Abstract
Using US data drawn from the 1978 young men NLS sample, this paper replicates a key finding reported by Rees and Shah in their analysis of self‐employment using large‐scale British data. A wage equation, which allows for sample selectivity, is estimated for subsamples of employees and the self‐employed. Despite differences between the data sets and the variables analysed, the NLS results support Rees and Shah's conclusion that there is positive selection bias in the observed earnings of employees.