Why you need to know this: if you have ever been rated differently by two managers for the same work, or you rate other people yourself, this briefing explains what a rating actually measures, what the person giving it brings to it, and what changes it.
The comfortable assumption about a performance rating is that it describes the person being rated, with a little noise around the edges. The evidence points the other way: in two samples of more than 2,000 managers each, rated by bosses, peers and staff, the raters’ own tendencies accounted for 62 and 53 per cent of the differences between ratings, and the performance of the person being rated for 21 and 25 per cent (Scullen, Mount & Goff, 2000). How large that share is remains argued over, and recent work puts agreement between supervisors higher than the older figures, but no estimate makes the rater a small part of the rating (Zhou, Sackett, Shen & Beatty, 2024; Jackson et al., 2022).
What the rater brings. A temperament: across 28 studies, raters’ own personalities accounted for between 6 and 22 per cent of the differences in the ratings they gave, with agreeable raters the most generous (Harari, Rudolph & Laginess, 2015).
A purpose: across 22 studies covering 57,775 people, ratings that decided pay or promotion were more generous than ratings given for development (Jawahar & Williams, 1997). An investment in the people they hired (Schoorman, 1988), a first impression that resists revision (Heslin, Latham & VandeWalle, 2005), and a self to protect (Fast, Burris & Bartel, 2014).
Who seems similar. The similarity that counts is not the visible kind. Across 82 samples, similarity in age, gender, race, and education showed no link to performance, while similarity in values and personality did, especially when it was perceived rather than measured (Condrea & Iliescu, 2026). Contact builds that perception: in one bank, smokers moved to a smoking manager advanced further without working harder (Cullen & Perez-Truglia, 2023), and personality shapes who is seen at all (Fang et al., 2015).
Where merit is the story. When a company was described as meritocratic, 445 participants with management experience gave an equally rated man a larger bonus than a woman (Castilla & Benard, 2010), and in one large employer the same rating bought smaller pay rises for some groups (Castilla, 2008).
What personality adds. On the rated side, people rate themselves more generously than their managers do, across 37,752 people (Heidemeier & Moser, 2009), and those higher in narcissism do so most for traits like leadership (Grijalva & Zhang, 2016). On the rater’s side the evidence about the darker traits is thin, and the briefing says how thin.
What changes it. If you are rated, read a rating as information about two people, let more than one person see your work and keep your own record. If you rate, write down why, and ask whether you would rate the same work the same way if the person did not remind you of yourself. If you run the organisation, use more than one rater, make raters explain themselves, and do not let fit stand in for performance.