Teaching Econometrics

Hillebrand, E. and W.E. Griffiths (Editors): Teaching Econometrics - A Tribute to R. Carter Hill. Advanced Studies in Theoretical and Applied Econometrics, Vol. 56, 2026, Springer (doi)

Chapter 1 Introduction

Chapter 2 William Greene: Teaching Applied Econometrics

In this chapter, I will describe some devices that I have used in teaching Applied Econometrics. The essential outline for this course is widely agreed upon. I have used some specific devices and extensions to engage my students. Most of these are examples, applications, methods or vignettes that expand the typical curriculum.

Chapter 3 Gary Koop: Reflections on the Teaching of Bayesian Econometrics

This chapter discusses my experience of teaching Bayesian econometrics over the last 30 years. It offers an overview of how the teaching of Bayesian econometrics has developed since the 1980s due to the huge improvement in computing power that has occurred. Prior to the 1980s, Bayesian research was mostly theoretical with empirical work being limited to a narrow range of models and priors. Advances in computing power, along with algorithms, which could exploit this extra computer power, revolutionized Bayesian econometrics since it could now handle a vast range of empirically-interesting models, including complicated high-dimensional models capable of handling the large data sets economists now have available. These developments in research were matched by a change in the way Bayesian econometrics is taught. This chapter offers my personal reflections on this change. My Bayesian course has evolved to place much more emphasis on Bayesian computation. Strategies are discussed for simple ways of teaching Bayesian econometrics and the associated computational methods in a relatively non-technical manner.

Chapter 4 Stan Hurn, Vance Martin, Peter C.B. Phillips, Jun Yu: Teaching Financial Econometrics to Students Converting to Finance

Financial econometrics is a dynamic discipline that began to take on its present form around the turn of the century. Since then it has found a permanent position as a popular course sequence in both undergraduate and graduate teaching programs in economics, finance, and business schools. Because of the breadth of the subject’s foundations, its extensive coverage in applications and because these courses attract a wide range of students with accompanying interests and skill sets that cover diverse areas and technical capabilities, teaching financial econometrics presents many challenges to the university educator. This chapter addresses some of these challenges, provides helpful guidelines to educators, and draws on the combined experience of the authors as teachers and researchers of modern financial econometrics as well as their recent textbook Financial Econometric Modeling (Hurn et al., 2021). The focus throughout is on students who are converting to finance and econometrics with limited technical background.

Chapter 5 Badi H. Baltagi: Teaching Panel Data Econometrics

The theme of this volume, dedicated to a great econometrics teacher, is the teaching of econometrics and its pedagogics. In this spirit, this paper revisits some of the basics in the econometrics of panel data and gives my style of teaching this material using tricks from matrix algebra that I found useful to illustrate their value especially for applied researchers who normally want to skip the derivations and focus on applications. I use the Frisch-Waugh-Lovell theorem discussed in Davidson and MacKinnon (1993, p. 19) to explain why the fixed effects estimator can be obtained from the within regression, which is easier to compute and is the standard practice in panel data software. Next, I use a beautiful trick from Wansbeek and Kapteyn (1982) to derive the random effects estimator as a weighted least squares as suggested by Fuller and Battese (1974). The paper then emphasizes that rejecting the Hausman (1978) specification test based on fixed versus random effects is not necessarily an endorsement of the fixed effects estimator as is done regularly in practice. Rejection of the null means that there is misspecification, and Baltagi (2024) provides alternatives to selecting the fixed effects estimator when the Hausman test rejects. Next, the paper demonstrates that for unbalanced panel data with missing observations, things get messy, and the tricks get fancier, and the students complain.

Chapter 6 Jan R. Magnus: The Harm that Good Teachers Do and Other Stories

Happy birthday, Carter! You have been a big inspiration to the econometrics community, and I feel honored and privileged that Eric and Bill have invited me to be part of this collection of essays on teaching. Below, I offer some of my thoughts; however personal and misguided they might be.

Chapter 7: Anson T.Y. Ho, Kim P. Huynh, David T. Jacho-Chávez, Katie Leinenbach, Carson H. Rea: Teaching Reproducibility and Replicability while Teaching Econometrics in the Classroom

This research discusses how reproducibility and replicability can be taught to economists and social scientists while learning econometrics. Instructors can use standard tools from data science and machine learning to teach classical undergraduate Econometrics curriculum. This paper emphasizes the usage of self-contained computing environments for students to complete and submit their econometric practice exercises using open-source software. Demonstrations highlight how instructors can create computer-based assignments that can be distributed electronically to students, and how researchers can easily replicate and reproduce research using the same tools. For students, assignments are accompanied by code that automatically deploys a computing environment in the cloud where the assignment can be completed without the need for further software installation or a hardware upgrade. This teaches students how to prepare their work to be reproducible and replicable.

