The influence of ChatGPT on artificial intelligence related crypto assets: Evidence from a synthetic control analysis

Article Source

Abstract

The introduction of OpenAI's large language model, ChatGPT, catalyzed investor attention towards artificial intelligence (AI) technologies, including AI-related crypto assets not directly related to ChatGPT. Utilizing the synthetic difference-in-difference methodology, we identify significant "ChatGPT effects,” with AI-related crypto assets experiencing average returns ranging between 10.7% and 15.6% (35.5% to 41.3%) in the one-month (two-month) period after the ChatGPT launch. Furthermore, Google search volumes, a proxy for attention to AI, emerged as critical pricing indicators for AI-related crypto assets post-launch. We conclude that investors perceived AI-related crypto assets as possessing heightened potential or value after the launch, resulting in higher market valuations.

Introduction

The launch of OpenAI's ChatGPT on November 30, 2022, marks a significant milestone in the advancement of Artificial Intelligence (AI).1 ChatGPT is a state-of-the-art transformer-based large language model with sophisticated natural language processing capabilities, enabling real-world applications across diverse domains. Trained on a vast textual corpus, the ChatGPT model can generate human-like responses to user inputs known as prompts (OpenAI, 2022). ChatGPT exemplified its potential through superior performance in a Google interview for a highly sought-after role (Bhaimiya, 2023) and successful completion of examinations for law and business schools (Murphy Kelly, 2023). Its utility has been demonstrated by its impressive popularity, exceeding 100 million active users in January 2023, establishing a record for the fastest-growing user base in history (Savitz, 2023).2

The ground-breaking ChatGPT technology has stimulated commercial AI development (Tyrrell, 2022) and catalyzed digitalization initiatives (Brown, 2022). Media coverage predicts integration of ChatGPT into Microsoft's Bing search engine could position it as a formidable online search market competitor, compelling other tech giants like Google and Baidu to prioritize AI development (Elias, 2023). These companies' reactions to ChatGPT signaled an elevated perceived value of AI technology among investors. For example, c3.ai, a software company specializing in AI, saw its share price surge 28% following the announcement of ChatGPT integration into its suite of AI tools (Fox, 2023).3 The influence of ChatGPT transcends directly related companies to the wider AI sector, as evidenced by the rising share prices of BigBear.ai and SoundHound AI following the ChatGPT launch, despite their lack of association to ChatGPT (Singh and Biwas, 2023), signaling that investors perceived AI-related companies as possessing heightened potential or value. Moreover, empirical research has further shown that AI-focused exchange-traded funds exhibited a name-premium of around 0.4% even before the ChatGPT launch, indicating that investors recognized the growing importance of AI as a valuable technology (Wu and Chen, 2022).

Previous academic studies have examined ChatGPT's significance in different fields, including research (Dowling and Lucey, 2023; Thorp, 2023; Zaremba and Demir, 2023), software engineering (Sobania et al., 2023; Stokel-Walker, 2023), and education (Susnjak, 2022). However, despite growing interest, there is a noticeable lack of empirical studies exploring the potential impact of ChatGPT and its subsequent hype on AI-related financial assets. A salient study revealed 90% positive abnormal returns for a select group of AI-themed cryptocurrencies (Ante and Demir, 2023).

In light of this context, this paper aims to evaluate the impact of the ChatGPT launch on the narrative trading of AI-related crypto assets (hereafter referred to as AI-assets). This focus is crucial because, while some companies listed on the stock market may integrate AI into their offerings, only a limited number specialize in AI product and service provision. In contrast, a distinct cohort of crypto assets prioritize AI as a fundamental aspect of operations.4 Consequently, this study investigates how the introduction of ChatGPT influenced price discovery and market perceptions of AI-crypto-assets. In pursuance of this objective, we empirically investigate if: (1) returns of AI-assets and non-AI-assets were equivalent before the ChatGPT launch, (2) a disparity in investor perception between AI and non-AI-assets post-launch, and (3) this divergence can be explained by changes in Google search volumes (financial news wires) as a proxy for retail (institutional) investor attention towards AI.

