Effects of market conditions at time of enrollment on PhD completions.

***pStandard errors robust to arbitrary heteroskedasticity are presented in parentheses below each coefficient estimate.Results are from instrumental variables estimation, with dependent variable the first-differenced, logged ratio of PhD completions to first-time graduate student enrollment in U.S. biomedical sciences graduate programs six years prior. All explanatory variables are likewise measured six years prior, relative and first-differenced, and use the sixth lags of our two measures of relative industry R&D expenditures as instruments. As in Table 5, here the combination of first-differencing and use of lagged industry R&D estimates, for which data are only available 1998 onwards, limits the number of observation-years available (with their attendant lags) for time-series estimation.