Research
Working Papers:
Contagious crime: How cryptocurrency manipulation spills into stocks ”, Working Paper (Under Review).
(with Talis Putnins and Anirudh Dhawan)
We show that a type of market manipulation coordinated on pseudo-anonymous online forums (e.g., Telegram) and popularized in cryptocurrency markets has now spread to stock markets. These manipulations occur in Australia, India, and the US during 2020–2022 and generate millions of dollars of volume and large returns of 17%–107%. These manipulations are unique in that manipulators do not always explicitly deceive their victims. We also show that direct regulatory intervention in forums is effective in curbing this activity. The findings illustrate a negative externality of unregulated markets: breeding market manipulation that spills into other markets.
Experimental evidence on helping scam victims using persuasive messaging, Working Paper.
(with Talis Putnins, Robert Slonim, Inigo De Juan Razquin, and Zoey Isherwood)
A substantial proportion of scam victims refuse to acknowledge having been scammed, imposing material costs on banks in trying to help victims. This issue is far less understood than why individuals fall victim to scams in the first place. We use controlled experiments to test the effectiveness of different ways of framing communications to victims (i) reducing stigma of being scammed, (ii) reducing distrust of the financial institution, and (iii) increasing awareness about the consequences of denial. We find significant differences in how men and women respond to the stigma treatment, with a larger effectiveness for women. However, in the pooled sample of all participants, we find no significant differences between these three treatments. Using a stylized framework, we argue that individuals may engage in motivated reasoning to preserve short-term psychological comfort effectively choosing self-deception to avoid the immediate pain of acknowledging a loss.
Obscure disclosures: Strategic information concealment by substantial shareholders, Working Paper.
(with Talis Putnins, Sean Foley, and Vinay Patel)
Using a broad sample of substantial shareholder trades, we assess the concealment of private information through complex disclosures and its impact on market quality. Trades which are obscured in a handwritten format display characteristics expected of an informed trader, resulting in a significant market reaction upon announcement (i.e., stocks associated with handwritten disclosures have a cumulative abnormal return exceeding 12% over one year). Such complex disclosures are more common in stocks with higher levels of information asymmetry, are more likely to be released during high market volatility, and have longer release delays. Further, we observe a significant post-announcement drift in returns following handwritten filings, indicating that obscure informed disclosures undermine market efficiency.
From Disclosure to Deception: Detecting Greenwashing using LLM, Working Paper.
(with Talis Putnins)
In Australia’s $4T superannuation market, competition over sustainable products incentivizes greenwashing. We exploit the Australian Securities and Investments Commission’s (ASIC) inaugural greenwashing prosecution as a quasi-natural experiment, analyzing Product Disclosure Statements (PDS). Post-prosecution, 35% of funds reduce their ESG-related disclosures, a pattern consistent with unwinding prior greenwashing. We then test Large Language Models (LLMs) as supervisory tools for regulators to assess greenwashing risk. We find that advanced reasoning models like Gemini 2.5 Pro accurately identify high-risk documents, whereas base models like GPT 4o provide inconsistent or mediocre risk assessments. Our results have implications for regulators and investors to better scrutinize sustainability claims.
On the predictive power of tweet sentiments and attention on Bitcoin*
(with Sandy Suardi and Bin Liu)
This paper investigates the predictive power of information contained in social media tweets on bitcoin market dynamics. Using Valence Aware Dictionary for Sentiment Reasoning (VADER), we extract useful information from tweets and construct two factors – sentiment dispersion (SD) and investor attention (IA) – to test their predictive power. We show that investors face greater return volatility for rising sentiment dispersion associated with more significant market uncertainty. Further, IA is found to predict bitcoin trading volume but not returns and volatility. Finally, we design an IA-induced trading strategy that yields superior performance to the passive buy-and-hold strategy in 2018. However, it does not deliver superior performance in other years during the sample period suggesting that investor attention alone as a trading parameter does not produce superior performance over the long term.