In an apparent attempt to make Michael Masters look good, Rice University’s Baker Institute has released a report on oil speculation, that concludes that increasing volumes of speculation have caused oil prices to be higher than they would otherwise be, and in a novel twist, have caused a higher correlation between the dollar and the price of oil. It also lays blame on everybody’s favorite whipping boy, the Commodity Futures Modernization Act of 2000.
In a nutshell, this report is bilge. It relies on no rigorous economic analysis to support its contentions. Moreover, it is unscientific, eschewing any serious statistical analysis.
The report first sets out its evidence that speculation has increased, and that this increase is due to the CFMA. It attempts to argue that speculation increased 15-fold, while commercial trading only doubled, by adding non-commercial outright futures, options, and spread positions, as measured by the CFTC Commitment of Traders Reports. The vast bulk of the increase in speculation is attributable to spread positions.
But it is odd indeed to argue that a massive increase in speculation caused higher oil prices when the bulk of this increase in speculation occurred in spread positions. Such positions, which involve the purchase and sale of oil for different delivery dates, are speculations on the shape of the oil forward curve, not on the absolute price level. Indeed, it is very difficult to imagine how such spread positions would cause prices to rise since they involve purchases and sales in equal magnitude. Certainly the authors of the Baker study provide no rigorous justification for any such connection. For that matter, they don’t provide any non-rigorous justification either.
Tying the massive increase in speculation as they measure it to CFMA is particularly bizarre. As they note, the main provisions of the CFMA eased restrictions on trading on the OTC market, and in non-US markets. As a result, if anything, the CFMA favored these other markets over NYMEX. But all of the data in the “study” pertain to NYMEX futures markets. How can you attribute a rise in trading on NYMEX, and a change in the composition of that trading, to regulatory changes that would tend to shift trading to venues not covered by the data? (You could make an argument that the increase in off-exchange activity lifted NYMEX volumes due to arbitrage, or to the use of futures to hedge OTC positions. But the Baker boy and girl don’t do that. And if some non-commercial NYMEX positions are actually hedges of OTC positions, you can’t conclude the futures positions are speculative without knowing about the related OTC positions).
In other words, report authors Ken Medlock and Amy Jaffee play the drunk looking for a wallet under the lamppost. The COT data is the only data relating to speculation that they can get their hands on, and they use it even though it is peripherally related to what they want to demonstrate, namely, that a set of regulatory changes that mainly benefitted non-NYMEX markets led to a dramatic increase in speculation.
This is the crudest post hoc, propter hoc reasoning. CFMA happened in 2000. Speculative futures positions increased, especially post-2003. Therefore, the former caused the latter. (Why the big lag? Don’t bother looking for answers in the report. You won’t get one.) This is about as weak as it gets.
When they get to analyzing the effects of speculation on prices, the spreaders are out the window, and Medlock and Jaffee focus on net non-commercial positions. (This could be viewed as a tacit admission that there is no basis to believe that spreaders could have an effect on price levels.)
Do the study authors utilize statistical techniques, like Granger Causality or the like, to document such a connection? Surely you jest.
They use a much more sophisticated diagnostic instrument. Their eyeballs.
They present a graph of net non-commercial positions overlaid on prices, and say:
Generally, movements in price over the last few years have coincided with trends in open interest by noncommercial traders. We can see that during periods where [sic] speculators have been net short, prices typically declined, even if only slightly. When speculators are net long, the general shift in the market has been upwards, in some cases to a dramatic extend. Some exceptions have occurred when speculators were generally in a net long position, but were moving to liquidate positions. In this case, such as the late spring/early summer 2008, prices responded by moving sharply lower.
Well, when I look at their Figure 5 with my eyeballs, I note numerous instances when the changes in net non-commercial positions were of a similar magnitude to what occurred in spring-summer 2008, and the price movements were nowhere as large as in the late-summer/fall of 2008. I see other periods (fall, 2008) when net long positions spiked, but prices fell.
