Motivation Mix and Agency Reputation: A Person-Centered Study of Public-Sector Workforce Composition

Abstract

Identifying what motivates public servants and how those motives vary across agencies is essential for both theory and practice, yet most existing “types of bureaucrats” remain untested against real workforces. Drawing on reputation theory, which posits that external audiences’ beliefs shape who seeks and retains employment in an organization, we theorize that agency reputation will systematically sort employees into distinct motivational profiles. We analyze survey data from 13,471 U.S. federal employees merged with an externally derived, 40-year measure of agency reputation based on congressional speeches. A multi-level latent class analysis uncovers four robust motivation types—All-rounders (35%), intrinsically focused Job-motivated (25%), Self-interested (24%), and Amotivated (16%)—and two clusters of agencies distinguished by their profile mix. Reputational standing predicts profile membership: employees in highly reputed agencies are significantly more likely to be Job-motivated and less likely to be Self-interested or Amotivated, consistent with self-selection and socialization mechanisms highlighted in the extant literature. These findings validate classic typologies while demonstrating the value of integrating organizational-level reputation into motivation research, and they imply that recruiting and retention strategies should be tailored to the reputational context of each agency.

1. Introduction

Understanding what motivates public servants—and how those motives vary across agencies—is central to public administration (Perry & Wise, 1990). Classic and contemporary perspectives recognize heterogeneous motivations and urge integration across traditions (Downs, 1967; Ali et al., 2021; Andersen et al., 2018; Esteve & Schuster, 2019; Robertson & Tang, 1995), complemented by behavioral insights from psychology (Grimmelikhuijsen et al., 2017; Tummers et al., 2016). Yet much of what we know about “types of bureaucrats” still rests on theory rather than on the composition of actual public workforces.

Two gaps motivate this study. First, influential typology-building efforts often infer types from general premises, narrow policy domains, or a single theoretical lens (e.g., Gailmard & Patty, 2007; Bowling et al., 2004; Moynihan, 2013), leaving limited evidence on which motivational profiles are most prevalent in real organizations (Baekgaard et al., 2022; Nørgaard, 2018). Second, although organizational reputation—audiences’ beliefs about an agency’s capacities, intentions, history, and mission—should shape attraction, socialization, and retention (D. Carpenter, 2001; Heclo, 1978; Kingdon, 1984; Moe, 1989; see also Lasswell, 2018), its link to motivational profiles remains underdeveloped. Emerging work suggests that reputation-oriented mechanisms can bolster engagement and extra-role behaviors (Rho et al., 2015; Hameduddin & Lee, 2021), and classic arguments imply that mission and culture matter for attraction and performance (Rainey & Steinbauer, 1999). Building on self-selection logic in public service (Gailmard, 2010), we expect motivational compositions to vary systematically with agency reputation.

We advance these literatures by mapping public employees’ motivational profiles across a representative federal workforce and linking those profiles to agency reputation using an external, longitudinal measure derived from congressional speeches. Using a person-centered approach, we characterize profile prevalence within and across agencies and examine individual and organizational correlates. Organizational reputation does more than insulate agencies; it structures who is attracted to, socialized in, and retained by them (D. Carpenter, 2001; D. P. Carpenter & Krause, 2012). Two mechanisms are especially salient. First, self-selection: in the attraction–selection–attrition logic, prospective employees infer mission quality, professional standards, and work conditions from reputational cues and gravitate toward agencies whose standing signals a better match with their intrinsic and prosocial motives (Rainey & Steinbauer, 1999; Gailmard, 2010; Kjeldsen & Jacobsen, 2013; Wright & Christensen, 2010; Goodsell, 2011). Second, socialization and identification: once inside, employees internalize reputational narratives—prestige, identity, image—linking self-concept to organizational standing, which sustains engagement and extra-role behavior (Albert & Whetten, 1985; Dutton et al., 1994; Brown et al., 2006; Rho et al., 2015; Hameduddin & Lee, 2021). This reputational channel complements integrative theories of public-sector motivation spanning intrinsic, extrinsic, and prosocial drivers (Andersen et al., 2018; Esteve & Schuster, 2019) and aligns with political–institutional accounts in which reputation enables resources, discretion, and insulation that shape everyday work (Busuioc & Lodge, 2016; Moffitt, 2010; Salomonsen et al., 2021). Because insider and outsider views can diverge (Brown et al., 2006; Helm, 2013), we pair a subjective, insider-based index (Partnership for Public Service, 2016) with an external, audience-based measure constructed from congressional speech over four decades (Bellodi, 2022; Overman et al., 2020) to test whether an agency’s reputational standing is systematically associated with its motivational profile mix.

This yields the following set of questions:

(1)

What are public employees’ primary workplace motives, and how are they distributed across individuals and agencies?

(2)

Which motivational profiles are most prevalent, and how do they vary by sociodemographic characteristics?

(3)

How is agency reputation related to the likelihood that employees belong to one profile versus another?

In brief, we identify four robust profiles—All-rounders, Job-motivated, Self-interested, and Amotivated Bureaucrats—and two agency clusters that differ in their mix. Reputational standing is positively associated with the prevalence of Job-motivated Bureaucrats and negatively associated with Self-interested and Amotivated Bureaucrats. We conclude with implications for typology building, reputation theory, and recruitment/retention strategy.

