Social experiments are powerful sources of information about the effectiveness of interventions. In practice, initial randomization plans are almost always compromised. Multiple hypotheses are frequently tested. "Significant" effects are often reported with p-values that do not account for...

"It is well known that children reared in non-intact families on average have less favorable educational outcomes than children reared in two-parent families. Evidence from the United States and Sweden indicates that living in a non-intact family is correlated with lower educational attainment....