and individually negotiated between employees and employers.Because they imply that work is tailored to individual needs, I-deals are assumed to strengthen the person-job fit.A basic rationale for their implementation is thus that they are mutually beneficial for both the employer and the employee (Rosen, Slater, Chang, & Johnson, 2013).More specifically, I-deals have the potential to improve employees' working conditions overall and also signal interest on the part of the employer in retaining and developing high-quality employment relationships.As such, they are likely to positively affect employees' work attitudes, preferences, and actual behavior-including their opportunities for work in later life (Rousseau, Ho, & Greenberg, 2006).Notwithstanding the empirical basis supporting I-deal implementation to achieve desired organizational outcomes in general, and prolonged working lives in particular, is still in its infancy.Furthermore, little is currently known about the function and effects of I-deals in labor markets characterized by strong collective agreements and HR policies geared toward universal solutions, and there is also a need for more studies focusing on I-deals' antecedents, contextual as well as individual (Hornung, Rousseau, Glaser, Angerer, & Weigl, 2010).In other words, the question of when and where opportunities for such individually negotiated work arrangements exist, and for whom, remains to be further explored.In relation to retirement, for instance, the importance of such endeavors is underlined by a number of recent studies indicating that, when applied, measures to retain older workers are often applied selectively and/or in an ad hoc manner.Because this appears to primarily benefit employees that the employer would like to keep, and not seldom already privileged workers (e.g., those in higher socioeconomic positions and/or with already beneficial working conditions; Hofäcker, Hess, & König, 2016;Wainwright et al., 2019), it risks reinforcing the inequalities that prolonged working lives bring to the fore (Kadefors, Nilsson, Östergren, Rylander, & Albin, 2019;König, Lindwall, & Johansson, 2018;Radl, 2013;van Solinge & Henkens, 2017).This risk is accentuated by the fact that early retirement preferences/intentions are already known to be more widespread in certain groups, for example, among women, low-skilled workers, and employees with generally strenuous working conditions (Bockerman & Ilmakunnas, 2017;Solem et al., 2016;Örestig, Strandh, & Stattin, 2013).
To sum up, the identification of factors that could strengthen older workers' motivation and ability to continue working is critical, and I-deals could constitute a complementary approach to more formalized age-management practices in efforts to retain older workers by improving the person-job fit.However, opportunities for such flexible and personalized work arrangements are likely to be heavily contingent on both individual-level factors as well as on more general working conditions and labor market regulations.In turn, this implies that an ill-advised implementation of I-deals could give rise to unequal opportunities in terms of possibilities to extended working lives.Therefore, the present study seeks to explore five areas of I-deals (i.e., Task and Work Responsibilities, Workload Reduction, Schedule Flexibility, Location Flexibility, and Financial Incentives;Rosen et al., 2013;Rousseau & Kim, 2006) and their relationships with retirement preferences among Swedish public-sector employees.Furthermore, it sets out to advance current research by exploring the presence of I-deals across different groups of workers.The more specific hypotheses are presented in the following sections.

