Consumption-Based Asset Pricing

Guanglian Hu

S1, 2020

Guanglian Hu S1 2020 S1 2020 1 / 26

Overview

Consumption-Based Asset Pricing: Theory

Equity Risk Premium Puzzle

Risk-Free Rate Puzzle

Hansen-Jagannathan Bound

Guanglian Hu S1 2020 S1 2020 2 / 26

Consumption Based Asset Pricing

Consider the intertemporal choice problem of an investor, who can

trade freely in some asset i and try to maximize the expectation of a
time-separable utility function
∞
X
max Et [ β j U (Ct +j )]
j =0

where β is the time discount factor.

First order condition or Euler equation describing the investor’s

optimal consumption and and portfolio plan is:

U 0 (Ct ) = Et [βU 0 (Ct +1 )(1 + Ri,t +1 )],

where 1 + Ri,t +1 is the gross simple rate of return on the asset held
from time t to time t + 1.
Guanglian Hu S1 2020 S1 2020 3 / 26
Consumption Based Asset Pricing

Consumption-based asset pricing starts from the Euler equation

U 0 (Ct ) = Et [βU 0 (Ct +1 )(1 + Ri,t +1 )] (1)

The left hand side of equation (1) is the marginal utility cost of
consuming one dollar less at time t.

The right hand side is the expected marginal utility benefit from
investing the dollar in asset i at time t, selling it at time t + 1 for
1 + Ri,t +1 dollars, and consuming the proceeds.

The marginal cost equals the marginal benefit.

Guanglian Hu S1 2020 S1 2020 4 / 26

Consumption Based Asset Pricing

Dividing equation (1) by U 0 (Ct ) yields:

U 0 (Ct +1 )
1 = Et [ β (1 + Ri,t +1 )]
U 0 (Ct )

U 0 ( Ct + 1 )
Denote Mt +1 = β U 0 ( Ct )
, we have:

1 = Et [Mt +1 (1 + Ri,t +1 )]

Mt +1 is the intertemporal marginal rate of substitution of the investor, also

known as the stochastic discount factor (SDF), pricing kernel, or simply
marginal utility.
Mt +1 is a random variable and always positive. High Mt +1 means low
consumption.
marginal utility is high when consumption Ct +1 is low.

Guanglian Hu S1 2020 S1 2020 5 / 26

Consumption Based Asset Pricing

The derivation for 1 = Et [Mt +1 (1 + Ri,t +1 )] assumes the existence of

an investor maximizing a time-separable utility function, but in fact
the equation holds more generally.

The existence of a positive stochastic discount factor is guaranteed by

the absence of arbitrage opportunities in markets.

Different asset pricing models use different Mt +1 . Think of Mt +1 as

an index for bad times.

Guanglian Hu S1 2020 S1 2020 6 / 26

Consumption Based Asset Pricing

We just derived the fundamental equation in asset pricing:

1 = Et [Mt +1 (1 + Ri,t +1 )] (2)

One can manipulate this equation to get a pricing formula. Note that
Xi,t +1
1 + Ri,t +1 =
Pi,t

where Xi,t +1 is the payoff of the asset at time t + 1.

This implies:
Xi,t +1
1 = Et [ M t + 1 ] ⇒ Pi,t = Et [Mt +1 Xi,t +1 ]
Pi,t

The value of an asset is equal to the expected value of discounted

future payoffs (by Mt +1 ).

Guanglian Hu S1 2020 S1 2020 7 / 26

Consumption Based Asset Pricing
Explore the implications of the fundamental equation in the return
space.
Expanding the expectation,
Et [Mt +1 (1 + Ri,t +1 )] = Et [Mt +1 ]Et [(1 + Ri,t +1 )] + Covt [Ri,t +1 , Mt +1 ]
Substituting into (2) yields:
1 = Et [Mt +1 ]Et [(1 + Ri,t +1 )] + Covt [Ri,t +1 , Mt +1 ]
Rearranging gives:
1 − Covt [Ri,t +1 , Mt +1 ]
Et [(1 + Ri,t +1 )] = (3)
Et [ M t + 1 ]
An asset with a high expected return must have a low covariance with
the stochastic discount factor. So if M covaries negatively with R,
then the return is low in bad times. It is a risky asset to have and
should offer higher expected return.
Guanglian Hu S1 2020 S1 2020 8 / 26
Consumption Based Asset Pricing

Equation (3) must hold for any asset, including a riskless asset whose
return has zero covariance with the stochastic discount factor (or any
other random variable).