2. Literature Review

2.1. Theoretical Foundations of Bureaucratic Motivation: From Downs (1967) to Perry (2020)

A typology is a “complex theoretical statement developed to identify ideal types” (Doty & Glick, 1994, p. 243). Downs’ (1967) classic model remains foundational: Climbers (extrinsically oriented toward power, income, prestige), Conservers (security/convenience), Zealots (narrow, policy-focused loyalty), Advocates (organizational domain expansion with high commitment), and Statesmen (broad public goals, akin to PSM/altruism). While grounded in rational choice (Downs, 1967, p. 2), Downs drew on Simon’s bounded rationality and organizational identification (Simon, 1947/1997), emphasizing that real officials mix motives. Despite its influence (Rainey, 2014, pp. 259–260), empirical tests are limited (Arapis & Bowling, 2020; Brewer & Maranto, 2000), and the field often presumed the prevalence of self-interested types (Pandey et al., 2008), dovetailing with bureaucratic control and political economy perspectives (Meier et al., 2006).

Subsequent work, frequently in budgeting, reaffirmed heterogeneity. Bowling et al. (2004) mapped Conservers to “budget minimizers”, Climbers to “maximizers”, and Statesmen to “altruists”, with variation by self-interest, ideology, and agency attachment. Countering Niskanen’s (1971) archetype, Moynihan (2013) showed that budget maximization can reflect public-spirited motives (PSM). Gains and John (2010), building on Dunleavy’s (1991) bureau-shaping, distinguished “classic policymakers” from “implementers”, contingent on task environment and political interaction. Beyond budgeting, scholarship contrasted self-interest with public-regarding motives (Le Grand, 2003, 2010; Piatak & Holt, 2020; Ritz et al., 2020), and incorporated role perception (Pfiffner, 1987), civil service ethic (Hollibaugh et al., 2020; O’Leary, 2010), and mission/culture (Dilulio, 1994; Goodsell, 2011), challenging the pure utility-maximizer view (Downs, 1967; Fiorina, 1981; Mayhew, 1974; Niskanen, 1971). Typology efforts within PSM likewise identified diverse ideal types (Brewer et al., 2000) and role images (Delegate, Trustee, Politico) conditioned by motivational assumptions (Wise, 2004; Pitkin, 1967). In sum, theory implies multiple coexisting motivational logics; what remains underexplored is how those logics are distributed in actual workforces across agencies.

2.2. Syntheses and Integration of Theoretical Perspectives

Calls to bridge traditions are longstanding (Robertson & Tang, 1995; Shepsle, 1991). Recent syntheses specify complementary drivers—extrinsic, intrinsic, and prosocial—and their theoretical anchors (Andersen et al., 2018). Esteve and Schuster (2019) assemble a two-dimensional map (self vs. other; intrinsic vs. extrinsic) drawing on self-determination, principal–agent, social identity, prosocial motivation, and commitment theories; Ali et al. (2021) broaden the managerial lenses. These works clarify constructs and mechanisms, but they remain largely conceptual or variable-centered. Two limitations follow. First, they seldom empirically map the composition of motivation at the person level across a representative public workforce, leaving open which profiles are actually prevalent. Second, they under-specify cross-level linkages from organizational attributes to employees’ motivational profiles. Perry (2020), for example, advances leadership and management practices to “sustain passion”, yet provides limited guidance on how organization-level conditions (beyond managerial levers) sort or socialize employees into distinct motivational mixes. Likewise, while Esteve and Schuster (2019) integrate theories, they do not test boundary conditions under which organizational contexts selectively amplify intrinsic/prosocial over extrinsic orientations. The present study addresses both limitations by (a) using a person-centered approach to recover latent motivational profiles and (b) theorizing and testing a specific organization-level driver—agency reputation—that should shape profile prevalence via attraction–selection–attrition and socialization processes.

2.3. Motivation and Organizational Reputation

At the organization level, determinants of motivation span structure, leadership, and climate (Esteve & Schuster, 2019), yet reputation—audiences’ beliefs about an organization’s capacities, intentions, history, and mission—remains comparatively undertheorized as a motivational force (D. P. Carpenter & Krause, 2012; D. Carpenter, 2002). Reputation is a political asset that conditions support, delegated authority, and insulation (Heclo, 1978; Kingdon, 1984; Moe, 1989), and it increasingly features in public administration (Busuioc & Lodge, 2016; Moffitt, 2010; Salomonsen et al., 2021). However, reputational research has tilted toward institutional persistence and exogenous threats (Maor, 2015) and often bifurcates into political-science and organizational strands (Meier & O’Toole, 2006; Bustos, 2021), leaving micro-level motivational consequences understudied. Recent work underscores this linkage, showing that institutional design and reputation cues jointly shape motivation (Yu, 2023).

