Measuring risk for large hedgers and large speculators in major US futures markets

Ikhlaas Gurrib

Abstract

ABSTRACT

Using the Commodity Futures Trading Commission's Commitments of Traders data, considering both the generalized autoregressive conditional heteroskedasticity (GARCH) and the power ARCH volatility-based models, it has been found that the lagged volatility and the news about volatility from the previous month are significant in explaining large hedger and speculator volatilities. A greater reliance on the ARCH term by speculators suggests their greater reliance on past information for making their current decisions. Furthermore, the hedger volatility in treasury bonds and coffee and the speculator volatility in gold and Standard & Poor's 500 futures have experienced increasing volatility persistence to shocks during the 1990s. In all remaining markets, hedger and speculator volatilities have shown a tendency  to decay over time in response to shocks, supporting the belief that both players are informed and react well to news volatility. The PARCH model explains the volatility of both players more accurately by exhibiting a greater number of negative components of volatility than theGARCH model. Both models, under the normal and t-distributions, support the fact that most futures returns in the 29 US markets were leptokurtic. The PARCH model, under a normal distribution, was ranked first in explaining actual returns and forecasting the one-month futures return for both players. The PARCH model, under the t-distribution, was ranked last due to its high sensitivity for the standard deviation of returns. Idiosyncratic volatility was found to be a poor proxy of volatility in forecasting the onemonth futures return.