1
1 + Rf ,t +1 = (4)
Et [Mt +1 ]

The above equation can be used to rewrite equation (3):

Et [(1 + Ri,t +1 )] = (1 + Rf ,t +1 )(1 − Covt [Ri,t +1 , Mt +1 ]) (5)

An asset with a negative (positive) covariance with the stochastic

discount factor will earn expected rates of return that are higher
(lower) than the risk-free rate

Guanglian Hu S1 2020 S1 2020 9 / 26

Consumption Based Asset Pricing

Assume that the joint conditional distribution of asset returns and the
stochastic discount factor is lognormal and homoskedastic
When a random variable X is conditionally lognormally distributed, it has
the following convenient property that

1
log Et X = Et log X + Vart log X
2
With joint conditional lognormality and homoskedasticity of asset returns
and consumption, we can take logs of equation (2):

1
0 = Et ri,t +1 + Et mt +1 + (σi2 + σm
2
+ 2σi,m ) (6)
2
where mt +1 = log(Mt +1 ), ri,t +1 = log(1 + Ri,t +1 ), σi2 is the unconditional
2 is the unconditional variance of the stochastic
variance of log returns, σm
discount factor, and σi,m is the unconditional covariance between returns
and stochastic discount factor

Guanglian Hu S1 2020 S1 2020 10 / 26

Consumption Based Asset Pricing

Consider the log return on the risk-free asset:

2
σm
rf ,t +1 = −Et mt +1 − (7)
2
which is the log counterpart of equation (4)
Subtracting equation (7) from equation (6) yields an expression for the
expected excess return on risky assets over the riskless rate:

σi2
Et [ri,t +1 ] − rf ,t +1 + = −σi,m (8)
2
which is the log counterpart of equation (5)
The right hand side of equation (8) says that the risk premium is determined
by the negative of the covariance of the asset with the stochastic discount
factor.
An asset with a negative covariance with the stochastic discount factor must
have larger risk premium

Guanglian Hu S1 2020 S1 2020 11 / 26

Jensen’s Inequality

σi2
Et [ri,t +1 ] − rf ,t +1 + = −σi,m
2

σi2
What is with the 2 term?
σi2
2 arises from the fact that we are describing expectations of log return.

σi2
log Et [1 + Ri,t +1 ] = Et [ri,t +1 ] +
2
Two ways to write equity risk premium equation:

σi2
Et [ri,t +1 ] − rf ,t +1 + = −σi,m
2
log Et [1 + Ri,t +1 ] − log(1 + Rf ,t +1 ) = −σi,m

Guanglian Hu S1 2020 S1 2020 12 / 26

International Stock Market Data

Guanglian Hu S1 2020 S1 2020 13 / 26

International Consumption and Dividend Data

Guanglian Hu S1 2020 S1 2020 14 / 26

Data Summary

Stock markets have delivered high average returns with high standard
deviations

Short term debt (risk-free asset) has delivered low returns with low
standard deviation

Consumption growth is smooth with low standard deviation

Guanglian Hu S1 2020 S1 2020 15 / 26

Consumption Based Asset Pricing with Power Utility
Assume that there is a representative agent who maximizes a time-separable
power utility function defined over aggregate consumption Ct :

Ct1−γ − 1
U (Ct ) =
1−γ
where γ is the coefficient of relative risk aversion. With constant relative risk
aversion(CRRA), fraction allocated to risky asset is independent of wealth.

Power utility implies:

U 0 (Ct +1 ) Ct +1 −γ
U 0 (Ct ) = Ct−γ ; Mt +1 = β = β( ) ;
U 0 ( Ct ) Ct

The log of the stochastic discount factor is

mt +1 = log β − γ∆ct +1

where ∆ct +1 = ct +1 − ct and ct = log(Ct ).

