Designing Social Inquiry

Gary King is the Albert J. Weatherhead III University Professor at Harvard University. His books include A Solution to the Ecological Inference Problem (Princeton). Robert O. Keohane is professor emeritus of international affairs at Princeton University. His books include After Hegemony (Princeton). Sidney Verba (1932–2019) was the Carl H. Pforzheimer University Professor Emeritus and research professor of government at Harvard. His books include Unequal and Unrepresented (Princeton). King, Keohane, and Verba have each been elected as members of the National Academy of Sciences.

• Preface to the 2021 EditionPreface to the First Edition1 The Science in Social Science1.1 Introduction1.1.1 Two Styles of Research, One Logic of Inference1.1.2 Defining Scientific Research in the Social Sciences1.1.3 Science and Complexity1.2 Major Components of Research Design1.2.1 Improving Research Questions1.2.2 Improving Theory1.2.3 Improving Data Quality1.2.4 Improving the Use of Existing Data1.3 Themes of This Volume1.3.1 Using Observable Implications to Connect Theory and Data1.3.2 Maximizing Leverage1.3.3 Reporting Uncertainty1.3.4 Thinking like a Social Scientist: Skepticism and Rival Hypotheses2 Descriptive Inference2.1 General Knowledge and Particular Facts2.1.1 “Interpretation” and Inference2.1.2 “Uniqueness,” Complexity, and Simplification2.1.3 Comparative Case Studies2.2 Inference: The Scientific Purpose of Data Collection2.3 Formal Models of Qualitative Research2.4 A Formal Model of Data Collection2.5 Summarizing Historical Detail2.6 Descriptive Inference2.7 Criteria for Judging Descriptive Inferences2.7.1 Unbiased Inferences2.7.2 Efficiency3 Causality and Causal Inference3.1 Defining Causality3.1.1 The Definition and a Quantitative Example3.1.2 A Qualitative Example3.2 Clarifying Alternative Definitions of Causality3.2.1 “Causal Mechanisms”3.2.2 “Multiple Causality”3.2.3 “Symmetric” and “Asymmetric” Causality3.3 Assumptions Required for Estimating Causal Effects3.3.1 Unit Homogeneity3.3.2 Conditional Independence3.4 Criteria for Judging Causal Inferences3.5 Rules for Constructing Causal Theories3.5.1 Rule 1: Construct Falsifiable Theories3.5.2 Rule 2: Build Theories That Are Internally Consistent3.5.3 Rule 3: Select Dependent Variables Carefully3.5.4 Rule 4: Maximize Concreteness3.5.5 Rule 5: State Theories in as Encompassing Ways as Feasible4 Determining What to Observe4.1 Indeterminate Research Designs4.1.1 More Inferences than Observations4.1.2 Multicollinearity4.2 The Limits of Random Selection4.3 Selection Bias4.3.1 Selection on the Dependent Variable4.3.2 Selection on an Explanatory Variable4.3.3 Other Types of Selection Bias4.4 Intentional Selection of Observations4.4.1 Selecting Observations on the Explanatory Variable4.4.2 Selecting a Range of Values of the Dependent Variable4.4.3 Selecting Observations on Both Explanatory and Dependent Variables4.4.4 Selecting Observations So the Key Causal Variable Is Constant4.4.5 Selecting Observations So the Dependent Variable Is Constant4.5 Concluding Remarks5 Understanding What to Avoid5.1 Measurement Error5.1.1 Systematic Measurement Error5.1.2 Nonsystematic Measurement Error5.2 Excluding Relevant Variables: Bias5.2.1 Gauging the Bias from Omitted Variables5.2.2 Examples of Omitted Variable Bias5.3 Including Irrelevant Variables: Inefficiency5.4 Endogeneity5.4.1 Correcting Biased Inferences5.4.2 Parsing the Dependent Variable5.4.3 Transforming Endogeneity into an Omitted Variable Problem5.4.4 Selecting Observations to Avoid Endogeneity5.4.5 Parsing the Explanatory Variable5.5 Assigning Values of the Explanatory Variable5.6 Controlling the Research Situation5.7 Concluding Remarks6 Increasing the Number of Observations6.1 Single-Observation Designs for Causal Inference6.1.1 “Crucial” Case Studies6.1.2 Reasoning by Analogy6.2 How Many Observations Are Enough?6.3 Making Many Observations from Few6.3.1 Same Measures, New Units6.3.2 Same Units, New Measures6.3.3 New Measures, New Units6.4 Concluding RemarksReferencesIndex