Abstract
Research on destructive leadership has largely treated authoritarian leadership as a stable individual difference, rooted in personality traits. We argue instead that authoritarian leadership may take shape over time as leader respond to organizational climates that alter the perceived legitimacy of control. We use a 4-year longitudinal panel from an Eastern European service organization, and we forecast shifts in authoritarian behavior using weighted dynamic feature selection and lagged regression models. Results reveal that deteriorating justice climate and rising perceptions of organizational politics consistently predict future increases in leader control, whereas dark traits such as narcissism and Machiavellianism, while theoretically relevant, exhibit weak standalone predictive power. Crucially, narcissism predicts control escalation only under conditions of low voice climate, suggesting a conditional activation effect. By modeling authoritarian drift as a gradual, climate-contingent pattern rather than a static trait profile, this study challenges trait-dominant perspectives and reframes despotism as an emergent response to climate erosion. The findings offer a time-sensitive diagnostic framework for anticipating leadership derailment and inform HR practices aimed at preventing control intensification before it institutionalizes.
Keywords
Introduction
Leader sometimes increasingly relies on authoritarian form of control, even in the absence of formal mandates. Prior research has largely treated despotic leadership as a fixed trait or dispositional risk (De Hoogh and Den Hartog, 2008). Scholarship also suggests that such behaviors can drift into prominence under specific organizational climates (Thoroughgood et al., 2018). This study investigates the dynamic interplay between stable leader characteristics (e.g., narcissism, Machiavellianism) and shifting workplace signals, such as justice erosion, organizational politics, and muted voice climates, to shape the behavioral expression of despotism over time.
Traditionally destructive leadership have often emphasized reactive, episodic abuse (Tepper, 2000; Tepper et al., 2017), portraying subordinates as targets of leaders’ stress or insecurity. Authoritarian leadership centers on the exercise of hierarchical authority and control (Liu and Ling, 2025). Despotic leadership, by contrast, reflects a more strategic and persistent pattern of control rooted in moral disengagement and ideological dominance (Naseer et al., 2016). Yet little is known about how despotic tendencies escalate, when they become behaviorally activated, and under what conditions they are socially reinforced. Psychological theories of trait activation (Tett and Burnett, 2003) and moral disengagement (Bandura et al., 1996) suggest that stable traits become behaviorally salient only when situational cues legitimize their expression. However, unfortunately no study modeled this activation process in longitudinal, real-world leader data.
It important to study that leaders are not simply vessels of personality; they are also interpreters of organizational signals. Organizational climates, particularly those characterized by perceived injustice, eroding voice, and political maneuvering, send implicit cues about what is tolerated, rewarded, or ignored (Blickle et al., 2020; Judge et al., 2009). In such climates, leaders with only moderate levels of dark traits may nonetheless adopt more controlling and restrictive leadership behaviors as organizational cues increasingly legitimize authority and constraint. Destructive leadership is not always negative; it may be “functional” within climates that equate control with competence or outcomes with virtue (De Clercq et al., 2018).
Building on this perspective, we argue that despotic leadership is not merely selected, it is made. It emerges from a dynamic reinforcement loop between leader predispositions and organizational climates that subtly, yet systematically, reward dominance, silence dissent, and obscure procedural fairness. Specifically, we propose that authoritarian drift can be anticipated by tracing shifts in climate variables (e.g., justice erosion, political climate) and by modeling how these shifts interact with leader traits like narcissism or Machiavellianism. Our framework departs from static trait models by treating authoritarian behavior as a conditional trajectory rather than a fixed disposition.
We test these claims in a longitudinal panel study involving real-world leaders across multiple organizational units, combining dynamic machine learning techniques with contextual moderation analyses. We examine not only which factors predict future authoritarian behavior but also when and under what climates those predictors become salient. Our moderation analyses further explore how climates of employee silence (i.e., low voice climate) amplify the behavioral expression of narcissism, revealing that trait activation is not merely probabilistic, but context-contingent.
This study advances leadership research in three ways. First, we advance a temporally sensitive account of despotic leadership by modeling drift, rather than assuming presence. Second, we integrate trait activation theory with dynamic climate signaling to illuminate how dark traits become behaviorally consequential only under legitimizing environmental conditions. Third, we offer methodological innovation by applying weighted dynamic feature selection (WDF) to detect early warning signs of authoritarian escalation, an approach increasingly recommended in organizational science for tracing nonlinear behavioral emergence (Son and Pak, 2024; Thoroughgood et al., 2018). By foregrounding the climate-trait interface, we move beyond “bad apple” narratives and toward a more contextual, process-based understanding of how despotism emerges, not as a singular act, but as an accumulating trajectory legitimized by the organization itself.
Climate-induced activation of authoritarian drift
Research on destructive leadership has increasingly acknowledged that such behaviors are not merely trait-driven but contextually emergent, shaped by the climate in which leaders operate (Mehmood et al., 2023; Thoroughgood et al., 2018). Tow climates features are especially consequential for the emergence of control-oriented leadership: perceptions of organizational politics and procedural injustice. Political climates, characterized by favoritism, hidden agendas, and informal power networks, signal that influence and outcomes are detached from competence or fairness (Erkutlu and Chafra, 2018; Manara et al., 2020). Likewise, low justice climates reduce shared expectations about how authority should be exercised and evaluated, increasing leaders’ reliance on formal control and rule enforcement (Colquitt et al., 2002; Tremblay et al., 2018). These climate conditions foster what has been termed “ideological dominance” (De Hoogh and Den Hartog, 2008), where leaders justify increased control not as moral failure, but as rational adaptation to a deteriorating context. In such environments, despotic behavior becomes not just possible, but legitimized, a survival strategy cloaked in strategic necessity (Khan et al., 2022; Naseer et al., 2016).
