This research explores whether the risk of having extreme losses in the European stock market can be predicted by incorporating macro-economic variables. To answer this question, we investigated the tail distribution of weekly stock losses on six different European indices (AEX, DAX, CAC40, PSI-20, IBEX35 and FTSE MIB). We set up an Extreme Value Theory (EVT) machine learning framework using shrinkage regression techniques, such as LASSO. Our results show that when adding a limited amount of macro-economic covariates to the tail distribution of weekly losses, the prediction for the VaR is improved for five of the six indices. Inflation, short-term interest rate, industrial production and the USD/EUR exchange rate appear to have predicting power for the tail risk. No predicting power is found when using unemployment and the long-term interest rate. This study implies the importance of macro-economic information when estimating financial risks of investments in the European stock market, which can be of added value for investors and regulators.

Additional Metadata
Keywords Extreme Value Theory, Peaks-Over-Threshold, Generalized Pareto distribution, Poisson point process, Macro-economic covariates, Value-at-Risk, European stock market, Shrink-age regression
Thesis Advisor Zhou, C.
Persistent URL
Series Econometrie
Galle, M.J. (2020, July 14). Modeling extreme European Market Risk. Econometrie. Retrieved from