Work on Term structure modeling

Sankha Banerjee

Introduction

Background

Modeling

Results

Summary

The? Panic of 2008 Whatassetshaveheldtheirvaluebest US Treasury bills (stable value) Nominal US Treasury bonds (increasing value) WhyhavenominalTreasuriesbeensuch good hedges? Flight to quality helps safe assets, but why are nominal Treasuries regarded as safe? They have no credit risk, but they do have inflation risk Have nominal Treasuries always hedged investors against other risks? Understanding Bond Risks Thislectureexplorestime-variationininflation risk and its effect on the nominal Treasury yield curve TheanalysismakessomeuseofTIPSdata but TIPS are not the main focus Attheend,abriefanalysisofcurrencies along the same lines
2

Introduction

Background

Modeling

Results

Summary

What do we know from the 2008?

US Treasury bills and Nominal US Treasury bonds hold their value during a high volatility event They act as good hedge, and the phenomenon of "Flight to quality helps safe assets. Why are nominal Treasuries regarded as safe? They have no credit risk, but they do have inflation risk

Introduction

Background

Modeling

Results

Summary

Time series of the stock-bond covariance and the CAPM beta of the 10 year nominal bond
4

Introduction

Background

Modeling

Results

Summary

Time series of the stock-deflation covariance and the CAPM beta of deflation
5

Introduction

Background

Modeling

Results

Summary

Time series of the US 10-year inflation-indexed yields Estimated time series of the real rate
6

Introduction

Background

Modeling

Results

Summary

Time series of the real bond second moments

Introduction

Background

Modeling

Results

Summary

Time series of the nominal bond second moments

Introduction

Background

Modeling

Results

Summary

Estimated time series of permanent and transitory components of expected inflation

Introduction

Background

Modeling

Results

Summary

Estimated time series of covariance of the inflation with interest rate. Response of the nominal expected excess returns to covariance of the inflation with Interest rate.

10

Introduction

Background

Modeling

Results

Summary

Estimated time series of expected excess returns for 10 year nominal bonds.
11

Introduction

Background

Modeling

Results

Summary

Response of the yield curves to covariance of the inflation with Interest rate.
12

Introduction

Background

Modeling

Results

Summary

Changing Conventional Wisdom Late1970 sandearly1980 s: Bonds are exposed to the risk of stagflation Avoid them unless the term premium is high 2000 s: Bonds are hedges against the risk of deflation Anchor to windward Hold them even at a low term premium Changing CW reflects changing reality Bonds as hedges in 2007-2008 Changing Inflation Behavior The changesinmeasuredbondrisksappear to be related to changing behavior of the Phillips Curve WhenthePhillipsCurveisstable(early 1960 s, 2000 s), inflation falls when unemployment rises Then bonds do well in bad times and hedge macroeconomic risk WhenthePhillipsCurveisunstable(1970 s and early 1980 s), inflation and unemployment move together (stagflation) 13 Then bonds do badly in bad times and are risky

Introduction

K Real stochastic discount factor (SDF): K Real stochastic discount factor (SDF): K Real stochastic discount factor (SDF):

Real Term Structure Real Term Structure Real Term Structure Term structure modeling
Background Modeling Results

Summary

K K x xtt is is real real rate: rate: K xt is real rate: K xt is real rate: K K z ztt drives drives time-variation time-variation in in volatility volatility of of SDF: SDF: K zt drives time-variation in volatility of SDF: K zt drives time-variation in volatility of SDF: K K x xtt and and z ztt follow follow AR(1) AR(1) processes: processes: K xt and zt follow AR(1) processes: K xt and zt follow AR(1) processes:
18 18 18

14

18

Introduction

Background

Modeling

Results

Summary

Modelling the Yield Curve Changingbondriskdoesseemtomatterover the long run Intheshortrun,however,thereareother influences on the yield curve Tocaptureitsmovements,weneedto consider more traditional factors as well: The real interest rate Investor attitudes towards risk Expected inflation Campbell,Sunderam,andViceira2008 undertakes this project A Bond Pricing Model Weconsiderfivefactorsthatmoveindifferent ways: Real interest rate xt (transient) Risk aversion zt (persistent) Long-run expected inflation t (permanent) Temporary expected inflation t (transient) Covariance of inflation with recession t (persistent, can change sign) Thefivefactorsarenotdirectlyobserved,so we back out their implied values from data we do observe Nonlinear Kalman filtering
15

Introduction

Background

Modeling

Results

Summary

eal (inflation-indexed) term time-variation structure is affine in the K zt drives exogenous in real risk premia. ort-term real interest rate and aggregate risk ersion: K exogenous time-variation in real risk premia. Kz Real bonds: t drives

Risk Premia on Real Assets

K Real bonds:

K Equities: mple pricing structure, yet risk premium on real nds varies over time. K Equities:

19
20

16

K Log nominal SDF:

Nominal Term Structure K Log nominal short rate:

17

K Log nominal SDF:

