Derivatives and Internal Models: Modern Risk Management

Part IFundamentals

© The Author(s) 2019

H.-P. Deutsch, M. W. BeinkerDerivatives and Internal ModelsFinance and Capital Markets Serieshttps://doi.org/10.1007/978-3-030-22899-6_1

1. Introduction

Hans-Peter Deutsch¹   and Mark W. Beinker²  

(1)

Niedernhausen, Germany

(2)

Friedrichsdorf, Germany

Hans-Peter Deutsch (Corresponding author)

Email: hans-peter.deutsch@d-fine.de

Mark W. Beinker

Email: mark.beinker@d-fine.de

Electronic supplementary material The online version of this chapter (https://​doi.​org/​10.​1007/​978-3-030-22899-6_​1) contains supplementary material, which is available to authorized users.

The explosive development of derivative financial instruments continues to provide new possibilities and increasing flexibility to manage finance and risk in a way specifically tailored to the needs of individual investors or firms. This holds in particular for banks and financial services companies who deal primarily with financial products, but is also becoming increasingly important in other sectors as well. Active financial and risk management in corporate treasury can make a significant contribution to the stability and profitability of a company. For example, the terms (price, interest rate, etc.) of transactions to be concluded at a future date can be fixed today, if desired even giving the company the option of declining to go ahead with the transaction later on. These types of transactions obviously have some very attractive uses such as arranging a long-term fixed-rate credit agreement at a specified interest rate a year in advance of the actual transaction with the option to forgo the agreement if the anticipated need for money proves to have been unwarranted (this scenario is realized using what is known as a payer swaption) or providing a safeguard against fluctuations in a foreign currency exchange rate by establishing a minimum rate of exchange today for changing foreign currency into euros at a future date (using a foreign currency option).

However, the complexity of today’s financial instruments and markets and the growing pace of technological progress have led to an almost uncontrollable increase in the risks involved in trading and treasury while simultaneously reducing drastically the time available for decision making. Thus, the inappropriate use of financial instruments may quickly result in losses wiping out gains achieved over years in a company’s primary business (for example, the production and sale of automobiles or computer chips).

In recent years we have seen an increase in the number of sizable losses incurred in consequence of derivative transactions which, in some cases, have resulted in bankruptcy. This phenomenon has not been restricted to banks but has involved companies in various other sectors as well. Spectacular examples include Metallgesellschaft (oil futures), Procter & Gamble (speculation with exotic power options), Orange County (interest rate derivatives, highly leveraged portfolio), Barings (very large, open index futures positions), Daiwa Bank (short-term US Bonds), NatWest Markets (incorrect valuation of options in consequence of incorrect volatility assumptions), the hedge fund LTCM, the Subprime-Crisis in the US 2007/2008 (ABS structures), Societe Generale (speculations of a single equity trader), etc., etc.

Not only have financial instruments become increasingly risky (more volatile), so have the markets themselves. Since the beginning of the 1980s we have seen a fundamental change to the economic framework in the financial world. In today’s investment environment, yields, foreign currency exchange rates, commodity and stock prices can shift daily to an extent which would have been inconceivable in previous years. Increased market fluctuation (volatility) is the financial markets’ reaction to developments such as the accumulation of capital, the globalization of financial markets, an increase in the budget deficits of the leading industrialized nations and the dismantling of government regulations, to name just a few.

The main prerequisite for the continued success of a bank or treasury department in an environment of highly volatile markets and extremely complex (and thus very sensitive) financial instruments is a sound understanding of the products being traded and efficient management of the risk involved in these transactions. Derivatives are therefore the main reason for and the most effective means of conducting risk management, and thus can be viewed as the be-all and end-all of risk management.

The pricing of derivatives and structured financial instruments is the foundation, the conditio sine qua non, of all risk management. To this end, all derivatives under consideration, regardless of their complexity, are broken down into their basic components and systematically categorized as spot, futures and option transactions. This process is called stripping. Understanding the fundamental instruments and their valuation, as well as stripping poses a serious challenge for those involved, requiring skills which in the past were not called for to any extent in the financial world, such as the following

Theoretical demands: statistics, probability theory, time series analysis, etc., essential to the understanding of concepts such as value at risk and the estimation of underlyings and their correlation; differential equations, martingale theory, numerical analysis, financial mathematics, etc., for pricing financial instruments.

Trading demands: understanding the function of increasingly complicated securities and transactions (e.g., by stripping); making imaginative choices of derivatives or constructing new instruments to realize a particular strategy while taking into consideration the often surprising side effects; developing complicated hedging strategies; operating highly complex computer systems.

IT demands: client-server computing, distributed systems, object-oriented programming, new operating systems, network engineering, real time information distribution, Internet and intranet architecture and increasingly support for parallel algorithms on computer grids, multi-core CPUs and graphic processors (GPUs).

The theoretical and trading aspects in particular pervade all business hierarchies, since it is equally important for the trader/ treasurer to be able to understand the risk involved in his or her portfolio as it is for a managing director or the board of directors whose decisions are based on this information.

