What predicts a voluntary departure

The exit interview is conducted by the employer, after the decision, with a person who may have an interest in preserving the relationship. The reasons given there are systematically more consensual than in anonymous surveys.

Studies comparing the two sources find an over-representation of external reasons, notably an opportunity elsewhere, and an under-representation of reasons relating to colleagues and management.

That does not make the interview worthless. It can reveal problems the organisation was unaware of, provided it is read as an account addressed to the employer, and not as a neutral record of why the person left.

Reviews of the predictors of voluntary departure find associations of moderate size with reported satisfaction and engagement, and a stronger association with intention to leave, which is to be expected since the two measures are close.

The most powerful predictor remains external to the person: the availability of alternatives in the local labour market, a variable company surveys never measure. James March and Herbert Simon already distinguished in 1958 between the desire to leave and the ease of doing so, two dimensions that do not move together.

Thomas Lee and Terence Mitchell proposed in the 1990s a model in which departure is triggered by an identifiable event rather than by gradually growing dissatisfaction. The shock may be professional, such as a reorganisation, or personal, such as a partner's relocation or an unsolicited offer.

The model accounts for a fact that satisfaction surveys explained poorly: satisfied people leave, and dissatisfied people stay for a long time.

The same authors later developed the notion of job embeddedness, which describes the whole set of ties holding a person in their job and their area: relationships, fit with the post, what leaving would cost. It explains part of the staying that satisfaction alone does not.

Estimates of the cost of replacing an employee, widely quoted in management, mostly come from consultancies and rest on unpublished methods.

Academic work on the subject gives orders of magnitude that vary widely by occupation and skill level, which makes single figures of little use. It also points out that a departure does not only bring costs: it can open up an internal promotion or bring in new skills.

Administrative data on labour movements allow exit rates to be measured by sector and establishment size, independently of what people declare.

Without relying on what the people concerned say, they show strong variation between sectors and a marked sensitivity to the economic cycle, two dimensions that analyses centred on individual factors leave out.