K Log nominal short rate: Fisher equation

Inflation risk premium

K The conditional covariance between the real economy (log real SDF) and log inflation determines risk premium on short-term nominalInflation bonds: Fisher equation risk premium
K The conditional covariance between the real economy (log real SDF) and log inflation determines 23 risk premium on short-term nominal bonds:

Introduction

CAPM beta
-0.4 0.1 0.2 0 -0.3 -0.2 -0.1

0.3

Background

Modeling

CAPM beta of bonds (2002.06-2008.09)

Date

Results

3-month centered beta, 10-year Treasury on S&P500

8

06 /2 0 10 /20 /1 02 0 01 /20 /3 02 0 05 /20 /2 03 2 09 /20 /1 03 1 01 /20 /0 03 1 04 /20 /2 04 2 08 /20 /1 04 2 12 /20 /0 04 2 03 /20 /2 04 4 07 /20 /1 05 4 11 /20 /0 05 3 02 /20 /2 05 3 06 /20 /1 06 5 10 /20 /0 06 5 01 /20 /2 06 5 05 /20 /1 07 7 09 /20 /0 07 6 12 /20 /2 07 7 04 /20 /1 07 7 08 /20 /0 08 7/ 20 08

Summary

18

Introduction

Background

Modeling

Results

Summary

19

Introduction

Background

Modeling

Results

Summary

Nominal Nominal Yields Yields

20
28

Introduction

Background

Modeling

Results

Summary

Real Yields

29

21

Introduction

Background

Modeling

Results

Summary

Bond Second Moments Bond Second Moments

Bond-Equity Covariance Bond-Equity Covariance

Bond Variance Bond Variance

22
31

31

Introduction

Real State State Variables Variables Real

Background

Modeling

Results

Summary

Real Interest Interest Rate Rate Real

Risk Aversion Aversion Risk

32 23 32

Introduction

Inflation Components Components Inflation

Background

Modeling

Results

Summary

PermanentExpected Expected Permanent Inflation Inflation

TemporaryExpected Expected Temporary Inflation Inflation

33 24
33

Introduction

Background

Modeling

Results

Summary

Inflation-Recession Covariance
Stagflation risk

Deflation risk

34

25

Introduction

Background

Modeling

Results

Summary

Implications for the Yield Curve Weplottheyieldcurveatthesamplemeanofall the state variables Thenwevaryeachstatevariabletoitssample minimum and maximum, while holding the other state variables at their sample mean

26

Introduction

Background

Modeling

Results

Summary

This variable is the main innovation of our model, and plays important role: (t)2 drives time variation in the conditional volatility of both realized inflation and expected inflation. zt t drives time variation in the covariance of the real economy with inflation, and thus determines nominal bond risk premia. This covariance (and thus bond risk premia) can switch sign as t takes positive or negative values (zt is always positive in the data)

27

Introduction

Background

Modeling

Results

Summary

Nested Models Both zt and t constant Two-factor affine yield model of Campbell and Viceira (2001, 2002). Both real bond risk premia and nominal risk premia are constant. zt varying and t constant Three-factor affine yield model ((Bekaert et al., 2004, Buraschi and Jiltsov 2006, Wachter 2006). Both real bond risk premia and nominal bond risk premia vary with aggregate risk aversion zt constant and t varying Single-factor affine yield model for the real term structure, and a linear-quadratic model for the term structure for nominal interest rates. Constant real bond risk premia, time-varying nominal bond risk premia. 25
28

Introduction

Background

Modeling

Results

Summary

Estimation Maximum likelihood via nonlinear Kalman filter because state variables are unobserved. UnscentedKalmanfilter(JulierandUhlmann 1997, Wan and van der Merwe 2000, Koijen and Binsbergen 2008) 26 Observed Variables Nominalyieldcurveatmaturities3months,1 year, 3 years, 10 years TIPSyield Realized inflation Equityreturnsanddividendyield(proxyforrisk aversion) Realized bond variance and bond-equity covariance in daily data

29

Introduction

Background

Modeling

Results

Summary

Implications for the Yield Curve Realinterestrateandtemporaryexpected inflation move the short end Riskaversionmovesthelongend Permanentexpectedinflationmovestheyield curve up and down in parallel Inflation-recessioncovariancedrivesthe curvature of the yield curve Implications for the Yield Curve Fixed-incomepractitionersanalyzetheyield curve using level, slope, and curvature factors Theydonotrelatethesefactorstoexternal market conditions Ourmodeldoes: Real interest rate and permanent expected inflation drive the level factor Real interest rate, risk aversion, and temporary expected inflation drive the slope factor Inflation-recession covariance drives the curvature factor 41 Implications for Bond Returns Whathappenswheninvestorsbecomemorerisk averse? Ifbondsarerisky,theninvestorssellbothstocks and bonds Ifbondsarehedges,theninvestorssellstocks and buy bonds (flight to quality) Thusmovementsinriskaversionamplifythe covariance of bonds and stocks If the covariance is positive, it becomes more positive If the covariance is negative, it becomes more negative 42 Implications for Term Premia Expectedexcessbondreturns(termpremia)are determined by Price of risk quantity of risk

30