Chapter 8 William Griffiths: Learning Basic Bayesian Econometrics Using EViews

The use of Bayesian econometrics as a research tool has exploded over recent decades, but, despite this explosion, it is absent from many introductory econometrics’ courses, often being introduced only as an advanced specialist course. The objective of this paper is to provide a basis for including some Bayesian econometrics in introductory econometrics courses at the level of Hill et al. (2018). It is assumed students have had a prior course in statistics that gives them some knowledge of probability and distributions. Matrix algebra is used sparingly; some examples will require further explanation if students do not have a matrix algebra background. Topics covered are (1) characteristics of the Bayesian approach that distinguish it from the frequentist approach, (2) the requirement for simulation when analytical approaches are inadequate, (3) posterior distributions of nonlinear functions of parameters, (4) the Metropolis algorithm, and (5) Gibbs sampling. Simple examples taken from Hill et al. (2018) are used to illustrate the various concepts; EViews code is provided for each of the examples. Such code will be particularly useful for courses that use EViews as their main software platform.

Chapter 9 Helmut Lütkepohl: A Software for Teaching Multivariate Time Series Analysis

In the early days of teaching multiple time series analysis, a menu-driven software was developed for use in the classroom and for examinations. The software and its follow-up version are briefly motivated and reviewed. A personal view of the author about future software tools for time series analysis is expressed.

Chapter 10 Katherine Hauck, Tiemen Woutersen: Explaining Ridge Regression and LASSO

Machine learning is a tool that uses a computer’s analytic power to make decisions and predictions from data. Two common machine learning techniques are Least Absolute Shrinkage and Selection Operator (LASSO) and Ridge regression. We provide intuition to identify cases in which a researcher may prefer these models to least squares. We discuss the application, implementation, and uses of LASSO and Ridge regression, relative to (i) each other and (ii) least squares, including splitting the data and the choice of tuning parameter. Further, we use an example to compare least squares, LASSO, and Ridge regression to demonstrate how machine learning techniques select the most important regressors for prediction analysis.

Chapter 11 Katherine Hauck, Tiemen Woutersen: Nonparametric Identification

Using a nonparametric model allows empirical researchers to more closely align their research design to economic theory and to obtain more robust results. We show the importance of nonparametric identification by contrasting it with parametric identification. In particular, we show the numerical instability of the estimated parameters in a model that is parametrically identified but fails to be nonparametrically identified, and we demonstrate this lack of identification with a proof.

Chapter 12 Thomas B. Fomby: Teaching Econometrics Students How to Model Bivariate Time Series Using Monte Carlo Data for the Purpose of Validating Leading Indicators

This paper presents a decision tree approach for teaching econometrics students how to properly model bivariate time series data sets for the purpose of determining if a target variable (Y) can be more accurately forecast using a supplementary (leading indicator) variable (X) in a multivariate model compared to forecasting Y using a univariate model, the determination coming from an out-of-sample forecasting experiment. Students are given Monte Carlo data sets and asked to choose one of four bivariate models on a training set portion of the data. The decision tree takes the student through various pre-test procedures that help the student choose a correct bivariate model. Thereafter, the student proceeds to cross-validate the worthiness of the supplementary variables vis-à-vis the chosen bivariate model. The construction of the Monte Carlo data sets and how they can be obtained from the author are discussed in the paper. In addition, an example cross-validation involving a target variable and two supplementary variables is discussed.

Chapter 13 James Davidson: Reflections on a Life of Teaching Econometrics

This chapter reviews the experience of teaching econometrics over a span of 44 years and at a range of institutions worldwide. The computing revolution gave rise to huge changes in lecturing style, in teaching methods generally, and not least in methods of textbook writing and typesetting. Major changes in the econometrics syllabus over the period reflected the corresponding progress of research. Notably, the same period has featured profound changes in the style of university administration.

Chapter 14 Jianghao Chu, Tae-Hwy Lee, Aman Ullah: Asymmetric AdaBoost for Maximum Score Estimation of High-Dimensional Binary Choice Regression Models

Carter Hill’s numerous contributions (books and articles) in econometrics stand out especially in pedagogy. An important aspect of his pedagogy is to integrate “theory and practice” of econometrics, as coined into the titles of his popular books. The new methodology we propose in this paper is consistent with these contributions of Carter Hill. In particular, we bring the maximum score regression of Manski (1975, 1985) to high dimension in theory and show that the “Asymmetric AdaBoost” provides the algorithmic implementation of the high-dimensional maximum score regression in practice. Recent advances in machine learning research have not only expanded the horizon of econometrics by providing new methods but also provided the algorithmic aspects of many of traditional econometrics methods. For example, Adaptive Boosting (AdaBoost) introduced by Freund and Schapire (1996) has gained enormous success in binary/discrete classification/prediction. In this paper, we introduce the “Asymmetric AdaBoost” and relate it to the maximum score regression in the algorithmic perspective. The Asymmetric AdaBoost solves high-dimensional binary classification/prediction problem with state-dependent loss functions. Asymmetric AdaBoost produces a nonparametric classifier via minimizing the “asymmetric exponential risk” which is a convex surrogate of the non-convex 0-1 risk. The convex risk function gives a huge computational advantage over non-convex risk functions of Manski (1975, 1985) especially when the data is high dimensional. The resulting nonparametric classifier is more robust than the parametric classifiers whose performance depends on the correct specification of the model. We show that the risk of the classifier that Asymmetric AdaBoost produces approaches the Bayes risk which is the infimum of risk that can be achieved by all classifiers. Monte Carlo experiments show that the Asymmetric AdaBoost performs better than the commonly used LASSO-regularized logistic regression when parametric assumption is violated and sample size is large. We apply the Asymmetric AdaBoost to predict business cycle turning points as in Ng (2014).

