Determinants of entrepreneurial intention: a GEM data based study in China

Purpose The main purpose of this study is to investigate whether a series of factors such as human capital, social capital, and environmental and demographic variables affect the entrepreneurial intention of individuals in China. Design/methodology/approach To address this problem, a conceptual model is developed and empirically tested using data from a sample of 2,814 people from China. Unlike other studies that examine the determinants of entrepreneurial intentions in the Chinese context, the large sample data in this study comes from the widely recognized Global Entrepreneurship Monitor (GEM) national data. Furthermore, this study uses the CatBoost algorithm to enhance the model’s accuracy and to test the hypotheses, which optimizes the logistic regression approach adopted in studies in this area. Findings First, risk perception ability was not found to be associated with individuals’ entrepreneurial intentions, which is not consistent with the existing literature. Furthermore, the results show that social capital in terms of role model can be positively associated with entrepreneurial intentions, which provides support for the previous findings. Meanwhile, perceptions of the desirability of entrepreneurship, along with its high status and respect, can be related to individuals’ intention to start a business. Finally, this study optimizes the logistic regression (LR) model commonly used in this field and validates the hypotheses using the CatBoost algorithm. Research limitations/implications First, with respect to the sample data, like many other empirical studies that use secondary data (e.g. GEM data), this study is subject to some specific limitations. Secondly, in terms of research content, this paper lacks verification of the conceptual model’s impact mechanism, revealing some optimization space for its construction. Originality/value First, in the Chinese context, we use GEM large sample data to analyze and validate the relationship between external (i.e. environmental factors) and internal (i.e. human, social capital) factors and entrepreneurial intentions. Secondly, the original LR model was optimized using the CatBoost algorithm, and the accuracy of the model was significantly improved. In addition, the degree of each independent variable’s contribution and hypothesis can be analyzed and verified through the SHAPley Additive exPlanations (SHAP) algorithm. This approach may represent a methodological innovation to introduce these two aforementioned algorithms into research in the field of entrepreneurship.