Transcript Consumption Based Asset Pricing After 25 Years

Consumption-Based Asset
Pricing After 25 Years
Douglas T. Breeden*
*Dean and William W. Priest Professor of Finance,
Duke University, Fuqua School of Business
Reference notes, tables and graphs for
June 20, 2005 Western Finance Association Talk
Perspective and Goal of the Paper

Rip Van Winkle (Austin Powers?) academic career:
Intertemporal consumption, portfolio theory and asset
pricing research 1976-1989. Left Duke 1992 to build
Smith Breeden, did applied research on mortgages and
corporate bonds. Returned to academia in 2000. Dean at
Duke 2001-present, with usual IQ drop of Dean.

Paper will be a follow-up look at the numbers for some
major results of consumption-based asset pricing that I
began working on 25 years ago. There are many areas
for future research that we’ll see.
2
Preceding Work on Intertemporal
Asset Pricing and the Term Structure

Markowitz (1952), Sharpe (1964), Lintner (1965)
and Mossin (1966) developed diversification and
the market-based CAPM.

Samuelson (1969), Merton (1969, 1971, 1973),
Hakannson (1970), Fama (1970), Pye (1972) and
Long (1975) pioneered intertemporal investments.

Hirshleifer (1970 book), Cox, Ingersoll and Ross
(1985) and Garman (1976) on term structure.
3
First Decade of Selected Research on
Consumption-Based Asset Pricing

Rubinstein (1976 BJEMS), Breeden-Litzenberger
(1978 JB), Lucas (1978 Ec), Breeden (1979 JFE)

Hall (1978), Breeden (1980), Stulz (1981), Grossman-Shiller
(1982) , Marsh-Rosenfeld (1982), Mankiw-Shapiro (1986),
Mehra-Prescott (1985), Wheatley (1986), Hansen-Singleton
(1982, 1983), Ferson (1983), Breeden (1984),Gibbons-Ferson.

Chen, Roll and Ross (1986), Grossman, Melino, Shiller
(1987), Campbell-Shiller (1988), Breeden, Gibbons and
Litzenberger(1989)

Term Structure: Garman (1976), Cox, Ingersoll, Ross (1985),
Breeden (1986), Harvey (1988, 1989, 1991), Dunn-Singleton
(1986), Sundaresan (1989)
4
Second Decade of Selected Research on
Consumption-Based Asset Pricing

Constantinides (1990), Abel (1990), Epstein-Zin(1991), HansenJagannathan (1992), Cochrane(1991,1994,1996), Campbell (1991)

Campbell-Shiller(1990), Shanken (1990), Fama (1991), KandelStambaugh(1991),Ferson-Constantinides (1991), Mankiw-Zeldes
(1991), Fama-French (1992), Brennan-Schwartz-Lagnado(1997)

Heaton (1995), Elton-Gruber-Blake (1995), He-Modest (1995),
Constantinides-Duffie (1996), Jagannathan-Wang (1996),
Campbell-Cochrane (1999), Campbell-Viceira (1999), FersonHarvey(1999)
5
Third Decade of Selected Research on
Consumption-Based Asset Pricing

Campbell (2000), Heaton-Lucas (2000), Lettau-Ludwigson
(2001a,b), Santos-Veronesi (2001),Brav-Constantinides-Geczy
(2002), Wachter (2002), Barberis-Huang-Santos (2003)

Verdelhan (2003), Lustig-Verdelhan(2004), Piazzesi-SchneiderTuzel (2003),Bansal-Yaron(2005), BansalDittmar,Lundbad(2005),Bansal-Dittmar-Kiku (2004)

Jagannathan and Wang (2004), Parker-Julliard (2005),
Campbell-Vuolteenaho (2004),Hansen-Heaton-Li(2005)

In total, 179 articles with “consumption, asset pricing” in the
abstract, far more than mentioned here. Apologies.
6
Consumption Based Asset Pricing
Outline of Paper
1. Consumption and marginal utility.
2. Consumption risks of corporate profits & cash
flows. Capital budgeting.
3. Consumption betas vs. market betas for industries.
4. Term structure slope and consumption growth.
5. Risk and return and the “Maximum Correlation
Portfolio” for consumption.
6. Consumption deviations from wealth and the
investment and income opportunity sets.
7. Volatility of family consumption and the “Equity
Premium Puzzle.”
7
I.
Consumption and
Marginal Utility
Consumption and Marginal Utility

