1

The Driving Force of Swap Spreads

--An empirical analysis of the U.S. dollar interest rate swap spreads

Masters of Science in Business Administration

Bing Liang
Oct 26, 2007

Title: The Driving Force of Swap Spread Author: Bing Liang Supervisor: Anders Hederstierna Department: Business Administration, Blekinge nstitute of Technolog! Course: "aster#s thesis in $usiness administration, %& credits' Background and Problem Discussion: The determinants of interest rate swap have $een a pu((le for man! researchers in the past !ears' n this paper, use the )'S' interest rate swap market as an e*ample' Purpose: The purpose of this thesis is to create a model to $e a$le to anal!(e the determinants of interest rate swap' Method: +ointegration model and ,rror +orrection model' Theory: Based on the ,conomic theor!, have looked at different concepts that come up when anal!(ing how the determinants of interest rate swap' Therefore have chosen the most $asic theor! of determining the interest rate swap' Analysis: have made up m! own model to determinate the driving force of found out that the interest rate swap can#t see ver! strong interest rate swap' Conclusion: Based on m! test results, spread is positivel! correlated with the default premium in corporate $ond market' Due to the limited si(e of availa$le data set, correlation' -verall it is ver! hard for me to draw an! strong conclusions' However, there are two relativel! two strong correlations. the interest rate swap spread is correlated with default premium in corporate $ond market and the government $udget deficit inde*' will

Ta$le of +ontents
Abstract.......................................................................................................................... 1. !ntrod"ction................................................................................................................6 1.1 #ac$%ro"nd..........................................................................................................6 1.2 Ob&ecti'e of this (aper.........................................................................................6 1.) *isposition of this paper......................................................................................7 2. +iterat"re ,e'iew.......................................................................................................2. 1 #asics of interest rate swaps................................................................................2.2 !nterest rate swap spread bac$%ro"nd .................................................................2.). +iterat"re ,es"lts..............................................................................................10 ).*ata ..........................................................................................................................1) ).1 So"rce *ata .......................................................................................................1) ).2 !nterest rate swap spread....................................................................................16 ).) ./planatory 'ariables.........................................................................................17 0. 1ethodolo%y............................................................................................................10.1. 2ypotheses3.......................................................................................................10.2 ./planation of 4ariables....................................................................................21 0.). 5he 6A,72 model...........................................................................................21 0.0 5he 7ointe%ration 1odel..................................................................................2) 0.0.1 1ethodolo%y of 7ointe%ration....................................................................20 0.0.2 7ointe%ration analysis.................................................................................20 0.) 6A,72 1odel with three different 'ersions of .rror 7orrection 5erms.........2 0.).1..rror 7orrection 1odel 1 8.7119..............................................................2 0.).2 .rror 7orrection 1odel 2 8.7129..............................................................2 0.).) .rror 7orrection 1odel ) 8.71)9..............................................................26 . .mpirical res"lts......................................................................................................2: .1 Unit ,oot 5est ,es"lt.........................................................................................2: .2 7orrelation amon% e/planatory 'ariables..........................................................2.) Analysis of 2ypothesis 5ests.............................................................................)1

6. 7oncl"sion ...............................................................................................................) 7. Ac$nowled%ements..................................................................................................)6 ,eference.....................................................................................................................)7

Abstract
5he interest rate swap is one of the most pop"lar topics that researchers wor$ on since 1-:0s. .'en tho"%h there are so many research papers that are abo"t the determinant factors of interest rate swap, it shows the limited e/planation. ! ha'e cond"cted the cointe%ration test on each pair of 'ariables based on the financial and macroeconomic theory. 1y testin% res"lts show that the interest rate swap spread is correlated with defa"lt premi"m in corporate bond mar$et and the %o'ernment b"d%et deficit inde/ while other ass"med determinant 'ariables ha'e wea$ impacts on the swap spread. .

1 !ntroduction
1.1 Background
An interest rate swap is a contract"al a%reement between two parties to e/chan%e a series of interest rate payments witho"t e/chan%in% the "nderlyin% debt based on an a%reed "pon notional principal, mat"rity and predetermined fi/ed rate of interest or floatin% money mar$et inde/. A lar%e n"mber of empirical researches show that interest rate swaps are one of the ma&or financial inno'ations since 1-:0s and the most pop"lar deri'ati'e contracts "sed by U.S. firms 8!n, #rown and ;an%, 200)9. 5he interest rate swap mar$et is one of the most important fi/ed-income mar$ets in the tradin% and hed%in% of interest ris$. !n 1---, the notional o"tstandin% 'ol"me of transactions of swaps was reached to amo"nt to o'er US< 00.7- trillion 8;an% = 1"li&ono, 20019. Accordin% to the se'eral s"r'eys of income mar$ets in the tradin% and hed%in% res"lts, the ann"al amo"nt of new b"siness in interest rate swaps has increased from US <):: billion in 1-:7 to o'er US< 17 trillion in 1--7, an increase o'er 0200> 8#ondnar, 2ayt, 1arston, and Smithson, 1-- 9. 5here is more than US< 1 trillion amo"nt o"tstandin% in interest rate swaps by ?"ne 2001 based on the #an$ for !nternational Settlements 820029. 1any academics ha'e done a lot of research to try to "nderstand the theoretical and empirical analysis of swap pricin% models. *"ffie and 2"an% 81--79 ar%"es that swap spreads as a ris$ premi"m to compensate swap co"nterparties for 'ario"s ris$s bein% "nderta$en s"ch as interest rate ris$s, defa"lt ris$s and li@"idity ris$s. At the same time, +an%, +itAenber%er and and +i" ar%"e for the swap as a non-red"ndant sec"rity creates s"rpl"s and swap co"nterparties share this s"rpl"s to compensate their ris$ which affect swap spreads. 2owe'er, most of the literat"re on interest rate swap spread is only foc"sed on identifyin% determinant ris$ factors that affect the mo'ements of the swap spread since people still try to fi%"re o"t why the interest rate swaps spreads fl"ct"ates so m"ch.

1.2 Objective of this Paper

1y paper is to ree/amine the dri'in% force of swap spread chan%es in the U.S. interest

rate mar$et. ! try to find o"t the dri'in% force of interest rate swap in the different mat"rity by applyin% the econometric techni@"es s"ch as cointe%ration test, "nit root test, and 6A,72 model with different error correction models. 5here are three important inno'ations in this paper. ;irst, ! added two additional important determinant factors that are b"d%et deficit and b"siness cycle in the re%ression of determinant of the swap spread beca"se ! belie'e that b"d%et deficit and b"siness cycle are the determinant factors in the re%ression model. Second, ! belie'e, based on the economic theory, that there mi%ht ha'e stron% cointe%ration between b"d%et deficit and b"siness cycle. .'en tho"%h the test res"lts show that cointe%ratin% relationship between the dependent 'ariable and independent 'ariables across mat"rities is not e/actly one-to-one as ! e/pected, it still shows a close relationship between b"d%et deficit and b"siness cycle. 5hird, ! add three error correction models 8.719 in the 6A,72 model to analysis the short-r"n dynamic relationships between those 'ariables in order to catch the tr"e nat"re of interest rate swap.