Drawing from trait activation theory (Tett and Burnett, 2003), we propose that authoritarian tendencies, especially among leaders predisposed to dominance, are more likely to be activated when the surrounding climate signals instability, distrust, or competitive threat. Without such cues, the same traits may lie dormant or manifest in more benign forms of assertiveness (Blickle et al., 2020; Liu et al., 2017; Zhang and Liu, 2018).
In organizational climates characterized by high perceived politics and low procedural justice, leaders with dominance-prone dispositions are more likely to exhibit authoritarian behaviors over time.
Voice climate as a contextual moderator of narcissistic expression
Building on Trait Activation Theory (Tett and Burnett, 2003), we contend that narcissistic tendencies do not operate in isolation but become behaviorally expressed when organizational contexts signal weak normative constraints on authority and limited social accountability. In organizational settings, this suggests that personality traits like narcissism function less as fixed dispositions and more as probabilistic tendencies shaped by environmental constraints and expectations (Forsyth et al., 2012; Liu et al., 2012; O’Boyle et al., 2015). Although narcissism has long been associated with traits such as grandiosity, dominance, and interpersonal exploitation (Blickle et al., 2020; Judge et al., 2009), recent scholarship highlights that the expression of these traits is not automatic but depends on contextual conditions that regulate scrutiny, accountability, and acceptable authority use (Forsyth et al., 2012; Khizar et al., 2023; Liu et al., 2012; O’Boyle et al., 2015). In line with trait activation theory (Tett and Burnett, 2003), narcissistic tendencies require situational cues to be behaviorally expressed and are more likely to manifest in environments that do not constrain self-aggrandizing impulses.
One contextual factor that may significantly regulate narcissistic expression is voice climate (the shared belief that it is safe and constructive to speak up within a team or organization) (Colquitt et al., 2002; Khan et al., 2022; Tremblay et al., 2018). In high voice climates, norms of open communication and participatory leadership are reinforced, which can heighten reputational monitoring and reduce tolerance for authoritarian control (Erkutlu and Charfra, 2018; Manara et al., 2020). For narcissistic leaders, whose sense of self is often tethered to admiration and social dominance, such climates may signal risk of public scrutiny, thereby prompting strategic restraint and behavioral masking.
Conversely, low voice climates, characterized by silence, conformity, and limited upward feedback, may activate narcissistic tendencies by signaling weak normative boundaries and minimal interpersonal accountability (Liu et al., 2012; Restubog et al., 2015). In these environments, the absence of challenge can be interpreted by narcissistic leaders as implicit license to assert control, suppress dissent, and impose unilateral decisions (Mehmood et al., 2023). Thus, voice climate does not merely serve as a background condition but functions as a situational moderator that shapes whether narcissism remains latent or transforms into authoritarian enactment.
The relationship between leader narcissism and authoritarian behavior is moderated by team-level voice climate, such that narcissism is more strongly associated with authoritarian behavior in low voice climates than in high voice climates.
Trait activation and the behavioral manifestation of despotic leadership
Although a growing body of research has examined the antecedents and outcomes of despotic leadership (Khan et al., 2022; Khizar et al., 2023; Naseer et al., 2016), far less attention has been paid to its intra-individual variability, namely, why some leaders are more likely to exhibit despotic behaviors in certain contexts than in others. In line with trait activation theory (Tett and Burnett, 2003), we argue that individual predispositions such as narcissism and Machiavellianism may remain latent until triggered by cues embedded in the leader’s social or structural environment (Blickle et al., 2020; Liu et al., 2012).
Trait activation theory posits that personality traits are expressed as behaviors when relevant situational cues are present (Tett and Burnett, 2003). In the leadership context, traits such as narcissism are not uniformly expressed across settings, but are instead elicited by specific contextual affordances, such as a permissive or ambiguous moral climate (Thoroughgood et al., 2018), the absence of upward accountability (Judge et al., 2009), or normalized deviance within the organizational culture (Mehmood et al., 2023). This means that the same narcissistic individual may behave quite differently in two structurally similar organizations depending on how the local climate signals what is acceptable, rewarded, or tolerated.
Despotism, as a leadership style, is distinguished by its ideological dominance, moral corruption, and the instrumental use of fear (Khizar et al., 2023; Naseer et al., 2016). However, such behaviors may be more likely to surface when the organizational environment offers what Manara et al. (2020) calls “moral license for control,” a perceived sanctioning of rigid control, exclusionary tactics, and power-hoarding practices. For example, in cultures that prioritize hierarchical obedience and tolerate unethical political tactics, narcissistic tendencies are more likely to manifest as overt domination or emotional exploitation (Liu et al., 2017).
Furthermore, studies have shown that leaders high in Machiavellianism are particularly sensitive to the cost-benefit structure of their environments (Blickle et al., 2020). When climates convey that controlling or deceptive behaviors are not only possible but functionally adaptive (e.g., through unchallenged power asymmetries or weak checks and balances), these individuals are more likely to activate their manipulative scripts and engage in despotic conduct.
This framework helps explain the escalation of despotic tendencies over time. When unethical behaviors go unchecked, they produce a “slow corruption dynamic” where initial actions, perhaps subtle, gradually transform into sustained patterns of authoritarian control (Manara et al., 2020). From a trait activation perspective, each instance of tolerated aggression or manipulation reinforces the perceived utility of trait-congruent behaviors, making subsequent expressions of despotism more likely and more extreme.
In this way, despotic leadership should not be viewed solely as a fixed disposition or emergent response to structural pressures. Rather, it may be understood as a dynamic interaction between dispositional traits and contextual cues, described as a “trait-context congruence model of dark leadership” (Liu et al., 2012; Thurgood, 2014). Leaders with latent narcissistic or Machiavellian tendencies are not always despotic; they become despotic when environments activate those traits and reward their expression.
The emergence and intensification of despotic leadership will be amplified when organizational environments activate leaders’ narcissistic or Machiavellian traits through contextual cues that signal permissiveness, reward dominance, or fail to penalize moral violations.