I just wonder: think the financial crisis might have had something to do with the collapse in prices? What financial crisis you might ask, after reading the report: if this was the only information you had, you’d never know there was a financial crisis.
Also, why the asymmetry? Why do net shorts often result in “slight declines” while net longs can lead to big spikes? It is not obvious a prior that this should be the case, and Medlock and Jaffee certainly provide no explanation.
There is no way in hell that such a casual, visual analysis of that Figure can provide a basis for their grandiose conclusions. To say that this is unscientific is far too kind. It would be too kind to call it junk science. It’s just junk.
Medlock and Jaffe also interpret the data, uhm, flexibly, in order to make their case. On the one hand, they claim, prior to 2006 “the open interest of noncommercial players” was a “lagging indicator” of prices. (If it lags, how can it cause?) But then, in a “striking” development, it becomes a “leading indicator of price around January 2006.”
Do they present a rigorous test of this assertion? No.
Do they provide a coherent explanation as to why it would go from a lagger to a leader? They provide an explanation, but it is not coherent. It is in effect the story of a perpetual motion machine. Speculators buy so prices go up so people think prices will go up and so they buy and force up prices more yadda yadda. This is, at best, ex post rationalization, fitting their story to the data rather than coming up with a rigorous hypothesis without peeking at the data and then testing it.
How do they address the problem that if COT data is really a leader, and leads by months (according to their eyeballs, anyways), why did all of these greedy speculators leave money on the table? As soon as they recognized that COT data lead oil prices, they would trade on the COT data, and voila!, the relationship would disappear.
Again, the more obvious answer: there is no causal relationship between COT numbers and prices.
A more reasonable explanation is there is no connection between movements in net noncommercial positions and prices. Sometimes positions lead price, then sometimes the reverse is true, but there is no causal connection, just chance in action. This makes Jaffee and Medlock look like fools for randomness.
The asserted connection between oil and the dollar resulting from increased speculation is bizarre. For starters, I contest their statistical results, such as they are. They report a correlation between the dollar and the price of oil of -.82. I can reproduce that number by calculating the correlation between the levels of the dollar index and the price of oil. However, since both time series have unit roots, such high correlations are extremely suspect: highly persistent time series can be highly correlated even if, in fact, there is no economic relation between them. This is the spurious regression problem.
A more reasonable method is to look at correlations between percentage price changes. This number is much smaller (in absolute value) for 2001-2009: only -.22. It is bigger than observed in 1986-2000, but so what? What’s more, that -.22 isn’t nearly as eye-popping as -.8, is it?
And what theory do these people advance to explain why an increase in speculation would increase the correlation? They tell a story–and it’s no more than that–about speculation causing high oil prices causing higher trade deficits causing a lower dollar. There is no evidence supporting any one of these alleged causal connections. And considering the trade deficit–like, have these two heard of China? Or any other factor that might have influenced the trade deficit? This is more post hoc, propter hoc mumbo jumbo, with myriad omitted variables. A serious study would (a) estimate the relation between the trade deficit and the dollar index, controlling for other factors that could affect the dollar; (b) estimate the effect of the price of oil on the deficit, and combine this with (a) to get an effect of the oil price movements on movements in the dollar, and (c) estimate the effect of speculation on the price of oil, and combine with (a) and (b) to get an estimate of the contribution of speculation to changes in the value of the dollar. Or include oil speculation as an explanatory variable in a model of the dollar that also includes other relevant explanatory variables.
But that would be, as I say, a serious study. This one is a joke.
There’s also a serious issue regarding the direction of causality. Why should it go from oil prices to the dollar? A decline in the dollar represents a fall in its value relative to other currencies. The factors that lead to such a fall (e.g., a US monetary policy that is looser than the monetary policies of other nations) also plausibly lead to a decline its value relative to goods, including oil. A decline in the purchasing power of a dollar means that the prices of goods, including oil, rise in nominal terms.
Here’s a final, telling example of the level of rigor in the report: “While correlation does not imply causation [really?], the trends evident in the open interest data are impossible to ignore.”
Is that so? Just watch me.