1.3 Disposition of this paper

;irst, ! ha'e empirical e'idence pro'ided by S"n et al 81--)9 and *"ffie and Sin%leton 81--79 that the importance of credit ris$s in pricin% interest rate swap contract. !n the *"ffie and Sin%leton 81--79 paper, they disc"ss the spread in the terms of defa"lt ris$ in the swap mar$et by "sin% the empirical findin% to de'elop a term str"ct"re model of swap yields. Second, researchers ha'e shown that swap spreads beha'e differently from corporate bond spreads. S"n et al 81--)9 find that swap spread are not as cyclical as A-rated corporate spreads which means AAA bid rates are si%nificantly lower that the A bid rates. On the other hand, !n, #rown and ;an% 8200)9, in'esti%ates the relationship between the chan%e of swap spread and chan%e in the %eneral le'el of interest, +i@"idity 'olatility defa"lt premi"m slope of yield c"r'e, and 5reas"ry 'olatility. 5hird, besides the literat"re on the determinants of the interest rate swap spread and the term str"ct"re of swap yields, this paper is ree/amine empirically the determinants of swap spread chan%es in the U.S. interest rate mar$et. 5he a"thor tries to b"ild a deeper

"nderstandin% of the dynamic e'ol"tion of swap spread. to mat"rity.

5he re%ressions of the

determinants of the swap spread are r"n indi'id"ally for swap spreads of different terms ;o"rth, ! will disc"ss the determinants of swap prices and spreads with my data description. ;ifth, ! wo"ld show my empirical methodolo%y to test swap spread chan%e. Si/th, ! will present the main empirical res"lts and disc"ssions. +astly ! will draw concl"sion based on my research res"lts and %i'e f"rther research s"%%estions at the same time.

" #iterature $evie%

2. 1 Basics of interest rate swaps
A Bplain 'anillaC interest rate swap is a company a%rees to pay cash flows e@"al to interest at a predetermined fi/ed rate on a notional principal for a n"mber of years 8?. 2"ll, 200 9. !n other words, one party promised to pay a fi/ed rate of interest 8swap rate9 while the other party promises to pay a fi/ed rate of interest rate at each periodic inter'al in a simple fi/edDfloatin% rate swap. An interest rate swap can be "sed to transform a floatin%-rate loan into a fi/ed-rate loan or 'ice 'ersa. 5he swap rate is "s"ally determined by the mar$et forces and is in effect the price of the swap. 5he ma&or empirical research abo"t the interest swap spread is what determines interest rate swap spreads since these spreads ha'e 'aried from a low ro"%hly 2 basis points to more than 1 0 basis points sometimes mo'in% 'iolently. 5here are a lot of early research papers showin% that the interest rate swaps lowers financin% costs by pro'in% the opport"nity to arbitra%e mispricin% of credit ris$. 2owe'er, #ric$sler pointed o"t that the inception of interest rate swap was coincided with a period of tremendo"s 'olatility in U.S. mar$et interest rates which res"lt in the rapid %rowth of interest rate deri'ati'es on the part of firms to hed%e cash flow a%ainst the impact of interest rate 'olatility 8#ic$sler, 20009

2.2 Interest rate swap spread background

1ost literat"re defined the swap spread as the ma&or pricin% 'ariable that is the difference between the interest rate swap rate and the par 'al"e of the 5reas"ry bond rate of same constant mat"rity. 2owe'er, it is not always this case based on my obser'ation in the interest rate swap mar$et. 5he main ar%"ments of e/istin% theoretical and empirical wor$ on the determinant factors on swap spread are "s"ally boiled down to the defa"lt probability of the co"nterparties, %eneral le'el of interest rate, s"pplyDdemand shoc$s of the swap-specific-mar$et, and 'olatility of interest rate as well as 5reas"ry bill-+!#O, spread and the corporate bond @"ality spread. 5he most pop"lar e/planation of the defa"lt ris$ is the differential between corporate bond spread and the slope of the yield. As to the slope of the yield c"r'e which is "sed as the predicted f"t"re interest rate

10

presents the ne%ati'e relationship with the swap spread on the condition of the economic de'elopment sit"ation.to the ban$ lendin%, which %i'es the commercial ban$s the ad'anta%e to dominate the interest rate swap mar$et 8Smith, et al, 1-:69.

2.3. Literature esu!ts

1ost e/istin% literat"re on interest rate swap spread is based on identifyin% determinant ris$ factors in the swap spread. 5here are two main contrib"tions from the pre'io"s literat"re which are 5reas"ry-swap spread in 5reas"ry mar$et and the spread in terms of defa"lt ris$ in the swap mar$et. *"ffie and Sin%leton 81---, 1--79 de'eloped a m"lti-factor econometric model of the term str"ct"re of interest rate swap yields, which demonstrates the impact of co"nterparty defa"lt ris$ and li@"idity differences between Swap and 5reas"ry Sec"rities mar$ets on the spread. At the same time the most representati'e framewor$ for st"dyin% the determinants dri'in% interest rate swap spreads is pro'ided by S"n et al. 81--)9 obser'ed the effect of co"nterpartiesE different credit ratin% 8AAA and A9 on swap rate bid Foffer spread. Altho"%h the AAA bid rates are si%nificantly lower than the A bid rates, the AAA offer rates are si%nificantly hi%her than the A offer rates. ;"rthermore, Sorensen and #olllier 81--09 ar%"ed that the price of the defa"lt ris$ depends on the 'al"e of two options, which in t"rn relies on the slope of yield c"r'e and the 'olatility of the short term interest rate. On the contrary, Smith et al. 81-::9 e/amined the interest rate swap spread "nder the ass"mption of no defa"lt, no li@"idity ris$, and the fi/ed rate of interest rate swap presented as the yield on a co"pon-payin% %o'ernment bond. Also 6rinblatt 81-- 9 presented a framewor$ to analyse the spreads "nder the ass"mption that simple interest rate swaps are defa"lt free. 2e ar%"ed that the li@"idity difference between %o'ernment bonds and short-term ."rodollar borrowin% is the reason for the spread between disco"nt rates "sed to 'al"e swaps and %o'ernment bonds. 5he hi%h li@"idity of %o'ernment bonds res"lts in a li@"idity premi"m, which is lost to an in'estor who recei'es fi/ed payment in a swap a%reement. 5herefore, the swap spreads are determined by the present 'al"e of c"rrent and f"t"re li@"idity premi"m in his framewor$. !n addition to that, there are some other models tryin% to relate the corporate yields to the swap rates, which are called

