Uneven Grounds: Class, Gender, and the Social Distribution of Work Flexibility

Abstract

This article analyzes the unequal distribution of perceived work flexibility in Chile, drawing on data from the 2023 National Time Use Survey (ENUT). Using multilevel multinomial logistic regression models, this study explores how individual and contextual variables shape the likelihood of perceiving jobs as fully flexible, partially flexible, or non-flexible. The findings reveal that flexibility is a stratified experience: women are more likely to access partial—but not full—flexibility; workers with higher educational attainment report lower perceptions of flexibility, and longer working hours are associated with diminished flexibility. Among the contextual factors, occupation stands out as the strongest predictor, explaining a substantial share of variance across the models. Employers are more likely to report full flexibility, while dependent workers tend to experience only limited forms of autonomy. These results challenge the notion of flexibility as a universally empowering feature of modern labor markets, underscoring the need for policies and organizational practices that address the structural inequalities embedded in flexible work arrangements.

1. Introduction

Work flexibility has become a central concept in contemporary labor debates, promoted as a mechanism for enhancing worker autonomy, improving work–life balance, and adapting to dynamic economic conditions. However, growing evidence indicates that access to flexibility is unevenly distributed, often reflecting and reproducing existing inequalities in labor markets. Rather than being a universal resource, flexibility can function as a stratified good—more accessible to certain groups based on their position in occupational, gender, and income hierarchies.

In Chile, the debate around labor flexibility is particularly relevant given the country’s segmented labor market and persistent gender and class-based disparities. While some occupations offer high autonomy and adaptable schedules, others are characterized by rigid structures and limited control over work conditions. Moreover, flexibility is not merely an objective feature of jobs; it is also a matter of perception, shaped by individuals’ socio-demographic positions and the organizational context in which they work.

This article explores the social distribution of perceived work flexibility in Chile using data from the 2023 National Time Use Survey (ENUT). We examine how individual characteristics—such as gender, age, education, income, and family responsibilities—interact with occupational and sectoral structures to shape subjective assessments of work flexibility. A key innovation of our study is the application of multilevel multinomial logit models that allow us to account for the hierarchical nature of the data and identify the contextual influence of occupation, sector, and macrozone.

Far from being a randomly distributed perk, our research is grounded in the premise that the lived experience of flexibility is profoundly shaped by structural forces. For instance, the ‘flexibility’ available to a senior manager is likely worlds apart from that of a part-time retail worker. Our work, therefore, seeks to explore this very tension—to challenge the notion of flexibility as a universal good and instead investigate it as a socially embedded phenomenon, one that often reflects and reinforces the very inequalities it is presumed to solve.

Based on the theoretical and empirical gaps identified in the preceding review, this study is guided by a central overarching question: How is the perception of work flexibility socially distributed across the Chilean labor market? To unpack this complex issue, we pose three more specific research questions:

To what extent do individual-level characteristics—such as gender, educational attainment, income, and family responsibilities—predict a worker’s perceived level of job flexibility?

What is the relative importance of structural and contextual factors, namely a worker’s occupation and economic sector, in shaping these perceptions?

How do these individual and contextual factors interact to create stratified patterns of access to, and experience of, work flexibility?

By combining a sociological focus on inequality with robust statistical modeling, this paper contributes to the literature on labor flexibility by challenging narratives that present it as uniformly positive or universally accessible. Instead, we conceptualize flexibility as a socially embedded phenomenon shaped by broader patterns of inequality. The following section reviews the theoretical and empirical literature on flexibility, stratification, gender, and occupation, as well as methodological approaches to modeling labor inequalities.

2. Literature Review

2.1. The Stratified Experience of Work Flexibility

Workplace flexibility has evolved from being viewed as a neutral accommodation to being recognized as a stratified phenomenon that mirrors and often deepens social inequalities. Scholars have identified the so-called “flexibility paradox,” where the formal availability of flexible arrangements does not guarantee equitable outcomes [1,2,3]. Research shows that flexibility can carry penalties: employees may internalize a “flexibility stigma” [4], and occupational gender segregation distorts wage perceptions and limits access to meaningful flexibility [5,6].

Educational attainment and occupational roles further structure these disparities. Petković et al. [7] argue that less educated workers are often excluded from flexibility-enhancing policies, while Carney and Junor [8] show that professional norms condition whether mothers experience flexibility as a resource or a constraint. Higher education may sometimes open access, but also increase expectations and rigidity [7,8]. Other studies underscore occupational tensions: teachers experience stress despite formal flexibility [9], while organizational demands can transform autonomy into pressure to conform [10].

Gendered expectations remain central. Nakano (in Close et al. [11]) shows how workplace policies can reinforce traditional roles when women bear disproportionate care burdens. Glauber [6] and Chung and Booker [12] confirm that access varies with the gender composition of occupations and can reinforce domestic divisions rather than challenge them.

Taken together, this literature demonstrates that flexibility is not a universal good but a socially embedded experience that consolidates hierarchies of gender, education, and occupation [1,2,3,4,5,6,7,8,9,10,11,12]. Yet despite these advances, research has paid limited attention to how these dynamics operate in Latin American labor markets characterized by segmentation and informality. Existing studies rarely integrate individual and contextual determinants within the same model, leaving a gap that this article addresses through a multilevel multinomial analysis of Chilean survey data [13].

2.2. The Gendered Foundation of Flexibility Experience

The most consistent finding across decades of research is that flexibility experiences are fundamentally gendered. Early work by Peterson and Wiens-Tuers [14] examining work time and gender inequality established that control over work time operates differently for men and women, though their analysis provided limited detail on specific mechanisms. This foundational observation has been substantially developed through subsequent research that reveals the complexity of gendered flexibility experiences.

