Inter-Firm Similarity and Influence in the Latent Space of Business Strategy: A Textspatial Modeling Approach

Firm strategy is widely recognized as a key determinant of firm performance. But firm strategy is an intangible quantity whose measurement, modeling and analysis have proven to be a challenge. Since firms most often express and communicate firm strategy in descriptive terms and in textual form, we develop and demonstrate an objective and scalable approach to measure and analyze the impact of stated firm strategy on firm performance, within a generalized linear model framework. We do so by (i) invoking text analysis procedures to tokenize and extract strategic content from raw text data, (ii) using latent space models to transform similarities and differences in firms' strategic vocabulary into firm locations in a large-dimensional latent, strategy space, and (iii) drawing upon procedures in spatial econometrics for estimation, inference and interpretation. Specifically, we invoke a class of spatial autoregressive models applied to large-dimensional latent spaces. We label our approach textspatial modeling. We implement our approach on data from the Fortune 1000 public US firms and find a statistically significant performance differential across firms attributable to elements of firm strategy.