Regression Table 01 - Including Well-Groomed & Messy

Table No.1 - Baseline Regression
Dependent variable:
Release-Final-Outcome
(1) (2) (3) (4) (5) (6) (7)
Well-groomed 0.0181*** 0.0147* 0.0074* 0.0053
(0.0110, 0.0251) (0.0016, 0.0279) (0.0004, 0.0144) (-0.0077, 0.0183)
Messy -0.0152*** -0.0037 -0.0065 -0.0023
(-0.0218, -0.0087) (-0.0159, 0.0086) (-0.0129, 0.00004) (-0.0144, 0.0097)
ML-Face 0.6895*** 0.6786*** 0.6805*** 0.6785***
(0.6224, 0.7567) (0.6107, 0.7465) (0.6127, 0.7482) (0.6105, 0.7464)
Constant 0.6740*** 0.8269*** 0.7058*** 0.2472*** 0.2195*** 0.2817*** 0.2400***
(0.6391, 0.7089) (0.7978, 0.8560) (0.5939, 0.8177) (0.1965, 0.2978) (0.1624, 0.2765) (0.2203, 0.3430) (0.1203, 0.3596)
F(Messy + Groomed) X X 9.0361*** (0.00012) X X X 1.5598 (0.2102)
Observations 8,479 8,479 8,479 8,479 8,479 8,479 8,479
Adjusted R2 0.0020 0.0016 0.0019 0.0325 0.0327 0.0326 0.0326
F Statistic 17.8319*** (df = 1; 8477) 14.6675*** (df = 1; 8477) 9.0361*** (df = 2; 8476) 285.3710*** (df = 1; 8477) 144.2279*** (df = 2; 8476) 144.0506*** (df = 2; 8476) 96.1761*** (df = 3; 8475)
Note: p<0.1; p<0.05; p<0.01

Regression Table 02 - Including Demographic Variables

Table No.2 - Including Demographics
Dependent variable:
Release-Final-Outcome
(1) (2) (3) (4) (5) (6) (7)
SexM -0.1043*** -0.1079*** -0.1064*** -0.0276** -0.0308** -0.0325*** -0.0319**
(-0.1230, -0.0857) (-0.1266, -0.0891) (-0.1252, -0.0875) (-0.0480, -0.0071) (-0.0514, -0.0103) (-0.0532, -0.0118) (-0.0527, -0.0111)
RaceB -0.0278 -0.0298 -0.0283 -0.0268 -0.0238 -0.0248 -0.0241
(-0.1132, 0.0576) (-0.1151, 0.0556) (-0.1137, 0.0571) (-0.1113, 0.0576) (-0.1083, 0.0606) (-0.1093, 0.0596) (-0.1085, 0.0604)
Age 0.0007 0.0007 0.0008 0.0008* 0.0010** 0.0010** 0.0010**
(-0.00004, 0.0015) (-0.0001, 0.0015) (-0.00001, 0.0016) (0.00002, 0.0015) (0.0003, 0.0018) (0.0002, 0.0018) (0.0003, 0.0018)
† skin-tone 0.5980*** 0.5769*** 0.5889*** 0.3946** 0.4126** 0.4021** 0.4091**
(0.2846, 0.9114) (0.2637, 0.8902) (0.2753, 0.9024) (0.0838, 0.7053) (0.1016, 0.7235) (0.0914, 0.7128) (0.0981, 0.7202)
Well-Groomed 0.0213*** 0.0117 0.0105** 0.0064
(0.0140, 0.0286) (-0.0020, 0.0254) (0.0032, 0.0178) (-0.0071, 0.0200)
Messy -0.0198*** -0.0105 -0.0096** -0.0045
(-0.0266, -0.0130) (-0.0233, 0.0022) (-0.0164, -0.0027) (-0.0172, 0.0082)
ML-Face 0.6479*** 0.6281*** 0.6282*** 0.6265***
(0.5724, 0.7234) (0.5513, 0.7048) (0.5514, 0.7050) (0.5496, 0.7033)
Constant 0.7260*** 0.9194*** 0.8184*** 0.2920*** 0.2460*** 0.3406*** 0.2868***
(0.6228, 0.8292) (0.8233, 1.0156) (0.6661, 0.9707) (0.1786, 0.4053) (0.1282, 0.3638) (0.2220, 0.4591) (0.1226, 0.4510)
F(Messy + Groomed) X X 12.528*** (3.692e-06) X X X 2.958 (0.05198)
Observations 8,466 8,466 8,466 8,466 8,466 8,466 8,466
Adjusted R2 0.0138 0.0137 0.0139 0.0338 0.0343 0.0343 0.0343
F Statistic 15.7672*** (df = 8; 8457) 15.7495*** (df = 8; 8457) 14.2213*** (df = 9; 8456) 38.0392*** (df = 8; 8457) 34.4500*** (df = 9; 8456) 34.4199*** (df = 9; 8456) 31.0370*** (df = 10; 8455)
Note: p<0.1; p<0.05; p<0.01
† We encode skin-tone as a continuous variable increasing in brightness

