January 27 – February 14, 2014
University of São Paulo
São Paulo, Brazil
The goal of this program was to provide scholars of the social sciences with access to high-quality, cutting edge, advanced training in qualitative and quantitative social science methods. The target group was high-potential, upper-level scholars of political science, international relations, and related disciplines. Preference was given to current faculty members, post-doctoral and doctoral students. Truly exceptional masters’ students in political science, international relations, and closely related fields were also considered.
Academic objectives:
This program aimed to provide basic training in three general areas:
- quantitative data analysis,
- qualitative data analysis, and
- research design and methods.
It is hoped that participants were able to return to their home institutions with greatly enhanced methodological skills, having built bridges with peers from other institutions and nations, and with the potential to disseminate these skills effectively among their peers.
Institutional Partners:
The Summer School is a partnership between the International Political Science Association (IPSA), the Department of Political Science (DCP) and the Institute for International Relations (IRI) at the University of São Paulo.
IPSA-USP 2014 Course OfferingsMathematics for Social Scientists
Prof. Glauco Peres da Silva, University of São Paulo
Refresher Course in Statistics
Prof. Lorena Barberia, University of São Paulo
Predicting Elections
Prof. Clifford Young, IPSOS-Washington
The Philosophy and Methodology of the Social Sciences
Prof. Patrick Jackson, American University
Case Study Methodology
Prof. Derek Beach, Universtity of Aarhus, Denmark
Comparative Research Designs and Comparative Methods
Prof. Dirk Berg-Schlosser, Philipps University Marburg, Germany
Introduction to Network Analysis using Pajek
Prof. Vladmir Batagelj, University of Ljubljana
Mixed Methods
Prof. Max Bergman, University of Basel
Multi-Level Analysis
Prof. Iñaki Sagarzazu, Nuffield College, University of Oxford, UK
Quantitative Methods for Public Policy Analysis
Prof. Bruno Cautrès, Sciences Po, Paris
Time Series Analysis and Pooled Time Series Analyses
Prof. Lorena Barberia, University of São Paulo
Prof. Guy Whitten, Texas A&M University, USA
IPSA-USP 2014 Summer School Course Descriptions
Mathematics for Social Scientists
Glauco Peres da Silva, University of São Paulo and FECAP (Brazil).
OUTLINE
This course is designed for students who require basic training in mathematical concepts, which are essential for understanding formal and quantitative analysis in political science research. The course will take place in the week preceding the commencement of all other courses in the Summer School. It will cover topics such as mathematical notation, basic set theory, various number systems, the algebra of numbers, the notion of a function, several important classes of functions, and solutions to systems of linear equations. The course will also provide an introduction to differential and integral calculus and applied matrix algebra. To complement lectures, students will apply the concepts taught in lectures to analyze problems using the mathematical and software packages commonly used in quantitative social science research including Excel and Stata, as well as complete exercises.
DATES
This course runs January 27-31, 2014.
TEACHING FELLOW: Guilherme Jardim Duarte
DETAILED DESCRIPTION
This course will aim to review fundamental mathematics principles and skills that are needed for students wishing to study game theory, experimental methods, formal modeling and statistics. Similar to the Refresher Course in Statistics, this course will adopt a hands-on approach in which morning lectures will be followed by afternoon laboratory sessions in which student will apply the concepts taught in the morning to solve problems.
PREREQUISITES
A background in algebra is helpful and students should feel comfortable with engaging with mathematics, formulas, and data analysis. Students who will be taking two-week courses in the Summer School requiring background in mathematics should strongly consider enrolling in this course.
Refresher Course in Statistics
Lorena Barberia, University of São Paulo
OUTLINE
This course is designed for students who are interested in reviewing their training in statistics. It prepares students for courses offered in the IPSA Summer School that require statistical training. The course will take place in the week preceding the commencement of the Summer School. It would be a refresher course in statistics focusing on probability, and data analysis. It reviews basic probability; random variables and their distributions; confidence intervals and tests of hypotheses for means, variances, proportions from one or two populations, and concludes with an introduction to multiple regression analysis. Similar to the Mathematics for Social Scientists course, the course takes a “hands on” approach. To complement lectures, students apply the concepts taught in lectures to analyze problems using software packages commonly used in quantitative social science research including Excel, R and Stata.