11

+!#O, swap spread. #rown et al. 81--09 stated that interest rate swap spreads as f"nctions of pro/ies for e/pected f"t"re le'els of +!#O, o'er 5reas"ry sec"rities spreads and different meas"res of credit ris$ and hed%in% costs of the swap co"nterparties. All these 'ariables are rele'ant to the swap spreads, b"t their relati'e importance fl"ct"ates with the mat"rities of swaps. *"fresne and Solni$ 820019 de'eloped a model in which the defa"lt ris$ is enclosed in the swap term str"ct"re that is s"fficient to e/plain the +!#O,swap spread. Altho"%h the corporate bonds carry ris$, they ar%"ed that the swap contracts are free of ris$, since those contracts are inde/ed on credit-@"ality +!#O, rate. Accordin%ly, the swap spread between corporate yields and swap rates sho"ld e/press the mar$etEs e/pectations on credit @"ality of corporate bond iss"ers. +ater on there are recent researchers who e/tend the in'esti%ation of the 'ariation of interest rate swap spread by incl"din% some other possible determinant factors for e/ample e/pected +!#O, spreads and swap mar$et str"ct"re. #rown, 2arlow, and Smith 81--09 and Gielsen and ,onn 81--69 introd"ced the +!#O, spread component. +!#O, rates are "s"ally hi%her than the 5reas"ry rates of the same mat"rity, as a res"lt that the difference is often referred to as the +!#O, spread. !n this conte/t, the swap sellers e/pect to pay a rate which is hi%her than the 'ariable rate by the amo"nt of +!#O, spread, th"s the need to be compensated by a hi%her fi/ed rate leads to a positi'e swap spread. 5he demandDs"pply shoc$s in the swap mar$et will infl"ence the swap spread d"e to the nat"re of swap mar$et str"ct"re on swap spread. S"ch shoc$ is deri'ed from the ori%inal moti'ation of the swap mar$et participants who try to arbitra%e the debt mar$et imperfection. Hall 81-:-9 and 5itman 81--29 show that some borrowers prefer the to pay the fi/ed rate of swap when facin% the mar$et imperfection, s"ch as the potential distress costs and asymmetrical information. Accordin%ly, the partly decreased or eliminated debt mar$et imperfection creates the swap spread s"rpl"s for the swap contract co"nterparties. O'erall the determinant factors ca"sin% the 'ariation of the interest rate swap spread is still 'ery complicated and p"AAled for researchers. #esides the indi'id"al determinant 'ariable, the mat"rity of the swap contract can also acco"nt for the chan%es in the swap spread. S"n et al. 81--)9 fo"nd that the swap spreads %enerally increase with mat"rity in

12

his empirical st"dy of the swap rate. At the same time, 1inton 81--79 e/tended his st"dy and drew the concl"sion that interest rate swap spreads are not only related to the le'el and slope of the term str"ct"re, credit ris$ and li@"idity premi"m, b"t also to the mat"rities.

1)

& Data
&.1 "ource Data
;irst of all, ! obtained the monthly data of interest rate swap rates and constant mat"rity 5reas"ry rates of mat"rity of 2, , 7 and 10 years from *atastream and .cowin in order to e/amine the beha'io"r of determinants of interest rate swap spread empirically. At the same time, ! obtained -0-day 5reas"ry rates, do"ble A and triple A corporate bond rates from .cowin. *"e to the data limitations, there are only a total of 106 obser'ations from the sample period startin% from ?"ne )0, 1--: to 1arch )1, 2007. Secondly, ! determine the interest rate swap spread by ta$in% the difference between interest rate swap rate and the constant mat"rity 5reas"ry rates of the same mat"rity. !n ;i%"re 1 and ), ! plot the mo'ement of interest rate swap rates and interest rate swap spreads for all mat"rities. 5he slope of the yield c"r'e is calc"lated as the difference between 10-year 5reas"ry rate and -0-day constant mat"rity 5reas"ry rate. ! "se the "nemployment rate as pro/y for the 'ariation of b"siness cycle. 5he difference between do"ble A and triple A corporate bond rates "sed as the pro/y of defa"lt premi"m. ;inally, the implied 'olatility of .@"ity mar$et and 5reas"ry mar$et are obtained by calc"latin% the daily obser'ations of S=( 00.

10

;i%. 1. 5he mo'ements of swap rates, @"arterly data.

6 5 4 3 2 1 0 / % & 7un8 65 Dec8 65 7un8 66 Dec8 66 7un8 && Dec8 && 7un8 &% Dec8 &% 7un8 &/ Dec8 &/ 7un8 &0 Dec8 &0 7un8 &1 Dec8 &1 7un8 &2 Dec8 &2 7un8 &3 Dec8 &3

)SS9A:/ +urnc!

)SS9A:2 +urnc!

)SS9A:4 +urnc!

)SS9A:%& +urnc!

;i%. 2. 5he mo'ement of 5reas"ry ,ates

5 4 3 2 1 0 / % & 7un8 Dec8 7un8 Dec8 7un8 Dec8 7un8 Dec8 7un8 Dec8 7un8 Dec8 7un8 Dec8 7un8 Dec8 7un8 Dec8 65 65 66 66 && && &% &% &/ &/ &0 &0 &1 &1 &2 &2 &3 &3
H%2T/! inde* /;+"T< H%2T2! inde* 2;+"T< H%2T4! inde* 4;+"T< H%2T%&! inde* %&;+"T<

;i%"re 1 shows the swap rates of 2-, -, 7- and 10-year mat"rity d"rin% the period of ?"ne 1--: to *ecember 2006, which represents the short, medi"m and lon% term swap

rates, respecti'ely. As the %raph shows, swap rates ha'e climbed "p since *ecember 1--:. Short-term swap rates pea$ed "p to lon%-term swap rates while the short-term yield c"r'e mo'ed abo'e the lon%-term yield c"r'e d"rin% *ecember 1--- to ;ebr"ary 2001 showed by the %raph in ;i%"re 2. ;rom 1arch 2001 to September 200 , the short-term swap rates ha'e %enerally declinedI the short-term swap rate mo'e alon% the same direction as yield c"r'es. 2ence, it is 'ery clear that the swap rates 'ary with the chan%es in yield c"r'es of 5reas"ry bonds. ;i%. ). 5he mo'ement of Swap spreads
%'1 %'/ %'& &'5 &'3 &'1 &'/ &'& %& /& 0& 1& 2& 3& 4& 5& 6& %&&

Swap spread /; Swap spread 2;

Swap spread 4; Swap spread %&;

;i%"re ) %raphs the mo'ements of swap spreads d"rin% the sample period. 7learly, the mo'ements in short, medi"m and lon% term mat"rities show the stron% tendency to mo'e to%ether.

