White by Default: Bias in Criminal Racial Assignment (uncorrelated.xyz)

A recent study revealed significant racial misclassification in criminal records, where a model trained on 1.5 million datasets identified that 29% of individuals predicted to be Hispanic were incorrectly classified as White by authorities. Utilizing a multinomial logistic regression model with an impressive 92.76% accuracy in classifying individuals into three racial categories (Black, White, Hispanic), the research demonstrated that correcting these mislabels could increase Hispanic representation in criminal records by 31% while reducing White rates by 6%. This issue persisted even at high confidence levels, raising concerns about systemic biases that skew crime statistics. The significance of this study lies in its methodological rigor, employing advanced statistical techniques and comprehensive data collection from U.S. Department of Corrections databases. The findings challenge the narrative of intentional racial bias against White individuals, suggesting that the misclassification is more reflective of administrative errors rather than malicious intent. Furthermore, the research highlights underlying biases in the criminal justice system, marking an important step in understanding and addressing disparities in racial categorization and its implications for public policy and social justice.