Percentile survey

People rate their own skill too high. The workbook gives a simple example to show this.

Ask employees at a workplace one question: which percentile of performance do you think you are in, next to your peers? Most say somewhere between the 60th and 75th percentile. Almost nobody says below the 50th.

But that cannot be true of the whole group at once. Half of any real group must sit below the 50th percentile. The workbook calls this an "arithmetical impossibility."

The mechanism the survey exposes is overconfidence. If everyone genuinely ranked themselves accurately, the answers would spread out evenly around the 50th percentile — half above, half below. Instead, the workbook records a clear skew towards the 60-75 range, meaning the average respondent believes they are meaningfully above average, which cannot be true of the group as a whole.

The workbook connects this directly to investing behaviour: overconfidence of this kind shows up as investors attributing skill to bets that happened to come off, which can lead them to take on more risk than their actual ability to judge markets would justify.

Illustrative figures, following the workbook's own survey structure. Suppose a firm surveys 100 employees, asking each to place themselves in a performance percentile.

If answers were accurate, roughly 50 employees would place themselves below the 50th percentile and 50 above it. Instead, following the workbook's pattern, perhaps only 8 employees rate themselves below the 50th percentile, while the great majority cluster in the 60-75 range. Mathematically, at most 50 of the 100 can genuinely be above the 50th percentile — so the survey itself is the evidence of the bias, independent of any single employee's actual performance.

Chapter 17 (Behavioural Finance in Practice), the sub-section on "Overconfidence and dilution in risk management," uses this exact percentile-survey illustration. Expect a question testing whether a candidate recognises the 60-75 percentile clustering as an example of overconfidence bias, and why it is mathematically impossible for the group as a whole.

• The bias is about the group pattern, not any one individual's honesty. Any single employee could genuinely be in the 90th percentile — the impossibility only appears once most of the group claims to be above average.
• This is an overconfidence example, not a herding or anchoring example — it belongs with biases about self-assessment of skill, not about following others or fixating on a reference number.
• The workbook links this directly to attributing skill to a bet that came off — the connection to investor behaviour is the exam-relevant point, not the survey statistic in isolation.