Statistical Methods for Public Policy and Administration

Spring 2008
Lecture (PubP&Adm 608)
Monday and Wednesday, 4:00-5:30 PM
Herter Hall room 202
Lab (PubP&Adm 697AA)
Wednesday, 6:00-7:30 PM
Open Source Lab (First Floor Thompson Hall)

Instructor Michael Ash Teaching Assistant Dave Celata
814 Thompson Hall Thompson Low-Rise 20
Office hours
Tuesday, 9:30-10:30 AM
Wednesday, 2:30-3:30 PM
Thursday, 10:30-11:30 AM
Or by appointment
Office hours
By arrangement
Telephone: 413-545-6329 Email: celata@gmail.com
Fax: 413-545-2921
Email: mash@pubpol.umass.edu
UMass Spark login: http://spark.oit.umass.edu
Personal web page: http://people.umass.edu/maash
CPPA web page: http://www.masspolicy.org

Objectives

Statistics and other quantitative methods constitute a key device by which public decision-makers and policy advocates understand reality, communicate their understanding, and campaign to change it.

This course will introduce statistical methods and modelling for policy analysis and public administration. Students will learn how to apply statistical methods to interesting questions in public affairs. They will also develop the capacity to identify and critique good and bad statistical practice.

Content and Methods

The effective use of statistical methods in policy and administration demands persuasive writing to introduce quantitative analysis and to present quantitative results. Students will learn how to operationalize variables and to write up results in readable prose that will convince educated lay readers, decision-makers, and social scientists.

We will apply the standard methods of descriptive and inferential statistics. Students will learn to describe and test the distribution of variables with both quantitative and graphical methods. They will also learn to model and test association between variables.

The course emphasizes research design as the basis for plausible claims of causal relationships between variables. By the end of the course, students should understand the conditions that make a plausible case for an association to be considered causal.

Comfort with quantitative methods also implies the capacity to critique their misuse. Students will learn to criticize both poorly elaborated models and overstated claims of causality based on statistical assocation.

The course includes an extensive lab component. Students will gain comfort with statistical methods in both theory and practice. The course will also introduce students to the use of graphics as an effective tool for communicating quantitative information.

Teaching Assistant

Dave Celata, a second-year student in the MPPA program, will be the teaching assistant for the course. Dave will lead the lab section of the course. He will also hold office hours to discuss course material and problem sets with you.

Grading

There are three graded portions of the course: problem sets; class participation and an in-class presentation; and a final exam.

The problem sets are heavily weighted in the grade because they most closely reflect the type of thinking you will do as practitioners. Problems indicated as ``Exercises'' are at the end of each chapter in the textbook. Problems indicated as ``Empirical Exercises'' can be found under ``Exercises and Empirical Projects'' in the Student Resources section of the website for the textbook http://wps.aw.com/aw_stock_ie_2/. Problems that ask you to replicate and interpret tables or figures in the text require data from ``Data Sets,'' again in the Student Resources section.

For the in-class presentation, each student will present either a good use of statistics (in late March) or a bad use of statistics (in late April) in the literature on policy or administration. Each presentation will be 10 minutes and must describe a statistical analysis. In the course, students will develop the criteria for judgment about good or bad use.

The final exam will be cumulative but weighted towards the second half of the semester.

Points
Problem sets 50
Presentation and participation 15
Final examination
35
Total 100

Grades will be assigned according to the following schedule. Please note that your grade depends on a fixed standard of comprehension and expression, not comparisons to other students. Therefore, you should feel comfortable discussing and sharing your notes and ideas with your fellow students as well as collaborating on problem sets.

Point Cut-Off Grade
92 A
90 A-
88 B+
80 B
78 B-
76 C+
72 C
70 C-
68 D+
50 D
F

The required textbook for the course is:

Despite the misleading name, econometrics is widely applied in sociology, political science, and policy and administration, as well as in economics. The textbook is organized around four interesting questions in public policy.