DATES
This course runs January 27-31, 2014.
TEACHING FELLOW: Natália de Paula Moreira / Francisco Urdinez
INSTRUCTOR
Lorena G. Barberia joined the Department of Political Science at the University of São Paulo as a member of its faculty in 2011. Her primary fields of interest are comparative political economy, economic development and Latin American politics. Her work is aimed at analyzing redistributive politics in Latin America. Her current projects include: the analysis of the impact of democracy and electoral institutions in developing democracies in Latin America on social services; the application of political budget cycles models to understand the impact of electoral politics on state-level expenditures in Argentina and Brazil; and the exploration of the impact of local politics on corruption in health spending in Brazilian municipalities. At USP, Professor Barberia’s undergraduate and graduate courses include: Quantitative Methods and Techniques in Political Science.
DETAILED DESCRIPTION
This course departs from the premise that the most effective way to learn statistics is by actively engaging in doing the statistical analysis. For each topic, we will have lectures that will be followed by sessions in which students will use data to answer questions that are important to political scientists. Although the goals of the class are primarily conceptual rather than narrowly mathematical, students should feel comfortable with engaging with mathematics, formulas, and data analysis.
By the end of the intensive one-week course, students should be able to apply basic concepts in probability theory to social science research questions, to make inferences about the distribution of populations based on a sample and to correctly conduct and interpret hypothesis tests. For those students who will be studying multivariate regression analysis in the Summer School, the course will provide an intuitive and basic review of linear regression in theory and practice.
PREREQUISITES
The course presumes students have some basic training in mathematics including arithmetic and algebra operations.
Predicting Elections: Analytical Techniques and Illustrative Case Studies
Clifford Young, IPSOS-Washington
OUTLINE
Election forecasting is an activity which attempts to predict the outcome of an election. Such prediction can either be very far out from the election (a year or more) or very close to election-day (a few days before). Election forecasting can also employ the most simple of analytical frameworks to highly-sophisticated multivariate statistical models. So how is election forecasting employed in practice today? What methods are employed? Are some better under certain conditions? And what are different methods’ relative strengths and weaknesses? The course will attempt to answer these questions through the use of both an examination of the extant literature on the subject on election forecasting as well as using illustrative case studies. The primary objective is to provide the student with a general understanding of election forecasting and a simple tool box to work from. It is important to stress that the course will focus on national elections (in both presidential and parliamentary systems) not on downstream races at the legislative, state, and local levels
DATES
This is a one-week course and runs February 3-7, 2014.
TEACHING FELLOW: Natália de Paula Moreira
INSTRUCTOR
Clifford Young is Managing Director Ipsos’ Public Sector and Polling in the US and leads its global elections and polling risk practice. His specialties include social and public opinion trends, crisis management, communications research, and election polling. He also currently oversees Ipsos’ US presidential election polling for Thomson Reuters, and is the spokesperson for Ipsos Public Affairs in the US. Cliff is considered an expert on polling in emerging markets, as well as polling in adverse and hostile conditions. Before coming to Ipsos Public Affairs North America, he was Managing Director of Ipsos Public Affairs Brazil where he started the practice for Ipsos. Cliff has polled on over 80 elections around the world. He is a frequent writer, analyst, and commentator on elections, electoral polling, and public opinion.
DETAILED DESCRIPTION
Election forecasting is increasingly finding its way into the national media coverage of elections. The best example of this is the 2012 US Presidential election which pitted Obama versus Romney. During this election cycle, Nate Sliver as well as other election forecasters became important stakeholders in the national narrative about the elections and ultimately predicted the election spot on. Many are talking about the rise of the “forecaster-pundit” who will replace traditional journalist and politics in election analysis.
Election prediction has many applications in practice. Indeed, reducing uncertainty about which candidate will win on election-day can be an important strategic advantage for decision-makers in the private sector as we all in politics. The examples are many but include financial services firms making bets on currency or equities, ‘brick and motor’ companies making long-term capital investments, and political parties choosing the most viable candidate. Ultimately, election- forecasting is a central tool in assessing the political risk associated with any specific decision.
