Using a dynamic panel threshold approach across Chinese provinces, this study finds that institutional improvements are harmful to growth at very low levels of institutional quality up to a first critical threshold, beyond which better institutions raise growth until a second threshold where diminishing marginal returns set in.