Commodity Speculation Debate

...

For this reason, the CFTC issued its first ‘special call’ in June 2008 to assess the total investment activities (on- and off-exchange) of commodity index funds. According to Irwin & Sanders (2012), the index investment data (IID) collected through the special call provides the most accurate measure of total commodity index investment available to date.

IID collected the total notional value of the firm’s commodity index business and the equivalent number of futures contracts. Gathered data from 43 entities engaging in index activities in commodity markets. These entities included index funds, swap dealers, pension funds, hedge funds, mutual funds, exchange traded funds (ETFs), and exchange traded notes (ETNs). Each firm’s entire “book of business” in futures and OTC markets that is related to CII is reported not just the netted amount that may be managed ultimately in the futures markets. Data is cross-checked by comparing it to positions in the CFTC’s LTRS and by engaging the firms in extensive discussions.

IID is collected at the end of each quarter from December 31, 2007 through to March 31, 2011 so essentially 14 quarter-ends. Which is too short for time series analysis and renders powerful statistical analysis models useless.

Data comparison of IID, Supplement COT and Disaggregate COT highlighted the gravity of over-estimation. Comparison between IID and Supplement COT agricultural markets showed that the index trader positon was overestimated by 3% and average absolute error of 10%. This is because Supplement COT classifies a trader in a single category even if an index trader held positions other than index trading.

Data comparison of 12 agricultural markets between IID and Disaggregate COT Swap dealer positons, which is used as a proxy for index investment funds, had an average absolute error of 29%. This is because of Swap dealers excludes the Managed Money and Other Reportables category which include index trading. Netting effect adds further measurement error to the swap dealer positions.

Finally Data comparison between IID and Disaggregate COT energy and metals markets has an absolute error of 71%. This as mentioned previously was due to the netting effect.

Masters (2008) and other studies such as Gilbert (2010) and Singleton (2011), impute positions particularly for energy and metal markets from agricultural markets. This is mainly because the separate category created in Supplement COT report is for 12 agricultural markets only. The way they impute index positions is by scaling minor market positions up to the major market position. This creates a substantial scaling measurement error. This means any error already existing from Supplement COT report has increased the error substantially by this scaling procedures to impute the energy market futures position.

Proponents of Masters Hypothesis

Gilbert (2010) ran first order Augmented Dickey-Fuller (ADF) regression and found no statistically significant relationships between Index Traded Funds and energy metal and food price changes. The data used was daily closing prices used from Jan 2006 to Dec 2008. Measures of index positions in energy markets imputed from positions held in agricultural commodities as used as a proxy.

Singleton (2011) ran a regression and correlation tests and found index investment flows are an important determinant of price changes along with several other conditioning variables. Estimates that 1 million contract increase in Index Fund positons in WTI Crude oil over 13 week period results in 0.27% increase in nearby futures prices in the next week. Singleton’s evidence stated that financial flows precede changes in the oil futures price, but correlation does not necessarily imply causality. The data used was CFTC’s Supplement Commitment of Traders (COT) reports (also known as Commodity Index Traders: CIT Reports). The date range was from September 12, 2006 through to January 12, 2010. The short horizon was due to data constraint. What amplifies the data constraint in Singleton’s case is that he uses the short horizon data as measure of index fund positions in oil futures is imputed from agricultural futures.

Both papers that find econometric evidence that support Master’s Hypothesis’ use of agricultural positions which rely on measures of index positions in energy markets imputed from positions held in agricultural commodities, this is mainly because there is more data available for agricultural markets that there is for energy markets e.g. the Supplement COT Report.

Against Masters Hypothesis

Stoll & Whaley (2010) ran a battery of Granger Causality tests and found no statistical relationship for agricultural and energy markets. They used the Disaggregated COT report (US only data) for which they have a category called Swap Dealer and essentially that is used as a proxy for Commodity Index fund positions. A more accurate data would have been IID “survey call” data for which they highlighted the importance of the IID report but due to the lack of data points available it made it unfeasible to statistically test it as most time series analysis and methods require a large sample size and essentially a major limitation finance and accounting analysis. Granger Causality tests, though inappropriately named, is a test of precedence rather than causality. It also lacks in statistical power to reject null hypothesis of non-causality because changes in commodities futures prices is extremely volatile. Furthermore this research was supported by a grant from Gresham Investment

Management LLC, which is an index investment firm, which essentially gives off an air of researcher bias.

Irwin and Sanders have consistently published empirical studies to disprove that Commodity Index Investing were the cause of the increase in commodities futures prices in 2007-2008. They published papers previously in 2010, and two in 2011 in which they found no statistical relationships for both the agricultural and energy market.

Kaufman & Ulman (2009) ran Two-Step DOLS Regression and found Causal relationships between the prices of 10 crude oils analysed, suggest that the rise in oil prices leading up to March 2008 were generated by both changes in market fundamentals and speculation

Increasing demand, along with stagnant levels of non-OPEC production changed supply-demand balance, causing prices to increase. This fundamental change was recognised by speculators, who took positions accordingly. Increased anticipation in futures market transmitted to the spot market, driving prices beyond justified levels by supply/demand

...