To assess the model's ability to identify members of each ethnic group, we examined voters with the highest probability scores. For each ethnic category, we look at voters scoring in the top 20% (highest scores) and measure what percentage of them actually belong to that ethnicity. For majority groups like White voters (59% of the validation sample), the top 20% of White probability scores is highly precise: 97% of voters in this segment are actually White–a 64% relative increase compared to sampling from the population at large (97% ÷ 59% = 1.64× lift). This high precision is possible because the group is large enough (59%) that the model can selectively rank the highest-confidence members above the 80th percentile threshold. For smaller demographic groups, the interpretation differs due to their lower base rates. For example, AAPI voters represent 9% of the validation sample. Even if the model perfectly identified every AAPI voter and ranked them highest, the top 20% of scores would still contain non-AAPI voters (since 20% of the population exceeds the 9% AAPI base rate). Our model achieves 41% AAPI concentration in the top 20% of AAPI scores–meaning the model has enriched AAPI representation by 4.6× compared to the overall population rate (41% ÷ 9% = 4.6× lift). Similar patterns hold for Black voters (17% base rate, 77% in top quintile, 4.5× enrichment), Hispanic voters (15% base rate, 69% in top quintile, 4.6× enrichment), and Native American voters (1% base rate, 2.4% in top quintile, 2.4× enrichment). At the other end of the distribution, individuals in the bottom 20% of scores for each ethnicity show significantly lower probabilities to belong to that ethnic category (7.4% for White, <1% for all other groups).