Contextualizing Idiosyncratic Deals Under the Umbrella of Collective Bargaining
Idiosyncratic deals, a term originally coined by Rousseau (2001), can be defined as "voluntary, personalized agreements of a non-standard nature negotiated between individual employers and their employees regarding terms that benefit each party" (Rousseau, 2001).The I-deal literature frequently views employment relations as a (social) exchange relationship in which the workplace is an arena for employers and employees to exchange social and economic resources (Blau, 1964, Conway & Coyle-Shapiro, 2015).In a work setting, exchange relations between an employee and employer are usually stipulated in employment contracts that are subordinated to labor market laws and regulations in the countries where the contracts are made.The idea of I-deals has migrated from an American context of employment relations to a more diverse international research environment, crossing cultural boundaries and accommodating differences in national labor market institutions.Such contextual aspects will likely affect the prevalence of I-deals within organizations, while also shaping the extent and application of formal HR practices.In Sweden, individuals have traditionally found strength in the collective (through union association), and the traditional bargaining model constitutes a special case distinct from those in many other European countries (Visser & Kaminska, 2009), being based on "strong employer organizations and unions negotiating collective agreements with a high degree of autonomy from the state" (Furåker & Larsson, 2020).In the Swedish labor market, work accommodations have historically been regulated primarily through the lens of collective agreements or labor market law.However, the heterogeneous needs of older employees challenge this institutionalized practice and put I-deals on the agenda.
On solving the Person-Job Fit Puzzle: Idiosyncratic Deals as Means to Accommodate the Heterogeneous Needs of an Aging Workforce?
The lifespan psychology literature states that aging encompasses both growth and decline in psychological and physical functions (Baltes & Baltes, 1993) as well as a transition to selecting more emotionally meaningful goals and activities (Carstensen, 1993, Kooij, De Lange, Jansen, Kanfer, & Dikkers, 2011).To minimize losses and remain functioning over their life course, individuals must adjust, optimize, set, pursue, and maintain goals that they find meaningful.Thus, the older people get, the more their unique character traits are reinforced (Kooij & Van de Voorde, 2015).It follows that heterogeneity in terms of needs and motivations at the aggregate level also tends to be greater in older age groups (Hansson, DeKoekkoek, Neece, & Patterson, 1997;Kanfer & Ackerman, 2004;Rowe & Kahn, 1987).Consequently, several researchers have concluded that an extended working life for more than just privileged groups requires both employee involvement and workplace-specific interventions (Bal & Jansen, 2015, Kadefors, Wikström, & Arman, 2020).This is not least because individuals' retirement decisions, although sometimes framed as a matter of personal choice, are known to be constrained by their opportunity structure, ranging from the resources they possess (e.g., health, education, and finances) to the contextual and institutional elements where they live (e.g., pension system eligibility age, norms, and culture; Fisher, Chaffee, & Sonnega, 2016;Wang & Shultz, 2010).
I-deals entail an adaptation of work tasks that is neither collective nor standardized; instead, these arrangements are determined by individual and, therefore, heterogeneous needs.As such, they are intended to maintain motivation and uphold performance in order to facilitate the achievement of common goals (Kooij & Van de Voorde, 2015).As stated by Rosen et al. (2013, p. 714), "almost any monetary or nonmonetary resource that is valued by an employee can be negotiated in an I-deal," which might partly explain why considerable variation in focus and operationalization exists across I-deal studies (Conway & Coyle-Shapiro, 2015;Rosen et al., 2013;Sun, Song, Kong, & Bu, 2019).However, most existing studies have focused on I-deals granted in one or more of the following four dimensions: work hour flexibility, workload reductions, career development (Rousseau & Kim, 2006), work tasks/job content (Hornung et al., 2010, Rosen et al., 2013, see also Liao, Wayne, & Rousseau, 2016), and location flexibility (Rosen et al., 2013).Although previous research on I-deals and their potential role within organizations is relatively limited, this type of arrangements have been demonstrated to have positive effects on the ability to attract, retain, and motivate employees (Bal & Jansen, 2015, Conway & Coyle-Shapiro, 2015).Likewise, their implementation has been found to promote positive relations within the organization (Rousseau, Hornung, & Kim, 2009), organizational commitment (Ng & Feldman, 2010), employability (Oostrom, Pennings, & Bal, 2016), and enhanced productivity (Hornung, Rousseau, & Glaser, 2008).With respect to working beyond retirement, Bal, De Jong, Jansen, and Bakker (2012) found that both "flexibility deals" (e.g., flexibility in terms of work tasks/schedule) and "developmental deals" (e.g., opportunities for skill development) increased employees' motivation to work longer.
Drawing on the Ex Post-Ideals Scale, originally developed and validated by Rosen et al. (2013), the present study examines the associations between retirement preferences and the following four I-deal dimensions: Task and Work Responsibilities (TWR), Schedule Flexibility (SF), Location Flexibility (LF), and Financial Incentives (FI).The Ex-Post I-deals Scale was considered particularly suitable as it includes items related to several aspects well known to influence retirement timing, for example, opportunities for flexibility in terms of job control (Browne, Carr, Fleischmann, Xue, & Stansfeld, 2019) and competence development (Van Solinge & Henkens, 2014;Zaniboni, Sarchielli, & Fraccaroli, 2010) as well as financial situation (Fisher et al., 2016).However, the scale does not include any questions relating to workload.Therefore, inspired by Rousseau and Kim (2006), and because factors such as work stress, functional limitations (Wahrendorf, Dragano, & Siegrist 2013), and physical workload (Lahelma et al., 2012)

