Christopher Adolph is Associate Professor of Political Science and Adjunct Associate Professor of Statistics at the University of Washington, Seattle, where he is also a core faculty member of the Center for Statistics and the Social Sciences. His book, Bankers, Bureaucrats, and Central Bank Politics: The Myth of Neutrality (Cambridge University Press), won the International Political Science Association’s Charles H. Levine Prize for best contribution in comparative public policy and administration. He is a former Robert Wood Johnson Scholar in Health Policy Research and won the American Political Science Association’s Mancur Olson Award for best dissertation in political economy. His research on comparative political economy and quantitative methods has appeared in American Political Science Review, Political Analysis, Social Science & Medicine, and other academic journals.
Course Content This course provides a survey of regression models for time series (TS) and time series cross-section (TSCS) data, with an emphasis on modeling dynamics and panel structures. After a review of the theory and estimation of linear regression and maximum likelihood, we will cover the following topics: modeling time series dynamics using ARIMA models, lagged dependent variables, and distributed lags; cointegration and error correction models; modeling cross-sectional variation using fixed and random effects; coping with panel heteroskedasticity, and presentation and interpretation of TS and TSCS models. Time permitting, we will cover advanced topics based on student interest, which in past years have included TSCS models for binary, ordered, or categorical data, multiple imputation for missing data in panel datasets, and linkages between panel data and hierarchical linear models.
Course Objectives For many political science subfields, including political economy, international relations, and comparative politics, panel data are ubiquitous, and training in TSCS analysis essential for quantitative research. Participants will gain an introductory understanding of the theory behind TSCS models and a working understanding of how to estimate, select, and interpret these models. Emphasis is placed on the development of both conceptual understanding and the ability to apply tools learned in the class using a variety of packages available for the R statistical language.
Course Prerequisites Students should enter the course with a solid understanding of first year statistics as taught in a standard political science doctoral program, an interest in data with either a time series or time series cross-sectional (panel) data structure, and either exposure to, or willingness to try, the R statistical package. Specifically, students should be familiar with basic data structures and computing, elementary matrix algebra, and the basic theory and application of the linear regression model.
Required texts All Readings will be provided; however, students unfamiliar with R will benefit from reading this book ahead of the class: Alain F. Zuur, Elena N. Ieno, and Erik H.W.G. Meesters. 2009. A Beginner’s Guide to R. Springer-Verlag.
Background knowledge required Statistics OLS = m
Computer Background R = e (or willingness to quickly learn)
e = elementary, m = moderate, s = strong
Goals This course provides a survey of regression models for time series (TS; also called longitudinal)and time series cross-section (TSCS; also called panel) data, with an emphasis on modelling dynamics and panel structures. Because panel data are ubiquitous in many political science subfields, including political economy, comparative politics, and international relations, training in TSCS analysis is essential preparation for performing and understanding quantitative research. Participants will gain an introductory understanding of the theory behind TSCS models and a working understanding of how to estimate, select, and interpret these models.
Overview After a review of the theory and estimation of linear regression, we will cover the following topics: diagnosing dynamic behavior in time series; modeling time series dynamics using arima models, lagged dependent variables, and distributed lags; cointegration and error correction models; modeling cross-sectional variation using fixed and random effects; appropriate techniques for fixed effects models used on data with many time periods available versus those used with few time periods; coping with panel heteroskedasticity; and presentation and interpretation of ts and tscs models, especially using dynamic simulation techniques. Time permitting, we will cover advanced topics based on student interest, which in past years have included multiple imputation for missing data in panel datasets, models for binary tscs data, and linkages between panel data and more general approaches to hierarchical linear models.
Prerequisites Students should enter the course with a solid understanding of first year statistics as taught in a standard political science doctoral program, an interest in data with either a time series or time series cross-sectional (panel) data structure, and either exposure to, or willingness to learn, the R statistical package. R. Extensive in-class code examples will use the R statistical package, which is powerful, free, open source, widely used, and rapidly becoming the standard for quantitative work in political science and other fields. You can obtain R at http://www.r-project.org/. Throughout the course, I will provide example code in R, and can only promise detailed homework help for R, not other statistics packages.
Course Readings Selections from the following books and articles will be provided as part of the course materials. Students seeking to get a head start should focus on the assigned readings for the first week; those without a background in R should especially concentrate on Zuur.
Nathaniel Beck and Jonathan N. Katz. 1995. “What to Do (And Not to Do) With Time Series Cross-Section Data.” American Political Science Review 89(3): 634-647 .
