Selection Bias and the Degree Effect
利用1972年美国高中班级纵向调查数据，评估大学学位效应（即完成大学相比前三年有特别高的回报），并采用有序概率模型校正选择偏差，发现结果高度依赖于识别模型的排除限制条件。
This paper evaluates the evidence for a college degree effect-a particularly high return to completing college compared to the first three years-in the National Longitudinal Study of the High School Class of 1972 (NLS-72). An ordered probit model of schooling choice is developed to correct for selection bias. If conventional exclusion restrictions are adopted, selection bias does not appear to account for the degree effect. However, use of Leamer's (1978, 1982) extreme bounds analysis shows that the results are very dependent on acceptance of the exclusion restrictions used to identify the model.