Statistical Methods in Finance 2019

---

Abstract

Forecasting Market Risk of Indian Equity and Bond Market Indices

by Santanu Dutta, Tezpur University, India

Forecasting market risk involves three components, viz. a risk measurement procedure based onhistorical data, a measure of its predictive performance, and finally backtesting the validity of the riskforecasts. In the context of point forecasting, a risk measure has to be elicitable without which it is notclear how to choose a scoring function for measuring the predictive accuracy of the forecasts. Unfor-tunately, several coherent risk measures (including the expected shortfall) are not elicitable. Using twoelicitable risk measures, viz. the Value at Risk (VaR) and the median shortfall (MS), we forecast themarket risk of the equity and bond market indices in the National Stock Exchange (NSE) in India. Apiecewise linear scoring function, consistent for the VaR, is used to measure the predictive performance.Out of sample backtesting is used to check the validity of the risk forecasts. The Basle recommendationof forecasting risk based on most recent twelve month daily return data, and updating the risk forecastsquarterly, does not seem to work well. We overcome this problem. This three step method of making,evaluating and back-testing point forecasts of the daily market risk in the Indian equity and bond marketscan be applied widely.

---