Math skills and supplementary books

The main analytic skills that you need for this course are (1) putting things in categories; and (2) computing averages.

I recommend a review of basic math skills: how to graph y against x and how to determine and interpret the slope of a line; how to do unit analysis, e.g., 8 hours x 5.15 dollars/hour = 41.20 dollars; and how to compute a percent change, e.g., 1,600 gallons decreasing to 1,400 gallons is a -12.5 percent change.

If you would like to review basic math skills, you might want to browse chapters 2, 3, 4, and 5 of Jeffrey O. Bennett and William L. Briggs, Using and Understanding Mathematics: A Quantitative Reasoning Approach (Addison Wesley)

If you would like more background in statistics to supplement topics in the first weeks of the course, the following textbooks may be useful

Software

Getting started with Stata.

The supported and recommended (although not required) software for the course is Stata 10. The CPPA computer lab, the Open Source Lab, and the computer lab in Dubois 720 have Stata for student use.

The optional lab course will offer instruction and practice on the use of Stata and other statistical applications. Much of your success in using statistical software will depend on collaboration with each other and posing questions to me and the TA.

You may purchase a copy of Stata for your PC or Mac at an educational price. Place the order on-line at http://www.stata.com/order/new/edu/gradplan.html, and pick up your copy at the UMass Bookstore.

Sources of help within Stata

Internet Help

Logs and Scripts (Please print and read)

Working with Spreadsheets

You may also use spreadsheet applications. (Excel is available in the CPPA cluster and many other labs on campus, many computers come with a spreadsheet, and the free, downloadable OpenOffice.org office suite includes a high-quality spreadsheet program.)

Spreadsheet skills are an invaluable skill for both policy analysts and public managers. In addition to budgeting, data management, and graphic displays of quantitative information, modern spreadsheets offer an impressive array of statistical tools. Because I assume that you have already learned how to use spreadsheets elsewhere, I will offer less specific instruction in class on the use of spreadsheets than of the statistics application, but I am always willing to meet with students during office hours to work with spreadsheets.

Other

You must receive email from the pubpadm-608-01-spr08@courses.umass.edu mailing list. If you are enrolled, then you are automatically subscribed to the list.

