Valuation Risk and Asset Pricing∗

Rui Albuquerque†, Martin Eichenbaum‡, Victor Luo,§and Sergio Rebelo¶

December 2015

Abstract
Standard representative-agent models fail to account for the weak correlation be-
tween stock returns and measurable fundamentals, such as consumption and output
growth. This failing, which underlies virtually all modern asset-pricing puzzles, arises
because these models load all uncertainty onto the supply side of the economy. We
propose a simple theory of asset pricing in which demand shocks play a central role.
These shocks give rise to valuation risk that allows the model to account for key asset
pricing moments, such as the equity premium, the bond term premium, and the weak
correlation between stock returns and fundamentals.

J.E.L. Classification: G12.

Keywords: Equity premium, bond yields, risk premium.

∗
We benefited from the comments and suggestions of Fernando Alvarez, Ravi Bansal, Frederico Belo,
Jaroslav Borovička, John Campbell, John Cochrane, Lars Hansen, Anisha Ghosh, Ravi Jaganathan, Tasos
Karantounias, Howard Kung, Junghoon Lee, Dmitry Livdan, Jonathan Parker, Alberto Rossi, Costis Ski-
adas, Ivan Werning, and Amir Yaron. We thank Robert Barro, Emi Nakamura, Jón Steinsson, and
José Ursua for sharing their data with us and Benjamin Johannsen for superb research assistance. Al-
buquerque gratefully acknowledges financial support from the European Union Seventh Framework Pro-
gramme (FP7/2007-2013) under grant agreement PCOFUND-GA-2009-246542. A previous version of this
paper was presented under the title “Understanding the Equity Premium Puzzle and the Correlation Puzzle,”
http://tinyurl.com/akfmvxb.
†
Boston College, CEPR, and ECGI.
‡
Northwestern University, NBER, and Federal Reserve Bank of Chicago.
§
Northwestern University.
¶
Northwestern University, NBER, and CEPR.
1. Introduction

In standard representative-agent asset-pricing models, the expected return to an asset re-

flects the covariance between the asset’s payo§ and the agent’s stochastic discount factor.
An important challenge to these models is that the correlation and covariance between
stock returns and measurable fundamentals, especially consumption growth, is weak at both
short and long horizons. Cochrane and Hansen (1992), Campbell and Cochrane (1999), and
Cochrane (2001) call this phenomenon the correlation puzzle. More recently, Lettau and Lud-
vigson (2011) document this puzzle using di§erent methods. According to their estimates,
the shock that accounts for the vast majority of asset-price fluctuations is uncorrelated with
consumption at virtually all horizons.
The basic disconnect between measurable macroeconomic fundamentals and stock re-
turns underlies virtually all modern asset-pricing puzzles, including the equity-premium
puzzle, Hansen-Singleton (1982)-style rejection of asset-pricing models, violation of Hansen-
Jagannathan (1991) bounds, and Shiller (1981)-style observations about excess stock-price
volatility.
A central finding of modern empirical finance is that variation in asset returns is over-
whelmingly due to variation in discount factors (see Campbell and Ammer (1993) and
Cochrane (2011)). A key question is: how should we model this variation? In classic
asset-pricing models, all uncertainty is loaded onto the supply side of the economy. In Lucas
(1978) tree models, agents are exposed to random endowment shocks, while in production
economies they are exposed to random productivity shocks. Both classes of models abstract
from shocks to the demand for assets. Not surprisingly, it is very di¢cult for these models
to simultaneously resolve the equity premium puzzle and the correlation puzzle.
We propose a simple theory of asset pricing in which demand shocks, arising from sto-
chastic changes in agents’ rate of time preference, play a central role in the determination
of asset prices. These shocks amount to a parsimonious way of modeling the variation in
discount rates stressed by Campbell and Ammer (1993) and Cochrane (2011). Our model
implies that the law of motion for these shocks plays a first-order role in determining the
equilibrium behavior of variables like the price-dividend ratio, equity returns and bond yields.
So, our analysis is disciplined by the fact that the law of motion for time-preference shocks
must be consistent with the time-series properties of these variables.
In our model, the representative agent has recursive preferences of the type considered by

1
Kreps and Porteus (1978), Weil (1989), and Epstein and Zin (1991). When the risk-aversion
coe¢cient is equal to the inverse of the elasticity of intertemporal substitution, recursive
preferences reduce to constant-relative risk aversion (CRRA) preferences. We show that, in
this case, time-preference shocks have negligible e§ects on key asset-pricing moments such
as the equity premium.
We consider two versions of our model. The benchmark model is designed to highlight the
role played by time-preference shocks per se. Here consumption and dividends are modeled
as random walks with conditionally homoscedastic shocks. While this model is very useful
for expositional purposes, it su§ers from some clear empirical shortcomings, e.g. the equity
premium is constant. For this reason, we consider an extended version of the model in which
the shocks to the consumption and dividend process are conditionally heteroskedastic. We
find that adding these features improves the performance of the model.1
We estimate our model using a Generalized Method of Moments (GMM) strategy, im-
plemented with annual data for the period 1929 to 2011. We assume that agents make
decisions on a monthly basis. We then deduce the model’s implications for annual data, i.e.
we explicitly deal with the temporal aggregation problem.2
It turns out that, for a large set of parameter values, our model implies that the GMM esti-
mators su§er from substantial small-sample bias. This bias is particularly large for moments
characterizing the predictability of excess returns and the decomposition of the variance of
the price-dividend ratio proposed by Cochrane (1992). In light of this fact, we modify the
GMM procedure to focus on the plim of the model-implied small-sample moments rather
than the plim of the moments themselves. This modification makes an important di§erence
in assessing the model’s empirical performance.
We show that time-preference shocks help explain the equity premium as long as the risk-
aversion coe¢cient and the elasticity of intertemporal substitution are either both greater
than one or both smaller than one. This condition is satisfied in the estimated benchmark
and extended models.
Allowing for sampling uncertainty, our model accounts for the equity premium and the
volatility of stock and bond returns, even though the estimated degree of agents’ risk aversion
is very moderate (roughly 1.5). Critically, the extended model also accounts for the mean,
1
These results parallel the findings of Bansal and Yaron (2004) who show that allowing for conditional
heteroskedasticity improves the performance of long-run risk models.
2
Bansal, Kiku, and Yaron (2013) pursue a similar strategy in estimating a long-run risk model. They
estimate the frequency with which agents make decisions and find that it is roughly equal to one month.

2
variance and persistence of the price-dividend ratio and the risk-free rate. In addition, it
accounts for the correlation between stock returns and fundamentals such as consumption,
output, and dividend growth at short, medium and long horizons. The model also accounts
for the observed predictability of excess returns by lagged price-dividend ratios.
We define valuation risk as the part of the excess return to an asset that is due to
the volatility of the time-preference shock. According to our estimates, valuation risk is a
much more important determinant of asset returns than conventional risk. Valuation risk is
an increasing function of an asset’s maturity, so a natural test of our model is whether it
can account for the bond term premia. We show that the model does a very good job at
accounting for the level and slope of the nominal yield curve, as well as the standard deviation
of nominal yields. The upward sloping nature of the nominal yield curve reflects the fact
that our model predicts an upward sloping yield curve for ex-ante real yields on nominal
bonds. In fact, in our model the slope of the yield curve is entirely driven by valuation risk.
The last result contrasts sharply with leading alternatives, such as the long-run risk model
which generally predicts a downward-sloping real yield curve. With this distinction in mind,
we discuss evidence on the slope of the real yield curve obtained using data for the U.S. and
the U.K. We also show that the estimated nominal SDF for our model has the properties
that Backus and Zin (1994) argue are necessary to simultaneously account for the slope of
the yield curve and the persistence of bond yields.
There is a literature that models shocks to the demand for assets as arising from time-
preference or taste shocks. For example, Garber and King (1983) and Campbell (1986)
consider these types of shocks in early work on asset pricing. Stockman and Tesar (1995),
Pavlova and Rigobon (2007), and Gabaix and Maggiori (2013) study the role of taste shocks
in explaining asset prices in open economy models. In the macroeconomic literature, Eg-
gertsson and Woodford (2003) and Eggertsson (2004), model changes in savings behavior
as arising from time-preference shocks that make the zero lower bound on nominal interest
rates binding.3 Hall (2014) stresses the importance of variation in discount rates in explain-
ing the cyclical behavior of unemployment. Bai, Rios-Rull and Storesletten (2014) stress the
importance of demand shocks in generating business cycle fluctuations.
Time-preference shocks can also be thought of a simple way of capturing the notion that
fluctuations in market sentiment contribute to the volatility of asset prices, as emphasized
3
See also Huo and Rios-Rull (2013), Correia, Farhi, Nicolini, and Teles (2013), and Fernandez-Villaverde,
Guerron-Quintana, Kuester, Rubio-Ramírez (2013).

3
by authors such as in Barberis, Shleifer, and Vishny (1998) and Dumas, Kurshev and Up-
pal (2009). Finally, in independent work, contemporaneous with our own, Maurer (2012)
explores the impact of time-preference shocks in a calibrated continuous-time representative
agent model with Du¢e-Epstein (1992) preferences.4
Our paper is organized as follows. In Section 2 we document the correlation puzzle using
U.S. data for the period 1929 to 2011 and the period 1871 to 2006. In Section 3, we present
our benchmark and extended models. We discuss our estimation strategy in Section 4. In
Section 5, we present our empirical results. In Section 6, we study the empirical implications
of the model for bond term premia, as well as the return on stocks relative to long-term
bonds. Section 7 concludes.

2. The correlation puzzle

In this section we examine the correlation between stock returns and fundamentals as mea-
sured by the growth rate of consumption, output, dividends, and earnings.

2.1. Data sources

We consider two sample periods: 1929 to 2011 and 1871 to 2006. For the first sample, we
obtain nominal stock and bond returns from Kenneth French’s website. We use the measure
of real consumption expenditures and real Gross Domestic Product constructed by Barro and
Ursúa (2011), which we update to 2011 using National Income and Product Accounts data.
We compute per-capita variables using total population (POP).5 We obtain data on real
S&P500 earnings and dividends from Robert Shiller’s website. We use data from Ibbotson
and Associates on the nominal return to one-month Treasury bills, the nominal yield on
4
Normandin and St-Amour (1998) study the impact of preference shocks in a model similar to ours.
Unfortunately, their analysis does not take into account the fact that covariances between asset returns,
consumption growth, and preferences shocks depend on the parameters governing preferences and technology.
As a result, their empirical estimates imply that preference shocks reduce the equity premium. In addition,
they argue that they can explain the equity premium with separable preferences and preference shocks. This
claim contradicts the results in Campbell (1986) and the theorem in our Appendix B.
5
This series is not subject to a very important source of measurement error that a§ects another commonly-
used population measure, civilian noninstitutional population (CNP16OV). Every ten years, the CNP16OV
series is adjusted using information from the decennial census. This adjustment produces large discontinuities
in the CNP16OV series. The average annual growth rates implied by the two series are reasonably similar:
1.2 for POP and 1.4 for CNP16OV for the period 1952-2012. But the growth rate of CNP16OV is three
times more volatile than the growth rate of POP. Part of this high volatility in the growth rate of CNP16OV
is induced by large positive and negative spikes that generally occur in January. For example, in January
2000, 2004, 2008, and 2012 the annualized percentage growth rates of CNP16OV are 14.8, −1.9, −2.8, and
8.4, respectively. The corresponding annualized percentage growth rates for POP are 1.1, 0.8, 0.9, and 0.7.

4
intermediate-term government bonds (with approximate maturity of five years), and the
nominal yield on long-term government bonds (with approximate maturity of twenty years).
We convert nominal returns and yields to real returns and yields using the rate of inflation
as measured by the consumer price index. We also use inflation-adjusted U.K. government
bonds (index-linked gilts) with maturities 2.5, 5, 10, 15, and 25 years.
For the second sample, we use data on real stock and bond returns from Nakamura,
Steinsson, Barro, and Ursúa (2013). We use the same data sources for consumption, expen-
ditures, dividends and earnings as in the first sample.
In our estimation, we use two measures of real bond returns. Our primary measure
is the ex-ante real return on nominal, one-year, five-year and twenty-year Treasury bonds
taken from Luo (2014). Luo (2014) constructs alternative models of expected inflation for
one, five and twenty-year horizons. He argues that the random walk model does a better
job at forecasting one-year inflation than more sophisticated models like time-varying VAR
methods of the sort considered by Primiceri (2005). It also does better than Bayesian vector
autoregressions embodying Minnesota priors. However, he argues that the latter does best
at forecasting inflation at the five- and twenty-year horizons.
We also report results for a second measure of real bond returns: the realized real return
on nominal one-year, five-year and twenty-year Treasury bonds. The one-year rate is the
measure of the risk-free rate used in Mehra and Prescott (1985) and the associated literature.
We compute dynamic correlations between stock returns and consumption growth for the
G7 and OECD countries. The data on annual real stock returns for the G7 and the OECD
countries are from Global Financial Statistics. The data on real per capita personal consumer
expenditures and real per capita Gross Domestic Product (GDP) for these countries comes
from Barro and Ursúa (2008) and was updated by these authors until 2009. The countries
(beginning of sample period) included in our data set are: Australia (1901), Austria (1947),
Belgium (1947), Canada (1900), Chile (1900), Denmark (1900), Finland (1923), France
(1942), Germany (1851), Greece (1953), Italy (1900), Japan (1894), Korea (1963), Mexico
(1902), Netherlands (1947), New Zealand (1947), Norway (1915), Spain (1941), Sweden
(1900), Switzerland (1900), United Kingdom (1830), and United States (1869).

