Dan Norcini and Gold Spreads
Yesterday Dan Norcini wrote a nice commentary on “gold spreads”, which are essentially a form of the gold basis expressed in the prices of different futures contracts. His explanation of how spreads, the basis, contango and backwardation work is very straightforward and easy to understand and everybody who seeks to understand gold and silver as well as the markets in general should read it. In particular I would focus on the following:
However, there is one thing that no amount of market intervention and price manipulation can succeed in doing and that is in changing the basic structure of the futures market as evidenced by the relationship of the front month contracts to the later dated contracts.
In trading terms, we refer to the “spread” between the front month and a back month/months or the difference in price between the two, as a gauge of demand for that particular commodity. As a general rule, when the front month trades at a discount to the next month or to a later-dated month, the structure of that particular commodity futures market is normal or in contango. A market in contango will see those distant month contracts trading at enough of a premium to the front month to account for any storage charges, insurance against loss and interest rates. Simply put, a seller has to be recompensed for his/her expense in storing a commodity while they wait to sell it into the market at some point in the future.
Whenever a market begins to see this “spread” between the front month and the next month or more distant months begin to tighten or narrow, then something is beginning to change regarding the demand/supply picture in that particular commodity. Why is this? Because the market is ratcheting up the front month price and attempting to send a signal to potential sellers that demand is increasing and that they are better served by selling sooner rather than later. Economically speaking, the incentive to store the commodity, pay all those storage costs, insurance costs, etc,. is not worth the increased cost that they might hope to receive at some point in the future. “Sell it to us now and we will pay you more for it than if you try to sell it later”, is the message the market is sending.
When markets begin moving in this direction, narrowing the spread, they are said to be moving towards a condition known as, “backwardation”. True backwardation occurs when the front month moves to a PREMIUM over the next month and particularly over the next set of three or four different contract months ( a note here - generally a market will not go into backwardation more than a few distant contracts out because it is assumed that the increased demand will result in increased production at some point and induce producers of that particular commodity to increase production on out into the more distant future bringing the demand/supply picture into more of an equilibrium. That will serve to bring the market back into a more normal structure of contango).
Backwardation is a powerful signal of very strong demand that is attempting to send a signal to the market that it needs more of that commodity to satisfy existing levels of demand. While market price manipulation can be somewhat effective short term for fogging signals generated from a rising price in gold for example, it is generally unable to affect the spread structure of the entire set of futures contracts listed on the board at any given time.
Even ignoring what Mr. Norcini says about Central Bank and bullion bank manipulation of gold prices, he is absolutely right that you cannot manipulate both the price and the basis at the same time (although see below for an exception). This is because manipulating the basis requires being in the market with both legs (long and short), which prevents effective price manipulation, whereas manipulating the price requires being in the market with one leg (long or short), which prevents effective basis manipulation. There is more to it than that but that is the simple explanation.
As far as what Mr. Norcini has noticed (and charted), I will repeat what I said last December when Professor Fekete and others got excited about (alleged) backwardation in gold, which is that the phenomenon was largely the result of falling interest rates and the use of illiquid spot month contracts to perform the spread or basis calculations.
First let’s talk about interest rates. With gold (and silver to a lesser extent), one of the “costs” of holding inventory is the interest foregone on holding cash that could otherwise be invested. Not only that, but in comparison to other commodities, the storage, insurance and transportation costs of gold and silver are minimal. Indeed, this is one of the pillars why gold is money. Thus, there is a direct — although not proportional — relationship between the gold basis and interest rates (on fiat money). Namely, the lower the rate of interest (in the case of gold it is appropriate to use LIBOR as the rate of interest), the lower the spread in gold.
For example, LIBOR fell precipitously and so did the gold basis last December. The 1-month LIBOR went from around 4.50% on October 13 to 0.45% on December 30, a historic drop of 90% in under 3 months. With the opportunity cost of holding gold for 1-month being essentially zero, it is no wonder that the 1-month LBMA forward rate (which is essentially equivalent to the spread between a spot COMEX gold contract and a contract one month in the future) fell from 2.16% to 0.16% during the same time. Moreover, 1-month LIBOR peaked around 0.55% in the middle of March 2009 after bottoming in January and has since dropped back down to 0.45%. As a result, the LBMA forward rate, COMEX futures spreads, and even the traditional basis calculation (cash vs. front-month futures) have shown a shrinking contango. This same pattern is clearly visible in Mr. Norcini’s charts.
Bottom line, for a basis or spread trend in gold or silver to make the most sense, we must recognize and isolate the influence of interest rates. This is something that I have been researching and analyzing for a while now and I hope eventually to share the summary of my work with all of you (details and trading strategies will be reserved for Founding Members of Metal Augmentor). In the meantime, I am happy to share these observations with Dan Norcini and everybody else doing important work on the gold/silver spread and basis.