Thus, the aim of this book is to provide a detailed presentation of the methods and procedures used in pricing financial instruments and identifying, quantifying and controlling risk. The book is structured as follows:

Part I introduces the fundamental market parameters which govern the price and risk of financial instruments. These include the price of stocks, commodities, currencies, etc. and naturally interest rates. A stochastic model, known as the random walk and its generalization, the Ito process, will be used to model these fundamental risks, also known as risk factors, and will serve to describe both the deterministic and the random aspects of the risk factors. In doing so, an additional parameter, called volatility will, similar to the above mentioned risk factors, play an important role in the price and risk accorded to a financial instrument.

Finally, in Part I, the most common instruments for trading the risk factors are introduced. Trading is defined as the acceptance of risk or risk factors, through a financial instrument, in return for a certain yield from the instrument and/or specified payments from a contracting party. Conversely, a financial instrument can of course also be used to transfer risk to a contracting party in exchange for paying a specified yield and/or amount to the risk-taking party. In Part I, the functionality of several of the most important financial instruments will be introduced without entering into a discussion of their valuation, which is frequently very complicated.

Once the fundamental risks and the financial instruments used for trading them have been introduced, the building blocks for further analysis will have been defined. This will allow us to proceed with the actual topic of this book, namely determining the prices and the risks of these building blocks. For this, the same fundamental methods will be repeatedly applied to many different financial instruments. In order to make this clear, Part II introduces the most important methods for pricing and hedging in their full generality—independent of any specific instrument. As a result, this section is rather theoretical and technical. The concepts elaborated upon in Part II will then be applied to specific financial instruments in Part III. This separate treatment of general methods and specific financial instruments will contribute to a clearer understanding of this rather complex material.

Once the pricing of the most common financial instruments has been dealt with in Part III, the determination of the risks associated with these instruments will be presented in Part IV. The information about the prices and risks of financial instruments can then be used for decision making, specifically of course for trading decisions and the management of investment portfolios. This is demonstrated in Part V. Finally, methods for determining and analyzing the market data and historical time series of market risk factors will be the topic of Part VI.

© The Author(s) 2019

H.-P. Deutsch, M. W. BeinkerDerivatives and Internal ModelsFinance and Capital Markets Serieshttps://doi.org/10.1007/978-3-030-22899-6_2

2. Fundamental Risk Factors of Financial Markets

Hans-Peter Deutsch¹   and Mark W. Beinker²  

(1)

Niedernhausen, Germany

(2)

Friedrichsdorf, Germany

Hans-Peter Deutsch (Corresponding author)

Email: hans-peter.deutsch@d-fine.de

Mark W. Beinker

Email: mark.beinker@d-fine.de

The fundamental risk factors in financial markets are the market parameters which determine the price of the financial instruments being traded. They include foreign currency exchange rates and the price of commodities and stocks and, of course, interest rates. Fluctuations in these fundamental risks induce fluctuations in the prices of the financial instruments which they underlie. They constitute an inherent market risk in the financial instruments and are therefore referred to as risk factors. The risk factors of a financial instrument are the market parameters (interest rates, foreign currency exchange rates, commodity and stock prices), which, through their fluctuation, produce a change in the price of the financial instrument. The above mentioned risk factors do not exhaust the list of the possible risk factors associated with a financial instrument nor do all risk factors affect the price of each instrument; for example, the value of a 5 year coupon bond in Swiss Francs is not determined by the current market price of gold. The first step in risk management is thus to identify the relevant risk factors of a specified financial instrument.

2.1 Interest Rates

Various different conventions are used in the markets to calculate interest payments. For example, interest rates on securities sold in the US money markets (T-bills, T-bill futures) are computed using linear compounding, whereas in the European money market, simple compounding is used. Interest rates in the capital markets are calculated using discrete or annual compounding while option prices are determined using the continuous compounding convention. While these conventions are not essential for a principle understanding of financial instruments and risk management they are of central importance for the implementation of any pricing, trading or risk management system. Before entering into a general discussion of interest rates in Sect. 2.1.5, we will therefore introduce the most important compounding conventions here.

2.1.1 Day Count Conventions

Before one of the many compounding conventions are applied to calculate the interest on a certain amount over a period between the time (date) t and a later time T, the number of days between t and T over which interest is accrued must first be counted. The beginning, the end and the length of this time period (measured in years) must be precisely specified. To do this there are again different conventions used in different markets, known as day count conventions, DCC for short.

Therefore, a time difference T − t measured in years depends in general on the chosen day count convention. To make this explicit, we will often use the expression τ(t, T) (year fraction ) for the time period from time t to T measured in years with a given day count convention. These market usances are usually specified by making use of a forward slash notation: the method for counting the days of the month are specified in front of the slash, the number of days of the year after the slash. A list of the most commonly used conventions is presented in Table 2.1. These conventions compute the length of an interest rate period as follows:

Actual/365f: The actual number of calendar days between t and T are counted and divided by 365 to obtain the interest period in years regardless of whether the year concerned is a leap year. This distinguishes the Actual/365f convention from the Act/Act in years.

Actual/360: The actual number of calendar days between t and T are counted and divided by 360 to obtain the interest period in years.

30/360: The days are counted as if there were exactly 30 days in each month and exactly 360 days in every year. In addition, the following holds:

If the beginning of the interest period falls on the 31st of the month, the beginning is moved from the 31st to the 30th of the same month for the purpose of the calculation.

If the end of the interest period falls on the 31st of the month, it is moved forward for the purposes of the calculation to the 1st of the next month unless the beginning of the interest period falls on the 30th or 31st of a month. In this case, the end of the interest period is moved from the 31st to the 30th of the same month.

