Selection bias: high-GPA students may be inherently more motivated and would improve regardless of scholarship

Regression to the mean: students at the threshold (near 3.5) are more likely to improve naturally

Confounding: unmeasured variables (family support, prior tutoring) may explain both GPA threshold qualification and subsequent improvement

RDD eliminates these biases by using the sharp cutoff; all above statements are incorrect