Some likely statistical indicators of times
when the marginal utility of $1 is quite
high, are the following:
1. Unemployment rate is increasing.
2. Job growth is less rapid than normal.
3. Businesses are failing more often,
risky bonds’ yield spreads are high.
4. Banks are charging off more loans.
9
United States Unemployment Rate
EOQ, 1960-Q4 2004
12
11
10
9
8
7
6
5
4
3
2
51
21
991
961
931
901
871
841
811
781
751
721
691
661
631
601
10
U.S. Real Consumption Growth
Last Four Quarters Percent, 1960-2004 Q4
10
8
6
4
2
0
05Q1
02Q1
99Q1
96Q1
93Q1
90Q1
87Q1
84Q1
81Q1
78Q1
75Q1
72Q1
69Q1
66Q1
63Q1
60Q1
-2
-4
11
Unemployment Changes (6mo) and
Consumption Growth 1959-2004
Change in Unemployment Rate (6 Months)
2
1.5
1
y = -0.13x + 0.46
2
t(slope)=-5.6, R = 0.25
0.5
0
-6
-4
-2
0
2
4
6
8
10
12
-0.5
-1
-1.5
-2
Real Total Consumption Growth (Last 6 months, annualized)
12
Unemployment Changes (6mo) and
Real S&P 500 Returns: 1959-2004
2
Change in Unemployment Rate (6 Months)
1.5
1
0.5
0
-40
-30
-20
-10
0
10
20
30
40
-0.5
-1
y = -0.01x + 0.013
2
t=-0.7, R = 0.005
-1.5
-2
Real S&P 500 Return (last 6 months)
13
Unemployment Rate vs. Consumption
and Stock Prices 1959-2004 (6 month changes)
Independent
Variable
Slope
t-statistic
CRSQ
S&P 500
-0.004
-0.65
-0.01
Total Real
Consumption
-0.13
-5.56
0.25
NDS
Consumption
-0.15
-4.58
0.18
14
Employment Growth vs. Consumption
and Stock Prices 1959-2004 (6 month changes)
Independent
Variable
Slope
t-statistic
CRSQ
S&P 500
-0.000
-0.00
-0.02
Total Real
Consumption
0.31
3.27
0.18
NDS
Consumption
0.49
3.68
0.22
15
Conclusion on Marginal Utility

Consumption’s percentage changes likely
represent more correctly changes in the marginal
utility of $1 to individuals than do changes in real
stock prices.

This is no failing of stock prices, for as the present
value of future dividends, they should reflect
future profit growth and changing risks and risk
aversion.
16
II. Consumption and Market
Risks of Corporate Cash Flows
Remember the Discounted Cash
Flow Approach to Valuation?

We teach our students to value an asset by
discounting expected cash flows at their proper
risk-adjusted discount rates.

Breeden-Litzenberger (Oct 1978, J. Business)
derived risk-adjusted discount rates in a
multiperiod economy with power utility, jointly
lognormal cash flows. Correct discount rates were
derived as the Consumption-based CAPM.
18
Consumption and Market Risks:
Earnings, Cash Flows vs. Stock Prices

Major problem applying CCAPM to stock prices
is imprecision of consumption beta estimates vs.
very precise market betas.