2.2. Empirical results

Table 1, panel A presents results for the sample period 1929 to 2011. We report correlations
at the one-, five- and ten-year horizons. The five- and ten-year horizon correlations are

5
computed using five- and ten-year overlapping observations, respectively. We report Newey-
West (1987) heteroskedasticity-consistent standard errors computed with ten lags.
There are three key features of Table 1, panel A. First, consistent with Cochrane and
Hansen (1992) and Campbell and Cochrane (1999), the growth rates of consumption and
output are uncorrelated with stock returns at all the horizons that we consider. Second, the
correlation between stock returns and dividend growth is similar to that of consumption and
output growth at the one-year horizon. However, the correlation between stock returns and
dividend growth is substantially higher at the five and ten-year horizons than the analogue
correlations involving consumption and output growth. Third, the pattern of correlations
between stock returns and dividend growth is similar to the analogue correlations involving
earnings growth.
Table 1, panel B reports results for the longer sample period (1871-2006). The one-
year correlation between stock returns and the growth rates of consumption and output
are very similar to those obtained for the shorter sample. There is evidence in this sample
of a stronger correlation between stock returns and the growth rates of consumption and
output at a five-year horizon. But, at the ten-year horizon the correlations are, once again,
statistically insignificant. The results for dividends and earnings are very similar across the
two subsamples.
Table 2 assesses the robustness of our results for the correlation between stock returns
and consumption using three di§erent measures of consumption for the period 1929 to 2011,
obtained from the National Product and Income Accounts. With one exception, the corre-
lations in this table are statistically insignificant. The exception is the five-year correlation
between stock returns and the growth rate of nondurables and services which is marginally
significant.
Table 3 reports the correlation between stock returns and the growth rate of consumption
and output for the G7, and the OECD. We report correlations at the one-, five- and ten-
year horizons. We compute correlations for the G7 and the OECD countries pooling data
across all countries. This procedure implies that countries with longer time series receive
more weight in the calculations. Again, there is also a relatively weak correlation between
consumption and output growth and stock returns at all the horizons we consider.
We have focused on correlations because we find them easy to interpret. One might
be concerned that a di§erent picture emerges from the pattern of covariances between stock

6
returns and fundamentals. It does not. For example, using quarterly U.S. data for the period
1959 to 2000, Parker (2001) argues that one would require a risk aversion coe¢cient of 379
to account for the equity premium given his estimate of the covariance between consumption
growth and stock returns.6
We conclude this subsection by considering the dynamic correlations between consump-
tion growth and stock returns. Authors like Parker (2001), Campbell (2003), and Jaganathan
and Wang (2007) find that in the U.S. post-war quarterly data, stock returns are correlated
with future consumption growth.7 Figure 1 reports the correlation between returns at time
t and consumption growth at time t + τ , where τ ranges from −10 years to +10 years. The
dotted lines are the limits of the 95 percent confidence interval. The figure also reports the
analogue correlations involving output growth. Three key features are worth noting. First,
for the U.S. and Canada there does appear to be a significant correlation between returns
at time t and both consumption and output growth at time t + 1. Second, this result is
not robust across the other G7 countries. Third, there is no other value of τ for which the
correlation is significant for any of the countries. So, we infer that, even taking dynamic cor-
relations into account, there is a weak correlation between stock returns and fundamentals,
as measured by consumption and output growth. Nevertheless, because we estimate our
structural model using U.S. data, we assess its ability to be consistent with the correlations
between stock returns at t and consumption growth at t + 1.
Viewed overall, the results in this section serve as our motivation for introducing shocks
to the demand for assets. Classic representative-agent models load all uncertainty onto the
supply-side of the economy. As a result, they have di¢culty in simultaneously accounting
for the equity premium and the correlation puzzle.8 This di¢culty is shared by the habit-
formation model proposed by Campbell and Cochrane (1999) and the long-run risk models
proposed by Bansal and Yaron (2004) and Bansal, Kiku, and Yaron (2012). Rare-disaster
models of the type proposed by Rietz (1988) and Barro (2006) also share this di¢culty
because all shocks, disaster or not, are to the supply side of the model. A model with a time-
6
Parker (2001) finds that there is a larger covariance between current stock returns and the cumulative
growth rate of consumption over the next 12 quarters. He shows that using this covariance measure, one
would require a risk aversion coe¢cient of 38 to rationalize the equity premium (see also Grossman, Melino
and Shiller (1987)).
7
In related work, Backus, Routledge, and Zin (2010) show that equity returns is a leading indicator of
the business-cycle components of consumption and output growth.
8
Lynch (1996) and Gârleanu, Kogan, and Panageas (2012) provide interesting analyses of overlapping-
generations models which generate an equity premium, even though the correlation between consumption
growth and equity returns is low.

7
varying disaster probability, of the type considered by Wachter (2013) and Gourio (2012),
might be able to rationalize the low correlation between consumption and stock returns
as a small-sample phenomenon. The reason is that changes in the probability of disasters
induce movements in stock returns without corresponding movements in actual consumption
growth. This force lowers the correlation between stock returns and consumption in a sample
where rare disasters are under represented. This explanation might account for the post-war
correlations. But we are more skeptical that it accounts for the results in Table 1, panel B,
which are based on the longer sample period, 1871 to 2006, which includes disasters such as
the Great Depression and two World Wars.
Below, we focus on demand shocks as the source of the low correlation between stock
returns and fundamentals, rather than the alternatives just mentioned. We model these
demand shocks in the simplest possible way by introducing shocks to the time preference of
the representative agent. Consistent with the references in the introduction, these shocks can
be thought of as capturing changes in agents’ attitudes towards savings or, more generally,
investor sentiment.

3. The model

In this section, we study the properties of a representative-agent endowment economy modi-

fied to allow for time-preference shocks. The representative agent has the constant-elasticity
version of Kreps-Porteus (1978) preferences used by Epstein and Zin (1991) and Weil (1989).
The life-time utility of the representative agent is a function of current utility and the cer-
∗
tainty equivalent of future utility, Ut+1 :
h " ∗ #1−1/ i1/(1−1/ )
1−1/
Ut = max λt Ct + δ Ut+1 , (3.1)
Ct

where Ct denotes consumption at time t and δ is a positive scalar. The certainty equivalent
∗
of future utility is the sure value of t + 1 lifetime utility, Ut+1 such that:
" ∗
#1−γ " 1−γ #
Ut+1 = Et Ut+1 .

The parameters and γ represent the elasticity of intertemporal substitution and the coef-
ficient of relative risk aversion, respectively. The ratio λt+1 /λt determines how agents trade
o§ current versus future utility. We assume that this ratio is known at time t.9 We refer to
9
We obtain similar results with a version of the model in which the utility function takes the form:
h " ∗ #1−1/ i1/(1−1/ )
1−1/
Ut = Ct + λt δ Ut+1 . The assumption that the agents knows λt+1 at time t is made

8
λt+1 /λt as the time-preference shock. Propositions 6.9 and 6.18 in Skiadas (2009) provide a
set of axioms that applies to recursive utility functions with preference shocks.10

3.1. The benchmark model

To highlight the role of time-preference shocks, we begin with a very simple stochastic process
for consumption:
log(Ct+1 /Ct ) = µ + σ c "ct+1 . (3.2)

Here, µ and σ c are non-negative scalars and "ct+1 follows an i.i.d. standard-normal distribu-
tion.
As in Campbell and Cochrane (1999), we allow dividends, Dt , to di§er from consumption.
In particular, we assume that:

log(Dt+1 /Dt ) = µ + π dc "ct+1 + σ d "dt+1 . (3.3)

Here, "dt+1 is an i.i.d. standard-normal random variable that is uncorrelated with "ct+1 . To
simplify, we assume that the average growth rate of dividends and consumption is the same
(µ). The parameter σ d ≥ 0 controls the volatility of dividends. The parameter π dc controls
the correlation between consumption and dividend shocks.11
The variable Λt+1 = log(λt+1 /λt ) determines how agents trade o§ current versus future
utility. This ratio is known at time t. We refer to Λt+1 as the time-preference shock. This
shock evolves according to:
Λt+1 = ρΛ Λt + σ Λ "Λt , (3.4)

where, "Λt is an i.i.d. standard-normal random variable. In the spirit of the original Lucas
(1978) model, we assume, for now, that "Λt is uncorrelated with "ct and "dt . We relax this
assumption in Subsection 3.4.
to simplify the algebra and is not necessary for any of the key results.
10
Skiadas (2009) derives a parametric SDF that satisfies the axioms in proposition 6.9 and 6.18 (see his
equation 6.35). This SDF can be modified to obtain a generalized Epstein and Zin (1989) parametric utility
function with stochastic risk aversion, intertemporal substitution, and time-preference shocks. We thank
Soohun Kim and Ravi Jagannathan for pointing this result out to us.
11
The stochastic process described by equations (3.2) and (3.3) implies that log(Dt+1 /Ct+1 ) follows a
random walk with no drift. This implication is consistent with our data.

9
The CRRA case In Appendix A we solve this model analytically for the case in which
γ = 1/ . Here preferences reduce to the CRRA form:

X
1
1−γ
Vt = Et δ i λt+i Ct+i , (3.5)
i=0

with Vt = Ut1−γ .
We denote the risk-free rate and the rate of return on a claim on consumption by Rf,t+1 ,
and Rc,t+1 , respectively. The unconditional risk-free rate depends on the persistence and
volatility of time-preference shocks:
& '
σ 2Λ /2
E (Rf,t+1 ) = exp δ −1 exp(γµ − γ 2 σ 2c /2).
1 − ρ2Λ
The unconditional equity premium implied by this model is proportional to the risk-free
rate:
( " # )
E (Rc,t+1 − Rf,t+1 ) = E (Rf,t+1 ) exp γσ 2c − 1 . (3.6)

Both the average risk-free rate and the volatility of consumption are small in the data.
Moreover, the constant of proportionality in equation (3.6), exp (γσ 2c ) − 1, is independent
of σ 2Λ . So, time-preference shocks do not help to resolve the equity premium puzzle when
preferences are of the CRRA form.

3.2. Solving the benchmark model

We define the return to the stock market as the return to a claim on the dividend process.
The realized gross stock-market return is given by:
Pd,t+1 + Dt+1
Rd,t+1 = , (3.7)
Pd,t
where Pd,t denotes the ex-dividend stock price.
It is useful to define the realized gross return to a claim on the endowment process:
Pc,t+1 + Ct+1
Rc,t+1 = . (3.8)
Pc,t
Here, Pc,t denotes the price of an asset that pays a dividend equal to aggregate consump-
tion. We use the following notation to define the logarithm of returns on the dividend and
consumption claims, the logarithm of the price-dividend ratio, and the logarithm of the

10
price-consumption ratio:

rd,t+1 = log(Rd,t+1 ),
rc,t+1 = log(Rc,t+1 ),
zdt = log(Pd,t /Dt ),
zct = log(Pc,t /Ct ).

In Appendix B we show that the logarithm of the stochastic discount factor (SDF) implied
by the utility function defined in equation (3.1) is given by:
θ
mt+1 = θ log (δ) + θΛt+1 − ∆ct+1 + (θ − 1) rc,t+1 , (3.9)

where θ is defined as:

1−γ
θ= . (3.10)
1 − 1/
When γ = 1/ , the case of CRRA preferences, the value of θ is equal to one and the
stochastic discount factor is independent of rc,t+1 .
We solve the model using the approximation proposed by Campbell and Shiller (1988),
which involves linearizing the expressions for rc,t+1 and rd,t+1 and exploiting the properties
of the log-normal distribution.12
Using a log-linear Taylor expansion we obtain:

rd,t+1 = κd0 + κd1 zdt+1 − zdt + ∆dt+1 , (3.11)

rc,t+1 = κc0 + κc1 zct+1 − zct + ∆ct+1 , (3.12)

where ∆ct+1 ≡ log (Ct+1 /Ct ) and ∆dt+1 ≡ log (Dt+1 /Dt ). The constants κc0 , κc1 , κd0 , and
κd1 are given by:

κd0 = log [1 + exp(zd )] − κd1 zd ,

κc0 = log [1 + exp(zc )] − κc1 zc ,

exp(zd ) exp(zc )
κd1 = , κc1 = ,
1 + exp(zd ) 1 + exp(zc )
where zd and zc are the unconditional mean values of zdt and zct .
12
See Hansen, Heaton, and Li (2008) for an alternative solution procedure.

11
 The Euler equations associated with a claim to the stock market and a consumption
claim can be written as:
Et [exp (mt+1 + rd,t+1 )] = 1, (3.13)

Et [exp (mt+1 + rc,t+1 )] = 1. (3.14)

We solve the model using the method of undetermined coe¢cients. First, we replace
mt+1 , rc,t+1 and rd,t+1 in equations (3.13) and (3.14), using expressions (3.11), (3.12) and
(3.9). Second, we guess and verify that the equilibrium solutions for zdt and zct take the
form:

zdt = Ad0 + Ad1 Λt+1 , (3.15)

zct = Ac0 + Ac1 Λt+1 . (3.16)

This solution has the property that price-dividend ratios are constant, absent movements
in Λt+1 . This property results from our assumption that the logarithm of consumption and
dividends follow random-walk processes. We compute Ad0 , Ad1 , Ac0 , and Ac1 using the
method of undetermined coe¢cients.
The equilibrium solution has the property that Ad1 , Ac1 > 0. We show in Appendix B
that the conditional expected return to equity is given by:

Et (rd,t+1 ) = − log (δ) − Λt+1 + µ/ (3.17)

* +
(1 − θ)
+ (1 − γ)2 − γ 2 σ 2c /2 + π dc (2γσ c − π dc ) /2 − σ 2d /2
θ
, -
+ (1 − θ) (κc1 Ac1 ) [2 (κd1 Ad1 ) − (κc1 Ac1 )] − (κd1 Ad1 )2 σ 2Λ /2.

Recall that κc1 and κd1 are non-linear functions of the parameters of the model.
Using the Euler equation for the risk-free rate, rf,t+1 ,

Et [exp (mt+1 + rf,t+1 )] = 1,

we obtain:

rf,t+1 = − log (δ) − Λt+1 + µ/ − (1 − θ) (κc1 Ac1 )2 σ 2Λ /2 (3.18)

* +
(1 − θ) 2
+ (1 − γ) − γ σ 2c /2.
2
θ
Equations (3.17) and (3.18) imply that the risk-free rate and the conditional expectation
of the return to equity are decreasing functions of Λt+1 . When Λt+1 rises, agents value the

12
future more relative to the present, so they want to save more. Since risk-free bonds are
in zero net supply and the number of stock shares is constant, aggregate savings cannot
increase. So, in equilibrium, returns on bonds and equity must fall to induce agents to save
less.
The approximate response of asset prices to shocks, emphasized by Borovička, Hansen,
Hendricks, and Scheinkman (2011) and Borovička and Hansen (2011), can be directly inferred
from equations (3.17) and (3.18). The response of the return to stocks and the risk-free rate
to a time-preference shock corresponds to that of an AR(1) with serial correlation ρΛ .
Using equations (3.17) and (3.18) we can write the conditional equity premium as:

Et (rd,t+1 ) − rf,t+1 = π dc (2γσ c − π dc ) /2 − σ 2d /2 (3.19)

+κd1 Ad1 [2 (1 − θ) Ac1 κc1 − κd1 Ad1 ] σ 2Λ /2.

We define the compensation for valuation risk as the part of the one-period expected
excess return to an asset that is due to the volatility of the time preference shock, σ 2Λ . We
refer to the part of the one-period expected excess return that is due to the volatility of
consumption and dividends as the compensation for conventional risk.
The component of the equity premium that is due to valuation risk, vd , is given by the
last term in equation (3.19). Since the constants Ac1 , Ad1 , κc1 , and κd1 are all positive, θ < 1
is a necessary condition for valuation risk to help explain the equity premium (recall that θ
is defined in equation (3.10)).13
The intuition for why valuation risk helps account for the equity premium is as follows.
Consider an investor who buys the stock at time t. At some later time, say t + τ , τ > 0,
the investor may get a preference shock, say a decrease in Λt+τ +1 , and want to increase
consumption. Since all consumers are identical, they all want to sell the stock at the same
time, so the price of equity falls. Bond prices also fall because consumers try to reduce their
holdings of the risk-free asset to raise their consumption in the current period. Since stocks
are infinitely-lived compared to the one-period risk-free bond, they are more exposed to this
source of risk. So, valuation risk, vd , gives rise to an equity premium. In the CRRA case
(θ = 1) the e§ect of valuation risk on the equity premium is generally small (see equation
3.6)).
13
As discussed in Epstein et al (2014), the condition θ < 1 also plays a crucial role in generating a high
equity premium in long-run risk models. Because long-run risks are resolved in the distant future, they are
more heavily penalized than current risks. For this reason, long-run risk models can generate a large equity
premium even when shocks to current consumption are small.

