Another set of four-day-week results lands. The write-ups draw the conclusion they drew last time. Output per employee rose at the firms that cut the week to four days. Their staff, asked afterwards, said they liked it better. The coverage has moved on from arguing that the idea is worth testing to arguing that the test is over. I have no quarrel with the counting. Each firm measured its own output for six months before the change and six months after it, by the same definition at both ends, and where a firm altered how it counted mid-way the analysts said so. The gains are real, and nobody cooked them. That is a before-and-after design: every comparison runs between a firm and its own earlier self.

The trouble starts with who was in the trials. Nobody was assigned a four-day week. Every firm that took part asked to take part, and asking is not a neutral act. A company puts itself forward once its managers already believe it can absorb the loss of a day, because the work compresses, or because the year has slack somewhere to swallow a bad month. The confidence came first and the application came second. That confidence is rarely idle. Managers who expect the coming year to go well are usually looking at an order book that has filled up or a product that has begun to move. A firm in that position had every reason to post better figures over the following six months with or without a day off the week, and its own earlier figures are the wrong thing to measure it against. Firms that expected a shorter week to sink them never applied. Nobody hears from them. The filter here is applying, and it does not care what a firm makes or how many people it employs.

None of this makes a four-day week a bad idea. It may well raise productivity, and I would be glad to be shown that it does. But what these trials caught was a set of firms on the way up, doing what firms on the way up do, with a shorter week laid over the top. The extra output has an obvious rival explanation and the design cannot rule it out. Settling the question would take firms that never volunteered, weeks shortened whether the idea appealed to them or not, and the failures published as loudly as the wins. Until somebody runs that study, the productivity case for the four-day week rests on the judgement of the firms most likely to succeed at it.