Knowledge flows through text data: the case of Milan’s tech ecosystem

Abstract Investigating the role of proximity between firms is crucial, as it can influence performance through spillover effects. This paper makes two main contributions. First, we introduce a replicable web-based methodology for measuring different types of proximity among firms. We draw on text data from company websites to assign firms vectors of keywords reflecting their industrial specializations, knowledge and competencies, and adopted technologies. These vectors are then used to compute cosine similarity. To account for knowledge exchange, the resulting linear similarity measures are transformed into an inverted U-shaped relationship, as suggested by theory. Second, we assess and compare the spillover effects arising from non-geographical proximities and their impact on firms' performance. We construct three proximity matrices (industrial, cognitive, and technological) and employ spatial models to distinguish and compare the informational contribution of each matrix in capturing spillovers. The findings confirm that proximity, through both knowledge exchange and competitive pressures, significantly affects firm performance. This indicates that spillovers arising from non-geographical proximities are important indirect drivers of firm performance and should not be underestimated.