We begin with two introductory chapters. The first describes the type of research agenda for which the material in subsequent chapters is most likely to be useful. The second discusses the sense in which randomized trials of the sort used in medical research provide an ideal benchmark for the questions we find most interesting. After this introduction, the three chapters of part II present core material on regression, instrumental variables, and differences-in-differences. These chapters emphasize both the universal properties of estimators (e.g., regression always approximates the conditional mean function) and the assumptions necessary for a causal interpretation of results (the conditional independence assumption; instruments as good as randomly assigned; parallel worlds). We then turn to important extensions in part III. Chapter 6 covers regression discontinuity designs, which can be seen as either a variation on regression-control strategies or a type of instrumental variables strategy. In chapter 7, we discuss the use of quantile regression for estimating effects on distributions. The last chapter covers important inference problems that are missed by the textbook asymptotic approach. Some chapters include more technical or specialized sections that can be skimmed or skipped without missing out on the main ideas; these sections are indicated with a star. A glossary of acronyms and abbreviations and an index to empirical examples can be found at the back of the book.