AK 2024 Primary — logit
Likely-voter model for AK 2024 PRIM, training target election held Aug 20, 2024.
Methodology
Likely-voter model for 2024 PRIM, fitted as a logistic regression on voter-history and demographic features from the AK Division of Elections statewide voter file. Trained on n=461,685 eligible voters (target positive rate 18.7%); test ROC AUC 0.8892; Brier 0.0939. Features were standardized (zero-mean, unit-variance). Top positive coefficients: voted_2022_primary (+0.398), voted_2022_general (+0.384), voted_2020_primary (+0.256), ever_voted_municipal (+0.255), n_prior_special (+0.151). Top negative coefficients: years_since_last_vote (-15.496), share_polls (-13.104), share_absentee (-8.836), share_mail (-6.946), share_questioned (-5.283).
Standardized coefficients (sorted by |magnitude|)
| Rank | Feature | coef (std-scaled) |
|---|---|---|
| 1 | years_since_last_vote | -15.4959 |
| 2 | share_polls | -13.1037 |
| 3 | share_absentee | -8.8362 |
| 4 | share_mail | -6.9464 |
| 5 | share_questioned | -5.2829 |
| 6 | share_early | -4.9050 |
| 7 | voted_2022_primary | 0.3976 |
| 8 | voted_2022_general | 0.3842 |
| 9 | voted_2016_general | -0.2629 |
| 10 | voted_2020_primary | 0.2562 |
| 11 | ever_voted_municipal | 0.2546 |
| 12 | voted_2018_general | -0.1851 |
| 13 | n_prior_special | 0.1514 |
| 14 | years_registered | 0.1298 |
| 15 | ever_voted_primary | 0.1165 |
| 16 | age_at_election | 0.1156 |
| 17 | n_prior_municipal | 0.1107 |
| 18 | party_D | 0.0840 |
| 19 | n_prior_elections | 0.0799 |
| 20 | party_U | -0.0684 |
| 21 | n_prior_primaries | 0.0603 |
| 22 | ever_voted_general | 0.0601 |
| 23 | years_at_district | 0.0460 |
| 24 | party_OTHER | -0.0428 |
| 25 | party_R | 0.0416 |
| 26 | pra_flag_int | 0.0371 |
| 27 | party_L | -0.0305 |
| 28 | voted_2018_primary | 0.0256 |
| 29 | n_prior_generals | -0.0233 |
| 30 | voted_2020_general | -0.0232 |
| 31 | gender_M | -0.0183 |
| 32 | gender_F | 0.0182 |
| 33 | party_G | -0.0036 |
| 34 | gender_U | 0.0015 |
| 35 | party_N | 0.0000 |
Intercept: -2.3888. Coefficients are in standardized units (features were zero-mean / unit-variance scaled before fit).
Calibration — predicted vs. actual turnout by decile
| Decile (10 = highest LV) | n | Pred. p | Actual turnout |
|---|---|---|---|
| 10 | 57711 | 0.724 | 0.759 |
| 9 | 57711 | 0.498 | 0.474 |
| 8 | 57711 | 0.309 | 0.288 |
| 7 | 57710 | 0.161 | 0.159 |
| 6 | 57711 | 0.077 | 0.076 |
| 5 | 57711 | 0.036 | 0.038 |
| 4 | 57710 | 0.024 | 0.025 |
| 3 | 57711 | 0.020 | 0.018 |
| 2 | 57711 | 0.017 | 0.018 |
| 1 | 57710 | 0.009 | 0.018 |
Eligible scored voters bucketed into 10 equal-size deciles by predicted probability. A well-calibrated model has predicted p ≈ actual turnout in every decile.
Hyperparameters
{'C': 1.0, 'seed': 42, 'max_iter': 200, 'test_size': 0.2}
Trained on AK Division of Elections statewide voter file snapshot AK DoE 2025-05 at 2026-05-23 04:58 UTC.