Supplemental Appendix for I M Not Voting for Her: Polling Discrepancies and Female Candidates

Supplemental Appendix for I’m Not Voting For Her: Polling Discrepancies and Female Candidates

Supplemental Appendix Table A: Predicting Polling Accuracy for Female Candidates using Elazar’s Political State Culture

Traditionalist Model
Traditionalist States / -3.137
(1.460)*
Women in State Leg / 0.038
(0.071)
Competitiveness / 0.123
(0.029)**
% Undecided / 0.121
(0.129)
Days to Election / -0.009
(0.059)
Incumbent in Race / -1.349
(1.119)
Third Party/DK / 0.020
(0.068)
Governor / 1.306
(1.093)
VAP Turnout / 0.117
(0.636)
Partisanship / -0.108
(0.088)
Constant / -2.696
(2.612)
Observations / 117
R-squared / 0.26

** p<0.01, * p<0.05, + p<0.1 DV: Latest Pre-Election Poll Margin of Victory-Final Result Margin of Victory (Positive direction signals that polls overestimated margin of victory). A positive number indicate that the candidate performed better in pre election polls than in the election. Negative results indicate a Richards Effect. Standard Errors are presented within parenthesis.

Supplemental Appendix Table B: Predicting Polling Accuracy for Male and Female Senate and Gubernatorial Candidates from 1989-2008 W/ Outliers

Mosteller #5
Female / -2.127
(0.976)*
Competitiveness / 0.117
(0.022)**
% Undecided / -0.055
(0.112)
Days to Election / -0.022
(0.051)
Incumbent in Race / -1.310
(1.015)
Third Party/DK / -0.008
(0.067)
Democrat / 0.149
(1.008)
Governor / 1.612
(0.969)
VAP Turnout / 0.071
(0.565)
Women in State Leg / 0.015
(0.069)
AAUW Score / 0.018
(0.024)
Partisanship of State / -11.744
(9.373)
Female Labor Force / 0.314
(0.133)*
Constant / -19.179
(7.652)*
Observations / 209
R-squared / 0.23

** p<0.01, * p<0.05, + p<0.1 DV: Latest Pre-Election Poll Margin of Victory-Final Result Margin of Victory (Positive direction signals that polls overestimated margin of victory). A positive number indicate that the candidate performed better in pre election polls than in the election. Negative results indicate a Richards Effect. Standard Errors are presented within parenthesis.

Supplemental Appendix Table C: Predicting Polling Accuracy for Male and Female Senate and Gubernatorial Candidates from 1989-2008 Using Measure A

Measure A
Female Candidate / -0.068 (0.025)**
Competitiveness / 0.004 (0.001)**
% Undecided / -0.001 (0.003)
Days to Election / -0.001 (0.001)
Incumbent in Race / -0.049 (0.026)
Third Party/DK / -0.001 (0.002)
Democrat / 0.003 (0.026)
Governor / 0.031 (0.025)
VAP Turnout / 0.000 (0.015)
Women in State Leg. / 0.003 (0.002)
AAUW Scores / 0.001 (0.001)
Partisanship of State / -0.100 (0.239)
Female Labor Force / 0.009 (0.003)*
Constant / -0.567 (0.198)**
Observations / 205
R-squared / 0.31

** p<0.01, * p<0.05, + p<0.1 DV:Log ((Poll support for Candidate A/Poll support for Candidate B)/(Election support for Candidate A/Election support for Candidate B)) A positive number indicate that the candidate performed better in pre election polls than in the election. Negative results indicate a Richards Effect. Standard Errors are presented within parenthesis.

Supplemental Appendix Table D: Predicting the Richards Effect: OLS Regression Predicting the Difference between the Latest Pre-Election Poll and the Final Results for Male and Female Candidates from 1988-2008 Using an Alternate Matching Technique

