Table of Contents
Preface ix
Acknowledgments xiii
About the Author xv
Part I Estimating Electoral Support 1
1 Polling 101 3
1.1 Simple Random Sampling 5
1.2 One Estimator of π: The Sample Proportion 6
1.3 Reasonable Values of π 12
1.4 A Second Estimator of π: Agresti-Coull 21
1.5 SRS without Replacement 26
1.6 Conclusion 28
1.7 Extensions 29
1.8 Chapter Appendix 30
2 Polling 399 33
2.1 Stratified Sampling 35
2.2 The Mathematics of Estimating π 38
2.3 Confidence Intervals 49
2.4 Conclusion 52
2.5 Extensions 53
2.6 Chapter Appendix 54
3 Combining Polls 57
3.1 Simple Averaging of Polls 59
3.2 Weighted Averaging of Polls 61
3.3 Averaging of Polls over Time 64
3.4 Looking Ahead 70
3.5 South Korean 2017 Presidential Election 73
3.6 Conclusion 77
3.7 Extensions 78
3.8 Chapter Appendix 79
4 In-Depth Analysis: Brexit 2016 83
4.1 Knowing Your Data 85
4.2 Combining the Polls 89
4.3 Discussion: What Went Wrong? 90
4.4 Conclusion 92
Part II Testing Election Results 94
5 Digit Tests 95
5.1 History 97
5.2 The Benford Test 99
5.3 The Generalized Benford Test 108
5.4 Using the Generalized Benford Distribution 112
5.5 Conclusion 119
5.6 Extensions 120
5.7 Chapter Appendix 121
6 Differential Invalidation 125
6.1 Differential Invalidation 127
6.2 Regression Modeling 132
6.3 Examining Côte d'Ivoire 141
6.4 Conclusion 143
6.5 Extensions 144
6.6 Chapter Appendix 145
7 Considering Geography 149
7.1 Detecting Spatial Correlation 151
7.2 The Spatial Lag Model 158
7.3 Casetti's Spatial Expansion Model (SEM) 161
7.4 Geographically Weighted Regression 164
7.5 The Spatial Lagged Expansion Method 169
7.6 Conclusion 172
7.7 Extensions 173
7.8 Chapter Appendix 174
8 In-Depth Analysis: Sri Lanka since 1994 177
8.1 Differential Invalidation 179
8.2 Methods and Data 179
8.3 Results by Election 182
8.4 Discussion 192
8.5 Conclusion 193
Bibliography 195
Index 207