I-Deals for Whom?
As to who receives I-deals, it has been proposed that employers might be more inclined to grant certain employees, such as high performers, this type of individual arrangements (Conway & Coyle-Shapiro, 2015;Rousseau et al., 2006).Additionally, it has been shown that employees who have better relationships with their supervisors and/or who are "politically skilled" are more likely to strike I-deals (Rosen et al., 2013).However, in more general terms, opportunities for flexible and personalized work arrangements are not solely dependent on individual-level factors.Rather, they are known to also be heavily influenced by factors such as job type.Specific work tasks might, for example, require physical presence and/or need to be carried out during specific hours (Damman & Henkens, 2018;Hornung, Rousseau, & Glaser, 2009).Furthermore, as recently demonstrated by Damman and Henkens (2018), perceptions of schedule and location flexibility, both of which are factors suggested to promote prolonged working lives, differ markedly between groups of employees.Their results suggest that women generally experience less flexibility, in terms of both schedule and work location, than do men, and similar differences were observed between occupational skill levels.From this follows that it is crucial to also acknowledge the structural conditions under which I-deals, specifically, are granted to employees-a matter that, to our knowledge, has received relatively little attention in previous literature.As to flexibility and workload reduction deals, for example, Hornung et al. (2009) demonstrated that the extent to which they are authorized by supervisors depends inter alia on constraints such as required onsite presence.Regarding task-and career-related deals, Hornung et al. (2010) found the former to be associated with both job level and organizational tenure, whereas Hornung, Rousseau, Weigl, Mueller, and Glaser (2014) suggested that career-related deals are primarily influenced by job level.In light of these findings, and because an uneven distribution of I-deals could give rise to unequal opportunities in terms of ability and willingness to continue working, the second hypothesis guiding this study reads as follows: H2: Socioeconomic position (SEP), self-rated health, and organizational tenure will be positively associated with all I-deal dimensions, while age and female sex will be negatively associated with them

Retirement Preferences
During the retirement decision-making process, people's intentions and deliberations are affected by a range of environmental factors.We therefore consider the term "preferences" to be suitable to describe employees' views of their future retirement timing.Preferences, as a theoretical concept, are exogenous and, as such, adaptable to social circumstances, temporal opportunities, and constraints (Feldman & Beehr, 2011;Halleröd, Örestig, & Stattin, 2013;Lukes, 2004).This is relevant in an organizational context in which there is striving, by creating the right opportunities and work conditions via personnel policies, to enable individuals to prolong their working lives.In this study, retirement preferences were measured by the following prospective question: "

I-deals
The latent variables were estimated and defined through confirmatory factor analysis (described in more detail below) and consist of the covariance between a set of manifest items (Little, 2013).As indicated above, most of the manifest items (13 of 15) were obtained from the Ex Post I-Deals Scale (Rosen et al., 2013), albeit translated into Swedish and slightly modified (i.e., superlative wording was toned down) to better fit the contextual and cultural aspects of Swedish employment relations and work life.The remaining two questions concerning workload were intended to capture employer-employee negotiations regarding alternative and/or fewer work tasks.Each of the five I-deal dimensions was captured by a latent variable.All 15 questions (presented in full in Table 2) were formulated such that respondents could specify the extent to which they had negotiated agreements with their employers, using a seven-point Likert scale ranging from 1 ("Not true at all") to 7 ("Completely true").

Covariates
The demographic variables age, sex (male = 1, female = 2), self-rated health, and socioeconomic position (SEP) were included to examine the prevalence of I-deals across different groups of employees as well as to control for their potentially confounding associations.Self-rated health was assessed using a six-point Likert scale ranging from 1 ("Very bad") to 6 ("Very good") and organizational tenure was indicated by a dichotomous variable where 1 = ≤5 years and 2 = >5 years.(Oesch, 2003(Oesch, , 2006)).
Compared with traditional ways of operationalizing class by means of hierarchical division, the class schema proposed by Oesch (2003Oesch ( , 2006) ) puts greater emphasis on differences in skills and "work logics" (i.e., technical vs. organizational vs. interpersonal).As such, it is better designed to also capture horizontal (and often gendered) class cleavages such as that between workers in routine sales or service occupations, and production workers.Since the study sample consists solely of public-sector workers, this schema was considered the most suitable option.For employees, it distinguishes between the following four socioeconomic classes: (1) unskilled workers, (2) skilled workers (e.g., craft workers, clerks, and skilled service workers), (3) lowergrade service class (e.g., semiprofessionals and associate managers), and (4) higher-grade service class (e.g., professionals and managers).