13
 It is interesting to highlight the di§erences between time-preference shocks and conven-
tional sources of uncertainty, which pertain to the supply-side of the economy. Suppose that
there is no risk associated with the physical payo§ of assets such as stocks. In this case,
standard asset pricing models would imply that the equity premium is zero. In our model,
there is a positive equity premium that results from the di§erential exposure of bonds and
stocks to valuation risk. Agents are uncertain about how much they will value future divi-
dend payments. Since Λt+1 is known at time t, this valuation risk is irrelevant for one-period
bonds. But, it is not irrelevant for stocks, because they have infinite maturity. In general,
the longer the maturity of an asset, the higher is its exposure to time-preference shocks and
the larger is the valuation risk.
We conclude by considering the case in which there are supply-side shocks to the economy
but agents are risk neutral (γ = 0). In this case, the component of the equity premium
that is due to valuation risk is always positive as long as < 1. The intuition is as follows:
stocks are long-lived assets whose payo§s can induce unwanted variation in the period utility
1−1/
of the representative agent, λt Ct . Even when agents are risk neutral, they must be
compensated for the risk of this unwanted variation.

3.3. Relation to the long-run risk model

In this subsection, we briefly comment on the relation between our model and the long-run-
risk model pioneered by Bansal and Yaron (2004). Both models emphasize low-frequency
shocks that induce large, persistent changes in the agent’s stochastic discount factor.14 To
see this point, it is convenient to re-write the representative agent’s utility function, (3.1),
as:
h " ∗ #1−1/ i1/(1−1/ )
1−1/
Ut = C̃t + δ Ut+1 , (3.20)
1/(1−1/ )
where C̃t = λt Ct . Taking logarithms of this expression we obtain:
. /
log C̃t = 1/ (1 − 1/ ) log(λt ) + log (Ct ) .

Bansal and Yaron (2004) introduce a highly persistent component in the process for log(Ct ),
which is a source of long-run risk. In contrast, we introduce a highly persistent component
into log(C̃t ) via our specification of the time-preference shock. From equation (3.9), it is
clear that both specifications can induce large, persistent movements in mt+1 . Despite this
14
See Dew-Becker (2014) for a version of a long-run risk model in a production economy.

14
similarity, the two models are not observationally equivalent. First, they have di§erent
implications for the correlation between observed consumption growth, log(Ct+1 /Ct ), and
asset returns. Second, the two models have very di§erent implications for the average return
to long-term bonds, and the term structure of interest rates. We return to these points
when we discuss our empirical results in Section 6. There we present empirical evidence for
the implications of di§erent models for the slope of the yield curve. In addition, we assess
whether the SDFs of the di§erent models have the properties that Backus and Zin (1994)
argue are necessary for explaining the slope of the yield curve and the persistence of bond
yields.

3.4. The extended model

The benchmark model just described is useful to highlight the role of time-preference shocks
in a§ecting asset returns. But its simplicity leads to two important empirical shortcomings.15
First, since consumption is a martingale, the only state variable that is relevant for asset
returns is Λt+1 . This property means that all asset returns are highly correlated with each
other and with the price-dividend ratio. Second, and related, the model displays constant
risk premia and so it cannot generate predictability in excess returns.
In this subsection, we address the shortcomings of the benchmark model by allowing for
a richer model of consumption and dividend growth:
" #
log(Ct+1 /Ct ) = µc + ρc log(Ct /Ct−1 ) + αc σ 2t+1 − σ 2 + π cΛ "Λt+1 + σ t "ct+1 , (3.21)
" #
log(Dt+1 /Dt ) = µd +ρd log(Dt+1 /Dt )+αd σ 2t+1 − σ 2 +σ d σ t "dt+1 +π dΛ "Λt+1 +π dc σ t "ct+1 , (3.22)

and
" #
σ 2t+1 = σ 2 + v σ 2t − σ 2 + σ w wt+1 , (3.23)

where "ct+1 , "dt+1 , "Λt+1 , and wt+1 are mutually uncorrelated standard-normal variables. We
restrict the unconditional average growth rate of consumption and dividends to be the same:
µc µd
= .
1 − ρc 1 − ρd
Relative to the benchmark model, equations (3.21)-(3.23) incorporate three new features.
First, motivated by the data, we allow for serial correlation in the growth rates of consump-
tion and dividends. Second, as in Kandel and Stambaugh (1990) and Bansal and Yaron
15
The shortcomings of our benchmark model are shared by other simple models like model I in Bansal
and Yaron (2004), which abstract from conditional heteroskedasticity in consumption and dividends.

15
(2004), we allow for conditional heteroskedasticity in consumption. This feature generates
time-varying risk premia: when volatility is high the stock is risky, its price is low and its ex-
pected return is high. High volatility leads to higher precautionary savings motive so that the
risk-free rate falls, reinforcing the rise in the risk premium. According to our specification,
increases in volatility a§ects the growth rate of consumption and dividends. This interaction
is stressed in the DSGE models proposed by Basu and Bundick (2015), Kung (2015), and
Fernandez-Villaverde, Guerron-Quintana, Kuester, and Rubio-Ramírez (2015). A common
property of these models is that an increase in the volatility of shocks to the economy a§ects
precautionary savings and reduces consumption. Third, we allow for a correlation between
time-preference shocks and the growth rate of consumption and dividends. In a production
economy, time-preference shocks would generally induce changes in aggregate consumption.
For example, in a simple real-business-cycle model, a persistent increase in Λt+1 would lead
agents to reduce current consumption and raise investment in order to consume more in
the future. Taken literally, an endowment economy specification does not allow for such a
correlation. Importantly, only the innovation to time-preference shocks enters the law of
motion for log(Ct+1 /Ct ) and log(Dt+1 /Dt ). So, equations (3.21)-(3.23) do not introduce any
element of long-run risk into consumption or dividend growth.
Since the price-dividend ratio and the risk-free rate are driven by a single state variable
in the benchmark model, they have the same degree of persistence. But, according to our
empirical estimates, the price-dividend ratio is more persistent than the risk-free rate. A
straightforward way to address this shortcoming of the benchmark model is to assume that
the time-preference shock is the sum of a persistent shock and an i.i.d. shock:

Λt+1 = xt + σ η η t+1 , (3.24)

xt+1 = ρΛ xt + σ Λ "Λt+1 .

Here "Λt+1 and η t+1 are uncorrelated, i.i.d. standard normal shocks. We think of xt as
capturing low-frequency changes in the growth rate of the discount rate. In contrast, η t+1
can be thought of high-frequency changes in investor sentiment that a§ect the demand for
assets (see, for example, Dumas et al. (2009)). If σ η = 0 and x1 = Λ1 , we obtain the
specification of the time-preference shock used in the benchmark model. Other things equal,
the larger is σ η , the lower is the persistence of the time-preference shock.

16
4. Estimation methodology

We estimate the parameters of our model using the Generalized Method of Moments (GMM).
Our estimator is the parameter vector Φ̂ that minimizes the distance between a vector of
empirical moments, ΨD , and the corresponding model moments, Ψ(Φ̂). Our estimator, Φ̂, is
given by:
Φ̂ = arg min [Ψ(Φ) − ΨD ]0 Ω−1
D [Ψ(Φ) − ΨD ] .
Φ

We found that, for a wide range of parameter values, the model implies that there is small-
sample bias in terms of various moments, especially the predictability of excess returns. We
therefore focus on the plim of the model-implied small-sample moments when constructing
Ψ(Φ), rather than the plim of the moments themselves. For a given parameter vector, Φ, we
create 500 synthetic time series, each of length equal to our sample size. For each sample,
we calculate the sample moments of interest. The vector Ψ(Φ) that enters the criterion
function is the median value of the sample moments across the synthetic time series.16 In
addition, we assume that agents make decisions at a monthly frequency and derive the
model’s implications for variables computed at an annual frequency. We estimate ΨD using
a standard two-step e¢cient GMM estimator with a Newey-West (1987) weighting matrix
that has ten lags. The latter matrix corresponds to our estimate of the variance-covariance
matrix of the empirical moments, ΩD .
When estimating the benchmark model, we include the following 19 moments in ΨD : the
mean and standard deviation of consumption growth, the mean and standard deviation of
dividend growth, the contemporaneous correlation between the growth rate of dividends and
the growth rate of consumption, the mean and standard deviation of real stock returns, the
mean, standard deviation and autocorrelation of the price-dividend ratio, the mean, standard
deviation and autocorrelation of the real risk-free rate, the correlation between stock returns
and consumption growth at the one, five and ten-year horizon, the correlation between stock
returns and dividend growth at the one, five and ten-year horizon. We constrain the growth
rate of dividends and consumption to be the same. In practice, when we estimate the
benchmark model we find that the standard deviation of the point estimate of the risk-free
rate across the 500 synthetic time series is very large. So here we report results corresponding
to the case where we constrain the median risk-free rate to exactly match our point estimate.
16
As the sample size grows, our estimator becomes equivalent to a standard GMM estimator so that the
usual asymptotic results for the distribution of the estimator apply.

17
 For the benchmark model, the vector Φ includes nine parameters: γ, the coe¢cient of
relative risk aversion, , the elasticity of intertemporal substitution, δ, the rate of time pref-
erence, σ c , the volatility consumption growth shocks, π dc , the parameter that controls the
correlation between consumption and dividend growth shocks, σ d , the volatility of dividend
growth shocks, ρΛ , the persistence of time-preference shocks, σ Λ , the volatility of the innova-
tion to time-preference shocks, and µ, the mean growth rate of dividends and consumption.
In the extended model, the vector Φ includes nine additional parameters: ρc and ρd ,
which control the serial correlation of consumption and dividend growth, αc and αd , which
control the e§ect of volatility on the growth rate of consumption and dividends, respectively,
π cΛ and π dΛ , which control the e§ect of time preference shock innovations on consumption
and dividends, respectively, ν, which governs the persistence of volatility, σ w , the volatility
of innovations to volatility, and σ η , the volatility of transitory shocks to time preference. In
estimating the extended model, we add to ΨD the following six moments: the first-order
serial correlation of consumption growth, the average yield for 5 and 20-year bonds, and the
slope coe¢cients from the regressions of the excess-equity returns over holding periods of 1,
3 and 5 years on the lagged price-dividend ratio.

5. Empirical results

In the main text, we discuss the results that we obtain estimating the models using the ex-
ante measures of real bond yields. In appendix D, we report the analogue results obtained
estimating the models with ex-post measures of real bond yields.
Table 4 reports our parameter estimates along with standard errors. Several features
are worth noting. First, the coe¢cient of risk aversion is quite low, 1.5 and 2.4, in the
benchmark and extended models, respectively. We estimate this coe¢cient with reasonable
precision. Second, for both models, the intertemporal elasticity of substitution is somewhat
larger than one. Third, for both models, the point estimates satisfy the necessary condition
for valuation risk to be positive, θ < 1. Fourth, the parameter ρΛ that governs the serial
correlation of the growth rate of Λt is estimated to be close to one in both models, 0.991 and
0.992 in the benchmark and extended model, respectively. Fifth, the parameter ν, which
governs the persistence of consumption volatility in the extended model, is also quite high
(0.997). The high degree of persistence in both the time-preference and the volatility shock
are the root cause of the small-sample biases in our estimators.

18
 Table 5 compares the small-sample moments implied by the benchmark and extended
models with the estimated data moments. Recall that in estimating the model parameters
we impose the restriction that the unconditional average growth rate of consumption and
dividends are the same. To assess the robustness of our results to this restriction, we present
two versions of the estimated data moments, with and without the restriction. With one
exception, the constrained and unconstrained data moment estimates are similar, taking
sampling uncertainty into account. The exception is the average growth rate of consumption,
where the constrained and unconstrained estimates are statistically di§erent.

Implications for the equity premium Table 5 shows that both the benchmark and
extended model give rise to a large equity premium, 5.81 and 3.88, respectively. This result
holds even though the estimated degree of risk aversion is quite moderate in both models.
In contrast, long-run risk models require a high degree of risk aversion to match the equity
premium.
Recall that in order for valuation risk to contribute to the equity premium, θ must be less
than one. This condition is clearly satisfied by both our models: the estimated value of θ is
−1.6 and −2.6 in the benchmark and extended model, respectively (Table 4). In both cases,
θ is estimated quite accurately. Taking sampling uncertainty into account, the benchmark
model easily accounts for the equity premium, while the extended model does so marginally.
We can easily reject the null hypothesis of θ = 1, which corresponds to the case of constant
relative risk aversion.
The basic intuition for why our model generates a high equity premium despite a low
coe¢cient of relative risk aversion is as follows. From the perspective of the model, stocks
and bonds are di§erent in two ways. First, the model embodies the conventional source of an
equity premium, namely a di§erential covariance of bonds and stocks with the SDF. Since γ
is relatively small, this traditional covariance e§ect is also small. Second, the model embodies
a compensation for valuation risk that is particularly pronounced for stocks because they
they have longer maturities than bonds. Recall that, given our timing assumptions, when
an agent buys a bond at time t, the agent knows the value of Λt+1 , so the only source of risk
are movements in the marginal utility of consumption at time t + 1. In contrast, the time-t
stock price depends on the value of Λt+j , for all j > 1. So, agents are exposed to valuation
risk, a risk that is particularly important because time-preference shocks are very persistent.
To assess the importance of valuation risk in accounting for the equity and bond premium,

19
we set the variance of the time-preference shock in the benchmark model to zero. The equity
premium falls from 5.8 percent to −0.2. So, in that model the equity premium is driven solely
by valuation risk.
Next, we set the variance of various shocks to zero in the extended model. Because these
shocks are correlated, the resulting decomposition is not additive. First, we set the variance
of the persistent component of the preference shock (σ Λ ) to zero. As a result, the equity
premium falls from 3.9 to 2.2. The bond premium, defined as the di§erence between yields
on 20-year bonds and 1-year bonds falls from 2.2 percent to −0.6. Second, we make all the
shocks in the model conditionally homoscedastic by setting σ w to zero. The equity premium
drops from 3.9 to 2.2, while the bond premium rises from 2.2 to 2.8. Finally, we set both
σ Λ and σ w to zero. The equity premium falls from 3.9 to −0.5, while the bond premium
falls from 2.2 to zero. Based on these results, we infer that both conditional volatility and
valuation risk play important roles in generating an equity premium in the extended model.
In contrast, the bond premium is entirely driven by valuation risk.

Implications for the risk-free rate A problem with some explanations of the equity
premium is that they imply counterfactually high levels of volatility for the risk-free rate
(see e.g. Boldrin, Christiano and Fisher (2001)). Table 5 shows that the volatility of the
risk-free rate and stock market returns implied by our model are similar to the estimated
volatilities in the data.
An empirical shortcoming of the benchmark model is its implication for the persistence
of the risk-free rate. Recall that, according to equation (3.18), the risk-free rate has the same
persistence as the growth rate of the time-preference shock. Table 5 shows that the AR(1)
coe¢cient of the risk-free rate, as measured by the ex-ante real returns to one-year treasury
bills, is only 0.52, with a standard error of 0.07. This estimate is substantially smaller than
the analogue statistic implied by the benchmark model (0.90). The extended model does a
much better job at accounting for the persistence of the risk-free rate (0.51). In this model
there are both transitory and persistent shocks to the risk-free rate. The former account for
roughly 44 percent of the variance of Λt .

