discount rates, Почетная лекция по финансам при поддержке...
TRANSCRIPT
Discount Rates
John H. Cochrane
University of Chicago Booth School of Business
April 11, 2011
Discount rates
1. Facts: How risk discount rates vary over time and acrossassets.
2. Theory: Why discount rates vary.I “Macro,”“Behavioral,”“Segmented/institutional,”“Liquidity”
3. ApplicationsI Portfolio theory, Active/passive management, Accounting,Corporate Finance
4. Apology — see long paper for citation, documentation
Forecasting with DPHorizon k b t(b) R2 σ [Et (Re )]
σ[Et (R e )]E (R e )
1 year 3.8 (2.6) 0.09 5.5 0.765 years 20.6 (3.4) 0.28 29.3 0.62
Ret→t+k = a+ bDtPt+ εt+k ; σ [Et (Re )] ≡ σ
(b× Dt
Pt
)
1950 1960 1970 1980 1990 2000 2010
0
5
10
15
20
25
4 x D/P and Annualized Following 7Year Return
4 x DPReturn
Long-Horizon Regression Coeffi cients and Price VolatilityI Identity: (dpt ≡ log(Dt/Pt ); ρ = 0.96)
dpt ≈k
∑j=1
ρj−1rt+j −k
∑j=1
ρj−1∆dt+j + ρkdpt+k
I Long-run regressions, and coeffi cient identity
k
∑j=1
ρj−1rt+j = a+ b(k )r × dpt + εrt+k , etc.
⇒ 1 ≈ b(k )r − b(k )∆d + b(k )ρkdp .
b(k )r b(k )∆d b(k )ρkdp
Direct regression , k = 15 1.01 -0.11 -0.11Implied by VAR, k = 15 1.05 0.27 0.22
VAR, k = ∞ 1.35 0.35 0.00
I Why do prices (p/d) move? 100% (135%!) discount rates, 0%(-35%!) dividend growth. 0% “Rational Bubble.”
A Pervasive Phenomenon, and cycles
I A pervasive phenomenon:
1. Stocks. DP → Return, not dividend growth2. Treasuries. Yield → Return, not rising rates3. Bonds/CDS. Yield → Return, not default4. Foreign Exchange. Interest spread → Return, not devaluation5. Sovereign Debt, Foreign Assets. → Return, not repayment,exports
6. Houses. Price/Rent → Return, not rent growth.
Houses —Price and Rent
1960 1970 1980 1990 2000 2010
6.8
7
7.2
7.4
7.6
7.8
20 x Rent
Price
Date
log
scal
e
20 x rentCSW priceOFHEOprice
Houses: b t R2 Stocks: b t R2
rt+1 0.12 (2.52) 0.15 0.13 (2.61) 0.10∆d t+1 0.03 (2.22) 0.07 0.04 (0.92) 0.02dpt+1 0.90 (16.2) 0.90 0.94 (23.8) 0.91
A Pervasive Phenomenon, and cycles
I A pervasive phenomenon:
1. Stocks. DP → Return, not dividend growth2. Treasuries. Yield → Return, not rising rates3. Bonds/CDS. Yield → Return, not default4. Foreign Exchange. Interest spread → Return, not devaluation5. Sovereign Debt, Foreign Assets. → Return, not repayment,exports
6. Houses. Price/Rent → Return, not rent growth.
I Common element, business cycle association:low prices, high returns in recessions. High prices, lowreturns in booms
I “Bubble?”"Prices too high" ⇐⇒ Discount rate “too low”
Multivariate Challenges: More variables
1. Many forecasters. Multiple regression? Common forecastersacross assets?
r stockt+1 = as + bs × dpt + cs × yst + d ′szt + εst+1?
rbondt+1 = ab + bb × dpt + cb × yst + d ′bzt + εbt+1?
