If a bank lends you money for one year and lends money to someone else for ten years, it is very likely that the rate of interest charged for the one-year loan will di\ufb00er from that charged for the ten-year loan. Term-structure theory has as its basis the idea that loans of di\ufb00erent maturities should incur di\ufb00erent rates of interest. This basis is grounded in reality and allows for a much richer and more realistic theory than that provided by the

for studying term-structure models. In these notes we will often assume that there arem compounding periods
per year, but it should be clear what changes need to be made for continuous-time models and di\ufb00erent
compounding conventions. Time can be measured in periods or years, but it should be clear from the context
what convention we are using.

the spot rate,st, is the rate of interest charged for lending money from today (t= 0) until timet. In particular, this implies that if you lendA dollars fort years2 today, you will receiveA(1 + st/m)mt dollars when thet years have elapsed. Theterm structure of interest rates may be de\ufb01ned to constitute the sequence of spot rates,

bonds. Government bonds should be used as they do not beardefault risk and so the contracted payments are sure to take place. For example the price,P , of a2-year zero-coupon government bond with face value$100, satis\ufb01esP= 100/(1 + s2)2 where we have assumed an annual compounding convention.

1Nor does the YTM framework preclude arbitrage opportunities.
2We assume thatt is a multiple of 1/m both here and in the de\ufb01nition of the discount factor,dt, above.
3It is quoted on an annual basis unless otherwise stated.

where time is measured in periods and there arem periods per year. In particular, the present value at datei of
a cash\ufb02ow,xj , that occurs at datej > i, is given bydi,j xj . It is also easy to see that these discount factors
satisfydi,k= di,j dj,k fori < j < k and they are consistent with earlier de\ufb01nitions.

forward rates. There is no inconsistency in this de\ufb01nition as the forward rates de\ufb01ne the spot rates and the spot rates de\ufb01ne the forward rates. We also remark that in ann-period model, there aren spot rates andn(n+ 1)/2 forward rates. The set4 of short forward rates,{rf

spot rates,sj\u2212i, that are expected to prevail at timei. While this has some intuitive appeal, if the hypothesis was true then the fact that the spot rate curve is almost always upwards sloping would mean (why?) that the market is almost always expecting spot interest rates to rise. This is not the case.

bonds to longer maturity bonds. This is because longer-maturity bonds are generally more sensitive to changes in the general level of interest rates and are therefore riskier. In order to persuade risk-averse individuals to hold these bonds, they need to be sold at a discount, which is equivalent to having higher interest rates at longer maturities.

interest rates at datet2 fort1\ue000= t2. The rationale for this is that short-term securities might be of interest to one group of investors, while longer term securities might be of interest to an altogether di\ufb00erent group. Since these investors have nothing in common, the markets for short- and long-term securities should be independent of one another and therefore the interest rates that are set by the market forces of supply and demand, should also be independent. This explanation is not very satisfactory and explains very little about the term structures that are observed in practice.

Two bonds,A andB both mature in ten years time. BondA has a7% coupon and currently sells for$97, while bondB has a9% coupon and currently sells for$103. The face value of both bonds is$100. Compute the price of a ten-year zero-coupon bond that has a face value of$100.

(i)preclude arbitrage possibilities
(ii)betractable in the sense thatrelevant security prices in the model are computable
(iii)provide an adequate \ufb01t torelevant empirical data.

The models we will develop in this course will, by construction, usually satisfy (i) automatically. We will spend a lot of time showing that they also satisfy (ii) though this will of course depend on the particular security prices we are seeking to compute. We will not focus on (iii) though we will sometimes discuss the issue ofcalibration.

We now develop some further notation and review some of the important concepts and results of \ufb01nancial engineering that will be used throughout the course. These de\ufb01nitions are initially given in the context of discrete time models. (We will give their continuous-time analogues later in the course.)

Consider a \ufb01nancial market withN+1 securities and true probability measure,P . We assume that the investment horizon is[0,T ] and that there are a total ofT trading periods. Securities may therefore be purchased or sold at any datet fort= 0,1,...,T\u22121.

In order to respect the evolution of information as time elapses, it is necessary that\u03b8t be apredictable
stochastic process. In this context, \u2018predictable\u2019 means that\u03b8t cannot depend on information that is not yet
available at timet\u22121. We also say that\u03b8t isFt\u22121-measurable where we useFt\u22121 to denote all the information
in the \ufb01nancial market that is know at datet\u22121.