The eighty per cent myth

The formula appears in training materials, management articles and sales presentations, almost always without a reference. When a source is given, it points to a popular book, which in turn points to an intuition rather than a measurement.

No published research establishes this split. It is not a disputed or fragile result: it is a figure that has never existed in this form.

The starting point is real. Occupational psychology studies report that measures of cognitive ability explain a limited share of the variance in job performance, varying by occupation and by the criteria used.

The error consists in treating the remainder as a block and attributing it to a single factor. That remainder contains social background, the training received, health, seniority, the quality of management, the state of the sector and a share of chance. Nothing authorises giving it a single name.

The meta-analyses published around 2010, notably by Dana Joseph and Daniel Newman and then by Ernest O'Boyle and colleagues, find a positive link between emotional intelligence and work performance. It is weak, and it is clearer for so-called mixed instruments, which contain personality traits already measured elsewhere.

They also find a context effect more interesting than the overall average: the link strengthens in jobs where emotional labour is central, and becomes close to zero elsewhere. Assessing all of a company's candidates with the same instrument therefore amounts, for most posts, to measuring something unrelated to the task.

A round, memorable figure that overturns an established hierarchy and flatters those whom school served less well meets every condition for circulation. It also provides an immediate selling point for training courses that are hard to evaluate otherwise.

Its longevity also rests on the fact that nobody checks it: the source is always assumed to be known to the audience, and the formula is repeated in good faith by people who heard it elsewhere.

A defensible formulation would look like this: there are stable interindividual differences in the processing of emotional information, they are imperfectly measured, they have a positive and modest link with performance in occupations with high relational demands, and they are largely covered by personality.

That is markedly less saleable. It is also what one can maintain before someone who asks for the references, which is the only criterion that counts.