Machine Learning Insights into Structural Constraints and Scale Advantages: Total Factor Productivity-performance Dynamics in Pakistani and NASDAQ-listed High-tech Firms

This study investigates a productivity-regime mechanism in the total factor productivity (TFP)–performance relationship for high-tech firms. Using machine learning on a panel of 77 Pakistani and 271 NASDAQ-listed firms (2018–2022), the study finds that TFP enhances firm performance primarily after entering key model-identified productivity regimes (above approximately 1.0 in Pakistan) and when combined with sufficient firm size. This productivity–scale synergy is particularly pronounced in Pakistan, where structural constraints limit firms’ ability to convert efficiency gains into profitability without adequate scale. In contrast, NASDAQ-listed high-tech firms exhibit a more direct and stronger TFP–performance association, consistent with more supportive institutional environments. Interpretable machine learning (Random Forests, SHapley Additive exPlanations and partial dependence plots) uncovers nonlinearities and interaction effects hidden in conventional models, showing that firm size acts as a critical multiplier of efficiency gains in emerging markets. These insights provide actionable policy guidance for fostering growth in high-tech sectors across different development contexts.