This paper analyses the determinants of different innovation types. Beside a wide range of firm characteristics also the effects of regional factors are estimated using three-level random effect logit models which account for the clustered and longitudinal structure of the data. The analyses contain three regional variables: the unemployment rate, the assessment of the region with reference to proximity to research and technology centres and universities and the rate of graduates in mathematics, informatics, natural sciences and technological sciences (MINT-graduates). The empirical basis is the IAB-Establishment Panel Survey 2006 to 2010. Process and radical innovations are significant affected by the unemployment rate and the share of MINT-graduates. The unemployment rate has also for some of the innovation combos a significant effect. The proportion of MINT-graduates is relevant for the probability of all 4 innovation types simultaneously.