Modeling the dynamic escalation of despotic leadership
This study examines how authoritarian leadership behaviors emerge and escalate over time, treating despotism not as a static trait but as a context-contingent trajectory. Drawing on a longitudinal, multi-source dataset from a large Eastern European service organization, this study traces organizational dynamics over time. We traced how organizational signals, such as perceived injustice, internal politics, and suppressed voice, interact with leader dispositions to forecast behavioral drift. Our focus is on identifying when, how, and under what organizational conditions supervisory behaviors shift toward sustained authoritarian control.
Our analytic goals are fourfold: (a) Identify early patterns that signal the onset of escalating control behavior; (b) differentiate stable leader traits (e.g., narcissism, Machiavellianism) from changing organizational climate conditions (e.g., voice climate, perceived justice); (c) examine nonlinear and interaction-based predictors of future authoritarianism; and (d) identify early warning indicators that precede into increases in authoritarian control.
In doing so, we reconceptualize leader derailment as a process that unfolds over time in response to different organizational conditions, rather than as a direct reflection of stable personality traits. This perspective situates despotic behavior within a temporal process; practically, by identifying climate signals that forecast authoritarian drift before it institutionalizes.
Study context and data structure
We analyzed a quarterly panel dataset constructed from three organizational systems: (1) the Human Resource Information System (HRIS), which tracks personnel records, quarterly performance appraisals, and formal employee complaints; (2) a psychometric onboarding platform used to assess traits such as Machiavellianism and self-confidence at the point of promotion; and (3) a 360-degree feedback system coupled with biannual team climate surveys, capturing subordinate evaluations of leadership behavior and the work environment.
All data were collected for developmental, not evaluative, purposes. The dataset was accessed under a strict nondisclosure agreement. Personal identifiers were encrypted; the research team had no contact with employees or access to identifiable unit-level information.
Each observation in our panel corresponds to a leader-quarter pairing. Psychometric trait assessments (measured once) were merged with time-varying performance reviews, HR records, and climate survey data using an encrypted supervisor ID maintained by the firm’s analytics division. Quarterly aggregation served as the temporal unit of analysis, selected to balance signal stability with sensitivity to intra-year variation.
To ensure temporal coherence, we verified merge accuracy through timestamp checks, alignment with survey administration logs, and cross-referencing complaint submission windows with feedback cycles. Observations with timestamp mismatches or missing values were excluded, ensuring a clean and temporally ordered structure for forecasting analyses.
Analytical strategy: Modeling escalation trajectories
Analytical focus
Our central analytic objective was not merely to classify who displays despotic leadership, but to model how and when such behavior escalates over time. We conceptualized despotism as a trajectory, an emergent within-person pattern that unfolds under contextual pressure rather than a stable disposition. The outcome variable was each supervisor’s authoritarianism score at quarter t+1, derived from 360-degree feedback combining subordinate, peer, and senior evaluations.
To preserve temporal integrity, all predictors, including leader traits, team climate scores, and HR incidents, were lagged by one quarter (i.e., time t) to test whether conditions in a given quarter forecasted escalation in the subsequent period. This design allowed us to model drift rather than status, identifying early conditions that precede behavioral intensification.
Standard regression models assume fixed effects and linear influence, which may obscure inflection points or nonlinear accelerations in behavior. To overcome these limitations, we employed WDF, a permutation-based approach optimized for temporal forecasting. WDF ranks predictors based on their marginal contribution to forecast accuracy, does not assume linearity, and accommodates complex interactions without requiring functional form assumptions. Its flexibility makes it especially suitable for identifying early signals in dynamic organizational settings.
For each observation in our panel, we constructed a feature matrix of lagged variables and trained time-aware random forest models. Recursive feature elimination was applied to remove low-importance predictors, and feature weights were calculated based on the average mean decrease in impurity across 50 bootstrapped iterations. To prevent overfitting, we used 10-fold time-series cross-validation, maintaining the temporal structure of the data and avoiding information leakage across folds. Model parameters were tuned using grid search within each fold, and all scripts were executed in Python using the tsfresh, scikit-learn, and statsmodels libraries.
To probe conditional dependencies, we supplemented WDF with interaction analyses between leader traits and team-level moderators. For instance, we tested whether the effect of narcissism on future authoritarianism was amplified under conditions of low voice climate. These interactions were modeled using conditional mutual information techniques, enabling us to examine the contextual activation of dark traits.
Forecasting results revealed that contextual variables, particularly perceived politics and justice climate, were the most robust predictors of escalation. Dispositional predictors, including narcissism and Machiavellianism, exhibited consistently weaker signal strength. This pattern reinforces our theoretical position: authoritarian drift is not merely a personality defect, but a context-sensitive outcome shaped by deteriorating normative environments.
Variable operationalization and measurement sources
Organizational scale–academic scale mapping.
Note. Exact item wordings were adapted by the organization to ensure contextual relevance and legitimate neutrality. Mapping was verified by three organizational psychology scholars using a blind matching protocol. Internal consistency values for all multi-item scales exceeded .75.
Measures
All variables were sourced from institutional records and standardized survey instruments embedded within the organization’s HRIS and analytics platforms. Leader-level constructs included Perceived Narcissism, Machiavellianism, Self-Confidence, 360° Authoritarian Scores, Complaints Received, and Performance Ratings. Perceived Narcissism was derived from 360° feedback cycles and reflected subordinate evaluations of leaders’ grandiosity and need for admiration (e.g., “Seeks admiration”). Machiavellianism and Self-Confidence were collected via a proprietary onboarding diagnostic, indexing manipulative tendencies (e.g., “It’s wise to keep others in the dark about plans”) and general efficacy beliefs, respectively. These were treated as stable, time-invariant traits. Authoritarian Scores were computed from composite 360° ratings (subordinates, peers, supervisors) covering unilateral decision-making and intolerance for dissent. Complaints Received captured the count of HR-reported incidents tied to each leader per quarter. Performance Ratings were drawn from formal quarterly review cycles and served as a benchmark for distinguishing authoritarianism from legitimate high performance.