Hofäcker and König’s [15] multilevel analysis of European workers during the economic crisis demonstrated that women and men use flexibility for fundamentally different purposes. Women predominantly mobilize flexible arrangements to manage work-family balance, while men more frequently use flexibility to intensify their work commitment. This differential usage pattern suggests that flexibility does not necessarily challenge traditional gender roles but may actually reinforce them.

The complexity of these gendered patterns is further illuminated by what Lopez-Martinez [1] identifies as the “flexibility paradox.” This paradox reveals that women often gain access to partial flexibility arrangements that, rather than enhancing their autonomy, serve primarily to facilitate their accommodation of family caregiving responsibilities. Kim et al. [16] provide empirical support for this paradox, showing that while flexible schedules may benefit overall well-being, women can simultaneously experience increased work-family conflict, suggesting that the purported benefits of flexibility are not equally distributed across gender lines.

Further complicating the gender flexibility phenomenon is the pervasive cultural concept of the “ideal worker” norm—the notion of a worker, implicitly male, who is perpetually available and prioritizes work above all else [17]. Research demonstrates that this norm powerfully shapes how flexibility is perceived and utilized within organizations. Even when gender-neutral flexibility policies are formally available, they often fail to disrupt this underlying cultural script. Consequently, women who use flexibility may be seen as less committed to their careers, reinforcing their position outside the “ideal worker” model, while men who use it for caregiving purposes can face a “flexibility stigma” for deviating from traditional masculine roles [3,18]. Recent scholarship suggests that overcoming such a problem requires more than just policy implementation; it demands active and supportive leadership that transparently challenges the ideal worker myth and frames flexibility as a gender-neutral tool for productivity and well-being [19].

In a more recent approach, researchers have extended this understanding by examining flexibility stigma. Thébaud and Pedulla [3] demonstrate through experimental methods that men face particular penalties for using work–family policies, especially in high-stigma organizational contexts. This finding reveals that flexibility experiences are shaped not only by access but also by the social and professional costs associated with utilization. The persistence of these gendered patterns across different national contexts suggests that gender operates as a fundamental axis of stratification in flexibility experiences.

2.3. Educational Credentials and the Complexity of Human Capital

The relationship between educational attainment and flexibility experiences reveals one of the most intriguing paradoxes in the literature. Conventional human capital theory predicts that higher education should translate into greater workplace autonomy and flexibility. Indeed, Galinsky et al. [20] documented that educational credentials provide access to jobs with formal flexibility policies.

However, more recent research has complicated this straightforward relationship. Lopez-Martinez [1] notes that in certain professional contexts, higher educational attainment can correlate with lower perceived flexibility due to the intensified expectations and demanding work cultures that accompany high-status positions. Women outnumber men in the U.S. college-educated workforce, now making up 51% of those ages 25 and older, yet this educational advantage has not translated into equivalent flexibility benefits. This paradox was amplified in the post-pandemic era, where educational attainment has become a primary sorting mechanism for access to remote work. Analysis shows that those with postgraduate degrees are vastly more likely to have remote work options compared to those with only a high school education, creating a new form of credential-based stratification [21].

This educational paradox suggests that credentials operate in complex ways within organizational hierarchies. While education may provide formal access to flexibility policies, it can simultaneously create expectations for availability and responsiveness that undermine workers’ actual sense of autonomy. The literature suggests that the relationship between education and flexibility is contingent on occupational context, organizational culture, and broader labor market conditions, necessitating analysis that moves beyond simple linear relationships.

2.4. Economic and Life Course Dimensions of Flexibility

Economic resources and life course dynamics jointly shape how flexibility is accessed and experienced. Evidence shows that the relationship between income and flexibility is far from straightforward. Ray and Pana-Cryan [22], using longitudinal data from the General Social Survey, demonstrate that the impact of income on flexibility satisfaction varies across demographic groups and sectors. Persistent gender pay gaps further complicate this relationship: despite women surpassing men in higher education attainment, they continue to earn only 82 cents for every male dollar, a ratio virtually unchanged since 2002 [23]. This enduring disparity limits women’s bargaining power and highlights that economic resources alone cannot explain flexibility outcomes.

Income operates as both facilitator and constraint. Higher earnings may provide security to negotiate flexible arrangements, but high-income positions often impose demanding expectations that undermine actual autonomy. Conversely, low-wage workers frequently encounter “flexibility” imposed by employers—manifesting as unstable schedules, last-minute changes, and on-call work—which is a stronger predictor of material hardship, including food insecurity, than low wages themselves [24]. These findings underscore the dual nature of economic resources and their intersection with occupational segregation and workplace discrimination.

Age and life stage further diversify flexibility needs. The life course perspective suggests that demands evolve as workers transition through different family and career stages. Younger cohorts, particularly Generation Z, express stronger preferences for flexibility aimed at work–life balance, valuing control over workspace and task variety rather than over hours [25]. Parental status intensifies these dynamics: mothers, in particular, often face additional constraints that convert flexibility from a potential resource into a limited or conditional choice. Such patterns illustrate that flexibility is not static but changes with age, family responsibilities, and broader socio-economic conditions.

Taken together, the literature shows that flexibility is deeply conditioned by both economic inequality and life course trajectories. Income and age interact with gender, occupation, and family status to produce stratified outcomes, yet systematic analyses integrating these dimensions remain scarce—especially in Latin American labor markets. This gap motivates our study’s focus on Chile, where ENUT 2023 data allow for simultaneous examination of these intersecting determinants.

