Deconstructing Burnout at the Intersections of Race, Gender, and Generation in Local Government

Abstract

In recent years, there have been multiple calls for public administration scholars to adopt an intersectional approach to the study of diversity within public organizations. This paper empirically examines the simultaneous influence of multiple dimensions of individual identity on employee burnout. We advance a better understanding of disparities in individual well-being outcomes for public servants. Using conservation of resource (COR) theory and applied intersectionality, we systematically measure and model differential vulnerabilities to emotional exhaustion, depersonalization, and loss of personal accomplishment for individuals at the intersection of gender, racial, and generational identities. Using survey data on local government employees across two neighboring large cities in California, we use ordinary least squares and ordered logit models to estimate the impact of intersectional identities on different dimensions of burnout. Our results show that younger generations of women of color are particularly vulnerable to burnout, but the experience of burnout is not uniform across groups, with each dimension of burnout revealing different vulnerable groups. These findings highlight the importance of deconstructing burnout into its discrete dimensions to better understand the experience of different socio-demographic groups of employees and develop culturally competent strategies to better support an increasingly diverse public workforce.

Introduction

Much scholarship in public administration research has looked at diversity management and administrative outcomes for focal demographics through the lens of representative bureaucracy theory. That scholarship suggests recruiting and retaining a talented and diverse workforce reflective of the community being served is of paramount importance to improve public service and public trust (Ding, Lu, and Riccucci 2021). These organizational outcomes partially hinge on whether diverse employees feel engaged and committed to the work and that the focal demographic characteristic is salient to the bureaucrat, their peers, management, and clientele (Marvel and Resh 2015).

Yet, representation can come at a cost to the very people who carry a particular demographic trait of interest. People have multiple identities whose intersection may not neatly align to organizational intent of representation or the perceptions of organizational peers or the public. These misperceptions can place unintended burdens on particular intersectional identities in the workplace that might then lead to disengagement or burnout. In this paper, we explore which socio-demographic groups in two large local government workforces are most vulnerable to disengagement as measured by burnout on the job.

Hofhuis, Van der Zee, and Otten (2014) show that minority employees are more difficult to retain than majority employees, which reduces organizations’ ability to benefit from diversity. It is well documented in meta-analytic work that burnout is strongly associated with turnover intentions and reduces organizational commitment, both of which impact the quality of public service delivery (Alarcon 2011; Lee and Ashforth 1996). Addressing burnout in local government is increasingly important to improve organizational well-being and outcomes, including employee retention and commitment. At the local level, recent work by Headley et al. (2021) shows that “passive representation without an organizational commitment to positive or fair treatment by bureaucrats can limit the potential symbolic benefits of demographic representation,” highlighting the potential consequences of employee burnout on the communities they serve (p. 6).

Lu and Guy (2019) argue public service work, while varied across job types, has especially high emotive demands associated with serving the public. Particularly for those who frequently engage with residents seeking services, employees must be skilled emotionally to manage their own emotions and the emotional states of others (Lu and Guy 2019). This emotional labor may come at a cost, depleting psychological resources and necessitating exploration of burnout among local government employees. In the broader management literature, there is evidence that interactions characterized by customer aggression and unreasonable or “ambiguous” customer expectations are more likely to provoke employee burnout (Goussinky 2011). In public management scholarship, ambiguous citizen expectations (Bryer 2011; Lipsky 1980) and negative valence attributions to government service (Marvel 2016) are more likely the norms. As well, modes of communication can be limited and interactions rule-bound, despite the complexity of extemporaneous citizen needs (Resh, Barboza-Wilkes, and Marvel 2021).

Indeed, recent work shows that public administrators’ willingness to engage with citizens directly is a function of their past interactions (Resh, Barboza-Wilkes, and Mooradian 2020; Resh, Barboza-Wilkes, and Marvel 2021). How those past interactions were perceived by the employee (negatively or positively) will induce expectations of similarly valanced future interactions. Burnout has been linked to repeated customer service encounters through the emotional dissonance (i.e., the separation between felt and expressed emotions) that occurs with emotional labor (Grandey et al. 2013; Lee and Ashforth 1996). Burnout is likely further influenced by the characteristics of the interaction itself (such as length and frequency), the organizational context, and personal characteristics intrinsic to the individuals involved (Grandey et al. 2013). In both their external citizen-facing roles and in their internal professional contexts, emotional dissonance may be more prevalent for those of minority or multiple identities through “code-switching”—i.e., behavioral adjustments of “one’s style of speech, appearance, behavior, and expression in ways that will optimize the comfort of others in exchange for fair treatment, quality service, and employment opportunities” (McCluney et al. 2019).

Importantly, Maslach and Jackson (1981) argue that emotional exhaustion is only one component of burnout. We know far less about the relationship between emotionally taxing public service work and the depersonalization and self-efficacy components of burnout. Building upon Conservation of Resource (COR) Theory, experiences of oppression along racial, gender, and generational lines influence the resource reserves and coping mechanisms available to different groups (Hobfoll 2012). In recent years, there have been multiple calls for public administration scholars to adopt an intersectional approach to the study of diversity within public organizations. This paper examines the simultaneous influence of multiple dimensions of individual identity on employee burnout. Using COR theory and applied intersectionality, we systematically measure and model differential vulnerability to emotional exhaustion, depersonalization, and loss of personal accomplishment for individuals at the intersection of gender, racial, and generational identities.

Breslin, Pandey, and Riccucci (2017) advocate for future public management research to acknowledge that individuals exist at the intersection of multiple identities simultaneously. In their review of public administration scholarship over 25 years, they find most studies examine gender or race separately in a way that obscures the “unique experiences of individuals who occupy multiple marginalized social categories” (Breslin et al. 2017, p. 160). Despite public organizations and agencies working to diversify their workforce to become more representative of the public they serve, limited work exists to address the multiplicative effects of racial and gender disparities in job-related experiences for public servants beyond explicit discrimination and prejudicial treatment (Choi and Rainey 2013) Our study explores which socio-demographic groups in the local government workforce are most vulnerable to the different dimensions of burnout.

We begin by reviewing the existing literature on diversity management and burnout from a conservation of resources perspective, and then introduce the theoretical framework of applied intersectionality to explain our hypotheses and modeling decisions. We then review our results and discuss (1) why an intersectional approach is critical to understanding a diverse workforce, (2) how the multiple dimensions of burnout are experienced by different groups, and (3) why a multidimensional intersectional approach to studying burnout is critical for the development of culturally responsive diversity management practices and empirical applications of relevant theory-building. Throughout our analysis, we emphasize how experiences of burnout differ across different identity groups, and we conclude with the implications for managing a diverse public workforce.

Managing Burnout Among Diverse Employees

Diversity Management

Studies of the position of specific socio-demographic groups in organizations date back to the 1970s. Such studies often posit that public sector organizations are not meritocracies and document how inequalities in organizations are structured along gender and racial lines as organizing principles (Zanoni et al. 2010). In the domains of mentoring, satisfaction, performance evaluations, promotion opportunities, and income, scholars find extensive evidence of unequal treatment of women and racial minority employees, with negative effects on their work satisfaction and careers (Zanoni et al. 2010). These outcomes are generally explained as the effect of prejudice and discrimination, which largely overlooks structural, context-specific elements that lead to organizational inequalities (Roberson et al. 2017).

Acker (2006) argues that interlocked practices and processes in organizations result in continuing inequalities that influence employee well-being. While the basis for inequality in organizations varies by context, most organizations will show that class, gender, and race processes are present (Acker 2006). Hierarchies are usually gendered and racialized, especially where prominent leadership positions at the top are almost always occupied by white men in the United States, organizing work in the image of a class-privileged white man (Acker 2006). Ray (2019) argues seeing organizations as racial structures in which cultural schemas connect to social and material resources moves beyond models using race as merely a demographic variable, focusing instead on the mechanisms reproducing racial inequality.

Rodriquez et al. (2016) argue that intersectionality is a lens through which to interrogate mainstream human resource management and to reveal implicit assumptions and systems that sustain inequalities in the workplace. Empirical work employing an intersectional lens shows that women of color fall below white women and co-racial men on several key indicators of labor market inequality (Browne and Misra 2003). Concerning wage inequality, there are distinct patterns of disadvantage for women of color, but also similarities to co-ethnic men within the race stratification system and white women within the gender stratification system (Browne and Misra 2003). Feminist intersectional approaches argue gender, race, and other markers of differences inform not only employers’ decisions about hiring, promotions, training, and wages but also the interactions with coworkers and customers in ways that may contribute to inequitable institutions (Browne and Misra 2003).