2. Are Et (r it+1) = bi × xt correlated across assets? Factorstructure of time-varying expected returns?
3. Relate mean to covariance
Et(r it+1
)= covt (r it+1f
′t+1)λt
4. Can’t just run big regressions!
5. Back to prices (price/dividend) — long-run forecasts?
Understanding prices. Short and long-run forecasts
Coeffi s t-stats Otherdpt cayt dpt cayt R2 σ [Et (yt+1)]
rt+1 0.12 0.071 (2.14) (3.19) 0.26 8.99∆dt+1 0.024 0.025 (0.46) (1.69) 0.05 2.80dpt+1 0.94 -0.047 (20.4) (-3.05) 0.91cayt+1 0.15 0.65 (0.63) (5.95) 0.43
r lrt 1.29 0.033 0.51∆d lrt 0.29 0.033 0.12
r lrt =∞
∑j=1
ρj−1rt+j ; ∆d lrt =∞
∑j=1
ρj−1∆dt+j
Understanding prices. Short and long-run forecasts
1950 1960 1970 1980 1990 2000 201020
10
0
10
20
30
40
dp and caydp onlyac tual r
t+ 1
rt+1 =a+ b× dpt [+c × cayt ] + εt+1;
1950 1960 1970 1980 1990 2000 20105
4.5
4
3.5
3
2.5E(r l r|dp,cay)E(r l r|dp)dp
∑∞j=1 ρj−1rt+j =
a+ b× dpt [+c × cayt ] + ε
dpt =∞
∑j=1
ρj−1rt+j −∞
∑j=1
ρj−1∆dt+j
0 5 10 15 20
0
0.2
0.4
0.6
0.8
1
R es pons e to∆d s hoc k
returndiv grow ths hoc k da te
0 5 10 15 200.05
0
0.05
0.1
0.15R es pons e to dp s hoc k
Σρj1 rt+j
=1.29
Σρj1 ∆dt+j=0.29
rt = 0 .96
returndiv grow th
0 5 10 15 200.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
R es pons e to 1σ c ay s hoc k
Σρj1 rt+j=0.033
Σρj1 ∆dt+j=0.033
returndiv grow th
0 5 10 15 20
0
0.2
0.4
0.6
0.8
1
R es pons e to∆d s hoc k
pr ic ediv idends hoc k da te
0 5 10 15 20
1
0.8
0.6
0.4
0.2
0
0.2
R es pons e to dp s hoc k
pr ic ediv idend
0 5 10 15 20
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16R es pons e to 1σ c ay s hoc k
pr ic ediv idend
The “cross section”
1. Chaos
2. CAPM E (Rei ) = βiE (Rem)
3. Chaos again E (Rei ) = αi + βiE (Rem) (value)
4. Fama and French
E (Rei ) = βiE (Rem) + hiE (hml) + siE (smb)
3. Chaos again
E (Rei ) = αi + βiE (Rem) + hiE (hml) + siE (smb)
(Market, value, size), momentum, accruals, equity issues,beta-arbitrage, credit risk, bond & equity market timing, carrytrade, put writing, “liquidity provision,”...
Value effect and factor4. Fama and French
E (Rei ) = βiE (Rem) + hiE (hml) + siE (smb)
Grow th Value0.2
0
0.2
0.4
0.6
0.8E(r)
β x E(rmrf)
b x E(rmrf)
h x E(hml)
Aver
age
retu
rnAv erage returns and betas
Fama - French 10 B/M sorted portfolios. .
Value (size, and bond factors)4. Fama and French
E (Rei ) = βiE (Rem) + hiE (hml) + siE (smb)
a. Theories (m) only need to explain the factor
E (Rei ) = ...+ hiE (hml) (Fama-French)
E (hml) = cov(hml ,m) (Theory)
b. Value stocks rise and fall together; mean⇔covariance. (APT).But theories must now explain covariance!
c. Value betas explain other E (Re ) sorts, e.g. sales growth.
5. Chaos again..(Market, value, size), momentum, accruals,issues, beta-arbitrage, credit risk, bond & equity markettiming, carry trade, put writing, “liquidity provision,..
E (Rei ) = αi + βiE (Rem) + hiE (hml) + siE (smb)
6. How to repeat FF?
The Multidimensional Challenge
I (Market, value, size), momentum, accruals, equity issues,beta-arbitrage, credit risk, bond & equity market timing, carrytrade, put writing, “liquidity provision,”...