2.5. Temporal and Occupational Structures of Flexibility

The configuration of working time and occupational hierarchies represent two of the most powerful structural determinants of flexibility experiences. Research indicates that full-time employment may paradoxically provide greater access to certain forms of flexibility compared to part-time work, which often carries precarity and reduced benefits [20]. These findings highlight that temporal arrangements interact with employment status in ways that complicate assumptions about work intensity and autonomy. The capacity to exercise “boundary control” is especially critical: workers who can manage their temporal boundaries report higher levels of engagement, satisfaction, and work–life balance, underscoring how working hours function as both a resource and a constraint in realizing flexibility benefits [20].

Beyond temporal dynamics, occupational position fundamentally structures access to workplace autonomy. Kossek and Lautsch [2] conceptualize flexibility as a form of job inequality, with advantaged managers and professionals reporting substantially higher levels than those in routine or service jobs [20]. For lower-level workers, flextime could provide significant benefits, yet they are often subject instead to unstable scheduling practices—including last-minute changes and on-call shifts—that are directly linked to psychological distress, poor sleep, and reduced well-being [26,27].

Ethnographic evidence underscores these divides. Gerstel and Clawson [28] show how even within healthcare, autonomy varies widely among doctors, nurses, technicians, and support staff, reflecting professional status and organizational power. Comparative studies extend this insight across Europe, where female-dominated occupations systematically offer less flexibility than male-dominated ones, perpetuating both occupational segregation and gender inequality [29]. Employment relations add another layer: employers may enjoy formal autonomy but face market pressures; employees often encounter organizational limits on policy use; and freelancers, despite nominal control, are constrained by economic insecurity [28,29].

Taken together, the literature demonstrates that flexibility outcomes are inseparable from both temporal constraints and occupational hierarchies. Long working hours, unstable schedules, and segmented occupational structures reinforce stratification, positioning autonomy as a privilege of higher-status groups while subjecting lower-wage and feminized occupations to precarious forms of flexibility [2,20,26,27,28,29].

2.6. Sectorial and Regional Variations and Contemporary Developments and Emerging Patterns

The Chilean case illustrates vividly how global debates on flexibility are mediated by national structures of inequality. Since the neoliberal reforms of the 1980s, labor relations in Chile have been reorganized around subcontracting, outsourcing, and temporary contracts, which provided firms with numerical flexibility while transferring risks onto workers [30,31]. This process deepened a dual labor market, long described in Latin American scholarship, where a privileged segment of workers access stability, protection, and career mobility, while a secondary segment remains confined to low-quality jobs marked by insecurity, poor wages, and limited rights [32,33,34]. Approximately one-third of the workforce falls within this secondary segment, making segmentation a structural determinant of how flexibility is experienced [35]. In this context, the distinction between “positive” flexibility—autonomy over time and place of work—and “precarious” flexibility—employer-driven instability—maps directly onto class divides.

Gender inequalities reinforce and complicate these dynamics. Persistent occupational segregation means that women remain concentrated in feminized and undervalued sectors such as domestic service, commerce, and health, and within them disproportionately occupy precarious and low-status roles [36,37,38]. The 2023 National Time Use Survey (ENUT) shows that women dedicate more than two additional hours per day to unpaid care and household work compared to men [13]. This asymmetry channels many women into part-time employment and flexible schedules intended to reconcile work and family responsibilities. Yet, rather than enhancing autonomy, such arrangements often consolidate women’s disadvantage, limiting wages, social protection, and career progression [39]. The sharp fall of female labor force participation below 50% during the pandemic [40] underscored how fragile these arrangements can be when structural care burdens collide with crises. In this sense, the Chilean evidence strongly resonates with international debates on the “flexibility paradox,” where arrangements formally aimed at work–life balance reproduce traditional gender norms [1,41,42].

Educational and occupational hierarchies further stratify access to meaningful flexibility. ENUT 2023 confirms that university-educated workers disproportionately benefit from remote work and other forms of schedule autonomy [13]. During the COVID-19 pandemic, teleworking opportunities were almost exclusively available to white-collar professionals, while lower-skilled workers experienced flexibility through unstable shifts, on-call hours, or the need to combine multiple informal jobs [43,44]. The OECD [45] similarly notes that precarious forms of flexibility remain concentrated among low-income groups across Latin America, reinforcing inequality rather than reducing it. In Chile, temporal autonomy thus emerges less as a universal right than as a privilege reserved for those with higher education and stronger bargaining power.

These patterns are not only sociological but institutional. Chile’s Labor Code historically sanctioned long working hours and broad exemptions, favoring employer discretion [46]. Law 21.220 on Telework recognized remote work as a legitimate modality but allowed employers and employees to exempt themselves from ordinary working-time limits, raising concerns about unpaid overwork [47,48]. In 2023, the 40-Hour Law (Law 21.561) reduced the statutory workweek to 40 h, introduced conciliation rights such as a two-hour daily band for parents of young children, and narrowed exemptions for teleworkers, requiring most remote workers to record their hours [49]. While these reforms seek to balance flexibility with protection, some analysts have noted that stricter regulation could discourage employers from offering telework [50,51]. Taken together, these institutional adjustments reveal Chile’s search for a regulatory equilibrium between empowering and precarious flexibility.

Anchoring the analysis in the Chilean context thus demonstrates that flexibility is not an abstract or neutral innovation but a phenomenon embedded in segmented labor markets, gendered divisions of labor, educational hierarchies, and national regulatory frameworks. The results from ENUT 2023 confirm that women, subcontracted workers, and those with lower education are systematically restricted to constrained and often precarious forms of flexibility, while professionals and high-status groups benefit from autonomy. Far from leveling inequalities, flexibility in Chile mirrors and magnifies the country’s entrenched social divides.

2.7. Synthesis and Implications

The literature reveals that workplace flexibility is not a uniform experience but a stratified phenomenon that reflects and reproduces existing social inequalities. Gender emerges as the primary axis of differentiation, with women more likely to experience partial rather than full flexibility and to use flexibility for family accommodation rather than career advancement. Educational credentials show complex, sometimes paradoxical relationships with flexibility, providing access to formal arrangements while potentially increasing expectations that limit actual autonomy.

