Political Science 271b

Quantitative Methods

Thad Kousser                                                                                                                      Fall 2004

Where to Find the Readings

› I’ve assigned a number of different treatments of each subject in the hope that you will find the one that works best for you, and know where to find the others.  So you should not think of it all as required reading, and you certainly will not be quizzed on it, but you should probably try to put your eyes on all of it at some point. 

› William H. Greene’s Econometric Analysis is the standard reference for economists and political scientists, and the fifth edition is available at the bookstore.  It is expensive, but worth it.  If you already own an earlier edition, let me know which one, and I’ll try to translate the assignments.  Michigan Press’ reprint of Gary King’s Unifying Political Methodology is also available at the bookstore, and should be identical to the original Cambridge version.

› Applied articles are available for download through JSTOR, and I’ll post the ones that are not.

› We will use Stata 8.0 in class and in the problem sets.

› The course webpage, located at http://weber.ucsd.edu/~tkousser/PS271b.htm, will contain information such as my lecture notes, special guest star (mostly Neal Beck) lecture notes, links to STATA user guides, and other hopefully helpful information.

Course Assignments

› 60% from four problem sets that mix a little math and substance with a lot of data

    Poli 271b Problem Set 1   Recall Exit Poll

    Zipped Edited Foreign Data

    Poli 271b Problem Set 2 Fall 2004

    Girlie Men   new rules and innovation dataset

    legislative leadership, duration format   leaderbeck

    Poli 271b Problem Set 3 Fall 2004

    Civil War Duration Dataset

    1987 Assembly Votes

    Monetary7.0   Poli271Medicaid7.0

    Congress Time Series

› 40% for a 10-15 page analysis of whatever data you have.  It is due Friday, Dec. 10th.

            The goal of this course is to make you sophisticated shoppers in the mall of quantitative methods.  You may not be able to sew your own clothes or recite every ingredient of a hotdog-on-a-stick, and we’ll stay away from most of today’s fads, but you should become intelligent consumers who are familiar with a broad range of approaches and models and know how to buy the one that is perfect for you.  We’ll talk about why the old hand me down of OLS might not fit your data anymore.  Then the emphasis will be on matching substantive questions to appropriate methods, and trying a few on for size.  I will expose you to some of the mathematical ideas and notations that drive different approaches, but I’ll focus on training you as researchers rather than econometricians. 

            I will assume that your math skills are as rusty as mine, but that you are willing to sharpen them.  The first part of the course, the survey of maximum likelihood models, builds on some basics of probability theory and uses high school-level calculus.  The second part will involve learning the basics of linear algebra, which I’ll teach you.  None of the math here should be a barrier to entry, so please let know at any point if it becomes an obstacle.  If you have taken 204b, you should be sufficiently prepared.    

            My goal is to alternate lectures that introduce a new method with ones that apply the method, using real data and Stata.  The balance may not remain perfect, but each set of lectures will attempt to answer the following questions:  How does the structure of this dataset differ from classical regression assumptions?  What kinds of substantive issues would lead to this sort of data generating process? How can we determine what’s going on with our data?  What is the theory behind an alternative approach?  How can we implement the alternative approach?  What is the clearest and most informative way to interpret and present our results?

Course Outline

Part I. More Maximum Likelihood

Thursday, September 23^rd. Course Introduction and Stretching of Math Muscles.         

Tuesday, September 28^th

i.                    Unifying Political Methodology. Read Chapters 1 and 2.

ii.                  Neal Beck’s Likelihood notes.

iii.        September 28 The Logic of Likelihood 

Thursday, September 30^th. Linear and Dichotomous Variable Models in ML Framework

i.                    Unifying Political Methodology, skim chapter 3, read Chapter 4..

ii.                  Greene, Chapters 17 and 21.1-21.6, and Appendix E.6.

iii.                September 30 Constructing a Likelihood Function

Tuesday, October 5^th. Interpreting the Results of Maximum Likelihood Estimations

i.                    Unifying Political Methodology, Chapter 5.1-5.2.

ii.                  King, Tomz and Wittenberg. 2000. Making the Most of Statistical Analysis. AJPS 44:341-355 and the documentation for CLARIFY.

iii.        October 5 What Good is a Likelihood Function?

iv.        October 5 Lab Output  mylogit1_lf.ado myprobit_lf.ado  mynormal1_lf.ado

Thursday, October 7^th. Multiple Category Models: Multinomial Logit and Ordered Probit

i.                    Unifying Political Methodology, Chapter 5.3-5.6.

ii.                  Greene, Chapter 21.7-21.8.

iii.                Neal Beck’s Limited Dependent Variables notes.

iv.                October 7 Interactions and Ordered Categorical Variables

Tuesday, October 12^th. Applications of Multiple Category Models

i.                     FOCUS ON THIS! Alvarez, R. Michael, and Jonathan Nagler. 1998. When Politics and Models Collide: Estimating Models of Multiparty Elections. American Journal of Political Science 42:55-96.

ii.                 Alvarez, R. Michael, and Jonathan Nagler. 2000. Issues, Economics, and the Dynamics of Multi-Party Elections. APSR 94:131-49.

