Censorship makes new ideas less available to others, but also reduces the share of people choosing to develop non-compliant ideas. We propose a new method to measure the effect of censorship on knowledge growth, accounting for the agents' choice between compliant and non-compliant occupations. We apply our method to the Catholic Church's censorship of books written by members of Italian universities and academies over the period 1400-1750. We highlight two new facts: once censorship was introduced, censored authors were of better quality than the non-censored authors, but this gap shrank over time, and the intensity of censorship decreased over time. These facts are used to identify the deep parameters of a novel endogenous growth model linking censorship to knowledge diffusion and occupational choice. We conclude that censorship reduced by 34% the average log publication per scholar in Italy, while adverse macroeoconomic processes are responsible for another 9% reduction. Interestingly, the induced reallocation of talents towards compliant activities explains half the effect of censorship.