Breeden paper presented at the French Finance
Ass’n at U. Paris Dauphine on “Capital Budgeting
with Consumption” June 1989 showed opposite
results for earnings risks. Updated slides follow.
Consumption risks are more precisely estimated
for earnings than are market risks, making
CCAPM use natural for capital budgeting. Just in
textbooks, e.g., Elton-Gruber.
19
NIPA U.S Real BT Earnings Growth vs Total Consumption:
Annual Data: 1930-2003 (Excl. 1939-1947)
U.S. Real
Earnings
Growth
60
y = 4.51x - 6.25
40
2
t=9.8, R = 0.61
20
0
-10
-8
-6
-4
-2
0
-20
2
4
6
8
10
-40
-60
Real Total Consumption Growth Per Capita
20
NIPA U.S Real BT Earnings Growth vs NDS Consumption:
Annual Data: 1930-2003 (Excl. 1939-1947)
U.S. Real
Earnings
Growth
60
y = 5.46x - 7.5
40
2
t=8.0, R = 0.51
20
0
-10
-8
-6
-4
-2
0
-20
2
4
6
8
10
-40
-60
Real Nondurables and Services Consumption Growth
21
NIPA U.S Real BT Earnings Growth vs S&P 500 Real
Return: Annual Data: 1930-2003 (Excl. 1939-1947)
U.S. Real
Earnings
Growth
60
y = 0.56x - 1.80
2
t=5.2, R = 0.31
40
20
0
-40
-30
-20
-10
0
-20
10
20
30
40
50
60
-40
-60
S&P 500 Real Return
22
NIPA U.S Real BT Earnings Growth vs Total Consumption:
Postwar Annual Data: 1948-2003
U.S. Real 40
Earnings
Growth 30
20
y = 3.62x - 4.56
2
t= 5.0, R = 0.32
10
0
-2
-1 -10 0
1
2
3
4
5
6
-20
-30
Rl Total Consumption Growth PC
23
NIPA U.S Real BT Earnings Growth vs S&P 500
Real Return: Postwar Annual Data: 1948-2003
U.S. Real
Earnings
Growth
50
y = 0.30x + 1.07
2
t=3.0, R = 0.14
40
30
20
10
0
-30
-20
-10 -10 0
10
20
30
40
50
-20
-30
S&P 500 Real Return
24
NIPA Profits and Cash Flows:
Average RSQ vs. SP500 and
Consumption
0.70
Adj. RSQ
0.60
0.50
0.40
SP500
PCE Total
PCE NDS
0.30
0.20
0.10
0.00
Annual
1930-2003
Annual
1948-2003
Quarterly
1949-2003
25
Conclusion on Cash Flow Risks

As we teach our students and in practice, it is both
more correct and more intuitive to measure cash
flow risks in terms of sensitivity to fluctuations in
aggregate real consumption, rather than in terms
of their relationship to stock market fluctuations.

Of course, if P/E multiples were constant, stock
prices would be proportionate to earnings, and 1%
higher earnings would give 1% higher stock
prices. However, in reality, stocks’ price/earnings
multiples fluctuate also with interest rates,
economic risk and with growth prospects.
26
III. Relative Consumption Betas For
Industries Are Different From
Their Market Betas
Market Betas vs. Consumption
Betas: Estimation Procedure

Stock market betas are estimated from industry
returns data from Professor Kenneth French’s
website, using quarterly data from 1948-2004.

Consumption betas are from NIPA “coarse
industries” quarterly profit data, using 2-quarter
percentage changes in real profits. Actual
calculation is of changes in real profits/employee
compensation vs. real consumption growth,
divided by average profits/employee
compensation.
28
Risks by Industry 1948-2004: Part 1
Market Betas vs. Consumption Betas
Industry
Beta vs
S&P500
Beta vs
TotCons
Relative
C-Beta
Beta vs
NdsCon
Relative
C-Beta
Utilities
0.61
-0.4
-0.10
-1.0
-0.20
Oil
0.76
-1.4
-0.35
1.6
0.40
Food &
Bev’ge
U.S.
Total
Banks
0.84
1.8
0.45
2.9
0.70
1.00
4.0
1.00
4.3
1.00
1.03
1.2
0.30
1.7
0.40
29
Risks By Industry 1948-2004: Part 2
Market Betas vs. Consumption Betas
Industry
Beta vs
SP 500
Beta vs
TotCons
Relative
C-Beta
Beta vs
NdsCon
Relative
C-Beta
Motor
Vehicles
Retail
1.08
24.5
6.10
15.1
3.50
1.09
7.0
1.70
9.2
2.20
Wholesale
1.10
2.8
0.70
3.6
0.90
Durables
1.16
11.2
2.80
10.4
2.40
Construction
1.23
5.2
1.30
8.5
2.00
30
Market Betas vs. Relative Consumption
Betas for Selected Industries
Semiannual data 1948-2004.
Stock returns on stocks, Profits on Consumption.
7
Utilities
Oil
Food&Bev
Aggregate
Banks
Motor Vehicles
Retail
Wholesale
Durables
Construction
6
5
4
3
2
1
0
-1
Market Beta
Ctot Beta
Cnds Beta
31
IV. The Term Structure of Interest Rates
as a Predictor of Economic Growth
Theory: Slope of the Term Structure Optimally
Related to Changes in Real Economic Growth