Team-level (employee-aggregated) constructs included Justice Climate, Voice Climate, Perceived Politics, Turnover Intentions, and Absenteeism Days. Justice Climate and Voice Climate were based on semi-annual surveys and reflected aggregated perceptions of procedural fairness and psychological safety for voice (e.g., “It is safe to speak up with concerns”), modeled after Colquitt et al. (2001) and Detert and Burris (2007). Perceived Politics was assessed through the firm’s cultural audit tool, using items conceptually aligned with the POPS scale (Kacmar and Carlson, 1997), though no external validation was performed. Turnover Intentions (e.g., “I am actively looking for another job”) and Absenteeism Days were extracted from HRIS logs and engagement survey records and reflect employee behavior under each leader’s supervision.
All constructs were linked to a unique, encrypted Leader_ID across quarterly (Q1–Q4) observations, enabling a multilevel and longitudinal design. Internal audit records confirmed that all multi-item constructs exhibited satisfactory internal consistency (α ≥ .75). No measures were created post hoc; all were sourced from preexisting institutional tools and protocols.
So this analysis seeks to uncover not static levels of authoritarianism, but the trajectory by which supervisory behavior shifts over time toward increasingly despotic patterns. Our focus is not on identifying high authoritarian scores in isolation, but on detecting escalation, the gradual, compounding drift in behavior that signals leadership derailment in real-time. By modeling behavioral changes quarter-over-quarter, we aim to surface early, temporally ordered signals that precede authoritarian inflection points.
Rationale for forecast-oriented modeling
Each observation in our dataset represents a unique pairing between a focal leader and a quarterly time point. To preserve causal inference and minimize simultaneity bias, all predictors were lagged by one quarter (t–1), ensuring temporal precedence over the criterion variable, namely, the subordinate-rated authoritarian score at time t, captured via 360-degree feedback systems.
The modeling process began with data harmonization. Inputs from performance management systems, personality audits, and climate surveys were first temporally aligned, cleaned for inconsistencies, and merged into a unified panel dataset. Observations with missing leadership identifiers or overlapping reporting structures were excluded to ensure internal consistency and avoid inflating variance estimates.
We then constructed a lag structure that shifted all predictors back by one quarter. This approach allowed us to assess how prior-quarter conditions forecast shifts in leader behavior. For example, narcissism, justice climate, turnover intention, and team instability, all originally collected at time t–1, were used to predict despotic leadership manifestations at time t.
Next, we employed a permutation-based WDF model to quantify the relative forecasting value of each predictor. WDF assigned importance scores to each t–1 variable based on its contribution to improving predictive accuracy for t+1 authoritarian scores. These scores were computed across 1000 reshuffled samples to mitigate bias from data noise and to produce stable feature rankings.
To further investigate interaction effects, particularly under adverse climate conditions, we conducted CMI analyses. This allowed us to evaluate whether specific combinations of traits and contextual variables, such as narcissism under low voice climate, exhibited conditional dependence in forecasting leadership drift. These analyses helped isolate predictors that were not only strong in isolation but also contingent on hostile or enabling contexts.
Robustness checks were implemented through holdout validation. The dataset was partitioned into temporally staggered training and test sets to examine generalizability across quarters. Additional sensitivity analyses were conducted on stratified subsamples defined by unit size, tenure composition, and change rate in team membership. These steps ensured that our findings were not artifacts of overfitting or isolated subsample idiosyncrasies.
Forecasting target and interpretive focus
The outcome of interest was the 360-degree authoritarian score at quarter t+1. Unlike classification tasks focused on thresholds or categories, our emphasis was on trajectory detection, whether authoritarianism scores increased meaningfully from one quarter to the next. We interpreted higher WDF importance scores as evidence that a given lagged variable served as a reliable early indicator of behavioral escalation.
Consistent with this orientation, our modeling results (reported in Table 4) elevate team climate variables, particularly perceived politics and justice climate, as the most temporally predictive features. Trait-based factors, by contrast, consistently failed to forecast future authoritarian increases. These findings reinforce a shift in theoretical emphasis: despotism may be conditioned into existence, not merely expressed from stable traits.
Results
Means and standard deviations.
Perceived Politics (M = 2.88, SD = 1.21) also displays substantial dispersion, reflecting divergent experiences of favoritism, manipulation, or opacity across units. Importantly, 360° Authoritarian Scores (M = 3.05, SD = 1.16) reveal a relatively broad range of perceptions regarding leader control behaviors. Employees’ turnover Intentions (M = 2.82, SD = 1.01) and Absenteeism Days (M = 4.38, SD = 2.92) indicate modest but nontrivial behavioral withdrawal from the organization.
Temporal predictors of authoritarian escalation
Rationale for temporal modeling
In response to concerns regarding oversimplified lag structures, we expanded our temporal modeling strategy beyond single-lag (t–1) designs. We implemented a dynamic forecasting framework using WDF and CMI to capture nonlinear, interaction-driven patterns that traditional linear panel regressions overlook. Crucially, WDF prioritizes predictors based on their temporal contribution to behavioral change, rather than mere explained variance at a single point in time.
Model tuning
To prevent overfitting, WDF models were implemented with nested cross-validation, and all hyperparameters (e.g., window size, decay rate) were optimized via grid search. Predictor importance scores were bootstrapped across 100 iterations to ensure stability. In parallel, CMI was used to probe moderation effects, including hypothesized trait-by-climate interactions (e.g., narcissism × low voice climate), enabling more nuanced interpretability of conditional risk factors.
Forecasting accuracy and reliability
Model fit metrics.
