Commodity Index Traders -- The New Whipping Boys?*

An August CFTC conference in Washington DC was organised to highlight and discuss recent academic research on key issues affecting commodity markets. The conference came at a time of intense debate surrounding recent CFTC rulemaking.  Several conference panelists argued that speculators in general, and commodity index traders (CITs) in particular, have affected the functioning of commodity markets and caused oil price swings that cannot be explained by energy market fundamentals – especially during the 2008 financial crisis.  However, in the presentations of that set of papers, we failed to see any detailed accounting for those very fundamentals.  In contrast, the few papers at the conference that focused on fundamentals found no clear-cut evidence of speculators driving prices away from their fundamental values.

Perhaps one of the most discussed papers in the follow-up press coverage was delivered by Stanford University’s Kenneth Singleton.  It is worth noting here that Professor Singleton’s paper, which has been seized upon by supporters of limits on speculative commodity positions, started as a survey that he conducted for the Air Transport Association of America (ATA).  For this reason and because the paper has received wide publicity, it is worth discussing some of the paper’s limitations and methodological shortcomings – which its author himself has acknowledged.

Professor Singleton’s paper comprises two parts – a formal model of how speculation could, in theory, temporarily drive commodity prices away from their fundamental value, and an empirical analysis.  The theory part assumes that investors are a heterogeneous bunch. Those market participants, including speculators, try to anticipate their competitors’ move given imperfect information on economic fundamentals and on speculative activity. Their decisions are therefore sometimes affected by other participants’ behaviour (herding); that is to say, investors sometimes mimic their competitors’ moves. Investors might also have different opinions about the future course of economic fundamentals. Although these investors might be using common knowledge, their interpretation of common information might be different, which leads to higher trading volumes and co-movements among different asset classes. This has the effect of moving prices away from their fundamental values, inducing higher volatility and generating booms and busts in prices.

Methodology Questions Abound

Professor Singleton next seeks to provide empirical evidence for his analysis. For his empirical work, he uses CIT positions imputed on the basis of investment advisor Michael Master’s methodology.  This extrapolation method has significant problems.  First, the imputation is based on CFTC Supplemental Commitments of Traders (SCOT) report, which does not contain any data for crude oil. To estimate crude oil CIT positions, one must therefore use data from 12 agricultural markets for which the CFTC gathers CIT position information, and then make a number of unlikely assumptions about the relationship between CIT positions in agricultural and energy markets. Second, the SCOT dataset itself is problematic when measuring CITs activity – even for the 12 agricultural commodities. This is because the SCOT, which initially identified 32 (now 43) CITs, classifies all the positions of a trader engaged in commodity index trading as commodity index investment – regardless of the trader’s actual trading strategy.  Therefore, as has been pointed out by the CFTC itself, “the published aggregate futures position in the COT-Supplemental may overstate or understate the actual amount of index trading (overstate it to the extent the positions reflect other trading strategies, and understate it to the extent that index positions are internally netted against non-index positions before the net position is brought to the futures markets).”

As was pointed out in another paper presented at the CFTC conference by the University of Illinois’ Scott Irwin, the imputation methodology employed by Professor Singleton can lead not merely to some measurement errors but to huge measurement errors: 

§ The netting effect might not be important for agricultural commodities where the swap dealers’ futures positions are generally limited to long futures hedges offsetting their short OTC exposure to those pension funds or other index-based traders. However, the netting effect might be very important where many swap dealers (as is the case in energy products), in addition to their commodity index-related OTC activity, enter into other OTC derivative transactions in individual commodities, both with commercial firms hedging price risk and with speculators taking on price risk. 

§ Professor Irwin and his co-author Dwight Sanders show that the level of errors is quite large. Comparing the imputed position with Index Investment Data (IID) provides a glimpse of the extent of measurement errors (52% mean absolute errors).  The measurement error at some point leads to a categorisation of some 70-75% of long positions as “index investments”, implying that not only all swap dealers’ long positions but also managed money traders’ positions are categorised as commodity index trading. This is not possible: the CFTC’s own “ Special Call” data indicates that at most 15-20% of index investment is carried out by money managers.

While some of the other papers presented at the CFTC conference proposed less imprecise alternatives to estimate CIT activity in commodity futures markets, they all relied on time series that the CFTC does not make public. If anything, the data difficulties faced by Professor Singleton and others in proxying for CIT activity points to the need for the CFTC to release more disaggregated position information.  For example, there is evidence that the long positions of commodity swap dealers in near-maturity contracts provide a reasonable proxy for commodity index investment. Clearly, publicly available position data disaggregated by maturity would help underpin more detailed analysis in this sphere.

Finally, the interpretation of Professor Singleton’s estimation results also presents some difficulties.

First, to estimate realised excess returns in crude oil prices using weekly data from 12 September 2006 to 12 January 2010, the paper uses the change in the long positions of commodity index funds and money managers as well as some financial variables (such as the change in repo positions on Treasury bonds and lagged returns on emerging market equity positions as explanatory variables). As another paper presented at the conference by the University of California’s Professor James Hamilton showed, energy market fundamentals are crucial to understanding crude oil prices. Professor Singleton’s estimation procedure, however, does not directly account for oil market fundamentals such as demand growth in emerging market economies (especially in China), inventories outside of the OECD, or rising costs of production. Furthermore, as we have argued in previous issues of the OMR, both excess returns and commodity index flows might be responding to some common shocks (such as expectation of higher growth in China and other emerging countries). The point estimates in the regressions might thus be biased due to this endogeneity problem.

Significance: Economic versus Statistical

Second, the paper should provide descriptive statistics on the variables so that one can assess the impact of a change in index investment on realised returns. Professor Singleton’s paper argues that it presents evidence of “an economically and statistically significant effect of investor flows on futures prices, after controlling for returns in US and emerging-economy stock markets, the futures/spot basis, and lagged returns on oil futures.” The index investment coefficient is statistically significant, although it is difficult to assess the economic significance of commodity index investment on realised return, given the information provided on the paper.

Finally, as suggested by the University of Houston’s Craig Pirrong in his blog, suppose that Professor Singleton’s findings do indicate that excess returns on crude oil futures are predictable, conditional on measures of speculative activity. Nonetheless, such a predictability of returns would not imply that speculation has distorted prices. Rather, predictability is the result of market frictions that might create hedging demand, leading to an increase in the risk premium. Professor Pirrong suggests that, in such circumstances, speculative positions can predict changes in futures prices. To prevent the predictability of returns, it might be advisable to reduce constraints on the flow of speculative investment to commodity markets, rather than limiting them.


*IEA Oil Market Report-September 2011