Economic resources, measured through income and working hours, interact with other factors in complex ways that defy simple predictions. Age and parental status introduce life course dynamics that shape flexibility needs and experiences over time. Most importantly, structural factors including occupation, economic sector, and regional context appear to be the strongest determinants of flexibility experiences, suggesting that individual characteristics operate within powerful contextual constraints.

The employment relationship itself creates fundamental differences in flexibility experiences, with employers, employees, and freelance workers facing different forms of autonomy and constraint. Geographic and sectoral variations indicate that flexibility is embedded within broader institutional contexts that shape both access and utilization. These patterns suggest that empirical analysis of flexibility experiences requires attention to both individual characteristics and contextual factors, with particular emphasis on the structural determinants that appear to exercise the strongest influence on worker experiences. The stratified nature of flexibility indicates that policies aimed at promoting workplace flexibility must address the underlying inequalities that shape access to and benefit from flexible work arrangements.

3. Materials and Methods

3.1. Data Analysis and Model Fitting

The complexity revealed in this literature suggests that flexibility should be understood not as a simple organizational benefit but as a socially embedded experience that both reflects and shapes broader patterns of workplace inequality. This understanding provides the foundation for empirical analysis that can illuminate how these patterns operate within specific national and institutional contexts.

The analysis is based on the results of the National Time Use Survey (ENUT). The main goal of this tool is to characterize how individuals spend their time on various activities in their daily lives. The results of the survey provide insights into the activity patterns of men and women in the workplace, household maintenance, and broader societal participation [52].

As stated above, this study uses microdata from the 2023 Chilean National Time Use Survey [13], a nationally representative dataset that captures labor market dynamics and household conditions. From the full dataset, we constructed a set of variables to capture flexible work arrangements and key sociodemographic factors. The features employed for the analysis are shown in

Table 1. Description of variables from 2023 ENUT used in the analysis.

Feature	Type	Levels
Flexibility level	Categorical	Not flexible: 0 Partially flexible: 1 Very flexible: 2
Gender	Categorical	Male: 0 Female: 1
Age	Numerical
Education level	Categorical	Primary education or no education: 0 Secondary education: 1 Technical education: 2 Undergraduate or graduate education: 3
Income	Numerical
Working hours	Numerical
Regional zone	Categorical	North: 1 Center: 2 South: 3 Metropolitan Region: 4
Economic sector	Categorical	Primary: 1 Industrial: 2 Construction: 3 Trade/Transportation: 4 Services: 5 Public administration: 6 Other: 7
Occupation	Categorical	Employer: 1 Freelance worker: 2 Employed or familiar worker: 3

. The dependent variable was defined as a multinomial indicator with three categories: not flexible (0), partially flexible (1), and highly flexible (2). It was based on the presence of up to three conditions: (a) telework on weekdays or weekends, (b) shift work, and (c) working fewer than 30 h per week. Explanatory variables included gender (male/female), age (continuous), educational attainment (basic or less, secondary, tertiary), household income quintiles, presence of children in the household (yes/no), macro-zone of residence (North, Center, South, Metropolitan Region), sector of economic activity (seven categories), and employment status (employer, self-employed, dependent/family worker).

As shown in Table 1, the flexibility variable includes more than two categories. To model this outcome, multinomial logit models are employed, as they can handle variables with multiple levels. Furthermore, due to the hierarchical structure of the data, the regional zone, economic sector, and occupation are treated as higher-level variables. Consequently, multilevel multinomial logit models are used. Model selection is based on a comparison using information criteria to identify the model that best represents the phenomena. Data processing, analysis, and model fitting are conducted using version 4.4.2 of the R programming language.

3.2. Multilevel Multinomial Logistic Regression Model

Finally, recent research has identified several emerging trends that are reshaping flexibility experiences. The percentage of women in the C-suite has increased from 17% to 28% from 2015 to 2023, the research shows, while the representation of women at vice president and senior vice president level has also improved; yet women continue to face barriers in reaching senior roles despite increased flexibility options. The study confirms that work flexibility plays a crucial role in job satisfaction, with employees preferring structured autonomy over rigid or excessively flexible work arrangements. This finding suggests that the optimal flexibility experience involves clear boundaries and predictable autonomy rather than unlimited flexibility. The psychological dimensions of flexibility have also received increased attention. Results show that a flexible workplace culture, but not access to flextime or flexplace, is associated with lower psychological distress, indicating that organizational culture may be more important than formal policies in determining flexibility outcomes.

The Multilevel Multinomial Logit (MMNL) model, also referred to as the hierarchical multinomial logit model, is a statistical approach designed to analyze categorical outcomes with more than two categories. It is particularly suitable for data with a hierarchical structure, such as individuals nested within groups. Unlike the standard multinomial logit model, the MMNL incorporates random effects to capture variability across higher-level units [53,54,55]. This feature helps reduce parameter overestimation during the estimation process [55]. Moreover, multilevel models yield more accurate and precise parameter estimates in the presence of clustered data structures [56], as they account for group-level effects and exploit information shared across similar clusters to better capture underlying phenomena [57]. Given that our outcome variable has multiple categories and the survey data are hierarchically organized, the MMNL framework is especially appropriate. Neglecting such structures would result in underestimated standard errors, inflated Type I errors, and biased coefficients due to unmodeled group-level variation [56].