Chapter 15 Daniel J. Henderson, Christopher F. Parmeter: Using Replication to Teach and Understand Econometrics

This paper is written as a guide for instructors/students on how to assign/complete a successful replication paper for an econometrics course. It discusses the benefits of replication, the pitfalls that students may encounter, and suggestions for which papers to choose to replicate. A suggested timeline is included for such an assignment. The paper ends with a successful replication (including all tables, figures, and R code) of Mankiw et al. (1992).

Chapter 16 Walter Enders: Reflections of an Applied Econometrician

I came to econometrics through the back door. I had published a number of articles involving overlapping generations models and had no thoughts about being an applied econometrician...

Chapter 17 Zhongjian Lin, Esfandiar Massoumi: Imputing Missing Waves for Pseudo Panels: A Generalized Scoring and Matching Method

We identify statistical “matches” for missing individual observations in cross-section waves to construct or complete pseudo-panels. Unlike propensity score matching, our method is applicable even when a classifier outcome (e.g., treatment status) is not observed. In non-panel cross-sections, agents are assessed as similar relative to several observable characteristics, which are optimally aggregated. This is model-free, unlike “cohort”, “synthetic variable”, and other imputation methods. Observed covariates are employed as surrogates, not cohort averages or synthetic variables that must satisfy a given model. Our aggregate score is “information efficient”, utilizing all of the probability laws generating the observed variables. Applications include panel construction, network membership, treatment effects, and missing data. We examine private return to R&D in the presence of spillovers using macropanels, and female labor force participation using micropanel data (PSID).

Chapter 18 Jeffrey S. Racine: Quarto the Assassin

My interest in open source research tools has a long arc, and the tools I adopt evolve as new and improved approaches surface (Racine & Hyndman, 2002; Meredith & Racine, 2009; Racine, 2012). A recent development brings to mind overused superlatives like “game-changer”, among others, and I suspect you will agree “100%” that one recent development belongs in either the “Assassin” or “Saturn” category, or perhaps both (the latter being associated with consuming one’s progeny as prophecy had it that Saturn would be overthrown by one of his sons, so in response, he devoured them upon birth). In this chapter we examine how Quarto can be used to create dynamic reveal.js presentations that are simply unrivaled, anywhere, period (“100%”, a “game-changer”). As such, and particularly for those accustomed to the Beamer slide format, the adoption of these new tools may be of particular interest. This document, naturally, is authored in Quarto, and the Quarto script for this document and the reveal.js slides discussed below are available at https://github.com/JeffreyRacine/rch.

Chapter 19 Mohamad A. Khaled, Alicia N. Rambaldi, Christiern Rose: Teaching Statistical Learning in Econometrics

Statistical learning techniques are increasingly being used by practising economists. Some bring new tools while others have existed in different forms as part of the standard econometrics toolbox. The increasing use of those methods is gradually bringing into focus a dichotomy between an emphasis on prediction and algorithms for statistical learning on the one hand and a focus on causal inference and identification in applied econometrics on the other, based on the fact that many causal inference methods rely on a prediction stage (e.g., instrumental variables first stage). Students, both at the undergraduate and graduate levels, should complete their training with an understanding of statistical learning models and their interaction with econometrics and statistics.

Chapter 20 Eric Hillebrand: Teaching Mathematics for Economists (arXiv version, tex-file with tikz code)

In this chapter, I discuss teaching mathematical tools specifically tailored for economics students. A typical one-semester course in this area seeks to blend a range of topics: from foundational elements of subjects such as linear algebra and multivariate calculus to intermediate areas like real and convex analysis and further into advanced topics such as dynamic optimization in both continuous and discrete time. This breadth of coverage corresponds to material usually spread across multiple years in traditional mathematics programs. Given the comprehensive nature of these courses, careful selection of topics is essential, balancing numerous trade-offs. I discuss potential course sequences and instructional design choices. I then focus on conceptualizing and explaining mathematical modeling in economics. I reflect on three years of teaching an advanced undergraduate course in mathematical methods online. The latter part of the chapter offers examples and visualizations I have found particularly beneficial for imparting intuition to economics students. They cover a range of topics at different degrees of difficulty and are meant as a resource for instructors in Mathematics for Economists. Among these, I use the Ramsey model as a recurring example, especially relevant when designing a mathematical tools course with an orientation toward preparing students for macroeconomic analysis.