Author: Moorad Choudhry
2 Basis Trading and the Implied
Repo Rate
In this chapter we look in more detail at some fundamentals behind the basis,
including the factors that drive its behaviour, and we also consider implications of
the short future’s delivery option. There is also, in an appendix at the back of this
book, recent delivery history for the LIFFE long gilt future, for illustrative purposes
and to observe delivery patterns.
2.1 Analysing the basis
Having discussed, in the previous chapter, the theoretical foundation behind
futures prices, it is nevertheless the case that they move out of sync with the no-
arbitrage price and present arbitrage trading opportunities. A review of the US
Treasury or the gilt bond basis relative to the bond carry would show that the basis
has frequently been greater than the carry, and this would indicate mis-pricing in
the futures contracts.1 The anomalies in pricing are due to a number of factors, the
principal one of which is that the short future has the option of delivery. That is,
the short picks which bond to deliver from the basket, and the time at which it is
delivered. The long future simply accepts the bond that is delivered. It is this
inequality that is the option element of the contract.
We will take a look at this, but first let’s consider the principle behind no-
arbitrage delivery.
2.1.1 No-arbitrage futures price2
At the expiration of a futures contract, as the exchange delivery settlement price is
being determined, there should be no opportunity for a market participant to
generate an arbitrage profit by buying bonds and selling futures; by definition,
because on the last day of trading there is no uncertainty with regard to the carry
costs of the bond to delivery. In fact certain exchanges arrange it so that the time
between the last trading day and date for delivery is identical to the settlement
process in the cash market.3 On the last day, someone buying cash bonds will
receive value on the same day, and with the same accrued interest, as a long future
being delivered into. Thus carry cost is no longer an uncertainty and the price of
1
Burghardt (ibid.), page 27.
2
We adopt the approach also used in Plona (ibid, Chapter 3).
3
For instance, the Bund contract on Eurex.
39
40 Author: Moorad Choudhry The Futures Bond Basis
the futures contract in theory must equate that of the cash bond. In other words,
the basis is zero at this point.
Consider Figure 2.2, the delivery basket for the December 2001 long gilt
contract. The bonds have been priced so that they all yield 7%, the notional
coupon. Under these conditions, only one futures price will satisfy the no-
arbitrage principle. As carry is not an issue on expiry, the no-arbitrage condition is
met provided there is a zero basis for one of the deliverable bonds and no negative
basis for any of the other bonds. For instance, at a futures price of 100.09,
following the price factor conversion the equivalent bond price would be below
the market price of the 8% Treasury 2013 (the cheapest-to-deliver bond at this
level) and thus maintain a positive basis. However, for the 6.25% Treasury 2010
bond, this futures price would be equivalent to a converted bond price of 95.1443.
The market price of this bond is lower than this, at 94.9685. In theory a trader can
buy the bond at this price, sell the futures contract at 100.09 and realise a trading
gain of 0.1758 (the difference between the two prices). This is the arbitrage profit.
So the initial suggested price for the futures contract is too high. At a price of
100.06 the future no longer presents an opportunity for profit if buying the basis;
however, in theory, selling the basis against the 9% 2011 bond still generates profit.
The long future must accept delivery of any of the bonds in the basket however,
and will not be delivered this bond. So the adjusted futures price is too low.
Figure 2.1: Dec01 long gilt delivery basket as at 17 September 2001.
#Bloomberg L.P. Reproduced with permission
Author:
Basis Trading and the Implied Moorad
Repo Rate Choudhry 41
Figure 2.2: Dec01 long gilt delivery basket with yields all set at 7%.
