La Follette School of Public Affairs

PA 819 Syllabus: Quantitative Methods for Public Policy

Time and location:

Spring 2007, Monday and Wednesday, 9:30-10:45 AM

Social Science 5231

Instructor: 

Carolyn Heinrich, Associate Professor of Public Affairs

E-mail: cheinrich@lafollette.wisc.edu

Office phone: (608)262-5443 or (608)265-9377

Office location: La Follette School of Public Affairs and Social Sciences Bldg., Rm. 3438

Office hours: Monday and Wednesday, 11:00 AM –12:00 PM at SS 3438 and by appointment

Teaching assistant: Jonathan Hore, office hour TBA                        

Course description:

Quantitative Methods for Public Policy is the second course of a two-course sequence in quantitative analysis at the La Follette School of Public Affairs.  The course begins with a brief review of material covered in the pre-requisite course, Quantitative Tools for Public Policy Analysis (PA 818), in particular, some of the basic tools of statistical analysis that one uses to describe and interpret data. We will quickly advance to more complex methods of statistical analysis, as you will inevitably find that the complexities of human and institutional activities frequently demand a more sophisticated strategy for the production and analysis of information that can inform public policy decision-making and our understanding of policy impacts.

The amount of data collected from or about individuals, communities, and institutions has increased rapidly in recent years, and these data are increasingly being used to make major decisions regarding people's health, education, employment, environment and other aspects of social welfare. This course will emphasize the application of statistical concepts and methodologies to the analysis of public policy issues, including the limitations of quantitative methods in generating answers to policy and research questions. 

In a project that will span the duration of the course, you will take on the role of a researcher or policy analyst and choose a dataset (one among those posted on our course website, or with special permission, a dataset of  your choosing), and these data will become the basis for the empirical work you will do in this class.  You are encouraged to pair up with another student in class on this project.  In particular, you will:

1.     Formulate one or more policy or research questions to address

2.      Explore your data (using descriptive statistics) and identify limitations of your data, refining your research question(s) if needed

3.      Specify hypotheses that you will test empirically

4.      Identify statistical methods appropriate for your data and analysis

5.      Specify statistical models to test

6.      Conduct sensitivity analyses (if appropriate) of alternative model specifications

7.      Interpret the results of your statistical analyses in terms of the research questions and hypotheses you defined at the onset of the study

8.      Make a presentation to your peers of your study findings, including a discussion of your analytical approach

To undertake this work, it will be essential for you to become skilled in using a statistical processing program such as Stata or SAS.  You will have considerable support for your statistical programming/processing activities via sample programs, supporting documentation, in-class demonstrations, hands-on lab sessions, and a teaching assistant who will help you with the “mechanics” of using the statistical software in empirical analysis.  In addition, in the first 2 weeks of class, the discussion sections will be used for SAS and Stata training sessions set up specifically for this class in room 3218 of the Social Sciences building.  The course TA will also be available in discussion sections to help you with questions about the course material and data analysis. It will be up to you to take advantage of these resources for your project.

Datasets are posted on our website (NLSY, CPS and NSAF), ready to download and use.  I strongly encourage you to use one of these datasets, which are already cleaned and well-documented, and for which it will be easier for us to provide you with technical assistance.  We can’t be responsible for helping you with technical details of other datasets you might choose, e.g., how a particular variable was coded, etc..

Texts and materials:

The textbook for this course is Introduction to Econometrics, by James H. Stock and Mark W. Watson (Addison Wesley, 2007, second edition) and is available at Underground Textbooks.  There is also a small reading packet that has been prepared for you at Underground Textbooks, which includes readings required for this course.  In addition, you may be asked to read other journal articles or papers that will provide you with examples of empirical research that is well-designed, executed and communicated.  In fifteen weeks, you will not become an expert in all of the statistical methodologies we will study.  However, it is expected that you will develop a good understanding of how to design and undertake an empirical analysis in response to a policy or research question, a respectable level of skill in the application of multiple regression analysis, and some experience in one or more advanced statistical approaches.

