Working Paper: NBER ID: w11739
Authors: Bronwyn H. Hall; Jacques Mairesse; Laure Turner
Abstract: The identification of age, cohort (vintage), and period (year) effects in a panel of individuals or other units is an old problem in the social sciences, but one that has not been much studied in the context of measuring researcher productivity. In the context of a semi-parametric model of productivity where these effects are assumed to enter in an additive manner, we present the conditions necessary to identify and test for the presence of the three effects. In particular we show that failure to specify precisely the conditions under which such a model is identified can lead to misleading conclusions about the productivity-age relationship. We illustrate our methods using data on the publications 1986-1997 by 465 French condensed matter physicists who were born between 1936 and 1960.
Keywords: scientific productivity; age effects; cohort effects; period effects; panel data
JEL Codes: C23; O31; J44
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| age (J14) | productivity (O49) |
| cohort (C92) | productivity (O49) |
| period (C41) | productivity (O49) |
| age + cohort + period (J11) | productivity (O49) |
| cohort + period (C41) | age (spurious) (J14) |
| model choice (C52) | peak age of productivity (J26) |