Outcome variable
The primary dependent variable was the 360-degree authoritarian score at t
Key contributions of the forecasting approach
Our study modeled authoritarian drift as a temporally unfolding process rather than a static trait display. By identifying nonlinear predictors such as rising perceived politics and declining justice climate, we demonstrate that shifts in control behavior are not random but forecastable. We also find that the behavioral expression of narcissism intensifies under structurally silent conditions, such as low voice climates, underscoring the conditional nature of trait activation. Rather than framing despotism as an episodic or reactive phenomenon, we conceptualize it as cumulative degradation, an accretive pattern reinforced by climate cues.
Although limited by quarterly resolution, our approach offers a theory-consistent and replicable method for anticipating authoritarian escalation. These findings hold practical implications for leadership risk monitoring, suggesting that early climate signals, especially those indicating political turbulence or muted dissent, can serve as actionable warning indicators.
Cultural framing and generalizability
Our study was conducted in a large Eastern European service organization where hierarchical structures are institutionalized. While this enhances internal validity for modeling top-down control dynamics, it may limit cross-cultural generalizability, particularly in flatter or participative organizational cultures. We further acknowledge the limitations of subordinate-rated trait indicators (e.g., perceived narcissism), which may conflate observed behavior with inference, particularly under ambiguous or strained team climates.
Lag structure and model validity
To further strengthen the temporal validity and interpretability of our model, we conducted a structured series of robustness checks. The revised analysis includes five major enhancements, the most central of which is the implementation of flexible lag windows.
Lag window comparison
Beyond the original t–1 structure, we tested t–1, t–2, and t–3 lags using both stepwise and multivariate regressions. Results revealed: Perceived Politics remained a significant and stable predictor across both t–1 and t–2, suggesting sustained sensitivity to political climates. Justice Climate peaked in predictive strength at t–1 but decayed by t–3, indicating a more immediate temporal influence. Trait-based predictors (e.g., narcissism, Machiavellianism) remained non-significant across all lags, reinforcing WDF findings that traits, in isolation, carry limited temporal lift.
These findings suggest that despotic drift is more tightly coupled to recent contextual signals than to distant or static factors. While distal predictors retain some informational value, their effects appear diluted over time, highlighting the importance of proximal environmental cues, particularly organizational opacity and justice erosion, in shaping authoritarian trajectories.
Model fit metrics
We report the predictive accuracy of both standard linear regression (in Table 3) and a WDF-based random forest model using three standard metrics: RMSE, MAE, and R2.
The inclusion of nonlinear feature weighting yielded meaningful gains in prediction accuracy. Lower RMSE and MAE values indicate improved model fit, while the increase in R2 reflects enhanced explanatory power through interaction effects and nonlinear dependencies.
Weighted dynamic feature selection
To isolate predictors of despotic leadership, we applied WDF modeling. Predictors were lagged across t–1 to t–3 and entered into a time-aware random forest. Recursive feature elimination (RFE) discarded weak signals, and variable importance was assessed via mean decrease in impurity, averaged over 50 bootstrap samples. This approach balances forecast accuracy and interpretability, capturing how early cues, rather than concurrent conditions, shape future authoritarian drift. Partial dependence plots and cross-validation confirmed model stability and minimized overfitting risk.
Intra-class correlation
To ensure that quarterly aggregation was methodologically appropriate, we computed intra-class correlation coefficients (ICCs) for repeated constructs: Voice Climate: ICC = .73, Justice Climate: ICC = .68, Authoritarian Score: ICC = .79.
These values indicate strong within-leader temporal consistency, justifying the use of quarter-level units. A multilevel random-intercepts model further showed that over 70% of the variance in Authoritarian Scores resided within leaders across time, supporting our focus on within-person behavioral drift rather than static between-person differences.
Analysis update: Trait activation and contextual moderation
Moderation analysis: Narcissism × voice climate on authoritarian behavior
To test whether the effect of leader narcissism on subsequent authoritarian behavior (t+1) is moderated by team-level voice climate at time t.
In our Analysis, DV was Authoritarian Score (t+1), IV was Narcissism (t), Moderator was Voice Climate (t), and we used Interaction of Narcissism × Voice Climate. Control variables were Machiavellianism, Self-Confidence, and Performance Rating.
Regression results
Regression results.
Predicting authoritarian drift: Results from dynamic forecasting
Predicting authoritarian drift.
Authoritarian drift.
Trait-based predictors offered weaker predictive utility. Narcissism (M = 0.114) and Machiavellianism (M = 0.076) registered some signal, yet their effects were secondary to environmental cues. This pattern challenges dark trait determinism and reinforces a conditional activation view, where traits require conducive climates to materialize behaviorally. Notably, performance ratings ranked fourth in importance (M = 0.093), suggesting that high-performing leaders may receive implicit behavioral latitude, consistent with a performance–protection paradox.
Static and temporal forecasting salience.
Feature importance.
Robustness check: Multilevel model
To test whether our key predictors retained explanatory power in a conservative model, we estimated a multilevel lagged regression model with leader-level random intercepts. Findings reaffirmed that: (a) Perceived Politics (t–1) and Justice Climate (t–1) remained statistically significant predictors (p < .05), (b) trait predictors again failed to reach significance, and (c) no evidence emerged of reverse causality from performance or complaints predicting climate indicators. A three-quarter lag window was chosen for both theoretical and organizational reasons. The firm operates on a quarterly review cycle, aligning naturally with leadership behavior and climate shifts. Evidence suggests that control-oriented behavior typically escalates within two to three quarters when triggered by environmental stressors, but tends to plateau or reverse thereafter (Kacmar et al., 2013; Tepper et al., 2017). Preliminary diagnostics confirmed that predictor variance and signal strength declined beyond t–3, supporting this as a meaningful upper bound.
Multilevel model.
