Modelling the Return Distribution of Rare Earth Element Stocks using Quantile Regression

Abstract

In this article, we model the dependency structure between rare earth industry returns and changes in five macroeconomic risk factors. We utilize the quantile regression methodology which further enables us to calculate and stress test Value-At-Risk (VaR) from the estimated conditional quantiles. Our focal point is on the daily logarithmic return of three equally weighted portfolios: Exploration, Mining & Production and Fully Integrated, which we regress on the log return of the market portfolio, changes in the volatility index, changes in the USD exchange rate, changes in the S&P GSCI Industrial metal index and changes in the usage weighted price index for rare earth metals. In the analysis, we find differences in factor effects across quantiles for all variables, indicating varying risk exposure for dissimilar market conditions. This implies that the OLS regression may be insufficient in uncovering the risk-return relation for the portfolios, especially in the upper and lower part of the distributions. Further, we forecast the next-day VaR levels with the beta estimates retrieved from quantile regression, followed by a scenario analysis for each macro factor. This provide valuable insights to financial risk- and portfolio managers on how VaR is affected in the left and right tails by changing the values of the independent variables. We find, for long and short positions alike, that the mining & production portfolio holds the highest sensitivity for upper and lower tail risk, followed by the fully integrated- and exploration portfolio, respectively. As there is little prior research on risk modelling in the industry, we are the first to model the dependence between rare earth element stocks and a set of pre-specified macroeconomic factors for the entire distribution of conditional returns.