Research

Abstract: The introduction of ChatGPT has changed how humans process textual data. We devise a prompting strategy for ChatGPT to identify and analyze unusual aspects of financial communication, focusing on earnings calls of S&P 500 firms. Utilizing the latest GPT-4-Turbo model, we identify and categorize unusual financial communication across 25 dimensions, which fall into four categories: unusual communication by executives, by financial analysts, unusual content, and technical issues. A significant portion of earnings calls displays unusual financial communication, which correlates with certain firm characteristics and fluctuates with the business cycles. The stock market reacts negatively to unusual communication, with an elevated trading activity. We highlight the potential of large language models like ChatGPT in financial analyses, offering new insights into the interpretation of complex textual data and its economic consequences on market impacts. 

Abstract: In this paper we study the dynamics of stocks and options with lottery characteristics. In particular, we show that the availability of options for retail investors displaces lottery stocks, and only lottery stocks without options sell at premium since retail investors can easily invest in options. We further show that at-the-money lottery options and lottery stocks are substitutes, while deep-out-of-the-money options complement lottery stocks, as they are the lottery choice when lottery stocks are not available. These results allow us to interpret order imbalances in lottery stocks and options. We also analyze the factors that explain the “lotteryness” of stocks and options. 

Abstract: We establish the out-of-sample predictability of monthly exchange rates via machine learning techniques based on 70 predictors capturing country characteristics, global variables, and their interactions. To better guard against overfitting in our high-dimensional and noisy data environment, we make additional adjustments to “off-the-shelf” implementations of machine learning techniques, including imposing economic constraints. The resulting forecasts consistently outperform the no-change benchmark, which has proven difficult to beat. Country characteristics are important for forecasting, once they interact with global variables. Machine learning forecasts also markedly improve the performance of a carry trade portfolio, especially since the global financial crisis. 

Outstanding Paper Award in 2022 Chinese Finance Annual Meeting

Semi-finalist for the Best Paper Award of 2022 FMA

Abstract: This paper identifies a unique dimension of the currency carry trade that is related to the intensity of technology spillover across countries. Particularly, we provide empirical evidence to show that technology diffusion is a fundamental determinant of currency risk premium. Technology diffusion is measured by the R&D ingredient embedded in intermediate goods trade. To rationalize this fact, we build a two-country growth model with Epstein-Zin preference and one-side technology spillover. Intuitively, endogenous technology adoption provides insurance to high-interest rate countries against the risk of innovations. However, carry traders require a risk premium for holding the high-technology-diffusion currency as compensation for financing the technology adoption, while the low-technology-diffusion currency provides a hedge against the downside movements of carry trade profitability. 

Abstract: We investigate the effect of ETF ownership on stock market anomalies and market efficiency. We find that low ETF ownership stocks exhibit higher returns, greater Sharpe ratios, and highly significant alphas in comparison to high ETF ownership stocks. We show that high ETF ownership stocks demonstrate more pronounced information flows than low ETF ownership stocks which reduces their mispricing as they are more informationally efficient. We find similar results when we match the two groups based on size, volume, book-to-market and momentum. Our results are robust to different matching methods and to a wide array of controls in Fama-MacBeth regressions. 

Abstract: In this paper, we propose a cross-sectional option momentum strategy that is based on the risk component of delta-hedged option returns. We find strong evidence of risk continuation in option returns. Specifically, options with high risk components significantly outperform those options with lower risk components over time. The risk-based option momentum strategy is highly profitable for different formation and holding periods, and it is more profitable than the option momentum strategy of Heston, Jones, Khorram, Li, and Mo (2023). We show that risk-based option momentum is unrelated to standard option momentum and mostly subsumes its performance. The strategy is not subject to crash risk, and it is not followed by long-term reversals. Our results are robust to alterations of the empirical setup.

Presentations: EFA* 2024 (Bratislava, scheduled), Inquire UK-Inquire Europe* 2024 (Southampton), FMA Consortium on Asset Management* 2024 (Cambridge), FMA* 2023 (Chicago), Wolfe Research Annual Global Quantitative and Macro Investment Conference 2023 (New York), DGF* 2023 (Hohenheim), Human Normal University,  Jiangxi University of Finance and Economics* 2023, Nanjing University* 2023, Peking University* 2023, Renmin University of China* 2023, Tongji University* 2023 , Tsinghua University* 2023, Washington University in St. Louis* 2023, Xian Jiaotong University* 2023, Hong Kong Conference for Fintech* 2023, AI and Big Data in Business* 2023.

Abstract: We show that investment decisions of ETF market participants when trading country ETFs are mostly driven by shocks to U.S. fundamentals, rather than local risks. Investors react only to negative news about local economies. When U.S. economic uncertainty increases, investors switch to Cash ETFs. We demonstrate that ETF arbitrage mechanism is one of the key channels through which U.S. shocks propagate to local economies leading to increased return correlation with the U.S. market, limiting the benefits from international diversification. We find that countries with stronger ETF price discovery and lower limits to arbitrage have a higher comovement with the U.S. market. 

Presentations: AFA* 2022 (Boston), Queen Mary University of London* 2022, FMA 2019 (New Orleans), CICF 2019 (Guangzhou, China), Finance Forum* 2019 (Madrid), Fulcrum Asset Management* (London) 2019, The Hebrew University of Jerusalem* 2019, Jupiter-Bristol PhD Seminar* 2019 (London), Lancaster-Warwick Workshop* 2018, Brown Bag Series - Warwick Business School* 2018. 