Statistical Analysis
In this study, confirmatory factor analysis (CFA) and structural equation modeling (SEM) with robust full-information maximum likelihood (FIML) estimation were applied to test the hypothesized relationships between retirement preferences and five types of idiosyncratic deals (Rhemtulla, Brosseau-Liard, & Savalei, 2012).This methodological approach is essentially hypothesis driven and has several advantages.First, it enables the incorporation of both observed and unobserved variables (i.e., latent factors).Latent factors represent theoretical constructs that are assumed to underlie outcome variation of some sort.As such, they are not readily observed and therefore difficult to measure.However, by means of CFA, the covariance between a set of observed (or manifest) variables can be used to indicate such more abstract phenomena.By extension, this means that the latent factors' antecedents and outcomes can be further explored.Second, unlike traditional regression techniques, CFA/SEM allows the analyst to explicitly estimate and model measurement error (Byrne, 2013).Following standard CFA/SEM procedures, we relied on the following fit indices and cutoff criteria to evaluate model fit: a root mean square error of approximation (RMSEA) of close to 0.06 or below, a comparative fit index (CFI) and Tucker-Lewis index (TLI) of close to 0.95 or greater, and a standardized root mean square residual (SRMR) of less than 0.08 (Hu & Bentler, 1999).In addition, χ 2 is also reported, but because it is known to be inflated when n is large (Browne & Cudeck, 1993, Byrne, 2013), significant values were not considered to be of any major concern.For latent factors with only two indicators, an equality constraint was put on the factor loadings (Little, Lindenberger, & Nesselroade, 1999).Fully standardized coefficients are reported for continuous predictors (including latent factors), while coefficients of binary predictors were standardized only regarding Y. Data management, data cleaning, and descriptive statistics were carried out in Stata 15.1, and the CFA/SEM analyses were estimated in MPlus version 8.1.

Post Hoc Face Validity
After the analysis, a face validity analysis was conducted to improve our understanding of how the I-deal questionnaire was interpreted by different occupational groups.Respondents were sampled using a strategic sampling approach to include a variety of occupations.The sample comprised employed individuals aged ≥55 years living in Sweden (n = 13).We asked the respondents to (1) complete the questionnaire; (2) describe how they interpreted the questions and why they had responded as they did; and (3) explain how they understood the questions in relation to the response scales.The post hoc face validity results indicated that some respondents interpreted the questions as intended, that is, as referring to explicit negotiations, whereas other respondents answered the questions having in mind the general and present employment conditions at their workplace.The face validity findings are discussed together with results from the empirical analysis below.

R E SULTS
The analyses were performed in three steps, and the subsequent sections are organized in accordance with these steps.First, we conducted several descriptive analyses.Second, using CFA, we estimated a measurement model to confirm that all manifest indicators were related to the latent factors as posited.Finally, we expanded the measurement model by specifying a structural model, which concurrently tested our two main hypotheses (see Figure 1).Thus, the full structural model included both the potential I-deal predictors and the outcome of interest, that is, retirement preferences.To control for potentially confounding effects, direct paths between all predictors and retirement preferences were also specified, as were paths between SEP, age, sex, and self-rated health.

Descriptive Analyses
Overall, the mean values for all 15 I-deal indicators were relatively low (min.WLR2, x̅ = 1.7; max.FI2, x̅ = 3.5, both measured on a 7-point scale), suggesting that the presence of I-ideals is somewhat limited in the present sample (Table 1).Nonetheless, significant, yet relatively weak, correlations were observed between retirement preferences and all individual I-deal indicators, except those pertaining to Workload Reduction (Table 2).Moreover, the descriptive analyses demonstrated that the prevalence of such individual arrangements varies between different groups of employees, confirming that women, as well as individuals occupying lower SEPs and reporting poorer health, would prefer to retire earlier (Table 1).