Breeden’s (1986, JFE) article on “Consumption, Production,
Inflation and Interest Rates: A Synthesis”, generalized Garman
(1976), Rubinstein (1976), Fisher (1907) and Hirshleifer (1970)
and derived and illustrated optimal relations of the term structure
with expected consumption growth and its volatility.

Higher expected consumption growth is consistent with higher
real rates. Higher volatility is consistent with lower real rates.
With forecasted declines in expected growth, the term structure
should have a negative slope. A positive slope should presage an
expected strengthening in economic growth.
33
Tests and Uses of the Term Structure Slope
to Forecast Changes in Economic Growth

Harvey (JFE 1988, 1989,1991,1993) tested Breeden’s
equilibrium model and found it to be quite powerful,
forecasting economic growth better than most professional
economists and working in many countries.

Reflecting this research, in 1996, the slope of the term
structure was added as a variable in the Index for Leading
Economic Indicators.

Dotsey (1998) of the Federal Reserve Bank of Richmond
found a negative term structure slope gave 18 true
quarterly signals and only 2 false signals of recession in
quarterly growth during the 1955-1995 period.
34
US Yield Curve Inverts Before Last Six US Recessions
(5-year US Treasury bond - 3-month US Treasury bill)
Source: Campbell R. Harvey, Professor, Duke University
Annual
GDP growth
or Yield Curve
8
% Real annual GDP growth
6
4
2
Yield curve slope
0
Yield curve accurate
in recent forecast
1
9
M
ar
-0
7
M
ar
-9
5
M
ar
-9
3
M
ar
-9
1
M
ar
-9
M
ar
-9
9
7
M
ar
-8
M
ar
-8
5
M
ar
-8
M
ar
-8
3
Data though 1/12/03
1
9
Recession
Correct
M
ar
-8
M
ar
-7
7
5
M
ar
-7
M
ar
-7
3
1
M
ar
-7
M
ar
-7
M
ar
-6
9
-2
Recession
Correct
-4
Recession
Correct 2 Recessions
Correct
-6
35
Slope of the Term Structure Predicts
Real Consumption Growth 1959-2004
400
y = 11.58x + 61.7
2
R = 0.11, t(slope)=3.93
300
3 Yr-3Month Treasury Yields
200
100
0
-6
-4
-2
0
2
4
6
8
10
12
-100
-200
-300
Real Total Consumption Growth (Next 6 months, annualized)
36
Forecasts of Growth from Term Spread
Quarterly Data, 1953 : 1978
Horizon
1 qrt
2 qrts
4 qrts
GDP
PCE Total
PCE NDS
Slope
0.51
0.37
0.18
Robust t-stat
3.82
3.21
2.16
adj. RSQ
0.09
0.09
0.05
Slope
1.08
0.78
0.39
Robust t-stat
4.36
3.89
2.80
adj. RSQ
0.16
0.16
0.10
Slope
1.90
1.13
0.52
Robust t-stat
4.87
0.22
3.45
0.13
2.38
0.07
adj. RSQ
37
Forecasts of Growth from Term Spread
Quarterly Data, 1979 : 2004
Horizon
1 qrt
2 qrts
4 qrts
GDP
PCE Total
PCE NDS
Slope
0.29
0.19
0.12
Robust t-stat
4.43
3.46
3.59
adj. RSQ
0.23
0.14
0.13
Slope
0.52
0.36
0.23
Robust t-stat
4.11
3.83
3.87
adj. RSQ
0.26
0.20
0.18
Slope
0.94
0.66
0.41
Robust t-stat
4.00
0.30
3.81
0.23
3.61
0.20
adj. RSQ
38
Global Term Structure Slopes: 10 Year-3 Month
Ends of Years 1989-2003
400
300
200
100
0
1989
1991
1993
1995
1997
1999
2001
2003
U.K.
U.S.A.
Japan
Germany
-100
-200
-300
39
V. Consumption Risk and Returns and
the Maximum Correlation Portfolio
Maximum Correlation Portfolio Elements
S&P 500, Baa-Treasury Bonds, Term Spread

