Prediction and causal inference are the central pursuits of empirical political science. Critical to reliable prediction and causal inference is an understanding of the structural relationships in the social and political systems under study. Graphical models are excellent tools for conceptualizing and representing relationships, such as relationships among the individual actors under study, or relationships among the variables in a model. Understanding the first type of relationship is critical particularly in the study of relational data, such as data on international conflict. Understanding the second type is at the core of causal inference aiming at clarifying the structural relationships among various quantities of interest in the system under study. This project explores the use of graphical methods and models in political data analysis, particularly in the study of international and civil conflict. International events take place not in isolation, but within a network of intricate interdependence. The structural characteristics of the international network, in addition to individual state or dyad level attributes, therefore likely hold important explanatory and predictive power. Systematic measurement and modeling of the characteristics and dependence structure of the network has largely eluded previous research. In this project, the international network of a given relation (e.g., the relation of "in conflict with") is modeled as a (random) graph. Dependence structure is modeled endogenously, and a rich array of well defined graph theoretic measures capturing the structure of the entire network are obtained to augment the usual state/dyad level attribute variables in the statistical model. Because causal models are more useful than associational models for policy analysis, the project explores the identification of causal structure from observed associational data using methods from causal graph theory, and improves the assessment of causal effects of key policy variables, such as level of democratization, by ensuring correct specification of control variables with the aid of the causal graph.

Understanding and forecasting international and civil conflict are of general interest, and the methods investigated in this project are broadly relevant to other areas of political science and indeed across disciplinary boundaries as prediction and causal inference are central to most fields of scientific inquiry. The graphical methods and models investigated in this project also will enrich the investigator's teaching of political methodology.

Agency
National Science Foundation (NSF)
Institute
Division of Social and Economic Sciences (SES)
Type
Standard Grant (Standard)
Application #
0318275
Program Officer
Cheryl L. Eavey
Project Start
Project End
Budget Start
2003-09-01
Budget End
2005-06-30
Support Year
Fiscal Year
2003
Total Cost
$50,312
Indirect Cost
Name
George Washington University
Department
Type
DUNS #
City
Washington
State
DC
Country
United States
Zip Code
20052