To understand the MMNL we stand first with the basic multinomial logit model. Let $y_i\in \{1,\dots ,K\}$ be the categorical outcome for observation $i$, let ${\mathrm{x}}_i\in {\mathbb{R}}^p$ w here $p$ denotes the number of features. To model the probability of each flexibility level, we start with the basic multinomial logit formulation as shown in (1).

$$P\left(y_i=k\right)={\displaystyle \frac{1}{1+{\sum }_{j=1}^{K-1}\mathrm{e}\mathrm{x}\mathrm{p}\{-{{\mathrm{x}}_i}^T{\beta }_j\}}}$$

(1)

Suppose the data are nested in groups, so the individual $i$ belongs to the group $j$. We allow the coefficients ${\beta }_j$ shown in (1) to vary by group $j$. The probability of choosing a category $k$ for the individual $i$ in the group $j$ can be written as (2).

$$P\left(y_{ij}=k\right)={\displaystyle \frac{\mathrm{e}\mathrm{x}\mathrm{p}\left\{{-{\mathrm{x}}_{ij}}^T{\beta }_{jk}\right\}}{{\sum }_{l=1}^K\mathrm{e}\mathrm{x}\mathrm{p}\left\{{-{\mathrm{x}}_{ij}}^T{\beta }_{jl}\right\}}}$$

(2)

where ${\beta }_{jk}={\beta }_k+u_{jk}$, such that $u_{jk}\sim 𝒩(0,{\mathsf{\Sigma }}_k)$ represents the random effects for group $j$ and category $k$. This structure captures group–level variations in category–specific effects [58].

Since the data is hierarchical, with observations nested within groups, the Intraclass Correlation Coefficient (ICC) is computed to measure the proportion of total variability in the target variable that is attributable to differences between groups [58]. The ICC quantifies how much of the variability is due to differences between groups (e.g., occupations) and is presented in (3).

$$\mathrm{ICC}={\displaystyle \frac{{{\sigma }_{group}}^2}{{{\sigma }_{group}}^2+\frac{{\pi }^2}{3}}}$$

(3)

The values of the ICC range from 0 to 1, where a value near 0 indicates most of the variability is within groups, and a value near 1 indicates that most of the variability is between groups. It is worth noting that in (3) the term ${{\sigma }_{group}}^2$ represents the cluster variance, and ${\pi }^2/3$ is the variance of the logistic distribution [59].

3.3. Information Criteria

As stated above, a comparison of different model specifications is conducted using information criteria The metrics employed for this purpose are the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). Both criteria assess the complexity and fit of each model, helping to identify the best option based on its specification and overall performance. The AIC calculation is given by (4) where $\mathcal{L}$ denotes the likelihood function of the model fitted with a set of parameters $p^{\prime }$ [60,61].

$$\mathrm{AIC}=-2\log \left(\mathcal{L}\right)+2p^{\prime }$$

(4)

The BIC metric calculation is given by (5), where $n$ denotes the number of observations presented in the data [62,63].

$$BIC=-2\log \left(\mathcal{L}\right)+p^{\prime }\log \left(n\right)$$

(5)

The Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) are commonly used measures to assess model complexity and fit. Their main difference lies in the penalty applied to model parameters: BIC imposes a stronger penalty, thereby favoring simpler models, while AIC is more efficient for predictive accuracy. In both cases, lower values indicate a better balance between goodness of fit and model complexity [63].

3.4. Balancing Underrepresented Categories

Balancing class weights in the MMNL model involves adjusting the loss function to account for class imbalance, particularly when some outcome categories are underrepresented or occur more frequently than others. This is achieved by reweighting the likelihood function, ensuring that underrepresented classes are not overshadowed by the overrepresented ones. As shown in (6) with the term $w_k$ that represents the weight of the class $k$ with respect to the rest of the data and is measured as the inverse marginal probability weight of the category. This ensures that each class contributes equally to the likelihood, being a form of balancing that simulates a pseudo-population where the outcome classes are equally likely [64]

$${\mathcal{L}}_{weighted}={\sum }_j{\sum }_i{\sum }_k{w}_{k}1\left(y_{ij}=k\right)\log \left({\displaystyle \frac{\exp \left\{-{{\mathrm{x}}_{ij}}^T{\beta }_k\right\}}{{\sum }_{l=1}^K\exp \left\{{-{\mathrm{x}}_{ij}}^T{\beta }_{jl}\right\}}}\right)$$

(6)

It is worth noting that the weights $w_k$ apply to the outcome category, not to the group–level effects.

4. Results

To ensure consistency across models, the analytic sample was restricted to individuals with complete information on the dependent variable and the main explanatory variables. After applying this criterion, the final working sample consisted of 375 individuals. This subsample does not represent the entirety of ENUT 2023 but rather those cases with full data availability; therefore, results should be interpreted within this analytical scope

Figure 1 shows the distribution of individuals by gender, categorized by their perceived level of flexibility. Males and females each represent approximately 50% of the sample. Among females, there is a higher concentration in the “partially flexible” category compared to males. Conversely, a greater proportion of males fall into the “not flexible” category.

Figure 2 shows the distribution of surveyed individuals by education level. Individuals with technical education represent approximately 30% of the sample, followed by those who completed secondary education (29%) and those with complete or incomplete primary education (25%). The distribution of individuals in the “very flexible” category is relatively uniform across all education levels. However, the technical and secondary education groups exhibit a higher concentration of individuals who perceive their flexibility as “not flexible,” compared to those with primary or undergraduate education.

The distribution of individuals with and without children is shown in Figure 3. Approximately 55% of the individuals do not have children. Moreover, the distribution of flexibility levels is nearly identical across both groups. In both the “with children” and “without children” categories, around 5% of individuals fall into the “not flexible” category, approximately 15% into the “partially flexible” category, and about 80% into the “very flexible” category.

Most individuals are concentrated in the southern and central regions of Chile, accounting for approximately 80% of the sample, as shown in Figure 4. The northern region represents 37% of the data, while the Metropolitan Region accounts for about 4%. The southern region shows a higher concentration of individuals reporting a “not flexible” level, and it also has the highest concentration of individuals in the “partially flexible” category.