We advance an explicit two-mechanism account linking reputation to motivation. First, self-selection (attraction–selection–attrition): reputational cues signal mission quality, professional standards, and work conditions; applicants with stronger intrinsic/prosocial motives sort into agencies with higher standing (Rainey & Steinbauer, 1999; Gailmard, 2010; Kjeldsen & Jacobsen, 2013; Wright & Christensen, 2010; Goodsell, 2011). Second, socialization and identification: inside the organization, employees internalize reputational narratives—prestige, identity, image—aligning self-concept with organizational standing and reinforcing engagement and extra-role behavior (Albert & Whetten, 1985; Dutton et al., 1994; Brown et al., 2006; Rho et al., 2015; Hameduddin & Lee, 2021). These mechanisms situate motivation within political–institutional dynamics by which reputation enables resources, discretion, and insulation that shape day-to-day work experiences (Busuioc & Lodge, 2016; Moffitt, 2010; Salomonsen et al., 2021) and complement integrative motivation theories (Andersen et al., 2018; Esteve & Schuster, 2019).

Measurement complicates this linkage. Much evidence relies on self-reported or perceived external reputation (Helm, 2013), which may diverge from audiences’ assessments (Brown et al., 2006; Overman et al., 2020). To reduce same-source bias and capture audience beliefs, we employ an external, demand-side measure constructed from four decades of congressional speech (Bellodi, 2022), while also considering insider-based indices (e.g., Partnership for Public Service, 2016). This dual lens responds to calls for validated, multi-source operationalization of public-sector reputation and shifts emphasis from supply-side reputation management (Anastasopoulos & Whitford, 2019; Lee & Whitford, 2013; Maor et al., 2013) to reputational standing as an organizational condition that can shape the motivational profile mix of agencies. On this view, highly reputed agencies should exhibit a higher prevalence of employees whose motivation is anchored in intrinsic aspects of work (interest, contribution, expertise) and a lower prevalence of employees primarily oriented toward extrinsic security—or those who are amotivated.

3. Empirical Strategy

3.1. Data

To examine (1) the distribution of individuals’ motivational sources and (2) how those are related to agency reputation, this study estimates a multilevel latent class analysis (LCA) on a set of indicators (

Table 1. Summary statistics.

Variables	N	Mean	SD	Hist.	Min	Median	Max
Key motives
Interesting work	13,471	0.78	0.41	▂▁▁▁▇	0	1	1
Feeling of respect	13,471	0.72	0.45	▃▁▁▁▇	0	1	1
Opportunity to exercise expertise	13,471	0.72	0.45	▃▁▁▁▇	0	1	1
Being included in decision making processes	13,471	0.64	0.48	▅▁▁▁▇	0	1	1
Making positive contributions	13,471	0.78	0.42	▂▁▁▁▇	0	1	1
Job security	13,471	0.74	0.44	▃▁▁▁▇	0	1	1
Pay	13,471	0.66	0.48	▅▁▁▁▇	0	1	1
Performance-based bonuses	13,471	0.49	0.50	▇▁▁▁▇	0	0	1
Benefits	13,471	0.79	0.41	▂▁▁▁▇	0	1	1
Opportunity to use innovative technology	13,471	0.41	0.49	▇▁▁▁▆	0	0	1
Learning/development opportunities	13,471	0.51	0.50	▇▁▁▁▇	0	1	1
Opportunity for greater responsibility	13,471	0.53	0.50	▇▁▁▁▇	0	1	1
Opportunity for advancement	13,471	0.46	0.50	▇▁▁▁▆	0	0	1
Work-life balance	13,471	0.75	0.44	▃▁▁▁▇	0	1	1
Work flexibility	13,471	0.62	0.48	▅▁▁▁▇	0	1	1
Demographics (level 1)
Agency tenure (1 = 3 years or less, 2 = 4 years or more)	12,022	1.91	0.29	▁▁▁▁▇	1	2	2
Supervisory status	12,010	2.25	1.30	▇▂▅▂▁	1	2	5
Salary	11,917	2.53	1.02	▅▅▁▇▅	1	3	4
Minority status (0 = non-minority, 1 = minority)	11,741	0.33	0.47	▇▁▁▁▃	0	0	1
Sex (1 = male, 2 = female)	11,863	1.42	0.49	▇▁▁▁▆	1	1	2
Age (1 = 39 and under, 2 = 40 and over)	11,881	1.86	0.35	▁▁▁▁▇	1	2	2
Education (1 = Less than AA, 2 = AA, 3 = Graduate)	11,938	2.22	0.73	▃▁▇▁▇	1	2	3
Agency variables (level 2)
Reputation (Bellodi, 2022)	13,471	0.61	0.16	▁▁▂▇▆	0.05	0.61	0.82
Reputation (Partnership for Public Service, 2016)	13,471	61.9	7.9	▂▃▇▂▂	45.8	63.1	80.7
Agency political ideology	13,471	0.26	0.96	▃▇▅▇▅	−1.5	0.09	1.93
Decision making independence (Selin, 2015)	13,058	−0.23	0.37	▇▂___	−0.69	−0.36	1.31
Political independence (Selin, 2015)	13,058	0.45	0.70	▇▆___	−0.36	0.37	3.57
Agency politicization	13,058	30.7	28.2	▇▂___	0	23.6	201

) that reflect various workplace motivations of public employees. LCA is used to identify unobserved types of employees characterized by distinct motivational portfolios and to estimate each type’s share in the population. After deriving the bureaucratic typologies, each type is set as the dependent variable and modeled as a function of agencies’ reputations, key agency features, and individual demographic characteristics.

