Lyft priced its riders' minutes: 3.7M fares turned a marketplace into a value-of-time lab

Governments weigh roads and railways by the value of the time they save, so the number matters. In 1997 the U.S. Department of Transportation, using the best research then available, set personal travel time at half the typical household's hourly wage — about $14 an hour today. The figure rested on surveys, where people answer hypotheticals: what would you pay to save five minutes?

Lyft and a team of academic and in-house economists — Ariel Goldszmidt, John A. List, Robert D. Metcalfe, Ian Muir, V. Kerry Smith and Jenny Wang — ran a different kind of study. Before the pandemic they experimented on 3.7 million Lyft customers across nine American cities, tweaking prices and wait times and watching when riders requested rides and when they did not. Because every data point was a real purchase decision with real money and real waiting, the data revealed what time is actually worth to the person spending it.

The result: an average value of time of $19 an hour in the nine cities — significantly above the government's $14 — rising 50% during peak commute, and higher still in rain or snow, when waiting hurts most. The authors adjusted their sample against nationally representative government surveys to counter the skew of big-city taxi users. Metcalfe's conclusion: policymakers are underestimating people's time, which means projects that cut travel time are being undervalued in cost-benefit analysis.

Survey-based value-of-time numbers are hypothetical answers; Lyft's riders were spending real money on real waits, so every observation was a decision rather than an opinion.

The platform already varied prices and ETAs in ordinary course — the laboratory existed; only the measurement had to be designed in.

Lyft had its own commercial reason to know what a minute costs riders — pricing optimization — so the experiment ran at a scale (3.7M customers) no academic survey could recruit.

A platform's ordinary price experiments are an economic instrument no survey can match. Instrument them, and questions governments estimate from stale formulas get answered with observed behavior.

The study's headline was methodological as much as numerical: revealed-preference measurement of time at scale. Its policy implication — that the DOT's formula undervalues travelers' minutes and distorts infrastructure cost-benefit analysis — was the authors' stated target; Metcalfe said policymakers 'are not giving it the full weight that you fully give it.'