Trait variables, narcissism, Machiavellianism, and self-confidence, remained non-significant across all lags, supporting the view that traits lack forecasting power without climate triggers. To validate these patterns, we used WDF. WDF ranks lagged predictors based on forecasting accuracy loss when each variable is permuted. We applied recursive feature elimination (RFE) with 50-fold stratified bootstrapping, resampling on supervisor-quarter units while maintaining the proportion of high and low authoritarian ratings within each fold. This approach ensured that the training and validation sets reflected the underlying behavioral distribution. Hyperparameters were tuned via grid search, and feature importance scores represent the averaged predictive contribution across all iterations (Table 5). It identifies early predictors of drift. Across all runs, climate signals, especially perceived politics, consistently outperformed trait-based variables. Authoritarian drift, in this data, is best forecasted by shifting climates, not stable dispositions.
Robustness check using classical linear regression
To validate the robustness of our forecasting results, we estimated an OLS model using concurrently measured predictors. This provides a static benchmark to complement the temporal precision of our WDF and lagged regression models.
Lagged effects.
These results reinforce a key asymmetry: climate features retain predictive salience even in static models, while traits require temporal context to influence control behavior. This analysis serves not as a substitute but as a boundary test, demonstrating that climate remains consequential across models, but temporal approaches are more sensitive to inflection points in authoritarian drift.
OLS regression testing concurrent predictors.
OLS regression.
Discussion
OLS robustness model testing.
COR-based coping tactics under despotic and abusive leaders.
In line with prior scholarship on organizational politics as a stress-inducing and power-concentrating force (Kacmar and Carlson, 1997; Naseer et al., 2016), perceived politics emerged as the most potent predictor of escalating authoritarianism. In environments depicted by favoritism and hidden agendas, leaders may respond with micro-control as a self-protective strategy (Blickle et al., 2020). Psychologically, political conditions reduce clarity about influence and standing, which may prompt greater reliance on centralized decision-making (Pandey & Wright 2006). Likewise, erosion of justice climate, especially procedural and relational justice, was associated with rising control behaviors, consistent with meta-analytic findings linking injustice to diminished prosociality and increased leader defensiveness (Colquitt et al., 2001; Colquitt et al., 2002; Tremblay et al., 2018). When justice climate erodes, leaders may view the social order as less stable and respond be emphasizing conformity and control (Osborne et al., 2023).
In contrast, personality traits often implicated in toxic leadership-narcissism and Machiavellianism, showed only modest predictive weight across models. While these traits have long been associated with dominant behavior (Christie and Geis, 1970; Ames et al., 2006; O'Boyle et al., 2015), our results suggest their effects may be conditional rather than assessor. This aligns with prior work arguing that dark traits require situational activation to translate into behavioral control (Liu et al., 2017; Manara et al., 2020; Tett and Burnett, 2003). Despotism, then, is less a function of fixed internal dispositions and more the product of leader–context interactions that unfold over time (Schyns and Schilling, 2013; Thoroughgood et al., 2018).
A second insight emerges from the moderate signal strength of performance ratings and turnover intentions. This points to a performance–protection paradox: high-performing leaders may be granted discretionary space to intensify control behaviors, particularly in metric-driven systems where outcomes overshadow process (De Hoogh and Den Hartog, 2008; Khan et al., 2022). Such protection can reduce external constraints on leadership behavior, allowing authority-based practices to persist and intensify with limited scrutiny. Conversely, subordinate exit intentions may reflect silent resistance to oppressive leadership climates, a finding that echoes past research linking authoritarian climates to reduced voice and heightened withdrawal (Detert and Burris, 2007; Tepper, 2000).
These findings clarify the temporal architecture of despotic leadership. Authoritarian behavior does not erupt in abrupt episodes; rather, it develops gradually through repeated exposure to unfairness, silence, and distrust (Rasool et al., 2018; Tepper et al., 2017). By integrating both time-sensitive predictors and conventional robustness checks, our results move beyond the “bad apple” narrative and instead highlight how authoritarian leadership takes root in organizational climates (Khizar et al., 2023; Mehmood et al., 2023).
Contributions to literature
Our study contributes to the literature on destructive leadership (Schyns and Schilling, 2013; Tepper, 2000; Thoroughgood et al., 2018) and the dynamic emergence of leader behavior by reframing despotic leadership not as a fixed personality configuration but as a temporally evolving outcome of leader–context interplay. While prior research has predominantly examined despotism through a trait-centric lens (Judge et al., 2009; O’Boyle et al., 2015), our dynamic forecasting approach illuminates how contextual cues, particularly perceived organizational politics and justice climate, are stronger and more temporally persistent predictors of authoritarian escalation than dispositional factors such as narcissism or Machiavellianism. This shift encourages a move beyond static trait-pathology models and toward frameworks that track how despotic tendencies accumulate incrementally over time in response to shifting organizational climates.
Second, our findings extend context-activation theories of the dark triad (Manara et al., 2020; Tett and Burnett, 2003) by empirically demonstrating that narcissistic and Machiavellian tendencies fail to predict authoritarian drift unless coupled with climate-level triggers. These results challenge the sufficiency of trait-based explanations and support a situated activation model, in which personality dispositions only gain behavioral traction in permissive or volatile environments. By applying a longitudinal lens to this interactionist logic, we bring temporal precision to what has largely been a cross-sectional conversation in dark leadership research (De Hoogh and Den Hartog, 2008; Liu et al., 2017).
Third, we add to the organizational climate literature by identifying perceived politics and justice climate not merely as correlates of workplace strain (Colquitt et al., 2001; Kacmar and Carlson, 1997) but as causal, compounding forces in the incubation of authoritarianism. These climates appear to legitimize control, suppress dissent, and erode trust, thus functioning as behavioral accelerants for leaders predisposed, or pressured, to assert dominance. Our evidence supports calls to view climate as an active shaper of leadership conduct (Naseer et al., 2016; Tremblay et al., 2018), and provides a robust empirical basis for conceptualizing despotic behavior as a climate-contingent drift, rather than a fixed leadership style.