Data come from the 2016 Merit Principles Survey administered by the U.S. Merit Systems Protection Board (MSPB). MSPB uses a random sample stratified by agency and supervisory status to provide a representative sample of permanent full-time federal employees. The sample was drawn in 2015 and included nearly 126,000 employees from 25 federal agencies; data were collected from July through September 2016. The MPS 2016 comprised three paths—Path 1, Path 2, and Path L—with distinct samples (e.g., Path L for supervisory positions) and topical modules (e.g., ethics in Path 2). This study uses Path 1, which includes items on the perceived importance of job and work-environment attributes. Path 1 had a response rate of 38.8% (14,515/37,452). To preserve a common analytic sample across model comparisons, observations with missing values on the motivation indicators or on level-1/level-2 covariates (e.g., selected demographic items and agency-level measures such as decision-making independence, political independence, politicization, or ideology) were excluded listwise. Item nonresponse on the indicators was limited relative to the full Path 1 file, and the resulting analytic sample continues to span the set of agencies included in the survey frame.

3.2. Measuring Bureaucratic Motivations Based on Key Motives

To measure primary workplace motives, respondents were asked: “For the following questions, please indicate how important each of these is to you in a job or work environment”, with examples such as “Job security”, “Pay”, and “Opportunity for greater responsibility within my area of expertise.” Items are measured on a 5-point Likert scale ranging from (1) “Very unimportant” to (5) “Very important.”

Each item is transformed into a dichotomous variable (1 = “Very important”, 0 = all other responses). This coding serves the study’s substantive aim of identifying salient motives that respondents deem essential rather than merely desirable and improves measurement for the person-centered model given the negatively skewed distributions (Appendix A Figure A1 and Figure A2). Because these items are not trade-offs, respondents have little reason not to endorse “important”; selecting “very important” thus provides a stronger signal of priority. With mass in the top categories, finer ordinal distinctions add limited discrimination across persons at the upper end; binary indicators reduce sparse cells, align with the logit link in the LCA, and yield class-specific item–response probabilities that are straightforward to interpret (e.g., the probability that members of a class rate “Job security” as very important). While dichotomization necessarily sacrifices within–top-category variation, it improves class separation and interpretability in this setting. This practice is common in LCA with categorical survey items and is recommended when items are highly skewed (e.g., Maas et al., 2018).

To assess potential information loss, two families of sensitivity analyses are conducted. First, alternative codings (4–5 vs. 1–3; and 5 vs. 1–4 with agency fixed effects) reproduce the same number and interpretation of classes, with high class-assignment agreement and negligible changes in entropy. Second, an ordinal LCA (graded-response parameterization) and a factor-mixture variant yield the same four-class solution with near-identical class prevalences and profiles. These checks indicate that results are robust to the dichotomization choice.

Agency-level measures come from multiple sources. Organizational reputation is measured using Bellodi’s (2022) external, audience-based index constructed from four decades of U.S. congressional speeches. To provide an insider perspective, we also use the 2016 Best Places to Work (BPW) index based on the Federal Employee Viewpoint Survey, which taps perceived external/internal reputation. Additional agency controls include politicization (OPM FedScope), operationalized as the ratio of Schedule C, Noncareer-SES, and limited-term SES appointments to career SES (Lewis, 2008; Limbocker et al., 2021), decision-making and political independence (Selin, 2015), and agency ideology (Richardson et al., 2018).

3.3. Multilevel Latent Class Analysis

As the main analysis, this study applied a latent class analysis (LCA) to identify bureaucratic typologies regarding primary workplace motives. Scholars in political science are increasingly adopting person-centered approaches to depict behavioral/perceptual heterogeneity. For instance, Mack et al. (2021) categorized three types of citizens (grumpy, supporting, and indifferent) based on perceived administrative burden. Nielsen et al. (2021) identified five types of citizens’ coping behaviors when encountering street-level bureaucrats. Also, Van Parys and Struyven (2018) developed a typology of interaction styles of street-level bureaucrats. The current study selected this method because a person-centered approach, compared to the more commonly used variable-centered approach (e.g., regression, factor analyses, and structural equation modeling), is more suitable when the goal is to categorize heterogeneous groups of individuals into classes (Muthén & Muthén, 2000, p. 882).

While a fixed-effects LCA is suitable for situations where population membership can be observed directly by a group variable, a re-classification at the group level is necessary when there are many levels in the observed group (Kim & Chung, 2024). Vermunt (2003) suggested a non-parametric random effect LCA as a solution, where group-level latent variables (or latent clusters) are defined by random means from the level 1 latent class solution. This differs from the traditional fixed effects LCA approach, which assumes that parameters are identical for the whole sample. Latent clusters at level 2 have different distributions of the random means, which means that they have different probabilities of membership at level 1 (Henry & Muthén, 2010, p. 198). That is, a random-effects model enables latent class intercepts (type of bureaucrat) to vary across latent clusters (type of agency); the probability of an individual belonging to one latent class over another differs across agencies. In this way, we examined how agencies influence level 1 indicators that define one’s bureaucratic type.