30E/360: The days are counted as if there were exactly 30 days in each month and exactly 360 days in each year. In addition, the following holds:

If the beginning of the interest period falls on the 31st of the month, the beginning is moved from the 31st to the 30th of the same month for the purpose of the calculation.

If the end of the interest period falls on the 31st of the month, the end of the period is moved to the 30th of the same month for the purpose of the calculation (this differentiates this day count convention from the 30/360 convention).

Act/Act: The actual number of calendar days are counted and divided by the actual number of days in the year, i.e. parts of the time period falling in leap years are divided by 366, otherwise the divisor is 365. Often, the same notion is used for other day count conventions as well, so you need to be cautious.

ICMA is a specific convention proposed by the International Capital Market Association, which is frequently used as standard convention for bond emissions. The number of days equals the actual number of days. The calculation of the divisor is based on the coupon payment frequency m (i.e. number of coupon payments per year) of the bond and the rolling day DR, which is the start date for the roll out of all coupon periods, which are subsequently adjusted due to business day rules (see Sect. 2.1.2). Next, the natural length L (i.e. without any further adjustments) of the considered time period is calculated. The divisor is set to the product mL of natural length L and coupon frequency m. This method guarantees that any coupon period, which start and end days fall together with the rolled out and according to business day rules adjusted dates, has exactly the length L∕12 in years. Only periods, which have been shortened or lengthened by some days may differ.

BD/252 counts only the business days, divided by the number of business days per year, which is fixed to 252.

Table 2.1

The commonly used day count conventions

These conventions for calculating the time between two dates can also be expressed in formulas:

(2.1)

The notation should be read as follows: D 1 denotes the start date of the interest period, D 2 the end date. D 1 consists of the number T 1 for the days, M 1 for the months and J 1 for the years; D 2 is defined analogously. m is the number of coupon payments per year and DR the rolling date, which is the start date for the roll out of all coupon start and end dates. The function Date delivers the running number for each given date. The counting of this running number begins at some time in the distant past, for example January 1, 1900. Then the date function (as defined in Microsoft Excel, for instance; please note that for compatibility reasons Microsoft Excel counts the (actually non-existing) date February 29th, 1900 by default also as a valid date) yields the value 35,728 for Date (1997, 10, 25). This is the running number corresponding to October 25, 1997. The function "feb(x)" is defined as equal to zero for x ≤ 0, and 1 otherwise. The function $$\min (x,y)$$ yields the smaller of the two values x and y, and the function $$\max (x,y)$$ yields the larger of the two. In general, the difference D 2 − D 1 is actually the number of days between D 1 and D 2, including D 1 but excluding D 2. The single exception is the last premium period of a Credit Default Swap ( CDS), which includes both, D 1 and D 2 and is thus one day longer than, e.g., the equivalent interest rate period. The function Adjusted(D) adjusts the date D according to a given business day rule (see Sect. 2.1.2) to fall on a business day.

The computation of time periods using these day count conventions is demonstrated in Table 2.2 and in the Excel-Sheet Usance.xls available in the download section [50] accompanying this book. The time period has been chosen to yield different results for different conventions. The period starts at February 15th, 2012 and ends on December 31st, 2012. The chosen business day calendar is TARGET2 (the official calendar for the EUR zone) with February 15th, 2012 as roll date and period length of one month. The time period has been intentionally selected to contain the 29th of February in order to demonstrate the difference between the Act/Act and Act/365f conventions. Furthermore, the interest period was chosen to end on the 31st of the month to illustrate the difference between the 30/360 and 30/E360 conventions.

Table 2.2

Determining the length of a time period using different day count conventions. The time period was chosen to yield a different length (in years) in each day count convention

2.1.2 Business Day Conventions

Establishing a day count convention is not sufficient to uniquely determine interest periods. The value dates of the cash flows must also be defined, i.e., the number of days following the end of an interest period T on which the interest payment must be settled. Several different conventions govern this calculation as well. Furthermore, there are conventions to account for weekends and holidays. If the value date falls on a bank holiday, for example, should payment be made on the day before or after the holiday? And finally, bank holidays themselves vary from country to country. The conventions governing these questions are called business day conventions, BDC for short. These conventions transfer the value dates of a cash flow away from weekends and bank holidays in accordance with the rules compiled in Table 2.3.

Table 2.3

Business day conventions

Table 2.4 shows as an example the adjustment of March 31st, 2012 and January 1st, 2012. These example dates fall on either a weekend or bank holiday (according to TARGET2 calendar). For the EUR zone, TARGET2 is the official bank holiday calendar. TARGET2 is the short form of Trans-European Automated Real-Time Gross Settlement Express Transfer System (2nd version), the settlement system for money transfers between banks operating in the EUR zone. TARGET2 replaced on May 19th, 2008 the preceding system (or collection of systems) TARGET. The TARGET2 holidays are January 1st, Good Friday, Easter Monday, May 1st, and December 25th and 26th. All other days, except for Saturdays and Sundays, are business days.

Table 2.4

Effects of business day conventions

For interest rate instruments, a further distinction is made between whether the adjustment convention holds solely for the payment date of an interest period or for its maturity date as well. The maturity date determines the length of the interest rate period and thus affects the amount of the interest payment (if the length of an interest period is longer, the amount of the interest payment to be made is naturally higher). The payment date determines when the interest payments are actually made (usually one or two business days after the maturity date) and therefore affects how strongly a payment is discounted, in other words, today’s value of the payment (a later payment is obviously worth less than an earlier one); it is thus relevant when the payment is actually made and not when it was due.