iii.                Quinn, Martin, and Whitford. 1999. Voter Choice in Multi-Party Democracies. AJPS 43:1231-1247.

iv.                Nagler. 1994. Scobit. AJPS 38:230-255.

v.         Jonathan Nagler’s Discrete Choice Models notes.

vi.        October 12, Unordered Choice Models

Thursday, October 14^th. Censored and Truncated Data

i.                    Greene, Chapter 22.1-22.4.

ii.                  Unifying Political Methodology, Ch. 9.

iii.                Neal Beck’s Censored Data notes.

iv.        October 14, Nonrandom Selection Models

Tuesday, October 19^th.  Event Count Models

i.                    Greene, Chapter 21.9.

ii.                  Unifying Political Methodology, Chapter 5.7-5.10.

iii.        October 19 The Poisson Event Count Model

iv.        October 19 Lab, Censored Data and Event Counts

v.         Courtesy of Kristian, here are some cites to look at to determine when event count models or OLS are appropriate:

Dixon, William J., Edward N. Muller, and Mitchell A. Seligson. 1993.  "Response [to T.Y. Wang's Inequality and Political Violence Revisited]."  American Political Science Review 87:983-993.

Thursday, October 21^st. Event History I: Survival Functions, Hazard Rates, and Models

i.                    Greene, Chapter 22.5.

ii.                  Kousser, “The Stability of Leadership: How Long do First Among Equals Last?” pages 3-8, 18-19, 33-39.

iii.              October 21 Interpreting and Extending the Poisson Model

Tuesday, October 26^th.  Event History II: Models with Duration Data.

            i.          Neal Beck’s Duration Data notes.

            ii.         October 26 Modeling Event History via Duration Data

            iii.        October 26 Event History Lab

Thursday, October 28. Event History III: Applications of Models

i.                    King, Gary, James E. Alt, Nancy Elizabeth Burns, and Michael Laver. 1990. A Unified Model of Cabinet Dissolution in Parliamentary Democracies. American Journal of Political Science 34:846-71.

ii.                  Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. Taking Time Seriously: Time-Series—Cross-Section Analysis with a Binary Dependent Variable. American Journal of Political Science 42:1260-1288.

iii.                Neal Beck’s Event History as Time Series notes.

iv.        October 28 Event History as a Binary Time Series Cross Section Analysis

Part II. Regression in the Language of Linear Algebra

Tuesday, November 2. Linear Algebra 101

i.                    Greene, Appendix A.1-A.3.

ii.                  Neal Beck’s Linear Algebra notes.

iii.        November 2 Linear Algebra 101

Thursday, November 4. The Classical Regression Model Restated

i.                    Greene, Chapters 2 and 3.

ii.                  Neal Beck’s OLS in Matrix Form notes.

iii.        November 4 Restating the Classical Regression Model with Matrices

Tuesday, November 9. Heteroskedasticity Revisited: Tests and Ways to Leverage Information

i.                    Greene, Chapter 11.

ii.                  Alvarez and Brehm. 1995. American Ambivalence Towards Abortion Policy: Development of a Heteroskedastic Probit Model of Competing Values. AJPS 39:1055-1082.

iii.        November 9 Heteroskedasticity

iv.        November 9 Lab    mynormal1_lf.ado     ols1.do    ols2.do

Tuesday, November 16. Instrumental Variables and Two-State Least Squares

i.                    Unifying Political Methodology, Chapter 8.1-8.2.

ii.                  Greene, Chapter 15.1-15.5.

iii.                Jacobson, Gary. 1990. The Effects of Campaign Spending in House Elections: New Evidence for Old Arguments.” AJPS 34:334-62.

iv.        November 16 Instrumental Variable Models

Thursday, November 18. Time Series I: Autoregression and Moving Average

i.                    Greene, Chapter 12.

ii.                  Quinn and Jacobson. 1989. Industrial Policy Through Restrictions on Capital Flows. AJPS 33:700-737.

iii.        November 18 Autocorrelation and Time Series Data I

Tuesday, November 23. Time Series II: Stationarity, Nonstationarity, Cointegration.

i.                    Greene, Chapter 20.

ii.                  Neal Beck’s Time Series notes.

iii.                Neal Beck. 1994. The Time Series Method of Cointegration. Political Analysis 4:237-248.

iv.        November 23 Time Series II Autoregression and Moving Average

v.         November 23 Lab Time Series Processes

Tuesday, November 30.  Time Series, Cross Sectional Data: Substantive Assumptions

i.                    Greene, Chapter 13.

ii.                  Neal Beck’s Longitudinal Data and Panel Data notes.

iii.        November 30 Time Series Cross Sectional Analysis

Thursday, December 2. Time Series, Cross Sectional Data: Fixed Effects and Random Effects.

            i.          Chris Zorn. 2001. Estimating Between and Within Cluster Covariate Effects, with an Application to                     Models of International Disputes. International Interactions 27:433-445. 

ii.                  Neal Beck and Jonathan Katz. 1995. What to Do (and What Not to Do) With Time Series, Cross Section Data. APSR 89:634-47.

iii.        December 2 Fixed and Random Effects