20
Implications for the correlation puzzle Table 6 reports the model’s implications for
the correlation of stock returns with consumption and dividend growth.17 Recall that in
the benchmark model consumption and dividends follow a random walk. In addition, the
estimated process for the growth rate of the time-preference shock is close to a random walk.
So, the correlation between stock returns and consumption growth implied by the model is
essentially the same across di§erent horizons. A similar property holds for the correlation
between stock returns and dividend growth.
In the extended model, persistent changes in the variance of the growth rate of consump-
tion and dividends can induce persistent changes in the conditional mean of these variables.
As a result, this model produces correlations between stock returns and fundamentals that
vary across di§erent horizons.
The benchmark model does well at matching the correlation between stock returns and
consumption growth in the data, because this correlation is similar at all horizons. In
contrast, the empirical correlation between stock returns and dividend growth increases with
the time horizon. The estimation procedure chooses to match the long-horizon correlations
and does less well at matching the yearly correlation. This choice is dictated by the fact that
it is harder for the model to produce a low correlation between stock returns and dividend
growth than it is to produce a low correlation between stock returns and consumption growth.
This property reflects the fact that the dividend growth rate enters directly into the equation
for stock returns (see equation (3.11)).
The extended model does better at capturing the fact that the correlations between
equity returns and dividend growth rises with the horizon for two reasons. When volatility
is high, the returns to equity are high. Since αd < 0, the growth rate of dividends is low. As
a result, the one-year correlation between dividend growth and equity returns is negative.
The variance of the shock to the dividend growth rate is mean reverting. So, the e§ect
of a negative value of αd becomes weaker as the horizon extends. The direct association
between equity returns and dividend growth (see equation (3.11)), which induces a positive
correlation, eventually dominates as the horizon gets longer.
An additional force that allows the extended model to generate a lower short-term cor-
relation between equity returns and dividend growth, is that the estimated value of π dΛ is
17
The moments reported in this table are slightly di§erent from those reported in Table 1. The reason is
that the moments in the two tables were computed using di§erent samples. In Table 6, we compute a leaded
and a lagged correlation, so we lose one observation in the beginning of the sample and another in the end
of the sample.

21
negative. The estimation algorithm chooses parameters that allow the model to do rea-
sonably well in matching the one- and five-year correlation, at the cost of doing less well
at matching the ten-year correlation. Presumably, this choice reflects the greater precision
with which the one-year and five-year correlations are estimated relative to the ten-year
correlation. Taking sampling uncertainty into account, the extended model matches the cor-
relation between stock returns and consumption growth at di§erent horizons. Interestingly,
the correlation between stock returns and consumption growth increases with the horizon.
We conclude by highlighting an important di§erence between our model and two leading
alternatives. The first is the external-habit model proposed by Campbell and Cochrane
(1999).18 Working with their parameter values, we find that the correlation between stock
returns and consumption growth are equal to 0.53, 0.69, and 0.62 at the one-, five- and
ten-year horizon, respectively.
The second alternative is the long-run risk model proposed by Bansal, Kiku, and Yaron
(2012). Working with their parameter values, we find that the correlation between stock
returns and consumption growth are equal to 0.34, 0.49, and 0.54 at the one-, five- and
ten-year horizon, respectively. Their model also implies correlations between stock returns
and dividend growth equal to 0.64, 0.90, and 0.93 at the one-, five- and ten-year horizon,
respectively.
Our estimates reported in Table 1 imply that, for both models, the correlations between
fundamentals and returns are counterfactually high. The source of this empirical shortcoming
is that all the uncertainty in the long-run risk model stems from the endowment process.

Implications for the price-dividend ratio In Table 5 we see that both the benchmark
and the extended models match the average price-dividend ratio very well. The benchmark
model somewhat underpredicts the persistence and volatility of the price-dividend ratio. The
extended model does much better at matching those moments. The moments implied by
this model are within two standard errors of their sample counterparts.
Table 7 presents evidence reproducing the well-known finding that excess returns are
predictable based on lagged price-dividend ratios. We report the results of regressing excess-
equity returns over holding periods of 1, 3 and 5 years on the lagged price-dividend ratio.
The slope coe¢cients are −0.10, −0.27, and −0.42, respectively, while the R-squares are
18
As in Wachter (2005), we consider a version of the model in which equities are a claim to consumption.
We thank Jessica Wacher for sharing her program for solving the Campbell-Cochrane model with us.

22
0.07, 0.14, and 0.26, respectively.
The analogue results for the benchmark model are shown in the top panel of Table 7.
In this model, consumption is a martingale with conditionally homoscedastic innovations.
So, by construction, excess returns are unpredictable in population at a monthly frequency.
Since we aggregate the model to annual frequency, temporal aggregation produces a small
amount of predictability (see column titled “Model (plim)”).
Stambaugh (1999) and Boudoukh et al. (2008) argue that the predictability of excess
returns may be an artifact of small-sample bias and persistence in the price-dividend ratio.
Our results are consistent with this hypothesis. The column labeled “Model (median)”
reports the plim of the small moments implied by our model. The slope coe¢cients for the
1, 5, and 10 year horizons, are −0.05, −0.14 and −0.21, respectively. In each case, the
median Monte Carlo point estimates are contained within a two-standard deviation band of
the respective data estimates.
Table 7 also presents results for the extended model. Because of conditional heteroskedas-
ticity in consumption, periods of high volatility in consumption growth are periods of high
expected equity returns and low equity prices. So, in principle, the model is able to generate
predictability in population. At our estimated parameter values this predictability is quite
small. But, once we allow for the e§ects of small-sample bias, the extended model does quite
well at accounting for the regression slope coe¢cients. Finally, we note that the benchmark
model understates the regression R-squares. Taking sampling uncertainty into account, the
extended model does better on this dimension.
Cochrane (1992) proposes a decomposition of the variation of the price-dividend ratio
into three components: excess returns, dividend growth, and the risk-free rate.19 While this
decomposition is not additive, authors like Bansal and Yaron (2004) use it to compare the
importance of these three components in the model and data.
In our sample, the point estimate for the percentage of the variation in price—dividend
ratio due to excess return fluctuations is 102.2 percent, with a standard error of 30 percent.
Dividend growth accounts for −14.5 with a standard error of 13 percent. Finally, the risk-
free rate accounts for −20.4 percent with a standard error of 14.8 percent. These results are
similar to those in Cochrane (1992) and Bansal and Yaron (2004).
19
The fraction of the volatility of the price-dividend ratio attributed to variable x is given by
15
X
Ωj cov(zdt , xt+j )/var(zdt ), where Ω = 1/(1 + E(Rd,t+1 )). See Cochrane (1992) for details.
j=1

23
 Based on the small-sample moments implied by the benchmark model, the fraction of
the variance of the price-dividend ratio accounted for by excess returns, dividend growth,
and the risk-free rate is 34.6, −2.8, and 54.2, respectively. So, this model clearly overstates
the importance of the risk-free rate and understates the importance of excess returns in
accounting for the variance of the price-dividend ratio.
The extended model does substantially better than the benchmark model. Based on the
small-sample moments implied by the extended model, the fraction of the variance of the
price-dividend ratio accounted for by excess returns, dividend growth, and the risk-free rate
is 51.2, −38.5, and 5.6, respectively. The fraction attributed to excess returns is just within
two standard errors of the point estimate. The fraction attributed to dividend growth is well
within two standard errors of the point estimate.
It is interesting to contrast these results to those in Bansal and Yaron (2004). Their
model also attributes a large fraction of the variance of the price-dividend ratio to excess
returns. At the same time, their model substantially overstates the role of dividend growth
in accounting for the variance of the price-dividend ratio.

Correlation between the price-dividend ratio and the risk-free rate We now dis-
cuss the implications of our model for the correlation between the price-dividend ratio and
the risk-free rate. Table 5 reports that our estimate of this correlation is 0.27 with a stan-
dard error of 0.11. The benchmark model predicts that this correlation should be sharply
negative (−0.95). The intuition for this result is clear. When there is a shock to the rate of
time preference, agents want to save more, both in the form of bonds and stocks. So, the
price-dividend ratio must rise to clear the equity market and the risk-free rate must fall to
clear the bond market. The extended model does much better along this dimension of the
data (−0.20). Recall that a positive shock to "Λt drives the risk-free rate down. Since π dΛ is
negative, the same shock lowers the growth rate of dividends which causes the price dividend
ratio to fall. So, the extended model embodies a force that generates a positive correlation
between the risk-free rate and the price-dividend ratio. This forces helps to substantially
lower the sharp negative correlation between these variables in the benchmark model.
It is interesting to compare our model with Bansal, Kiku and Yaron (2012). The long-run
risk shock in that model induces a counterfactually sharp positive correlation: 0.65. The
intuition is straightforward. A shock that induces a persistent change in the growth rate of
dividends causes stock prices to go up since equity now represents a claim to a higher flow of

24
dividends. Dividends do not change much in the short run, so the price-dividend ratio goes
up. At the same time, the shock causes agents to want to shift away from bond holdings and
into equity. So, the risk-free rate must rise to clear the bond market. Other things equal,
the long-run risk shock induces a positive correlation between the price-dividend ratio and
the risk-free rate. In contrast to our model, the Bansal, Kiku and Yaron (2012) does not
embody any forces that would lead to a negative correlation between the price-dividend ratio
and the risk-free rate. The net result is that their model overstates this correlation.
We are hesitant to use this correlation as a litmus test to choose between our model
and Bansal, Kiku and Yaron’s for two reasons. First, the point estimate of this correlation
is sensitive to sample period selection. The correlation is positive (0.39) before 1983 but
negative (−0.42) after 1983. Second, the sign of this correlation varies substantially across
the G7 countries and is again sensitive to breaking the sample in 1983 (see Table 8).

6. Bond term premia

As we emphasize above, valuation risk plays an important role in generating an equity

premium in the benchmark and extended models. Since the valuation premium increases
with the maturity of an asset, a natural way to assess the plausibility of our models is to
evaluate their implications for nominal and real bond yields.
This section is organized as follows. First, we price nominal bonds using the model-
implied nominal SDF. Second, we estimate a stochastic process for inflation and modify the
endowment process to allow consumption growth to be correlated with inflation. We then
show that the extended model does a good job at accounting for the nominal yield curve.
Third, we derive the ex-ante expected real return associated with nominal bonds of di§erent
maturities. We argue that, in our model, the upward sloping nature of the nominal yield
curve reflects the upward sloping nature of the ex-ante real yield curve. Fourth, we evaluate
whether the estimated nominal SDF in our extended model has the properties that Backus
and Zin (1994) argue are necessary to simultaneously account for the slope of the yield curve
and the persistence of bond yields.

Model implications Let m$t+1 denote the logarithm of the nominal SDF:
& '
$ λt+1 θ
mt+1 = θ log (δ) + θ log − ∆ct+1 + (θ − 1) rc,t+1 − ιt+1 .
λt

25
Here, ιt+1 denotes the rate of inflation computed as log(Pt+1 /Pt ) where Pt is the economy-
wide nominal price level.
Motivated by the specification in Stock and Watson (2007), we assume that inflation
evolves according to the following stochastic process:20

ιt+1 = µι + χt + σ ι "ιt+1 ,
χt+1 = ρχ χt + σ χt "χt+1 ,
σ 2χt+1 = (1 − vχ ) σ 2χ + vχ σ 2χt + σ χe et+1 .

Here, "ιt+1 and "χt+1 are uncorrelated, i.i.d. standard normal shocks. We modify the en-
dowment process in the extended model to allow inflation to influence the growth rate of
consumption:
" #
∆ct+1 = µc + ρc · ∆ct + αc σ 2t+1 − σ 2 + π cλ "λt+1 + σ t "ct+1 + π cχ σ χt "χt+1 .

The only new element relative to our extended-model specification is the term π cχ σ χt "χt+1 .
Let Θπ denote the vector of parameters related to the inflation process as well as π cχ :
( )
Θπ = µι , σ 2ι , ρχ , vχ , σ 2χ , σ 2χe , π cχ .

Our procedure for estimating Θπ is as follows. Because we are interested in evaluating

the implications of the extended model for the nominal yield curve, we hold fixed all the
parameters that we estimated for that model. We estimate Θπ using the simulated method
of moments. Given a value for Θπ , we simulate π t and ∆ct for 10, 000 × 12 periods. We
time-aggregate the monthly series and compute the model-implied annual moments, Ψπ (Θπ ).
Our estimator is the parameter vector Θ̂π that minimizes the distance between a vector of
empirical moments, Ψπ , and the corresponding model moments, Ψπ (Θ̂π ):

Θ̂π = arg min [Ψπ (Θπ ) − Ψπ ]0 Ω−1

π [Ψπ (Θπ ) − Ψπ ] .
Θπ

Here Ωπ is the variance-covariance matrix of Ψπ . The vector Ψπ is given by:

h , " #- 1 i
Ψπ = E (π t ) , σ (π t ) , corr (π t , π t−j ) , corr (zt , zt−j ) , E [π t − E (π t )]4 4 /σ (yt ) , cov (π t , ∆ct ) .
20
The main di§erences between our specification and Stock and Watson’s (2007) are as follows. First, we
assume that shocks to the transitory component of inflation, σ ι "ιt+1 , are homoscedastic. This simplification
is motivated by Stock and Watson’s finding that there is little variation in the volatility of the transitory
component of inflation. Second, we assume that χt and σ 2χt+1 follow AR(1) processes instead of random
walks. Finally, Stock and Watson (2007) work with a quarterly specification while we work with a monthly
specification.

26
Here, zt is the squared deviation of inflation from its mean: zt = [π t − E (π t )]2 and j =
1,2,3. In practice, we found that it is di¢cult to simultaneously estimate µι and the other
parameters of the model. So, we set µι equal to the average monthly rate of inflation in our
data.
Table 9 reports our results. The parameters ρχ and vχ are precisely estimated and exceed
0.9, so the level and volatility of inflation are quite persistent. Interestingly, the spillover
parameter between inflation and consumption growth is negative (−0.50). However, it is
imprecisely estimated with a standard error of 0.39. This imprecision is consistent with the
large standard error (1.09) of our estimate of the covariance between consumption growth
and inflation (0.68). Clearly, we cannot reject the hypothesis that this covariance is zero.
We reach a similar conclusion when we focus on the correlation between consumption
growth and inflation. While the point estimate of this correlation is 0.41, the standard error
is 0.20 (see Table 11). So, we cannot reject the hypothesis that this correlation is close to
zero.
The correlation for the sample period considered by Piazzesi and Schneider (2007), 1952-
2006, is −0.27 with a standard error of 0.14. The analogue numbers for the sample period
considered by Wachter (2006), 1952-2004, is −0.27 with a standard error of 0.14. So, for
their sample periods, we cannot rule out the hypothesis that this correlation is zero.