If the maturity date of a financial instrument is specified as fixed, i.e., non-moveable, it is not adjusted. However, the payment date is still adjusted in accordance with the business day convention for the instrument concerned. The rollover day of an interest rate instrument determines on which day and month of each year the rollover from one interest period into the next is to take place, i.e., when the maturity and payment dates of individual interest payments occur. Depending on the selected business day convention, a decision is made as to how the maturity and payment dates, derived from the rollover date, are to be adjusted. For example, federal bonds are commonly agreed to be fixed. This means that only the payment date is adjusted to the next valid business day while the maturity date is always the same as the rollover day. For swaps, on the other hand, both the payment and maturity dates are adjusted to the next valid business day subsequent to the rollover day.

All these conventions make trading substantially more complicated without causing a fundamental change in the properties of the instruments being traded. They are actually unnecessary but the markets are inconceivable without them because of strong historical ties. This is particularly true of holiday calendars, some of which even have religious roots.

2.1.3 Discount Factors

In order to concentrate on the essentials, a general notation will be observed in this book which holds for all compounding, day count, business day and other market conventions. To accomplish this, discount factors rather than interest rates will be employed throughout. The discount factor is the value by which a cash flow to be paid at a time T is multiplied in order to obtain the value of the cash flow at an earlier time t (for example, today). In general, an amount of money today is worth more than the same amount of money which is accessible not before a future date T, because of the missing choice to either spend the money, invest it in a profit-gearing asset, or just leave it alone until T. This option has a positive value, the so-called time value of money.¹ Thus, the discount factor is generally less than 1 (but greater than 0). A discount factor for discounting from a time T back to an earlier time t will be referred to using the notation

(2.2)

The letter B is used since a discount factor is identical to the price at time t of a zero bond with maturity T and nominal of one currency unit (i.e. 1 EUR).

The discount factor yields the value of a future payment today (discounting). Conversely, the future value of a payment today (compounding) is obtained by multiplying the payment by B −1(t, T). The interest accrued between times t and T is thus the difference between the compounded value and the original amount, i.e., the original amount multiplied by the factor (B −1(t, T) − 1).

2.1.4 Compounding Methods

Over time, various methods for calculating interest have been established. The explicit form of the discounting and compounding factors and of the interest accrued (based on a notional=1) are shown for the four most common compounding methods in Table 2.5. At each stage in this book, the results expressed in the general notation given in the first line of Table 2.5 can be converted directly into the explicit expressions of the desired compounding method by replacing the general expression with the appropriate entries in Table 2.5, given the fact that discount factors offer a convention-independent way to express these factors.

Table 2.5

Interest rate factors in general notation and their specific form for the three most commonly used compounding methods. Linear compounding is an exception, since it should not be seen as an additional compounding method rather than an approximation for short time periods of the simple compounding method and is only mentioned for sake of completeness. Calculation of interest requires always the specification of a compounding method, therefore, it is not possible to calculate accrued interest if only the discount factor is given

These discount factors are obtained from intuitive considerations which are now described in detail for each compounding method. Note that interest rates quotes are always relate to a specific time unit. Typically, interest rates are given as a percentage per year (i.e. per annum), quoting interest rates per month or even per day is the exception. The actual length of the time period over which interest is paid, and which has to be measured in terms of day count convention consistent with the quoted interest rate, is independent from the above mentioned time unit. For example, an interest rate quote of 6% is only complete if the related time unit (e.g. per annum, if nothing else is given) and day count convention (e.g. act/365) is also specified.

Simple Compounding

For simple compounding the interest paid at the end of the term agreed upon is calculated by simply applying the rule of three: if an interest rate R per time unit has been agreed upon and the interest period (T − t) spans n time units, the interest payable is simply the product n multiplied by R. If the period (T − t) is measured in the same unit (e.g., years) as that used to quote the interest rate (e.g., interest rate is quoted per annum) then we simply have n = (T − t) and the interest paid simply equals R(T − t) as indicated in Table 2.5. If capital in the amount K 0 has been invested, the interest earned resulting from simple compounding is obtained by multiplying K 0 by R(T − t). The capital held at maturity from an investment paying, for example, 6% per annum over a quarter of a year is thus

The compounding factor is thus 1 + R(T − t). The discount factor is the reciprocal of the compounding factor. Simple compounding is often used in money markets where interest periods (T − t) are usually less than one year or for fixed income instruments with periodic interest rate payments (without compounding interest rate over more than on period).

