Overview and details of the sessions of this conference. Please select a date or room to show only sessions at that day or location.
Please select "Show Presentations" and "List View" or a single session for detailed view (with abstracts, discussants, downloads). PDF downloads are accessible to registered participants only (after Login).
The expected returns on machine-learning strategies
Vitor Azevedo1, Christopher Hoegner2, Mihail Velikov3
1RPTU Kaiserslautern-Landau, Germany; 2Technische Universität München; 3Smeal College of Business, Penn State University
Discussant: Stefan Scharnowski (University of Mannheim)
We estimate the expected returns of machine learning-based anomaly trading strategies and quantify the impact of three factors often overlooked in the previous literature: transaction costs, post-publication decay, and the post-decimalization era of high liquidity. Despite a cumulative performance reduction averaging about 57% when accounting for these three factors, sophisticated machine learning strategies remain profitable, particularly those employing Long Short-Term Memory (LSTM) models. We estimate that our most effective strategy, the one based on an LSTM model with one hidden layer, has an expected gross (net) Sharpe Ratio of 0.94 (0.84). We rationalize these findings in a simple theoretical framework in which technological diffusion gradually erodes trading profits while superior signal processing capabilities allow the extraction of alpha from increasingly complex information.
2:30pm - 3:00pm
Dealer quid pro quo in the municipal bond market
Casey Dougal2, Daniel A Rettl1, Vasiliy Yakimenko1
Dealers intermediate trades in OTC markets through trading networks. In the municipal bond market, we document greater complexity than the typical core-periphery structure. Analyzing dealer reciprocity-the tendency to repay favors-we find reciprocity generally reduces markups. Dealers trade lower markups today for future liquidity. However, in small trading communities, reciprocity can foster collusion via quid-pro-quo agreements, inflating transaction chain markups. Among high-centrality dealers in large communities, high reciprocity lowers average markups by 80 basis points, while among low-centrality dealers in small communities, it raises markups by 72 basis points. Although only around 2% of transaction chains suggest collusive behavior, these significantly affect regression results, highlighting the importance of controlling for such outliers to accurately estimate centrality premiums or reciprocity discounts.
3:00pm - 3:30pm
Hidden Liquidity - Evidence from the Introduction of Iceberg Orders
Stefan Scharnowski
University of Mannheim, Germany
Discussant: Daniel A Rettl (University of Georgia)
This paper analyzes the effects of hidden liquidity by studying the introduction of iceberg orders at a large cryptocurrency exchange. Compared to other assets, cryptocurrencies often trade against both fiat currencies and pegged stablecoins. Considering the introduction of iceberg orders for trading pairs against the US dollar but not against a dollar-pegged stablecoin, this study finds that hidden liquidity is associated with increased quoting and trading activity. Larger average trade sizes suggest greater institutional participation. Quoted liquidity improves while the price impact of trades declines. Realized spreads increase, indicating improved revenues for market makers. Price discovery also shifts significantly toward the markets accepting iceberg orders. Overall, our results suggest that hidden liquidity has positive effects on market quality.