6.1. The model’s implications for the nominal yield curve

Table 10 reports the mean and standard deviation of nominal yields on short-term (one
year) Treasury Bills, intermediate-term government bonds (with approximate maturity of
five years), and long-term government bonds (with approximate maturity of twenty years).
This table also reports the implication of the estimated extended model for those moments
of the data. With one exception, the model does a very good at accounting for the level
and slope of the nominal yield curve as well as the standard deviation of nominal yields.
The exception is that the model somewhat understates the standard deviation of the 20-year
nominal yield.
The upward sloping nature of the nominal yield curve reflects the fact that in our model
the ex-ante real yield curve is upward sloping (see Table 13). We derive these ex-ante real
yields using the expected rate of inflation implied by the inflation process discussed above.
According to our model, long-term bonds command a positive risk premium that increases
with the maturity of the bond because longer maturity assets are more exposed to valuation

27
risk. The one-year, five-year and 20-year ex-ante expected real return on the corresponding
nominal bonds are 0.58, 1.46, and 2.81, respectively. The slope of nominal yield curve is
virtually identical to the slope of the ex-ante expected real yield curve.
The implications of our model do not depend sensitively on the value of π cχ . Suppose we
set this parameter to zero, so that there is no interaction between inflation and consumption
growth. As Table 10 shows, the implications of the model for the nominal yield curve are
virtually unchanged.
It is worth contrasting the predictions of our model for yield curves with those of leading
competitors. Long-run risk models of the type pioneered by Bansal and Yaron (2004) can
generate upward sloping nominal yield curves (see, e.g. Bansal and Shaliastovich (2013)).
But their ability to do so depends critically on the presence of a negative covariance between
consumption growth and inflation.
To understand this claim, recall that, as stressed by Piazzesi and Schneider (2007) and
Beeler and Campbell (2012), long-run risk models generally imply negative long-term bond
yields and a negative bond term premium. Indeed, Beeler and Campbell (2012) find that
the return on a 20-year real bond in the Bansal, Kiku and Yaron (2012) model is −0.88.
The intuition for this result is that in a long-run risk model agents are concerned that
consumption growth may be dramatically lower in some future state of the world. Since
real bonds promise a certain payout in all states of the world, they o§er insurance against
this possibility. The longer the maturity of the real bond, the more insurance it o§ers and
the higher is its price. So, the real term premium is downward sloping. For similar reasons,
standard rare-disaster models also imply a downward sloping term structure for real bonds
and a negative real yield on long-term bonds (see, for example the benchmark model in
Nakamura et al. (2013)).21
For these models to generate an upward sloping nominal yield curve there must be forces
at work to counteract the impact of a downward-sloping real yield curve. Consider for ex-
ample Bansal and Shaliastovich (2013) who generate an upward-sloping nominal yield curve
in a long-run risk model. The key to their result is that they work with a sample period
in which there is a negative correlation between the persistent components of consumption
growth and inflation (see their table 4).22 Given their specification, a shock that produces
21
According to Nakamura et al. (2013), these implications can be reversed by introducing the possibility
of default on bonds and assuming that probability of partial default is increasing in the maturity of the
bond.
22
In a similar vein, Wachter (2006) generates an upward sloping nominal yield curve in an external habits

28
persistent increases in inflation generates both low real yields on long-term bonds and per-
sistently low consumption growth. So, long-term nominal bonds are particularly risky and
the nominal yield curve is upward sloping.
A key question is: how robust is the sign of the correlation between inflation and con-
sumption growth? Table 11 displays that correlation, for di§erent horizons, over the sample
periods 1929-2011 and 1952-2011. For the full sample, the point estimates for the one, five,
10 and 20 year correlations are positive. We cannot reject that the one-year correlation is
close to zero. In all cases the correlations are not statistically di§erent from zero. For the
post-war period the point estimates are negative for the one and five-year correlation but
positive for the 10 and 20 year correlation. But, again, none of the correlations are statis-
tically di§erent from zero. We conclude that the strong negative correlation between the
growth rate of consumption and inflation is a weak foundation on which to base an upward
sloping nominal yield curve.
We now discuss other evidence regarding the slope of the real yield curve. Consider first
results based on inflation-index U.K. gilts. Table 12 reports the di§erence between the yield
of a bond with 2.5 years maturity and yields of bonds with 5, 10, 15, and 20 years of maturity.
Our benchmark sample period is 1985 to 2015. In every case, the longer duration bond has,
on average, a higher yield than the bond with duration 2.5 years.23 One might be concerned
about the e§ect of the recent financial crisis on these results. As Table 12 indicates, the
qualitative results are una§ected if we use a sample period from 1985 to 2006.24
Table 12 also reports the slope of the real yield curve for the U.S. based on monthly
TIPS data for the period from July 2004 to November 2015. Consistent with Alvarez and
Jermann (2005), we find that the term structure of real yields is upward sloping. Also, like
Campbell, Shiller and Viceira (2009) we find that the average real yield on long-term TIPS
is positive. Over our sample this average is equal to 1.54 with a standard error of 0.21.
We conclude with some suggestive calculations that are premised on our model-based
result that the covariance of consumption growth with inflation is not a very important
determinant of bond yields. Table 13 reports the mean and standard deviation of ex-ante real
model where innovations to consumption and inflation growth are negatively correlated.
23
The statistical significance of these average di§erences depends on whether we use zero or 12 lags in
computing Newey-West standard errors.
24
Evans (1998) and Piazzesi and Schneider (2006) finds that the U.K. real yield curve is downward sloping
for the periods January 1983-November 1995 and December 1995-March 2006. Our results indicate that
their findings depend on the sample period that they work with.

29
yields on short-term (one year) Treasury Bills, intermediate-term government bonds (with
approximate maturity of five years), and long-term government bonds (with approximate
maturity of twenty years). The ex-ante yields on five (twenty) year bonds are computed as
the di§erence between the five (twenty) year nominal yield and the ex-ante five (twenty)
year inflation rate (yields are expressed on an annualized basis). In all cases, we compute
expected inflation using the inflation process discussed above. The results in Table 13 are
consistent with our other evidence: the ex-ante real yield curve is upward sloping and the
long-term real yield is positive. This table also shows that our model does a good job at
accounting for the level and slope of this measure of the real yield curve.

Accounting for the slope of the yield curve and the persistence of bond yields
Backus and Zin (1994) investigate the properties of the log of the nominal SDF that are
consistent with two key features of nominal yields: the yield curve is upward sloping and
yields are very persistent. These properties can be summarized as follows. First, the log of
the nominal SDF must have negative serial correlation to account for the positive slope of
the nominal yield curve. Second, the log of the nominal SDF must be close to i.i.d. but still
have a small predictable component.25 The log of the nominal SDF in both our benchmark
and extended model satisfy these two properties.
Backus and Zin’s (1994) estimate various ARMA representations for the log of the nom-
inal SDF using data on the mean and autocovariance of bond yields.26 Their best statistical
fit for the nominal SDF is an ARMA(2,3). For such a representation, the properties that
they stress are most easily seen in the impulse response function of the extended model’s log
nominal SDF.
To deduce the implications of our model for the ARMA representation of the nominal
SDF we proceed as follows. First, we generate 1000 synthetic time series for the log of the
nominal SDF, each of length equal to our sample size. For each of the 1000 synthetic time
series, we estimate an ARMA(2,3) for the log of the nominal SDF and compute the impulse
response function to an innovation. Figure 2 displays the average impulse response functions
of the log of the nominal SDF along with a 95 percent confidence interval computed across
the 1000 synthetic time series.
25
See Ljungqvist and Sargent ( 2000) for a detailed exposition of these two properties.
26
Bansal and Viswanathan (1993) follow a similar approach but use a semi-nonparametric approximation
to the SDF.

30
 The estimated SDF for our extended model has the two characteristics that Backus and
Zin (1994) argue to be required to explain an upward sloping nominal yield curve and the
persistence of bond yields. We already showed that our model generates an upward sloping
nominal yield curve. So, we conclude by briefly discussing the serial correlation of the
nominal yields. The first-order autocorrelation of the one, five and 20 year yields are: 0.91
(0.05), 0.95 (0.04) and 0.97 (0.04), where the standard errors are indicated in parenthesis.
The analogue model statistics are 0.81, 0.92, and 0.94. So, our model clearly generates
empirically plausible persistence in nominal bond yields.

Long-term equity premium The benchmark and extended models imply that the dif-
ference between stock returns and 20-year ex-ante real bond yield is roughly 1 percent. In
the data, the di§erence between stock returns and the ex-ante real 20-year bond yields is
roughly 4.16 percent with a standard error of 2.39. So, taking sampling uncertainty into
account, both the benchmark and extended models are consistent with the data.
In our model, the positive premium that equity commands over long-term bonds reflects
the di§erence between an asset of infinite and twenty-year maturity. Consistent with this
perspective, Binsbergen, Hueskes, Koijen, and Vrugt (2011) estimate that 90 (80) percent
of the value of the S&P 500 index corresponds to dividends that accrue after the first five
(ten) years.
It is important to emphasize that the equity premium in our model is not solely driven
by the term premium. One way to see this property is to consider the results of regressing
the equity premium on two alternative measures of excess bond yields. The first measure is
the di§erence between yields on bonds of 20-year and 1-year maturities. The second measure
is the di§erence between yields on bonds of 5-year and 1-year maturities. Table 14 reports
our results. Not surprisingly, the benchmark model does poorly since the expected equity
premium and the term premia are constant. The extended model does better in the sense
that the model slopes are within one standard error of the slope point estimates. Also, both
models are consistent with the fact that the R2 in these regressions are quite low.
We conclude with an interesting observation made by Binsbergen, Brandt, and Koijen
(2012). Using data over the period 1996 to 2009, these authors decompose the S&P500 index
into portfolios of short-term and long-term dividend strips. The first portfolio entitles the
holder to the realized dividends of the index for a period of up to three years. The second
portfolio is a claim on the remaining dividends. Binsbergen et al (2012) find that the short-

31
term dividend portfolio has a higher risk premium than the long-term dividend portfolio, i.e.
there is a negative stock term premium. They argue that this observation is inconsistent with
habit-formation, long-run risk models and standard of rare-disaster models. Our model, too,
has di¢culty in accounting for the Binsbergen et al (2012) negative stock term premium.27
Of course, our sample is very di§erent from theirs and their negative stock term premium
result is heavily influenced by the recent financial crisis (see Binsbergen et al (2011)). Also,
Boguth, Carlson, Fisher, and Simutin (2012) argue that the Binsbergen et al (2012) results
may be significantly biased because of the impact of small pricing frictions.

7. Conclusion

In this paper we argue that allowing for demand shocks substantially improves the per-
formance of representative-agent, asset-pricing models. Specifically, it allows the model to
account for the equity premium, bond term premia, and the correlation puzzle with low
degrees of estimated risk aversion. According to our estimates, valuation risk is by far the
most important determinant of the equity premium and the bond term premia.
While our model is consistent with many features of stock and bond returns, it does
have a number of empirical shortcomings. For example, it is di¢cult for the model to
generate the non-negative correlation observed between the price-dividend ratio and the
risk-free rate. As discussed above, long-run risk models generate correlations that are much
larger than observed in the data. This result suggests that incorporating long-run risk into
our model would improve the model’s performance. Similarly, our model understates the
positive correlation between stock returns at time t and consumption growth at time t + 1.
We suspect that more sophisticated information structures in which agents receive news
about future movements in consumption could help resolve this problem.

27
Recently, Nakamura et al (2013) show that a time-vaying rare disaster model in which the component
of consumption growth due to a rare disaster follows an AR(1) process is consistent with the Binsbergen et
al (2012) results. Belo et al. (2013) show that the Binsbergen et al. (2012) result can be reconciled in a
variety of models if the dividend process is replaced with processes that generate stationary leverage ratios.

32
References

[1] Alvarez, Fernando, and Urban J. Jermann, “Using Asset Prices to Measure the Persis-
tence of the Marginal Utility of Wealth,” Econometrica, 73 (6): 1977—2016, 2005.

[2] Backus, David K., and Stanley E. Zin. Reverse engineering the yield curve. No. w4676.
National Bureau of Economic Research, 1994.

[3] Backus, David, Bryan R. Routledge, and Stanley E. Zin, The Cyclical Component of
US Asset Returns, manuscript, New York University, April 2010.

[4] Bai, Yan, José Victor Ríos-Rull and Kjetil Storesletten “Demand Shocks as Productivity
Shocks,” manuscript, University of Minnesota, 2014.

[5] Bansal, Ravi, and Ivan Shaliastovich. “A Long-run Risks explanation of predictability
puzzles in bond and currency markets,” Review of Financial Studies, 26.1 (2013): 1-33.

[6] Bansal, Ravi and S. Viswanathan “No Arbitrage and Arbitrage Pricing: A New Ap-
proach,” The Journal of Finance, 48:4, 1231-1262, 1993.

[7] Bansal, Ravi and Amir Yaron, “Risks for the Long Run: A Potential Resolution of
Asset Pricing Puzzles,” The Journal of Finance, 59 (4): 1481-509, 2004.

[8] Bansal, Ravi, Dana Kiku and Amir Yaron, “An Empirical Evaluation of the Long-Run
Risks Model for Asset Prices,” Critical Finance Review, 1: 183—221, 2012.

[9] Bansal, Ravi, Dana Kiku, and Amir Yaron. “Risks for the Long Run: Estimation with
Time Aggregation,” manuscript, Wharton School, 2013.

[10] Barberis, Nicholas, Andrei Shleifer, and Robert Vishny. “A Model of Investor Senti-
ment,” Journal of Financial Economics 49 (3): 307-43, 1998.

[11] Barro, Robert J. “Rare Disasters and Asset Markets in the Twentieth Century,” Quar-
terly Journal of Economics, 121: 823—66, 2006.

[12] Barro, Robert J., and José F. Ursúa. “Rare Macroeconomic Disasters,” working paper
No. 17328, National Bureau of Economic Research, 2011.

[13] Basu, Susanto and Brent Bundick “Endogenous Volatility at the Zero Lower Bound:
Implications for Stabilization Policy,” manuscript, Boston College, 2015.

33
[14] Beeler, Jason, and John Y. Campbell. “The Long-Run Risks Model and Aggregate Asset
Prices: An Empirical Assessment,” Critical Finance Review, 1: 141—82, 2012.

[15] Belo, Frederico, Pierre Collin-Dufresne, and Robert S. Goldstein. “Dividend Dynamics
and the Term Structure of Dividend Strips,” manuscript, University of Minnesota, 2013.

[16] Binsbergen, Jules H. van, Michael W. Brandt, and Ralph S.J. Koijen. “On the Timing
and Pricing of Dividends,” American Economic Review, 102 (4): 1596-618, 2012.

[17] Binsbergen, Jules H. van, Wouter Hueskes, Ralph Koijen, and Evert Vrugt. “Equity
Yields,” working paper N. 17416, National Bureau of Economic Research, 2011.

[18] Boldrin, Michele, Lawrence J. Christiano, and Jonas Fisher. “Habit Persistence, Asset
Returns, and the Business Cycle,” American Economic Review, 91: 149-66, 2001.

[19] Borovička, Jaroslav, and Lars Peter Hansen. “Examining Macroeconomic Models
through the Lens of Asset Pricing,” manuscript, University of Chicago, 2011.

[20] Borovička, Jaroslav, Lars Peter Hansen, Mark Hendricks, and José A. Scheinkman.
“Risk-Price Dynamics,” Journal of Financial Econometrics 9 (1): 3-65, 2011.

