This is the first paper systematically calculating, testing and explaining different definitions of the survivorship bias in fund performance. We document that the survival-performance-relation is stronger for small funds and we find significant under-performance of non-survivors but no significant out-performance of new funds. Survivorship bias is still a problem as well in other fields of research, e.g., in countries where survivorship bias-free data is not available and because certain methods require truncated data. This paper provides guidance on how to deal with and reduce survivorship bias in empirical studies.