Nathaniel Beck and Jonathan N. Katz. 2011 “Modelling dynamics in Time-Series–Cross-Section political economy data.” Annual Review of Political Science 14:331-52.
Nathaniel Beck, Jonathan N. Katz, and Richard Tucker. 1998 “Taking time seriously:Time-Series–Cross-Section analysis with a binary dependent variable.” American Journal of PoliticalScience 42(4) 1260-1288.
Janet M. Box-Steffensmeier, John R. Freeman, Matthew P. Hitt, and Jon C.W. Pevehouse. 2014. Time Series Analysis for the Social Science. Cambridge University Press.
Paul S.P. Cowpertwait and Andrew V. Metcalfe. 2009. Introductory Time Series with R. Springer-Verlag.
Andrew Gelman and Jennifer Hill. 2007. Data analysis using regression and multilevel/hierarchical models. University of Cambridge Press.
James Honaker and Gary King. 2010. “What to do about missing values in Time-SeriesCross-Section data.” American Journal of Political Science 54(2): 561-581.
Gary King, Michael Tomz, and Jason Wittenberg. 2000. “Making the Most of Statistical Analyses.” American Journal of Political Science 44(2): 347-361
Giovanni Millo. 2014. “Robust standard error estimators for panel models: a unifying approach.” MPRA Paper No. 54954.
Bernhard Pfaff. 2008. Analysis of Integrated Series with R. Springer-Verlag.
David Roodman. 2009. “How to do xtabond2: An introduction to difference and system GMM in Stata.” The Stata Journal. 9(I): 86-136.
Jeffrey M. Wooldridge. 2009. Econometric Analysis of Cross-Sectional and Panel Data. MIT Press. 2nd Edition.
Alain F. Zuur, Elena N. Ieno, and Erik H.W.G. Meesters. 2009. A Beginner’s Guide to R. Springer-Verlag.
Course outline Readings assigned for a particular day should be read in advance of class. Students seeking feedback on homework should turn in the listed problems each day. It is strongly recommend that students new to R work through the example code presented in Zuur as they read.
Week 1 Day 1 Introduction, Review of the Linear Model and its Properties; R Basics Readings:On R: Zuur, Ch. 1, 2, and 3 On linear regression: Woolridge, Ch. 4 On matrix algebra (opt.): Matrix Handout (Kevin Quinn)
Day 2 Estimation and Simulation; Linear Regression in R Readings: On R: Zuur, Ch. 4 and 5 On simulation: King, Tomz, and Wittenberg 2000 Homework: Problem 1
Day 3 Basic Concepts for Time Series: Trends, Lags, and Cycles Readings: On time series: Box-Steffensmeier et al, Ch. 1, 2 Alternate reading: Cowpertwait & Metcalf, Ch 1.1, 1.4, 1.6, 2.1-2.5 On R: Zuur, Ch. 6 Homework: Problems 2 and 3
Day 4 Modelling Stationary Time Series Readings: On time series: Box-Steffensmeier et al, Ch. 3 Alternate reading: Cowpertwait & Metcalf, Ch. 4, Ch. 5.1-5.4, 5.9-5.11 Ch. 6 Homework: Problem 4 or Bonus Problem A
Day 5 Modelling Nonstationary Time Series Readings: On time series: Box-Steffensmeier et al, Ch. 5. 6 Alternate reading: Cowpertwait & Metcalf, Ch. 7 On cointegration: Pfaff, Ch. 4 Homework: Problems 5 and 6
Day 6 Basic Concepts for Panel Data Readings: On random effects: Woolridge, Ch. 10.1-10.4 On hierarchical models: Gelman and Hill, Ch. 11, 12, 13 Homework: Problem 7
Day 7 Panel Data Models with Many Time Periods Readings: On fixed effects: Woolridge, Ch. 10.5-10.7 On panels with large-T: Beck & Katz 2011 Homework: Begin work on Problem 8
Day 8 Panel Data Models with Few Time Periods Readings: On panels with small-T: Roodman 2009(skip code examples) Homework: Continue work on Problem 8
Day 9 Heteroskedasticity in Panel Data / In-Sample Simulation for Panel Data Models Readings: On panel-corrected standard errors: Beck & Katz 1995 On heteroskedasticity and autocorrelation correction: Millo 2014 Homework: Complete Problem 8
Day 10 Course Wrap-up / Advanced Topics Possible topics: Multiple imputation; Binary outcomes; Course Q&A;
Student research presentations Readings: On multiple imputations for TSCS Honaker and King 2010 On binary outcomes in TSCS: Beck, Katz & Tucker 1998 Homework: No additional homework