Confirmatory Factor Analyses
When the measurement model, comprising 5 latent factors and 15 manifest indicators, was fitted to the data, the obtained factor loadings were generally high (none < 0.5 and 13 > 0.7) and significant.The fit indices for this model indicated moderate fit, i.e., χ 2 (df) = 1,681.625(83), RMSEA = 0.072, CFI = 0.917, TLI = 0.895, SRMR = 0.053 (Table 3), suggesting room for improvement.To identify localized areas of misfit, we turned to the obtained modification indices (MIs) and the corresponding estimates of expected parameter change (EPCs).The estimated MIs suggested that three constraints, all representing residual covariances, should be relaxed to improve model fit: (1) SM1 with SM2, (2) SF4 with SF3, and (3) SF4 with SF5.Accordingly, the model was respecified to include the aforementioned parameters, that is, to account for the fact that these items have something in common that is not captured by the latent factors.As recommended by Brown (2015), they were relaxed one by one (results not shown but available on request).Adding these three residual covariances ameliorated the model fit, as evidenced by the following fit indices: χ 2 (df) = 841.434(80), RMSEA = 0.053, CFI = 0.961, TLI = 0.948, SRMR = 0.040.The final measurement model, with five latent constructs, comprising the covariance between the 15 manifest indicators and 3 residual covariances, was thus considered to fit the data well (Table 3).

Structural Models
In accordance with H1, we expected all I-deal dimensions to be positively related to retirement preferences, that is, that more I-deals would Overall, the conceptual model fit the data poorly, as suggested by the following fit indices: χ 2 (df) = 5,725.710(171), RMSEA = 0.087, CFI = 0.784, TLI = 0.699, SRMR = 0.180 (Table 4, Model 1).Thus, areas of local misfit were identified by dint of the obtained MIs and the corresponding EPC estimates.The MI/EPC values suggested that the factor residuals should be allowed to covary.This was considered to make theoretical sense given that, for example, more specific work organization conditions not included in the present analyses might still influence the extent to which certain I-deal arrangements are possible and thus granted to employees.Consequently, the default constraints put on these parameters were relaxed stepwise (results not shown but available on request).Although the overall results remained virtually unchanged in comparison with the conceptual model, these modifications significantly ameliorated the model fit (Table 4, Model 2): χ 2 (df) = 1,297.394( 161), RMSEA = 0.041, CFI = 0.956, TLI = 0.935, SRMR = 0.030.As argued by Byrne (2013, p. 168), "final models in SEM should represent the best fitting, albeit most parsimonious model," which means that all parameters should be statistically significant.Thus, in the final step, we refined the model further through the stepwise exclusion of non-significant paths, starting with the one with the highest p-value ( Jöreskog, Yang, Marcoulides, & Schumacker, 1996; results not shown but available on request).The fit indices obtained from the resulting final model indicated a slight improvement compared with Model 2, that is, χ 2 (df) = 236.989(50), RMSEA = 0.030, CFI = 0.983, TLI = 0.974, SRMR = 0.017 (Table 4, Model 3, and Figure 2), but the overall conclusions remained the same.