16

3.2 Interest rate swap spread

Interest rate swap spread is determined by the difference between swap rate and 5reas"ry rate of same constant mat"rity. !n my thesis paper, ! defined swap rate as the fi/ed rate of interest that ma$es the 'al"e of the swap e@"al to Aero at the contract date ;i%"re 0 shows the mo'ements of swap spreads of 2-, - 7-, and 10-year mat"rity d"rin% the chosen sample period. !t is clearly shown that the mo'ements in short, medi"m and lon% term mat"rities ha'e tendency to mo'e to%ether all the time. 2owe'er, ! it is obser'ed that there is a pea$ at the end of -0s and be%innin% of 2000. 5he e/planation co"ld be that the financial crisis in 1--: mi%ht imply both a defa"lt ris$ e'ent and a li@"idity e'ent. ;i%"re 03 5he mo'ement of Swap spreads

%'3 %'1 %'/ %'& &'5 &'3 &'1 &'/ &'& 7un8 65 Dec8 65 7un8 66 Dec8 66 7un8 && Dec8 && 7un8 &% Dec8 &% 7un8 &/ Dec8 &/ 7un8 &0 Dec8 &0 7un8 &1 Dec8 &1 7un8 &2 Dec8 &2 7un8 &3 Dec8 &3

Sw ap spread /;=t>

Sw ap spread 2;=t>

Sw ap spread 4;=t>

Sw ap spread %&;=t>

17

3.3 #$p!anator% variab!es

Changes in slope of the yield curve. #ased on the pre'io"s research res"lts, there is a need for 'ariables that can s"mmariAe the information in the 5reas"ry Sec"rities yields. !n line with those research res"lts, ! incl"ded the slope of yield c"r'e into my model and defined the slope of the yield c"r'e as the difference between 10-year and -0-day 5reas"ry sec"rities yields. #esides the e/pected contrib"tions to 'ariation of swap spreads, this pro/y 'ariable is also interpreted as an indication of e/pected f"t"re shortterm interest rate as well as an indication of o'erall economic health.

Changes in implied volatility of Treasury market. 5here is stron%ly close correlation between swap rates and %o'ernment bond yields o'er lon%-term mat"rity based on economic theory. ! belie'e the 'olatility in 5reas"ry mar$et ha'e impact on mo'ement of swap spreads. 5o better inte%rate this 'ariable into swap spread in'esti%ation, ! "se the implied 'olatility as a pro/y of the 'olatility of 5reas"ry mar$et. 5he implied 'olatility of an option contract is the 'olatility implied by the mar$et price of the option based on an option-pricin% model. 1ore specifically, the 'olatility, %i'en a partic"lar pricin% model s"ch as #lac$-Shole model, yields a theoretical 'al"e for the option e@"al to the c"rrent mar$et price. 5his allows some non-option financial instr"ments s"ch as 5reas"ry bonds ha'in% embedded optional, to also ha'e an implied 'olatility. Changes in default premium. .'en tho"%h the defa"lt premi"m has been considered the basic determinant factor on 'ariation of swap spread, there are no stron% statistically consistent empirical e'idence on the relationship between defa"lt ris$ premi"m and swap spread chan%es has not been pro'ed e'en tho"%h. 2owe'er, ! wo"ld still li$e to incl"de this 'ariable into my model. 5he standard way is to ass"me that defa"lt ris$ in swaps can be precisely pro/ied with the information from the corporate bond mar$et as noted by 1ilas 820019. !n my paper, ! define the defa"lt premi"m as the difference between do"ble A and triple A corporate bond yields of same constant mat"rities. Changes in implied volatility of Stock market. Similar rationale as abo'e disc"ssion, !

1:

need a 'ariable which can catch the information in stoc$ mar$et. 5heoretically, there is ne%ati'e co-mo'ement between the defa"lt probability and the stoc$ price. 5herefore, ! incl"de the 'olatility in the stoc$ mar$et has its role in the swap spread chan%es. ! obtain this pro/y 'ariable by calc"latin% the standard de'iation of the daily obser'ations from S=( 00, the theoretically rationale of "sin% implied 'olatility is similar to that of implied 'olatility of 5reas"ry mar$et as e/plained abo'e. ;"rthermore, since the 'al"e of the option increases with the 'olatility, it implies that the swap spread sho"ld increase with the 'olatility as well. On the other hand, an increased 'olatility implies that the probability of defa"lt increases as well. Changes in Budget Deficit. 5here is only a limited st"dy on this 'ariable as a determinant ris$ factor on swap spread. 5he reason ! want to incl"de this 'ariable into o"r model is the iss"ance of %o'ernment bond increases with the increases in %o'ernment b"d%et deficit based on the economic theory,. 5herefore, ! consider that the 5reas"ry rates mi%ht climb "p or decline d"e to the demandDs"pply shoc$ in 5reas"ry bond mar$et. Accordin%ly, ! predict that a chan%e in swap spread is related to the chan%e in b"d%et deficit. !n my paper, ! define this e/planatory 'ariable by "sin% the monthly %o'ernment b"d%et deficit inde/. Changes in Business Cycle. +iAenber%er 81--29 ar%"ed that defa"lt ris$ allocation between swap co"nterparties 'aries with the b"siness cycleI hence this 'ariable sho"ld be controlled while testin% the impact of defa"lt ris$ on swap spread. 2owe'er, he did not show how e/actly defa"lt ris$ allocation 'aries with b"siness cycle. ;"rthermore, there are really limited researches re%ardin% to this 'ariable. 5hese @"estions moti'ate me to incl"de this 'ariable into my model. !n this paper, ! "se U.S monthly "nemployment rate as a pro/y 'iable for b"siness cycle.

1-

' Methodology
! present the empirical hypothesis, e/planations of 'ariables, 6A,72 model, cointe%ration model and methodolo%y and 6A,72 model with three different 'ersions of .rror 7orrections terms. ! arri'e at the followin% empirical hypothesis based on the literat"re re'iew, a'ailability of reliable data, and my research ob&ecti'es.