Burnout and Conservation of Resource (COR) Theory

Moving beyond individual experiences of discriminatory or prejudicial treatment, we look at burnout to explore the compounded effects of both individual episodes and systemic forces that influence the overall experience of diverse employees in the public workforce. Maslach and colleagues conceive of burnout as an individual-level stress response rooted in the interpersonal context of relational transactions in the workplace (Maslach et al. 2008; Maslach and Jackson 1981, 1986). Importantly, burnout is a multidimensional construct, with any one dimension necessary but not sufficient to constitute burnout (Maslach et al. 2008).

Burnout is described as “a psychological syndrome in response to chronic interpersonal stressors on the job” and is identified through three signs: (1) emotional exhaustion, (2) depersonalization (also cynicism), and (3) reduced personal accomplishment/efficacy (Maslach et al. 2008; Maslach and Jackson 1981, 1986). Exhaustion is a stress dimension referring to feelings of overextending and depleting emotional resources (Maslach et al. 2008). Depersonalization (or cynicism) represents interpersonal distancing that manifests as a detached response reflective of lowered emotional or cognitive involvement in work (Maslach et al. 2008). Reduced accomplishment (or efficacy) is a self-evaluation reflecting feelings of incompetence and lack of achievement in work (Maslach et al. 2008).

Based on COR theory, burnout is a process whereby (1) emotional exhaustion occurs when resources are depleted, (2) then cynicism develops as a response to exhaustion when maladaptive coping mechanisms are used to account for the lost resources, and (3) this maladaptive coping leads to a lack of personal accomplishment combining to create a spiral of resource loss (Alarcon 2011; Hobfoll et al. 2018). When interpersonal interactions on the job demand emotional effort, emotional resources are depleted, leaving employees emotionally exhausted if those resources are not replenished. By contrast, depersonalization is a coping mechanism by which employees can psychologically distance themselves from others in an interaction to reduce stress and cope with the resource loss associated with emotional job demands (Hochschild 1983). The cynicism and disengagement seen among those engaging in depersonalization to cope with the job is believed to be associated with reduced organizational commitment, poor performance, and loss of personal accomplishment (Hobfoll et al. 2018).

Analyzing burnout as a multidimensional process allows for a more nuanced understanding of the experience of diverse members of the workforce, and we argue that studying burnout in both its aggregate and disaggregated forms will be more useful for public managers to identify specific vulnerabilities to burnout for different groups of employees. COR theory is distinct from other resource-adaptation models in its emphasis on sociocultural framing of resources, which asserts that many resources are common among members who share a cultural niche (Hobfoll 2002). This cultural rather than individualistic orientation further suggests that resources, or their absence, aggregate in resource caravans within groups (Hobfoll 2002). This orientation appropriately places resources as a currency in historical context (Hobfoll 2002). In other words, COR theory’s foundational assumption of the burnout process might unfold differently across groups as each may have different contextually adaptive competencies. COR theory thus leads us to emphasize the potential importance of specific cultural resources, or lack thereof, that may manifest in different vulnerabilities to burnout.

As a fundamental theory in the field of burnout, COR emphasizes common appraisals of stressors and resources held jointly by groups, organizations, and cultures (Hobfoll 2011). Resource caravans are believed to be the consequence of “nurturance and learned adaptation,” meaning they emerge from common social, environmental, and developmental conditions that make sets of resources highly correlated within groups (Hobfoll 2011; Hobfoll et al. 2018, p. 107). Hobfoll et al. (2018) note that commonly valued resources such as well-being, self-esteem, and sense of purpose are expressed differently across cultures.

Starting in early childhood, individuals begin to accumulate experiences that influence how they navigate the world around them (Wilkins and Pace 2014). Historically, communities of color have prepared their children to deal with a racist society by instilling specific emotional skills and resources aimed at coping with different manifestations of racism and discrimination (Wilkins and Pace 2014). Hobfoll (2011, 2012) asserts that different cultures will operate under the same rules of conserving resources, but the ranking and valuing of resources will be different (Hobfoll et al. 2018). Additionally, prior literature on emotional capital suggests that there is a gendered dimension to emotional resources, warranting an intersectional approach (Cottingham 2016). Cottingham (2016) argues that gendered socialization within the culture of the home environment influences the use of emotional resources by men and women in different ways. Across racial and ethnic groups, we suspect there are meaningful differences in the ways communities have prepared women to deal with sexism over time.

Given the cultural origins of resource caravans, we anticipate that different socio-demographic groups will have varied resource sets available that may make them differentially vulnerable to (or tolerant of) the multiple dimensions of burnout and subsequent resource loss cycles. From a structural perspective, individuals of lower status are held to a higher degree of interpersonal effort than their higher status counterparts (Grandey et al. 2013). Resource-based approaches suggest that the interpersonal burden in an exchange varies as a function of interpersonal power, with high-power people less affected by the emotions of others given their substantial resource advantage that provides them with the ability to act at will without serious consequences (Grandey et al. 2013).

Atewologun and Sealy (2014) find privilege is “fluid and changeable” for individuals over time in organizations with added professional experience and promotion opportunities (p. 426). However, it is important to note time spent in organizations does not guarantee an increase in power or status from positional seniority. Snape and Redman (2006) found that perceiving age discrimination, for being either too old or too young, has negative consequences for affective commitment across hierarchical positions. Sex and age combine to create unique forms of age discrimination for older women in the workforce (Porter 2003). For younger women, difficulty combining work and family responsibilities during childbearing years has been associated with disadvantages in employment and earning (Plickert and Sterling 2017). Thus, we expect that local government employees vulnerable to sexism, racism, and ageism simultaneously experience burnout in ways that are similar to and also different from peers who do not stand at the same intersection of identities. This allows us to expand upon an undertheorized component of COR theory, which has yet to explore the temporal dimensions of resource caravans as generational cohorts experience age-based discrimination and privileges differently across other social identities.

Resource Caravan Passageways in Organizational Context

According to COR theory, resources do not exist in isolation but rather they travel in caravans for both individuals and organizations while existing in ecological conditions that either foster the development of additional resources or limit the creation and sustenance of resources (Hobfoll 2011; Hobfoll et al. 2018). The passageways concept helps explain the high correlations among resources that travel together in caravans, and suggests that caravans will be further accentuated in organizations because individuals will share an organizational setting and culture (Hobfoll 2011). Hobfoll et al. (2018) argue “whether we are speaking of classes of people, such as women or ethnic minorities, or of any given set of individuals, when we look at their resources, stress, and productivity, we are actually seeing a reflection of the greater organization and culture’s stage setting, allowances, and facilitations—the passageways they create, maintain, and foster” (p. 107). Critical for human resource management, Hobfoll (2011) argues that creating and sustaining ecologies of organizational support, stability, and safety are necessary to keep pace with the challenges of the workplace.

We suspect that socio-demographic differences between employees and residents or coworkers will create more opportunities for interpersonal stressors by heightening the opportunity for cultural misunderstanding. Robeson et al. (2017) argue positive workplace climates and interactions may help an employee who faces isolation and alienation in their community to better cope with stress because of the support, validation, and understanding they receive from their workgroup. Inclusion is defined as “the degree to which individuals feel part of critical organizational processes” (Mor-Barak and Cherin, 1998, p. 48) and involves valuing and leveraging the strengths of diverse employees (Choi and Rainey 2013; Sabharwal 2014). More recent research has found that employee perception of inclusion is associated with feeling valued and recognized for their efforts in the organization, whereas feelings of exclusion and unfair treatment are associated with a loss of job interest (Ferdman et al. 2010; Sabharwal 2014). Nishii (2013) argues an important aspect of inclusive climates is that they facilitate the engagement of whole selves, drawing attention to the need to explore the experiences of multiple identities in the study of public organizations aiming to recruit and retain a workforce representative of the communities they serve. While most representative bureaucracy scholarship explores race and gender, an employee’s sense of inclusion, belonging, and representation in an organization involves multiple simultaneous and intersecting identities (Vinopal 2020).

Intersectionality as a Research Paradigm

Scholars have called for more work that integrates knowledge about racism, power, and inequality into public administration research to better understand the more subtle underlying institutional and structural forces that endure in public organizations (Frenkel and Shenhav 2006). Atewologun and Sealy (2014) advocate for an intersectional lens to the study of organizations for its ability to demonstrate the dynamic and multifaceted experiences of situationally privileged non-dominant social group members. To support the well-being of diverse public employees, public agencies should understand differential vulnerability to the experiences of burnout, and how burnout manifests differently across groups. Such research can help inform more equitable and culturally appropriate mechanisms to improve a climate of inclusion that can enhance employee well-being and public service outcomes (Nishii 2013). Thus, this paper empirically examines the simultaneous influence of multiple dimensions of individual identity on employee burnout to better understand disparities in individual well-being outcomes for employees in our public institutions.

