Extracting causal claims from observational economic data.
10 topics · 13 curated works
No prior grounding assumed.
The Credibility Revolution in Empirical Economics: How Better Research Design Is Taking the Con out of Econometrics
Joshua D. Angrist & Jörn-Steffen Pischke · 2010
Argues that applied microeconomics grew more credible over the last three decades not through fancier estimators but through research designs built…
Assumes you know the vocabulary.
14.32 Econometrics
Joshua D. Angrist (MIT OpenCourseWare) · 2007
Teaches regression as the working tool of applied economics, moving from the classical linear model through heteroskedasticity, endogeneity and…
+2 more at this level
Primary sources and full treatments.
Does Compulsory School Attendance Affect Schooling and Earnings?
Joshua D. Angrist & Alan B. Krueger · 1991
Quarter of birth predicts years of schooling through compulsory attendance laws but not ability, giving an instrument that isolates schooling's…
+8 more at this level
12 of 13 works
Joshua D. Angrist & Jörn-Steffen Pischke
Argues that applied microeconomics grew more credible over the last three decades not through fancier estimators but through research designs built around a specific, defensible source of variation.
22 pageslink checked 17 Sept 2026Joshua D. Angrist (MIT OpenCourseWare)
Teaches regression as the working tool of applied economics, moving from the classical linear model through heteroskedasticity, endogeneity and instrumental variables using real economic datasets throughout.
one-semester courselink checked 17 Sept 2026Thomas Herndon, Michael Ash & Robert Pollin
Replicates a widely cited debt-and-growth study and finds a spreadsheet coding error, selective exclusion of data, and an unconventional weighting scheme behind its headline result.
link checked 17 Sept 2026David Card
Argues that empirical economics improved most not from new statistics but from a change in habits of mind, toward designs that isolate a specific comparison the data can actually answer.
Nobel Prize lecturelink checked 17 Sept 2026Joshua D. Angrist & Alan B. Krueger
Quarter of birth predicts years of schooling through compulsory attendance laws but not ability, giving an instrument that isolates schooling's effect on earnings from who chooses to stay in school.
36 pageslink checked 17 Sept 2026Manuel Arellano & Stephen Bond
Develops a generalised method of moments estimator for dynamic panel models that uses lagged levels of the variables as instruments, and shows by Monte Carlo simulation that it outperforms the estimators then in standard use.
19 pageslink checked 17 Sept 2026Guido W. Imbens & Joshua D. Angrist
When an instrument only nudges some people to treatment, what an instrumental-variables estimate identifies is the effect on those compliers specifically, not on everyone.
9 pageslink checked 17 Sept 2026Francis X. Diebold & Roberto S. Mariano
Provides a test for whether one forecasting model's errors are genuinely smaller than another's, rather than just smaller in the sample that happened to be used to compare them.
10 pageslink checked 17 Sept 2026James J. Heckman
Argues that because people respond differently to the same policy, an average treatment effect estimated on one population is not automatically the right number for predicting what a different policy would do to a different population.
Nobel Prize lecturelink checked 17 Sept 2026Daniel McFadden
Sets out how discrete choice models recover the underlying preferences behind a decision between a finite set of options, from which car to buy to which route to commute, from data on which option people actually pick.
Nobel Prize lecturelink checked 17 Sept 2026Clive W. J. Granger
Explains why regressing one non-stationary economic series on another can produce a statistically significant relationship that is entirely spurious, and how cointegration identifies the cases where the relationship is real.
Nobel Prize lecturelink checked 17 Sept 2026Robert F. Engle
Argues that volatility itself is predictable even when returns are not, and traces how the ARCH family of models turned that observation into the standard tool for forecasting financial risk.
Nobel Prize lecturelink checked 17 Sept 2026This subject genuinely sits in more than one domain. These fields approach the same ground with different methods.