Key points are not available for this paper at this time.
Entrepreneurship represents a pivotal pathway for transforming knowledge into innovation and economic value; however, the factors underlying this process remain a “black box”. Drawing on the Cognition, Capability, and Commitment (3C) framework, this study employs a data-driven machine learning approach—specifically, a decision tree algorithm—to systematically identify the potential predictors of Total Early-stage Entrepreneurial Activity (TEA). Using primary data from more than 70,000 individuals in the 2020 Global Entrepreneurship Monitor (GEM), supplemented with 2021 data for robustness checks, the results suggest that various forms of entrepreneurial cognition, such as necessity-driven motivations, family tradition, and perceptions of opportunity scarcity, emerge as prominent potential predictors. In contrast, capability-related variables exhibit comparatively lower predictive weights. These findings may advance theoretical understanding of entrepreneurial patterns and offer potential actionable implications for policymakers seeking to foster entrepreneurship. Moreover, they highlight the utility of machine learning in uncovering complex, non-linear interactions between motivational and contextual factors, thereby providing complementary exploratory insights beyond conventional GEM-based analyses.
Huang et al. (Mon,) studied this question.