This document contains the two latest versions of the election table 01:

  1. Config 2: This has the state*party interaction
  2. Config 3: This has the party_win_share * state interaction

Table 01 - Config 2

This table summarizes the results from a new parameterization of the election_lm as well as the inclusion of two new features.

  1. Election LM now includes a state*party interaction
  2. Win_Rate_Prior captures a parties win record in that state over the years prior to the current (i.e only the past)
  3. Win_Rate_Total captures a parties win record over all years excluding the current (i.e. past and future)

I include these two new variables as controls and use the new election_lm as a benchmark below:

Table 01 - Version 02 - Election Regressions
Fit measured in adjusted R squared and AUC
Model Configuration Adjusted R Squared ROC AUC
Single Variable Model
Election LM 0.0666 0.6517163
Lower 95% C.I. 0.0429 0.6180325
Upper 95% C.I. 0.0928 0.6854001
Win_Rate_Prior 0.0016 0.5533115
0.0001 0.5186098
0.0110 0.5880133
Win_Rate_Total 0.0037 0.5665316
0.0003 0.5315549
0.0148 0.6015084
Sex 0.0153 0.5604767
0.0065 0.5314525
0.0331 0.5895009
Skine-Tone −0.0049 0.5454082
0.0156 0.5106945
0.0428 0.5801219
MTurk Features 0.0017 0.5433649
0.0024 0.5077725
0.0193 0.5789574
P_hat_cnn 0.0972 0.6814615
0.0707 0.6487163
0.1303 0.7142067
Combined Variable Model
Election LM + Win_Rate_Prior 0.0664 0.6566925
0.0461 0.6231589
0.0957 0.6902261
Election LM + Win_Rate_Total 0.0673 0.6624640
0.0472 0.6290332
0.0964 0.6958948
Election LM + Win_Rate_Prior + Win_Rate_Total 0.0665 0.6592791
0.0478 0.6257532
0.0982 0.6928049
+ Sex 0.0762 0.6725071
0.0578 0.6394033
0.1115 0.7056108
Election LM + Win Rates + Sex + P_hat_cnn 0.1510 0.7301885
0.1263 0.6994458
0.1939 0.7609312
Election LM + Win Rates + Sex + Skin-Tone 0.0682 0.6730208
0.0800 0.6399697
0.1348 0.7060720
Election LM + Win Rates + Sex + Skin-Tone + P_hat_cnn 0.1447 0.7348479
0.1489 0.7043409
0.2173 0.7653548
Election LM + Win Rates + Sex + Skin-Tone + MTurk 0.0687 0.6764715
0.0862 0.6435739
0.1438 0.7093691
Election LM + Win Rates + Sex + Skin-Tone + MTurk + P_hat_cnn 0.1455 0.7375033
0.1539 0.7071345
0.2247 0.7678721

Table 01 - Config 3

I now re-specify our baseline-lm to include the party-state specific win rate in place of the state variable.

Table 01 - Version 03 - Election Regressions
Fit measured in adjusted R squared and AUC
Model Configuration Adjusted R Squared ROC AUC
Single Variable Model
Election LM 0.1717 0.7385170
Lower 95% C.I. 0.1387 0.7082288
Upper 95% C.I. 0.2105 0.7688053
Win_Rate_Prior 0.0016 0.5533115
0.0000 0.5186098
0.0100 0.5880133
Win_Rate_Total 0.0037 0.5665316
0.0003 0.5315549
0.0148 0.6015084
Sex 0.0153 0.5604767
0.0059 0.5314525
0.0311 0.5895009
Skine-Tone −0.0049 0.5454082
0.0160 0.5106945
0.0422 0.5801219
MTurk Features 0.0017 0.5433649
0.0025 0.5077725
0.0198 0.5789574
P_hat_cnn 0.0972 0.6814615
0.0704 0.6487163
0.1282 0.7142067
Combined Variable Model
Election LM+ P_hat_cnn 0.2331 0.7800592
0.1397 0.7520213
0.2139 0.8080971
Election LM + Win_Rate_Prior 0.1723 0.7403064
0.1393 0.7100965
0.2117 0.7705162
Election LM + Win_Rate_Total 0.1789 0.7456350
0.1459 0.7155917
0.2199 0.7756782
Election LM + Win_Rate_Prior + Win_Rate_Total 0.1818 0.7456074
0.1503 0.7155438
0.2246 0.7756710
+ Sex 0.1856 0.7500246
0.1549 0.7201937
0.2300 0.7798555
Election LM + Win Rates + Sex + P_hat_cnn 0.2414 0.7871633
0.2121 0.7594227
0.2881 0.8149040
Election LM + Win Rates + Sex + Skin-Tone 0.1789 0.7564142
0.1767 0.7268846
0.2514 0.7859438
Election LM + Win Rates + Sex + Skin-Tone + P_hat_cnn 0.2369 0.7935588
0.2324 0.7662178
0.3087 0.8208999
Election LM + Win Rates + Sex + Skin-Tone + MTurk 0.1791 0.7581897
0.1827 0.7287806
0.2570 0.7875989
Election LM + Win Rates + Sex + Skin-Tone + MTurk + P_hat_cnn 0.2377 0.7958363
0.2377 0.7686166
0.3172 0.8230561