Finally, our study contributes to dynamic leadership scholarship by elucidating the micro-patterns through which authoritarian behavior crystallizes. Unlike prior work focused on episodic or crisis-induced toxicity (Khan et al., 2022; Tepper et al., 2017), we model how despotism accumulates gradually via recursive transactions between leader conduct, subordinate withdrawal (e.g., turnover intentions), and system neglect (Detert and Burris, 2007; Rasool et al., 2018). This longitudinal process model broadens current understanding of how organizational systems fail to detect or interrupt authoritarian drift. In doing so, we respond to recent calls for leadership research that integrates temporal emergence, climate dynamics, and behavioral interdependence across levels (Khizar et al., 2023; Mehmood et al., 2023).
Implications for coping mechanisms
This study refines the conceptual distinction between despotic leadership and abusive supervision, arguing that the psychological demands imposed by each require different coping architectures. Despotic leadership is not merely abusive; it is ideological, institutionalized, and structurally embedded.
Despotic leadership is rooted in strategic power consolidation, anchored in a belief system that subordinates ethical constraints to personal or ideological dominance (De Hoogh and Den Hartog, 2008). Despots tend to score high on narcissism and Machiavellianism, but unlike episodic aggressors, they employ manipulation and control proactively and with foresight (Thoroughgood et al., 2018). Followers are not merely victims; they are cast as instruments in the leader’s agenda. By contrast, abusive supervision (Tepper, 2000) reflects reactive emotional dysregulation, a breakdown in impulse control under stress. These behaviors are often volatile, inconsistent, and context-dependent. Abusive leadership is an emotionally charged response to frustration, not a calculated strategy of domination (Tepper et al., 2017).
Despotic leadership saturates the organizational climate: it breeds fear, institutional silence, and chronic moral injury (Rasool et al., 2018; Schyns and Schilling, 2013). It is rarely experienced as interpersonal mistreatment; rather, it is encountered as an ambient regime. In contrast, abusive supervision is more often constrained to the leader–subordinate dyad and moderated by HR safeguards or peer alliances. Critically, employee attributions differ. While abusive acts may be rationalized as personality flaws or temporary lapses, despotic leadership is perceived as deliberate and systemic, eroding psychological distance and increasing moral fatigue.
Why coping mechanisms are not just useful—but essential: Under despotic leadership
This study advances a critical reorientation in the way scholars conceptualize employee coping under oppressive leadership. Specifically, we argue that status striving, communion striving, and autonomy striving are not discretionary adaptations but functional necessities under despotic regimes. Drawing from COR theory (Hobfoll, 1989), we contend that despotic leadership depletes fundamental psychological resources, status, belonging, and control, triggering motivational architectures aimed at preserving identity, self-worth, and viability.
Unlike episodic mistreatment, despotic leadership manifests as institutionalized moral corruption, structurally reinforced, ideologically coherent, and strategically manipulative (De Hoogh and Den Hartog, 2008; Son and Pak, 2024). Employees under such regimes are not simply recipients of aggression but are embedded in a climate of chronic psychological erosion (Manara et al., 2020; Rasool et al., 2018). The system itself rewards compliance and penalizes dissent, dissolving moral boundaries and warping organizational values (Schyns and Schilling, 2013; Vickers, 2014). This saturation makes passive endurance untenable, demanding a shift toward active, goal-directed responses.
Status striving emerges as a rational safeguard in hierarchies where formal advancement is decoupled from merit. Rather than signaling narcissistic ambition, it becomes a preemptive defense strategy, an attempt to regain influence that shields against unchecked domination (Kim and Pettit, 2019; Manara et al., 2020; Zeigler-Hill et al., 2019). COR theory conceptualizes this as proactive resource accumulation in anticipation of systemic threat (Foulk et al., 2019).
Communion striving reflects employees’ effort to reconstruct relational bonds ruptured by fear-based climates. Despotism erodes interpersonal trust and fragments social cohesion, necessitating affiliation-oriented coping to restore affective connection (Abele and Wojciszke, 2007; Kil and Grusec, 2022; Varma et al., 2016). Communion, in this context, is not sentimentality, it is resilience through solidarity.
Autonomy striving becomes critical in response to surveillance, micro-regulation, and ideological rigidity. By proactively reshaping role boundaries and crafting discretionary zones of influence, employees restore a sense of agency (Rousseau, 2005; Zhang et al., 2021). This aligns with COR’s assertion that control is a foundational psychological resource, and its depletion provokes compensatory goal-striving (Spector et al., 2002).
While both despotic leadership and abusive supervision are corrosive, they activate distinct motivational architectures. Abusive supervision, characterized by reactive emotional volatility (Tepper et al., 2017), tends to elicit short-term relational repair and localized boundary assertion (May et al., 2015). Despotic leadership, by contrast, constitutes a stable power structure that demands more strategic, future-oriented adaptations (Spain et al., 2014; Zettler et al., 2015).
Notably, status striving is disproportionately activated under despotic leadership, not merely to gain esteem, but to construct a shield of indispensability. Employees intuit that visibility and symbolic capital may be their only insulation from arbitrary control (Anderson et al., 2012; Vedel and Thomsen, 2017). Such striving is not ego-driven but structurally induced. As our predictive modeling shows, despotic drift is shaped more by contextual features than dispositional traits. This substantiates our theoretical claim: under despotism, coping is not passive endurance, but an orchestrated motivational response. By linking status, communion, and autonomy striving to distinct COR domains, we extend the theory’s relevance from acute stress episodes to institutionalized oppression. Employees are not merely surviving difficult leaders; they are resisting systems of domination.
This reframing disrupts pathologizing views of coping. It positions employees not as fragile actors but as psychologically agile agents who deploy motivational strivings to navigate hostile structures. In climates where leadership is unaccountable and moral boundaries are suspended; such strivings are not optional; they are essential.