Regression Table 03 - Including Psychological Features

Table No.3 - Including Todorov Labels
Dependent variable:
Release-Final-Outcome
(1) (2) (3) (4) (5) (6) (7)
SexM -0.0992*** -0.1024*** -0.1013*** -0.0255** -0.0274** -0.0287** -0.0284**
(-0.1181, -0.0803) (-0.1214, -0.0833) (-0.1205, -0.0822) (-0.0461, -0.0049) (-0.0481, -0.0066) (-0.0497, -0.0078) (-0.0494, -0.0074)
RaceB -0.0243 -0.0265 -0.0252 -0.0237 -0.0215 -0.0224 -0.0219
(-0.1097, 0.0611) (-0.1119, 0.0589) (-0.1106, 0.0603) (-0.1082, 0.0608) (-0.1061, 0.0630) (-0.1069, 0.0621) (-0.1064, 0.0627)
Age 0.0009* 0.0009* 0.0009* 0.0011** 0.0012** 0.0012** 0.0012**
(0.0001, 0.0017) (0.0001, 0.0017) (0.0001, 0.0017) (0.0003, 0.0019) (0.0004, 0.0020) (0.0004, 0.0020) (0.0004, 0.0020)
† skin-tone 0.5912*** 0.5753*** 0.5835*** 0.3971** 0.4099** 0.4036** 0.4071**
(0.2775, 0.9048) (0.2619, 0.8887) (0.2697, 0.8973) (0.0862, 0.7080) (0.0986, 0.7211) (0.0926, 0.7145) (0.0957, 0.7184)
Attractiveness 0.0006 0.0015 0.0008 0.0037 0.0021 0.0025 0.0022
(-0.0094, 0.0105) (-0.0084, 0.0113) (-0.0092, 0.0107) (-0.0059, 0.0134) (-0.0077, 0.0120) (-0.0072, 0.0122) (-0.0076, 0.0121)
Competence 0.0018 0.0013 0.0013 -0.0001 -0.0010 -0.0012 -0.0012
(-0.0086, 0.0122) (-0.0091, 0.0117) (-0.0091, 0.0117) (-0.0103, 0.0102) (-0.0112, 0.0093) (-0.0115, 0.0091) (-0.0115, 0.0091)
Dominance -0.0077* -0.0062 -0.0067 -0.0055 -0.0056 -0.0050 -0.0052
(-0.0152, -0.0002) (-0.0137, 0.0013) (-0.0143, 0.0008) (-0.0129, 0.0019) (-0.0130, 0.0018) (-0.0124, 0.0025) (-0.0127, 0.0023)
Trustworthiness 0.0141** 0.0150** 0.0145** 0.0129** 0.0121** 0.0125** 0.0123**
(0.0042, 0.0240) (0.0051, 0.0249) (0.0045, 0.0244) (0.0032, 0.0227) (0.0023, 0.0219) (0.0027, 0.0223) (0.0025, 0.0221)
Well-groomed 0.0167*** 0.0078 0.0068 0.0031
(0.0086, 0.0247) (-0.0067, 0.0222) (-0.0013, 0.0149) (-0.0112, 0.0174)
Messy -0.0154*** -0.0096 -0.0063 -0.0040
(-0.0227, -0.0082) (-0.0226, 0.0034) (-0.0136, 0.0009) (-0.0169, 0.0089)
ML-Face 0.6313*** 0.6215*** 0.6209*** 0.6202***
(0.5553, 0.7073) (0.5446, 0.6983) (0.5439, 0.6978) (0.5432, 0.6972)
Constant 0.7120*** 0.8520*** 0.7936*** 0.2545*** 0.2369*** 0.2946*** 0.2717***
(0.6058, 0.8183) (0.7440, 0.9601) (0.6406, 0.9466) (0.1358, 0.3731) (0.1164, 0.3574) (0.1673, 0.4219) (0.1070, 0.4364)
F(Messy + Groomed) X X 6.5335** (0.001461) X X X 1.0896 (0.3364)
Observations 8,466 8,466 8,466 8,466 8,466 8,466 8,466
Adjusted R2 0.0148 0.0149 0.0149 0.0348 0.0349 0.0349 0.0348
F Statistic 11.5991*** (df = 12; 8453) 11.6582*** (df = 12; 8453) 10.8217*** (df = 13; 8452) 26.4166*** (df = 12; 8453) 24.5349*** (df = 13; 8452) 24.5453*** (df = 13; 8452) 22.7990*** (df = 14; 8451)
Note: p<0.1; p<0.05; p<0.01
† We encode skin-tone as a continuous variable increasing in brightness