As shown in Figure 5, almost half of the sample belongs to the trade and transportation sector, followed by services (22%) and the industrial and construction sectors (around 11% each). Regarding the economic sector in which individuals are employed, approximately 49% of the sample is concentrated in trade and transportation, followed by services (22%), and both the industrial and construction sectors, each representing around 11%. All responses from individuals in the primary or other sectors fall into the “very flexible” category. In the construction sector, about 86% of individuals report a “very flexible” level. Trade and transportation show the lowest proportion of responses in the “very flexible” category.

Figure 6 shows the distribution of individuals by occupation. Half of the responses come from individuals in the “Employer” category, while “Own Business” accounts for 23%, and “Dependent/Familiar” represents 18% of the sample. The distribution of flexibility levels is similar for the “Own Business” and “Dependent/Familiar” categories, whereas the “Employer” category shows a higher concentration of individuals reporting a “Not Flexible” level.

Figure 7 presents the distribution of numerical features across perceived flexibility levels using boxplots for each category. In terms of age, individuals who perceive their flexibility as “partially flexible” tend to be older than those in the other two categories. Income levels do not show substantial differences in central tendency; however, the spread of income is wider among individuals who perceive their work as “very flexible.” Finally, working hours per week follow a clear trend: as weekly hours increase, the perceived level of flexibility tends to decrease.

4.1. Model Estimation

The model estimation parameters are presented in

Table 2. Parameters result for the MMNL models fitted considering different variables in the higher level.

	Variable at the High Level
Parameter	Macrozone		Occupation		Economic Sector
	Partially Flexible	Very Flexible	Partially Flexible	Very Flexible	Partially Flexible	Very Flexible
ICC	0.0128	0.0437	0.2001	0.2129	0.0325	0.2569
AIC	2045.4		1986.9		2044.1
BIC	2151.0		2092.5		2149.6

. The ICC, AIC, and BIC metrics are calculated to compare models with different specifications of random effects. The model that includes macrozone as a random effect yields ICC values close to zero, indicating that only about 1% to 4% of the variance is explained by differences between macrozones. In contrast, the model incorporating the economic sector as a random effect explains between 3% and 25% of the variance across groups. The model that includes occupation as a random effect is the most consistent in explaining variance due to cluster-level differences, with ICC values of 20% for the “partially flexible” category and 21% for the “very flexible” category.

Regarding information criteria, the model with macrozone as a random effect has the worst AIC and BIC values. The model with the economic sector shows similar performance to the macrozone model. In contrast, the model with occupation as a random effect produces the lowest AIC and BIC values, indicating the strongest grouping effect. Therefore, occupation is selected as the random effect variable for the final inference procedure.

4.2. Model Interpretation and Inference

The parameters estimated from the MMNL, along with the significance and respective odds ratios are shown in

Table 3. Results on the MMNL model for the partially flexible level outcome.

Feature	Group	Estimate	p-Value	Odds–Ratio Probability
(Intercept)		$-1.14$	$0.53$	$0.32$
Gender (male)	Female	$0.86$	$~0.0*$	$2.36$
Age		$0.02$	$~0.0*$	$1.02$
Log(income)		$0.32$	$0.01*$	$1.38$
Children (No)	Yes	$-0.12$	$0.47$	$0.89$
Working hours		$-0.75$	$~0.0*$	$0.47$
Education level (Primary or incomplete)	Secondary	$-1.43$	$~0.0*$	$0.24$
	Technical	$-1.56$	$~0.0*$	$0.21$
	University	$-1.08$	$~0.0*$	$0.34$

“*” indicates statistical significance at the 0.05 level. “~0.0 *” denotes p-values smaller than 0.001 *.

for the comparison of partially flexible vs. not flexible. Females have 2.36 times higher odds of reporting a partially flexible than males, holding other variables constant. Each additional year of age slightly increases the odds of reporting a partially flexible perception of 2%, a significant result with small effect. Higher income is associated with higher odds of flexible work by 38%. Having children shows results that are not statistically significant. As the working hours increase the lower the odds of reporting a partially flexible perception by 53% less. All the categories in education level are significant. Individuals with secondary education have 76% lower odds of reporting a partially flexible perception. Technical education is associated with 79% lower odds compared to primary education. University–educated individuals have 66% lower odds of reporting a partially flexible perception compared to those with primary or incomplete education.

The parameters estimated for the comparison of very flexible vs. not flexible of the MMNL are shown in

Table 4. Results on the MMNL model for the very flexible level outcome.

Feature	Group	Estimate	p-Value	Odds–Ratio Probability
(Intercept)		$5.16$	$~0.0*$	$174.16$
Gender (male)	Female	$0.10$	$0.59$	$1.10$
Age		$0.02$	$0.01*$	$1.02$
Log(income)		$0.48$	$~0.0*$	$1.62$
Children (No)	Yes	$-0.34$	$0.06*$	$0.71$
Working hours		$-2.89$	$~0.0*$	$0.06$
Education level (Primary or incomplete)	Secondary	$-1.17$	$~0.0*$	$0.31$
	Technical	$-1.21$	$~0.0*$	$0.29$
	University	$-1.14$	$~0.0*$	$0.32$

“*” indicates statistical significance at the 0.05 level. “~0.0 *” denotes p-values smaller than 0.001 *.

. The gender shows no significant results. Like the previous results, each additional year of age increases the odds of reporting a very flexible perception by 2%. The income shows a stronger effect compared to the partially flexible outcome. Having children decreases the odds of reporting a very flexible perception by 29%. Each additional working hour per week shows 94% lower odds of reporting a very flexible perception. All the education levels show significant results with lower odds about of 70% reporting very flexible compared to primary or incomplete education.

