A software company asks every employee to complete a self-assessment before the annual review, rating themselves from 1 to 5 on six competencies. Managers see the self-ratings before writing their own.
Oliver, a mid-level developer of average output, gives himself 5s across the board and writes four confident paragraphs about his impact. Yuki, whose code underpins the company’s main product and who is the person colleagues turn to when something breaks, gives herself 3s and writes: “Met objectives. Need to improve estimation.” Their manager, short of time, starts from the self-ratings and adjusts. Oliver ends on 4. Yuki ends on 3.5. Later, in the pay review, Oliver is flagged as a high performer. Nobody misrepresented anything. Both wrote down what they believed. The process took two people’s opinions of themselves and treated the difference between them as a difference in performance.
Self-assessment is a standard part of performance reviews, development planning and, increasingly, pay and promotion cases. It rests on a reasonable idea: nobody knows your work better than you do.
The difficulty is that knowing your work and judging its quality are different abilities, and the second is distorted in predictable ways. This guide looks at what research shows about how people rate themselves, why some overrate and others underrate, and how to use self-assessment without importing those distortions into decisions.
Self-ratings are the weakest measure.
They agree only modestly with managers and peers, who agree much more with each other.
Errors run both ways.
Most people overrate themselves; some of the most able underrate or under-describe themselves.
Collect evidence, not scores.
Ask for achievements and reflection, and have managers rate before they read.
How accurate are self-ratings?
Not very, if accuracy means agreement with other observers.
A meta-analysis of studies comparing rating sources found that self-ratings correlated only modestly with supervisor ratings (around .35) and with peer ratings (around .36), while peers and supervisors agreed with each other considerably more (around .62).
So the two outside perspectives resemble each other much more than either resembles the person’s own. Self-ratings also tend to be higher. On average, people place themselves above where their managers and colleagues place them.
This does not make self-assessment worthless. It contains information nobody else has: what was hard, what was done out of sight, what the person is trying to improve. But as a measurement of performance level, the self-rating number is the weakest input available.
Why do most people think they are above average?
Because of a cluster of tendencies that psychologists group under “self-enhancement”.
The classic demonstrations are well known. In one survey of university teaching staff by Patricia Cross, 94% rated themselves above average as teachers. In a study by Ola Svenson, a large majority of drivers rated their driving skill and safety as above the median. This “better-than-average effect” has been found for intelligence, ethics, leadership and job performance.
Several mechanisms contribute.
- Privileged access. You know about all your own effort and good intentions and only some of other people’s.
- Flattering definitions. “Good performance” is ambiguous, and people define it around their own strengths. The careful worker defines quality as accuracy. The fast one defines it as speed.
- Self-serving attribution. Successes are attributed to ability, failures to circumstances.
- Selective memory. Achievements are recalled more readily than lapses.
In most groups, most members believe they are above the group’s average. Any system that tells half its people they are below the middle will therefore contradict the self-image of most of them, a point with consequences for forced ranking and pay transparency.
What is the Dunning-Kruger effect, really?
It is the finding that people with the least skill in an area tend to overestimate their performance the most.
In tests of logical reasoning, grammar and humour, participants scoring in the bottom quarter estimated on average that they had performed better than most of their peers; those at around the 12th percentile placed themselves near the 62nd. Top scorers slightly underestimated their relative standing.
The authors’ explanation was that the skills needed to do a task well are often the same skills needed to recognise good performance. Someone with a poor grasp of grammar cannot see their own errors. Top performers, for their part, tended to assume that what was easy for them was easy for everyone.
The effect is often exaggerated in popular accounts. The original data do not show that the least competent believe themselves the most competent; estimates rose with actual performance, just not steeply enough. Statisticians have also pointed out that part of the pattern arises from measurement noise and regression to the mean. What remains is a modest and useful point: lack of skill limits insight into lack of skill, so self-assessment is least reliable where development is most needed.
Why do some strong performers rate themselves low?
For several different reasons, which call for different responses.
Impostor feelings. Pauline Clance and Suzanne Imes described in 1978 what they called the impostor phenomenon: capable people who believe their success is due to luck or to deceiving others and who fear being found out. It was first observed among high-achieving women and has since been documented in men and across many professions. Estimates of how common it is vary enormously between studies, depending on how it is measured. It is an experience, not a clinical diagnosis.
High standards. Experts know how much they do not know. Yuki compares her work with what it could be, not with what her colleagues produce.
Cultural norms. Modesty is expected in many cultures, and self-promotion is regarded as poor manners. Cross-cultural research has found weaker self-enhancement, and sometimes self-criticism, in East Asian samples compared with North American ones.