Discrete Compounding

In contrast to simple compounding, discrete compounding takes compounding effects into account. It is used for long term investments (longer than one interest period), if during the investment period no interest is actually paid out. Therefore, the length of the interest period or the frequency of assumed interest periods per year is required as an additional parameter. Typical interest periods are annual, semi-annual, quarterly, monthly, or daily, though again the interest rate quote refers to interest paid per annum, with the full investment period T − t given in fraction of years. In the following, m denotes the number of interest periods per year. Then, interest is calculated as if after each interest period the interest is immediately re-invested at the same interest rate. Or, to put it otherwise, the invested capital is increased by the interest due. The assumed interest amount for each period is calculated by means of simple compounding. An interest rate of 5% with m = 2 (semi-annual compounding) would thus yield after half a year (one interest period) an invested capital K 1 of

This capital is reinvested over the next time unit (the second year) in exactly the same manner. Thus, after one year the investor’s capital has increased to

In the second period, interest is accrued on the interest earned in the first period in addition to the initial investment capital. This is called compounded interest. Likewise, after the third period, the investor’s capital is given by

and so on. The compounding factor thus obtained is (1 + R∕m)mn where n is the length of the full investment period in years and m the number of interest periods per year, or, more general, the compounding factor is equal to (1 + R∕m)m(T−t) as in Table 2.5. Here, m(T − t) is not necessarily a whole number. The calculation remains the same for T − t = 3.5, for example.

Continuous Compounding

In the case of continuous compounding, the calculation is performed as if interest payments were made after each infinitesimal small time increment (each payment calculated using simple compounding) with the accumulated interest being immediately reinvested at the same rate. Therefore, it can be considered as the limit of the discrete compounding method for infinitesimal small interest periods. This method is frequently used for modeling derivatives, because of its convenient mathematical features, easing the life of financial mathematicians significantly. Since different compounding methods can be easily converted into each other, you have the free choice of using the most convenient method for doing your math. However, the interest rates for real interest coupons paid by fixed income instruments are never quoted as continuous compounding rates.

The total capital accumulated on an investment over a time period of T − t calculated by this method is then

(2.3)

The compounding factor is thus given by e R(T−t) as indicated in Table 2.5. Here, Euler’s number e, also called the natural number, arises. Its value is approximately

Euler’s number to the power of some number x is called the exponential function, which is defined by means of the limit in Eq. 2.3:

Linear Compounding

Linear compounding is justified for very short periods of time T − t. For such times, the product R(T − t) is also very small. For example, if R = 3% per annum and the time to maturity T − t is one month = 0.083 years, the product R(T − t) = 0.0025. The square of this product is considerably smaller, namely R(T − t)² = 0.00000625. Thus, in the case of linear compounding, only terms of order R(T − t) are of any importance, i.e., all non-linear terms are simply neglected. If we represent the discount factor, which is always the inverse of the compounding factor, as a geometric series² neglecting all terms of higher order, we obtain the discount factor given in Table 2.5:

It follows, in strict linear approximation, that the general rule that the reciprocal is identical to the inverse discount factor holds also for linear compounding. However, since this condition is only approximatively true, linear compounding is inferior to other compounding methods and should be used only as an approximation for simple compounding. Since the advent of computers, it’s practical importance has diminished. Therefore, we won’t use this method within the rest of the book.

Convention-Dependent Interest Rates

Today’s value of a future cash flow is determined by the discount factor for the proper time period. On the other hand, the discount factor could easily be calculated as today’s value of the cash flow divided by its amount (or its value at pay date). However, expressing discount factors in terms of interest rates requires the application of compounding methods and day count conventions, which is demonstrated by the very different formulas for discount factors in Table 2.5. After all, today’s value of a monetary unit paid in the future must be independent of the convention used for discounting.

Other than discount factors, interest rates depend on the methods and conventions used. The impact of conventions is fully absorbed in the interest rates. As a useful consequence, equalizing discount factors expressed in different compounding methods as shown in Table 2.5 enables the transformation of a given interest rate from one compounding convention into another:

For example, the interest rate necessary to generate a discount factor in discrete discounting (with m = 1, i.e. annual compounding) with the same numerical value as a given discount factor in continuous compounding is

If for discrete compounding the interest period is identical to the compounding period, the relation is very simple, i.e. if $$\frac {1}{m}=T-t$$ we have:

or

(2.4)

In this special case, the expression for the discount factor is identical to the case of simple compounding. This relation will be used frequently throughout the book.

However, not only the effects of the compounding methods but also the effects of the day count convention are absorbed in the interest rates. This is demonstrated in the Excel-sheet Usance.xlsx available from the download area and in Table 2.6. The interest period is the same as in Table 2.2, but the different day count conventions generate different time lengths (measured in years). The discount factor must be the same for all conventions. The interest rates associated with this single discount factor, however, are strongly influenced by both the day count convention as well as the compounding convention. They vary between 1.955% and 2.027%.

Table 2.6

Interest rates for the same discount factor based on different day count and compounding conventions. The interest period is the same as in Table 2.2. The value of the discount factor for this period is 0, 98261858

As already mentioned, all of the conventions introduced here are actually unnecessary for understanding financial instruments. In order to concentrate on the essentials, discount factors rather than interest rates will be predominantly used in this book. For any concrete calculation, the reader should be able

to write down explicitly the required discounting factor using Table 2.5 and

to calculate the time length T − t using Eq. 2.1 (along with the Excel-sheet Usance.xlsx)

after the exact dates t and T in the appropriate business day convention as specified in Table 2.3 have been determined.

In what follows, we will therefore work only with the general discount factor given in Eq. 2.2.

2.1.5 Spot Rates

Spot rates are the current yields on securities which generate only one single payment (cash flow) upon maturity. A zero coupon bond is an example of such a security as are coupon bonds whose last coupon payment prior to maturity has already been made. The spot rates as a function of time to maturity T is called the spot rate curve or the term structure. These spot rate curves can be represented by the discount factors BR(t, T).

