Self-assessment is unreliable as a standalone measure of leadership effectiveness, and the gap between self-rating and observer rating is largest in the domains that matter most. Understanding what produces the gap, and why it is widest among some leaders, is essential for designing assessment systems that actually inform development.
The self-other rating gap, the consistent tendency for individuals to rate their own performance higher than those who observe them rate it, is one of the most replicated findings in the leadership assessment literature. Atwater and Yammarino (1997) found that the gap between self and subordinate ratings was not random but systematically patterned: over-raters showed distinctly different developmental trajectories from accurate self-assessors and under-raters. Harris and Schaubroeck (1988) found in a meta-analysis that self-supervisor agreement was lower than self-peer agreement, and that both fell well below the convergent validity that would justify treating self-assessment as a substitute for observer assessment in high-stakes development or selection contexts. This article reviews the mechanisms producing the self-other gap, examines the individual and situational characteristics that widen it, addresses its implications for development program design, and considers the assessment design features that most effectively harness self-assessment as a useful rather than misleading component of leadership diagnostics.
The self-serving attribution bias, first systematically documented by Miller and Ross (1975), describes the tendency for individuals to attribute successes to their own stable qualities and failures to external circumstances, situational factors, or other parties. In the leadership assessment context, this attribution pattern produces inflated self-assessment of effectiveness: positive team outcomes are attributed to the leader's own capability and decisions, while negative outcomes are attributed to market conditions, team member limitations, or organizational constraints beyond the leader's control. Observers, and particularly subordinates, who have more direct and consistent observation of the leader's specific behaviors tend to attribute outcomes more accurately, connecting specific leader behaviors to specific performance effects in ways that the leader's own attribution processing systematically discounts.
Dunning and Kruger (1999) documented a specific self-assessment failure mode with direct implications for leadership development: individuals with genuine skill deficits in a domain are also, by the nature of those deficits, unable to accurately recognize the deficits, because the same competence that would enable excellent performance would also enable accurate recognition of the gap between one's own performance and excellent performance. In the leadership domain, a leader lacking the interpersonal awareness to recognize how their behavior affects others is also lacking the interpersonal awareness to recognize that their interpersonal behavior is producing negative effects. The self-assessment accuracy problem is therefore most severe in precisely the leaders most in need of accurate feedback for development, creating a systematic inverse relationship between need for accurate self-assessment and ability to provide it.
The social desirability dimension of self-assessment inflation is partially distinct from the attribution and metacognitive mechanisms. Leaders who complete self-assessments in organizationally visible contexts, including 360-degree feedback programs, performance management processes, or development program intake assessments, are aware that their self-assessments may be seen by others and may consciously or unconsciously shape those assessments to project the image of competence that organizational performance requires. The inflated self-assessment in visible contexts is therefore partly motivated by impression management, the desire to present oneself favorably to those who will see the assessment, rather than entirely by the unconscious attribution and metacognitive mechanisms that produce inflation even in genuinely private and anonymous contexts.
Atwater and Yammarino (1997) classified leaders into four categories based on their pattern of self-other rating correspondence: in-agreement/good performers, whose self-ratings aligned with observers and were both relatively high; over-estimators, whose self-ratings substantially exceeded observer ratings; under-estimators, whose self-ratings were substantially below observer ratings; and in-agreement/poor performers, whose self-ratings aligned with observers and were both relatively low. Their research found that the developmental trajectories across these four groups were substantially different: in-agreement leaders showed steady capability development, while over-estimators showed minimal development and high rates of derailment, not because they lacked capability but because they consistently received and rejected the feedback that would have enabled development.
The over-estimator profile is particularly consequential for organizational leadership development programs because it describes the leaders most resistant to the development interventions those programs provide. Leaders who believe their current performance level is substantially above what observers report are not in a mental state that makes them receptive to the feedback that would close the self-other gap. They experience observer feedback as inaccurate rather than as informative, and they direct their cognitive energy toward explaining the discrepancy in ways that protect their self-assessment rather than toward the behavioral change that would address the actual performance gaps observers are reporting. Development programs that do not address the over-estimator's self-other gap before delivering developmental content are delivering information to a recipient who is motivated to dismiss it.
The organizational cost of systematic over-estimation in leadership populations extends beyond individual development ineffectiveness to collective decision quality and organizational learning. Leaders who over-estimate their own effectiveness are less likely to seek feedback, less likely to recognize the value of external perspectives, and less likely to update their approaches in response to performance information that their self-serving attribution processing is discounting. They are also less likely to create the conditions for accurate upward communication in their teams, because leaders who are certain of their own effectiveness are less likely to actively solicit the challenging information that would contradict that certainty. The over-estimator's bias at the individual level therefore produces systematic information deficits at the team and organizational levels that compound the direct individual development costs.
