Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
We propose conditions under which parameters of fixed-effect dynamic models are identified with unequally spaced panel data. Under predeterminedness, weak stationarity, and empirically testable rank conditions, AR(1) parameters are identified given the availability of “two pairs of two consecutive time gaps”, which generalizes “two pairs of two consecutive time periods”. This result extends to models with multiple covariates, higher-order autoregressions, and partial linearity. Applying our method to the NLS Original Cohorts: Older Men, where personal interviews took place in 1966, 67, and 69, we analyze the earnings dynamics in the old time, and compare the results with more recent ones.