Le résumé en français n'a pas été communiqué par l'auteur. / Administrative registers maintained by statistical offices on vastly heterogeneous firms have much untapped potential to reveal details on sources of productivity of firms and economies alike. It has been proposed that firm-level shocks can go a long way in explaining aggregate fluctuations. Based on novel monthly frequency data, idiosyncratic shocks are able to explain a sizable share of the Finnish economic fluctuations, providing support to the granular hypothesis. The global financial crisis of 2007-2008 has challenged the field of economic forecasting, and nowcasting has become an active field. This thesis shows that the information content of firm-level sales and truck traffic can be used for nowcasting GDP figures, by using a specific mixture of machine learning algorithms. The agency problem lies at the heart of much of economic theory. Based on a unique dataset linking owners, CEOs and firms, and exploiting plausibly exogenous variations in the separation of ownership and control, agency costs seem to be an important determinant of firm productivity. Furthermore, the effect appear strongest in medium-sized firms. Enterprise group structures might have important implications on the voluminous literature on firm size, as large share of SME employment can be attributed to affiliates of large business groups. Within firm variation suggests that enterprise group affiliation has heterogeneous impacts depending on size, having strong positive impact on productivity of small firms, and negative impact on their growth. In terms of aggregate job creation, it is found that the independent small firms have contributed the most. The results in this thesis underline the benefits of paying attention to samples encompassing the total population of firms. Researchers should continue to explore the potential of rich administrative data sources at statistical offices and strive to strengthen the ties with data producers.