&.1. '%potheses(
5he !, swap spread will be related ne%ati'ely correlated with the slope of yield c"r'e of 5reas"ry Sec"rities. 5he !, swap spread will be positi'ely correlated with the implied Stoc$ mar$et 'olatility. 5he !, swap spread will be positi'ely correlated with the defa"lt premi"m in corporate bond mar$et. 5he !, swap spread will be positi'ely correlated with the b"siness cycle. 5he !, swap spread will be positi'ely correlated with the implied 5reas"ry mar$et 'olatility. 5he !, swap spread will be positi'ely correlated with the %o'ernment b"d%et deficit inde/. Accordin%ly, the appropriate re%ression e@"ation can be "sed to test the effects of determinant factor of interest rate swap spread is as follows3
swapspread i ,t = i , 0 + i ,1 DPt + i , 2 slopet + i , ) Tvolatilit y t + i , 0 Svolatility t

+ i , BDt + i , 6 BC t + i ,t

8 .69

i J1, 2, ), 0 imply the mat"rity of the swap contract 82K, K, 7K and 10K9 and i ,t
L

[0, ] .
2 t

DPt , slopet , Tvolatiltiy t , and Svolatility t are the Hhere, swapspread t ,

20

first difference of swapspread t , DPt , slopet , Tvolatility t , and Svolatility t respecti'ely.

21

&.2 #$p!anation of )ariab!es

swapspread t ,which is determined by the difference between swap rate and

5reas"ry yield of same constant mat"rities.

DP t , chan%es in the defa"lt premi"m which is determined by the difference

between interest rate of AA corporate bond and interest rate of AAA corporate bond of same constant mat"rities'
slopet , chan%es in the slope of yield c"r'e of 5reas"ry Sec"rities which is

determined by the difference between %o'ernment bond of 10 year mat"rity and -0 days mat"rity.
Tvolatility t , is the implied 5reas"ry mar$et 'olatility which is obtained by

calc"latin% the standard de'iation of the daily obser'ations from S=( 00.
Svolatilit y t , is the implied 'olatility of Stoc$ mar$et which is obtained by

calc"latin% the standard de'iation of the daily obser'ations from S=( 00.
BDt , is the monthly %o'ernment b"d%et deficit inde/.

BC t , is the b"siness cycle which is represented by a'era%e monthly "nemployment

rate.

&.3. *he +, -' .ode!

! "sed an 6eneraliAed A"tore%ressi'e 7onditionally 2eteroscedastic 86A,729 model de'eloped in a "ni'ariate form by #ollersle 81-:69 to test and @"antify the effect of determinant factors on interest rate swap spread which e/presses the conditional 'ariance chan%es o'er time as a f"nction of past 'al"es of the s@"ared errors and past conditional 'ariances lea'in% the "nconditional 'ariance constant. 5he basic specification of 6A,72 model is %i'en by3

2 t = + 1 2 t 1 + 1 2 t 1 ,

8 .19

22

5he error term t denotes the real-'al"ed discrete time stochastic process and t 1 is the information set a'ailable at time t 1 .
t t 1 L

80, 2 t 9

8 .29

Hhere,
0,
1 0 ,

1 0 ,
1 + 1 1 , is s"fficient for wide sense of stationary
t = t t

8 .)9

t L IID and

80, 2 t 9

5his is a 6A,72 81, 19 model, in which

2t

is $nown as the conditional 'ariance

since it is a one-period ahead estimate for the 'ariance calc"lated on the basis of any past information considered rele'ant. !t is possible to interpret the c"rrent fitted 'ariance, , as a wei%hted f"nction of a lon%-term a'era%e 'al"e dependent on

2t

, information

abo"t 'olatility d"rin% the pre'io"s period 8 1 2 t 1 9 and fitted 'ariance from the model d"rin% the pre'io"s period 8 1 2 t 1 9. Additionally, it is fo"nd that a 6A,72 81, 19 specification is s"fficient to capt"re the 'olatility dynamics in the data. 5herefore, the only one la%%ed s@"ared error and one la%%ed 'ariance is needed. 6A,72 model has se'eral ad'anta%es o'er the p"re A,72 model. ;irst of all, the 6A,72 model is more parsimonio"s. As a res"lt, the model is less li$ely to breach nonne%ati'ity constraints 8#roo$s, 20029. Secondly, a relati'ely lon% la% in the conditional 'ariance e@"ation is often re@"ired to a'oid problems with ne%ati'e 'ariance parameter estimates a fi/ed la% str"ct"re is called for in application of the A,72 model 8#ollersle', 1-:69. !n this li%ht, the 6A,72 specification allows for both a lon%er memory and a more fle/ible la% str"ct"re. 5hirdly, as pointed o"t by #ollersle', the conditional 'ariance is specified as a linear f"nction of past sample 'ariances only in the A,72 8@9 model, whereas the 6A,72 8p, @9 model allows la%%ed conditional 'ariances to enter as well,

2)

this process corresponds to some $ind of adapti'e learnin% mechanism. ;o"rthly, the 'irt"e of the 6A,72 model enables a small n"mber of terms appears to perform as well as or better than an A,72 model with many. Accordin%ly, in order to e/amine the effect of determinants of interest rate swap spread &ointly and pro'ide f"rther insi%ht of 'ariation of interest rate swap spread the abo'e re%ression e@"ation has been e/tended to a m"lti'ariate 6A,72 model of se'eral 'ariables.

2 t = + 1 2 t 1 + 1 2 t 1 + 2 2t 1 + 2 2 t 1 + + ! 2 t ! + p 2 t p
8 .09
! p

= + i u 2 t i + " 2 t "
i =1 " =1

8 . 9 Hhere,
! 0, p 0

0,
" 0 ,

i 0 ,
i =1, p.

i =1, , !,

&.& *he -ointegration /ode!

Accordin% to the economic theory, there are two 'ariables3 b"d%et deficit and b"siness cycle which are cointe%rated to some e/tent. A nonstationary 'ariable tends to wander e/tensi'ely b"t some pairs of nonstationary 'ariables can be e/pected to wander in a way that they donEt drift apart from each other 8Mennedy, 20019. Under s"ch consideration, ! cond"ct the cointe%ration test between b"d%et deficit and b"siness cycle since the data are ! 819 which means that .71 8.rror 7orrection 1odel9 estimatin% e@"ation co"ld be prod"cin% sp"rio"s res"lts, s"ch 'ariables are said to be cointe%rated. ! want to p"r%e and estimate the nonstationary 'ariables by differencin% and "sin% only differenced 'ariables if the data are shown to be nonstationary. ! am @"ite interested in the cointe%ration between b"d%et deficit and b"siness cycle since the cointe%ratin% combination is interpreted as an e@"ilibri"m relationship which can be shown that

20

'ariables in the error-correction term in an .71. #y testin% the cointe%ration of the abo'e two 'ariables, ! co"ld eliminate the "nit roots. !f the set of !819 'ariables is cointe%rated, then re%ressin% one on the others sho"ld prod"ce resid"als !809. 4.4.1 Methodo og! of "ointegration ;irst ! "se the "nit root tests to determine the order of inte%ration of the raw data series. Second, ! r"n the cointe%ratin% re%ression on b"d%et deficit and b"siness cycle. 5hird, ! apply an appropriate "nit root test to the resid"als from this re%ression to test for cointe%ration. ;o"rth, if cointe%ration is accepted, ! "se the la%%ed resid"als from the cointe%ratin% re%ression as an error correction term in an .71.

