Title and Abstract
Reassessment of classic case studies in labor economics with new instrument-free methods
It is well-known that instrument validity cannot decisively be assessed by over-identification restrictions tests. We demonstrate that even serious instrument invalidity will frequently not be detected by the Sargan test, whereas instrumental variables based inference is shown to be seriously vulnerable regarding even mild invalidity of the instruments. These disturbing findings are confronted with the qualities and limitations of recently developed instrument-free methods. These may provide much more robust subject matter inferences in empirical research. Besides, they provide more decisive evidence on the invalidity of exclusion restrictions than Sargan-Hansen tests can. This all is illustrated for two classic case studies in labor economics focusing on the effect of education on earnings.
Jan F. Kiviet
University of Amsterdam;
University of Exeter