Guanglian Hu S1 2020 S1 2020 16 / 26
Consumption Based Asset Pricing with Power Utility

Recall from equation (6)

1
0 = Et ri,t +1 + Et mt +1 + (σi2 + σm
2
+ 2σi,m )
2

With power utility, mt +1 = log β − γ∆ct +1 and the above equation

becomes:
1
0 = Et ri,t +1 + log β − γEt ∆ct +1 + (σi2 + γ 2 σc2 − 2γσi,c )
2

σc2 is the unconditional variance of log consumption growth (∆ct +1 ),

and σi,c is the unconditional covariance between returns and
consumption growth Cov (ri,t +1 , ∆ct +1 )

Guanglian Hu S1 2020 S1 2020 17 / 26

Consumption Based Asset Pricing with Power Utility

The risk-free rate in equation (7) now becomes:

γ 2 σc2
rf ,t +1 = − log β + γEt ∆ct +1 − (9)
2

Risk-free rate is high when β is low (investors are impatient)

Risk-free rate is linear in expected consumption growth, with slope

coefficient equal to the coefficient of relative risk aversion.

The conditional variance of consumption growth has a negative effect

on the riskfree rate which can be interpreted as a precautionary
savings effect.

Guanglian Hu S1 2020 S1 2020 18 / 26

Consumption Based Asset Pricing with Power Utility

The equity risk premium in equation (8) now becomes:

σi2
Et [ri,t +1 ] − rf ,t +1 + = γσi,c (10)
2

Equation (10) says that the risk premium on any asset is the
coefficient of relative risk aversion times the covariance of the asset
return with consumption growth

Intuitively, an asset with a high consumption covariance tends to have

low returns when consumption is low, that is, when the marginal
utility of consumption is high. Such an asset is risky and commands a
large risk premium.

Guanglian Hu S1 2020 S1 2020 19 / 26

Equity Risk Premium Puzzle

Note: first row RRA(1)= 0.08071/0.0003354. Power utility model can only fit the
equity premium if the coefficient of relative risk aversion is very large.

Guanglian Hu S1 2020 S1 2020 20 / 26

Risk Aversion

Consider the following gamble: I flip a coin and ...

If it comes up heads, I multiply your lifetime income by 1 million
If it comes up tails, I reduce your lifetime income by 10%

If you accept, your γ is less than 10

What about 20% reduction? If yes, γ < 5

What about 30% reduction? If yes, γ < 3

Many economists believe risk aversion γ should be less than 10.

Guanglian Hu S1 2020 S1 2020 21 / 26

Risk-Free Rate Puzzle

One response to the equity premium puzzle is to consider larger

values for the coefficient of relative risk aversion.

Recall equation (9) shows that the riskless interest rate is

γ 2 σc2
rf ,t +1 = − log β + γEt ∆ct +1 −
2

High values of γ would imply high values of γEt ∆ct +1 . This can be
reconciled with low interest rates only if the time discount factor β is
close to or even greater than one, corresponding to a low or even
negative rate of time preference.

Guanglian Hu S1 2020 S1 2020 22 / 26

Hansen-Jagannathan Bound

Recall from equation (8) the expected excess return on any risky asset is
given by
σ2
Et [ri,t +1 ] − rf ,t +1 + i = −σi,m
2
Expanding the covariance term,

σi,m = σi σm ρi,m

where ρi,m is the correlation between asset return and stochastic discount
factor.
Since ρi,m ≥ −1, −σi σm ρi,m ≤ σi σm , which then implies

−σi,m ≤ σi σm .

Guanglian Hu S1 2020 S1 2020 23 / 26

Hansen-Jagannathan Bound

A little manipulation yields the Hansen-Jagannathan Bound

σi2
Et [ri,t +1 ] − rf ,t +1 + 2
σm ≥ (11)
σi
Equation (11) says that the standard deviation of the log stochastic
discount factor must be greater than the Sharpe ratio for any
arbitrary asset i, that is, it must be greater than the maximum
possible Sharpe ratio obtainable in asset markets.

Guanglian Hu S1 2020 S1 2020 24 / 26

Hansen-Jagannathan Bound

The Sharpe ratio of the market portfolio is approximately 50% on an annual

basis.

This means that the standard deviation of the stochastic discount factor
must be equal to or greater than 50%.

Recall with power utility, the (log) of the pricing kernel is given by

mt +1 = log β − γ∆ct +1 ⇒ σ (m) = γσ (∆ct +1 )

In the data, σ (∆ct +1 ) is about 1%, which means we need a very large risk
aversion coefficient.

Guanglian Hu S1 2020 S1 2020 25 / 26

Revisiting Equity Risk Premium Puzzle

Another way to look at the equity risk premium puzzle is that the stochastic discount
factor implied from the standard consumption-based asset pricing models is not volatile
enough.

Guanglian Hu S1 2020 S1 2020 26 / 26