How does digital tax administration affect R&D manipulation? Evidence from dual machine learning

Abstract

Governing firms' R&D manipulation is crucial amid the prevalent opportunism during innovation in emerging economies. This study employs data on listed companies in China from 2010 to 2021 and uses the Golden Tax Project III as a quasi-natural experiment to estimate the effect of digital tax administration (DTA) on R&D manipulation via a dual machine learning model. Based on the capability, opportunity, and motivation-behavior (COM-B) theoretical framework, we explore the key pathways and identify the heterogeneous effects of DTA on various R&D manipulation motivations and directions. The results indicate the following: (1) DTA can regulate firms' R&D manipulation behavior and incentivize innovation quality and diversity. (2) DTA reduces R&D manipulation by enhancing accounting information quality and transmission efficiency in capital markets (internal and external information channels), strengthening tax audits (external monitoring channels), and curbing managerial self-interest and tax avoidance motives (internal motivation channels). (3) The DTA primarily functions as a governance instrument to address manipulation in various directions driven by tax avoidance and executive self-interest motives. These findings offer valuable insights for exploring governance models to address R&D manipulation in emerging economies.