#Bloomberg L.P. Reproduced with permission
Hence we know that the arbitrage-free futures price lies between these two
levels. In fact we obtain the no-arbitrage price by dividing the bonds’ market
prices by their respective conversion factor. These are shown in Table 2.1. The
prices in Table 2.1 are the futures prices at which there exists a zero basis for that
particular underlying bond. We can determine this relationship easily from the
definition of the basis, as shown below:
Basis ¼ Pbond Pfut 6CF ð2:1Þ
Pbond ¼ Basis þ Pfut 6CF ð2:2Þ
If we set the basis at zero, we obtain
Pbond
Pfut ¼ ð2:3Þ
CF
Bond Conversion factor Price at 7% yield Price divided by factor
6.25% Treasury 2010 0.9505874 94.9685 99.90506922
9% Converison 2011 1.1381240 114.0107 100.1742341
8% Treasury 2013 1.0793560 108.0365 100.0894817
5% Treasury 2012 0.8551727 85.3286 99.7793779
Table 2.1: December 2001 long gilt delivery basket, price at 7% notional
yield level, and zero-basis futures price
42 Author: Moorad Choudhry The Futures Bond Basis
which illustrates how the deliverable bond price divided by its conversion factor is
equal to the zero-basis futures price.
Taking this further, the futures price that would ensure that all the deliverable
bonds have a basis that is either zero, or greater than zero, is the lowest possible
zero-basis futures price. The price cannot exceed this otherwise there would be an
arbitrage opportunity. If we calculate the zero-basis futures price at different yield
levels, we will observe that when yields lie above the contract notional coupon,
generally the shortest-dated bond carries the lowest zero-basis futures price. If
yields lie below the notional coupon, frequently the longest-dated bond carries the
lowest zero-basis futures price, and so is the CTD bond. This has been observed
empirically by a number of authors, and formalised by Benninga (2001), for
instance.4 The observation reflects a bias in the conversion factor. We illustrate
this bias in Figure 2.3, for the shortest-dated and longest-dated deliverable bonds
for the June 2000 long gilt future. It shows where the futures price meets for both
bonds is at the contract notional coupon of 7%, this being known as the inflection
point. The conversion factor bias determines which bond is the cheapest-to-
deliver bond, based on yield levels of bonds in the basket and their position
relative to the contract notional coupon.
The relative bias of the conversion factor is a function of the duration property
of the bonds. For further information on this see Meisner and Labuszewski (1984)
and Arak et al (1986).
Figure 2.3: Illustrating bias in the conversion factor, June 2000 long gilt
future
4
See also Appendix 2.2 for a short chat on a general model of the CTD bond.
Author:
Basis Trading and the Implied Moorad
Repo Rate Choudhry 43
2.1.2 Options embedded in bond futures contracts
In the US Treasury bond contract, the short future has the following options with
regard to delivery:
j the wild card or time to deliver option;
j the quality option.
The short future may deliver during any business day of the delivery month,
although the administrative process involved takes place over three days. The first
day is the position day, the first day on which the short declares an intent to
deliver. The first possible day that this can occur is the second-to-last business day
before the first day of the delivery month. There is an important advantage to the
short future on this day: at 14:00 hours that day the exchange settlement price and
hence the invoice amount to be received by the short are fixed. However, the short
does not have to announce an intent to deliver until 20:00 hours on that day.
During the six hours after the settlement price, if interest rates rise and the bond
price falls, the short will earn the difference between the actual price received for
the bond, which is the EDSP set at 14:00, and the price that they will have to pay to
acquire the bond. Equally if the bond is in the short’s possession, the price
received for the short future delivery will be higher than the market price of the
bond. The second day is the notice of intention day, and during this day the
exchange clearing house identifies the short and long parties to each other. The
third day is the delivery day; during this day the short future must deliver the bond
to the clearing house. The long future pays the invoice amount based on the
settlement price fixed on the position day.
The short future has an option of when to exercise the time to deliver option
from the penultimate business day prior to the start of the delivery month up to
the last business day of the delivery month, known as the last trading day. At 14:00
hours on that day the final settlement price is determined, which stays constant to
the end of the month. The short future is left with one more wild card at this point,
known as the end-of-the-month option. Assuming that the short has not declared
an intent to deliver until now, the settlement price for the contract is now
irrevocably fixed. At this point the short still has five more days before having to
declare an intention. In this case the possibility still exists for profit generation if,
for example, the trader purchases the bond at the settlement price and holds it to
delivery, earning the accrued interest. This will generate a carry profit if the bond’s
running yield is higher than its specific repo offer rate.
