Stablecoin Depegging Risk Prediction
40 Pages Posted: 19 Jan 2024
Abstract
This paper extensively reviews empirical literature on stablecoins, systematically identifying key variables that could lead to depegging risks. Based on this, we construct predictive models using three machine learning algorithms (logistic regression, random forest, and XGBoost) that can accurately and timely predict stablecoin depegging events. Our main subjects of study are the top four stablecoins in daily trading volume: USDT, USDC, BUSD, and DAI. Unlike previous literature that used static depegging threshold values, we adopt a dynamic threshold adjusted for trading volume as the criteria for depegging. In addition to traditional on-chain price and volume data, this study is the first to incorporate sentiment indicators from news sources. The empirical period covers from January 1, 2022, to December 31, 2023. The results show that, as indicated in the literature, significant price and volume fluctuations of mainstream cryptocurrencies (BTC and ETH) indeed cause stablecoin depegging. Furthermore, measures of instability from past literature also provide longer-term early warning effects for stablecoin depegging. However, surprisingly, the sentiment indicators used in this study did not show a significant early warning effect for our research subjects. The models constructed in this study enable crypto asset investors to timely predict the risk of stablecoin depegging and make corresponding investment decisions, thereby reducing investment risks.
Keywords: Stablecoins, Depegging, Machine Learning
Suggested Citation: Suggested Citation