The self-other rating gap is not uniformly distributed across assessment dimensions or across individual leaders. The gap is consistently largest in the dimensions that are most invisible to the self: interpersonal impact, the effect of one's behavior on others' emotional states and engagement; organizational awareness, the accuracy of one's understanding of how the organization actually works beyond one's own direct experience; and development of others, the degree to which one's management behavior is producing growth in direct reports. These dimensions share a common characteristic: they require the perspective of others to assess accurately, and they are dimensions where self-assessment depends on inference rather than direct observation. Leaders can directly observe their own task execution quality; they cannot directly observe the interpersonal effects their behavior produces in others.
Individual leader characteristics predict the magnitude of the self-other gap above and beyond assessment dimension. Leaders higher in narcissism, as measured by the validated Narcissistic Personality Inventory, consistently show larger self-other rating gaps, particularly on the interpersonal dimensions where self-assessment accuracy is most dependent on the accurate reading of others' responses. Leaders lower in emotional intelligence similarly show larger gaps on interpersonal dimensions, not necessarily because of motivated self-inflation but because the emotional awareness that would allow accurate reading of others' responses to one's behavior is itself inadequate. Both of these individual characteristics predict larger self-other gaps through mechanisms that require targeted developmental responses rather than simply better assessment feedback delivery.
Organizational factors also predict the self-other gap. Leaders in cultures where challenging senior performance, providing upward negative feedback, and reporting problems to those responsible for them is organizationally costly will receive consistently inflated observer ratings as observers self-censor the most challenging assessments. The resulting reduction in the self-other gap does not reflect improvement in self-assessment accuracy or in actual leadership effectiveness but rather the distortion of observer ratings toward the leader's self-assessment. Organizations that want accurate self-other gap data, and the diagnostic value that gap provides, must create the organizational conditions in which observers feel safe providing accurate ratings, including anonymous administration and organizational norms that protect raters from retaliation for assessments that leaders find unflattering.
| Self-other pattern | Profile | Developmental trajectory | Intervention priority |
|---|---|---|---|
| In-agreement / high performer | Self and observers both rate high; gap minimal | Steady development; high receptivity to feedback | Development challenge; stretch assignments |
| Over-estimator | Self rates substantially above observers | High derailment risk; rejects developmental feedback | Gap revelation; targeted reality-testing; coaching |
| Under-estimator | Self rates substantially below observers | Undersells capability; imposter syndrome risk | Confidence building; gap revelation; positive evidence |
| In-agreement / low performer | Self and observers both rate low; gap minimal | High development need; often receptive | Direct development support; specific behavioral targets |
The assessment design features most reliably associated with reducing self-other rating gaps, and with producing self-assessments that are diagnostically useful alongside observer assessments, are those that increase the observability of the behaviors being assessed from the self's perspective. Item framing that describes specific, observable behaviors rather than global competencies increases self-assessment accuracy by reducing the inferential gap between the self's observations of their own behavior and the assessment item's implicit standard for that behavior. Items asking leaders to assess their effectiveness at team problem-solving produce larger self-other gaps than items asking them to assess how often they solicited input from all team members before reaching a decision, because the behavioral item is anchored to a directly observable behavior rather than to a global effectiveness judgment that is vulnerable to the attribution and impression management biases that produce self-other gap inflation.
The pre-feedback preparation approach developed in 360-degree feedback practice, presenting leaders with an explicit analysis of their self-other gap before delivering detailed feedback, has been shown to increase feedback acceptance and development intention formation by reducing the defensiveness that unexpected gap revelation produces during the feedback delivery itself. When leaders have cognitive time and framing support to understand that self-other gaps are normative rather than reflecting unique personal failure or assessor inaccuracy, they are more capable of engaging with the specific content of the gap productively rather than spending their cognitive energy on explaining the gap away. The pre-feedback preparation investment is small relative to the assessment and development investment it supports, and its return in development engagement and behavioral intention formation is consistently positive.
The organizational infrastructure most supportive of reducing self-serving bias in leadership self-assessment over time is systematic and repeated multisource assessment, combined with structured coaching support for understanding and acting on the findings. Single-point multisource assessment events produce improvements in self-assessment accuracy that are modest and temporary. Repeated assessment, which shows leaders whether their self-other gap is narrowing over time, and which makes the gap trajectory itself a developmental target rather than a single diagnostic finding, produces more durable improvement in self-assessment accuracy. Organizations that invest in this infrastructure are investing not only in diagnostic accuracy for current development decisions but in the ongoing self-awareness capability that makes leaders increasingly accurate self-assessors and increasingly effective seekers of the feedback that their self-assessment process cannot independently generate.