In summary, older individuals show higher odds of reporting flexibility, and income has a similarly positive effect, particularly at the “very flexible” level. In contrast, longer working hours and higher educational attainment are consistently associated with lower perceptions of flexibility. The random effects from the multilevel model, presented in

Table 5. Results of the random effects in the MMNL model for partially and very flexible outcomes.

	Partially Flexible			Very Flexible
Level	Estimate	Odds–Ratio	Level	Estimate	Odds–Ratio Probability
Employer	$-0.90$	$0.41$	Employer	$1.29$	$3.32$
Own business	$-0.58$	$0.56$	Own business	$-0.31$	$0.73$
Dependent/Familiar	$1.21$	$3.35$	Dependent/Familiar	$-0.72$	$0.49$

, reveal marked differences by occupational status. Employers are unlikely to perceive partial flexibility but are much more likely to report full flexibility, suggesting a polarization toward complete autonomy. Business owners exhibit reduced odds of perceiving either partial or full flexibility, possibly reflecting structural constraints or the burdens of responsibility. Finally, dependent workers show a strong association with partial flexibility but a negative association with full flexibility, underscoring their restricted access to broader forms of autonomy.

5. Discussion

The present study offers a robust empirical contribution to the growing literature on the stratification of work flexibility, reaffirming that flexibility is not a universally experienced benefit but rather a socially distributed resource shaped by structural inequalities. Drawing on nationally representative data from the 2023 Chilean Time Use Survey and leveraging multilevel multinomial logistic modeling, our results reveal that flexibility is systematically associated with both individual and contextual variables—most notably, occupation. This aligns with prior scholarship emphasizing the embeddedness of flexibility within occupational hierarchies and broader institutional arrangements [2,28].

In general terms, our results suggest that perceived flexibility is not randomly distributed but is strongly associated with structural variables. For example, being female increases the odds of perceiving partial flexibility but not full flexibility, while higher education levels are paradoxically associated with lower perceptions of flexibility. Among contextual variables, occupation emerges as the strongest cluster-level predictor, highlighting the importance of job type in shaping the lived experience of work autonomy.

A particularly salient finding is the asymmetric effect of gender on flexibility perceptions. While women are significantly more likely to report partially flexible work conditions, they are not more likely to report full flexibility. This finding lends empirical support to the “flexibility paradox” [1], in which access to flexible arrangements often facilitates caregiving responsibilities rather than expanding professional autonomy or advancement. Women’s access to flexibility thus appears to be structured by societal expectations around care work, reinforcing traditional gender roles even in contexts where formal flexibility is present [3,18]. In Chile, this asymmetry is further intensified by the country’s care regime. ENUT 2023 shows that women dedicate more than two additional hours per day to unpaid domestic and care work compared to men [13]. This unequal distribution of household responsibilities is particularly pronounced in rural and southern regions, where public childcare infrastructure is scarcer [36,37]. As a result, partial flexibility for women often translates into constrained arrangements aimed at accommodating family responsibilities rather than expanding career opportunities.

Moreover, the results complicate traditional assumptions drawn from human capital theory. Despite the expectation that educational attainment enhances access to better job conditions, including flexibility, we find that individuals with university or technical education levels report significantly lower odds of perceiving their work as flexible compared to those with only primary education. These findings resonate with emerging critiques that link higher education with heightened job demands and reduced boundary control, particularly in high-status occupations [21,24]. This suggests that the credentialing process may facilitate entry into work environments that paradoxically constrain rather than expand subjective experiences of autonomy. In Chile, this paradox has a territorial component: university-educated professionals in Santiago and other metropolitan centers are more likely to access telework and flexible schedules, yet under conditions of intensified workloads and “always-on” expectations [43,44]. Outside the capital, however, less-educated workers are disproportionately exposed to “flexibility” as employer-driven instability—on-call shifts, subcontracting, and rotating schedules—documented in national reports as drivers of insecurity and material hardship [31,39].

Another noteworthy pattern concerns working hours. Our analysis confirms that longer weekly hours are strongly associated with diminished perceptions of both partial and full flexibility. This finding aligns with research on temporal boundaries and burnout, where higher work intensity undermines perceived control over schedules [20]. It also reinforces the argument that flexibility should not be equated with mere schedule modification but understood in terms of broader structural conditions, including workload and expectations for constant availability. In Chile, such a phenomenon is consistent with the long-standing orientation of the Labor Code, which historically favored employer discretion over working-time regulation [46]. Recent reforms—such as Law 21.220 on telework [47] and the 40-Hour Law [49]—have sought to strike a balance by introducing conciliation rights and narrowing exemptions [47,49]. Still, analysts warn that these reforms may create new risks of unpaid overwork or employer reluctance to extend telework arrangements [51].

Among contextual factors, the multilevel model reveals that occupation explains the greatest proportion of variance in perceived flexibility, far surpassing the effects of macrozone or economic sector. Employers report a polarized pattern, with significantly higher odds of perceiving full flexibility but lower odds of perceiving partial flexibility. Conversely, dependent workers are more likely to perceive partial flexibility but not full autonomy. This polarization underscores the central role of job authority and organizational control in shaping subjective experiences of flexibility. While the public discourse often promotes flexibility as a neutral or egalitarian innovation, our findings demonstrate that such experiences are deeply conditioned by position within occupational hierarchies. In the Chilean case the segmented character of the labor market is long documented in Latin American scholarship [32,33,34]. The distinction between “positive flexibility”—autonomy over time and place of work—and “precarious flexibility”—employer-driven instability—maps directly onto class divides [35].