Schedule

Please complete reading assignments and problem sets before class on the day indicated.
SessionTopics and Assignments
Introduction Introduction to Statistics
Goals of the course
Game in a Box
Questions statistics can and cannot answer
Causality and research design
Counterfactuals
Forecasting
Math exercise: percent, percentages, and natual logarithm
Writing-up
Lecture 1 Introduction to Statistics
Operationalizing variables: outcomes, causal factors, and conditioning variables
The shape of data: observations and variables
Types of variables
Using data to answer questions
Source of variation
Notation
  • Chapter 1: Economic Questions and Data
  • Prepare for in-class discussion: propose a research question in policy and administration and discuss how you might answer it with data.
Lecture 2 Probability
Outcomes and probabilities
Expected value, mean, and variance
Other measures of central tendency, the median and mode
Graphing data: histograms
  • Chapter 2: Review of Probability, 2.1-2.2
Lecture 3 Probability
Joint distribution and correlation
  • Chapter 2: Review of Probability, 2.3
  • NOTE: Monday 13 February is the last day to drop without record.
Lecture 4 Probability
Common and important distributions
Random sampling
Distribution of the sample average
  • Chapter 2: Review of Probability, 2.4-2.6
Lecture 5 Statistics
Population mean, sample mean
Estimators, bias, and efficiency
Hypothesis testing
Statistical significance and practical significance
Lecture 6 Statistics
  • Chapter 3: Review of Statistics, 3.4-3.5
Lecture 7 Statistics
Relationships between variables
Graphing data: scatterplots
  • Chapter 3: Review of Statistics, 3.6
Lecture 8 Linear Regression with One Regressor
A regression model
Estimating regression coefficients
Graphing data: scatterplots
  • Chapter 4: Linear Regression with One Regressor, 4.1-4.2
Lecture 9 Linear Regression with One Regressor
Least Squares Assumptions
Sampling Distributions of Estimators
Lecture 10 Linear Regression with One Regressor
Inference about regression coefficients
Hypothesis Testing
Confidence Intervals
  • Chapter 4: Linear Regression with One Regressor, 4.5-4.6
Lecture 11 Linear Regression with One Regressor
Regression When X is a Binary Variable
Regression When X is a Categorical Variable
  • Chapter 5: Linear Regression with One Regressor, 5.1-5.3
Lecture 12 Linear Regression with One Regressor
Goodness of Fit
  • Chapter 5: Linear Regression with One Regressor, 5.4 and 5.7 (5.5 and 5.6 optional optional)
Spring Recess No meeting
Lecture 13 Linear Regression with Multiple Regressors
Omitted Variable Bias
  • Problem Set 3 Exercises 4.1, 5.2; Empirical Exercises on College Distance E4.3, E5.3 ; replicate and interpret Figure 4.3
  • Chapter 6: Linear Regression with Multiple Regressors, 6.1
Presentations Student Presentations: Good use of statistics
Linear Regression with Multiple Regressors
  • Chapter 6: Linear Regression with Multiple Regressors, 6.2-6.3
Lecture 14 Student Presentations: Good use of statistics
Linear Regression with Multiple Regressors
Fully-Specified Models
  • Chapter 6: Linear Regression with Multiple Regressors, 6.4-6.8
Lecture 15 Linear Regression with Multiple Regressors
Estimating and testing regression coefficients
Joint Hypotheses and Measures of Fit
Standard Error of the Regression
  • Chapter 7: Linear Regression with Multiple Regressors, 7.1-7.2, 7.5-7.7 (7.3 and 7.4 optional)
Lecture 16 Internal and External Validity in Regression Analysis
External Validity
Internal Validity
  • Chapter 9: Assessing Studies Based on Multiple Regression, 9.1-9.2
Lecture 17 Internal and External Validity in Regression Analysis
  • Problem Set 4 Exercises 6.2, 6.4, 7.2, 7.4, 7.6 ; Empirical Exercises on College Distance E6.2, E7.3 ; replicate and interpret Columns (1) and (5) of Table 7.2
  • Chapter 9: Assessing Studies Based on Multiple Regression, 9.3-9.5
Lecture 18 Some Notes on Data Management
Lecture 19 Experiments and Quasi-Experiments
  • Chapter 13: Experiments and Quasi-Experiments, 13.1-13.3
Lecture 20 Experiments and Quasi-Experiments
  • Chapter 13: Experiments and Quasi-Experiments, 13.4-13.8 (13.7 optional)
Presentations Student Presentations: Bad use of statistics
Limited Dependent Variables
  • Chapter 11: Regression with a Binary Dependent Variable 11.1-11.3
  • Problem Set 5 Exercises 9.1, 9.3; Empirical Exercise on College Distance E9.3; replicate and interpret Column (1) and (2) of Table 9.2
Lecture 21
Limited Dependent Variables
Binary Outcomes
  • Chapter 11: Regression with a Binary Dependent Variable 11.4-11.5, Appendix 11.3
Lecture 22 Limited Dependent Variables
The Linear Probability Model
Index models, logit and probit
Other limited dependent variable models
Stock and Watson Stata script for HMDA data
Lecture 23 Panel Data
Multiple observations
  • Chapter 10: Regression with Panel Data, 10.1-10.3
Lecture 24 Panel Data
Fixed Effect Models
  • Problem Set 6 Exercises 11.6, 11.7; Empirical Exercises on Smoking Bans E11.1, E11.2.
  • Chapter 10: Regression with Panel Data, 10.4-10.7
Lecture 25 Closing remarks
  • Problem Set 7 Exercises 10.1 and 10.3; Empirical Exercise on Seat Belt Laws, E10.2.


Copyright 2004-2008 by Michael Ash. All rights reserved.