How do Incumbents Respond to the Entry of AI? A Study in the Mutual Funds Industry

Abstract

This paper studies how incumbent firms respond to the emergence of AI as a process technology that performs tasks that humans can perform. Humans use different human capital skills to perform multiple tasks. Recent research suggests that AI could replace human labour in middle skilled jobs while capital and labour will be used by firms to perform low and high skill jobs respectively (Acemoglu and restrepo 2019). Yet, there is less clarity on firm level response to AI when the same skill can be used to perform multiple tasks. Hence, we adopt a question driven approach to answer the following questions (1) how do incumbents respond to the entry of AI/automation from outside the industry? (2) how does incumbent response vary by nature of the task subject to AI/automation? (3) how does incumbent response vary by skill specialization? (4) what are the performance consequences of incumbent sensitivity to AI/automation? We use the mutual funds industry as the context for our empirical examination. Using the emergence of algorithmic trading as an exogenous source of entry of AI, we study how mutual funds change their portfolio allocation in response to competition from AI. Further, we study how the change in holdings vary based on the nature of information available to managers to make decisions. We then study how the response of funds vary by the skill of the manager in picking stocks. Finally, we explore the difference in performance of funds based on managers’ response to the entry of AI. We find that (1) on average, mutual funds responded to the entry of algorithmic traders by reducing their exposure to affected stock; (2) this average response, however, masks considerable heterogeneity at both the task and skill-levels with funds exposure reduction increasing with greater information availability and decreasing with greater information ambiguity; (3) examining heterogeneity by skill specialization we observe that funds relying on timing skills of managers have a stronger response in reducing their exposure to AT affected stocks; (4) examining performance effects, we note that funds that move away from AT affected stocks modestly outperform stocks that are less responsive to algorithmic trading.