The course includes four components:
First, the course will review the existing literature on forecasting as well as best-industry practices. In particular, we will explore definitions of prediction and forecasting; will survey theories behind behavioral change and behavioral prediction; will examination of the median voter model and related concepts; and will review of different types of analyst biases and how they can hinder optimal prediction.
Second, the course will review the different forecasting models used today by election forecasters, including the most naïve ‘rule of thumb’ frameworks to highly sophistical multivariate statistical models. We will explore a number of different statistical models including those which include purely economic variables to those which employ public opinion variables as well. Finally, we will also examine statistical models using both classical and Bayesian frameworks. In essence, we want to compare and contrast different methods, assessing their accuracy and determining under what conditions they perform more optimally and under which conditions they don’t. Ultimately, we will find that ‘one size does not fit all’ and that the forecaster needs to have a broad tool box of methods in order to be effective across contexts and countries.
Third, we will use the US 2012 presidential election and the Brazilian 2010 election to assess the different forecasting methods. I choose these two elections because of their unique characteristics: the US is what I consider a data-intensive election environment, while Brazil is a data-scarce one. Such relative differences in the volume of information require different forecasting solutions. When appropriate, we will also examine other elections including the 2013 Kenyan, Italian, and Venezuelan elections; the 1992 British elections; 2011 Nigerian Elections as well as the upcoming Brazilian 2014 election.
Fourth, the course will also include lab sessions which will be used to run different models and methods. Specifically, we will analyze data both from the 2012 US presidential and 2010 Brazilian elections. The lab is meant to give ‘hands on’ data experience and practical tools.
PREREQUISITES
- Some familiarity with SPSS, STATA R, or SAS. We will be using a combination of SPSS, Excel, and R.
- At least one course in multivariate statistics.
The Philosophy and Methodology of the Social Sciences
Patrick Thaddeus Jackson, American University
OUTLINE
This course is a broad survey of epistemological, ontological, and methodological issues relevant to the production of knowledge in the social sciences. The course has three overlapping and interrelated objectives: to provide you with a grounding in these issues as they are conceptualized and debated by philosophers, social theorists, and intellectuals more generally; to act as a sort of introduction to the ways in which these issues have been incorporated (sometimes—often—inaccurately) into different branches of the social sciences and to serve as a forum for reflection on the relationship between these issues and the concrete conduct of research, both your own and that of others.
That having been said, this is neither a technical “research design” nor a “proposal writing” class, but is pitched as a somewhat greater level of abstraction. As we proceed through the course, however, you should try not to lose sight of the fact that these philosophical debates have profound consequences for practical research. Treat this course as an opportunity to set aside some time to think critically, creatively, and expansively about the status of knowledge, both that which you have produced and will produce, and that produced by others.
DATES
This course runs February 3-7, 2014.
TEACHING FELLOW: Vinicius Spira
DETAILED DESCRIPTION:
PREREQUISITES
No prerequisites except the willingness to work hard reading some complicated conceptual material.
Case Study Methodology – Comparing, Matching and Tracing
Derek Beach, University of Aarhus
OUTLINE
The aim of this course is to provide students with a set of methodological tools that enable the use of causal case study methods in your own research. A constant theme throughout the course will be on debating the strengths and limitations of different causal case study methods, illustrating the types and scopes of inferences that are possible, and whether and how they can be nested into mixed-methods research designs.
DATES
This course runs February 3-14, 2014.
TEACHING FELLOW: Terra Friedrich Budini
DETAILED DESCRIPTION
The course starts by introducing the debate on whether there is division between quantitative, large-n and qualitative case study methods. This is followed by a session on working with concepts, case selection principles in different case study methods, and discussions about causal inferences and causal relationships in case study methods. We discuss how we can develop strong empirical tests of our causal theories, and how your existing tests can be improved.
The course then turns to individual case study methods, including cross-case, comparative research designs (Mill’s methods, structured-focused comparisons and typological theorization) and within-case methods (congruence/pattern-matching, and process tracing). Particular emphasis will be given to process tracing as a core case study method that enables the research to open up the black-box between a cause and outcome by tracing causal mechanisms. In the final part of the course, we debate when and how different case study methods can be nested into mixed-methods designs.