2.1.6 Forward Rates

Forward rates are the future spot rates from today’s point of view which are consistent with the current spot rates. The pay-off profile of many interest rate products is based on a forward interest rate F(t, t fix, T, T ′), which is fixed at a future time t fix and which relates to the future interest period from T to T ′ with t fix ≤ T < T ′. Of course, from today’s point of view with t < t fix, the actual value of this interest rate is yet unknown.

The forward rate as of today can be derived from the following arbitrage argument: One monetary unit is to be invested today (time t) until a specified maturity date T ′. The investor can consider the following two investment strategies:

1.

Invest the monetary unit without adjusting the position until the maturity date T′, or

2.

Invest the monetary unit until some time T where T < T′, and at time T immediately reinvest the interest earned together with the original amount until the maturity date T′.

If the investor were able to fix today the interest rate for the future period between T and T ′ (see Fig. 2.1), then the total return for both strategies must be the same. If this were not the case, the investor would have an opportunity to earn a profit without risk, i.e., an arbitrage opportunity (see Sect. 6.​1). If the yield from the first strategy were higher, an investor could raise capital according to the second strategy and invest this capital according to the first. Financing the investment using the second strategy requires less interest than the total return on investing in the first. Thus, since all interest rates are fixed at time t, a profit would have been made without risk or investment capital at time t. Conversely, if the total return of the first strategy is lower than that of the second, the investor could finance the second investment strategy according to the terms of the first. Again, this gives the investor an arbitrage opportunity.

Fig. 2.1

The sequence of the times t, T and T ′

The interest rate fixed at time t for the future time period between T and T ′ which eliminates any possible arbitrage opportunity, i.e., which is consistent with the spot rates at time t, is the forward rate R(T, T ′|t). The notation for the analogous forward discount factor, i.e. the price of a zero bond at the future time T which pays one monetary unit at T ′ > T is

(2.5)

The condition eliminating the arbitrage opportunity can be summarized as follows: the compounding factor using the spot rate from t to T ′ must be equal to the compounding factor of the spot rate from t to T multiplied by the compounding factor of the forward rate from T to T ′. The compounding factors are the inverse of the corresponding discount factors as indicated in Table 2.5. Consequently, the following equality must hold to prevent arbitrage:

or equivalently,

(2.6)

The forward discount factors can thus be uniquely determined from today’s discount factors:

(2.7)

The last two equations are quite fundamental. They will be used repeatedly in what follows to simplify formulas involving terms with products and quotients of discount factors. With the aid of these equations, many essential properties of financial instruments depending on interest rates can be derived without ever having to specify the compounding convention. This helps to clarify questions as to whether certain properties under consideration are intrinsic properties of the instruments themselves or merely results of the compounding convention being used. For example, the difference between Macaulay Duration and Modified Duration (see Chap. 5) is not an inherent property of bonds but merely an effect resulting from applying a particular compounding convention. In the case of continuous compounding, for example, there is no difference between the two!

At this point we provide an example demonstrating the usefulness of Eq. 2.6 and Table 2.5. We give explicitly the forward rates as a function of the spot rates for each of the three common compounding co nventions:

This clearly demonstrates the advantage of using the general notation for discount factors 2.7 introduced above. In Fig. 2.2, the forward rates are calculated from the spot rates taken from the Excel sheet PlainVanilla.xls (see examples on accompanying website [50]) using the above formulas for annual, discrete compounding over a period of 15 years.

Fig. 2.2

Forward rates for periods starting in T = 1, 2, …, 15 years for terms T ′− T = 1, 2, …, 15 years. From line to line the start points T of the forward periods change by one year. From column to column the lengths T ′− T of the forward periods change by one year. The inset graphic shows, from top to bottom, the current term structure along with the forward term structures in 1, 5 and 10 years

Since coupon payments are generally calculated by means of the simple compounding methods, we demonstrate here explicitly how F(t, t fix, T, T ′) depends on the (forward) discount factor:

(2.8)

The forward rate calculated in this way is the risk neutral or risk free ³ forward rate, which can be calculated by means of risk neutral discount factors B(t, T) only. For actually traded forward rates, an additional tenor basis has to be taken into account, which is an add-on of a few basis points to the risk neutral forward rate. A major cause for the tenor basis is that unsecured interest payments bear credit default risk which is the higher the longer the term to payment date is. The tenor basis could be specified as an explicit add-on (the tenor basis spread) or by replacing B(t, T) by a forward yield curve or tenor curve (see Sect. 15.​2 also).

2.2 Market Prices

Let S(t) be the spot price at time t of a stock, a commodity, or a currency. The dividend-adjusted spot price $$\widetilde {S}(t,T)$$ at time t is the price net of the value of all dividends paid between the times t and T. It is given by

(2.9)

The dividend adjustment is thus accomplished by subtracting the value of dividends, discounted back to time t at the spot rate R, from the spot price, or—in the case of a dividend yield q—discounting the spot price from T back to t using the dividend yield q. In the following, the notation Bq(t, T) = exp[−qτ(t, T)] might be used as an alternative way to express the dividend yield, making use of the formal similarity to discount factors. For example, currencies are mathematically equivalent to stocks with dividend yields, where q represents the risk-free interest rate (in continuous compounding) in the foreign currency. Similar, for commodities q is replaced by the difference between the convenience yield and the cost of carry (expressed as a yield, see later). Relative or proportional discrete dividends are proportional to the underlying price at the dividend payment date. The dividend cash flow therefore depends on the forward price S(t, t ′).