The last option advantage of the short future is the quality option. This is the
option to deliver any of the bonds within the delivery basket. As we have seen the
bonds can vary widely in coupon and maturity, and hence yield, and despite the
conversion factor equalisation there is a bias in this factor that means at yield
levels above the notional coupon, long-dated bonds are the cheapest, and vice
versa if yields are below the notional level. The quality option also presents the
44 Author: Moorad Choudhry The Futures Bond Basis
long future with potential problems if there is a change in yields sufficient to
change the CTD from one bond to another. If this comes as a surprise to the basis
trader, it can be potentially very serious.
The delivery options available to the short future carry value, and this is
reflected in the difference between the gross basis and the net basis. In theory the
value of the delivery options, when added to the price of the futures contract,
should equal the value of the bond together with the carry.
2.2 Bond delivery factors
2.2.1 The cheapest-to-deliver
The deliverable bond is of course the cheapest-to-deliver bond. This is not, as a
junior trader once suggested to the author, the bond with the lowest market price
(although it might well be). The CTD bond is the one that maximises the return to
the arbitrageur engaging in buying the basis, that is, buying the bond and
simultaneously selling the future, holding the bond to expiry and then delivering it
into the futures contract. Some market practitioners use the implied repo rate to
identify the CTD, while others prefer the net basis method. We assess both now.
Essentially a good approximation for the CTD is to compare the basis for a bond
with its total carry costs to delivery. The difference between the two is the net
basis. However, this method may produce incorrect CTD rankings when the net
basis values for two bonds are very close. When this happens the net basis for the
bond that is actually the cheapest is higher than another bond in the basket,
despite the fact that it is the cheapest. This happens because since the net basis
method measures cheapness by comparing net basis to the dirty price of the bond,
and this price is related to coupon size, maturity and given yield levels, the actual
running cost is sometimes not captured. This can produce a lower net basis for a
bond that is not actually the cheapest. For this reason the most accurate method
by which to identify the CTD is to pick the bond with the highest implied repo rate.
From the previous chapter we saw that the IRR is the hypothetical return achieved
by going long the basis and running this position to expiry. It is in effect a repo
agreement with the futures market, and the calculation fully accounts for the
bond’s purchase price. The IRR method is recommended by academic writers
because of the way that it is calculated; to reiterate from Chapter 1:
Pinvoice M
IRR ¼ 1 6 ð2:4Þ
Pbond n
where
Pinvoice is the invoice price;
Pbond is the cash bond purchase price;
M is the day base (365 or 360);
n is the days to delivery.
Author:
Basis Trading and the Implied Moorad
Repo Rate Choudhry 45
From (2.4) we see that the bond currently trading at a price that results in the
highest ratio of futures invoice price to the purchase price will have the highest
implied repo rate. In other words, the bond with the lowest purchase price relative
to its invoice price is the CTD bond.
Nevertheless net basis is still popular amongst traders because it identifies the
actual loss (when negative) from the basis trade, and as such is a more quantitative
measure.
2.2.2 Selecting delivery time
Another advantage of the IRR measure is that it clearly indicates the time that the
short future should deliver the bond. Consider Table 2.2 for the December 2000
long gilt contract.
This shows the implied repo rates for delivery on the first day of the delivery
month and the last day of the month. For all of the bonds the IRR for delivery on
the last day is higher than that for delivery on the first day. It is apparent that the
(theoretical) return from a long basis trade would be higher if the delivery date was
delayed to the last possible moment, and so in this case the short future would
elect to deliver on the last business day.
In fact our illustration is a peculiar one, because the gilt yield curve was inverted
at this time. In a positively-sloped yield curve environment, higher IRRs will result
for longer-term trades and the decision of the short over when to deliver is an
obvious one. The other reason why the short future would prefer to delay delivery
is because early delivery eliminates the value of the option element that the short
future possesses. It is a bit like early exercise of an American option eliminating
the option’s time value. In a negative yield curve environment the decision is not
so clear cut, although early delivery still removes the short’s option advantage.
