Table of Contents
Abstract
This paper explores how the statistical phenomenon known as regression to the mean affects perceptions and decisions in multi-round job interviews. When candidates perform exceptionally well or poorly in initial interviews, their subsequent performances tend to move closer to their true average ability, not because of changes in competence, but due to normal statistical variation. This natural fluctuation is often misinterpreted by interviewers, who tend to seek causal explanations for changes in performance rather than recognizing them as statistical inevitabilities. By analyzing this phenomenon through the lens of cognitive psychology and statistical principles, this paper contributes to our understanding of biases in hiring processes and offers insights for both interviewers and candidates navigating multi-stage interviews.
1. Introduction
Job interviews are crucial evaluation tools in hiring decisions, with many organizations employing multiple interview rounds to assess candidates thoroughly. However, these evaluations are vulnerable to cognitive biases that can distort interviewers’ judgments about candidates’ true abilities.
One significant source of misinterpretation is the failure to recognize regression to the mean, a statistical phenomenon first documented by Sir Francis Galton in the 19th century. This principle states that when a variable is extreme on first measurement, it will typically be closer to the average on subsequent measurements. Despite its statistical inevitability, regression to the mean is frequently overlooked in favor of causal explanations for performance changes.
In his influential book “Thinking, Fast and Slow,” Daniel Kahneman illustrates how we often misinterpret regression to the mean in various contexts, from sports performance to academic achievement. This paper examines how this same phenomenon manifests in multi-round interviews and explores its implications for hiring practices. By understanding regression to the mean, both interviewers and candidates can better navigate the interview process.
2. Understanding Regression to the Mean
2.1 The Statistical Principle Explained
Regression to the mean occurs when variables that are extreme on first measurement tend to be closer to average on follow-up measurements. This happens because most measurements contain both a signal (true ability) and noise (random factors). When we observe an extreme result, it often represents both the true signal and favorable random factors. On subsequent measurements, these random factors are unlikely to align in the same way.
To illustrate this concept simply: if you flip a coin 100 times and get 70 heads (an extreme result), and then flip it another 100 times, the second result will likely be closer to 50 heads, not because the coin has changed, but because the first result included both the true nature of the coin and random luck.
2.2 Kahneman’s Flight Instructor Example
Kahneman provides a compelling example from his experience teaching Israeli flight instructors. One instructor claimed that praising cadets for good performance typically led to worse performance next time, while criticizing poor performance often led to improvement. The instructor concluded that punishment was more effective than praise.
Kahneman recognized this as a misinterpretation of regression to the mean. Exceptional performances (whether good or bad) tend to be followed by more average performances. The instructors were observing a statistical phenomenon but attributing it to their teaching methods.
As Kahneman explains, the flight instructors “were trapped in an unfortunate contingency: because they punished cadets whose performance was poor, they were mostly rewarded by a subsequent improvement, even if punishment was actually ineffective” (Kahneman, 2011, p. 175). This illustrates how regression to the mean can create an illusion of cause and effect where none exists.
2.3 The Statistical vs. Causal Thinking Divide
Our difficulty with regression to the mean stems from a conflict between two cognitive systems Kahneman identifies:
- System 1: Intuitive, automatic thinking that seeks causal narratives and explanations
- System 2: Analytical, deliberate thinking capable of understanding statistical concepts
“Our difficulties with the concept of regression originate with both System 1 and System 2,” Kahneman explains. “System 1 finds it difficult to understand and accept the principle of regression. System 2 may learn about regression, but this knowledge does not abolish the intuitive impressions of System 1” (Kahneman, 2011, p. 183).
This divide helps explain why even statistically knowledgeable people may misinterpret regression effects in real-world settings like job interviews.
3. Regression to the Mean in Multi-Round Interviews
3.1 Performance Variability in Interview Settings
Interview performance is determined by multiple factors:
- The candidate’s true ability and qualifications
- Variable factors including stress, question difficulty, interviewer rapport, and daily energy levels
- Random elements of luck and timing
This combination of factors makes interview performance inherently variable, creating ideal conditions for regression to the mean to occur.
3.2 Three Performance Scenarios
3.2.1 Exceptional First Interview Performance
When a candidate performs exceptionally well in an initial interview, this outstanding performance often reflects both their true ability and favorable random factors (perhaps they were asked questions that perfectly matched their expertise, or they were particularly well-rested).
Statistical principles predict that their performance in subsequent rounds will likely be closer to their true average, appearing as a “decline” even if their actual ability remains unchanged. Interviewers may incorrectly interpret this statistical regression as diminishing enthusiasm, overconfidence, or inconsistency.
3.2.2 Average First Interview Performance
Candidates who perform at an average level in their first interview provide the most accurate reflection of their true abilities. Their subsequent performances are likely to remain relatively consistent, creating an impression of stability and reliability.
Interestingly, these candidates might benefit in the long run, as they neither set unrealistically high expectations nor create initial negative impressions that must be overcome.
3.2.3 Poor First Interview Performance
Candidates who perform poorly in their first interview (perhaps due to nervousness, miscommunication, or simple bad luck) will statistically tend to improve in subsequent rounds, reverting closer to their true ability level.
Interviewers might attribute this improvement to the candidate’s learning ability, resilience, or response to feedback, when in fact it may simply represent regression to the mean. While this misattribution might benefit the candidate in the short term, it can create unrealistic expectations about their rate of growth or adaptability.