1. Which of these are independently important for E (Re )?(“multiple regression”)
2. Does E (Re ) spread correspond to new factors in cov?
3. Do we need all the new factors? Or again, fewer factors thanE (Re ) characteristics?
4. Why do prices move? —Long run.
I How to approach such a highly multidimensional problem?
Asset Pricing on Characteristics/Unification
1. Portfolio sorts are really cross-sectional regressions
Log(B/M)
E(R)
1 2 3 4 5Portfolio
PortfolioMean
Securities
E (Rei ) = a+ b log(b/mi ) + εi ; i = 1, 2, ...N
Asset Pricing on Characteristics/Unification
1. Portfolio sorts are really cross-sectional regressions
E (Rei ) = a+ b′Ci + εi ; i = 1, 2, ...N
2. Time series and cross-section are really the same thing —apanel.
Reit+1 = a+ b′Cit + εit+1
3. Result: Expected return is a function of characteristics
E (Reit+1|Cit )
Cit = [size, b/m, momentum, accruals, d/p, credit spread....]
4. Covariance with factors is also a function of characteristics.
covt (Reit+1, ft+1) = g(Cit )
E (Re |C ) = g(C )× λ?
The value/size premium as a panel regression
sizet bmt ∆sizet ∆bmt1. Cross section -0.030 0.272. Pooled -0.022 0.553. Time dummies -0.031 0.294. Portfolio dummies -0.087 1.485. Pooled -0.030 0.46 -0.38 1.11
FF 25 size and BM portfolios, monthly data.
Reit+1 = a+ (at ) + (ai ) + b× sizei ,t + c × bmi ,t+ (d × ∆12sizei ,t + e × ∆12bmi ,t ) .
Prices?
1. Why ER/β, not p,PV ?2. Prices/long run may simplify.
2.1 Campbell-Shiller:
∞
∑j=1
ρj−1rt+j =∞
∑j=1
ρj−1∆dt+j − dpt
2.2 One-period:
Rt+1 =Dt+1Pt
=
(Dt+1Dt
)/(PtDt
)rt+1 = ∆dt+1 − dpt
3. Long-run / price in the “cross-section”?
∞
∑j=1
ρj−1r it+j = a+ b′Cit + εi ?
Panel regressions and prices —simple example
Portfolio dummies Time dummies Pooledb b
1−ρφ b b1−ρφ dpt−1 ∆dpt−1
r 0.107 0.903 0.044 0.325 0.090 0.074∆d -0.005 -0.041 -0.083 -0.611 -0.004 0.073∆d∗ -0.011 -0.097 -0.092 -0.675 -0.012 0.076dp 0.92 0.90 0.94 0.002
FF 10 size portfolios, monthly data
r it+1 = a+ [ai ] + [at ] + b× dpit +[c × ∆dpit
]+ εit+1
dpit+1 = adp + φdpit + εdpit+1
Theory classification
1. Theories based on fundamental investors, with few frictions.
1.1 Macroeconomics — tie to macro data.
1.1.1 Consumption, Aggregate risks.1.1.2 Risk sharing/background risks (Hedging outside income)1.1.3 Investment and production.1.1.4 General equilibrium, including macroeconomics
1.2 Behavioral Irrational expectations. Tie to price data. Otherdata?
1.3 Finance. ER − β. Price data.
2. Theories based on frictions.
2.1 Segmented markets —different investors in different markets;limited risk bearing.
2.2 Intermediated markets..2.3 Liquidity. i) Idiosyncratic ii) Systemic iii) Information trading.
Risk-neutral probability theorem.
p = ∑s
πsmsxs =1R f ∑
sπ∗s xs
π∗s ≡ πsmsR f
ms = βu′(cs )u′(c)
Investor
Investor
Investor
Security class
Investor Investor
Investor
Investor
Intermediary
“Debt”“Equity”
?
?