[21] Boguth, Oliver, Murray Carlson, Adlai Fisher, and Mikhal Simutin. “Leverage and the
Limits of Arbitrage Pricing: Implications for Dividend Strips and the Term Structure
of Equity Risk Premia,” manuscript, Arizona State University, 2012.

[22] Boudoukh, Jacob, Matthew Richardson, and Robert F. Whitelaw. “The Myth of Long-
Horizon Predictability,” Review of Financial Studies 24 (4): 1577-605, 2008.

[23] Campbell, John Y. “Bond and Stock Returns in a Simple Exchange Model,” The Quar-
terly Journal of Economics 101 (4): 785-803, 1986.

[24] Campbell, John Y. “Consumption-Based Asset Pricing,” in Handbook of the Economics

of Finance, Eds. G.M. Constantinides, M. Harris and R. Stulz, 801-85, 2003.

[25] Campbell, John Y., and John Ammer. “What Moves the Stock and Bond Markets? A
Variance Decomposition for Long-Term Asset Returns,” The Journal of Finance 48 (1):
3-37, 1993.

34
[26] Campbell, John Y., and John H. Cochrane. “By Force of Habit: A Consumption-based
Explanation of Aggregate Stock Market Behavior,” Journal of Political Economy, 107
(2): 205-51, 1999.

[27] Campbell, John Y., and Robert J. Shiller. “The Dividend-price Ratio and Expectations
of Future Dividends and Discount Factors,” Review of Financial Studies 1 (3): 195-228,
1988.

[28] Campbell, John, Robert Shiller, and Luis Viceira. “Understanding Inflation-Indexed
Bond Markets,” Brookings Papers on Economic Activity, 79-120, 2009.

[29] Cochrane, John H. “Explaining the Variance of Price Dividend Ratios,” Review of Fi-
nancial Studies 5: 243-80, 1992.

[30] Cochrane, John H. Asset Pricing, Princeton University Press, 2001.

[31] Cochrane, John H. “Presidential Address: Discount Rates,” The Journal of Finance,
66 (4): 1047-108, 2011.

[32] Cochrane, John H., and Lars Peter Hansen. “Asset Pricing Explorations for Macro-
economics,” NBER Macroeconomics Annual 1992, Volume 7, pp. 115-82. MIT Press,
1992.

[33] Correia, Isabel, Emmanuel Farhi, Juan Pablo Nicolini, and Pedro Teles. “Unconven-
tional Fiscal Policy at the Zero Bound,” American Economic Review, 103 (4): 1172-211,
2013.

[34] Dew-Becker, Ian “A Model of Time-Varying Risk Premia with Habits and Production,”
manuscript, Northwestern University, 2014.

[35] Dumas, Bernard, Alexander Kurshev, and Raman Uppal. “Equilibrium Portfolio Strate-
gies in the Presence of Sentiment Risk and Excess Volatility,” The Journal of Finance,
64 (2): 579-629, 2009.

[36] Du¢e, Darrell, and Larry G. Epstein. “Stochastic Di§erential Utility,” Econometrica,
60: 353-94, 1992.

35
[37] Eggertsson, Gauti. “Monetary and Fiscal Coordination in a Liquidity Trap,” Chapter
3 of Optimal Monetary and Fiscal Policy in the Liquidity Trap, Ph.D. dissertation,
Princeton University, 2004.

[38] Eggertsson, Gauti B., and Michael Woodford. “Zero Bound on Interest Rates and Op-
timal Monetary Policy,” Brookings Papers on Economic Activity, 1: 139-233, 2003.

[39] Epstein, Larry G., and Stanley E. Zin. “Substitution, Risk Aversion, and the Temporal
Behavior of Consumption and Asset Returns: A Theoretical Framework,” Econometrica,
57 (4): 937-69, 1989.

[40] Epstein, Larry G., and Stanley E. Zin. “Substitution, Risk Aversion, and the Tempo-
ral Behavior of Consumption and Asset Returns: An Empirical Analysis,” Journal of
Political Economy, 99: 263-86, 1991.

[41] Epstein, Larry, Emmanuel Farhi, and Tomasz Strzalecki “How Much Would You Pay
to Resolve Long-Run Risk?,” American Economic Review, 104, (2014) 2680—2697

[42] Fernandez-Villaverde, Jesus, Pablo Guerron-Quintana, Keith Kuester, and Juan Rubio-
Ramírez. “Fiscal Volatility Shocks and Economic Activity,” forthcoming, American Eco-
nomic Review, 2015.

[43] Gabaix, Xavier, and Matteo Maggiori. “International Liquidity and Exchange Rate
Dynamics,” manuscript, New York University, 2013.

[44] Garber, Peter M., and Robert G. King. “Deep Structural Excavation? A Critique of
Euler Equation Methods,” working paper No. 31, National Bureau of Economic Re-
search, 1983.

[45] Gârleanu, Nicolae, Leonid Kogan, and Stavros Panageas. “Displacement Risk and Asset
Returns,” Journal of Financial Economics, 105 (3): 491-510, 2012.

[46] Gourio, François. “Disaster Risk and Business Cycles,” American Economic Review,
102 (6): 2734-66, 2012.

[47] Grossman, Sanford J., Angelo Melino, and Robert J. Shiller. “Estimating the Continu-
ous Time Consumption Based Asset Pricing Model”, Journal of Business and Economic
Statistics, 5: 315-28, 1987.

36
[48] Hall, Robert E. “High Discounts and High Unemployment,” working paper No. 19871,
National Bureau of Economic Research, 2014.

[49] Hansen, Lars Peter, John C. Heaton, and Nan Li. “Consumption Strikes Back? Mea-
suring Long-Run Risk,” Journal of Political Economy, 116 (2): 260-302, 2008.

[50] Hansen, Lars Peter, and Ravi Jagannathan. “Implications of Security Market Data for
Models of Dynamic Economies,” Journal of Political Economy, 99 (2): 225-62, 1991.

[51] Hansen, Lars Peter, and Kenneth J. Singleton. “Generalized Instrumental Variables
Estimation of Nonlinear Rational Expectations Models,” Econometrica, 50, 1269-86,
1982.

[52] Huo, Zhen, and José-Víctor Ríos-Rull. “Paradox of Thrift Recessions,” working paper
No. 19443, National Bureau of Economic Research, 2013.

[53] Kandel, Shmuel, and Robert F. Stambaugh. “Expectations and Volatility of Consump-
tion and Asset Returns,” Review of Financial Studies 3 (2): 207-32, 1990.

[54] Kreps, David M., and Evan L. Porteus. “Temporal Resolution of Uncertainty and Dy-
namic Choice Theory,” Econometrica, 46, 185-200, 1978.

[55] Kung, Howard “Macroeconomic Linkages between Monetary Policy and the Term Struc-
ture of Interest Rates,” Journal of Financial Economics, Volume 115, Issue 1, pages
42—57. 2015.

[56] Lettau, Martin, and Sydney C. Ludvigson. “Shocks and Crashes,” manuscript, New
York University, 2011.

[57] Lucas Jr, Robert E. “Asset Prices in an Exchange Economy,” Econometrica, 46, 1429-45,
1978.

[58] Luo, Victor. “Using Bond Yields to Discriminate Between Long-run and Valuation Risk
Models,” manuscript, Northwestern University, 2014.

[59] Lynch, Anthony W. “Decision frequency and synchronization across agents: Impli-
cations for aggregate consumption and equity return,” The Journal of Finance 51.4:
1479-1497, 1996.

37
[60] Ljungqvist, Lars and Thomas Sargent Recursive Macroeconomic Theory, 2nd edition,
Massachusetts Institute of Technology, 2000.

[61] Maurer, Thomas A. “Is Consumption Growth Merely a Sideshow in Asset Pricing?,”
manuscript, London School of Economics, 2012.

[62] Mehra, Rajnish, and Edward C. Prescott. “The Equity Premium: A Puzzle,” Journal
of Monetary Economics, 15 (2): 145-61, 1985.

[63] Nakamura, Emi, Jón Steinsson, Robert Barro, and José Ursúa. “Crises and Recover-
ies in an Empirical Model of Consumption Disasters,” American Economic Journal:
Macroeconomics, 5 (3), 35-74, 2013.

[64] Newey, Whitney K., and Kenneth D. West. “A Simple, Positive Semi-definite, Het-
eroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica, 55
(3): 703-8, 1987.

[65] Normandin, Michel, and Pascal St-Amour. “Substitution, Risk Aversion, Taste Shocks
and Equity Premia,” Journal of Applied Econometrics, 13 (3): 265-81, 1998.

[66] Parker, Jonathan A. “The Consumption Risk of the Stock Market,” Brookings Papers
on Economic Activity, 2: 279-348, 2001.

[67] Pavlova, Anna, and Roberto Rigobon. “Asset Prices and Exchange Rates,” Review of
Financial Studies, 20 (4): 1139-81, 2007.

[68] Piazzesi, Monika, and Martin Schneider. “Equilibrium Yield Curves,” NBER Macroeco-
nomics Annual 2006, 21: 389-472, 2007.

[69] Primiceri, Giorgio. “Time Varying Structural Vector Autoregressions and Monetary
Policy,” The Review of Economic Studies, 72: 821-852, 2005.

[70] Rietz, Thomas A. “The Equity Risk Premium a Solution,” Journal of Monetary Eco-
nomics, 22 (1): 117-31, 1988.

[71] Shiller, Robert. “Do Stock Prices Move Too Much to be Justified by Subsequent Divi-
dends,” American Economic Review, 71: 421-36, 1981.

[72] Skiadas, Costis. “Asset Pricing Theory,” Princeton University Press, 2009.

38
[73] Stambaugh, Robert F. “Predictive Regressions,” Journal of Financial Economics, 54,
375-421, 1999.

[74] Stock, James H. and Mark W. Watson “Why Has U.S. Inflation Become Harder to
Forecast?,” Journal of Money, Credit and Banking Vol. 39, pp. 3-33, Feb., 2007.

[75] Stockman, Alan C., and Linda L. Tesar. “Tastes and Technology in a Two-Country
Model of the Business Cycle: Explaining International Comovements,” The American
Economic Review, 85 (1): 168-185, 1995.

[76] Wachter, Jessica A. “Solving Models with External Habit,” Finance Research Letters
2.4: 210-226, 2005.

[77] Wachter, Jessica A. “A Consumption-based Model of the Term Structure of Interest

Rates,” Journal of Financial Economics 79.2: 365-399, 2006.

[78] Wachter, Jessica A. “Can Time-varying Risk of Rare Disasters Explain Aggregate Stock
Market Volatility?, Journal of Finance, 68, 987—1035, 2013.

[79] Weil, Philippe. “The Equity Premium Puzzle and the Risk-free Rate Puzzle,” Journal
of Monetary Economics, 24 (3): 401-421, 1989.

39
 Table61

Correlation6Between6Stock6Returns6and6Per6Capita
Growth6Rates6of6Fundamentals

Panel A, 1929-2011

Horizon Consumption Output Dividends Earnings

16year 80.05 0.05 0.05 0.10

(0.12) (0.10) (0.11) (0.10)
56years 0.00 0.00 0.30 0.20
(0.14) (0.12) (0.13) (0.13)
106years 80.11 80.09 0.59 0.30
(0.20) (0.14) (0.14) (0.11)

Panel B, 1871-2006

Horizon Consumption Output Dividends Earnings

16year 0.09 0.14 80.04 0.13

(0.09) (0.10) (0.10) (0.10)
56years 0.40 0.25 0.38 0.44
(0.18) (0.14) (0.15) (0.18)
106years 0.25 0.00 0.64 0.41
(0.18) (0.11) (0.17) (0.13)
 Table"2

Correlation"Between"Stock"Returns"and"Per"Capita
Growth"Rates"of"Fundamentals

NIPA measures of consumption, 1929-2011

Horizon Durables Non-durables Non-durables and services

1"year '0.04 0.05 0.27

(0.13) (0.14) (0.14)

5"years 0.07 '0.08 0.18

(0.11) (0.09) (0.08)

10"years 0.21 '0.20 0.06

(0.15) (0.13) (0.14)
 Table-3

Correlation-Between-Stock-Returns-and-Per-Capita
Growth-Rates-of-Fundamentals

Horizon Consumption- Output Horizon Consumption- Output

G7 OECD

1-year 0.013 0.178 1-year 0.038 0.125

(0.061) (0.079) (0.033) (0.045)

5-years 0.189 0.345 5-years 0.130 0.272

(0.105) (0.091) (0.067) (0.067)

10-years 0.280 0.397 10-years 0.151 0.262

(0.130) (0.117) (0.103) (0.104)
 Table 4

Parameter Estimates and Standard Errors

Models estimated with ex-ante real bond yields

Parameter Benchmark,model Extended,model Parameter Benchmark,model Extended,model

γ 1.516 2.3961 ρ 0.99132 0.99168

(0.029616) (0.11244) (0.00056369) (0.00078787)

ψ 1.4567 2.2107 σλ 0.00058631 0.00038594

(0.041753) ,(0.17437) ,(2.8373e?05) (1.8036e?05)

δ 0.99795 0.99796 Implied value of θ ?1.6459 ?2.5492

,(0.0017074) ,(8.0205e?05) ,(0.19603) (0.29743)

σc 0.0069004 0.0046033 ν 0.00 0.99717

,(0.00014956) (0.00013664) (0.00030196)

µ 0.0015644 0.0010428 σω 0.00 1.68E?06

,(4.4976e?05) ,(7.8579e?05) (4.9166e?08)

πcλ 0.00 ?0.0029185 αc 0.00 ?95.3767

(0.00023329) (4.1769)

σd 0.017103 0.025274 αd 0.00 ?151.8814

,(0.00039626) (0.19243) (22.7162)

πdc 0.00049127 ?1.3205 ρc 0.00 0.094708

,(0.00044218) (0.2236) (0.045391)

πdλ 0.00 ?0.010907 ρd 0.00 0.33427

(0.0015436) (0.11204)

ση 0.00 0.0077828
(0.00024356)
 Table*5

Model*estimated*with*exMante*real*bond*yields

Benchmark*
Moments Data*(constrained) Data*(unconstrained) Extended*model
model

Average*growth*rate*of* 1.4941 2.2197 1.87 1.38

of*consumption (0.30) (0.22413)

Average*growth*rate 1.4941 0.12538 1.87 1.38

of*dividends (0.30395) (0.65732)

Standard*deviation*of*the*growth 2.1865 2.1624 2.38 2.65

rate*of*consumption (0.34494) (0.31345)

Standard*deviation*of*the*growth 6.914 6.5181 5.85 7.25

rate*of*dividends (1.0409) (1.1219)

Contemporaneous*correlation 0.2215 0.23722 0.03 0.29

between*consumption*growth*and (0.1006) (0.10649)
dividend*growth

FirstMorder*serial*correlation 0.20014 0.1584 M0.01 0.30

of*consumption (0.1475) (0.1240)

Correlation*between*stock*returns M0.1555 M0.26065 0.00 M0.03

*and*oneMperiod*lagged*consumption*growth (0.1070) (0.0999)

Correlation*between*stock*returns 0.49504 0.50929 0.01 0.17

*and*oneMperiod*lead*consumption*growth (0.0670) (0.0490)