Breeden, Gibbons and Litzenberger (JF, 1989)
proved that the CCAPM also holds with regard to
betas measured against the maximum correlation
portfolio for consumption.

Three broad traded markets are for (1) stocks, (2)
Government bonds and (3) corporate bonds.
Consumption relates to each of these through
effects of wealth, the term structure, and relation
of credit risk to the economic cycle, respectively.
41
Maximum Correlation Portfolio
Semiannual Data (Dec-Jun) 1960-2004
S&P 500
PCETot
Slope
PCETot
t-stat
PCENDS
PCENDS
Slope
t-stat
.0685
3.30
.0549
3.55
-3.06
-0.85
-2.27
2.74
0.39
2.03
dBaa -10 -1.54
Yr Treas
(bp sprd)
Lagged
0.70
3Yr-3Mo
TS Slope
RSQ=.29
RSQ=.24
42
“Consumption Risk and the Cross-Section of
Expected Returns”, Parker-Julliard (JPE, 2005)
Consumption betas measured by contemporaneous
covariances of assets’ returns and consumption growth,
fail to explain the dispersion in risk premiums across
assets.

Parker-Julliard measure ultimate consumption risk as
covariation of return with current and future
consumption growth. Ultimate consumption betas,
therefore, measure the exposure of asset returns to longrun risks in consumption.

Consumption Risk and Expected Returns
Parker-Julliard (JPE, 2005) continued
Parker-Julliard
argue that ultimate consumption risk
measures are much more robust to measurement errors
in consumption, as well as slow and costly adjustments
of consumption to returns.
Using post-war quarterly data on 25 Fama-French
portfolios sorted by Book Equity/Market Equity and
Size, they show ultimate consumption risk measures are
able to explain from 44% to 73% of the variation in
expected returns.

VI. Consumption Deviations from Wealth
as a Predictor of Income and Investment
Opportunities
Consumption Deviations from Wealth Predict
Income and Investment Opportunities

Breeden (1984, JET) showed with relative risk aversion
>1, investors’ consumption levels are positively related
to income and investment opportunities. (If RRA<1,
reverse hedging occurs.)

In June 1989, at the French Finance Association in Paris
and in 1991 at Harvard, Breeden paper on “Capital
Budgeting with Consumption”, argued that as
consumption optimally is a function of wealth, income
and investment opportunities, consumption fluctuations
orthogonalized for wealth effects should be indicators
of the income and investment opportunity set. Test
results presented then are updated as follows:
46
Consumption Growth Predicted by Stock Returns
Quarterly Data, 1949 – 2003
Dependent
Variable
Slope Coefficients
Current
Lag1
Lag2
0.0099
0.0236
0.0233
t-stat
1.34
3.19
3.16
Robust t-stat
1.75
2.98
4.33
PCE Total
Dependent
Variable
PCE NDS
t-stat
Robust t-stat
Slope Coefficients
Current
Lag1
Lag2
0.0087
0.0143
0.0089
2.05
3.35
2.08
2.54
3.58
2.66
Adj.
RSQ
0.09
Adj.
RSQ
0.08
47
Consumption Growth Deviations and the
Income and Investment Opportunity Set

The lagged values of the residuals from the
above regressions are examined for predictive
ability with regard to income, wages and
corporate profits.

