Just as a drowning person will grab an anchor if thrown overboard, traders latch on to any piece of overweighted flotsam tossed their way. Take the weekly Commitments of Traders Report (COT) produced by the Commodity Futures Trading Commission (CFTC). The urge to believe backward-looking net positions held by various classes of traders can be useful in assessing market imbalances and in forecasting market turning points is extreme. However, unless a speedy visit to the ocean floor was on your to-do list, the temptation should be resisted.
This is not to say we are absent the occasional datum where, to continue our nautical metaphors, everyone was on one side of the boat and the damn thing flipped over. But a series of anecdotes does not a theory make. No, it is simply a case of confirmation bias where you simply remember what worked and forget about the rest. As discussed many years ago (see “Making A Commitment,” June 1998), the chain of causation for COT data is wrong.
- Economic value changes;
- Price moves to reflect these changes and generally overshoots;
- Trend-following traders accumulate positions in the direction of the trend and in grand market fashion become too long at the highs or too short at the lows; and then
- The market reverses as the trend-followers’ collective folly is exposed
In addition, the COT data behave differently across markets and cannot be pooled freely for cross-sectional analysis. To engage in a little anecdotal evidence, ask why managed funds are almost never net short commodities such as silver or RBOB gasoline or why non-reportable traders are never net long corn. Some futures markets are used primarily as vehicles for commercial traders to fix the price of forward deliveries, some are used primarily to speculate and some are used primarily in lieu of less-transparent cash markets for purposes of price discovery.
The eurodollar market
Eurodollar futures deliver 90-day time deposits of dollars held outside of the United States. The price of the contract determines the implied add-on yield paid or received for the three months following the expiration. As such, the contract can be used to bet directionally on the direction of short-term interest rates and this and trading the spread between Eurodollars and Treasury bills were its primary uses for a few years following its December 1981 introduction (see, “See TED Spread,” February 2004).
It did not take long for the market to find its primary use as a tool for pricing and hedging interest rate swaps by trading strips of contracts whose collective implied yield defined the fixed leg of an interest rate swap. A short position in one of these strips means the seller is a “fixed-rate payor” or borrower in the futures market at the strip’s implied rate; the long position is a “fixed-rate receiver” or lender at the strip’s implied rate.
As a result of this usage, the Eurodollar and similar futures such as Euribor have been the largest futures contracts in the world for years despite a near-complete lack of short-term interest rate volatility since the end of the financial crisis. The high-low range for three-month Libor has been less than six basis points over the year prior to this writing but average daily trading volume has been more than 2.32 million contracts. If this is speculative effort, we really ought to find some better speculators.
The ongoing effort by regulators to move interest rate and other swaps onto centralized clearing platforms including those operated by the CME Group, the InterContinental Exchange and Eurex will bring an increasing percentage of formerly over-the-counter swap positions into futures markets, albeit fitfully. Here they will be counted in the COT reports.
The CFTC classifies positions of fewer than 1,000 Eurodollar contracts as non-reportable; you may test this supposition for yourself by placing a 900-contract order and seeing whether your broker treats you as small-fry or not. Regardless, if we map the reportable positions for both commercial and non-commercial traders and the net difference between commercial and non-commercial reportable positions going back to January 1993, we see some very outsized net reportable positions for both commercial and non-commercial traders.
It looks as if non-commercial traders are, on aggregate, making a massive bet short-term interest rates are going to rise and commercial traders are betting already-low short-term interest rates are going to fall further. That interpretation, per H.L. Mencken, is neat, simple and wrong. The commercial traders are fixed-rate receivers in the futures market and therefore are most likely fixed-rate payors/floating-rate receivers in the swap market. The opposite applies for the non-commercial traders; they are fixed-rate payors in the futures market and therefore are most likely fixed-rate receivers/floating-rate payors in the swap market.
We can conceptualize the whole affair, borrowers taking a short position in Eurodollars in the futures market as a bearish position on interest rates and an assumption of the risk of having to pay higher rates on the cash market swaps. This position makes a lot of sense given the strong signals sent by global policymakers short-term interest rates will remain artificially low (see “Positions through time,” below).
While the public data are not disaggregated to contract level, a snapshot of open interest by quarterly Eurodollar contracts taken on May 21, 2014 shows a typical pattern of nearly 90% of the open interest held through the first three (white, red and green packs) years of the 10 years available (see “Short-term focus,” below). This distribution is typical and tells us in no uncertain terms the bulk of interest rate swaps being traded are of three years’ tenor or less. If we add the blue and gold packs, we have more than 99% of the open interest.
Now let’s ask whether the net reportable COT positions lead swap rates for tenors of one through five years, inclusive, and for the sake of completeness whether these swap rates lead net reportable COT positions. The correlations between swap rates and net reportable COT positions for lead and lag times up to 104 weeks, or two years can be calculated. The highest positive correlations are for the shorter- tenor swaps, those of one and two years, and at lead times of 53 and 57 weeks, respectively (see “Where’s the money,” below).
The direct relationship is easy to interpret: The more positive the net reportable COT position is today, the more likely a long position will be wrong and swap rates will be higher over the next year. Conversely, the net short non-commercial traders are more likely to profit over the next year on their futures positions than are net long commercial traders.
Those who like to brand commercial traders with the execrable label of “smart money” might wish to ask themselves why more than two decades of data demonstrate higher net commercial positions lead to higher swap rates and therefore to losses on commercials’ net long futures positions. The answer is a very neutral one: The commercials may very well be making much more with the swap positions than they are losing with the futures positions. They simply do not regard this as a directional bet on short-term interest rates at all.
A weak relationship
Finally, anyone who has an urge to convert this to a trading model should pause. The relationship, illustrated here by the net reportable positions leading two-year swap rates by 57 weeks, is a tepid one statistically (see “Two-year rates,” below). The r2, or percentage of variance explained, is only 0.20. Many factors affect swap rates significantly; alas, net reportable COT positions are not on that list.
Howard Simons is a long-time contributor to Futures and president of Rosewood Trading Inc. Reach him at email@example.com.