Firstly, we utilize the synthetic difference-in-difference (SDID) approach of Arkhangelsky et al. (2021) and Clarke et al. (2023) to examine the relative evolutionary performance of a large set of AI-assets relative to controls.5 Results demonstrate that AI-assets exhibited positive average treatment effects (ATTs) of 10.7% to 15.6% (35.5% to 41.3%) in the one-month (two-month) period following the ChatGPT launch. This positive effect is more meaningful in context, as the broader cryptocurrency market was characterized by a bear market and extreme investor uncertainty during that time, emphasizing the significant impact of ChatGPT on the narrative trading of AI-assets.6 Secondly, we reveal a spillover effect from attention generated by ChatGPT to AI-assets that have no direct relation to ChatGPT, contributing to ongoing research exploring the role of sentiment and hype in cryptocurrency markets (Nepp and Karpeko, 2022; Shahzad et al., 2022; Subramaniam and Chakraborty, 2020) as well as the literature on quality signals in price promotion and crypto market efficiency (Ante, 2023; Brunnermeier, 2005). Thirdly, a panel fixed-effects model furnishes empirical support for Google search volume data via Google Trends as a proxy for retail investor attention, serving as reliable price indicators for AI-assets only after the launch, highlighting the importance of incorporating such data into market analyses for effective decision-making. Lastly, we present empirical evidence that financial news wires, as a proxy for institutional investor attention, exerted limited influence on AI-crypto assets following the launch.

From a theoretical standpoint, there are several potential channels contributing to the positive price effect for AI-assets post-ChatGPT launch. Firstly, market efficiency theories assert that asset valuations are linked to perceived quality (Fama, 1970; Lo, 2004), and the comprehensive media coverage of the ChatGPT launch may serve as a quality proxy, leading retail investors to value AI-assets more highly in accordance with signaling theory (Spence, 1973). This aligns with the notion that retail investors, prone to cognitive biases and heuristics, tend to overreact to news or events (Ante, 2023). Secondly, institutional investors, acknowledging ChatGPT's potential and influenced by similar market psychology, may have directed capital toward AI-related projects, however institutional interest in cryptocurrencies waned. Thirdly, ChatGPT may have fostered information diffusion effects within the cryptocurrency market by equipping retail investors with the capacity to distill complex and technical concepts, thus facilitating more informed investment decisions (Binance, 2023). Finally, we acknowledge network effects arising from ChatGPT's launch may have spurred a self-sustaining growth cycle for AI-assets, potentially attracting more users and boosting demand and valuations.

Section snippets

Data

Our dataset spans October 1, 2022, to January 31, 2023. Asset returns, rt, are defined as the first difference of the natural log of the asset price, pt, relative to the previous day: rt = ln(pt/pt − 1). Panel A in Table 1 presents descriptive statistics for two cohorts of crypto asset returns: a curated cohort of 16 crypto assets classified as ‘AI’-related by coingecko.com (GAI) and a broader cohort of 86 crypto asset returns categorized as ‘AI and Big Data’-related by coinmarketcap.com (CAI)...

Baseline difference-in-difference estimates

We begin by estimating the traditional fixed-effects difference-in-difference (DID) model to provide preliminary estimates for the effects of the introduction of ChatGPT on AI-assets (outlined in Table 1, Panel A) as compared to corresponding non-AI asset control groups (outlined in Table 1, Panels B and C). The treatment groups GAI and CAI were exposed to the ChatGPT launch as of November 30, 2022. Models (1) to (6) in Table 2 reveal significant (mixed results) for ATTs of AI-assets in the...

Conclusion

This paper demonstrates the ChatGPT launch had a significant impact on the performance of AI-related crypto-assets, despite the overall cryptocurrency market being in a bearish state and risk-averse investor appetites. Using synthetic difference-in-differences, we found average price increases of at least 10.7% (35.5%) in the one-month (two-month) period following the launch. Firstly, we contribute to the literature on the impact of technology launches on financial markets and the role of...