The next point I’d like to make is that even though Mr. Norcini talks about the front month and outer months of COMEX gold, what he is actually comparing are the delivery month (April ‘09) and the active front month (June ‘09). This is important because the delivery month futures are typically quite illiquid. When they are not illiquid, it usually means something is afoot and that it might then make sense to use the delivery month in spread and basis calculation. There have been occasions in the past when this was the case. Otherwise, the general rule is that illiquid equals inaccurate.
For example, yesterday, April 21, the April ‘09 contract traded 655 times compared to 75,861 for the June contract, a difference of 100-to-1. Even if we assume these trades occurred only during the pit session — about 5 hours — that means the April contract traded at most once every 30 seconds (although there were several instance when it did not trade for many minutes) while the June contract traded much more frequently. Thus, any particular trade in the April contract cannot be effectively compared to any particular trade in the June contract. Moreover, Mr. Norcini appears to be using daily closing prices in his charts and analysis, and these prices are often not even based on actual trades, especially for contracts other than the active month (currently June is the active month).
From the NYMEX rulebook:
AMENDMENTS TO COMEX RULES 4.91, 4.92 AND 4.93Rule 4.91 – “Futures Settlement Prices”
(a) Active Month. The settlement price of the most active futures contract month shall be the average (rounded off to the nearest price tick) of the highest and lowest prices of the trades reported during the closing period, except as otherwise provided in this rule or in Rule 4.93 (”Use of Discretion to Establish Settlement Price”).
(b) In All Other Delivery Months. The settlement prices shall be determined based upon spread relationships between and among contract months, which relationships shall be determined in the judgement of the Settlement Price Committee with: (a) greater weight given to spreads executed later in the trading day in large volumes, and (b) lesser weight given to (i) spreads traded in lesser volumes, (ii) spread bids and offers actively represented later in the trading day, and (iii) spread transactions, bids and offers from earlier in the trading day.
In other words, the closing price that Mr. Norcini is using for April COMEX gold is “determined in the judgement of the Settlement Price Committee”. So much for no opportunity to manipulate spreads! Even if we don’t subscribe to conspiracy theories, it should be easy to see why using such “prices” can be problematic when we are dealing with sensitive calculations like spreads and the basis. This is why I no longer use “closing” or “last” prices exclusively in my analysis, preferring to use actual trading data whenever possible.
To better illustrate what I’m saying, just take a look at the open-high-low-close of the April and June ‘09 COMEX for yesterday, April 21:
Open: 886.9 - 886.5 (0.3 backwardation)
High: 895.1 - 896.4 (1.3 contango)
Low: 880.7 - 879.5 (1.2 backwardation)
Close: 882.1 - 882.7 (0.6 contango)
As you can see, there was quite a bit of variation in the prices of these two contract during the trading day, mostly because of the illiquid nature of the April contract (this illiquidity would also impact the accuracy of the April “closing” price determined by the Settlement Price Committee). Still, I can find plenty of relevance and importance in the fact that the spread between the high and low prices had an absolute variation of $2.50 whereas the closing price implied a spread of just $0.60. That leaves plenty of room for error.
The last point I’ll make is that charting the spread between two specific futures contracts over time will usually result in some degree of convergence between the prices — a narrowing of the spread — simply because gold spreads are not entirely linear. Why not? The same reason that interest rates are not linear but rather trace out an “interest rate curve”. Specifically, comparing two contracts such as June and December ‘09 COMEX gold over a period of time such as one or two years will involve each contract moving through a particular slope of the “spread curve” at different times. Since the curve is typically flatter further out in time, the impact on spreads is less noticeable (without making appropriate adjustments) the further away both contracts are from expiration. Conversely, the curve is steeper closer to expiration, and the impact of the differential movement of the contracts through the curve becomes more noticeable (without making appropriate adjustments). The net result is that a chart plotting the absolute value of the spread between two forward prices in gold, or between a forward price and the spot price, will always seem to converge over time even though the spread may actually be constant. I bolded that statement because a number of people I helped in the past to understand and track the basis have made this common but fundamental mistake, essentially rendering as meaningless their analysis of the basis or spread in gold and silver.
Now, please don’t get me wrong, there is still some value in the types of calculations done by Mr. Norcini and others as they can show roughly the relative size of the spread or basis over intermediate periods of time. It’s just that we cannot use this method to make accurate estimations of the basis over either very long time periods (more than a few months) or short time periods (less than a few weeks). That leaves quite a small window where this type of calculation is meaningful, but it is still a window nonetheless.
One solution to this small window is to create separate layers of filtering to (1) remove price discrepancies and trading artifacts associated with contract illiquidity, (2) adjust for the relative level of interest rates, and (3) normalize the spread curve to allow direct comparison of prices for contracts expiring at different times. Of course the simplest way to deal with several of these issues is to use a robust spot price to anchor one side of the basis calculation, but alas such price data is reserved for bullion banks and not available to mere mortals like us. Regardless, the research I have conducted to date, while not conclusive or ready to be shared, does support the notion that the contango has markedly shrunk over the past year independent of interest rates or other inputs and that we have indeed come close to “true backwardation” on a number of occasions. Right now is just not one of them (although tomorrow could be different, so stay tuned).
Recent Comments