Regional inequalities provide further nuance. Although macrozone random effects account for only a modest share of overall variance (1–4%), descriptive evidence shows that workers in the south are disproportionately concentrated in the “partially flexible” and “not flexible” categories. This outcome reflects Chile’s long-standing territorial disparities in development, where northern and southern regions are more dependent on extractive industries and precarious service jobs, while the Metropolitan Region concentrates opportunities for stable professional and public-sector employment. These findings resonate with comparative analyses showing that geographic disparities strongly shape access to flexibility in Latin America and Europe alike [29,40,45]. Thus, our results confirm that flexibility is not merely an organizational attribute but also a phenomenon embedded in Chile’s uneven regional labor markets.

Taken together, these findings suggest that work flexibility in Chile is a stratified and contextualized phenomenon that mirrors broader patterns of labor market inequality. Future research would benefit from longitudinal data to track changes over time, as well as qualitative methodologies to explore the meaning-making processes behind flexibility perceptions. Additionally, there is a need for organizational and policy interventions that not only promote formal access to flexibility but also mitigate the structural barriers that prevent its equitable use. Without such efforts, flexibility may continue to reproduce, rather than reduce, the inequalities it is often presumed to address.

While this study offers a robust analysis of the social stratification of perceived work flexibility in Chile, it is important to acknowledge its limitations, which in turn open avenues for future inquiry.

First, the sample size of 375 individuals, while sufficient for our multilevel modeling approach, is relatively modest. This means that while our findings are statistically significant and internally valid for the population sampled, caution should be exercised when generalizing them to the entire Chilean workforce. Furthermore, the survey’s sampling frame may not fully capture the experiences of those in the informal labor market. Given that informal work is often characterized by extreme precarity and a complete lack of worker control, its inclusion would likely reveal even starker patterns of inequality than those we have documented here.

Second, our operationalization of flexibility centers on workers perceived autonomy over their schedules. This is a critical dimension, but it does not encompass the full spectrum of what flexibility means in contemporary work. For instance, our measure does not distinguish between empowering, worker-initiated flexibility (e.g., choosing one’s own hours) and precarious, employer-initiated flexibility (e.g., last-minute shift changes or on-call work). Future research could employ more granular measures to differentiate between schedule control and schedule predictability, which are often inversely related, particularly in lower-wage service jobs. Similarly, the growing importance of remote work as a key facet of flexibility is a dimension not explicitly captured in our data.

Finally, our reliance on a subjective measure of “perceived” flexibility, while valuable for capturing lived experience, may mask hidden constraints. A worker might report high levels of flexibility because they are formally allowed to set their own hours, yet they may simultaneously be subject to an “always-on” culture that demands constant availability. This “illusory autonomy” is a phenomenon our survey cannot detect. Qualitative and ethnographic studies would be invaluable in complementing our quantitative findings, allowing researchers to explore the meaning-making processes behind flexibility and uncover the subtle negotiations and trade-offs that workers face, moving beyond perception to the lived reality of autonomy at work.

6. Conclusions

This study set out to examine the social distribution of perceived work flexibility in Chile, using nationally representative survey data and multilevel multinomial logit models to account for both individual and contextual determinants. So, after all the data has been analyzed, what is possible to learn about the promise of work flexibility in Chile? Findings suggest that flexibility is less of a universal workplace perk and more of a mirror, reflecting the deep and persistent fractures within the Chilean labor market. Far from being an equalizing force, it appears to be a resource that is hoarded, rationed, and distributed along the well-worn lines of gender, education, and, most powerfully, occupational class.

The results underscore the limitations of framing flexibility as a neutral, empowering organizational practice. The story for women, for instance, is a particularly bittersweet one. Yes, they are more likely to gain access to some flexibility, but our results reveal this is often a constrained version—a kind of flexibility on a leash, designed more to accommodate the enduring expectations of caregiving than to foster genuine professional advancement. In a society still grappling with traditional gender roles, this is not just a workplace issue; it is a social one, subtly reinforcing the very structures that limit women’s careers in the first place.

Additionally, this study shows that the very university and technical degrees that are held up as the primary engine of social mobility in Chile can paradoxically lead to less perceived autonomy. It is a bitter pill to swallow. This suggests that for many Chilean professionals, climbing the corporate ladder means stepping into a gilded cage—a world of higher pay, but also of crushing demands and an “always-on” culture that erodes the freedom that was supposed to come with success. Indeed, the strong association between longer working hours and lower flexibility perceptions indicates that time demands continue to act as a major constraint on autonomy.

These findings carry important implications for policy and organizational design. In the Chilean labor market, when it comes to flexibility, your occupation is your destiny. The fact that this single factor overshadowed everything else in our model speaks volumes about the country’s rigid social stratification. We saw this in the stark polarization between employers, who either enjoy complete autonomy or are shackled by market pressures, and their dependent workers, who are often granted only scraps of control over their time. This is not just a data point; it is the lived reality of the immense chasm between different tiers of the workforce.

What this all means is that simply creating more “flexible work” policies on paper is a profoundly inadequate solution. It is like applying a band-aid to a deep structural wound. Real change requires confronting the institutional and cultural forces that make flexibility a privilege, not a right. Policymakers must focus on protecting the most vulnerable—especially low-wage workers—from the tyranny of schedule instability. And organizations, for their part, must do the hard work of building genuinely inclusive cultures where flexibility is not a silent penalty but a normalized and respected way of working for everyone, regardless of their gender or their place in the hierarchy.

Ultimately, our study is a snapshot, a single frame capturing the uneven landscape of work in Chile today. We now need to see the movie—through longitudinal data that tracks careers over time and, just as importantly, through qualitative research that gives voice to the stories behind the statistics. By placing inequality at the very heart of the conversation, we can begin to move beyond the easy rhetoric and work toward a future where flexibility is not just another word for privilege but a genuine tool for a more just and humane way of working for all Chileans.