Statistical programming resources:

Stata and SAS software are available for use in the La Follette School computer lab or the Social Sciences computing lab (SS 3218).  The following books are reasonably good for assistance in learning SAS and/or Stata:

Applied Statistics and the SAS Programming Language by Ronald P. Cody, et al, 2006. 

Handbook of Statistical Analyses Using SAS, Second Edition, Geoff Der, Brian S. Everitt, 2001.

Statistics with STATA, Version 8 or 9

These can be ordered from Amazon.com or your favorite book-shopping alternative.  There is not a bookstore order for these books. 

Other resources for Stata are the manuals that can be purchased with the software.  See below for the University’s pricing on these.

To purchase Stata 9 through SSCC, see: http://www.stata.com/order/new/edu/gradplans/gp-campus.html

Purchasing SAS through DOiT:

http://techstore.doit.wisc.edu/

Course requirements and evaluation:

Course grades will be based on the following:

Class participation and effort: 10%

Two assignments: 30%

Midterm exam: 30%

Final project: 30%

Class participation – attendance, punctuality, coming prepared to ask questions (i.e., assigned reading done BEFORE lecture) – is essential for this course.  The amount of effort you put into preparation and study for this course will be the most important variable predicting how well you will do and what you will gain!  You are also expected to attend the discussion sections and to work on problems or assignments as assigned by the TA (although this additional work will not be graded).  

The two assignments are intended to help you apply the statistical concepts and methods that we are learning in class in ways that will advance your work on the final project.  You will need to use your statistical programming skills (Stata, SAS or another program) to complete these assignments.

• Assignment #1 due March 5:  Description of data, research/policy questions and hypotheses, limitations of data for statistical modeling and hypothesis testing, proposed model specification and approach to analysis

• Assignment #2 due April 11:  Preliminary results and their interpretation – how well does my model perform?  Plans for refining models – am I answering the question(s) I set out to answer?  What can I change (and what am I unable to change) to improve my analysis?

Your response to each of these assignments should be approximately 5 pages, including attachments.

The mid-term exam will be on March 14 and will cover all material up to the March 12 review session.

Final projects will be due no later than May 14 (early submissions welcome!)

COURSE OUTLINE:

Week 1

January 22/24:  Course introduction, getting started on course projects, review of PA 818 concepts,  measurement, formulating hypotheses for tests of causal relationships           

Stock and Watson, Chapter 1 (review Chapters 2 and 3 as necessary)

Wooldridge, Chapter 19, "Carrying Out an Empirical Project" [packet]

W. James Bradley and Kurt C. Schaefer, Chapter 6, “Limitations of Measurement in the Social Sciences,” Chapter 7, “Information for Inferences,” Chapter 8, "Causality" in The Uses and Misuses of Data and Models, Sage Publications, 1998. [packet]

January 25 discussion sections: SAS/Stata training session in Social Sciences room 3218

Week 2              

January 29/31:   Quantitative tools review, assumptions and properties of OLS 

   Stock and Watson, Chapters 4 and 5                  

February 1 discussion sections: SAS/ Stata training session in Social Sciences room 3218 

Week 3

February 5/7: Multiple regression and model specification

                    Stock and Watson, Chapters 6 and 7

                    Kennedy, Chapter 5 “Specification” and Chapter 11 on multicollinearity [packet]

Week 4

February 12/14:  Nonlinear regression functions 

                       Stock and Watson, Chapter  8, "Nonlinear Regression Functions"         

Week 5

February 19/21:

                       Multiple regression with qualitative information (binary/limited dependent variables)

Stock and Watson, Chapter 11, "Regression with a Binary Dependent Variable"

Week 6

February 26/28:  Threats to internal and external validity

                       Stock and Watson, Chapter 9, "Assessing Studies Based on Multiple Regression"

Week 7

March 5/March 7:   More on model specification, missing data problems, specification tests

                   Continuing material from previous chapters with applications in SAS/Stata 

                   Allison, Paul, Missing Data

                       Assignment #1 due March 5:  Description of data, research/policy questions and hypotheses, limitations of data for statistical modeling and hypothesis testing, proposed model specification and approach to analysis    

Week 8  

March 12:  Midterm Exam Review

March 14:  Midterm Exam

Week 9

March 19/21:  More on model specification, regression with panel data