4.4.# "ointegration ana !sis

5able ) 5est for cointe%ration f"nctions *( *(7 *(10 Slope #* #7 S=( Go Go Go Go Go Go Go -

SS2K Go 1 1 SS K 2 Go 1 1 SS7K Go Go 1 1 SS10K Go Go 1 1 Slope Go Go Go Go 2 #* 2 2 2 Go Go #7 Go Go Go 2 Go S=( 1 1 Go Go Go Go 5he res"lts of the cointe%ration tests are shown in 5able ). ! tested for cointe%ratin% f"nctions for all pairs of 'ariables with a "nit root. 5able ) shows the lar%est n"mber of cointe%ratin% f"nctions fo"nd allowin% for intercept and trendDno trend. Only swaps and defa"lt premi"ms of the same mat"rity are compared. 5he table ) res"lts are consistent with o"r prediction of the cointe%ration between b"d%et deficit and b"siness cycle. !t also indicates that there e/ists a cointe%ration relationship between b"d%et deficit and b"siness cycle.

&.3 +, -' /ode! with three different versions of #rror -orrection *er.s
4.$.1.%rror "orrection Mode 1 &%"M1' ! first re%ress the swap spread on b"siness condition "sin% the +east S@"ares method.
SS t = + BCt + BC ,t

5he resid"als
BC ,t = SS t BC t

;rom the re%ression are then p"t into the followin% model containin% first differences of all interestin% 'ariables.
SS t = + 1 BC ,t 1 + 2 BC t + ) BDt + 0 S#t + Svolt + 6 Tvol t + 7 DPt + t

5he coefficients are calc"lated "sin% 6A,72.

Ta( e 4 "oefficients) *+Statistics and R# for %"M1. 2K,

7oeff A-Stat 0.2:-10 -0.)0-:6 -1.07012 1. 222) 1.-02)0 -0.62 7 -2.0)6:7 7oeff -0.160-) -0.)2626 -0.-2.0.060:0 1.):2-1 1. 6)61 -0.16:-0 0.0--01

K,
A-Stat -1.)220-1.-60 0 -0.6 0:1 2.- : 0.:06 0.206)1 -1.77760

7K,
7oeff -0.1 0 1 -0.))0)0 -0.)-.0.06-): ).00) 1 -1 .6767 -0.1 -2 0.0-070 A-Stat -1. 27)7 -2.10206 -0. 12)).7-677 2.00)2) -2.)-1-: -2.60:7-

10K,
7oeff -0.11:72 -0.)0 6) -0.00017 0.0 1): 0.66102 -1.102--0.10077 0.1)770 A-Stat -1.022:-1. )-2-2.1-201 2. )0: 0.):-2: -0.1)1 7 -1. 1:2:

1 2 ) 0 6 7 $2

-0.0): 7 -0.0 707 -:.00.0.02)70 2. ---) -2.6)0: -0.167 0 0.1-:-:

4.$.# %rror "orrection Mode # &%"M#'

! first re%ress the swap spread on b"d%et deficit "sin% the +east S@"ares method.
SS t = + BDt + BD ,t

5he resid"als

26

BD ,t = SS t BDt

the re%ression are then p"t into the followin% model containin% first differences of all interestin% 'ariables.
SS t = + 1 BD ,t 1 + 2 BDt + ) BCt + 0 S#t + Svolt + 6 Tvol t + 7 DPt + t

5he coefficients are calc"lated by "sin% 6A,72. Ta( e , "oefficients) *+Statistics and R# for %"M#. 2K,
7oeff A-Stat -1.6 7:) 0.0 -07 -0.: 2: 2.: :07oeff -.-1.-0.26::0 1.7 .0.007)0 1.)7))1 6.661) -0.22:2) 0.0:07)

K,
A-Stat 0. 767-1.7002 0.1 -12 2.)-107 0.-2)6) 0.: 0: -2. 662

7K,
7oeff :.20.-0.2-072 1.71.0.0 0: ).1- :7 --.200 ) -0.1606) 0.0710A-Stat 0.0:207 -1.:2:20 0.10 :0 ).2)7:2 1.: )1 -1.2-012 -2.::)6)

10K,
7oeff -6.:-.-0.20 : -0.00016 0.0 2 0. 66-).007:6 -0.10062 0.116:A-Stat -0.)2:1: -1.27 0 -1.20670 2.- 00 0.0)1:1 -0.00677 -2.762)0

1 2 ) 0 6 7 $2

-0.07200.22.-06 -0.0:17: 0.00117

2.26066 2.0))0: -1.--0:0 -0.0:101 -0.20 7) -).0:670 0.20:0110

4.$.$ %rror "orrection Mode $ &%"M$' 5he resid"als from re%ressin% swap spread on b"siness condition from .711 and the resid"als from .712 8Swap spread re%ressed on #"d%et *eficit9 are both "sed in the same model.
SS t = + 1 BD ,t 1 + 2 BC ,t 1 + ) BDt + 0 BC t + S#t + 6 Svolt + 7 Tvol t + : DPt + t

5he coefficients are a%ain calc"lated "sin% 6A,72. Ta( e - "oefficients) *+Statistics and R# for %"M$. 2K,
7oeff A-Stat -0.6--71 0.2 172 7oeff 6.-6.-0.116)7

K,
A-Stat 0.00::2 -1.0:7 0

7K,
7oeff :.76.-0.1070 A-Stat 0.0-62-1. 16:)

10K,
7oeff -0.100-0 -6.)-.A-Stat -1.27:)) -0.)1:6

1 2

-0.11020.0)7):

27

) 0 6 7 : $2

2.0).-0.07-70.000 2 1.:620 -1.0) 10 -0.20201 0.20-6

0.21):0 -0.7101: 2.770:6 1.71602 -0.22:-).22 7:

1. 1.-0.)6) 0.06)00 1.)6227 0.:01-1 -0.20-01 0.0:- 2

0.1)062 -2.2:06 ).260 0 1.0000) 0.10 0 -2.)):06

7.2 .-6 -0.)02:) 0.06)-1 ).0107) -10.0)0-0.162)0 0.0:-06

0.0602) -2.01) ) ). :0:7 1.-71)0 -1.0) 0 -2.-011:

-0.00017 -0.27)67 0.06-1: -0.) 1-0 - .26)00 -0.60720.1)2:0

-1.):262 -1. 2:-6 ).0 0)0 -0.20062 -0.6072-2.00:22

2:

( )mpirical results
0.1 1nit oot *est esu!t

;irst, ! "se A"%mented *ic$ey-;"ller 8A*;9 "nit root test to e/amine whether these 11 time series are stationary. 5able 1 report the test res"lts that the n"ll hypothesis of a
swapspread , slope , Tvolatility , "nit root in le'el of ele'en 'ariables 8 DP , Svolatilit y 9

with a constant and a trend are not re&ected at 1> si%nificance le'el.