The concept of intersectionality highlights that rather than a single, easily stated, unitary identity, everyone has overlapping identities, loyalties, and allegiances that may at times come into conflict with one another (Collins 2008; Crenshaw 1991). The tendency to treat race and gender as mutually exclusive categories of experience and analysis is perpetuated by the single-axis framework dominant in antidiscrimination law, feminist theory, and antiracist politics (Crenshaw 1991). For Collins (2008), the “interlocking systems of race, class and gender” constitute a “matrix of domination” within which an individual can simultaneously experience disadvantage and privilege through the combined statuses of gender, race, and class. The “matrix” approach to intersectionality looks at inequality across activities at all levels and in all institutional contexts, making it difficult to imagine any social process having a singular “main effect” for anyone (Collins 2008). Hancock (2016) argues that permeable boundaries exist between the oppressed and oppressor necessitating a reconceptualization of the ontological relationships between analytical categories of differences. Ontological complexity refers to the idea that “analytical categories like ‘race,’ ‘gender,’ ‘class,’ and the hegemonic practices associated with them (racism, sexism, classism, to which imperialism and homophobia certainly could be added) are mutually constitutive, not conceptually distinct” (Hancock 2016, p. 71).

Rodriguez et al. (2016) argue that the fundamental contribution of intersectionality is its ability to critique and disrupt dominant logics in organizations that (re)produce inequalities, making it a key lens through which to interrogate taken-for-granted assumptions that sustain inequalities in human resource management theory and practice. Acker (2006, 2012) argues examining work organizations through an intersectional lens helps identify barriers to equality. We add that scholarship within public administration must move beyond oversimplified white/non-white binaries when interacting race with gender to highlight the nuanced differences across racially minoritized groups.

The dominant approach to the study of intersectionality in work and organizations has focused on intersections of social identities to describe the subjective experience of individuals and groups given their social membership (Rodriguez et al. 2016). As an example, McCall’s (2005) intercategorical approach places the focus on making comparisons between and within groups and categories that are considered to be non-static anchor points. Rodriguez et al. (2016) argue that a more fruitful approach for future research is to embed individual experiences within systemic dynamics of power and oppression, looking critically at how power is exercised and institutionalized simultaneously in multiple spheres of influence. Additionally, Rodriguez et al. (2016) call for future research to expand beyond the typical grouping of gender-race-class to a more complex set of intersecting categories of differences. Burnout as an outcome of interest allows us to understand the combined psychological impact of a range of subjective micro-level encounters, routine organizational practices, and institutional arrangements, whereas an intersectional approach helps us to highlight otherwise hidden experiences of those with multiple marginalized identities as they navigate organizational dynamics.

Intersectional Hypotheses

In a review of the workplace diversity literature, Kulik and Bainbridge (2006) suggest that sex, race, and age should all be considered “surface” diversity dimensions, given that they are typically reflected in physical features that trigger implicit and explicit biases. We expect that local government employees vulnerable to sexism, racism, and ageism simultaneously will experience burnout in ways that are similar to and also different from peers who do not hold the same combination of identities. The local government context is especially important for this intersectional approach because of the opportunities for mistreatment and microaggressions both in interactions with residents and in navigating public institutions that privilege non-minorities.

The concept of differential racialization highlights the ways the dominant society has racialized different minority groups at different times, necessitating an intersectional analysis that moves beyond the white/non-white binary (Collins 2008; Crenshaw 1991). We argue that this should include racialized and gendered ageism that can lead to differential vulnerabilities to burnout. A key principle of COR theory is that “people must invest resources in order to protect against resource loss, recover from losses, and gain resources,” making it critical to understand differences in the resources across socio-demographic groups (Hobfoll et al. 2018, p. 105). COR theory suggests that resource loss is more powerful than resource gain, and those with fewer resources available are believed to be more vulnerable to resource loss and less capable of resource gain—creating resource loss cycles (Hobfoll et al. 2018). Given the proposed cultural origins of resource caravans, we anticipate that different socio-demographic groups will have varied resource sets available that may make them differentially vulnerable to the multiple dimensions of burnout and subsequent resource loss cycles.

Studies that explore age-related biases in the workplace yield both positive and negative stereotypes for the young and the old: Younger workers are perceived as being more trainable, faster at information processing, more creative, and more flexible with new technologies, but also less reliable and weaker with interpersonal skills than older workers; older workers are perceived as being more reliable, experienced, and interpersonally savvy, but also prone to accidents, less flexible, and less productive than younger workers (Calo, Peterson, and Decker 2013). Generational cohorts have also been found to vary considerably in terms of their public service motivation and organizational commitment, with younger workers less motivated and committed, making them more prone to stress, burnout, and turnover (Mor-Barak et al. 2016; Ng et al. 2010; Yang and Guy 2006). A review by Mor-Barak et al. (2016) finds that age has a curvilinear relationship with work outcomes, with employees at younger and older ends of the age spectrum experiencing more negative outcomes than their peers in the middle of the age distribution. Yang and Guy (2006) add that “until job applicants are socialized into the organizational culture, their frame of reference is more likely to be generational” (p. 281).

Building upon COR theory and applied intersectionality, we approach this study with three hypotheses:

H1: Individuals with multiple marginalized identities (e.g., women of color) will experience higher levels of burnout compared to their white male peers.

H2: Among women, heterogeneity of experiences with burnout is based on both racial and generational differences.

H3: The salience of each dimension of burnout differs among socio-demographic groups.

While specific directional hypotheses for each subgroup are beyond the scope of this exploratory analysis, we suspect that the dominant groups—those who identify as white, male, and GenX—will experience situational privileges. Conversely, we anticipate women of color, particularly in older and younger generational cohorts will experience more burnout, with nuanced differences at specific combinations of race, gender, and generation. Importantly, COR theory is not explicit in delineating empirically whether each dimension of burnout has an additive or multiplicative effect on subsequent dimensions or if individuals enter and exit dimensions of burnout as discrete stages, making this research disaggregating burnout exploratory.

Data, Methods, and Measures¹

In May 2018, assisted by the personnel departments of two neighboring large cities in California, our research team electronically distributed survey instruments to the entirety of both city government workforces. The fact that they are neighboring cities allows us to account for heterogeneity but within homogeneous socioeconomic and geographic proximity. A total of 6,006 employees responded, a response rate of 25.8% for the first city and 23% for the second city. We narrowed our sample to those who answered all the relevant survey items for our analysis: 3,232 for overall burnout, 3,244 for emotional exhaustion, 3,240 for depersonalization, and 3,251 for loss of personal accomplishment. Our sample consisted of approximately 49% women, and the racial makeup was 33% white, 24% Mixed/Other, 15% Asian, 14% Hispanic, 11% black, 2% Native Hawaiian/Pacific Islander (NHPI), and 1% American Indian/Alaskan Native (AIAN). See Supplementary Appendix A for specific distribution by gender, race, and generation. Relative to the population demographics, we have oversampled women which account for roughly 30% of the city workforce across both cities. This is ideal given our particular interest in studying underrepresented groups within the workforce. Our survey instrument offered the ability to select more than one race, which resulted in a larger number of individuals identifying as Mixed/Other, a category that is unavailable in the population data set provided by the cities. These cities are in a county with a total population of over two million people, where 50.7% identify as women, 48.6% identify as Hispanic, 26.1% white, 15.4% Asian, 9.0% black, 3.1% Mixed, 1.4% AIAN, and <1% NHPI.

We utilize factor analysis and multivariate regression models to estimate the impact of intersectional identities on different dimensions of burnout while controlling for a series of other predictors. Our approach to factoring and grouping latent variables depends on theories in burnout literature. When appropriate, we rotate and modify factors to enhance interpretability and maximize differences between variables and their factor loading scores. We use eigenvalues to determine how much information each factor contains.

