The increasing complexity of contemporary law systems is a central topic today. Several disciplines, such as economics, law and political science, emphasize the drastic consequences of legislative complexity. Poor quality legislation cannot be easily implemented and enforced, and moreover, can be easily challenged. Despite the great attention to this topic, the literature has used simplistic measures of legislative quality and has not provided a robust analysis of its effects. By building on recent developments in NLP, we look at the syntactic and lexical features of the language of legal texts to measure complexity and ambiguity. We apply these measures to a dataset of EU legislation, containing detailed information on its implementation and decision-making process. We find that complex and ambiguous acts are less likely to be implemented.