A comment on "Network formation and efficiency in linear-quadratic games: An experimental study"

We replicate Horváth (2025), experimentally studying link formation and effort in a linear-quadratic game with positive externalities. Across five treatments, subjects exert 38-97 percent more effort than the Nash benchmark yet create too few links, depressing payoffs. In groups of five, the complete network appears in roughly 25 percent of final rounds (66-76 percent if deviations of ±2 links are allowed); in groups of nine it is almost never reached. Larger groups and lower link costs fail to improve connectivity. Following the original procedures and analysis step-for-step, our replication reproduces the sign, magnitude, and statistical significance of every reported effect. Robustness checks-learning, benefit salience, group benchmarking, alternative clustering, and multiple link-formation specifications- confirm the core pattern: persistent over-provision of effort coupled with under-provision of links, generating substantial efficiency losses.