Once composite scores have been created through factoring, we integrate them into our multivariate regression models. We rely on a series of ordinary least squares (OLS) regression models where our outcome of interest (Y) is a continuous variable (i.e., composite scores for burnout), our predictors of interests are varying social identities and their interactions, and our controls (x) are a set of individual- and organizational-level predictors related to burnout:

$${\displaystyle \begin{array}{l}{\displaystyle Y=B_0+B_{1}gender+B_{2}race+B_{3}generation+\dots +B_k{X}_k+\epsilon ,}\end{array}}$$

(1)

$${\displaystyle \begin{array}{l}{\displaystyle Y=B_0+B_1(gender\ast race)+B_{2}generation+\dots +B_k{X}_k+\epsilon ,}\end{array}}$$

(2)

$${\displaystyle \begin{array}{l}{\displaystyle Y=B_0+B_1(gender\ast race\ast generation)+\dots +B_k{X}_k+\epsilon .}\end{array}}$$

(3)

We use a series of ordered logit models when our outcome of interest has multiple outcomes on an ordinal scale:

$${\displaystyle \begin{array}{l}{\displaystyle Y=\frac{1}{1+e^{-(B_0+B_{1}gender+B_{2}race+B_{3}generation+\dots +B_k{X}_k+\epsilon )}},}\end{array}}$$

(4)

$${\displaystyle \begin{array}{l}{\displaystyle Y=\frac{1}{1+e^{-(B_0+B_1(gender\ast race)+B_{2}generation+\dots +B_k{X}_k+\epsilon )}},}\end{array}}$$

(5)

$${\displaystyle \begin{array}{l}{\displaystyle Y=\frac{1}{1+e^{-(B_0+B_1(gender\ast race\ast generation)+\dots +B_k{X}_k+\epsilon )}}.}\end{array}}$$

(6)

For each outcome, we produce three models that look at demographic identities as one-, two-, and three-dimensional with an intent to examine the outcomes of burnout reproduced at the intersection of gender, race, and generation.

Dependent Variable(s)—Burnout

Our dependent variable across our initial single-axis and intersectional models is overall burnout. Our overall burnout factor combines all three dimensions of burnout (i.e., emotional exhaustion, depersonalization, and loss of personal accomplishment) using eight items (eigenvalue = 2.36). We use three items verbatim from the Maslach Burnout Inventory—General Survey (MBI-GS, Maslach and Jackson 1986) and adapt the language of five additional MBI-GS items to the public service context to explicitly highlight resident engagement and impact on the community as unique components of local government work (see Supplementary Appendix A). Importantly, confirmatory factor analysis on all eight items yields three significant discrete factors, reflecting the three discrete dimensions of burnout. To further contextualize the experience of burnout, we deconstruct our dependent variable into its three dimensions and run our intersectional models separately for emotional exhaustion, depersonalization, and loss of personal accomplishment specifically to assess which groups are most sensitive to the different dimensions of the core construct of burnout. Our emotional exhaustion factor is composed of four items (eigenvalue = 2.13) explicitly about emotional experience. Our depersonalization factor is composed of two items (eigenvalue = 0.46),² measuring the extent to which employees feel dissimilar and disconnected from residents. Lastly, sense of personal accomplishment is measured using a single item “I feel I’m positively influencing other people’s lives through my work” on a 7-point scale from (1) strongly disagree to (7) strongly agree. Importantly, given the emotional exhaustion dimension contains more individual items than either depersonalization or loss or personal accomplishment, we are cautious that the overall burnout measure may disproportionately weight exhaustion over other burnout factors. This further emphasizes the need to explore the discrete dimensions of burnout.

Independent Variables of Interest—Intersectional Identity

Gender and race are the dominant axes used most commonly in intersectional studies (Hancock, 2016), but to assess the nuances of identity we operationalize intersectionality along single-, double-, and triple-axes to include generational differences. Adding a generational dimension to our analysis allows for an expanded consideration of the role that generational socialization and age play in the use, development, and maintenance of resources (Cottingham 2016; Wilkins and Pace 2014). Generation is modeled as a categorical variable and includes four groups: (1) Baby Boomers (born between 1946 and 1964), (2) Generation X (GenX; born between 1965 and 1980), (3) Millennials (born between 1981 and 1996), and (4) Generation Z (GenZ; born after 1996). In our single-axis model, we treat gender, race, and generation as separate independent variables. In our double-axis model, we interact gender with race yielding 14 total combinations and include generation as a separate independent variable. Lastly, our triple-axis model interacts gender with race and generation yielding 50 total combinations once netting out subgroups with insufficient observations.

Organizational and Individual Controls

We incorporate several workplace context variables believed to impact burnout through the fulfillment of the basic psychological needs for autonomy, competence, and relatedness (Deci, Olafsen, and Ryan 2017). Specific to relatedness, Nishii (2013) proposes three dimensions of a climate for inclusion: (1) fairness of employment practices to “level the playing field,” (2) integration of differences and allowing authentic expression, and (3) inclusion in decision-making. Similarly, Sabharwal (2014) conceptualizes a scale of organizational inclusion behaviors (OIB) with three categories, including (1) leadership commitment to foster inclusion, (2) employee influence in organizational decisions, and (3) fair/equitable treatment. We include an item on the perception of bias in management behaviors to assess the fairness of organizational practices, and multiple items surrounding autonomy and influence in both individual work tasks and what happens in the department to assess inclusion in decision making. Combined, these items allow us to assess the extent to which the environment makes diverse employees feel unfairly constrained by their social location within the organization or autonomous where their differences are integrated and they are included in decision making. To explore cultural competence, respondents are asked to what degree they feel cultural differences between citizens and employees in their department inhibit their collective ability to perform.

To look at the psychological need for relatedness, we include multiple items about the extent to which the employee feels their ethnic and gender backgrounds are adequately represented in their department. Additionally, we include two perceptual measures of the extent to which employees believe both ethnic and gender differences between themselves and citizens are difficult to overcome to assess confidence working across cultural boundaries. We also include a dummy variable that indicates whether the employee primarily interacts with residents of their own race, and a variable for time spent engaging with residents as another way of measuring the potential for fulfilling the need for relatedness in relationships external to the organization. We control for time spent engaging residents to account for the important differences in job types and characteristics.

At the individual level, we control for differences in prosocial motivation, confidence, and job satisfaction both overall and in terms of pay specifically. Our prosocial motivation factor is composed of three items (eigenvalue = 1.32) that look at the extent to which employees are motivated to help others and find their work to be meaningful. Our confidence variable looks explicitly at confidence in one’s ability to do their job, allowing us to measure sense of competence. Additionally, we control for the highest level of education and tenure in the organization measured through years of experience in the municipality.

Our sample includes employees from 30 distinct departments, which necessitates appropriate controls to address differences in job characteristics. While data limitations prevent us from including individual fixed effects at the position or unit level, we do include department fixed effects to address the heterogeneity of job characteristics associated with agency-specific tasks and goals.

Supplemental Analyses of Perceived Representation

Supplemental to our main analyses, we also examine how interpersonal interactions with residents and perceived racial (dis)similarity shape burnout outcomes across socio-demographic groups. We test the relationship across socio-demographic groups using Pearson Chi-Squared tests to determine if there are statistical differences in the degree to which participants “find ethnic differences between [themselves] and citizens are easy to overcome in [their] job.” The findings and discussion are available in Supplementary Appendix B.

Results

First, we look at the extent to which individuals with multiple marginalized identities might experience higher levels of burnout compared to their white male counterparts, with the hypothesis that women of color will experience more burnout, but meaningful differences will exist between women of different racial backgrounds and generational cohorts. We then examine how intersecting generational identities facilitate different burnout dimensions. Our models that examine overall burnout, emotional burnout, and depersonalization as outcomes utilize factor scores³ where a score of 0 means the observation’s rating on the latent measure is close to the sample average. A negative score indicates a lower rating relative to the sample average, whereas a positive score indicates a higher-than-average rating on the latent measure. Said differently, a positive coefficient indicates that the participant, on average, is more likely to experience burnout than the reference or comparison group. A negative coefficient indicates that the participant is less likely to experience burnout relative to the reference group. The variable measure for personal accomplishment uses an ordinal scale response, so our analysis considers the odds ratio. In other words, higher odds on this variable indicate that respondents are more likely to feel they are not making a positive impact on the lives of others through their work, indicating less personal accomplishment. As we are examining different socio-demographic groups and intersections, we utilize (1) men, (2) white men, and (3) white men of GenX as our baseline comparison groups as they make up the largest shares in our subsamples and the two cities’ workforce populations.

Gender, Racial, and Generational Disparities in Burnout Outcomes

Examining overall burnout, our results show no statistical difference between men and women in the single-axis model (see Supplementary Appendix A). Importantly, our third hypothesis suggests that the salience of each dimension of burnout will differ for different groups, and only when we deconstruct burnout into emotional exhaustion (see table 1), depersonalization (see table 2), and loss of personal accomplishment (see table 3) do we start to see statistical differences along gender lines. When burnout is disaggregated, table 2 shows women on average score .101 points higher than their male counterparts in depersonalization. Our results in table 3 also show that women have 47.5% higher odds of being in stronger disagreement that they are positively influencing other people’s lives through their work, suggesting a lack of personal accomplishment compared to men. These results for gender as a category highlight the limitations of a single-axis approach to studying diversity with an aggregated dependent variable, which will be discussed at length below. Full single-axis model results, including controls, are available in Supplementary Appendix A.

Table 1.