Regression Table 04 - Including Current Charge Dummies and Risk

Table No.4 - Including Risk & Charge
Dependent variable:
Release-Final-Outcome
(1) (2) (3) (4) (5) (6) (7)
SexM -0.0555*** -0.0582*** -0.0579*** 0.0010 -0.0003 -0.0018 -0.0019
(-0.0738, -0.0372) (-0.0766, -0.0398) (-0.0764, -0.0394) (-0.0189, 0.0210) (-0.0204, 0.0198) (-0.0221, 0.0185) (-0.0222, 0.0185)
RaceB 0.0293 0.0279 0.0283 0.0256 0.0270 0.0266 0.0264
(-0.0516, 0.1103) (-0.0530, 0.1089) (-0.0527, 0.1093) (-0.0548, 0.1061) (-0.0535, 0.1074) (-0.0538, 0.1070) (-0.0540, 0.1069)
Age -0.0003 -0.0003 -0.0003 -0.0001 -0.00002 0.000003 0.000001
(-0.0011, 0.0005) (-0.0010, 0.0005) (-0.0010, 0.0005) (-0.0008, 0.0007) (-0.0008, 0.0008) (-0.0008, 0.0008) (-0.0008, 0.0008)
† skin-tone 0.4937*** 0.4825*** 0.4850*** 0.3598** 0.3679** 0.3649** 0.3637**
(0.1965, 0.7908) (0.1856, 0.7795) (0.1877, 0.7824) (0.0643, 0.6553) (0.0721, 0.6638) (0.0693, 0.6605) (0.0677, 0.6597)
Attractiveness 0.0020 0.0025 0.0023 0.0041 0.0031 0.0032 0.0033
(-0.0073, 0.0114) (-0.0069, 0.0118) (-0.0071, 0.0117) (-0.0050, 0.0133) (-0.0062, 0.0125) (-0.0061, 0.0124) (-0.0061, 0.0126)
Competence 0.00004 -0.0005 -0.0005 -0.0011 -0.0017 -0.0020 -0.0020
(-0.0098, 0.0099) (-0.0104, 0.0093) (-0.0104, 0.0093) (-0.0108, 0.0086) (-0.0115, 0.0080) (-0.0118, 0.0077) (-0.0118, 0.0078)
Dominance -0.0053 -0.0041 -0.0043 -0.0037 -0.0037 -0.0032 -0.0032
(-0.0124, 0.0017) (-0.0112, 0.0030) (-0.0115, 0.0029) (-0.0107, 0.0034) (-0.0108, 0.0033) (-0.0103, 0.0038) (-0.0103, 0.0040)
Trustworthiness 0.0093 0.0099* 0.0097* 0.0080 0.0075 0.0077 0.0078
(-0.0001, 0.0187) (0.0005, 0.0192) (0.0003, 0.0191) (-0.0013, 0.0173) (-0.0018, 0.0169) (-0.0016, 0.0170) (-0.0016, 0.0171)
Felony-Dummie -0.2030*** -0.2029*** -0.2029*** -0.2012*** -0.2014*** -0.2014*** -0.2014***
(-0.2192, -0.1867) (-0.2191, -0.1867) (-0.2192, -0.1867) (-0.2173, -0.1851) (-0.2175, -0.1852) (-0.2175, -0.1853) (-0.2175, -0.1852)
Gun-Crime-Dummie 0.0422** 0.0420** 0.0420** 0.0383** 0.0382** 0.0381** 0.0381**
(0.0136, 0.0709) (0.0134, 0.0707) (0.0134, 0.0707) (0.0099, 0.0668) (0.0098, 0.0667) (0.0097, 0.0666) (0.0097, 0.0666)
Drug-Crime-Dummie 0.0363*** 0.0366*** 0.0367*** 0.0409*** 0.0413*** 0.0415*** 0.0415***