We first estimated a series of fixed-effects LCAs varying the number of classes and used information criteria and bootstrap likelihood ratio comparisons to select the number of level-1 classes (Kim & Chung, 2024). We then estimated non-parametric multilevel models that allow agency-level clustering (varying the number of latent clusters) and compared these to fixed-effects specifications, again using the same fit indices and tests, alongside interpretability of class profiles. To maintain comparability of class meaning across agencies, we contrasted models with and without measurement invariance on the item-response probabilities and retained the specification that preserves a common measurement structure while allowing class prevalences to vary across agencies (Henry & Muthén, 2010; Kim & Chung, 2024). Covariates at both levels were incorporated as in the two-level specification above to relate class membership to individual (level-1) and agency (level-2) predictors. Specifically, a two-level logistic random intercept regression model can be expressed as:

$$\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{i}\mathrm{t}\left(P_{ij}\right)={\beta }_{0j}+{\beta }_1{x}_{ij} (Level 1)$$

$${\beta }_{0j}={\gamma }_0+{\gamma }_1{w}_j+U_{0j} (Level 2)$$

where $P_{ij}$ is then expressed as the following logistic function.

$$P_{ij}={\displaystyle \frac{\mathrm{e}\mathrm{x}\mathrm{p}({\gamma }_0+{\beta }_1{x}_{ij}+{\gamma }_1{w}_j+U_{0j})}{1+\mathrm{e}\mathrm{x}\mathrm{p}({\gamma }_0+{\beta }_1{x}_{ij}+{\gamma }_1{w}_j+U_{0j})}}$$

The equations show that this model incorporates both predictors at level 1 ($x_{ij}$) and level 2 ($w_j$). In the current analysis, along with the job-level motivators, we use individual demographic variables such as agency tenure, supervisory status, age, education, gender, minority status, and salary as level 1 predictors of the log-odds of belonging to one type of bureaucrat over another. For level 2 predictors, we use agency reputation, political ideology, politicization, and agency independence. A two-level multinomial logistic regression is employed since there are more than two latent classes. This allows the probability that an employee falls under a particular level 1 latent class to vary across level 2 units (Henry & Muthén, 2010). Single filled circles in Figure 1. represent level 1 random means that vary across Level 2 latent clusters; there are T-1 random means where T equals the number of level 1 latent classes (Henry & Muthén, 2010, p. 199). Level 2 (agency-level) latent clusters are defined by the random means from the Level 1 latent class solution. Analyses were conducted by fitting a set of fixed effects LCA models, followed by a series of non-parametric LCA models to compare the model fit via the glca R package (Kim & Chung, 2024).

4. Results

4.1. Model Selection

To select the suitable number of bureaucrat types (latent classes) and agency types (latent clusters), we evaluate the model fit of a set of models based on several criteria, including Akaike’s Information Criterion (AIC), the Bayesian Information Criterion (BIC), and Bozdogan’s criterion (CAIC). Additionally, the R package glca enables to conduct bootstrap likelihood-ratio tests for absolute model fit.

First, four fixed-effects LCA models that ranged from a 2-class model to a 5-class model were considered.

Table 2. Fit statistics for models at Level 1 and Level 2.

Number of Class/Clusters	Log-Likelihood	AIC	CAIC	BIC
2 Class	−106,269.29	212,600.6	212,864.3	212,833.3
3 Class	−101,858.78	203,811.5	204,211.4	204,164.4
4 Class	−99,274.85	198,675.7	199,211.7	199,148.7
5 Class	−99,220.74	198,575.5	199,145.5	199,078.5
4 Class/2 Cluster	−96,800.45	193,790.9	194,599.2	194,504.2
4 Class/3 Cluster	−99,099.88	198,463.8	199,586.9	199,454.9
4 Class/4 Cluster	−99,099.88	198,463.8	199,586.9	199,454.9

indicates that overall model fit statistics improve as the number of classes increases. Although the five-class model shows a slightly better model fit than the four-class model, model selection should be chosen to strike a balance between parsimony/interpretability and model fit (Kim & Chung, 2024). Considering the theoretical interpretability of the typology and the marginal improvement in model fit, the four-class model is selected over the five-class model.

Second, starting from the four-class model at the individual level, we assess the heterogeneity at the agency level using a multilevel LCA model that incorporates latent clusters of agencies. The model with four classes and two clusters indicates the lowest AIC, CAIC, and BIC values, implying the most parsimonious and adequate model that fits the data. We select this model as the final model.

4.2. Description of Class and Clusters: Bureaucratic Typology and Agency Types

Table 3. Item-response probabilities by latent class.

	All-Rounder	Job-Motivated	Amotivated	Self-Interested
Prevalence (overall)	35%	25%	16%	25%
Prevalence (cluster 1)	29%	33%	16%	22%
Prevalence (cluster 2)	36%	21%	16%	26%
Indicators (key motives)
Interesting work	0.95	0.92	0.38	0.68
Feeling of respect	0.96	0.80	0.18	0.67
Exercise expertise	0.98	0.92	0.13	0.55
Decision making	0.93	0.80	0.10	0.43
Contribution	1.00	0.93	0.27	0.67
Job security	0.97	0.50	0.32	0.92
Pay	0.95	0.28	0.23	0.89
Performance bonus	0.85	0.15	0.10	0.57
Benefits	0.99	0.61	0.38	0.95
Innovation	0.83	0.24	0.05	0.24
Learning opportunities	0.94	0.36	0.08	0.33
Greater responsibility	0.93	0.53	0.05	0.27
Promotion	0.78	0.42	0.07	0.29
Work-life balance	0.97	0.63	0.34	0.84
Work flexibility	0.88	0.41	0.23	0.73

and Figure 2 represents four types of bureaucrats corresponding to the four latent classes. Each class represent patterns of item response probabilities of employees’ perceptions of important workplace motives.

