According to Angrist and Pischke (2008), not everything on the menu is equally valuable or important. Some of the more exotic items are needlessly complex and may even be harmful.   First, they believe that empirical research is most valuable when it uses data to answer specific causal questions, as if in a randomized clinical trial. This view shapes their approach to all research questions. In the absence of a real experiment, they look for well-controlled comparisons and/or natural ‘quasi-experiments’. Of course, some quasi-experimental research designs are more convincing than others, but the econometric methods used in these studies are almost always fairly simple.  Based on this idea, the most important items in an applied econometrician’s toolkit are: 1. Regression models designed to control for variables that may mask the causal effects of interest; 2. Instrumental variables methods for the analysis of real and natural experiments; 3. Differences-in-differences-type strategies that use repeated observations to control forum observed omitted factors.

　　To summarize, Angrist and Pischke (2008)’s perspective is as follows: you want to do a causal inference, you already have an outcome measure and a treatment indicator—in their examples, the outcome is almost always continuous and the treatment is almost always binary—and you also probably have some pretreatment measures. And then you run a regression. Angrist and Pischke (2008) explain how direct regression works, and then they discuss how various methods such as instrumental variables and discontinuity analysis can help you make use of any quasi-experimental structure in your data. Despite its broad title, Angrist and Pischke (2008)’s coverage is limited to one corner of econometrics.   The data structures considered are cross-sectional, with no time series and virtually no panel or time-series cross-sectional data.   And, in the world of Angrist and Pischke (2008), the problems of interest are all causal:  there is no forecasting, no descriptive modeling, and no theory testing. Finally, when it comes to methods, the book is all about least-squares estimation of regression coefficients:  there is little to nothing about nonparametric methods, Bayesian inference, or models other than linear regression (multi-level/hierarchical, sophisticated sampling method, etc.)