<p>Advances in technology and web-based survey services have expanded the use of online experiments in the social sciences. Although cost-effective and diverse, they raise concerns about integrating data with laboratory studies because of decisionmaking differences. This study examined whether cognitive ability measures improve treatment effect estimation in laboratory experiments in which random assignment is difficult. Using data from dictator and ultimatum game experiments with adult participants at Kansai University’s Economic Experiment Laboratory of the Research Institute for Socionetwork Strategies, we applied propensity score matching with demographic variables (age, gender, and education) and cognitive ability measures (Raven’s Progressive Matrices score and Cognitive Reflection Test score). Cognitive ability significantly influenced allocation behavior. Models including these measures yielded the highest treatment effects when comparing the dictator game and hot treatment of the ultimatum game, and reduced <i>p</i> values for comparisons between hot and cold decision-making methods. These findings demonstrate that cognitive ability enhances estimation accuracy and support the complementary roles of laboratory and online experiments.</p>