The course is designed for students in the early to mid-stages of the dissertation who have already defined their research question. The course will only indirectly with interpretive qualitative methods. The final exercise will be the production of a 5-7 page research design paper.
PREREQUISITES
Each student should arrive at the summer school with a 2-3 page description of their research question and tentative research design which will be presented and discussed during the course. Students are expected to have encountered basic qualitative, case study research methods in their graduate-level education (e.g. King, Keohane and Verba’s Designing Social Inquiry (1994) is a good starting point).
Comparative Research Designs and Comparative Methods
Dirk Berg-Schlosser, Philipps University Marburg (Gernamy).
OUTLINE
Emile Durkheim, one of the founders of modern empirical social science once stated that the comparative method is the only one which suits the social sciences. But Descartes already had reminded us that “comparaison n’est pas raison”, i. e. comparison is not reason (or theory) by itself. This course provides an introduction and overview of systematic comparative analyses in the social sciences and shows how to employ this method for constructive explanation and theory-building.
DATES
This course runs February 3-14, 2014.
TEACHING FELLOW: Adrian Albala
DETAILED DESCRIPTION
The course begins with comparisons of very few cases and specific “most similar” and “most different” research designs. A major part is then devoted to the often occurring situation of dealing with a small number of highly complex cases, e.g., when comparing Latin American political systems or particular policy areas.
In response to this complexity, new approaches and software have been developed in recent years (“Qualitative Comparative Analysis”, QCA, and related methods). These procedures are able to reduce complexity and to arrive at “configurational” solutions based on set theory and Boolean algebra, which are more meaningful in this context than the usual broad-based statistical methods. In a last section more common statistical comparative methods at the macro-level of states or societies are presented and the respective strengths and weaknesses discussed.
Participants are strongly encouraged to present their own research problems and data, if available. Some basic quantitative or qualitative methodological training is probably useful to get more out of the course, but participants with little methodological training should find no major obstacles to follow.
PREREQUISITES
Some basic quantitative or qualitative methodological training is probably useful to get more out of the course, but participants with little methodological training should find no major obstacles to follow.
Introduction to Network Analysis using Pajek
Vladimir Batagelj, University of Ljubljana (Slovenia).
OUTLINE
The course aims to provide an introduction into the main topics and concepts of social network analysis. It focuses on the analysis and visualisation of complete networks. Participants will get an understanding of basic network analysis concepts like centrality, cohesion, blockmodeling, etc. Special attention will be given to the analysis of large networks. After the course participants should be able to examine data in ‘social networks way’ – they should be able to identify and formulate their own network analysis problems, solve them using network analysis software and interpret the obtained results. The course is supported by Pajek – a program for analysis and visualisation of large networks.
DATES
This course runs February 3-14, 2014.
TEACHING FELLOW: Ana Carolina Andrada
DETAILED DESCRIPTION
The course will start with an overview of the history of social network analysis, followed by a presentation of some typical and well-known real-life networks. In the main part fundamental concepts and methods of network analysis will be explained. Lab sessions will be performed using the software package Pajek. Then the course will cover the following topics:
1) Basic network concepts: network representations: matrix, graph; types of networks: undirected networks, directed networks, multi-relational networks, 2-mode networks, temporal networks; size and density; small, large and huge networks, sparse and dense networks;
2) Program Pajek and other network analysis software: description of networks in Pajek input file; network layouts: automatic and manual drawing; Unicode; connection with statistical packages (R); utility programs: Text2Pajek, GSView, SVG, King, Inkscape;
3) Paths in networks: walk, chain and path; closed walk, cycle, closed chain, loop; length and value of path; the shortest path, diameter; k-neighbours; acyclic networks;
4) Centrality: degree, closeness, betweenness; hubs and authorities, clustering coefficient; Hummon-Doreian’s weights in acyclic networks; small world and scale-free networks;
5) Weights and properties: line and vertex cuts, sub-networks; regression; visualisation in Pajek;
6) Connectivity: weakly, strongly and bi-connected components; global and local views; contraction; extraction; skeletons: minimal spanning trees, Pathfinder;
7) Cohesion: triads, cliques, rings, cores, islands; strong and weak ties; pattern search (motifs);
8) 2-mode networks: examples of 2-mode networks; direct analysis of 2-mode networks; multiplication of networks; transforming 2-mode to 1-mode networks; analysis of bibliometric (citation, collaboration, keywords/tags,…) networks;
9) Blockmodelling: direct and indirect approaches; structural, regular equivalence;
10) Generalised blockmodelling and blockmodelling of 2-mode networks;
11) Temporal, spatial and multirelational networks: macros and operations on sequences of networks.