Practical implications
Our findings offer several important implications for organizational practice, particularly in environments where control-based leadership styles may go unchecked. First, by demonstrating that authoritarian drift is forecastable using real-time climate signals, this study equips organizations with early diagnostic tools. Specifically, the predictive strength of perceived politics and justice climate suggests that tracking changes in employees’ perceptions of fairness, transparency, and voice over time can function as an early warning system, allowing HR teams and senior leaders to intervene before control tendencies escalate into full blown despotism. This is especially relevant in high-stakes, high-discretion roles where formal oversight is limited, and relational climates serve as de facto behavioral regulators.
Second, the data point to a performance–protection paradox: leaders who deliver strong performance ratings are more likely to exhibit rising control behaviors. Organizations may inadvertently reward or tolerate such behaviors, mistaking dominance for competence. This may happen particularly when strong outcomes coincide with declining voice, rising complaints, or worsening justice perceptions. These results call into question performance appraisal systems that emphasize outcomes over methods and highlight the need for 360-degree review processes that capture relational fallout and climate perceptions alongside objective performance metrics.
Third, the finding that turnover intentions predict rising despotism introduces a new lens on employee withdrawal. Exit intentions may not merely reflect dissatisfaction, but could serve as silent protests against tightening control, especially in environments where direct confrontation feels unsafe. Rather than interpreting such signals as individual-level disengagement, leaders and practitioners should consider them as potential systemic feedback loops when increases in exit intentions coincide with declining voice and justice perceptions. Interventions aimed at improving voice mechanisms and restorative justice practices may be especially useful in buffering against further control escalation.
Finally, our results offer a more nuanced view of leadership development and derailment prevention. Traditional high-potential assessments often screen for trait liabilities (e.g., narcissism). Yet our this study found that authoritarian escalation is more strongly shaped by leaders ongoing interaction with their organization climates. Leadership pipelines should therefore emphasize contextual self-awareness, climate sensemaking, and feedback receptivity as core competencies, especially for leaders operating in politically volatile or procedurally unjust settings. So this research suggests that preventing authoritarian leadership is not merely a matter of selecting the right leaders but system level monitoring challenge, require organizations to attend to how everyday control practices are reinforced by the climates they allow to persist.
Limitations and future research directions
While this study offers a temporally sensitive model of authoritarian drift, several limitations warrant acknowledgment. These boundaries do not diminish the study’s contribution but rather signal productive terrain for future theorizing and empirical refinement.
First, the forecasting design, though rigorous in its temporal ordering and predictive logic, does not establish causal mechanisms. The WDF framework assigns importance based on forecasting utility, not directional influence or mediation. Thus, while predictors such as perceived politics and justice climate exhibit strong temporal salience, the model does not adjudicate whether their effects operate through cognitive, affective, or behavioral intermediaries. Future research could complement this diagnostic approach with mechanism-sensitive designs, including experience sampling or longitudinal to capture how these climate signals are internalized and enacted.
Second, our reliance on quarterly aggregation, while methodologically justified via intra-class correlations and organizational cadence, may mask finer-grained inflection points. Escalation trajectories likely unfold through micro-events and interactional ruptures that occur within weeks or even days. Emerging work in leadership micro-dynamics (e.g., McClean et al., 2019) suggests that behavioral drift may be punctuated rather than smoothly. High-frequency panel designs or digital trace data (e.g., communication logs) may enable future scholars to model these micro-escalation moments more precisely.
Third, although our sample spans multiple units and quarters within a large Eastern European firm, the cultural and structural context, marked by hierarchical norms and centralized authority, may constrain generalizability. In flatter or more participative organizations, the thresholds for authoritarian behavior may differ, and the predictive weight of climate cues may be attenuated. Cross-cultural replications and comparative institutional designs could clarify how organizational form and national culture modulate the activation of authoritarian drift.
Fourth, the psychometric trait measures, while collected prior to leadership emergence, were assessed through self-report at onboarding and not updated over time. While this temporal distancing helps isolate context-induced drift, it also precludes modeling potential trait malleability or dynamic self-perception. Future research could explore how leaders’ self-concept evolves in parallel with climate conditions, potentially forming feedback loops between environmental erosion and leader self-legitimation.
Fifth, the study focused on predicting escalation but not containment. Research could examine whether favorable organizational climates not only constrain authoritarian escalation but also foster constructive leadership trajectories (e.g., Riaz, 2024). While we identify antecedents of drift, we do not assess reversal mechanisms, which allow a leader on an authoritarian trajectory to recalibrate. Organizational interventions, upward feedback loops, or restorative justice climates may play corrective roles worth theorizing and testing. Understanding inflection in both directions, toward and away from despotism, remains a critical yet underexplored dimension of leadership dynamics. So, this study models authoritarian leadership not as a fixed trait but as an emergent trajectory shaped by climate degradation. Future research can build on this work by unpacking the psychological, relational, and structural processes through which early warning signals translate into sustained behavioral patterns, and by identifying the conditions under which those trajectories can be redirected before they ossify into regime-like control.
Footnotes
Ethical considerations
The study was approved by the departmental Ethics Approval Committee at the University of International Business and Economics (UIBE). The Business School of UIBE Research Ethics Board reviewed the “Forecasting Authoritarian Drift: How Climate Signals Trigger Escalating Leader Control” research proposal and considers the procedures, as described by the applicant, to conform to the University’s ethical standards and university guidelines. All data were collected from organizational records and structured evaluations conducted as part of the firm’s internal HR monitoring processes. No personal identifiers were accessed by the research team. Analyses were conducted on anonymized, aggregated data, and the study protocol complied with institutional and organizational ethical standards for secondary data use.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research was supported by the National Social Science Foundation of China, (Grant No. 23BGL142).
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.