Beware simplistic attempts to change behaviour

At the same time, we should be open-eyed about the challenges we’ll face as we make that shift. Take this example from Mexico City: City officials in 1989 banned the general public from driving one weekday per week, based on the last digit of their license plates. The intent was to encourage use of mass transit options and thereby improve air quality. It was a noble upstream effort to prevent air pollution.

It didn’t work. Many Mexicans bought a second car, often an old clunker, to keep costs down - so they could drive every day. Air quality did not improve.

Mexico City restricted each car from being driven on one weekday, using the final digit of its licence plate. Many drivers responded by buying a second car, often an older and more polluting one. The regulation was followed while air quality failed to improve.

The policy targeted vehicle eligibility, although pollution was the intended outcome. Once the rule changed, people adapted around it. Similar responses appear when call targets increase low-value calls, closed-ticket measures encourage premature closure, or feature targets reward smaller and less coherent units of work. No dishonesty is required. People pursue mobility, income or manageable workloads under the constraints they are given. A policy designed for behaviour that remains fixed after intervention is designed for people who do not exist.

Before launch, the intended outcome needs a direct measure and the proposed rule needs an adversarial test. How could someone satisfy its wording while defeating its purpose? Which adjacent behaviour becomes cheaper? What evidence would reveal the workaround early enough to change course? In Mexico City, the relevant result was the air, not the count of restricted number plates. Meanwhile, the second cars were joining the same traffic, many with older engines. The emissions monitors kept recording what followed.