5herefore, swap spread, slope, and S=( 'olatility sho"ld be considered as nonstationary while defa"lt premi"m and treas"ry 'olatility can be considered as stationary data e/cept *efa"lt (remi"m e/cept year 2. 5able 7 A"%mented *ic$ey-;"ller "nit root test
?ari$les Swap spread /; Swap spread 2; Swap spread 4; Swap spread %&; D: /; D: 2; D: 4; D: %&; Slope Tvolatilit! Svolatilit! Level +onstant onl! 8/'&33 8/'%/3 8/'/&4 8%'335 82'1%4 8/'352 8/'403 8/'662 8&'6&% 81'10& 8/'&&0 Level +onstant and linear trend 8/'4// 80'036 8/'526 8/'253 82'1/3 8/'306 8/'360 8/'6%5 8&'66& 81'42& 82'463 )nit <oot ;es ;es ;es ;es @o ;es ;es ;es ;es @o ;es

5he t-statistics from "nit root test, incl"din% either constant only or a constant and a linear trend. ,i%htmost col"mn shows if the test indicates a "nit root. 5able ) show the test res"lt that the n"ll hypothesis of a "nit root in the first
swapspread , slope , Tvolatilit y , difference of ele'en 'ariables 8 DP , Svolatilit y 9

with constant and a constant and a trend are re&ected at 1> si%nificance

le'el. 5herefore, these time series data sho"ld be considered as stationary if it ta$es difference.

2-

5able : A"%mented *ic$ey-;"ller "nit root test

?ari$les Swap spread /; Swap spread 2; Swap spread 4; Swap spread %&; D: /; D: 2; D: 4; D: %&; Slope Tvolatilit! Svolatilit! %st Difference +onstant onl! 8%%'46/ 8%%'0&0 8%/'&61 8%%'/26 8%0'00% 8%0'/6% 8%/'/20 8%%'5&6 86'%00 84'140 85'654 %st Difference +onstant and linear trend 8%%'410 8%%'/5& 8%/'&2% 8%%'/// 8%0'/34 8%0'/26 8%/'/11 8%%'5/3 86'001 84'054 85'611 )nit <oot @o @o @o @o @o @o @o @o @o @o @o

5he t-statistics from "nit root test, incl"din% either constant only or a constant and a linear trend. ,i%htmost col"mn shows whether the test indicates a "nit root. Gone of the 'ariables shows a "nit root in the first difference.

0.2 -orre!ation a.ong e$p!anator% variab!es

5able - 7orrelation matri/ for e/plainin% 'ariables.
#7 1.000 -0.070 -0.02 -0.0)-0.10) 0.01: -0.0-6 -0.07) -0.1-: #* -0.070 1.000 -0.10) -0.127 0.0): 0.0 0.0)6 0.0- 1 0.070 DP 2yr -0.02 -0.10) 1.000 0. 1) 0.0 : 0.)70 -0.0:: 0.11: 0.061 DP yr -0.0)-0.127 0. 1) 1.000 0.6 0 0. )7 -0.0:0 0.10: 0.100 DP 7yr -0.10) 0.0): 0.0 : 0.6 0 1.000 0. -6 -0.0:2 0.260.002 DP 10yr 0.01: 0.0 0.)70 0. )7 0. -6 1.000 -0.100.1:0 0.2)0 S+ S'ol Tvol -0.0-6 -0.07) -0.1-: 0.0)6 0.00.070 -0.0:: 0.11: 0.061 -0.0:0 0.10: 0.100 -0.0:2 0.26- 0.002 -0.10- 0.1:0 0.2)0 1.000 -0.06 0.)60 -0.06 1.000 0.1:: 0.)60 0.1:: 1.000

#7 #* DP 2yr DP yr DP 7yr DP 10yr S+ S'ol 5'ol

5he defa"lt premi"m for different mat"rities show stron% correlation b"t ! do not incl"de different mat"rities in the same re%ression so this is not a problem. 5he lar%est correlation otherwise enco"ntered is 0.)60 for 5reas"re 'olatility and Slope. 5his is still small eno"%h that it sho"ld ca"se problems in re%ressions.

)0

5he correlation matri/, see table -, shows how different e/plainin% 'ariable are correlated with each other, 1 or F1 means perfectly correlated and 0 means no correlation. 5wo or more stron%ly correlated e/plainin% 'ariables, incl"ded in a re%ression, can ca"se problems. ! can see some fairly stron%ly correlated 'ariables in 5able -, i.e defa"lt premi"ms of different mat"rity 8 and 7 year mat"rity of defa"lt premi"m ha'e a correlation of 0.6 09. 5his is pre'io"sly $nown fact abo"t defa"lt premi"ms, b"t it is not a problem since we ne'er incl"de different mat"rities in the same re%ression. 5he stron%est correlation to be fo"nd in the correlation matri/, that do not incl"de premi"ms of different mat"rity, is slope and treas"re 'olatility with a correlation of 0.)60, which is small eno"%h to not be of concern.

%'/ &'5 &'1 '&&3 '&&1 '&&/ '&&& 8'&&/ 8'&&1 8'&&3 %& /& 0& 1& 2& 3& 4& 5& 6& %&& T<,AS)<;AD FF SL-:,AD FF &'& 8&'1 8&'5

)1

0.3 ,na!%sis of '%pothesis *ests

5here are a total of :: entries in the tables with correlations and A-statistics. ! co"ld therefore e/pect one or two cases of false si%nificance on the 1> le'el, abo"t 2 to 0 cases of false si%nificance on the 2> le'el and e'en more cases of false si%nificance on the > le'el. #ased on st"dyin% s"ch a lar%e n"mber of correlations between 'ariables, the res"lts sho"ld therefore be done 'ery caref"lly. A concl"sion cannot be made "nless the si%nificance is 'ery stron%,