However the market repo rate would need to be considerably higher than the IRRs
to justify early delivery, and this was not the case here. Where the market repo rate
is higher, the short future will be running a carry cost each day the basis position is
maintained, so this will suggest early delivery.
IRR to first IRR to last
Closing day Conversion Gross Net delivery delivery Actual
Bond cash price factor basis basis day % day % repo rate %
5.75% 7 Dec 2009 106.0590 0.9174728 0.166 0.300 1.04 6.26 4.52
9% 12 Jul 2011 133.8585 1.1479281 0.951 0.509 3.32 2.16 4.52
6.25% 25 Nov 2010 110.8310 0.9467478 1.217 1.040 5.79 1.28 4.52
9% 6 Aug 2012 136.2685 1.1576368 2.237 1.808 6.78 3.74 4.52
8% 27 Sep 2013 129.9925 1.0835676 4.537 4.190 22.70 15.74 4.52
Table 2.2: Identifying the cheapest-to-deliver and the optimum delivery
time, December 2000 long gilt basket.
Source: LIFFE, JPMorgan Chase Bank
46 Author: Moorad Choudhry The Futures Bond Basis
In theory there should only be two days when the short future delivers: the start
or end of the delivery month; but changes in the yield curve, a particular bond
yield level, and market repo rates may make it necessary to deliver on dates in
between. Consider Table 2.3, which shows the delivery pattern for the September
1996 and December 1998 long gilt contract on LIFFE. In September 1996 the gilt
yield curve was conventional and positively sloping, and apart from a small
handful of deliveries just prior, all deliveries were made on the last eligible day of
the month. In December 1998 however, the yield curve was negatively sloping, and
this is reflected in the pattern of deliveries. It would appear that some market
participants had confused ideas, and although in general bonds were delivered
early in the month, some deliveries were still being made right in to the middle of
the month. This despite the fact the delivery parties would be experiencing
negative carry each day they did not deliver. Essentially, this would have been a
cost to all those that did not deliver on the first day.
For readers’ interest we list the delivery histories from March 1996 long gilt
contract through to the June 2001 contract, in the Appendix to this book.
September 1996 December 1998
Date 9% 2011 7.75% 2006 Date 9% 2008
2 1 1,238
3 2 1,787
4 3 26
5 4 4,116
6 7
9 8 200
10 9
11 10
12 11 8
13 14
16 15
17 16
18 17
19 18 3
20 21
23 22
24 23
25 16 24
26 29
27 3,500 4,515 30
Total 3,516 4,515 7,378
Table 2.3: Bond delivery patterns for two gilt futures contracts,
reflecting the shape of the yield curve at time of delivery
Author:
Basis Trading and the Implied Moorad
Repo Rate Choudhry 47
2.2.3 Changes in CTD status
A bond may be replaced as CTD if there are changes in relative yield levels of
deliverable bonds, if the shape of the yield curve changes or if specific repo rates
turn special for certain bonds and not others. Benninga (2001) amongst others has
identified the following general rules:
j where the yield level is below the notional coupon level, the bond with the
lowest duration will be the CTD. If the yield level is higher than the notional,
the bond with the highest duration will be the CTD;
j where bonds have roughly identical durations, the bond with the highest yield
will be the CTD.
Bond yields are relative however, and the bonds in the basket will trade at different
yield levels. A large relative shift can bring about a change in CTD status, while
overall yields remain roughly at the same level. A more significant yield shift or
change in the shape of the curve can also have this effect. The yield on any
particular bond is market-determined, and is a function of a number of factors,
such as its liquidity, benchmark status, and so on. If there are two or more bonds
in the delivery basket with approximately identical duration values, the bond with
the higher yield would have the lower converted price, and therefore would be the
CTD bond.