DISCUSS ION
Population aging and legislative actions to extend working life in many countries in the EU and OECD area have rapidly increased the proportion of older workers in European labor markets (Axelrad & Mahoney, 2017, Eurostat, 2018).In this study, we considered an alternative approach to formal universal age-management policies, so-called idiosyncratic deals (Rousseau, 2001), that by improving person-job fit could potentially enhance opportunities for work in later life.We investigated the potential impact of I-deals on retirement preferences in the Swedish public sector as well as the overall prevalence and distribution of such arrangements across groups of workers.In general, and although the overall presence of I-deals appears to be reasonably limited in the present sample, our findings suggest that Task and Work Responsibilities and Workload Reduction are the only I-deals that significantly affect retirement preferences, albeit in opposite directions.Moreover, they indicate that certain groups of employees are more frequently granted I-deals than are others.
With respect to the potential influence of I-deals on retirement preferences, H1 posited that being granted I-deals would make employees more inclined to extend their working lives.This hypothesis was only partially supported by the data, as Task and Work Responsibilities and Workload Reduction were the only types of deals that demonstrated a significant association with retirement preferences.Additionally, and in contrast to our expectations, the latter was negatively related to the preferred age at retirement.Regarding Task and Work Responsibilities, we found that employees who received such deals generally had preferences for later retirement.The importance of skill development for employability and motivation to delay retirement is known in the existing literature (van Solinge & Henkens, 2014;Zaniboni et al., 2010).Furthermore, due to its conceptualization in the present study, Task and Work Responsibilities captures a range of aspects related to the individual's work situation.For instance, it assesses the "match" between employees' competence/experience and their current jobs as well as their ability to handle tasks that extend beyond their ordinary responsibilities.It also includes items that indicate the employee's opportunity for competence and skill development.Accordingly, employees granted such opportunities within the context of their current jobs are likely to experience higher levels of overall work satisfaction and control in relation to the tasks at hand.By extension, this could potentially reduce job demands and lower work-related stress (Siegrist, Wahrendorf, Von dem Knesebeck, & Jürges, 2007), making individuals more committed to their organizations (Herrbach, Mignonac, Vandenberghe, & Negrini, 2009) and thus inclined to work in late age (Armstrong-Stassen & Schlosser, 2008).As to Workload Reduction, this type of deal was shown to be related to a lower preferred retirement age.This finding was inconsistent with our original hypothesis, yet appears reasonable considering the negative association between Workload Reduction and self-rated health and the fact that Swedish employers have a statutory and far-reaching responsibility for making work adjustments when employees get sick or suffer from long-standing functional impairments (The Swedish Work Environment Authority, 2015).Thus, bearing the cross-sectional design of the study, this negative association likely occurs because individuals who report poor health have received workload reductions of some sort, while they also tend to have preferences for earlier retirement (Siegrist et al., 2007;Van Rijn, Robroek, Brouwer, & Burdorf, 2014).
With respect to the insignificant associations between retirement preferences and the remaining three I-deal domains (i.e., Schedule Flexibility, Location Flexibility, and Financial Incentives), we suggest that these could be understood with reference to, on one hand, the variation in working conditions across different occupational groups and, on the other, the organization of the public sector.First, flexibility in terms of location and working hours is generally limited in occupational groups in which work must be executed at a particular location and/or during specific hours every day (e.g., in routine and/or lower-grade service work), as evidenced by the significant associations with SEP found here.By extension, this might explain why neither Schedule Flexibility nor Location Flexibility appears to affect retirement preferences when we control for SEP.Regarding this, it should also be noted that scholars have questioned the extent to which survey questions intended to measure the presence of I-deals capture an actual negotiation (Conway & Coyle-Shapiro, 2015).For instance, they note the methodological challenges posed by coexistent (or absent) flexibility options that are not individually negotiated but instead are regulated through personnel policies, collective agreements, or labor market law.In this study, all I-deal indicators contain words such as "agreed," "discussed," "offered," "granted," "requested," "permitted," and "accepted," which would ideally imply explicit negotiation or agreement.However, the face validity analysis suggested that even though some respondents interpreted the questions as intended, that is, as referring to actual negotiations with  4. Continued the employer/manager, others answered the questions with reference to what was possible and available at the workplace in general and/or stated in the employment contract.Therefore, many of the participants likely "failed" to distinguish I-deals from other forms of universal HR policies or age-management strategies and instead viewed the flexibility options as a single bundle.This is probably a consequence of existing overlaps between employment contracts, HR policies, and labor market laws that make it hard for employees to distinguish whether a flexibility option is indeed an I-deal.The I-deal survey questions used here to capture Task and Work Responsibilities are more likely than the others to capture true negotiations or "investments" on the part of the employer, as they do not pertain to working conditions that are usually regulated in collective agreements or employment contracts.Consequently, respondents might be less likely to answer with more general conditions in mind, which implies a potential source of bias that could partly explain why this is the only I-deal dimension that is positively related to retirement preference.Regarding financial incentive deals, the extensive implementation of performance-based pay in the Swedish public sector has brought about some flexibility in salary setting that goes beyond collective agreements (Ulfsdotter Eriksson, Larsson, & Adolfsson, 2019).However, all employees are restricted to a standardized process in which salaries are reviewed on one occasion annually, which could limit the potential influence of such deals on employees' willingness to delay their labor market exit.Finally, in support of H2, our findings suggest that I-deals are usually less common in older age groups and among women.They also indicate that the working conditions of individuals who occupy lower SEPs and report poorer health are less often adjusted as a result of individual negotiations.More specifically, our analyses demonstrate that self-rated health is positively related to all five I-deal domains except Workload Reductions where a negative association was detected (see above).Also, we found SEP to be positively associated with three out of five I-deal domains (i.e., Schedule Flexibility, Location Flexibility, and Financial Incentives).Similar to Hornung et al. (2010Hornung et al. ( , 2014)), we thus conclude that "job level" (in our case indicated by SEP) constitutes a plausible and potentially important I-deal antecedent.Likewise, because both age and female sex are negatively related to several types of I-deals, they should also be regarded as central predictors and thus be acknowledged as such-in practice as well as in future studies.

