Identification for prediction and decision / Charles F. Manski.

Format:

Book

Author/Creator:

Manski, Charles F.

Language:

English

Subjects (All):

Forecasting--Methodology.

Forecasting.

Social prediction.

Decision making.

Physical Description:

xiv, 348 pages ; 25 cm

Place of Publication:

Cambridge, Mass. : Harvard University Press, 2007.

Summary:

This book is a full-scale exposition of Charles Manski's new methodology for analyzing empirical questions in the social sciences. He recommends that researchers first ask what can be learned from data alone, and then ask what can be learned when data are combined with credible weak assumptions. Inferences predicated on weak assumptions, he argues, can achieve wide consensus, while ones that require strong assumptions almost inevitably are subject to sharp disagreements.

Contents:

The Reflection Problem 1

The Law of Decreasing Credibility 2

Identification and Statistical Inference 3

Prediction and Decisions 6

Coping with Ambiguity 6

The Developing Literature on Partial Identification 11

I Prediction with Incomplete Data

1 Conditional Prediction 17

1.1 Predicting Criminality 17

1.2 Probabilistic Prediction 18

1.3 Estimation of Best Predictors from Random Samples 22

1.4 Extrapolation 25

1.5 Predicting High School Graduation 28

Complement 1A Best Predictors under Square and Absolute Loss 30

Complement 1B Nonparametric Regression Analysis 32

Complement 1C Word Problems 34

2 Missing Outcomes 36

2.1 Anatomy of the Problem 37

2.2 Bounding the Probability of Exiting Homelessness 40

2.3 Means of Functions of the Outcome 42

2.4 Parameters That Respect Stochastic Dominance 44

2.5 Distributional Assumptions 45

2.6 Wage Regressions and the Reservation-Wage Model of Labor Supply 48

2.7 Statistical Inference 51

Complement 2A Interval Measurement of Outcomes 54

Complement 2B Jointly Missing Outcomes and Covariates 56

Complement 2C Convergence of Sets to Sets 60

3 Instrumental Variables 62

3.1 Distributional Assumptions and Credible Inference 62

3.2 Missingness at Random 64

3.3 Statistical Independence 66

3.4 Equality of Means 69

3.5 Inequality of Means 71

Complement 3A Imputations and Nonresponse Weights 73

Complement 3B Conditioning on the Propensity Score 75

Complement 3C Word Problems 76

4 Parametric Prediction 83

4.1 The Normal-Linear Model of Market and Reservation Wages 83

4.2 Selection Models 87

4.3 Parametric Models for Best Predictors 89

Complement 4A Minimum-Distance Estimation of Partially Identified Models 91

5 Decomposition of Mixtures 94

5.1 The Inferential Problem and Some Manifestations 94

5.2 Binary Mixing Covariates 98

5.3 Contamination through Imputation 102

5.4 Instrumental Variables 105

Complement 5A Sharp Bounds on Parameters That Respect Stochastic Dominance 107

6 Response-Based Sampling 109

6.1 The Odds Ratio and Public Health 110

6.2 Bounds on Relative and Attributable Risk 114

6.3 Information on Marginal Distributions 118

6.4 Sampling from One Response Stratum 119

6.5 General Binary Stratifications 122

II Analysis of Treatment Response

7 The Selection Problem 127

7.1 Anatomy of the Problem 128

7.2 Sentencing and Recidivism 134

7.3 Randomized Experiments 136

7.4 Compliance with Treatment Assignment 140

7.5 Treatment by Choice 148

7.6 Treatment at Random in Nonexperimental Settings 151

7.7 Homogeneous Linear Response 153

Complement 7A Perspectives on Treatment Comparison 157

Complement 7B Word Problems 160

8 Linear Simultaneous Equations 167

8.1 Simultaneity in Competitive Markets 167

8.2 The Linear Market Model 170

8.3 Equilibrium in Games 174

8.4 The Reflection Problem 177

9 Monotone Treatment Response 183

9.1 Shape Restrictions 183

9.2 Bounds on Parameters That Respect Stochastic Dominance 186

9.3 Bounds on Treatment Effects 189

9.4 Monotone Response and Selection 191

9.5 Bounding the Returns to Schooling 193

10 The Mixing Problem 198

10.1 Extrapolation from Experiments to Rules with Treatment Variation 198

10.2 Extrapolation from the Perry Preschool Experiment 200

10.3 Identification of Event Probabilities with the Experimental Evidence Alone 204

10.4 Treatment Response Assumptions 206

10.5 Treatment Rule Assumptions 207

10.6 Combining Assumptions 210

11 Planning under Ambiguity 211

11.1 Studying Treatment Response to Inform Treatment Choice 211

11.2 Criteria for Choice under Ambiguity 214

11.3 Treatment Using Data from an Experiment with Partial Compliance 218

11.4 An Additive Planning Problem 222

11.5 Planning with Partial Knowledge of Treatment Response 226

11.6 Planning and the Selection Problem 229

11.7 The Ethics of Fractional Treatment Rules 233

11.8 Decentralized Treatment Choice 235

Complement 11A Minimax-Regret Rules for Two Treatments Are Fractional 237

Complement 11B Reporting Observable Variation in Treatment Response 239

Complement 11C Word Problems 241

12 Planning with Sample Data 243

12.1 Statistical Induction 243

12.2 Wald's Development of Statistical Decision Theory 245

12.3 Using a Randomized Experiment to Evaluate an Innovation 250

III Predicting Choice Behavior

13 Revealed Preference Analysis 259

13.1 Revealing the Preferences of an Individual 260

13.2 Random Utility Models of Population Choice Behavior 263

13.3 College Choice in America 270

13.4 Random Expected-Utility Models 274

Complement 13A Prediction Assuming Strict Preferences 278

Complement 13B Axiomatic Decision Theory 282

14 Measuring Expectations 284

14.1 Elicitation of Expectations from Survey Respondents 285

14.2 Illustrative Findings 290

14.3 Using Expectations Data to Predict Choice Behavior 295

14.4 Measuring Ambiguity 298

Complement 14A The Predictive Power of Intentions Data: A Best-Case Analysis 300

Complement 14B Measuring Expectations of Facts 305

15 Studying Human Decision Processes 308

15.1 As-If Rationality and Bounded Rationality 309

15.2 Choice Experiments 312

15.3 Prospects for a Neuroscientific Synthesis 317.

Notes:

Includes bibliographical references (pages [321]-337) and indexes.

ISBN:

9780674026537

0674026535

OCLC:

85443934