Anticipated backlash. For some groups, confident self-description is received less well.
In a series of experiments, women described their own performance on a test less favourably than equally performing men when asked to rate it for a potential employer. The gap persisted even when participants were told exactly how well they had scored, and it did not appear when they rated another person’s performance.
That last detail is important. The women in the study were not mistaken about how they had done. They were more reluctant to describe it favourably. A self-assessment form therefore measures willingness to promote oneself alongside performance.
What happens when managers see self-ratings first?
They are influenced by them. This is anchoring, the tendency for a number seen first to pull subsequent estimates toward it, described in our guide to salary negotiation psychology. A manager who reads a self-rating of 5 before deciding their own will, on average, end up higher than one who has not seen it.
Field research on this has been done by Iris Bohnet and colleagues, who studied a company in which a system change meant managers in some cycles did not see employees’ self-ratings before rating them. Their work indicates that showing self-ratings to managers shapes the final ratings and can carry group differences in self-rating through to the outcome.
The mechanism is the one in the opening example. If Oliver inflates and Yuki deflates, and the manager adjusts part of the way from each starting point, the final ratings preserve some of the gap. The system has converted a difference in self-presentation into a difference in pay.
If self-ratings on a numerical scale are shown to managers before they rate, you are almost certainly rewarding self-confidence. This is among the easiest biases in performance management to remove.
How should self-assessment be used?
Keep the information and remove the anchor.
Drop the self-rating number. Ask instead for achievements with evidence, what was harder than expected, what the person would do differently and what they want to develop. These questions produce useful material from confident and modest people alike.
Sequence it. The manager forms and records a view before reading the self-assessment, or the self-assessment is shared without scores.
Prompt for completeness. Modest people omit things. Ask “What did you do this year that I may not know about?” and “What would your colleagues say you contributed?”
Use it for the conversation. The gap between the manager’s view and the employee’s is the most valuable thing a review can surface. A large gap in either direction means that feedback during the year did not get through, a problem addressed in our guide to feedback psychology.
How do you help people calibrate?
Self-assessment improves with clear standards and with feedback.
- Show what good looks like. Behavioural descriptions of each level, with examples, give people something to measure against. Overestimation shrinks when the standard is concrete.
- Give frequent, specific feedback. People who learn regularly how their work is seen have less room for illusion in either direction.
- Use multiple sources. Structured input from colleagues shows people how they come across. It should be used for development and handled with care.
- Teach the skill. Dunning and Kruger found that training low performers in the underlying skill improved their ability to judge their own performance. Competence brings calibration.
For the overconfident, the aim is accurate information delivered without humiliation. Describe the standard, show the evidence and agree a concrete next step. For those who underrate themselves, specific recognition matters. Telling an impostor-prone expert “you’re great” is ineffective; they discount it. Telling them exactly what they did and what it made possible is harder to dismiss.
In promotion and pay cases, do not rely on people nominating themselves or writing their own case. Have managers consider everyone at a level against the criteria. Processes that depend on self-advocacy select for it, as discussed in our guide to promotion bias.
Does it matter for hiring too?
Yes. Interviews, covering letters and “rate your proficiency” questions on application forms are all self-assessments. Confident candidates describe modest experience expansively. Others undersell substantial experience. Work samples and structured questions about specific past behaviour, as set out in the case for structured interviews, reduce the weight given to self-description.
The software company made two changes. Self-assessments no longer include scores, only achievements and reflections, and managers submit their ratings before the self-assessments are released to them. In the next cycle Yuki was rated at the top of her group. She said in her review that she thought the rating was generous, and her manager spent ten minutes going through the evidence with her.
Frequently asked questions
How accurate are employee self-assessments?
They show only modest agreement with ratings from managers and peers and are usually higher. They are valuable for the information and reflection they contain, but unreliable as a measure of performance level.
What is the Dunning-Kruger effect?
It is the finding that people with low skill in a domain tend to overestimate their performance, partly because the knowledge needed to perform well is also needed to recognise good performance. Highly skilled people tend to underestimate slightly.
What is impostor syndrome at work?
The impostor phenomenon describes capable people who doubt their abilities, attribute success to luck and fear being exposed as a fraud. It is a common experience, not a medical diagnosis, and it leads people to understate their contribution.
Should managers see self-ratings before rating employees?
Preferably not. Seeing an employee’s numerical self-rating first tends to anchor the manager’s rating toward it, which rewards confident self-presentation. Collect written evidence instead, and have managers form their view independently.
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.