Detecting speculation in volatility of commodities futures markets

This paper evaluates if and how speculation affects the volatility of commodity futures: it distinguishes between short term and long term measures of speculation and investigates if the impact on volatility is different. Speculation is measured by means of four indexes: scalping, Working’s T, the market share of non commercial traders and the percentage of net long speculators over total open interest in future markets. Data concern four energy commodities (light sweet crude oil, heating oil, gasoline and natural gas) and seven non-energy commodities (cocoa, coffee, corn, oats, soybean oil, soybeans and wheat) over the period 1986-2010 analyzed at weekly frequency in the US market. We use GARCH models and related modifications (GARCH-in-mean, EGARCH, TARCH) The paper finds that: (i) speculation significantly affects volatility of returns; (ii) scalping (which proxies short run speculation) has a positive and significant impact on volatility while the other three indexes (which proxy long run speculation) have generally a negative effect (when significant). Then we run a robustness exercise to verify how results change with different data frequency, econometric techniques and macroeconomic controls. We find that: (i) scalping is always positive and significant also at higher and lower frequency of data; (ii) results remain unchanged through different model specifications (GARCH-in-mean, EGARCH, TARCH); (iii) our choice of macroeconomic variables is the best that fits our data.