When considering stock indices, the dividend payments from the assets in the index are commonly averaged out to result in a dividend yield q of the index rather than considering each dividend of each stock as an individual payment. However, performance indices, such as the German DAX, which require that dividends be reinvested in the index should not be adjusted for dividend payments.⁴

2.3 An Intuitive Model for Financial Risk Factors

2.3.1 Random Walks as the Basis for Pricing and Risk Models

The random walk is a mathematical model which is frequently used to characterize the random nature of real processes. The concept of the random walk has attained enormous importance in the modern financial world. Most option pricing models (such as the Black-Scholes model) and several methods used in modern risk management (for example, the variance-covariance method) and, of course, Monte Carlo simulations are based on the assumption that market prices are in part driven by a random element which can be represented by a random walk. It is therefore worthwhile to acquire an understanding of random walks if only to develop an intuitive comprehension of stochastic processes which are at the heart of this book.

A random walk can be described as follows: starting from some point in space, we travel a random distance in a randomly selected direction. Having arrived at a new point, another such step of random length and direction is taken. Each individual step in the procedure has a length and direction and thus can be represented as a vector as shown in Fig. 2.3. The completed random walk is a series of such vectors. Each base point of a vector is the end point of its predecessor.

Fig. 2.3

A random walk with 8 steps in two dimensions

At this point, we ask the following important question: what is the distance from the starting point after having completed a random walk consisting of n steps?⁵In other words: how large is the end-to-end distance represented by the length of the vector R in Fig. 2.3? The length and direction of the vector R are random since R is the sum of random steps. As a result, only statistical statements are available to describe the properties of this vector. For example, the length of this vector cannot be determined with certainty but we could calculate its mean length. This requires a large number of random walks with the same number of steps.

For each of these random walks, the square of the Euclidean norm of the end-to-end vector R is determined and used to calculate the mean $$\left \langle \mathbf {R} ^{2}\right \rangle $$ . The mean end-to-end distance is than defined as the square root of this value. A Monte Carlo simulation (see Chap. 11) could be carried out to generate the random walks and obtain an estimate for the mean $$\left \langle {\mathbf {R}}^{2}\right \rangle $$ by measuring R ², the square of the end-to-end vector, of each simulated random walk and then take the average of these. In doing so, it can be observed that the square of the end-to-end vector is, on average, proportional to the number of steps taken in the random walk.⁶

Here E[x] denotes the expectation of a random variable x and 〈x〉 denotes its mean value.⁷ This holds irrespective of the dimension of the space in which the random walk occurs. The same result holds for a random walk on a line, in a plane as in Fig. 2.3 or in a 15-dimensional Euclidean space [52].

The expectation of the end-to-end vector itself is equal to zero, i.e., E[R] = 0. This is immediate since the end-to-end vector R points in any direction with equal likelihood and therefore all these vectors cancel each other out in the average. In consequence, the variance $$\operatorname {Var}[x]$$ of the end-to-end vector is given by:

The variance of the end-to-end vector is thus also proportional to n. The standard deviation, defined as the square root of the variance, is consequently proportional to $$\sqrt {n}$$ . This is the reason why the uncertainty in future market prices increases proportionally to the square root of time since, as will be shown in Sect. 2.3.2, the time corresponds to the number of steps when a random walk is used to model and interpret price movements. Many well-known rules in financial mathematics, for example, that the overnight value at risk is to be multiplied by a factor consisting of the square root of the liquidation period or that the conversion of daily volatilities into monthly volatilities is accomplished by multiplying by the square root of the number of days in the month, have their origin in the fact that the variance of the end-to-end vector of a random walk is proportional to the number of random steps!

The relation E[R ²] ∝ n is so fundamental, that it is worthwhile to understand its theoretical meaning. As mentioned above, a random walk consists of a series of ‘step’ vectors. According to the rules of vector addition, the end-to-end vector is merely the sum of these vectors:

. Since each individual step vector, r i points in any direction with the same likelihood, the expectation of each step is E[r i] = 0 just as for the end-to-end vector. Since each of the steps are independent of one another and therefore uncorrelated, we have

With this information about the individual steps, we immediately obtain

where the constant b denotes the mean length of a single step:

and as such is the constant of proportionality in the relation E[R ²] ∝ n. At no point did the dimension enter into the above derivation. The equation E[R ²] = b ²n thus holds in any dimension and is an expression of a fundamental property of random walks, namely their self-similarity: the statistical properties of a random walk are always the same, regardless of the degree of detail with which they are observed. In other words, a step in a random walk can itself be represented as the end-to-end vector of a random walk with smaller steps. Likewise, the end-to-end vector of a random walk can itself be considered a single step of a coarser random walk (coarse graining).