Coefficients for Emotional Burnout Across Single-, Double-, and Triple-Axis Models

Single-Axis	Double-Axis	Triple-Axis				Single-Axis	Double-Axis	Triple-Axis
Gender	Gender and Race	Gender, Race, and Generation				Gender	Gender and Race	Gender, Race, and Generation
Female 0.044 (0.029) [0.130]	NH White 0.065 (0.050) [0.200]	Baby Boomer	−0.177*	(0.082)	[0.031]	Male Comparison Group	NH White Comparison Group	Baby Boomer	−0.196**	(0.068)	[0.004]
		GenX	0.091	(0.077)	[0.234]			GenX	Comparison Group
		Millennial	0.171†	(0.098)	[0.082]			Millennial	0.063	(0.094)	[0.503]
		GenZ	1.860***	(0.096)	[0.000]			GenZ	−0.039	(0.387)	[0.920]
	Black −0.292*** (0.063) [0.000]	Baby Boomer	−0.458***	(0.104)	[0.000]		Black −0.286*** (0.079) [0.000]	Baby Boomer	−0.640***	(0.115)	[0.000]
		GenX	−0.313***	(0.094)	[0.001]			GenX	−0.181	(0.127)	[0.154]
		Millennial	−0.244*	(0.122)	[0.046]			Millennial	0.056	(0.182)	[0.758]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	Hispanic −0.153* (0.061) [0.013]	Baby Boomer	−0.099	(0.129)	[0.442]		Hispanic −0.103† (0.061) [0.094]	Baby Boomer	−0.550***	(0.117)	[0.000]
		GenX	−0.172*	(0.087)	[0.047]			GenX	−0.009	(0.088)	[0.918]
		Millennial	−0.141	(0.108)	[0.194]			Millennial	0.034	(0.117)	[0.770]
		GenZ	0.777***	(0.093)	[0.000]			GenZ	0.809***	(0.141)	[0.000]
	Asian 0.137* (0.061) [0.024]	Baby Boomer	−0.171	(0.105)	[0.103]		Asian 0.107† (0.059) [0.068]	Baby Boomer	−0.101	(0.094)	[0.284]
		GenX	0.172†	(0.102)	[0.094]			GenX	0.028	(0.098)	[0.772]
		Millennial	0.300**	(0.095)	[0.002]			Millennial	0.303**	(0.107)	[0.005]
		GenZ	0.791***	(0.118)	[0.000]			GenZ	0.764***	(0.138)	[0.000]
	NHPI 0.169 (0.145) [0.242]	Baby Boomer	−0.406*	(0.205)	[0.048]		NHPI 0.002 (0.183) [0.992]	Baby Boomer	−0.133	(0.311)	[0.667]
		GenX	0.295	(0.220)	[0.180]			GenX	−0.073	(0.237)	[0.756]
		Millennial	0.573†	(0.298)	[0.055]			Millennial	0.168	(0.661)	[0.800]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	AIAN −0.120 (0.168) [0.474]	Baby Boomer	−0.626***	(0.153)	[0.000]		AIAN −0.313* (0.140) [0.026]	Baby Boomer	−0.636*	(0.255)	[0.013]
		GenX	0.068	(0.294)	[0.816]			GenX	−0.394*	(0.186)	[0.034]
		Millennial	0.143	(0.393)	[0.715]			Millennial	0.146	(0.320)	[0.648]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	Mixed/Other −0.021 (0.053) [0.689]	Baby Boomer	−0.202†	(0.106)	[0.056]		Mixed/Other −0.107* (0.050) [0.032]	Baby Boomer	−0.271**	(0.096)	[0.005]
		GenX	−0.099	(0.078)	[0.209]			GenX	−0.132†	(0.073)	[0.072]
		Millennial	0.194*	(0.087)	[0.027]			Millennial	0.034	(0.094)	[0.719]
		GenZ	0.665***	(0.110)	[0.000]			GenZ	−0.255	(0.300)	[0.396]

Note: Standard errors in parentheses; p-values in brackets.

^†p < .1;

^*p < .05;

^**p < .01;

^***p < .001.

Table 2.

Coefficients for Depersonalization Burnout Across Single-, Double-, and Triple-Axis Models

Single-Axis	Double-Axis	Triple-Axis				Single-Axis	Double-Axis	Triple-Axis
Gender	Gender and Race	Gender, Race, and Generation				Gender	Gender and Race/Ethnicity	Gender, Race, and Generation
Female 0.101*** (0.022) [0.000]	NH White 0.086* (0.036) [0.017]	Baby Boomer	0.025	(0.054)	[0.640]	Male Comparison Group	NH White Comparison Group	Baby Boomer	−0.014	(0.051)	[0.783]
		GenX	0.130*	(0.057)	[0.023]			GenX	Comparison Group
		Millennial	0.017	(0.073)	[0.814]			Millennial	−0.029	(0.070)	[0.677]
		GenZ	0.960***	(0.073)	[0.000]			GenZ	−0.607***	(0.106)	[0.000]
	Black 0.216*** (0.046) [0.000]	Baby Boomer	0.284***	(0.074)	[0.000]		Black 0.059 (0.059) [0.311]	Baby Boomer	0.081	(0.096)	[0.397]
		GenX	0.214**	(0.069)	[0.002]			GenX	0.046	(0.077)	[0.552]
		Millennial	−0.009	(0.098)	[0.928]			Millennial	−0.053	(0.143)	[0.710]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	Hispanic 0.211*** (0.047) [0.000]	Baby Boomer	0.283**	(0.103)	[0.006]		Hispanic 0.013 (0.048) [0.783]	Baby Boomer	−0.037	(0.092)	[0.684]
		GenX	0.222**	(0.071)	[0.002]			GenX	0.004	(0.069)	[0.958]
		Millennial	0.086	(0.073)	[0.238]			Millennial	−0.000	(0.093)	[0.997]
		GenZ	0.040	(0.070)	[0.571]			GenZ	−0.160†	(0.095)	[0.094]
	Asian 0.133** (0.042) [0.001]	Baby Boomer	0.159*	(0.075)	[0.034]		Asian 0.066 (0.044) [0.132]	Baby Boomer	0.100	(0.067)	[0.134]
		GenX	0.137*	(0.066)	[0.038]			GenX	0.018	(0.076)	[0.817]
		Millennial	0.042	(0.069)	[0.543]			Millennial	0.014	(0.082)	[0.860]
		GenZ	−0.426***	(0.086)	[0.000]			GenZ	−0.317**	(0.119)	[0.008]
	NHPI 0.100 (0.080) [0.210]	Baby Boomer	0.146	(0.154)	[0.341]		NHPI 0.021 (0.103) [0.840]	Baby Boomer	0.123	(0.236)	[0.600]
		GenX	0.032	(0.113)	[0.778]			GenX	−0.001	(0.093)	[0.990]
		Millennial	0.137	(0.135)	[0.310]			Millennial	−0.278	(0.266)	[0.295]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	AIAN 0.027 (0.161) [0.869]	Baby Boomer	0.075	(0.227)	[0.742]		AIAN 0.364* (0.158) [0.021]	Baby Boomer	0.206	(0.256)	[0.421]
		GenX	−0.260	(0.234)	[0.266]			GenX	0.383	(0.238)	[0.108]
		Millennial	0.323	(0.335)	[0.335]			Millennial	0.485	(0.333)	[0.145]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	Mixed/Other 0.147*** (0.040) [0.000]	Baby Boomer	0.200*	(0.087)	[0.022]		Mixed/Other 0.064 (0.040) [0.106]	Baby Boomer	0.003	(0.071)	[0.962]
		GenX	0.130*	(0.060)	[0.029]			GenX	0.069	(0.060)	[0.252]
		Millennial	0.078	(0.066)	[0.236]			Millennial	0.039	(0.075)	[0.605]
		GenZ	−0.096	(0.081)	[0.236]			GenZ	−0.086	(0.248)	[0.728]

Note: Standard errors in parentheses; p-values in brackets.

^†p < .1;

^*p < .05;

^**p < .01;

^***p < .001.

Table 3.

