So far, yet so close. Using networks of words to measure proximity and spillovers between firms

Textual data are the last frontier in the empirical literature on proximity between firms. While there are a growing number of studies using textual data, no robust methodology has yet emerged, nor has any attempt been made to compare the result‐ ing findings with standard measures of proximity based on existing classification systems. The purpose of this paper is threefold. First, we propose a methodology that can be an effective and applicable tool for measuring proximity between com‐ panies. Second, we compare the resulting indicator of proximity, which we refer to as “business” proximity, with industrial and technological proximity scores based on activity codes and technology adoption, respectively. Third, we use business prox‐ imity to explain economic performance, assuming that knowledge sharing can occur between employees working in similar firms. Having established the soundness of the methodology, the empirical results confirm the substantial information content of the descriptive texts and provide evidence on the likelihood of spillover effects between firms that are close in the business and geographical dimension.