The Four-Day Week

What it is

Andrew Barnes ran the trial that made the idea famous, at Perpetual Guardian, a New Zealand wills and trusts firm of over two hundred and forty people across more than a dozen offices. It was a two-month trial, to be made permanent if productivity held. His arithmetic was that around three quarters of an hour more focused work a day would cover the missing day, which is the number that makes the whole proposition sound plausible, and it is his estimate rather than a measurement.

The measurement was done by Helen Delaney of the University of Auckland, and the findings should be credited to her rather than to Barnes or to the book that reports them. Time spent on social media at work fell, engagement, teamwork and stimulation rose, and stress fell.

In effect

The caveats matter more than the numbers, and there are five.

There was no control group, so nothing separates the four-day week from everything else that changed in those two months, including the attention the trial itself attracted. There was no comparison firm: one company, one sector, one country. The measures are partly self-report, since engagement, teamwork, stimulation and stress are asked rather than observed. The owner had publicly staked his reputation on the result and the staff had an obvious interest in a permanent four-day week, so everyone answering the questionnaire knew what answer produced the outcome they wanted. And Delaney's own quoted verdict is far more modest than the headline, amounting to the observation that it was not a monstrous failure, which is the most honest sentence in the account. Her fuller paper was still under peer review when Johann Hari drew on it in Stolen Focus.

Hari flags the vested interest himself and prints the numbers anyway. He also prints four further instances alongside Delaney's findings, none with a source and none distinguished from hers by evidential grade: a Microsoft Japan productivity improvement in 2019, which is a company press claim from a one-month trial rather than a study; a Gothenburg care home on six-hour days, with no numbers given; Toyota mechanics in Gothenburg, with an output figure and a profit figure but no source, no period and no definition of the baseline; and a Kellogg claim about accidents attributed to 1920s Britain, which is almost certainly wrong on its face, since the Kellogg six-hour-day experiment was American, in Michigan, and began in 1930. That last is recorded as printed and should not be repeated.

The pattern is worth naming because it recurs: one genuine piece of academic work, weakly designed, surrounded by press releases and half-remembered history, all presented in the same voice.

What it does not say

It does not show that a four-day week raises productivity. The trial had no control and no comparison, and the outcome that mattered commercially, whether the work got done, was assessed by a firm whose owner had predicted it would.

It does not show that shortening the week removes the pressure. Leslie Perlow's cycle of responsiveness explains why simply compressing the days will not work unless the expectation of availability goes with it, and the trial did not test that separately.

It does not carry figures this publication can vouch for. No sample size, instrument or analysis is printed in the source, Delaney's fuller work was unpublished at the time, and none of the four comparative cases has a source at all.

It is the exact case preregistration exists for. A trial whose sponsor had publicly forecast the result, measured largely by self-report, with no control group, is not a design that can tell you whether you were right.

---