(0.0167, 0.0560) (0.0170, 0.0563) (0.0170, 0.0563) (0.0213, 0.0604) (0.0218, 0.0609) (0.0220, 0.0611) (0.0220, 0.0611)
Violent-Crime-Dummie -0.1691*** -0.1696*** -0.1695*** -0.1623*** -0.1622*** -0.1624*** -0.1624***
(-0.1950, -0.1432) (-0.1955, -0.1437) (-0.1954, -0.1435) (-0.1881, -0.1365) (-0.1880, -0.1364) (-0.1882, -0.1366) (-0.1882, -0.1366)
Property-Crime-Dummie -0.0335*** -0.0338*** -0.0337*** -0.0281*** -0.0281*** -0.0282*** -0.0282***
(-0.0506, -0.0164) (-0.0508, -0.0167) (-0.0508, -0.0166) (-0.0451, -0.0112) (-0.0451, -0.0111) (-0.0452, -0.0112) (-0.0452, -0.0112)
Arrest Year - 2018 0.0340*** 0.0342*** 0.0342*** 0.0271*** 0.0273*** 0.0274*** 0.0274***
(0.0179, 0.0501) (0.0181, 0.0503) (0.0181, 0.0502) (0.0111, 0.0431) (0.0113, 0.0433) (0.0114, 0.0434) (0.0114, 0.0434)
Arrest Year - 2019 0.0893*** 0.0896*** 0.0896*** 0.0561*** 0.0569*** 0.0573*** 0.0572***
(0.0691, 0.1094) (0.0695, 0.1097) (0.0695, 0.1097) (0.0356, 0.0767) (0.0363, 0.0775) (0.0367, 0.0779) (0.0366, 0.0779)
Risk Prediction -0.8153*** -0.8141*** -0.8137*** -0.7759*** -0.7725*** -0.7716*** -0.7717***
(-0.9216, -0.7091) (-0.9203, -0.7078) (-0.9199, -0.7074) (-0.8815, -0.6704) (-0.8782, -0.6667) (-0.8773, -0.6658) (-0.8775, -0.6660)
Well-groomed 0.0122*** 0.0023 0.0044 -0.0011
(0.0045, 0.0198) (-0.0114, 0.0160) (-0.0033, 0.0121) (-0.0147, 0.0125)
Messy -0.0124*** -0.0107 -0.0052 -0.0059
(-0.0193, -0.0055) (-0.0230, 0.0016) (-0.0121, 0.0018) (-0.0182, 0.0063)
ML-Face 0.4965*** 0.4898*** 0.4873*** 0.4875***
(0.4217, 0.5713) (0.4140, 0.5655) (0.4115, 0.5631) (0.4116, 0.5634)
Constant 1.0282*** 1.1356*** 1.1181*** 0.6615*** 0.6492*** 0.6932*** 0.7010***
(0.9224, 1.1340) (1.0296, 1.2416) (0.9700, 1.2662) (0.5431, 0.7799) (0.5289, 0.7696) (0.5674, 0.8191) (0.5402, 0.8618)
F(Messy + Groomed) X X 4.427* (0.01198) X X X 0.7578 (0.4688)
Observations 8,466 8,466 8,466 8,466 8,466 8,466 8,466
Adjusted R2 0.1169 0.1172 0.1171 0.1285 0.1285 0.1286 0.1285
F Statistic 57.0530*** (df = 20; 8445) 57.1642*** (df = 20; 8445) 54.4399*** (df = 21; 8444) 63.4215*** (df = 20; 8445) 60.4423*** (df = 21; 8444) 60.4764*** (df = 21; 8444) 57.7215*** (df = 22; 8443)
Note: p<0.1; p<0.05; p<0.01
† We encode skin-tone as a continuous variable increasing in brightness

