AIFS0007
Written evidence submitted by Dr Arben Kita
Synopsis
The paper explores how algorithmic high-frequency trading (HFT) in the credit risk market enables the accurate forecasting of firms’ credit and equity risk premia using machine learning (ML). The study finds that credit-market-based risk measures are superior to those derived from stock markets in predicting future risk premia. Despite the rise of automated trading, short-term pricing discrepancies persist between equity and credit markets, suggesting the presence of latent arbitrage opportunities.
Research Questions
a) Do machine-learning-based high-frequency measures of credit and equity risk improve the prediction of firms’ risk premiums?
b) Has the advent of automated ML trading reduced market fragmentation?
Key Policy-relevant Findings
a) Despite the increase of high-frequency trading (HFT), price discrepancies between credit and equity markets persist.
b) Market fragmentation effects remain significant, challenging the assumption that arbitrage activities eliminate inefficiencies.
a) New risk measures can help investors manage risk and improve trading strategies.
b) Regulators can use these findings to assess market efficiency and potential regulatory interventions.
Other Empirical Findings
a) ML-based predictive models outperform traditional statistical methods in forecasting firms’ credit and equity risk premia.
b) Credit-market-based risk measures predict 47% of future credit risk premium changes, whereas equity-based measures predict only 11%.
a) Credit-based risk measures, particularly realised volatility (C-RVC) and jump risk (C-RJV), strongly predict risk premium changes.
b) Deep-out-of-the-money put options also predict changes in risk premia, reinforcing the idea that tail risk factors drive market movements.
Methodology
The paper used the following ML techniques: Lasso Regressions, Double Machine Learning and Random Forest predictive regressions.
The paper used high-frequency CDS and stock prices.
Conclusion
This paper's policy-relevant finding is that the rise of automated trading has not lessened the asynchronous movements between credit and equity market prices. Because a firm’s total value is weighted by its debt and equity, the finance theory predicts that these two quantities must move closely together. The violation of this prediction would allow for an arbitrage opportunity. Yet, empirical studies show a persistent breach of this prediction. The rationale put forward by researches to explain this anomaly is the slow capital moving theory. This theory predicts that because capitals flow slowly into markets, especially the credit market, these pricing discrepancies are to be expected. However, with the advent of ML high-frequency trading in credit markets, the expectation is that this pricing discrepancy should be minimal. Yet, this research shows that the pricing divergence has increased, when the tests are run on matched high-frequency credit and equity prices. Results reported in this paper show that the latency arbitrage is persistent; the front runners hike prices.
Yours,
Arben Kita
An early draft of the paper is available at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3800205
The most updated draft of the paper is available from the author.
February 2025