)2

The ./ swap spread wi (e re ated negative ! corre ated with the s ope of !ie d curve of Treasur! Securities. Only the .71) model shows si%nificance in the slope of yield c"r'e and then only for yr and 7yr mat"rity 8A-Stat. of -2.2: and F2.01 respecti'ely9. 5his is si%nificant at the 2> le'el. 5he correlation obser'ed is ne%ati'e, s"pportin% o"r hypothesis. So ! concl"de that chan%es in the !, swap spread will be related positi'ely to chan%es in the implied 5reas"ry mar$et 'olatility. The ./ swap spread wi (e positive ! corre ated with the imp ied Stoc0 mar0et vo ati it!. Got any of the .71s at any mat"rity is showin% si%nificance for a correlation with Stoc$ mar$et 'olatility. ! can draw no concl"sions abo"t this hypothesis from the data. The ./ swap spread wi (e positive ! corre ated with the defau t premium in corporate (ond mar0et. !t is in the defa"lt premi"m we find the stron%est correlation with !, swap, all three of o"r .71 models show stron% A-Statistics for this 'ariable 8with the e/ception of .711 and 10yr mat"rity9. .712 shows the stron%est correlation with A-statistics of F2. 6 or better, indicatin% si%nificance at the 1> le'el. 5he correlation is howe'er ne%ati'e, dispro'in% o"r hypothesis. The ./ swap spread wi (e positive ! corre ated with the (usiness c!c e. .711 shows a ne%ati'e correlation for the yr and 7yr mat"rity 8A-Stat of -1.-6 and -2.10 respecti'ely, si%nificant at the > le'el9. .71) shows a ne%ati'e correlation for yr and 7yr mat"rity 8A-Stat. of -2.2: and -2.01 respecti'ely, si%nificant at the 2> le'el9. 5he data dispro'es o"r hypothesis at the 2> le'el.

))

%'/ %'& &'5 &'3 &'1 &'/ &'& 8&'/ %& /& 0& 1& 2& 3& 4& 5& 6& %&& Swap spread 2;=t> Business +!cle
The ./ swap spread wi (e positive ! corre ated with the imp ied Treasur! mar0et vo ati it!. 5he 5reas"ry mar$et 'olatility shows only a si%nificant correlation with !, swap spread in .711 and for 7yr mat"rity 8A-Stat. of -2.)-, si%nificant at 2> le'el9. 7onsiderin% the n"mber of correlations e/amined, this may still be a statistical fl"$e. 5he coefficient is ne%ati'e, in conflict with o"r hypothesis. ;or other mat"rities and for the other two .71s, the coefficient is also mostly ne%ati'e, b"t none of these cases ha'e any hi%her si%nificance. ! can neither pro'e or dispro'e o"r hypothesis from this data. 2owe'er, it seems that any correlation sho"ld be @"ite wea$.

)0

The ./ swap spread wi (e positive ! corre ated with the government (udget deficit inde1. .711 shows a ne%ati'e correlation for the 10yr mat"rity 8A-Stat. of -2.1-, si%nificant at > le'el9 and .712 shows a ne%ati'e correlation for yr and 7yr mat"rity 8A-Stat. of -1.7 and -1.:) respecti'ely, si%nificant only at the 10> le'el9. 5he correlation is in conflict with o"r hypothesis, b"t a%ain the si%nificance is @"ite low.

* Conclusion
#ased on my test res"lts, ! fo"nd o"t that the interest rate swap spread is positi'ely correlated with the defa"lt premi"m in corporate bond mar$et. *"e to the limited siAe of a'ailable data set, ! canEt see 'ery stron% correlation. O'erall it is 'ery hard for me to draw any stron% concl"sions. 2owe'er, there are two relati'ely two stron% correlationsI the interest rate swap spread is correlated with defa"lt premi"m in corporate bond mar$et and the %o'ernment b"d%et deficit inde/. ! fo"nd o"t that the stron%est correlation with !, swap in the defa"lt premi"m. .712 shows the stron%est correlation with A-statistics of F2. 6 or better which indicates si%nificance at the 1> le'el. 2owe'er, this correlation is ne%ati'e. At the same time, it is shown that the stron% correlation with !, swap in he %o'ernment b"d%et deficit inde/. 2owe'er, the correlation is in conflict with o"r hypothesis. A%ain the si%nificance is @"ite low. Accordin% to the test res"lts, ! thin$ it mi%ht be some other relati'ely important factors which ! i%nored in the be%innin%. ! mi%ht ta$e the considerations of swap spreads &ointly not &"st separatin% in different mat"rity. Also ! sho"ld consider the international mar$et spill o'er effect for e/ample the UM and ?apanese mar$et since the U.S. mar$ets are affected by the rest of world .

)6

+ Ackno%ledgements
! wo"ld li$e to than$ (rofessor Anders 2ederstierna who is the one leadin% me to ;inance world for "sef"l comments and research s"pport. At the same time ! wo"ld li$e to say than$s to my dear h"sband 5orb&orn #lom@"ist for helpf"l s"pport and s"%%estions d"rin% my st"dy in Sweden. Also ! wo"ld li$e to than$ my parents who s"pport me with endless lo'e all the time.

)7

$e,erence
#ondnar,6.1., 2ayt,6.S., 1arston, ,.7., = Smithson,7.H. 81-- 9. Hharton s"r'ey of deri'ati'es "sa%e by US non-financial firms. ;inancial 1anan%ement,20,100-110 #ernadette A. 1inton 81--69 An empirical e/amination of basic 'al"ation models for plain 'anilla U.S interest rate swaps. *ai, N., Sin%letion, M.?., 1--7 Specification analysis of Affine term str"ct"re models, Hor$in% (aper3 Stanford Uni'ersity 2"ll,?.8200 9 Options, ;"t"res and Other *eri'ates, 10--1 0 !n, ;., #rown, ,., ;an%,4.,8200)9 1odelin% 'olatility and chan%es in the swap spread. !nternational ,e'iew of ;inancial Analysis, 8129 0 - 61. +itAenber%er,,.2.81--29. Swaps3 (lain and fancif"l. ?o"rnal of ;inance,07, -7-620 +an%, +., +itAenber%er, , and +i" 81--:9 *eterminants of interest rate swap spread, ?o"rnal of #an$in% = ;inance 22, 1 07-1 )2 5"rnb"ll, S.1. 81-:79. Swaps3 A Aero s"m %ameO ;inancial 1anan%ement,16, 1 -21 Sorensen, ..2., #ollier, 5.;., 1--0. (ricin% Swap *efa"lt ,is$. ;inancial Analsysis ?o"rnal 0,2)-)) Mennedy, (. 820029 A 6"ide to .conometrics. Mod&o 1. Aped&ino", Hhat dri'es interest rate swap spreadsO 200) An empirical

):

analysis of str"ct"ral chan%es and implications for modelin% t of the swap term str"ct"re.

Mod&o 1. Aped&ino", Hhat dri'es interest rate swap spreadsO 200) An empirical analysis of str"ct"ral chan%es and implications for modelin% t of the swap term str"ct"re. 4ictor ;an%, ,onny 1"l&ono, An empirical analysis of the A"stralian dollar swap spreads, ,ecei'ed 1) A"%"st 2001I accepted ?"ne 2002.