Specifically, we regress the growth rate of each
variable on its own lag and the lagged
consumption residuals.
48
Consumption Deviations Predict Real
Personal Income Growth: 1989 Results

Results in 1989 Breeden paper (Quarterly 1950-1988):

Personal Income Growth (t)=
= .0054+ .36 PI(t-1) + .31(PCETotal Residual)
(t=4.84)
(t=3.44)
RSQ=.27
= .0054 +.35 PI(t-1) + .47 (PCE NDS Residual)
(t=4.67)
(t=3.52)
RSQ=.27
49
Consumption Deviations Predict Real
Personal Income Growth: 2004 Results

Updating tests quarterly from 1949-2003:

Personal Income Growth (t)=
=
-.125 PI(t-1) + .33(PCETotal Residual)
(t=-1.38)
(t=3.89)
Adj RSQ=.06
=
-.133 PI(t-1) + .56 (PCE NDS Residual)
(t=-1.52) (t=4.90)
Adj RSQ=.05
50
Consumption Deviations Predict
Real Wage Growth

Using quarterly data from 1949-2003:

Real Wage Growth (t)=
=
0.29 RW(t-1) + .25 (PCETotal Residual)
(t=1.93)
(t=3.22)
Adj RSQ=.15
=
0.26 RW(t-1) + .58 (PCE NDS Residual)
(t=1.72) (t=4.15)
Adj RSQ=.15
51
Consumption Deviations Predict
Unemployment Rate Changes

Using semiannual data from 1949-2003:

6-Mo. Change in Unemployment Rate (t)=
=
=
-0.21 (PCETotal Residual)
(t=3.62)
Adj RSQ=.10
-0.26 (PCE NDS Residual)
(t=3.17)
Adj RSQ=.08
52
Consumption Deviations Predict
Real GDP Growth

Using quarterly data from 1949-2003:

Real GDP Growth (t)=
=
0.26 GDP(t-1) + .16 (PCETotal Residual)
(t=3.26)
(t=1.54)
Adj RSQ=.11
=
0.23 GDP(t-1) + .55 (PCE NDS Residual)
(t=3.22) (t=3.41)
Adj RSQ=.16
53
Consumption Deviations from Wealth
and the Investment Opportunity Set

Positive consumption deviations precede (1949-2003 Q):
Lower average stock returns and
Lower volatility (squared returns)
Higher corporate profits

Correlations of Lagged Consumption Residuals:
PCETot t
PCE NDS t
S&P 500 Real Return (Mean) -0.10 -1.5
-0.14
-2.1
S&P 500 Squared Return
-0.15 -2.2
-0.18
-2.7
S&P 500 Earnings/Share
0.29 4.5
0.23
3.5
54
Conclusion on Consumption and the
Income and Investment Opportunity Sets

Test results show strongly that, as consumption
and portfolio theory predict, consumption choices
do reflect knowledge about future income and
investment opportunities.

High consumption relative to wealth is followed
by high wage and personal income growth, and by
higher corporate profits and reduced volatility of
investment returns. Low consumption/wealth
reflects weak income opportunities, lower profits
and higher risks. Lettau-Ludwigson’s (2001a,b)
results confirm some of these effects.
55
V.
Volatility of Individual Consumption
vs. Aggregate Consumption and the
Equity Premium Puzzle
See also: Heaton-Lucas (1996 JPE) and
Brav, Constantinides, Geczy (2002 JPE)
Volatility of Family Consumption: Consumer
Expenditure Survey,1980 Q1-1998Q2
Vol(NIPA Total Consumption)=1.6%. Vol(Sample Aggregate=6.9%)
Low (1/3)
Income
Middle (1/3)
Income
High (1/3)
Income
11.0%
15.4%
19.1%
Medium (2- 8.9%
4 persons)
9.3%
14.2%
Large (>4
persons)
8.5%
14.6%
Small (<=2
persons)
8.9%
57
Equity Premium Puzzle and
Consumption Volatility

Individual consumption volatility is 5-10 times
larger than measured volatility of NIPA aggregate
consumption. High income individuals own the
most assets, have highest consumption volatility.

Reasons include incomplete markets and optimal
incentives for labor choices.

Marginal utility is indeed related to that
consumption volatility and likely helps explain the
equilibrium equity risk premium.
58