• Class 1: All-rounders

Members of Class 1 have high scores across all motives. With a few exceptions–the desire for promoting to supervisory/managerial roles (78 percent), performance-based bonuses (85 percent), Opportunity to use innovative technology (83 percent), and work flexibility (88 percent)—most motives score higher than 90 percent, meaning that conditional on one being categorized as a member of class 1, the probability of reporting (for example) making a positive contribution as “very important” is 100 percent. Members in this category have a strong desire for motives across intrinsic and extrinsic rewards. We label members of this class as All-rounders.

• Class 2: Job-motivated Bureaucrats

Class 2 comprises individuals who have high intrinsic motivations. The strongest motive for people in this group is making a positive contribution (93 percent), followed by doing interesting work (92 percent) and having opportunities to exercise job-related expertise (92 percent). We label these individuals as Job-motivated Bureaucrats, as they generally prioritize intrinsic aspects of work over pecuniary rewards. This is reflected in low scores on job security (50 percent), pay (28 percent), benefits (61 percent), and performance-based bonuses (15 percent).

• Class 3: Amotivated Bureaucrats

Individuals in Class 3 have relatively low scores across all motives. Even the highest-ranked motives–interesting work (38 percent), job security (32 percent), benefits (38 percent), and work-life balance (34 percent)–are far lower compared to that of members of other classes. Except for the desire to do interesting work, most motives are essential needs for anyone making a living. We label this class as Amotivated Bureaucrats. These individuals make up the most minor portion of the sample, i.e., the marginal class prevalence of 16 percent. However, it may be a startling figure considering the overall size of the federal workforce. We discuss more in the next section.

• Class 4: Self-interested Bureaucrats

Individuals in class 4 show similar patterns to Amotivated Bureaucrats in that they prioritize basic needs and monetary rewards over other motives. However, the desire for such extrinsic rewards is much stronger than Amotivated Bureaucrats. For example, the top five highest-ranked motives for these bureaucrats are job security (92 percent), pay (89 percent), and benefits (95 percent). The portfolio of motivations is similar to that of Amotivated Bureaucrats. However, the desire of each motive is much stronger. We label this class as Self-interested Bureaucrats. To sum, the derived bureaucratic typology is as Figure 3, where four types of bureaucrats are categorized based on the extrinsic/intrinsic aspects of workplace priorities.

• Agency types

Figure 4 presents two agency clusters based on the probability that employees belong to each motivational class. Roughly 30 percent of agencies fall in Cluster 1, with the remainder in Cluster 2. The clusters differ in their profile mix: All-rounders are less prevalent in Cluster 1 than Cluster 2 (29% vs. 36%), Job-motivated Bureaucrats are more prevalent in Cluster 1 (33% vs. 21%; 25% overall), Self-interested Bureaucrats are more prevalent in Cluster 2 (26% vs. 22%), and Amotivated Bureaucrats are similar across clusters (16% each).

In terms of mission environment, Cluster 1—which includes Commerce, Energy, EPA, Interior, NASA, SEC, and State—skews toward professionalized/regulatory–technical/diplomatic work, where specialized expertise and professional norms are central. Cluster 2—which includes Air Force, Agriculture, Army, Defense, Justice, Labor, Education, FDIC, GSA, Homeland Security, HUD, Navy, OPM, SSA, Transportation, Treasury, and VA—leans more operational–distributive/enforcement/defense–service delivery (see Appendix B). Consistent with these contexts, Cluster 1 contains a higher share of Job-motivated Bureaucrats and a lower share of Self-interested Bureaucrats than Cluster 2, while the Amotivated Bureaucrats is comparable across clusters. However, these interpretations should be read cautiously as descriptive patterns rather than causal claims, and the cluster labels are shorthand for broad mission emphases that individual agencies may straddle.

4.3. Individual and Agency Determinants of Bureaucratic Types

Next, subject/agency-specific covariates were incorporated to examine how those might influence the probability of the individual belonging to a specific class over another. The multinomial logistic regression is exploratory since there are no clear theoretical a priori expectations of the relationship between covariates and latent class probabilities. Individual-level covariates such as age, gender, agency tenure, supervisory status, salary, minority status, and education were included. At the agency level, measures of agency reputation (Bellodi, 2022; Partnership for Public Service, 2016), political/decision-making independence (Selin, 2015), political ideology (Richardson et al., 2018), and agency politicization were included.

Table 4. Multinomial regression results.