Abstract: We develop a novel measure of media attention to U.S. populism by extending an existing populist dictionary to capture the new form of populism. Our Aggregate Populist Rhetoric (APR) Index spikes around well-known events that spur populist sentiment, and exposure to APR is linked to financial globalization. We show that the APR Index is priced in the cross-section of currency excess returns. Currencies that perform well (badly) when attention to U.S. populism is high yield low (high) expected excess returns. Investors require a risk premium for holding currencies that underperform in times of rising attention to U.S. populism. Financial segmentation explains why friction to globalization in the form of populism affects the cross-section of currency returns.

Abstract: This paper investigates the cross-sectional predictive ability of text-based factors in the cryptocurrency market –an important asset class for retail and institutional investors. We employ Bidirectional Encoder Representations from Transformers (BERT) topic modeling to analyze news articles discussing the top 43 cryptocurrencies by market capitalization. We build text-based factors related to fundamentals and technical trading. We find that pessimism about technical news is positively priced in the cross-section of cryptocurrency returns, while pessimism about fundamental news is negatively priced. These factors provide information over and above existing factor models. Our results demonstrate the importance of considering text-based factors when analyzing cryptocurrency returns.

Abstract: Using textual analysis, we identify the set of Trump tweets that contain information on macroeconomic policy, trade, or exchange rate content. We then analyze the effects of Trump tweets on the intraday trading activity of foreign exchange markets, such as trading volume, volatility, and FX spot returns. We find that Trump tweets reduce speculative trading, with a corresponding decline in trading volume and volatility, and induce a bias reflecting Trump’s (optimistic) views on the U.S. economy. We rationalize these results within a model of Trump tweets revealing economic content as a public signal that reduces disagreement among speculators. 

Abstract: Given all the options with the same expiration written on a particular stock, Max Pain price is the strike price at which the total payoff of all the expiring options is the lowest. We construct a measure of Max Pain, sort stock prices according to this measure, and find that a spread portfolio that buys high Max Pain stocks and sells low Max Pain stocks generates large, positive, and statistically significant alphas. Our results provide strong evidence of stock price predictability at the expiration of the options. We also find that there is significantly higher abnormal stock volume and order imbalances for high Max Pain portfolios. The strategy is not related to reversals of price trends that might have explained initial options volume. Our results are robust to a large number of tests. 

Abstract: Short squeezes lead to sudden, large increases in stock prices. Using a novel measure of the likelihood of short squeezes we show that its volatility --which indicates possible high values of the probability of a short squeeze in the near future-- is a proxy for right-skewness and used by skewness-seeking investors. These investors buy call options instead of the underlying stocks, to maximize the right-skewness of their investment. In particular, they are willing to pay a premium for the upside potential of these call options, as it is the case for other lottery-like securities identified in the literature. 

Abstract: We investigate the out-of-sample predictability of daily cryptocurrency returns using modern machine-learning methods. We consider a large number of cryptocurrencies (41) and a rich set of predictors relating to a cryptocurrency's network value and activity, time-series momentum, technical signals, and investor attention and sentiment. Our results indicate that return predictability is an important feature of the cryptocurrency market: machine-learning methods significantly improve the statistical accuracy of cryptocurrency return forecasts and provide substantial economic value to an investor. We find that a diverse set of predictors contribute to cryptocurrency return predictability and that nonlinearities play a prominent role.

Abstract: Using close to 40 years of textual data from FOMC transcripts and the Federal Reserve staff’s Greenbook/Tealbook, we extend Romer and Romer (2008) to test if the FOMC adds information relative to its staff forecasts not via its own quantitative forecasts but via its words. We use methods from natural language processing to extract from both types of document text-based forecasts that capture attentiveness to and sentiment about the macroeconomy. We test whether these text-based forecasts provide value-added in explaining the distribution of outcomes for GDP growth, the unemployment rate, and inflation. We find that FOMC tales about macroeconomic risks do add value in the tails, especially for GDP growth and the unemployment rate. For inflation, we find value-added in both FOMC point forecasts and narrative, once we extract from the text a broader set of measures of macroeconomic sentiment and risk attentiveness.

Abstract: We develop a multivariate functional false discovery rate method that uses multiple informative covariates to examine the performance of predictive models while controlling for data snooping. This method has superior power and is robust to data dependence, estimation errors in covariates, and correlated covariates. We apply the new method to a large universe of currency technical trading rules, and construct a dynamic 30-currency portfolio that generates a Sharpe ratio around one for roughly 50 years. Technical trading profitability decreases with the computational power of traders, capital account openness, financial market development, and market liquidity, and increases with exchange rate volatility.

Presentations: World Finance Banking Symposium* 2023 (Vilnius), Pari Financial Management* 2023, FMA* 2022 (Atlanta). 

Abstract: We study firm-level characteristics that a manager would employ as signalling tools in order to time the market (i.e. repurchases and issues). Following the market timing framework, we develop a two-factor asset pricing model comprising a “market” and a “mispricing” factor, which is able to capture the cross-sectional variation of value and momentum. Specifically, loser (undervalued) portfolios provide a premium when market timing succeeds while winner (overvalued) portfolios provide a hedge under bad states of the world when market timing fails. The two factors contain important information regarding the time-variation of the strategies providing a unique explanation for momentum crashes.