Measurement Problem of Enterprise DigitalTransformation: New Methods and FindingsBased on Large Language Models

Jin Xingye1, Zuo Congjiang1, Fang Mingyue*2, Li Tao1, Nie Huihua3

1 School of Economics, Central University of Finance and Economics (CUFE), Beijing, China

2 College of Economics and Management, China Agricultural University (CAU), Beijing, China

3 School of Economics, Renmin University of China (RUC), Beijing, China

Abstract: Despite broad consensus on the importance of enterprise digital transformation,significant discrepancies persist regarding its actual effects. This divergence stems primarilyfrom two key measurement challenges: (1) a lack of clear and consistent definitions ofenterprise digital transformation, and (2) a lack of rigorous and accurate measurementmethodologies. These shortcomings lead to research findings that are incomparable,difficult to replicate, and often conflicting. To effectively address the aforementionedchallenges, this paper employs machine learning and large language models (LLMs) toconstruct a novel set of indicators for enterprise digital transformation. The work beginsby manually annotating sentences from annual reports of listed companies in Chinafrom 2006 to 2020. These labeled sentences are then used to train and fine-tune severalmachine learning models, including LLMs. The ERNIE model, demonstrating the bestclassification performance among the models tested, is selected as the sentence classifier topredict sentence labels across the full text of the annual reports, ultimately constructing theenterprise digital transformation metrics. Both theoretical analysis and multiple data crossvalidationsdemonstrate that the metrics developed in this paper are more accurate thanexisting approaches. Based on these metrics, the paper empirically examines the impact ofenterprise digital transformation on financial performance. Our findings reveal three keypoints: (1) enterprise digital transformation significantly enhances financial performance,with big data, AI, mobile internet, cloud computing, and the Internet of Things (IoT) allplaying a significant role; however, blockchain technology does not show a significanteffect; (2) the significant positive effect of digital transformation on financial performanceis primarily observed in firms with weaker initial financial performance; and (3) enterprisedigital transformation improves financial performance mainly through enhancing efficiencyand reducing costs. This research has practical implications for promoting enterprise digitaltransformation and fostering high-quality economic development.

Keywords: Enterprise digital transformation; digital economy; digital technology; AI; large

JEL Classification Codes: O33, D24

DOI: 10.19602/j.chinaeconomist.2025.03.04