This article examines the link between restrictions on the number of physicians and general practitioners' (GPs) earnings. Using a representative panel of 6016 French self-employed GPs over the years 1983-2004, we estimate an earnings function to identify experience, time and cohort effects. The estimated gap in earnings between 'good' and 'bad' cohorts can be as large as 25%. GPs who began their practices during the eighties have the lowest permanent earnings: they belong to the large cohorts of the baby-boom and face the consequences of an unlimited number of places in medical schools. Conversely, the decrease in the number of places in medical schools led to an increase in permanent earnings of GPs who began their practices in the mid-nineties. A stochastic dominance analysis shows that unobserved heterogeneity does not compensate for average differences in earnings between cohorts. These findings suggest that the first years of practice are decisive for a GP. If competition between physicians is too intense at the beginning of their careers, they will suffer from permanently lower earnings. To conclude, our results show that the policies aimed at reducing the number of medical students succeeded in buoying up physicians' permanent earnings. [Ed.]