New Backtests Improve Risk Forecasting with Expectiles
Financial institutions and regulators are gaining a more robust tool for evaluating market risk forecasts with the development of new backtesting approaches for expectiles. A recent study published in Risk Sciences details methods designed to address limitations in existing backtesting techniques, offering improved accuracy and reliability.
The Limitations of Traditional Risk Measures
Currently, Value-at-Risk (VaR) and Expected Shortfall (ES) are widely used to measure market risk. However, these measures have known drawbacks. VaR lacks coherence, while ES is not independently elicitable. Expectiles, a relatively newer risk measure, offer a compelling alternative as they are both coherent and elicitable – meaning they are logically consistent and can be accurately estimated from data.
What are Expectiles?
Expectiles represent the minimized asymmetric least squares criterion, functioning as a weighted average. They are particularly valuable as they align with the true indicate of a distribution under specific conditions. As explained by C.S. Philipps, expectiles also relate to quantiles, serving as a unifying framework for various statistical estimations.
New Backtesting Approaches
Researchers from Canada and the UK have developed new backtests specifically for expectile forecasts. These tests aim to overcome issues like size distortion and low test power that plague existing methods. The core innovation lies in separating two key properties: unconditional coverage (ensuring forecasts are accurate on average) and independence (verifying that forecast errors don’t exhibit problematic patterns over time).
“We wanted to introduce novel expectile backtests with better size and power properties,” said Yang Lu, the corresponding author of the study. The team employed a simplified Wald-style test for unconditional coverage, focusing on a single expectation condition to minimize distortion. For independence, they connected a Wald-testing framework to the Box–Pierce lack-of-autocorrelation test, adapting it for use with expectile forecasts.
Testing and Results
Simulation studies demonstrated promising results for the proposed tests, indicating improved performance in finite samples. The approach was then applied to S&P 500 return data, illustrating its practical application. Researchers noted that the independence testing construction is most effective under a location-scale framework and may be less suitable in scenarios with stochastic volatility.
Implications for the Financial Industry
These advancements in expectile backtesting provide financial institutions with more reliable tools for risk management and regulatory compliance. By offering a more accurate assessment of forecast performance, these new methods can contribute to a more stable and resilient financial system.
Further Research
Ongoing research continues to refine these techniques and explore their application in diverse financial contexts. As reported by Newswise, further development of backtesting tools for expectiles is crucial as their adoption grows within the financial industry.