Self-Report Bias in Estimating Cross-Sectional and Treatment Effects.