Average*return*to*equities 7.8252 6.3948 5.94 4.00

(1.7396) (1.8804)

Standard*deviation*of*return*to 17.2537 17.6831 17.37 19.23

equities (1.3783) (1.461)

Average*riskMfree*rate 0.12613 0.12087 0.13 0.13

(0.8088) (0.8088)

Standard*deviation*of*the*riskMfree 3.5615 3.4989 4.30 4.06

rate (0.81818) (0.83447)

FirstMorder*serial*correlation*of*the 0.51802 0.52154 0.90 0.51

riskMfree*rate (0.068326) (0.070816)

Equity*premium 7.6991 6.2739 5.81 3.88

(1.8141) (1.9485)

Average*priceMdividend*ratio 3.3761 3.3775 3.22 3.32

(0.14887) (0.14887)

Standard*deviaton*of*priceM 0.47324 0.44502 0.31 0.46

dividend*ratio (0.074827) (0.081116)

FirstMorder*serial*correlation*of 0.95041 0.93257 0.84 0.89

price*dividend*ratio (0.034499) (0.042679)

Contemporaneous*correlation *0.2736, 0.30705 M0.95 M0.20

between*risk*free*rate*and (0.11356) (0.11909)
priceMdividend*ratio
 Table*6

Correlation*Between*Stock*Returns*and*Per*Capita
Growth*Rates*of*Consumption*and*Dividends

Model*estimated*with*ex9ante*real*bond*yields

Consumption*growth

Data* Data* Benchmark*

Extended*model
(constrained) (unconstrained) model
Moments

19year*correlation*between 90.07 90.05 0.02 90.19

equity*returns*and (0.11) (0.12)
consumption*growth

59year*correlation*between 90.01 90.05 0.02 90.05

equity*returns*and *(0.16) (0.13)
consumption*growth

109year*correlation*between 90.08 90.21 0.03 0.11

equity*returns*and *(0.30) *(0.21)
consumption*growth

Dividend*growth

19year*correlation*between 0.08 0.05 0.34 90.10

equity*returns*and *(0.11) *(0.11)
dividend*growth

59year*correlation*between 0.22 0.29 0.32 0.03

equity*returns*and *(0.14) (0.14)
dividend*growth

109year*correlation*between 0.51 0.59 0.38 0.20

equity*returns*and *(0.22) (0.15)
dividend*growth
 Table)7

Predictability)of)Excess)Returns)by)Price1dividend)Ratio)at)Various)Horizons

Model)estimated)with)ex1ante)real)bond)yields

Benchmark*model

Model) Model) Model) Model)

Data (median) (plim) Data (median) (plim)
Slope)coefficient R1square (%)of)values)larger)than
R1square)in)data)
1)year 10.10 10.05 0.008 0.07 0.01 0.0002
(0.03) (0.03)

3)years 10.27 10.14 0.027 0.14 0.03 0.001

(0.06) (0.08)

5)years 10.42 10.21 0.034 0.26 0.04 0.001

)(0.12) (0.05)

Extended*model

1)year 10.10 10.07 10.04 0.07 0.03 0.01

(0.03) ()0.13)

3)years 10.27 10.19 10.10 0.14 0.08 0.04

(0.06) (0.23)

5)years 10.42 10.30 10.16 0.26 0.12 0.06

)(0.12) (0.15)
 Table-8

Correlation-of-price)dividend-ratio-and-ex)ante-risk)free-rate

Sample Full)sample Pre-1983 Post-1983

Canada 1952)2014 0.0223 0.288 )0.213

(0.139) (0.136) (0.239)

France 1932)2014 )0.484 )0.576 0.132

(0.152) (0.124) (0.179)

Germany 1968)2014 )0.115 )0.382 )0.117

(0.105) (0.189) (0.277)

Italy 1940)2014 )0.588 )0.725 0.344

(0.166) (0.0823) (0.2)

Japan 1960)2014 0.342 )0.128 0.682

(0.176) (0.16) (0.109)

U.K. 1934)2014 0.246 0.147 )0.253

(0.122) (0.0819) (0.216)

US 1939)2011 0.30705 0.39 )0.42

(0.11909) (0.113) -(0.227)
 Table 9

Parameter Estimates and Standard Errors

Inflation Process

Parameter Parameter

µi 0.003 σχ2 2.08E-07

(1.09E-07)

σi 2 4.45E-07 σχe2 3.38E-14

(5.21E-07) (7.45E-16)

ρχ 0.9111 πcχ -0.50433

(0.0027) (0.3767)

νχ 0.99625
(0.00048)
 Table 10

Term Structure of Nominal Bond Yields

Extended
Extended model,
Data
model
Moments πcχ = 0

Average yield

Long-term bonds 5.17 6.31 6.30

(0.89)

Intermediate-term bond 4.59 4.96 4.95

(0.95)

Short-term bond 3.53 4.18 4.16

(0.93)

Standard deviation of yield

Long-term bonds 2.63 1.48 1.48

(0.49)

Intermediate-term bond 2.87 2.52 2.52

(0.52)

Short-term bond 2.63 3.56 3.55

(0.49)
 Table 11

Correlation between inflation and real consumption growth

Sample: Sample:
Horizon 1929-2011 1952-2011

1 year 0.41 -0.17

(0.20) (0.14)

5 years 0.26 -0.08

(0.21) (0.16)

10 years 0.26 0.01

(0.19) (0.15)

20 years 0.23 0.23

(0.19) (0.28)
 Table 12

Slope of the yield curve on U.K. gilts and U.S. TIPS

Newey-West standard errors

Zero lags in 12 lags in

Sample Mean
weighting matrix weighting matrix

Difference between long-term yields and 2.5 year yields, inflation-indexed U.K. gilts

5 year minus 2.5 year 1985-2015 0.19 0.03 0.10

1985-2006.Q3 0.09 0.03 0.11

10 year minus 2.5 year 1985-2015 0.38 0.05 0.17

1985-2006.Q3 0.22 0.05 0.17

15 year minus 2.5 year 1985-2015 0.42 0.06 0.20

1985-2006.Q3 0.26 0.06 0.19

20 year minus 2.5 year 1985-2015 0.42 0.08 0.26

1985-2006.Q3 0.24 0.08 0.27

Difference between long-term yields and 5-year yields, U.S. TIPS

10 year minus 5 year 2004-2015 0.54 0.03 0.09

20 year minus 5 year 2004-2015 0.95 0.05 0.15

 Table 13

Term Structure of Ex-ante Real Bond Yields

Data (ex-ante Extended

Moments real yields) model

Average yield

Long-term bonds 2.90 2.81

(0.84)

Intermediate-term bond 1.93 1.46

(0.99)

Short-term bond 0.46 0.58

(0.78)

Return to equity minus long-term bond yield 2.54 1.13

(2.09)

Standard deviation of yield

Long-term bonds 2.59 1.38

(0.52)

Intermediate-term bond 3.14 2.40

(0.57)

Short-term bond 3.85 2.89

(0.77)

Return to equity minus long-term bond yield 20.09 18.55

(1.96)
 Table 14

Regressions of Excess Stock Returns on Long-term Bond Yields in Excess of Short Rate

Data 1929-2011 Extended model

Long-term government bonds (20 years)

R-square 0.03 0.03

Slope 0.91 0.98

(0.46)

Intermediate term government bonds (5 years)

R-square 0.02 0.02

Slope 0.72 1.00

(0.51)
 Figure 1: Correlograms of stock returns and fundamentals

Panel A

United States - Correlogram of Returns (t) and Consumption Growth (t + τ )

1

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
-10 -8 -6 -4 -2 0 2 4 6 8 10
τ

United States - Correlogram of Returns (t) and GDP Growth (t + τ )

1

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
-10 -8 -6 -4 -2 0 2 4 6 8 10
τ
 Figure 1: Correlograms of stock returns and fundamentals

Panel B

Canada: corr(rt , ∆ct+τ ) Canada: corr(rt , ∆yt+τ )

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
France: corr(rt , ∆ct+τ ) France: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
Italy: corr(rt , ∆ct+τ ) Italy: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
 Figure 1: Correlograms of stock returns and fundamentals

Panel C

Germany: corr(rt , ∆ct+τ ) Germany: corr(rt , ∆yt+τ )

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
Japan: corr(rt , ∆ct+τ ) Japan: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
United Kingdom: corr(rt , ∆ct+τ ) United Kingdom: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
 Figure 1: Correlograms of stock returns and fundamentals

Panel A

United States - Correlogram of Returns (t) and Consumption Growth (t + τ )

1

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
-10 -8 -6 -4 -2 0 2 4 6 8 10
τ

United States - Correlogram of Returns (t) and GDP Growth (t + τ )

1

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
-10 -8 -6 -4 -2 0 2 4 6 8 10
τ
 Figure 1: Correlograms of stock returns and fundamentals

Panel B

Canada: corr(rt , ∆ct+τ ) Canada: corr(rt , ∆yt+τ )

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
France: corr(rt , ∆ct+τ ) France: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
Italy: corr(rt , ∆ct+τ ) Italy: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
 Figure 1: Correlograms of stock returns and fundamentals

Panel C

Germany: corr(rt , ∆ct+τ ) Germany: corr(rt , ∆yt+τ )

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
Japan: corr(rt , ∆ct+τ ) Japan: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
United Kingdom: corr(rt , ∆ct+τ ) United Kingdom: corr(rt , ∆yt+τ )
1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-10 -5 0 5 10 -10 -5 0 5 10
τ τ
8. Appendix
8.1. Appendix A

In this appendix, we solve the model in Section 3 analytically for the case of CRRA utility.
Let Ca,t denote the consumption of the representative agent at time t. The representative
agent solves the following problem:
X
1 1−γ
Ca,t+i
i
Ut = max Et δ λt+i ,
i=0
1−γ

subject to the flow budget constraints

Wa,i+1 = Rc,i+1 (Wa,i − Ca,i ) ,

for all i ≥ t. The variable Rc,i+1 denotes the gross return to a claim that pays the aggregate
consumption as in equation (3.8), financial wealth is Wa,i = (Pc,i + Ci ) Sa,i , and Sa,i is the
number of shares on the claim to aggregate consumption held by the representative agent.
The first-order condition for Sa,t+i+1 is:

−γ
" −γ
#
δ i λt+i Ca,t+i = Et δ i+1 λt+i+1 Ca,t+i+1 Rc,i+1 .

In equilibrium, Ca,t = Ct , Sa,t = 1. The equilibrium value of the intertemporal marginal

rate of substitution is: & '−γ
λt+1 Ct+1
Mt+1 =δ . (8.1)
λt Ct
The Euler equation for stock returns is the familiar,

Et [Mt+1 Rc,t+1 ] = 1.

We now solve for Pc,t . It is useful to write Rc,t+1 as

& '
(Pc,t+1 /Ct+1 + 1) Ct+1
Rc,t+1 = .
Pc,t /Ct Ct
In equilibrium: * & '& '+
Pc,t+1 Ct+1 Pc,t
Et Mt+1 +1 = . (8.2)
Ct+1 Ct Ct
Replacing the value of Mt+1 in equation (8.2):
" & '−γ & '& '#
λt+1 Ct+1 Pc,t+1 Ct+1 Pc,t
Et δ +1 = .
λt Ct Ct+1 Ct Ct

40
Using the fact that λt+1 /λt is known as of time t we obtain:
* & '+
λt+1 " c
#1−γ Pct+1 Pct
δ Et exp µ + σ c "t+1 +1 = .
λt Ct+1 Ct

We guess and verify that Pct+1 /Ct+1 is independent of "ct+1 . This guess is based on the
fact that the model’s price-consumption ratio is constant absent time-preference shocks.
Therefore, & '
λt+1 ( ) Pc,t+1 Pc,t
δ exp (1 − γ) µ + (1 − γ)2 σ 2c /2 Et +1 = . (8.3)
λt Ct+1 Ct
We now guess that there are constants k0 , k1 ,..., such that
Pc,t 2
= k0 + k1 (λt+1 /λt ) + k2 (λt+1 /λt )1+ρ + k3 (λt+1 /λt )1+ρ+ρ + ... (8.4)
Ct
Using this guess,
& '
Pc,t+1
Et +1
Ct+1
. " " ## " " ##1+ρ /
= Et k0 + k1 (λt+1 /λt )ρ exp σ λ "λt+1 + k2 (λt+1 /λt )ρ exp σ λ "λt+1 + ... + 1
" # " #
= k0 + k1 (λt+1 /λt )ρ exp σ 2λ /2 + k2 (λt+1 /λt )ρ(1+ρ) exp (1 + ρ)2 σ 2λ /2 + ... + 1. (8.5)

Substituting equations (8.4) and (8.5) into equation (8.3) and equating coe¢cients leads to
the following solution for the constants ki :

k0 = 0,
( )
k1 = δ exp (1 − γ) µ + (1 − γ)2 σ 2c /2 ,

and for n ≥ 2
nh " #2 " #2 i 2 o
kn = k1n exp 1 + (1 + ρ)2 + 1 + ρ + ρ2 + ... + 1 + ... + ρn−2 σ λ /2 .

We assume that the series {kn } converges, so that the equilibrium price-consumption ratio
is given by equation (8.4). Hence, the realized return on the consumption claim is

Ct+1 k1 (λt+2 /λt+1 ) + k2 (λt+2 /λt+1 )1+ρ + ... + 1

Rc,t+1 = . (8.6)
Ct k1 (λt+1 /λt ) + k2 (λt+1 /λt )1+ρ + ...

The equation that prices the one-period risk-free asset is:

Et [Mt+1 Rf,t+1 ] = 1.

41
Taking logarithms of both sides of this equation and noting that Rf,t+1 is known at time t,
we obtain:
rf,t+1 = − log Et (Mt+1 ) .

Using equation (8.1),

λt+1 ( " " ##)
Et (Mt+1 ) = δ Et exp −γ µ + σ c "ct+1
λt
λt+1 " #
= δ exp −γµ + γ 2 σ 2c /2 .
λt
Therefore,
rf,t+1 = − log (δ) − log (λt+1 /λt ) + γµ − γ 2 σ 2c /2.

Using equation (3.4), we obtain

& '
( ) σ 2λ /2
E (λt+1 /λt )−1 = exp .
1 − ρ2
We can then write the unconditional risk-free rate as:
& 2 '
σ λ /2
E (Rf,t+1 ) = exp δ −1 exp(γµ − γ 2 σ 2c /2).
1 − ρ2
Thus, the equity premium is given by:
& 2 '
σ λ /2 ( " # )
E [(Rc,t+1 ) − Rf,t+1 ] = exp 2
δ −1 exp(γµ − γ 2 σ 2c /2) exp γσ 2c − 1 ,
1−ρ
which can be written as:
( " # )
E [(Rc,t+1 ) − Rf,t+1 ] = E (Rf,t+1 ) exp γσ 2c − 1 .

8.2. Appendix B

This appendix provides a detailed derivation of the equilibrium of the model economy where
the representative agent has Epstein-Zin preferences and faces time-preference shocks. The
agent solves the following problem:
h " #1−1/ i1/(1−1/ )
1−1/ ∗
U (Wt ) = max λt Ct +δ Ut+1 , (8.7)
Ct

( " #)1/(1−γ)
∗
where Ut+1 = Et U (Wt+1 )1−γ . The optimization is subject to the following budget
constraint:
Wt+1 = Rc,t+1 (Wt − Ct ) .