After listening to the lectures, participants will work individually in computer labs. Several data sets will be prepared to challenge their knowledge.
PREREQUISITES
Participants need to have basic knowledge of mathematics (set theory notation, computation with matrices and vectors) and statistics. Basic familiarity with at least one statistical package (R or SPSS) can be helpful. Participants are expected to attend computer labs daily, where the software package Pajek will be used. During the lab hours, students will perform several network analyses on different small and large networks individually.
Mixed Methods Design
Max Bergman, University of Basel
OUTLINE
Mixed methods research and design, sometimes also referred to methodological triangulation, is an innovative and increasingly important way to conduct research in the social sciences. It transcends the inherent limitations of mono-methods research, i.e. qualitative or quantitative methods. As a research and design strategy, it often yields results that go beyond the “sum” of a qualitative and quantitative component within a single research project. Finally, it presents one way of overcoming the often deplored quantitative – qualitative divide in social research methods.
DATES
This course runs February 3-14, 2014.
DETAILED DESCRIPTION
We will study when, why, and how to combine and integrate qualitative and quantitative research methods. Our focus is primarily practical, which means that our main emphasis will be on exploring the possibilities and limits of mixing qualitative and quantitative methods. The course includes lectures, practical exercises, computer lab work, and assignments.
In the first part of the course, more conventional mixed method designs will be covered, including sequential, convergent, and concurrent designs.
The second part of the course will be dedicated to the exploration of a more advanced approach to integrating methods, particularly holistic designs and the immediate context during data collection and analysis.
PREREQUISITES
A university course in qualitative methods (e.g. interviewing, qualitative data analysis, grounded theory) or (not and!) an introduction in multivariate statistics as well as an interest in exploring their combination are necessary to profit from this course.
Multi-Level Analysis
Iñaki Sagarzazu, University of Glasgow
OUTLINE
Many kinds of data used for social science research can have different levels of analysis; this is known as a multi-level, hierarchical or clustered structure. For example, two voters with similar incomes but who live in two different countries might have different preferences. Multi-level models recognize these hierarchies in the data. In this case, instead of assuming that countries don’t make a difference we can use a two-level model which allows for grouping of voter preferences within countries. Thus it can obtain two components: a between-country component, which allows comparing differences between countries, and a within-country component, which allows us to compare differences between voters. In this course you will learn how to go from traditional one level model to more complex two –or multi- level models.
Students will learn how to (a) determine the need for a hierarchical model in their own research; (b) operationalize the most appropriate model; and (c) assess model performance and fit.
DATES
This course runs February 3-14, 2014.
TEACHING FELLOW: Mauricio Izumi
DETAILED DESCRIPTION
| Morning | Afternoon | |
| 1 | Review of the Classical Linear Regression Model | Introducing R: The Basics, Organizing Data |
| 2 | Multilevel Processes and Structures | Programming in R |
| 3 | Model Specification | Presentation of Models and Results – Graphing in R |
| 4 | Basics of Multilevel Modeling | Estimating models in R |
| 5 | Models for Continuous Responses | Models for Continuous Responses |
| 6 | Models for Dichotomous Responses | Models for Dichotomous Responses |
| 7 | Multilevel Generalized Linear Models | Multilevel Generalized Linear Models |
| 8 | Model Checking—Statistical Power & Sample Size | Model Checking—Statistical Power & Sample Size |
| 9 | Understanding and Summarizing the Results | Understanding and Summarizing the Results |
| 10 | Causal Inference Using Multilevel Models | Causal Inference Using Multilevel Models |
PREREQUISITES
Students wishing to take this course need to have taken previously a course on Regressions and Statistical modeling. Students should also be confident in the use of statistical programs such as Stata or R; although the course will solely use R.