Remember that duration5 is an approximate measure of the percentage change
in the price of a bond for a 1% change in the bond’s yield. For a given change in
yield then, the prices of bonds of higher-duration will change by a greater amount
than lower-duration bonds. This is worth bearing in mind because it is behind the
bias in the conversion factor introduced in the previous section. Let’s reiterate this
here. A contract’s conversion factors are the approximate prices at which
deliverable bonds yield the notional yield level. For the Treasury long bond
then, all conversion factors are approximately neutral at a yield level of 6%. If every
bond in the delivery basket was trading at 6% yield, their converted prices (the
price at this yield, divided by the relevant conversion factor) should be equal to
100, or close to 100. At this yield level then, the short future is in theory indifferent
as to which bond to deliver. However, at yield levels above or below the notional
level of 6%, duration of the deliverable bonds becomes relevant.
At yields below 6%, as yields fall the rise in price of a lower-duration bond is
relatively lower than the price rise of a higher-duration bond. Thus, the low-
duration bond becomes the CTD bond. As yields rise above 6%, the higher-
duration bond experiences a smaller fall in price than the lower-duration bond,
and it becomes the CTD bond.
5
We refer here of course to modified duration, which is usually simply termed ‘‘duration’’
in the US Treasury market.
48 Author: Moorad Choudhry The Futures Bond Basis
Appendices
Appendix 2.1: General rules of the CTD bond
Beninnga (1997) has suggested some general rules in a non-flat yield curve
environment that may be taken to be a general model for basis trading. His study
analysed the character of the CTD bond under four different scenarios, as part of a
test of the following circumstances. When the term structure is flat, the CTD bond
is the one with:
j the highest duration if the market interest rate is higher than the notional
coupon;
j the lowest duration if the market interest rate is lower than the notional
coupon.
Benninga suggests that under certain scenarios, notably when:
j the market yield is higher than the notional coupon and there are no
deliverable bonds with a coupon lower than the notional coupon, and when
the market yield is higher than the notional coupon, and
j there are no deliverable bonds with a coupon higher than the notional,
the duration rule does not always apply.
The conclusions of his analysis are that:
j the CTD bond invariably has the highest coupon of the deliverable bonds,
where the market yield is lower than the notional coupon, otherwise it has the
lowest coupon of the deliverable bonds. The analysis assumes that the bonds
possess positive convexity, but the results are not dependent on the shape of
the yield curve;
j when market rates are lower than the notional coupon, the maturity of the
CTD is the shortest of all deliverable bonds; again, if the market rate lies above
the notional coupon, the CTD bond will have the longest maturity if it also has
a coupon greater than the notional coupon. If the coupon of the CTD bond is
lower than the notional coupon, Benningna concludes that the CTD will have
neither the longest or the shortest maturity in the delivery basket.
Certain anecdotal observation appears to confirm these generalities.
Appendix 2.2: A general model of the CTD bond
The price today (or at time 0) of a bond is generally given by (2.5):
T
P ¼ $0 Ce rt dt þ 100e rT ð2:5Þ
where C is the bond cash flow and T is the bond maturity date.
Author:
Basis Trading and the Implied Moorad
Repo Rate Choudhry 49
The discount factor at time t for one unit of cash at time s t when the time t
spot interest-rate is r is given by e rt . The value of the conversion factor for a bond
with maturity T and coupon C delivered at time F, the expiry date of the futures
contract, is given by (2.6):
T F
CF ¼ $0 Ce cs ds þ e cðT F Þ
ð2:6Þ
Ce cs T F cðT F Þ C 1 e cðT F Þ
¼ 0 þe ¼ þ e cðT F Þ
c c
where c is the notional coupon of the futures contract.
Selected bibliography and references
Arak, M., Goodman, L., Ross, S., ‘‘The Cheapest to Deliver Bond on the Treasury
Bond Futures Contract’’, Advances in Futures and Options Research 1, 1986, pp.
49–74
Benninga, S., Financial Modeling, MIT, 1997, Chapter 18
Kolb, R., Understanding Futures Markets, Kolb Publishing 1994, Chapter 9
Meisner, J., Labuszewski, J., ‘‘Treasury Bond Futures Delivery Bias’’, Journal of
Futures Markets, Winter 1984, pp. 569–572