Other assets
Segmented markets
Intermediated markets
Security class
Securities
Consumption/habits
1990 1992 1995 1997 2000 2002 2005 2007 2010
Surplus consumption (CX)/C and stocks
SPC (CX)/C
P/D
Xt ≈ k ∑∞j=0 φjCt−j ; risk aversiont = γ Ct
Ct−Xt
Investment and Q
1990 1992 1995 1997 2000 2002 2005 2007 20101
1.5
2
2.5
3
3.5
4Nonres. Fixed I/K and Q
I/KP/(20*D)ME/BE
1+ α itkt =markettbookt
= Qt
Challenges for theories
2007 2008 2009 2010 20110
1
2
3
4
5
6
7
8
9
10
BAA
AAA
20 Yr
5 Yr
1 Yr
baa,aaa
Bond yields
2007 2008 2009 2010
0.5
1
1.5
2
2.5
3
3.5
4
baaaaa and stocks
baaaaasp500p/d
Bonds and stocks
I Coordinated risk premium in all markets, incl. unintermediatedI Mean returns are associated with comovement.I Strong correlation with macroeconomics.I Macro: the recession is the risk premium.
“Arbitrages”
Feb 07 Jun 07 Sep 07 Dec 07 Apr 08 Jul 08 Oct 080
1
2
3
4
5
6
Bond
CDS
Perc
ent
Source: Fontana (2010)
“Arbitrages”
Feb 07 Jun 07 Sep 07 Dec 07 Apr 08 Jul 08 Oct 08 Jan 09 May 090
1
2
3
4
5
6
7
swaplibor
3 Month rates. Source: Baba and Packer (2009).
Price and volume in the tech “bubble.”
Feb98 Sep98 Mar99 Oct99 Apr00 Nov00 May01 Dec01
50
100
150
200
250
300
350
400
450
500
NYSE
NASDAQ
NASDAQ Tech
Feb98 Sep98 Mar99 Oct99 Apr00 Nov00 May01 Dec010
100
200
300
400
500
600
700
800
NYSE
NASDAQ
NASDAQ Tech
Dollar volume
I Price (discount rate) ⇒ Volume? Or some Volume ⇒ Price,like money?
I Why so much information trading?
Portfolio theory with many factorsI Sexy profit ideas? The average investor must hold the marketI Portfolio theory based on differences
Bonds —a cautionary tale
0 1 2 3 4 5 6 7 8 9 100
10
20
30
40
50
60
70
80
90
100
Price of a bond that matures in year 10 s imulation
time, years
bond
pric
e
Stocks (your endowment) in the crisis
2007 2008 2009 201060
50
40
30
20
10
0
10
What to do???
log
scal
e
perc
ent
S&P500
2007 2008 2009 2010
10
20
30
40
50
60
70
80
What to do??
Perc
ent
S&P500 volati l ity
montly vol.v ix
share =1γ
E (Re )σ2(Re )
. 0.6 =120.040.182
=⇒ 120.04
0.702= 0.04???
Payoff streamsI For long-run investors the coupons of an indexed perpetuityare the risk free asset, not the short term rate!
I Risky asset? If utility is quadratic,max{ct} E ∑∞
t=0 δt(− 12)(ct − c∗)2 and for any amount of
time-varying expected returns,
“Long run mean” E (x) = 11−β ∑∞
j=0 βjE (xt+j )
Alphas, betas, and performance evaluation
1994 1996 1998 2000 2002 2004 2006 2008 201050
40
30
20
10
0
10
20
30
40
50EquitMk tNeut
HFrmrf
Reit = αi + βi rmrft +hihmlt + si smbt +uiumdt + vol., carry, beta-arb, issues, ...+εit???
Procedures, corporate, accounting, regulation.
I Capital budgeting, valuation
value of investment =expected payout
R f + β [E (Rm)− R f ] ,
I Accounting, regulation, capital structure, if prices can changeon discount rate news?
I Macroeconoimcs. Recessions are rises in the risk premium,not “the” interest rate. Risk bearing, not allocation ofconsumption over time. .
Conclusion
I Discount rates vary over time and across assets a lot morethan you thought
I Empirical: how. Theoretical: why. Applications: at all.I We’ve only startedI How do you ask the right question?