Coefficients for Loss of Personal Accomplishment Across Single-, Double-, and Triple-Axis Models

Single-Axis	Double-Axis	Triple-Axis				Single-Axis	Double-Axis	Triple-Axis
Gender	Gender and Race	Gender, Race, and Generation				Gender	Gender and Race	Gender, Race, and Generation
Female 1.475*** (0.103) [0.000]	NH White 1.471*** (0.166) [0.001]	Baby Boomer	1.336†	(0.235)	[0.100]	Male Comparison Group	NH White Comparison Group	Baby Boomer	0.886	(0.146)	[0.462]
		GenX	1.474*	(0.248)	[0.021]			GenX	Comparison Group
		Millennial	1.278	(0.303)	[0.301]			Millennial	0.832	(0.177)	[0.387]
		GenZ	0.000***	(0.000)	[0.000]			GenZ	0.482	(0.606)	[0.561]
	Black 1.785*** (0.294) [0.000]	Baby Boomer	1.318	(0.378)	[0.335]		Black 1.170 (0.230) [0.423]	Baby Boomer	0.705	(0.203)	[0.224]
		GenX	1.860**	(0.445)	[0.009]			GenX	1.517	(0.494)	[0.201]
		Millennial	1.842†	(0.599)	[0.060]			Millennial	1.853	(0.766)	[0.135]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	Hispanic 1.572** (0.240) [0.003]	Baby Boomer	2.743*	(1.177)	[0.019]		Hispanic 1.162 (0.170) [0.303]	Baby Boomer	0.943	(0.252)	[0.827]
		GenX	1.317	(0.270)	[0.179]			GenX	0.963	(0.218)	[0.867]
		Millennial	1.406	(0.360)	[0.184]			Millennial	1.459	(0.362)	[0.127]
		GenZ	0.000***	(0.000)	[0.000]			GenZ	0.644	(0.195)	[0.146]
	Asian 2.273*** (0.324) [0.000]	Baby Boomer	2.506**	(0.707)	[0.001]		Asian 1.593** (0.233) [0.001]	Baby Boomer	1.539†	(0.381)	[0.082]
		GenX	1.633*	(0.340)	[0.018]			GenX	1.407	(0.344)	[0.163]
		Millennial	2.487***	(0.550)	[0.000]			Millennial	1.431	(0.360)	[0.154]
		GenZ	1.766†	(0.517)	[0.052]			GenZ	8.969***	(3.042)	[0.000]
	NHPI 4.155*** (1.378) [0.000]	Baby Boomer	7.304***	(4.378)	[0.001]		NHPI 1.198 (0.555) [0.697]	Baby Boomer	0.388	(0.256)	[0.152]
		GenX	2.735†	(1.557)	[0.077]			GenX	1.445	(0.984)	[0.589]
		Millennial	3.108***	(1.048)	[0.001]			Millennial	7.056***	(2.526)	[0.000]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	AIAN 2.099† (0.806) [0.054]	Baby Boomer	1.472	(1.093)	[0.602]		AIAN 0.814 (0.381) [0.660]	Baby Boomer	1.075	(1.128)	[0.945]
		GenX	1.084	(0.296)	[0.767]			GenX	0.478	(0.283)	[0.212]
		Millennial	8.314***	(4.987)	[0.000]			Millennial	1.293	(0.984)	[0.736]
		GenZ	Insufficient observations					GenZ	Insufficient observations
	Mixed/Other 2.078*** (0.260) [0.000]	Baby Boomer	1.608†	(0.418)	[0.068]		Mixed/Other 1.440** (0.173) [0.002]	Baby Boomer	1.202	(0.263)	[0.399]
		GenX	1.925***	(0.355)	[0.000]			GenX	1.518*	(0.269)	[0.018]
		Millennial	2.130***	(0.423)	[0.000]			Millennial	1.141	(0.248)	[0.544]
		GenZ	24.114***	(6.539)	[0.000]			GenZ	4.636***	(1.267)	[0.000]

Note: Standard errors in parentheses; p-values in brackets..

^†p < .1;

^*p < .05;

^**p < .01;

^***p < .001.

Our results show strong evidence of racial disparities in overall burnout outcomes. Compared to white respondents, all groups except for Asian and NHPI score on average lower in overall burnout. Black respondents score 0.358 points lower, Hispanic respondents score 0.167 points lower, AIAN respondents score 0.274 points lower, and Mixed/Other respondents score 0.102 points lower. Asian and NHPI respondents show no statistical difference. We see similar trends with emotional exhaustion where black respondents, on average, score 0.325 points lower, Hispanic respondents score 0.156 points lower, AIAN score 0.221 points lower, and Mixed/Other score 0.091 points lower than their white counterparts. Asian respondents score on average 0.092 points higher in emotional exhaustion. With depersonalization, white respondents are most likely to score lower. On average, black and Hispanic respondents score 0.101 and 0.072 points higher, respectively, than their white counterparts in depersonalization. In contrast, white respondents are more likely to achieve a sense of personal accomplishment than Asian, NHPI, and Mixed/Other respondents. On average, Asian respondents have 58.5% higher odds of experiencing a loss of personal accomplishment, whereas NHPI and Mixed/Other respondents have 111.4 and 43.1% higher odds, respectively, than white respondents.

Our results indicate that burnout outcomes also differ across generations. Compared to GenX respondents, Baby Boomers on average score 0.209 points lower on overall burnout. Millennials and GenZ respondents score 0.153 and 0.546 points, respectively, higher than their GenX counterparts. We see a similar relationship with emotional burnout where Baby Boomers score 0.211 points lower, Millennials score 0.141 points higher, and GenZ score 0.453 points higher than GenX respondents. In terms of depersonalization, only millennials are statistically different from GenX with 0.065 lower points. We find no statistical difference among generations in their perception of whether they believe their work positively influences people’s lives.

How Gender Disparities in Burnout are Jointly Mediated by Race

Although our results show that there is no statistical difference between women and men in overall burnout, we find there are nuanced differences among women—particularly women of color—when adding the gender*race interaction term. Compared to white men, the double-axis model shows that black and Hispanic women statistically score .347 and .184 points lower on overall burnout (see Supplementary Appendix A). Compared to their white counterparts, black men score .303 points lower, AIAN score .380 points lower, and Mixed/Other score .110 points lower.

This intersectional approach also reveals racial differences in emotional exhaustion across gender (see table 1). Compared to white men, black and Hispanic women, respectively, score 0.292 and 0.153 points lower, whereas Asian women on average score 0.137 points higher on emotional exhaustion. Black, AIAN, and Mixed/Other men on average score 0.286, 0.313, and 0.107 points lower than their white counterparts. This shows clear gender cleavages in experienced burnout within racial categories for Hispanic, Asian, AIAN men versus women, with only black men and women both showing significant results in the same direction.

For emotional exhaustion, we see very similar trends mirroring overall burnout in our double-axis column, but some interesting distinctions emerge for depersonalization and loss of personal accomplishment. With depersonalization (see table 2), we see that white (0.086), black (0.216), Hispanic (0.211), Asian (0.133), and Mixed/Other (0.147) women on average score higher than white men. Among men, AIAN men (0.364) score higher than their white counterparts, whereas other racial groups show no statistical difference.

Across all racial groups, women have higher odds than white men in experiencing a loss of personal accomplishment (see table 3). In particular, white women have 47.1% higher odds; black women have 78.5% higher odds; Hispanic women have 57.2% higher odds; Asian women have 127.3% higher odds; NHPI women have 315.5% higher odds; and Mixed/Other women have 107.8% higher odds. Among men, Asian and Mixed/Other respondents were the only groups that show statistical differences from white men in terms of loss of personal accomplishment. Asian men have 59.3% higher odds and Mixed/Other men have 44.0% higher odds of experiencing loss of personal accomplishment. Full double-axis model results, including controls, are available in Supplementary Appendix A.

The Impact of Generation on Gendered and Racialized Experiences With Burnout

When including the interaction term gender*race*generation, we find generational variation across gender and race. When compared to white men (GenX)—the largest subgroup in our sample—Baby Boomers and other GenX respondents across gender and race generally score lower on overall burnout with the starkest difference being AIAN (Baby Boomer) and black men (Baby Boomer) who, on average, score 0.715 and 0.678 points lower, respectively (see Supplementary Appendix A). When looking at Millennials, the results differ across gender and race with black women, on average, scoring 0.257 points lower than white men (GenX), whereas Asian women and Asian men score 0.300 and 0.296 points higher, respectively. Although our findings are sensitive to the small GenZ sample, these respondents across the groups with observations are more likely to score higher on overall burnout than White men (GenX), except for white and Mixed/Other men from GenZ.

We see similar trends with emotional exhaustion (see table 1) where statistical findings show Baby Boomers and other GenX respondents across gender and race score lower than white men (GenX). Among women, the starkest difference is with AIAN women (Baby Boomer) and black women (Baby Boomer) who score 0.626 and 0.458 points lower, respectively. Among men, the starkest differences include black men (Baby Boomer) who score 0.640 points lower and AIAN men (Baby Boomer) who score 0.636 points lower. Statistically significant estimates for Millennials show Asian women, Asian men, and Mixed/Other women to score 0.300, 0.303, and 0.194 points higher, respectively, on emotional exhaustion than white men (GenX). Similarly, GenZ respondents generally score higher on emotional exhaustion than white men (GenX), except for white and Mixed/Other men from GenZ.