Limitations and Methodical Considerations
This study is based on a cross-sectional sample of older public-sector employees.The public sector is dominated by women, which is reflected in the sample.However, the proportion of individuals with higher education was larger in the sample than in the sampled population as a whole.The study focus on retirement preferences which in research has been proven to be a relatively good proxy for actual retirement behavior (Nivalainen, 2020;Örestig et al. 2013), although this relationship may not be the same for all workers.For instance, a longitudinal study by Solem et al. (2016) found that older workers with poor health conditions and lower education leave the workforce earlier than they initially preferred.
The cross-sectional design also entails certain limitations; this was especially evident in our restricted ability to understand the negative association between Workload Reduction and retirement preferences.domains.For example, it is still somewhat unclear whether employees who are explicit with their employers about, for example, early retirement plans are more likely to strike deals concerning Task and Work Responsibilities, Financial Incentives, Schedule Flexibility, or Location Flexibility.These associations, as well as in-depth analyses of potential mechanisms linking certain types of I-deals to retirement intentions, would benefit from further explorations using a longitudinal approach with at least one follow-up.As to the measurement quality, some of the latent constructs were based on only two manifest indicators.Although some scholars think this is acceptable (Brown, 2015), others recommend three or more indicators (Little, 2013).In future studies, at least three indicators per I-deal construct would be desirable.Also, some I-deal domains that could potentially influence individuals' retirement preferences are not addressed here, for instance, social relations at work (e.g., social support and being allowed to work with certain teams/coworkers) or other forms of remuneration (e.g., car, insurance, and additional leave).Aside from the potential limitations concerning measurement validity described above, some of the respondents in our face validity analysis described how work tasks and working hours were occasionally negotiated with employer representatives other than their closest manager (e.g., foremen or schedulers).Therefore, in future studies, respondents should be provided with a detailed description that clarifies the meaning of explicit negotiations (and with whom such could take place), so that the I-deal measures omit task and work responsibilities, working conditions, flexibility options, and remuneration already included in employment contracts, available at the workplace in general, or stipulated in collective agreements.This is especially important within a Nordic context in which universal solutions are well established, encouraged, and part of many workers' mindset.To sum up, efforts to strengthen clarity and precision are essential in future empirical studies in order to achieve high measurement validity and reliability.