Regression Table 05 - Regressing on ML_Face

Table No.5 - Regression on ML Face
Dependent variable:
ML-Face
(1) (2) (3) (4) (5)
SexM -0.1154*** -0.1138*** -0.1189*** -0.1176***
(-0.1198, -0.1110) (-0.1182, -0.1094) (-0.1232, -0.1145) (-0.1220, -0.1132)
RaceB -0.0116 -0.0097 -0.0068 -0.0053
(-0.0316, 0.0084) (-0.0296, 0.0102) (-0.0264, 0.0129) (-0.0249, 0.0144)
Age -0.0009*** -0.0007*** -0.0004*** -0.0004***
(-0.0011, -0.0007) (-0.0009, -0.0005) (-0.0006, -0.0002) (-0.0006, -0.0002)
† skin-tone 0.2747*** 0.2681*** 0.2869*** 0.2845***
(0.2012, 0.3481) (0.1951, 0.3412) (0.2147, 0.3591) (0.2123, 0.3567)
Attractiveness 0.0011 -0.0024*
(-0.0011, 0.0034) (-0.0046, -0.0001)
Competence 0.0067*** 0.0040***
(0.0043, 0.0091) (0.0016, 0.0064)
Dominance -0.0033*** -0.0025**
(-0.0050, -0.0016) (-0.0043, -0.0008)
Trustworthiness 0.0052*** 0.0035**
(0.0029, 0.0075) (0.0013, 0.0058)
Well-groomed 0.0140*** 0.0085*** 0.0075***
(0.0106, 0.0174) (0.0053, 0.0116) (0.0042, 0.0108)
Messy -0.0019 -0.0096*** -0.0091***
(-0.0051, 0.0012) (-0.0126, -0.0067) (-0.0121, -0.0061)
Constant 0.8678*** 0.8245*** 0.6866*** 0.8485*** 0.8415***
(0.8459, 0.8897) (0.8008, 0.8483) (0.6576, 0.7155) (0.8135, 0.8836) (0.8063, 0.8767)
F(Messy + Groomed) X X 101.11*** (<2.2e-16) 157.16*** (<2.2e-16) 111.91*** (<2.2e-16)
Observations 8,466 8,466 8,479 8,466 8,466
Adjusted R2 0.2057 0.2162 0.0231 0.2340 0.2362
F Statistic 314.1893*** (df = 7; 8458) 213.2368*** (df = 11; 8454) 101.1098*** (df = 2; 8476) 288.3168*** (df = 9; 8456) 202.3818*** (df = 13; 8452)
Note: p<0.1; p<0.05; p<0.01
† We encode skin-tone as a continuous variable increasing in brightness

