Remote work and firm productivity : Evidence from french collective agreements

We estimate the causal effect of remote work adoption on firm-level total factor productivity (TFP) in France. We construct a novel dataset by applying a fine-tuned large language model to classify the universe of firm-level collective agreements filed with the French government since 2017, identifying those that establish remote work policies. Matching these agreements to administrative balance sheet data yields a panel of 20,000 firms over 2018-2025. Using a staggered difference-in-differences design with propensity score weighting, we find that remote work adoption increases TFP by 3.9 percent after five years, with gains materializing gradually over time. The productivity-remote work relationship exhibits an inverted-U shape : the optimal intensity is two days per week (+3.2%), while full-time remote work reduces productivity by 8.0 percent. Effects are concentrated among large firms and mid-sized enterprises, with no gains for SMEs. Sectoral heterogeneity is substantial : finance and insurance (+14.1%), information and communication (+9.4%), and professional services (+5.4%) benefit most, while manufacturing experiences productivity losses (-2.4%).