Financial Conditions Index Construction Through Weighted L_p-Norm Support Vector Regression

Abstract

This study proposes weighted L_p-norm support vector regression (WL_p-SVR) robust against both noise and outliers. Using L_p-norm enables WL_p-SVR to select financial variables for creating the financial conditions index (FCI) reliably. We use a weighted sum method to construct a Chinese FCI. We then evaluate our FCI’s ability to forecast real output based on the Granger-causality and Engle-Granger cointegration tests. Regression results show that our FCI has strong predictive power in forecasting real output, indicating that our FCI is a potential leading indicator of the future state of the economy.