Quantitative Methods for Public Policy Analysis
Bruno Cautrès, Science Po
OUTLINE
This course presents the major multivariate statistical methods that are fundamental to read advanced journals and to use statistics in the field of public policy analysis and public administration. Many specialists of public policy and administration, government/governance analysis, need modern quantitative methodologies when they are involved in evaluation of their actions and choices or in policy implementation. A key problem is then to show the net (and positive) impact of these policy changes and implementations. A policy expert that cannot prove this is somewhat in a bad position…The course aims at providing to participants a toolbox that may save their career…! The aim is making participants more effective users of modern statistical tools in analyzing public policies problems and to give diagnostics about policy solutions and implementations. The interaction between statistics, policy analysis and decision making will be highlighted.
DATES
This course runs February 3-14, 2014.
TEACHING FELLOW: Ivan Filipe de Almeida Lopes Fernandes
DETAILED DESCRIPTION
The course has two major objectives:
- To make the participants good experts of basic multivariate statistical methods (in particular modern methods of regression analysis, factor analysis);
- From there to cover recent extensions specially designed for public policy analysis.
Among these extensions some are very important for policy evaluation which aims at establishing a causal link between interventions and outcomes: randomized evaluations, natural experiments, the regression discontinuity design, instrumental variables, difference-in-differences regression, will be presented in details. These methods permit the experts and practionners of public policy analysis to study the net impact of their policy choices and implementations in a typical before/after quasi-experimental framework. The temporal change perspective will be privileged even if the methods could be applied to cross-national analysis. The course presents methods in simple terms, the level of mathematical presentation is basic since the use and practical understanding is the fundamental objective. Cases studies will be available through a selection of papers and applications. The first week will cover the basic tools of multivariate statistics (regression analysis, factor analysis); the second week will cover the specialized multivariate tools for public policy analysis (natural experiments, instrumental variables, difference-in-difference regression, the regression discontinuity design).
PREREQUISITES
The course is introductory, only some basic knowledge in descriptive statistics and statistical inference (up to hypothesis testing) may be helpful before to attend. In any case, recalls and remedial sessions in basic univariate statistics can be offered.
Time Series and Pooled Time Series Analyses
Guy D. Whitten, Texas A&M University and Lorena G. Barberia, University of São Paulo
OUTLINE
The course is an applied course focusing on the techniques for testing theories with time series and pooled time series data. Students will learn the theory and practical application of a wide range of techniques for making theoretical inferences about data with dynamic and pooled dynamic structures. While both types of data are pervasive and highly informative, they present a unique set of challenges to applied researchers. This course is designed to present clear explanations of these challenges and a series of strategies for how to overcome them. A strong emphasis will also be placed on strategies for presenting the results from such models through simulation-based graphics, tables, and effective writing.
DATES
This course runs February 3-14, 2014.
TEACHING FELLOW: Leonardo S. Barone
DETAILED DESCRIPTION
This is a hands-on course, meaning that a major goal is to have students learn about techniques by putting them to work with statistical software. To facilitate this, we will have lectures on each topic in the mornings followed by lab sessions in the afternoons with a teaching assistant. In these lab sessions students will have the option of working with their own data or working with data provided by the instructor. The main statistical software program that we will use for the labs is Stata. We will also look at some examples in EViews and WinRATS.
Although some knowledge of Stata or other statistical programs would be helpful, this is not required. Students are strongly encouraged to bring with them one or more time series or pooled time series data sets in which they have an interest. This suggestion inevitably leads to the question: “How many time points do I need in my data set to estimate the types of models covered in this course?” Although there is no definitive answer to this question, it is usually difficult to get useful estimates from data sets with less than 30 time points. The instructor will be able to convert most types of data sets into formats compatible with the programs used in the course lab exercises. If you have any questions about data that you might bring to the course, send them to the instructor.
PREREQUISITES
Students should have a basic knowledge of multiple regression and social science research design. There are many ways to obtain this, but one self-serving suggestion from the instructor is to read Kellstedt and Whitten’s The Fundamentals of Political Science Research, 2nd edition, with a focus on the materials covered in Chapters 8 through 10.
