What labour productivity measures

Hourly labour productivity relates value added to a volume of hours worked. It does not measure effort or intensity, but an accounting ratio between two aggregates.

It is not attributable to labour alone: it incorporates the effect of capital, organisation, technology and sectoral composition. The indicator's name suggests an attribution its computation does not support.

For that reason economists compute another quantity, total factor productivity, which attempts to isolate what is explained neither by labour nor by capital. Robert Solow brought it to light in 1957 as a residual, which makes plain that it measures first of all what we cannot explain.

Value added is measured in monetary units, which presupposes prices. In non-market services, for want of market prices, output is conventionally measured by costs, which makes productivity in those sectors hard to interpret.

Quality changes pose a second difficulty: an improvement in the service delivered without a change in price may appear as stagnant productivity.

As early as the 1960s William Baumol described a related mechanism: in certain activities, such as a concert or a medical consultation, working time is part of the service itself. Their measured productivity rises little, while their costs follow wages across the economy as a whole.

The volume of hours worked rests on estimates combining surveys and administrative data. For salaried staff on annual-day contracts and for the self-employed, it derives from declarations of low precision.

The treatment of undeclared overtime and of work at home varies by country, which directly affects international comparisons. A country that underestimates its hours mechanically shows higher hourly productivity, without its organisation of work being any different.

A country's measured productivity rises if the least productive jobs disappear, without any worker having changed how they work.

That mechanism explains part of the movements observed during periods of sharp employment variation, and it leads to mistaken readings when ignored. In a recession, average productivity may thus rise because the jobs cut were the least productive, which is no sign of a healthy economy.

Rich countries have seen since the 2000s a slowdown in productivity gains, a robust and debated finding. Proposed explanations range from insufficient measurement of digital services to the exhaustion of earlier innovations' effects, a thesis defended notably by Robert Gordon. The debate is not new: in 1987 Robert Solow already observed that you could see computers everywhere except in the productivity statistics, a paradox that faded in the following decade before reappearing in other forms.

None of those explanations commands consensus, and the mismeasurement hypothesis has been tested by several studies concluding that it explains only a limited share of the slowdown.