Deferred Carry - When Does Contango/Backwardation Occur?

Introduction

This post is the first in a series of posts on deferred calendar spreads in the general commodities universe. A calender spread is where we take opposing views on different parts of the futures curve of the same commodity. We define a bearspread by taking a short and long position in the front and deferred part of the futures curve respectively. A bullspread takes the opposite view, a long and short position and in the front and deferred parts of the curve respectively.

We define a spread

\[S = \text{Front} - \text{Deferred}\]

to be contango when the deferred part of the curve is trading at a premium compared to the front part, i.e. \(S < 0\). Conversely, when the spread \(S > 0\) the spread is said to be in backwardation.

Many commodities exhibit a collapse to contango as the front end of the curve nears expiry. In this post we study the onset of contango accross a range of commodities. For each commodity we determine a collection of calendar spreads, these mostly consits of consecutive contract months, but also the year on year spreads and some inter season spreads that we find particularly interesting. We use the daysdiff - number of days to expiry of the front contract - variable as a measure of time. In order to determine statistiscs we bin the daysdiff parameter into 30 day buckets. These then roughly correspond to monthly buckets. We use the convection that

  • 0 represents days 0 to 29
  • 1 represents days 30 to 59
  • \(\dots\)
  • \(n\) represents days \(30n\) to \(30(n+1) -1\)

Withing each of these buckets we determine the mean value of the spread. From historical data we then determine the fraction of the times the mean spread was in contango.

Below we consider data form 2009 up to the most recent.

Grains and Oilseeds

Below we show the evolution of contango for KW. We see that all the spreads shown tend to be in contngo from six months to expiry of the front month contract. Of these contracts the KN spread is the slowest to collapse to contango.

Below we show the evolution of contango for W. The evolution to contango of W occurs on a faster timescale compared to that of KW. Here we see the majority of the spreads above the 50% line at 12 months prior to expiry. The slowest spread to collapse to contango is HN.

Below we show the evolution of contango for C. Nine months prior to expiry all the spreads shown have an above 50% chance of being in contango. Here the NZ spread is one of the slowest to reach contango but ends up with a high percentage in contango close to expiry.

Below we show the evolution of contango for S. Here we split the data into those spreads that end in contango the majority of times and those that end in backwardation. Here the XX spread is interesting as it converges to contango at expiry. Many of the S spreads actually prefer to end in backwardation.

Below we show the evolution of contango for SM. Similar to the case of S we split the plots into those that have a majority of cases ending incontango and backwardation respectively. The backwardation cases are interesting and can be used as a hedge to a portfolio of bearspreads.

Below we show the evolution of contango for BO. This is a commodity that loves to trade in contango. Here we see that all the spreads shown decays into contango with 5 mongth until expiry of the front contract.

Soft Commodities

Below we show the evolution of contango for CC. Graphically it looks like all the spreads tend to end up in contango. The ones with the greatest fraction of contagoes are KN and UZ.

Below we show the evolution of contango for QC. Here we see that some spreads prefer contango while other prefer backwardation. This might be useful when thinking of ways to contruct butterflies with QC.

Below we show the evolution of contango for KC. This is a commodity that loves to trade in contango. Here we see that all the spreads prefer to trade in contango.

Below we show the evolution of contago for DF. Similar to KC we see that DF also likes to trade in a contango, but not quite to the same extent. From a relative value point of view we know that the driver of the spread is most likerly KC when stock levels are suppressed. However the data shown here, together with the long term roll characteristics of this relative value pair suggest that being long DF and short KC might be the better option.

Below we show the evolution of contango for SB. Interestingly we see that the HK spread prefers to trade in a backwardation while the other spreads shown tend to expiry in contango.

Below we show the evolution of contango for QW. From previous work we know that QW prefers to trade in a backwardation. Below we see that the majority of the spreads shown also agree with this observation. The ZH spread is particularly interesting where it shows a high fraction of contango at around 7 months to expiry before it eventually crashed down into a majority backwardatin.

Meats

Below we show the evolution of contango for FC. The majority of the spreads shown en in contango except FH and XF.

Below we show the evolution of contango for LC. Here we split the data into spreads that end in contango and backwardation respectively. Two interesting observations is that the JM spread always trades in backwardation while the VZ spread nearly always trades in a contango.

Below we show the evolution of contango for LH. This is a commodity that has a negative carry most of the time. Some of the more interesting spreads to consider are those that show early backwardation and eventually expiry in contango. One that is particularly interesting is MN.

Remarks

This post is a first stab at looking when different calendar spreads evolve into contango or backwardation. In the next post we will study the degree of contango.

Avatar
Mauritz van den Worm
Portfolio Manager and Quantitative Researcher

My research interests include the use of artificial intelligence in managing commodity portfolios

Related