Baseline Model / Election Context Model 1 / Female Turnout Model / Social Context Model
Female / -1.933
(0.931)* / -2.065
(0.929)* / -2.070
(0.930)* / -2.079
(0.931)*
Competitiveness / 0.110
(0.021)** / 0.100
(0.021)** / 0.091
(0.021)**
% Undecided / -0.013
(0.104) / -0.021
(0.103) / 0.011
(0.108)
Days To Election / -0.025
(0.048) / -0.027
(0.048) / -0.033
(0.049)
Incumbent in Race / -2.348
(0.944)* / -2.317
(0.944)* / -2.112
(0.965)*
Third Party/DK / 0.009
(0.063) / 0.011
(0.063) / 0.015
(0.065)
Democrat / 0.731
(0.967) / 0.752
(0.969) / 0.771
(0.971)
Governor / 1.439
(0.924) / 1.410
(0.925) / 1.584
(0.933)
VAP Turnout / 0.252
(0.534) / 0.256
(0.542)
Female Turnout / 0.014
(0.519)
Women in State Legislature / 0.027
(0.065)
AAUW Scores / 0.004
(0.023)
Partisanship of the State (Dem) / -7.935
(8.942)
Female in Labor Force / 0.159
(0.128)
Constant / -0.039
(0.706) / 0.746
(1.732) / 0.828
(1.739) / -10.192
(7.470)
Observations / 211 / 205 / 205 / 205
R-squared / 0.02 / 0.1 / 0.1 / 0.12

** p<0.01, * p<0.05, + p<0.1 DV: Latest Pre-Election Poll Margin of Victory-Final Result Margin of Victory (Positive direction signals that polls overestimated margin of victory). A positive number indicate that the candidate performed better in pre election polls than in the election. Negative results indicate a Richards Effect. Standard Errors are presented within parenthesis.

Supplemental Appendix Table E: Predicting the Richards Effect: OLS Regression Predicting the Difference between the Latest Pre-Election Poll and the Final Results for Male and Female Candidates from 1988-2008 Controlling for Election Year Effects.

Candidates / Polling Accuracy (Mosteller #5)
Female Candidates / -2.385 (0.880)**
% Undecided / -0.057 (0.106)
Competitiveness / 0.099 (0.020)**
Incumbent in Race / -0.198 (0.925)
Third Party/DK / -0.016 (0.065)
Days to Election / 0.006 (0.046)
Democrat / 0.537 (0.913)
VAP Turnout / 0.398 (0.657)
Governor / 1.799 (0.873)*
AAUW Scores / 0.042 (0.024)
Women in State Legislature / 0.031 (0.063)
Democratic Partisanship in the State / -10.333 (8.564)
Female in Labor Force / 0.327 (0.121)**
Year==1990 / 3.976 (3.796)
Year==1992 / 6.853 (3.893)
Year==1993 / 1.821 (5.327)
Year==1994 / 4.277 (3.860)
Year==1996 / 9.163 (4.034)*
Year==1997 / 11.583 (7.283)
Year==1998 / 4.999 (3.886)
Year==2000 / 6.489 (3.947)
Year==2002 / 6.917 (3.882)
Year==2003 / 2.467 (7.034)
Year==2004 / 5.662 (3.992)
Year==2006 / 3.078 (3.928)
Year==2008 / 0.708 (4.144)
Constant / -27.857 (7.726)**
Observations / 209
R-squared / 0.34

** p<0.01, * p<0.05, + p<01 DV: Latest Pre-Election Poll Margin of Victory-Final Result Margin of Victory (Positive direction signals that polls overestimated margin of victory). A positive number indicate that the candidate performed better in pre election polls than in the election. Negative results indicate a Richards Effect. Standard Errors are presented within parenthesis.

Supplemental Appendix Table F: Predicting the Richards Effect: OLS Regression Predicting the Difference between the Latest Pre-Election Poll and the Final Results for Male and Female Candidates from 1988-2008 Using a Random Sample of White Male Candidates.

Mosteller #5
Female Candidates / -2.338 (1.346)+
Competitiveness / 0.118 (0.028)**
Incumbent in Race / 0.694 (1.146)
Days to Election / 0.020 (0.063)
Democrat / 0.963 (1.146)
VAP Turnout / -0.356 (0.642)
Governor / 0.806 (1.106)
Womenin State Leg / -0.135 (0.076)
Democratic Partisanship / -3.924 (9.047)
AAUW Scores / -0.001 (0.017)
Female in Labor force / 0.285 (0.147)
Constant / -14.726 (8.562)
Observations / 196
R-squared / 0.17

** p<0.01, * p<0.05, + p<01 DV: Latest Pre-Election Poll Margin of Victory-Final Result Margin of Victory (Positive direction signals that polls overestimated margin of victory). A positive number indicate that the candidate performed better in pre election polls than in the election. Negative results indicate a Richards Effect. Standard Errors are presented within parenthesis.