With depersonalization (see table 2), women across races and generations tend to score higher than white men (GenX) except for Asian women (GenZ) who score 0.426 points lower. Among men, only white men (GenZ) and Asian men (GenZ) show statistical differences in depersonalization compared to white men (GenX). White men (GenZ) on average score 0.607 points lower, whereas Asian men (GenZ) score 0.317 points lower. As shown in table 3, women across race and generation are more likely to have higher odds than white men (GenX) in experiencing a loss of personal accomplishment with statistically significant findings for white women (GenX), black women (GenX), Hispanic women (Baby Boomer), Asian women (Baby Boomer, GenX, Millennial), NHPI women (Baby Boomer, Millennial), AIAN women (Millennial), and Mixed/Other women (GenX, Millennial, GenZ). Among men, the only statistical findings were for Asian men (GenZ), NHPI men (Millennial), and Mixed/Other men (GenZ) who, on average, have higher odds of experiencing a loss of personal accomplishment. Full triple-axis model results, including controls, are available in Supplementary Appendix A.

Discussion

Intersectional research in public administration is still nascent. Our results show that younger generations of women of color are particularly vulnerable to burnout, but the experience of burnout is not uniform across groups, with each dimension of burnout revealing different vulnerable groups. Cities are becoming increasingly diverse, and meeting the needs of the public necessitates a workforce more representative of the communities being served (Ding et al. 2021). With the goal of better supporting an increasingly diverse public workforce, these findings highlight the importance of deconstructing burnout into its discrete dimensions to better understand the experiences of different socio-demographic groups of employees and develop culturally competent managerial and organizational practices.

Low Utility of Single-Axis Approaches to Burnout

Our first hypothesis is partially supported as some, but not all, women of color seem to experience higher levels of burnout compared to their white male counterparts, with meaningful heterogeneity of experiences of burnout surfacing among women of color based on both racial and generational differences. In a single-axis model, we might ignore gender and race altogether because the results show no statistical difference. Even when we introduce a double-axis intersectional model that looks at the interaction of race and gender, we do not see significant differences across groups. However, we do see statistically significant and nuanced differences when we intersect gender, race, and generation. Groups that emerge as more vulnerable to burnout with a triple-axis lens include women who are white GenZ, Asian Millennial, Asian GenZ, Hispanic GenZ, and mixed/other GenZ, as well as men who identify as Hispanic GenZ, Asian Millennial, or Asian GenZ. Curiously, and contrary to our first hypothesis, we do not see black men or black women of any generation more vulnerable to burnout than the reference group when looking at burnout in its aggregate form.

Ultimately, this suggests that for burnout, race-only and gender-only analyses are prone to overlook vulnerable subgroups. We see greater utility emerging when we look at combined gender and racial identity along generational lines. In particular, we see how younger generations of women in the workforce, regardless of race, are experiencing more burnout in varied forms.

Utility of Disaggregated Burnout

We apply these same intersectional models to the discrete dimensions and find strong support for our third hypothesis that the salience of each dimension of burnout would differ for each socio-demographic group, with new groups emerging as vulnerable who did not appear when looking at burnout in its aggregate form (tables 4–6). Even within the single-axis approach, we see important differences based on the burnout dimension (see table 4). For emotional exhaustion, we see no significant difference based on gender but find Asians and younger generational cohorts (Millennials and GenZ) reporting higher levels. Interestingly, for depersonalization and loss of personal accomplishment, we find statistical differences in greater vulnerabilities across gender, but not generation. We see a much wider range of racial groups showing greater vulnerabilities beyond Asians, with black and Hispanic respondents reporting greater depersonalization and NHPI and those identifying as Mixed/Other having lower odds of feelings of personal accomplishment. The importance of disaggregating burnout into its multiple dimensions is most clear in our triple-axis model where we see statistically significant differences among groups in terms of their vulnerability to experience emotional exhaustion, depersonalization, or loss of personal accomplishment (see table 6).

Table 4.

Vulnerabilities to Different Dimensions of Burnout by Gender, Race, and Generation (Single-Axis)

	Single-Axis (ref: Men, White, GenX)
	Gender	Race	Generation
Agg	None	Asian	Millennials GenZ
EE	None	Asian	Millennials GenZ
Dep	Women	Black Hispanic	None
PA	Women	Asian NHPI Mix/Other	None

Agg = More Aggregated BurnoutEE = More Emotional Exhaustion.

Dep = More DepersonalizationPA = Less Personal Accomplishment.

Table 5.

Vulnerabilities to Different Dimensions of burnout by gender, race, and generation (Double-Axis)

	Double-Axis (ref: White men)
	Gender*Race
Agg	Asian women; Mix/Other men
EE	Asian women; Mix/Other men
Dep	Asian, Black, Hispanic, White, Mix/Other women; Asian, AIAN men
PA	Asian, Black, Hispanic, NHPI, White, Mix/Other women; Asian, Mix/Other men

Agg = More Aggregated BurnoutEE = More Emotional Exhaustion

Dep = More DepersonalizationPA = Less Personal Accomplishment

Table 6.

Vulnerabilities to Different Dimensions of Burnout by Gender, Race, and Generation (Triple-Axis)

	Triple-Axis (ref: GenX White men)
	Gender*Race*Generation
Agg	Millennials: Asian women; Asian men GenZ: Asian, Hispanic, Mix/Other, White women; Asian, Hispanic men
EE	Millennials: Asian, Mix/Other women; Asian men GenZ: Asian, Hispanic, White, Mixed/Other women; Asian, Hispanic men
Dep	Baby Boomers: Asian, Black, Hispanic, Mix/Other women; Asian men GenX: Asian, Black, Hispanic, Mix/Other, White women GenZ: White women
PA	Baby Boomers: Asian, Hispanic, NHPI women GenX: Asian, Black, Mix/Other women; Mix/Other men Millennials: Asian, AIAN, Mix/Other, NHPI women GenZ: Mix/Other women; Asian, Mix/Other men

Agg = More Aggregated BurnoutEE = More Emotional Exhaustion.

Dep = More DepersonalizationPA = Less Personal Accomplishment.

Through our application of COR theory, we posited that different socio-demographic groups have varied resource sets available to protect them from some dimensions of burnout while being vulnerable to others. While we do not find support for COR theory’s proposition that resource loss cycles happen in a universally ordered manner, we do find differential vulnerabilities to dimensions of burnout across intersectional socio-demographic groups. We find that, given the cultural origins of resource caravans, resource loss cycles and their sequence can differ across intersectional identities in both emphasis and order. For some groups, such as Asian women, we see greater vulnerability to all three dimensions of burnout simultaneously, whereas black women only appear more vulnerable to depersonalization and loss of personal accomplishment. Millennials show vulnerability to what COR theory would suggest are the first (i.e., emotional exhaustion) and final stages (i.e., loss of personal accomplishment) in the process of burnout while showing no increased vulnerability to the middle stage of depersonalization.

By deconstructing burnout, our data signal that different subgroups within broader categories of race, gender, and generation may be responding to interpersonal stressors on the job over time in quite different ways. Other scholars who approach burnout from a different theoretical orientation beyond COR theory still conceptualize burnout as a process but hypothesize depersonalization diminishes feelings of personal accomplishment, and that lack of accomplishment results in higher emotional exhaustion (Golembiewski 1989; Golembiewski and Munzenrider 1981) . Our work shows that the generalizability of these theories is dependent on both context and how one conceptualizes identity groups for analysis.

We do not believe that burnout and respective vulnerabilities as a function of intersectional identity will generalize across settings. The neighboring cities at the center of our study are embedded in a region that has a distinctly different socio-demographic makeup and history of racial and ethnic dynamics than other regions. Rather, we suggest that any human resource management strategy that fails to account for the multidimensionality of burnout in tandem with the intersectional nuances of individual identities in the workforce is likely to develop practices that only help some while largely missing the needs of others. Past studies of burnout that do not account for intersectional nuances may have masked evidence that the experience of burnout does not necessarily follow a predetermined sequence in isolated stages for all subgroups. We suspect this may be based on the cultural resources different groups have available to cope with job demands. If burnout is a process, the process may be culturally specific and sequencing of dimensions may vary if they are sequenced at all, warranting further exploration.

Other Covariates of Interest

Importantly, the varied signs and significance for the organization-level and individual-level controls in our model further suggest that an analysis that only looks at burnout in its aggregate form misses important contextual nuances for its various dimensions. While a full analysis of the differences between each control in our models across burnout dimensions is beyond the scope of this paper, we do want to highlight the nuances of how these controls impact depersonalization relative to the other dimensions of burnout. Our findings show that those high in autonomy report lower overall burnout and less emotional exhaustion but more depersonalization.

