This paper investigates whether firms innovate persistently or discontinuously over time using an innovation panel data set on German manufacturing and service firms for the period 1994?2002. We find that innovation behaviour is permanent at the firm?level to a very large extent. Using a dynamic random effects discrete choice model and a new estimator recently proposed by Wooldrigde (2005), we further shed some light on the driving forces for this phenomenon. The econometric results confirm the hypothesis of true state dependence for manufacturing as well as for service sector firms. In addition to past innovation experience, the results further highlight the important role of knowledge provided by skilled employees and unobserved individual heterogeneity in explaining the persistence of innovation.
innovation persistence state dependence unobserved heterogeneity dynamic random effects panel probit model