Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
SummaryIn panel data regression models, it is often not reasonable to expect all cross-sectional units to have identical responses to explanatory variables, or that all relevant variables have been properly accounted for. These concerns have recently motivated the use of interactive effects models with heterogeneous slopes. The workhorse of this literature is the common correlated effects approach, which assumes that both effects and slopes are randomly distributed. The current paper argues that the restrictions implied by this assumption are likely unreasonable in many applications, and that there is a need to allow for nonrandom heterogeneity.