Emotional labor scholars have shown that job autonomy that includes discretion about the emotions displayed will expand an individual’s behavioral choices so they can selectively engage colleagues and residents in ways that produce less emotional dissonance (Grandey et al. 2013). Curiously, the relationship with autonomy and depersonalization moves in the opposite direction, with those who experience more autonomy reporting more depersonalization and no significant relationship with a reduced sense of personal accomplishment. Following the process model of burnout from COR theory, it may be the case that some amount of depersonalization is adaptive, and having the autonomy to distance oneself from others is an effective protective mechanism against resource loss (Hochschild 1983).

Beyond the individual benefits, job autonomy and discretion are also key in activating representation. Research on conformity argues underrepresented groups are encouraged to adopt the same values, informal norms, and formal practices as their white male colleagues, which leaves unanswered what advantages of representation are lost through assimilation practices (Martin 2000). This finding, in particular, draws attention to the need for future research to explore autonomy and other organizational and job characteristics that influence burnout along each of its discrete dimensions using longitudinal data to understand cascading effects.

Limitations

Our analysis uplifts the significance of examining burnout outcomes reproduced at the intersection of multiple identities, as unique burnout experiences are facilitated by multiplicative mechanisms. As our results show, cultural, gender, and generational differences impact groups differently and are often hidden by overly aggregated analyses. However, it is worth noting how these outcomes can be further moderated by other socio-demographic dimensions we were unable to control due to data limitations, including immigration status and socioeconomic class. Although our current dataset does not have this information, we believe that they are important factors to further interrogate in future analyses of understanding cultural differences in mediating burnout.

Immigration status plays a significant role in studying burnout as public sector workers may be burdened with concerns for their safety and security during a time of heightened anti-immigrant rhetoric and state-sanctioned enforcement (Rathod 2010). At the same time, naturalized immigrants may also be more socially motivated to serve in local government and have unobserved characteristics correlated with their motivation to become citizens as barriers to work in local government are much greater than for their native-born counterparts (Lewis et al. 2014) Similarly, socioeconomic class can influence burnout outcomes as a proxy to financial stability and access to external resources (e.g., therapy). Those in more precarious financial situations are more vulnerable to shocks that can contribute to burnout. However, there is likely to be a nonlinear relationship as those with higher socioeconomic status due to their position may also face more responsibilities and expectations that can increase burnout (Kim and Wang 2018).

Additionally, research through an intersectional lens requires sampling power that can provide valid estimates. Given the nature of surveys and sampling, an intersectional approach may be vulnerable to bias if there is not an effective sample size that can capture variation. We can see this, for example, with certain GenZ groups as we further disaggregate by different socio-demographic dimensions. To address this issue in future research, a more intentional framework of intersectionality needs to be applied from the beginning of the research design to obtain a representative sample of individuals by intersectional dimensions of identities.

Another limitation of this study is the lack of sociodemographic administrative data from the cities, which would allow us to weight our sample more accurately at the intersection of gender, race, and generation. The personnel departments of the two cities were able to share the percentage of women and employees of color by race in the workforce, but not women of color. Additionally, only one of the two cities shared the ages of their employees, making it difficult to apply meaningful survey weights for our models. Furthermore, neither city collects nor reports educational level data also important in building accurate sample weights. We believe our sample is reflective of the diverse gender and racial identities and experiences of the local government workforce as we were intentional in sampling traditionally undersampled groups, such as women and people of color. Future researchers should work in collaboration with practitioners and partners in the field to ensure public organizations are collecting and reporting demographic data about their workforce in ways that would allow for more nuanced intersectional analyses. Lastly, though we include organizational and departmental level characteristics to serve as a proxy for job characteristics, those controls may not fully capture the nuanced job demands of individuals in the study.

Conclusions

As urban areas and the public organizations that serve them are becoming more diverse, there is a growing need to understand the existing racial, gender, and generational disparities in public employee well-being that are driven by institutionalized inequities. By identifying the specific mechanisms negatively impacting public administrators with multiple minoritized social identities, public organizations can strategize best practices to not only improve public employee well-being among diverse employees but also increase efficacy in navigating interactions with residents across cultural boundaries.

We show vulnerabilities to burnout differ across dimensions considerably for distinct groups. These findings have substantial implications for theories of representative bureaucracy, emotional labor, public engagement, and empirical investigations employing those theories. For instance, it is necessary to understand how passive representation can be activated for a focal demographic by first understanding the intersectional dynamics at play (Headley et al. 2021). If some intersectional identities that fall within a larger demographic of interest are more susceptible to burnout or one of its particular dimensions, active representation could be stymied.

The differences in burnout across socio-demographic groups and the distribution of perceived stress might be associated with (1) interpersonal interactions with residents, (2) perceived demographic (dis)similarity with co-workers, and (3) social support both within and outside the organization. Each could have profound impacts on the work of public administrators in occupations necessitating emotional labor in public engagement, where burnout is prevalent (Resh, Barboza-Wilkes, and Mooradian 2020). It could certainly complicate the ability to activate representation for any single demographic in an organization (Meier and Nicholson-Crotty 2006).

Public administration research and theory have ignored the link between emotional labor and race (Humphrey 2021). However, the extent to which particular cross-sections within an ethnic or racial group are more or less vulnerable to emotional exhaustion should be of concern in exploring that link. For instance, it has long been established that emotional labor carries gendered (if implicit) connotations and is often ignored in reward or remuneration in public organizations (Guy, Newman, and Mastracci 2008). Understanding what intersections between race and gender, and under what contexts, emotional exhaustion can be more prevalent should be an important consideration. As well, depersonalization has been shown to lead to individuals valuing minority status more than majority status (Brewer, Manzi, and Shaw 1993). Hence, a focus on how intersectional identities are distributed across a larger group might matter greatly in activating representation for the larger focal group, even if that focal group remains a minority in the larger organization.

Racial differences may be more or less salient based on gender and/or generation. An organization’s representative intentions must be coupled with creating and sustaining the resources necessary to buffer against the stress of interacting with residents across cultural boundaries. This should be better explored through a contextualized intersectional lens. We know little about how critical mass forms for a single group across multiple intersecting identities. More work needs to be done to explore whether gender and generational (dis)similarity to residents and co-workers might confound theoretical expectations and empirical evidence.

Our results show that the intersection of gender, race, and generation creates meaningful distinctions in the experiences of emotional exhaustion, depersonalization, and loss of personal accomplishment. In alignment with Maslach et al. (2008), we avoid trying to identify whether a group is “burnt out,” which implies crossing an artificial threshold, and instead emphasize making visible the differential vulnerabilities to various combinations of burnout dimensions across groups. Our results highlight important heterogeneity within gender, racial, and generational groups across the multiple dimensions of burnout. These findings have implications on a workforce’s ability to activate representation on behalf of any focal group. They reinforce the need to look at the different ways in which individual-level and organizational-level factors impact the development and maintenance of resources necessary to buffer against emotional exhaustion, depersonalization, or loss of personal accomplishment separately and in the aggregate.

The California context of our study may not be generalizable to other locales where racial, gender, and generational identities all have localized histories. Because of those regional nuances, we anticipate there will likely be differences in burnout for different subgroups across various geopolitical contexts. This paper reveals that differences between groups do exist, and further exploration is needed to understand these intersectional dynamics.

Acknowledgments

We thank the journal’s editor and anonymous reviewers for their constructive feedback in improving this paper. We thank Jason Coupet and Frank Zerunyan for their respective insights on this topic. We would like to also acknowledge the funding support from the John Randolph and Dora Haynes Foundation.

Data Availability

The data and instrument underlying this article are available in the Harvard Dataverse at https://doi.org/10.7910/DVN/QEZMPN.

Footnotes

1

Data collection and research approved by University of Southern California’s Institutional Review Board (Protocol #: UP-20-00286).

2

The two items used to measure depersonalization are measured on a Likert scale from 1 = strongly agree to 7 = strongly disagree. Question wording for the first item “I feel similar to residents in many ways” relates to general sense of relatability to residents, and the second item “I feel personally involved with residents’ problems” relates to more specific investment in the personal issues raised by residents in the work context. We suspect that the distinction between generalized depersonalization and task-based depersonalization is what drives the difference between the two items, resulting in a relatively low eigenvalue. Despite the low eigenvalue, we include this factor based on theory from the burnout literature which suggests depersonalization reflects a psychological dissimilarity from others, reflected in item one, and a detached response reflective of lowered emotional or cognitive involvement in work tasks specifically, reflected in item two (Maslach et al. 2008). Importantly, we run the models with and without the depersonalization items, and significance does not change the results for overall burnout.

3

Overall burnout factor scores range from −1.77 to 2.23, emotional burnout factor scores range from −1.57 to 1.93, and depersonalization factor scores range from −1.32 to 1.46.