Adressing Replication and Model Uncertainty: A Bayesian Averaging Approach Applied to Innovation Survey Data

Many fields of strategic management are subject to an important degree of model uncertainty. This is because the true model, and therefore the selection of appropriate explanatory variables, is essentially unknown. Drawing on the literature on the determinants of innovation, and by analyzing innovation survey data for France, Germany and the UK, we conduct a ‘large-scale’ replication using the Bayesian averaging approach of classical estimators. Our method tests a wide range of determinants of innovation suggested in the prior literature, and establishes a robust set of findings on the variables which shape the introduction of new to the firm and new to the world innovations. We provide some implications for innovation research, and explore the potential application of our approach to other domains of research in strategic management.