As yet, only the so called moments (expectation, variance, etc.) of the probability distribution p(R) have received mention. However, the distribution of R itself can be determined as well. In general, the end-to-end vector of a random walk in a d-dimensional space has a normal distribution. The concrete expression of this fact in one dimension (the most important case for the financial applications) is given by [52]

(2.10)

where

(2.11)

Equations 2.10 and 2.11, together with the principle of self-similarity are the quintessential properties of the theory of random walks introduced here. They comprise all that is necessary for their application in the field of finance. For example, the "normal distribution assumption" for price movements is an immediate consequence of the theory of random walks. Contrary to popular belief, the normality of the distribution of relative price changes need not be assumed. It follows automatically that if market prices (or more precisely, the logarithm of the market prices, see below) behave as random walks, they must be normally distributed with a density function as given in 2.10. In financial literature, random walks are often defined having normally distributed individual step vectors. By doing so the normal distribution assumption is injected into the definition of the random walk. This assumption is equally unnecessary! As mentioned above, a random walk is defined as a series of completely random steps. Absolutely no assumptions have been made on the distribution of the individual steps or their sum. The fact that the end-to-end vector has a normal distribution follows automatically as a consequence of the central limit theorem.

2.3.2 Risk Factors as Random Walks

A random variable z(t), whose values change randomly with time t is called a stochastic process. The process is called Markov if its future behavior is influenced solely by its value at the present time t. This means intuitively that the future behavior is independent of the path taken to reach the present value. Assuming that the current value of a risk factor, such as a stock price or an interest rate, contains all the information about its historical development (this is called weak market efficiency), it follows that the subsequent values taken on by such a risk factor depend only on the current price and other external effects, such as politics, but not on the past prices or rates. Market prices can then assumed to be Markov processes.

In order to derive a model for the Markov process S(t) representing the time-evolution of a risk factor, we assume that the process can be split into a random and a deterministic component. The deterministic component is called the drift. We will begin our discussion with an analysis of the random component.

The derivation of the model describing the random component given here is fundamentally different from that which is commonly found in the related literature and is based on the general properties of the random walk, Eq. 2.10 and Eq. 2.11. The literature usually begins with the introduction of a model and proceeds with a presentation of calculations and results following from the model assumption. It is more intuitive, however, to begin with a derivation of the model itself, i.e., to illustrate the steps and describe the motivation leading to its construction. Thus, this section serves the dual purpose of introducing a model for future analysis and an example of the modeling process. What follows is a detailed discussion of how we can construct a model describing a real phenomenon by means of abstraction (and intuition!) which is simple enough to allow a mathematical and/or computational analysis but still complex enough to capture the essential features of the real phenomenon.

A real process, for example the closing prices of a stock on 500 days, might look like the points shown in Fig. 2.4. We propose to model this process. To do so, several questions must first be answered:

Fig. 2.4

End-of-day values of a stock price over a period of 500 trading days. The values after 0, 100, 200, …, 500 days are denoted by S 0, S 1, S 2, …, S 5

What is the fundamental idea behind the model? As mentioned above, the random properties of many processes can be traced back to the general concept of the random walk. We therefore use the random walk as the basis for our model.

A random walk in which dimension? A market price can rise or fall, i.e., can change in only two directions. A random walk in d-dimensional space allows for an upward or downward change in d linearly independent directions, i.e., there are 2d possible changes in direction. Thus, the dimension required for the description of just upward or downward changes is d = 1.

Which real parameter is described by the number of steps n in the random walk? In order to observe a change in price (in other words for a step in a random walk to be taken), one thing must occur: time must pass. If the price is observed in regular, fixed time intervals dt (for example, every 100 days as in Fig. 2.4, or daily or hourly, etc.), then the amount of time passing between steps is dt. If the entire random walk occurs between t (=today) and a future date T then

(2.12)

Since dt is a constant, the number of steps n is proportional to the time in which the random walk occurs, i.e., proportional to T − t.

Which real parameter should be modeled by a random walk? At first glance, we might take the market price of a risk factor. The market price evolution S5 − S0 over the entire period in Fig. 2.4 can be decomposed into individual steps as follows:

or more generally,

If the market price itself were a random walk, then as a result of the self similarity property, the individual steps dSi would also be random walks. The price differences however are real cash amounts given in euros, for example. This would mean that a security costing 1000 euros would experience the same fluctuations (in euros) as one costing only 10 euros. This is surely not the case. It would make much more sense to consider relative fluctuations. Our next candidate for a step in our random walk could therefore be the ratio Si∕Si−1. The ratio of the last price to the first is given by

or more generally,

This is the product of the individual steps and not their sum. A random walk however, is a vector and as such must always be the sum of its component steps. In light of this fact, the ratios Si∕Si−1are completely unsuitable for the steps of a random walk as they are not even vectors! However, the ratios Si∕Si−1, which make economic sense, can be utilized by converting the products into sums by taking the logarithm of both sides of the above equation. The functionalequation for the logarithm is given by

Thus taking the logarithm of both sides of the above product yields

or more generally,

and thus,

This looks exactly like our first attempt with the sole exception that the market price S has been replaced with its logarithm $$\ln (S)$$ . This small change makes it possible to satisfy both the economic requirement that the proportional changes in the market price be modeled and the mathematical requirement that a random walk be the sum of its component steps.

We have thus completed the fundamental construction of our model: The random component of a market price is modeled by interpreting the logarithm of the price as a one-dimensional random walk with independently and identically distributed (iid) random steps and the number of these steps being proportional to the length of time during which the random walk takes place.

Now we are in a position to apply what we know about random walks to draw conclusions about the evolution of market prices. It follows from Eq. 2.10 that the end-to-end vector

is normally distributed with expectation and variance