<p>This paper employs machine learning to investigate the predictive power of educational attainment on new firm creation across U.S. counties. By using various supervised learning models trained on historical data (2018, 2019, and 2022) to predict 2023 establishment births, we find that the share of college-educated residents is consistently the most important predictor across all model types. In contrast, the share of high school dropouts is relatively unimportant and demonstrates inconsistent predictive power across different model specifications. These findings highlight that higher educational attainment serves as a robust predictive factor for establishment births at the local level, suggesting that counties with a high share of college-educated residents experience significantly higher rates of new firm creation.</p>