	All-Rounders	Self-Interested	Amotivated
Level 1 covariates
Agency tenure (1 = 3 years or less, 2 = 4 years or more)	0.26 ***	0.48	0.44 ***
	(0.03)	(0.27)	(0.02)
Supervisory status	−0.23	−0.40	−0.26
	(0.31)	(0.34)	(0.3)
Salary	−0.40 **	−0.34 ***	−0.32
	(0.12)	(0.02)	(0.18)
Minority status (0 = non-minority, 1 = minority)	1.26 ***	0.56 ***	0.26
	(0.13)	(0.11)	(0.28)
Sex (1 = male, 2 = female)	0.33	0.15	−0.64 ***
	(0.21)	(0.09)	(0.09)
Age (1 = 39 and under, 2 = 40 and over)	0.13 ***	−0.20	0.01
	(0.03)	(0.3)	(0.2)
Education (1 = Less than AA, 2 = AA, 3 = Graduate)	−0.48	−0.54	−0.31 ***
	(0.38)	(0.29)	(0.09)
Level 2 Covariates
Agency reputation (Bellodi, 2022)	−0.03 ***	−0.06 ***	−0.11 ***
	(0.00)	(0.01)	(0.03)
Agency reputation (insider-based; Partnership for Public Service, 2016)	−0.01	−0.01	−0.02
	(0.01)	(0.01)	(0.01)
Agency political ideology (Richardson et al., 2018)	0.16	0.16	0.14 **
	(0.22)	(0.09)	(0.07)
Decision making independence (Selin, 2015)	0.17	0.36 ***	0.1 ***
	(0.16)	(0.05)	(0.02)
Political independence (Selin, 2015)	0.01	0.06	0.06
	(0.08)	(0.06)	(0.2)
Agency politicization	0.00	0.01	0.00
	(0.08)	(0.11)	(0.01)

Reference group: Job-motivated Bureaucrats. Signif. codes: ** p < 0.01; *** p < 0.001.

shows the estimated coefficients compared to the baseline group, Job-motivated Bureaucrats, with the other three groups.

At the individual level, the results suggest that the divide between different types of bureaucrats is partially attributed to demographic characteristics. Agency tenure is positively associated with being an All-rounder (p < 0.001) and Amotivated Bureaucrat (p < 0.001) rather than a Job-motivated Bureaucrat; the longer one works in the agency, the more likely one has a chance to become an All-rounder or an Amotivated Bureaucrat over a Job-motivated Bureaucrat. Salary is negatively associated with membership in All-rounder (p < 0.01) and Self-interested Bureaucrat (p < 0.001) relative to Job-motivated Bureaucrats, indicating that higher-paid employees are comparatively more likely to be Job-motivated. Minority status is positively associated with being an All-rounder (p < 0.001) and Self-interested Bureaucrat (p < 0.001) relative to Job-motivated Bureaucrat. Being female is negatively associated with Amotivated Bureaucrat (p < 0.001) relative to Job-motivated Bureaucrat, and education shows a similar negative association with Amotivated Bureaucrat (p < 0.001). Age has a small positive association with All-rounders (p < 0.001), while other age contrasts are mixed. Taken together, the largest individual-level contrasts in Table 4 are for minority status (especially for All-rounders), salary (for All-rounders and Self-interested Bureaucrats), and tenure (for Amotivated).

The results of Table 4 also indicate that bureaucratic typology is a function of agency characteristics. Importantly, external agency reputation (Bellodi, 2022) is consistently related to a higher likelihood of being Job-motivated Bureaucrat: as reputation increases, the coefficients for All-rounders (p < 0.001), Self-interested Bureaucrats (p < 0.001), and Amotivated Bureaucrats (p < 0.001) relative to Job-motivated Bureaucrats are negative. This implies that Job-motivated Bureaucrats are more likely to be working in agencies with good reputations. Public employees driven by motives related to their work, e.g., mission orientations and prosocial motivations, may perceive that working in a highly reputable agency would increase their chances of meeting their needs. By contrast, the insider-based Partnership for Public Service (2016) index is not significantly related to class prevalence in this specification. Among other agency features, decision-making independence is positively associated with Self-interested (p < 0.001) and Amotivated (p < 0.001) relative to Job-motivated Bureaucrat, while agency ideology shows a modest positive association with Amotivated Bureaucrat (p < 0.01). Political independence and politicization do not show clear relationships with class membership. In sum, the regression indicates that both individual characteristics and organizational context are related to the composition of motivational profiles, with external reputation standing out as the clearest agency-level correlate of a Job-motivated workforce.

5. Discussion

Uncovering the motivational bases of bureaucrats is an enduring challenge in public administration, yet a complex and inconclusive area of inquiry. While there are studies that attempt to categorize bureaucratic behavior, most remain at the theoretical level, basing their argument on general premises or propositions. These approaches are limited in testing the degree to which types of bureaucrats are common in the general population (Nørgaard, 2018; Nowell & Albrecht, 2019). Notably, we know little of the comprehensiveness of these typologies in terms of how well they are representative of government employees. This is not to belittle the contributions of previous research. Instead, a more empirical-based approach would enable us to pick up where theory and typology building left off. In that spirit, the present study empirically recovers the patterns that prior work posits, showing which configurations actually appear in public organizations and how they distribute across agencies.

Based on such rationale, this study contributes to the broad literature on bureaucratic motivation by investigating how primary sources of workplace motivation are distributed across the federal workforce. Using a person-centered approach, this study considered the heterogeneity of workplace motivations within and between agencies. Contrary to previous studies that adopted variable-centered approaches, e.g., examining bivariate/multivariate relationships between variables, the current study aimed to accommodate a more holistic view reflecting a macro perspective of the motivational composition. Results produced largely four types of bureaucrats: All-rounders, Self-interested Bureaucrats, Job-motivated Bureaucrats, and Amotivated Bureaucrats. The inductively derived typology overall corroborates the hypothetical, deductive-driven, and in many cases, normative bureaucrats that appear in the literature (Downs, 1967; Le Grand, 2003; Perry & Wise, 1990; Esteve & Schuster, 2019; Andersen et al., 2018). Beyond corroboration, the four profiles provide an organizing map of coexisting motivational logics—mixed motives alongside autonomous, instrumental, and low-internalization patterns—and the two agency clusters show that their prevalence is not uniform but varies in ways somewhat consistent with agencies’ missions and environments, although this needs further exploration.