Implications
As the population ages, a larger proportion of retirees increases the need for welfare services.In Sweden, the public-sector estimates that over 500,000 workers will have to be recruited by 2025 to meet continuing welfare needs (Umegård, 2018).In sectors experiencing labor shortages, older workers constitute a potential labor reserve, and organizations' proactive behavior via HR management strategies is important to retain and attract older workers (Kadefors et al., 2020).In addition to labor market laws and organizational policies, I-deals constitute a potential informal and non-monetary intervention.I-deals could supplement traditional ways of meeting employees' needs by, for instance, giving lower managers the authority and resources to strike deals with their older employees.This study provides some guidance as to how public-sector organizations can prioritize and allocate idiosyncratic deals and what types of work arrangements have a potentially positive association with older workers' retirement preferences.Based on our results, employers interested in retaining older employees could improve the matching of competence and experience and offer opportunities for competence development as well as work tasks that extend beyond the employee's regular responsibilities.We argue that negotiating these aspects at work may increase motivation and facilitate older workers' person-job fit.Since the present data are based on public-sector employees, caution should be exercised in making inferences or claims regarding other labor market sectors and occupations.Finally, employers who set out to increase the prevalence of I-deals should be aware that doing so may challenge perceptions of fairness and equality (Conway & Coyle-Shapiro, 2015), which underlines the importance of paying close attention to whom they are granted.
Although efforts have been made to contextualize and compare I-deals in different contexts, they seem to have focused on situations within and between organizations and on the (dis)similarities in associations between factors and outcomes (Conway & Coyle-Shapiro, 2015, Liao et al., 2016).Apart from its empirical contribution, a theoretical implication of this study is the importance of contextualizing I-deals and considering the concept in relation to the pre-existing conditions that shape the opportunity structure for I-deals to take place.Earlier research on I-deals has primarily concentrated on what Liao et al. (2016) have referred to as Western (i.e., Germany, the United States, and the Netherlands) and Eastern (i.e., China, India, and South Korea) cultures.As far as the authors are aware, this is the first time that the full Ex Post I-Deals Scale developed by Rosen et al. (2013) has been empirically tested in Sweden and in a Nordic context.As previously noted, work accommodations in Sweden have historically been guided by collective agreements and strategic HR activities targeting aggregated groups rather than individual needs.However, such a universally oriented approach does not necessarily stop employees and employers from negotiating I-deals, given that doing so is sanctioned under the umbrella of collective agreements.At present, the large diversity between different sectors and branches in the Swedish labor market actualizes questions about whether collective agreements promote or hinder I-deals.In this respect, there is a need for more research.

CON CLUS ION
Five areas of idiosyncratic deals were tested to investigate their associations with employees' retirement preferences.We found that one I-deal domain-Task and Work Responsibilities-was positively associated with preferences for later retirement.Consequently, factors such as the matching of individuals' competence and experience, as well as growth opportunities, seem to be particularly important for publicsector employees' retirement preferences.
Downloaded from https://academic.oup.com/workar/article/7/3/240/6207921 by Umea universitet user on 21 July 2021 A similar inverse relationship between I-deals and retirement preferences cannot, theoretically, be entirely dismissed for other I-

Table 1 .
Demographics and Group Differences in Retirement Preference and I-Deal Indicators larger in the present study sample as well (78.3%,Table1).Likewise, the proportion of individuals with post-secondary education (73.0%) was higher than that estimated for the general population (35.2% among individuals 55-64 years old in 2018).The average respondent age was 59.8 years.Additionally, the participation rate varied somewhat between the different areas of municipal operations.For example, the participation rate among employees in social services was approximately 56%, whereas the corresponding figure among preschool workers was 37%.The study was approved by the Ethics Committee in the Gothenburg Region (EPN; Dnr: 2019-02934).
Note.Mean values refer to the average I-deal score measured on a 7-point Likert scale 1 (Not true at all) to 7 (Completely true) with the standard deviation for each indicator.Significance of group differences assed using the Kruskal-Wallis H test (>2 groups) and the Wilcoxon signed-rank test (2 groups).*p < .05. **p < .01. ***p < .001.Downloaded from https://academic.oup.com/workar/article/7/3/240/6207921 by Umea universitet user on 21 July 2021 All occupations were coded in accordance with the Swedish Standard Classification of Occupations 2012 (SSYK12; StatisticsSweden MIS 2012:1, 2012).These codes were later transformed into ISCO08 codes (International Labour Office, 2012), which, in turn, were used to create socioeconomic classes corresponding to the OESCH five-class schema The study utilized occupational class to indicate SEP, and information on respondents' occupations was obtained from the staff register of the municipality of Gothenburg.
tected in relation to three of the five dimensions: Location Flexibility (HSC vs. UW = 1.231, p < .001),Schedule Flexibility (HSC vs. UW = 0.345, p < .001),and Financial Incentives (HSC vs. UW = 0.229, p < .001).Furthermore, age was shown to be negatively and significantly associated with Location Flexibility (−0.063, p < .001)and Task and Work Responsibilities (−0.083, p < .001),but not with any of the other latent factors.Self-rated health was positively associated with Note. a Fully standardized coefficients (STDYX).*p < .05. **p < .01. ***p < .001.all Factor loadings excluded for clarity reasons, but were largely similar to those shown in Table3and are available on request.