Under what circumstances do new constitutions promote democracy? Between 1974 and 2011, the level of democracy increased in 62 countries following the adoption of a new constitution, but decreased or stayed the same in 70 others. Using data covering all 138 new constitutions in 118 countries during that period, we explain this divergence through empirical tests showing that overall increased participation during the process of making the constitution positively impacts post-promulgation levels of democracy. Then, after disaggregating constitution making into three stages (drafting, debating, and ratification) we find compelling evidence through robust statistical tests that the degree of citizen participation in the drafting stage has a much greater impact on the resulting regime. This lends support to some core principles of “deliberative” theories of democracy. We conclude that constitutional reformers should focus more on generating public “buy in” at the front end of the constitution-making process, rather than concentrating on ratification and referendums at the “back end” that are unlikely to correct for an “original sin” of limited citizen deliberation during drafting.