In a study of vocational training effects on earnings, researchers use propensity score matching combined with outcome regression (doubly-robust estimation). For a participant:

• Propensity score: P(Training=1|X) = 0.70
• Observed earnings (trained): Y = $36,000
• From outcome regression, predicted earnings if not trained: $\hat{Y}_0 = \$30,000$
• Matched non-participant's earnings: $31,500

Which estimates of the treatment effect are reasonable?