4. Cognitive Biases Amplifying Regression Effects
Several well-documented cognitive biases interact with regression to the mean, amplifying its misinterpretation in interview settings:
4.1 The Halo Effect and Expectation Management
The halo effect occurs when one positive trait or impression influences the perception of other traits. In multi-round interviews, this bias creates particularly problematic expectations:
When a candidate performs exceptionally well in a first interview, the interviewer develops an inflated overall impression of their abilities. This creates unrealistically high expectations for subsequent interviews. When the candidate’s performance inevitably regresses toward their true mean in later interviews (due to statistical principles, not declining ability), the interviewer experiences a heightened sense of disappointment. This disappointment is disproportionate because it stems from a comparison to an unsustainable benchmark rather than a fair assessment of ability.
As Thorndike (1920) first documented, this effect can significantly distort evaluation processes by creating a psychological contrast between first impressions and subsequent observations.
4.2 Confirmation Bias and Performance Trajectory Misinterpretation
Once interviewers form initial impressions, they interpret subsequent information through this lens, often reinforcing their original judgment regardless of actual performance changes.
For candidates who excel initially, confirmation bias leads interviewers to attribute any regression toward average performance to negative personal factors (decreased motivation, inconsistency, or overconfidence) rather than to statistical inevitability.
Conversely, for candidates who perform poorly at first, any improvement may be discounted as insufficient because the initial negative impression continues to influence perception, despite regression to the mean naturally improving their performance.
This powerful bias, which Nickerson (1998) identified as “perhaps the most problematic aspect of human reasoning,” prevents interviewers from recognizing the statistical nature of performance variations across multiple interviews.
4.3 The Fundamental Attribution Error
This bias describes our tendency to overestimate personal factors and underestimate situational factors when explaining others’ behaviors. In interview contexts, this leads to attributing performance changes to the candidate’s intrinsic qualities rather than to statistical variability.
When a stellar first-round performer shows more average results in subsequent rounds, interviewers typically attribute this to personal factors (“they don’t handle pressure well,” “they became overconfident”) rather than recognizing it as a predictable statistical phenomenon. Similarly, improvements from poor initial performance are attributed to personal growth rather than regression.
Ross (1977) identified this bias, which directly undermines proper understanding of regression to the mean by pushing evaluators toward causal rather than statistical explanations for performance fluctuations.
5. Practical Implications
5.1 For Interviewers
Understanding regression to the mean primarily offers awareness rather than definitive solutions, as this statistical phenomenon is inherent to evaluation processes:
- Statistical Awareness: The most important implication is simply developing awareness that changes in performance between interviews often reflect regression to the mean rather than meaningful changes in a candidate’s ability, motivation, or interest. This awareness can help interviewers resist drawing strong causal inferences from performance variations.
- Tempering Expectations: Recognizing that extreme performances (either exceptional or poor) in initial interviews will likely be followed by more average performances can help interviewers avoid setting unrealistic expectations or making premature judgments.
- Recognizing the Limits of Intervention: Even with structured interviews and standardized evaluation criteria, regression to the mean will still occur. The goal should not be to eliminate it (which is statistically impossible) but to recognize its influence when interpreting candidate performance.
- Conscious Debiasing: Interviewers can consciously question their interpretations of performance changes, asking themselves: “Am I seeing a meaningful pattern, or could this be regression to the mean?” Simply raising this question can help counter the automatic causal interpretations that System 1 thinking tends to generate.
5.2 For Candidates
Candidates can also benefit from understanding regression to the mean:
- Managing Expectations: Candidates who perform exceptionally well in their first interview should recognize that maintaining this peak performance may be challenging due to statistical factors, not personal deficiencies.
- Recovery Opportunity: Those who perform below their capabilities initially should take heart that statistical principles favor improvement in subsequent rounds.
- Consistency Focus: Aiming for consistent, representative performance across all interviews may be more effective than trying to create an exceptional first impression that sets unsustainable expectations.
- Strategic Performance: Some candidates might consider a measured first interview approach that demonstrates their baseline competence, with room to build upon in later rounds.
6. Theoretical Insights and Research Directions
This analysis of regression to the mean in job interviews contributes several theoretical insights:
- Cognitive-Statistical Tension: The tension between our cognitive preference for causal explanations and the statistical reality of regression to the mean is particularly acute in evaluative contexts like job interviews.
- Performance Attribution Bias: Interviewers exhibit a systematic bias in attributing performance changes to candidate characteristics rather than to statistical phenomena.
- System 1 Dominance: Even when interviewers intellectually understand regression to the mean (System 2), their intuitive judgments (System 1) often override this understanding in practice.
Future research could explore these dynamics through:
- Empirical Studies: Controlled studies comparing interviewer judgments across multiple rounds with objective measures of candidate ability.
- Debiasing Techniques: Development and testing of methods to help interviewers recognize and account for regression to the mean in their evaluations.
- Cross-Cultural Comparisons: Investigation of whether different cultural approaches to interviewing are more or less susceptible to regression-based misinterpretations.
7. Conclusion
Regression to the mean plays a significant yet often unrecognized role in multi-round job interviews. When extreme performances in initial interviews are followed by more average results, both interviewers and candidates tend to seek causal explanations rather than recognizing the statistical nature of these changes.
By understanding regression to the mean and its interaction with common cognitive biases, we can develop more nuanced approaches to the interview process. Interviewers can make more accurate assessments by considering statistical principles, while candidates can develop strategies that account for the inevitable pull toward the average.
As Kahneman notes in the context of performance evaluation: “Our screening procedure is good but not perfect, so we should anticipate regression. We shouldn’t be surprised that the very best candidates often fail to meet our expectations” (Kahneman, 2011, p. 184). This insight applies equally to job interviews, where recognizing the influence of regression to the mean can lead to fairer, more accurate hiring decisions.
References
Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.