The results of this study validate Downs (1967) and many others’ assumptions that bureaucrats have a diverse range of job preferences, suggesting that public managers consider variations in tasks that different types of jobs entail. The task of a bureaucrat is particularly important since it is what defines their daily activities and, thus, influences their evaluation of their job (Gains & John, 2010). Relatedly, findings serve as evidence that public personnel policies should pay more attention to job design (Piatak et al., 2020) and reject universal theories (e.g., PSM leads to positive outcomes). This also aligns with previous arguments that the effect of incentive schemes varies by context and individual characteristics (A. M. Bertelli, 2006; Weibel et al., 2010). Concretely, agencies can emphasize task significance and opportunities to exercise expertise where work is professionalized to attract and retain Job-motivated Bureaucrats, use realistic job previews and role clarity where operational demands dominate to reduce Amotivated profiles, and structure transparent pay and advancement where self-interested motives are salient while avoiding crowding-out of intrinsic motives—actions that resonate with classic job design (Hackman & Oldham, 1976) and contingent-incentive insights.

Another goal of this study was to examine motivation in sync with organizational reputation. According to reputation theory in the management literature, people want to be consistent with their beliefs about what reputable organizations do (Bromley, 1993). In political science, organizational reputation is a valuable asset used to generate public support and wield delegated authority from its political masters (D. Carpenter, 2001). Using a cross-disciplinary approach to motivation and reputation, this study demonstrated that employee motivation is linked to the reputation of their agency. Job-motivated Bureaucrats are most likely to belong to highly reputable agencies, particularly compared to Amotivated Bureaucrats. However, this does not hold for perceived external reputation, which turned out to have no significant influence on an individual’s chances of belonging to one type of bureaucrat over another. This suggests the presence of a gap between how insiders (employees) and outsiders (political principals and citizens) view a public bureaucracy’s reputation. Taken together with agency-cluster patterns, these results speak to self-selection into mission-centered, professionalized settings and to socialization/identification dynamics once inside; they also caution that audience-based and insider-based indicators may capture different facets of reputation and therefore different links to motivation.

Several limits bound this study. The bureaucratic motives of the current study lie mostly at the job level. Although job design is a critical feature of employee motivation (Hackman & Oldham, 1976), other features such as values and beliefs, personality, management-level determinants (e.g., leadership), and political-level determinants would all factor into one’s workplace motivation. Models that inform theories of public administration should treat seriously of the institutional arrangements (A. Bertelli et al., 2022). To make up for such limitations, this study incorporates bureaucratic reputation as a contextual factor along with a set of political controls at the agency level. Nonetheless, more attention should be directed to building an exhaustive list of bureaucratic motivations incorporating the settings in which bureaucrats are embedded (Jilke et al., 2019). In addition, the cross-sectional nature of the data limits causal leverage: the associations we document are consistent with both self-selection and socialization, and the reputation measures we use—external audience-based speech and insider Best Places to Work—capture different lenses that need not coincide in timing or emphasis.

Also, the cross-sectional nature of the data limits further analysis of the norm-shaping socialization processes within public bureaucracies. Although past studies sought to test Perry and Wise’s (1990, p. 370) proposition that individuals with high public service motivation are likely to seek membership in public organizations (Kjeldsen & Jacobsen, 2013; Wright & Christensen, 2010), we still know little of how key workplace motives change, and how this contributes to forming overall motivation. In a similar vein, the current study is limited in firmly insisting that highly motivated bureaucrats self-select into reputable agencies since it may be the socialization process within reputable agencies that drives employees’ motivation. Such challenges would have been more addressable if longitudinal data were available, e.g., implementing latent transition analysis to track changes in group membership over time and examine whether bureaucrats change types during one’s career. Building on this, future work can exploit plausibly exogenous changes in reputation (e.g., audits, scandals, reorganizations) or test recruitment and onboarding messages experimentally to separate sorting at entry from within-agency change, and can extend the design to state/local or cross-national settings to probe scope conditions.

Lastly, as the main purpose of this article is to provide a holistic view of public sector motivation, the nature of the study is more descriptive/positive than explanatory/normative. Therefore, this study is limited in providing normative suggestions of the desirable mix of motivational types. For instance, what is the desired proportion of each type of bureaucrat for a given agency? It is apparent that an agency consisting of fewer Amotivated Bureaucrats and more All-rounders would be desirable. Then, how do we advance amotivation or self-interestedness into a desire for intrinsic aspects of the job? Would it be more beneficial to achieve the mission in public organizations? This study does not aim to address these questions, however, pursued laying another brick on the extant literature by offering a model of bureaucratic types. Future research can use the current study’s approach as a foundation to predict and test untested propositions of bureaucratic behaviors, including longitudinal mapping of transitions across types, the motivational effects of managerial practices and incentive regimes, and whether politicization or independence conditions the reputation–motivation link identified here.