Verifying that we can interchange Dummies with P-hats

Table No.6 - Including Charge LM vs. Charge Dummies
Dependent variable:
Release-Final-Outcome
(1) (2) (3) (4)
sexM -0.0009 -0.0520*** 0.0010 -0.0539***
(-0.0205, 0.0188) (-0.0702, -0.0339) (-0.0189, 0.0210) (-0.0721, -0.0356)
raceB 0.0278 0.0259 0.0256 0.0257
(-0.0526, 0.1082) (-0.0549, 0.1068) (-0.0548, 0.1061) (-0.0553, 0.1067)
age -0.0002 -0.0005 -0.0001 -0.0005
(-0.0010, 0.0006) (-0.0013, 0.0002) (-0.0008, 0.0007) (-0.0012, 0.0003)
skin_tone_cont 0.3443* 0.4656*** 0.3598** 0.4754***
(0.0490, 0.6396) (0.1690, 0.7623) (0.0643, 0.6553) (0.1783, 0.7724)
risk_pred_prob -0.7587*** -0.8065*** -0.7759*** -0.8268***
(-0.8639, -0.6534) (-0.9122, -0.7008) (-0.8815, -0.6704) (-0.9329, -0.7208)
attractiveness 0.0048 0.0053 0.0041 0.0049
(-0.0044, 0.0139) (-0.0039, 0.0145) (-0.0050, 0.0133) (-0.0044, 0.0141)
competence -0.0014 0.0016 -0.0011 0.0018
(-0.0111, 0.0083) (-0.0082, 0.0113) (-0.0108, 0.0086) (-0.0080, 0.0115)
dominance -0.0037 -0.0052 -0.0037 -0.0052
(-0.0108, 0.0033) (-0.0122, 0.0019) (-0.0107, 0.0034) (-0.0123, 0.0018)
trustworthiness 0.0078 0.0099* 0.0080 0.0108*
(-0.0015, 0.0171) (0.0006, 0.0193) (-0.0013, 0.0173) (0.0014, 0.0201)
p_hat_charge 1.0713*** 1.1112***
(1.0036, 1.1389) (1.0434, 1.1791)
felony -0.2012*** -0.2026***
(-0.2173, -0.1851) (-0.2188, -0.1863)
gun_crime 0.0383** 0.0427**
(0.0099, 0.0668) (0.0140, 0.0713)
drug_crime 0.0409*** 0.0348***
(0.0213, 0.0604) (0.0151, 0.0544)
violent_crime -0.1623*** -0.1698***
(-0.1881, -0.1365) (-0.1958, -0.1439)
property_crime -0.0281*** -0.0339***
(-0.0451, -0.0112) (-0.0510, -0.0168)
arrest_year2018 0.0271*** 0.0337***
(0.0111, 0.0431) (0.0176, 0.0498)
arrest_year2019 0.0561*** 0.0883***
(0.0356, 0.0767) (0.0682, 0.1084)
p_hat_cnn 0.4685*** 0.4965***
(0.3957, 0.5412) (0.4217, 0.5713)
Constant -0.2360*** 0.1323* 0.6615*** 1.0780***
(-0.3639, -0.1081) (0.0171, 0.2474) (0.5431, 0.7799) (0.9768, 1.1791)
Observations 8,466 8,466 8,466 8,466
Adjusted R2 0.1291 0.1177 0.1285 0.1163
F Statistic 90.6426*** (df = 14; 8451) 87.8261*** (df = 13; 8452) 63.4215*** (df = 20; 8445) 59.6560*** (df = 19; 8446)
Note: p<0.1; p<0.05; p<0.01
† We encode skin-tone as a continuous variable increasing in brightness

Looking at distirbutional differences between Male - Female

Looking at distirbutional differences between Felony - Non Felony