To reconcile the mixed results emerging from the empirical literature, we first develop a theoretical model whose main implication is a concave impact of regulation on the probability of a crisis, and then we test this relationship by applying a Probit model of a non-linear specification to annual data from 1999 to 2011 drawn from 132 countries. Our key inference is that the probability of a financial crisis fits an inverted U-shaped curve: it rises as regulation stringency moves from low to medium levels and falls from medium to high levels. Countries located at the intermediate level of regulatory stringency face more financial instability than countries that are either loosely or severely regulated. We identify the latter two groups as falling in “liberalization traps”. Institutional quality interacts significantly with the regulatory environment, implying trade-offs between regulatory stringency and institutional quality.