Artificial intelligence, investment inefficiency and financial mismatch: causal inference using a double machine learning framework
Yue Zhang, Yu Wei, Jingyao Sa, Cheng ZhaoPurpose
This study aims to examine whether the adoption of artificial intelligence (AI) can alleviate financial mismatches and enhance capital allocation efficiency through firms' investment decisions, particularly in the context of rapid digitalization and structural financial reforms.
Design/methodology/approach
The study develops a “technology–efficiency–allocation” analytical framework to explain how AI promotes investment efficiency and mitigates financial mismatches. Using the Skip-gram model in Word2vec, firm-level indicators of AI application are constructed to overcome the biases of patent-based or industry-level measures. Inefficient investment is introduced as an instrumental variable, and a double machine learning (DML) framework is employed to control for high-dimensional covariates and specification errors, ensuring credible causal identification.
Findings
The study develops a “technology–efficiency–allocation” analytical framework to explain how AI promotes investment efficiency and mitigates financial mismatches. Using the Skip-gram model in Word2vec, firm-level indicators of AI application are constructed to overcome the biases of patent-based or industry-level measures. Inefficient investment is introduced as an instrumental variable, and a DML framework is employed to control for high-dimensional covariates and specification errors, ensuring credible causal identification.
Originality/value
This study provides firm-level empirical evidence clarifying AI's role in reducing financial mismatches and improving capital allocation efficiency. It extends the literature by integrating advanced textual measurement of AI application with causal machine learning methods and offers practical implications for promoting technological integration and enhancing financial efficiency in emerging economies.