42
The agent takes as given the stochastic processes for the return on the consumption claim
Rc,t+1 and the preference shock λt+1 . For simplicity, we omit the dependence of life-time
utility on the processes for λt+1 and Rc,t+1 .
The first-order condition with respect to consumption is,
−1/ " ∗ #−1/ ( " #)1/(1−γ)−1 " #
λt Ct = δ Ut+1 Et U (Wt+1 )1−γ Et U (Wt+1 )−γ U 0 (Wt+1 ) Rc,t+1 ,

and the envelope condition is:

" ∗ #−1/ ( " #)1/(1−γ)−1 " #
U 0 (Wt ) = U (Wt )1/ δ Ut+1 Et U (Wt+1 )1−γ Et U (Wt+1 )−γ U 0 (Wt+1 ) Rc,t+1 .

Combining the first-order condition and the envelope condition we obtain:

−1/
U 0 (Wt ) = U (Wt )1/ λt Ct . (8.8)

This equation can be used to replace the value of U 0 (Wt+1 ) in the first order condition:
" ∗ #−1/ ( " . /
−1/ 1−γ #)1/(1−γ)−1 1/ −γ −1/
λt Ct = δ Ut+1 Et U (Wt+1 ) Et U (Wt+1 ) λt+1 Ct+1 Rc,t+1 .
∗
Using the expression for Ut+1 , this last equation can be written after some algebra as,

1 = Et (Mt+1 Rc,t+1 ) . (8.9)

Here, Mt+1 is the stochastic discount factor, or intertemporal marginal rate of substitution,
which is given by:
−1/
λt+1 U (Wt+1 )1/ −γ Ct+1
Mt+1 = δ " ∗ #1/ −γ −1/
.
λt U Ct
t+1
We guess and verify the policy function for consumption and the form of the utility
function. As in Weil (1989) and Epstein and Zin (1991), we guess that:

U (Wt ) = at Wt ,
Ct = bt Wt .

Replacing these guesses in equation (8.8) and simplifying yields:

1−1/ −1/
at = λt b t . (8.10)

Using the same guesses in the Hamilton-Jacobi-Bellman equation (8.7) and simplifying, we
obtain:
2 0" 1 31/(1−1/ )
& '1−γ !#1/(1−γ) 1−1/
6 1−1/ Wt+1 7
at = 4λt bt + δ @ Et at+1 A 5 .
Wt

43
Finally, using the budget constraint to replace Wt+1 /Wt we obtain:
* .( " /1−1/ +1/(1−1/ )
1−1/ 1−γ #)1/(1−γ)
at = λ t b t + δ Et (at+1 (1 − bt ) Rc,t+1 ) . (8.11)

Equations (8.10) and (8.11) give a solution to at and bt .

Combining equations (8.10) and (8.11) we obtain:
.( " #)1/(1−γ) /1−1/
−1/
λt b t (1 − bt ) = δ Et (at+1 (1 − bt ) Rc,t+1 )1−γ ,

which we can replace in the expression for the stochastic discount factor together with (8.10)
to obtain:
& '(1−γ)/(1−1/ ) & '−((1/ −γ)/ )/(1−1/ ) & '−1/
λt+1 bt+1 Ct+1
Mt+1 = δ (1 − bt ) (Rc,t+1 )1/ −γ
.
λt bt Ct

Now note that θ = (1 − γ) / (1 − 1/ ), and that

Ct+1
Ct bt+1 Rc,t+1 (Wt − Ct ) / (bt Wt ) bt+1 (1 − bt )
= = ,
Rc,t+1 Rc,t+1 bt
to finally get,
& 'θ & '− θ
λt+1 Ct+1
Mt+1 = δ (Rc,t+1 )θ−1 .
λt Ct
Taking logarithms of both sides and equating the consumption of the representative agent
to aggregate consumption yields equation (3.9).
The nominal SDF is
$ Pt
Mt+1 = Mt+1 ,
Pt+1
and that the log-nominal SDF is m$t+1 = mt+1 − ιt+1 , where ιt+1 is the inflation rate.
The rest of the equilibrium derivation involves solving for the real return on the consump-
$
tion claim, rc,t+1 , the real return on the dividend claim, rd,t+1 , and rf,t+1 and rf,t+1 , the short
term real and nominal interest rates, respectively. Until this point, we did not need to spec-
ify the process for the time-preference, consumption growth, and dividend growth shocks,
or inflation. We solve the rest of the model assuming the most general processes used in
the paper, the inflation process and consumption growth of Section 6 and the processes in
equations (3.22) through (3.24) of Subsection 3.4. Recovering the equilibrium values for the
benchmark model is done by setting π cλ = π dλ = σ η = 0, ν = σ w = αc = αd = 0, and
π cχ = µι = σ ι = ρχ = ν χ = σ χ = σ χe = 0.

44
 To price a real claim on aggregate consumption, we must solve the pricing condition:

Et [exp (mt+1 + rc,t+1 )] = 1.

We guess that the logarithm of the price consumption ratio, zct ≡ log (Pc,t /Ct ), is

zct = Ac0 + Ac1 xt + Ac2 η t+1 + Ac3 σ 2t + Ac4 ∆ct + Ac5 σ 2χt ,

and approximate
rc,t+1 = κc0 + κc1 zct+1 − zct + ∆ct+1 . (8.12)

Replacing the approximation for rc,t+1 on the pricing condition gives

Et [exp (θ log (δ) + θ log (λt+1 /λt ) + (1 − γ) ∆ct+1 + θκc0 + θκc1 zct+1 − θzct )] = 1.

Computing this expectation requires some algebra and yields the equation

0 = θ log (δ) + (1 − γ + θκc1 Ac4 ) µc − (1 − γ + θκc1 Ac4 ) αc σ 2 + θκc0 + θκc1 Ac0 − θAc0 + θκc1 Ac5 (1 − vχ )
+ ((1 − γ + θκc1 Ac4 ) αc + θκc1 Ac3 ) (1 − ν) σ 2 + ((1 − γ + θκc1 Ac4 ) π cλ + θκc1 Ac1 σ λ )2 /2
+ (θκc1 Ac2 )2 /2 + ((1 − γ + θκc1 Ac4 ) αc + θκc1 Ac3 )2 σ 2w /2 + (θκc1 Ac5 σ χe )2 /2
+θxt − θAc1 xt + θκc1 Ac1 ρxt + θσ η η t+1 − θAc2 η t+1 − θAc4 ∆ct + (1 − γ + θκc1 Ac4 ) ρc ∆ct
−θAc3 σ 2t + ((1 − γ + θκc1 Ac4 ) αc + θκc1 Ac3 ) vσ 2t + (1 − γ + θκc1 Ac4 )2 σ 2t /2
−θAc5 σ 2χt + (1 − γ + θκc1 Ac4 )2 π 2cχ σ 2χt /2 + θκc1 Ac5 vχ σ 2χt .

In equilibrium, this equation must hold in all possible states resulting in the restrictions:
1
Ac1 = ,
1 − κc1 ρ
Ac2 = σ η ,
1−γ
αc v + 1−κc1 ρc
/2 1 − 1/
Ac3 = ,
1 − κc1 v 1 − κ1 ρc
(1 − 1/ ) ρc
Ac4 = ,
1 − κc1 ρc
(1 − γ) (1 − 1/ ) π 2cχ
Ac5 = .
2 (1 − κc1 vχ ) (1 − κc1 ρc )2

45
The sign of Ac5 is the same as that of θ and is independent of the sign of π cχ . Finally,

log (δ) + (1 − 1/ + κc1 Ac4 ) µc − (1 − 1/ + κc1 Ac4 ) αc σ 2 + κc0 + ((1 − 1/ + κc1 Ac4 ) αc + κc1 Ac3
Ac0 =
1 − κc1
θ ((1 − 1/ + κc1 Ac4 ) π cλ + κc1 Ac1 σ λ ) /2 + θ (κc1 Ac2 )2 /2 + θ ((1 − 1/ + κc1 Ac4 ) αc + κc1 Ac3 )2
2
+
1 − κc1
2
κc1 Ac5 (1 − vχ ) σ 2χ + θ (κc1 Ac5 σ χe ) /2
+
1 − κc1
To solve for the one period real risk-free rate that pays one unit of consumption next
period we again use the real SDF. In logarithms, the Euler equation is:

rf,t+1 = − log (Et (exp (mt+1 )))

& & & '''
θ
= − log Et exp θ log (δ) + θ log (λt+1 /λt ) − ∆ct+1 + (θ − 1) rc,t+1 .

Using equation (8.12), we get:

rf,t+1 = − log (Et (exp (θ log (δ) + θ log (λt+1 /λt ) − γ∆ct+1 + (θ − 1) (κc0 + κc1 zct+1 − zct )))) .

Substituting in the consumption process and the solution for the price-consumption ratio,
and after much algebra, we obtain,

rtf = −θ log (δ) + γ

e µc − γ eπ cλ )2 /2
eαc σ 2 − ((θ − 1) κc1 Ac2 )2 /2 − ((θ − 1) κc1 Ac1 σ λ − γ
eαc ) (1 − ν) σ 2
− (θ − 1) κc0 − (θ − 1) (κc1 − 1) Ac0 − ((θ − 1) κc1 Ac3 − γ
− ((θ − 1) κc1 Ac5 σ χe )2 /2 − (θ − 1) κc1 Ac5 (1 − vχ ) σ 2χ
eαc )2 σ 2w /2 − log (λt+1 /λt ) + (1/ ) ρc ∆ct
− ((θ − 1) κc1 Ac3 − γ
eαc ) vσ 2t + (θ − 1) Ac3 σ 2t − γ
− ((θ − 1) κc1 Ac3 − γ γ π cχ )2 σ 2χt /2 − (κc1 vχ − 1) (θ − 1) Ac5 σ 2χt ,
e2 σ 2t /2 − (e

where
e = γ − (θ − 1) κc1 Ac4 .
γ

Setting the relevant parameters to zero as detailed above, we get the benchmark-model
value of the risk-free rate (3.18).
Next, we price a claim that pays a real dividend. Again, we assume the price-dividend
ratio is given by

zdt = Ad0 + Ad1 xt + Ad2 η t+1 + Ad3 σ 2t + Ad4 ∆ct + Ad5 ∆dt + Ad6 σ 2χt ,

46
and approximate the log-linearized return to the claim to the dividend:

rd,t+1 = κd0 + κd1 zdt+1 − zdt + ∆dt+1 . (8.13)

The pricing condition is

Et [exp (mt+1 + rd,t+1 )] = 1.

Using the expressions for mt+1 , rc,t+1 and rd,t+1 ,

*
θ
1 = Et exp θ log (δ) + θ log (λt+1 /λt ) − ∆ct+1 + (θ − 1) (κc0 + κc1 zct+1 − zct + ∆ct+1 )

+κd0 + κd1 zdt+1 − zdt + ∆dt+1 ] .

Replacing the consumption and dividend growth processes and of the price-consumption and
price-dividend ratios, and repeating similar algebra as above, we obtain:
1
Ad1 = ,
1 − κd1 ρ
Ad2 = σ η ,
. /
(1 + κd1 Ad5 ) αd v − γ 1−γ 1/ −γ
bαc v − αc v + 1−κc1 ρ /2 1−κ c1 ρ
+ (1 + κd1 Ad5 )2 σ 2d /2 + ((1 + κd1 Ad5 ) π dc − γ
b )2 /
c c
Ad3 =
1 − κd1 v
− (1/ ) ρc
Ad4 = ,
1 − κd1 ρc
ρd
Ad5 = ,
1 − κd1 ρd
bπ cχ )2 /2 + (θ − 1) (κc1 vχ − 1) Ac5
((1 + κd1 Ad5 ) π dχ − γ
Ad6 = ,
1 − κd1 vχ
and

bαc ) σ 2 + (θ − 1) κc0 + (θ − 1) (κc1

bµc + (1 + κd1 Ad5 ) µd − ((1 + κd1 Ad5 ) αd − γ
Ad0 (1 − κd1 ) = θ log (δ) − γ
bαc + (θ − 1) κc1 Ac3 + κd1 Ad3 ) (1 − v) σ 2 + ((θ − 1) κc1 Ac2 + κd1 Ad
+ ((1 + κd1 Ad5 ) αd − γ
bπ cλ + ((θ − 1) κc1 Ac1 + κd1 Ad1 ) σ λ )2 /2
+ ((1 + κd1 Ad5 ) π dλ − γ
bαc + (θ − 1) κc1 Ac3 + κd1 Ad3 )2 σ 2w /2
+ ((1 + κd1 Ad5 ) αd − γ
+ ((θ − 1) κc1 Ac5 + κd1 Ad6 ) (1 − vχ ) σ 2χ + ((θ − 1) κc1 Ac5 + κd1 Ad6 )2 σ 2χe /2,

b = (γ − (θ − 1) κc1 Ac4 − κd1 Ad4 ) .

γ

47
 To complete the derivation of the equilibrium we must solve for the steady state values
of zc and zd :
µc
zc = Ac0 + Ac3 σ 2 + Ac4 + Ac5 σ 2χ ,
1 − ρc
and
µc
zd = Ad0 + Ad3 σ 2 + Ad4+ Ad6 σ 2χ .
1 − ρc
Having solved for these constants, we can compute the expected return on the dividend
claim, Et (rd,t+1 ). Setting the relevant parameters to zero, we obtain the benchmark-model
value of Et (rd,t+1 ) given in equation (3.17). We now derive the expression for the conditional
real risk premium in the benchmark model:

Et (rd,t+1 ) − rf,t+1 = Et (κd0 + κd1 zdt+1 − zdt + ∆dt+1 ) − rf,t+1 .

Replacing the values of zdt and ∆dt+1 and computing expectations,

Et (rd,t+1 ) − rf,t+1 = κd0 + (κd1 − 1) Ad0 − xt + µ − rf,t+1 .

Replacing the values of Ad0 and rf,t+1 and simplifying, we obtain expression (3.19).
Finally, we price a one period nominal risk free bond:

$
rf,t+1 = − log (Et (exp (mt+1 − ιt+1 ))) .

It can be shown that the di§erence between real and nominal interest rates is:

$
rf,t+1 − rf,t+1 = −σ 2ι /2 + µι + χt .

8.3. Appendix C

In this appendix we solve for the prices of real and nominal zero-coupon bonds of di§erent
(n)
maturities. Let Pt be the time-t price of a bond that pays one unit of consumption at t+n,
(1)
with n ≥ 1. The Euler equation for the one-period risk-free bond price Pt = 1/Rf,t+1 is
(1)
Pt = Et (Mt+1 ) .

The price for a risk-free bond maturing in n > 1 periods can be written recursively as:
. /
(n) (n−1)
Pt = Et Mt+1 Pt+1 .

In Appendix B we derive the value of the risk-free rate:

h i
(1)
rf,t+1 = − ln Pt .

48
It is useful to write the risk-free rate as:
& '
λt+1
rf,t+1 = − log + (1/ ) ρc ∆ct − p1 − q1 σ 2t − ζ 1 σ 2χt ,
λt
where

p1 = θ log (δ) − γ 1 µc + γ 1 αc σ 2 + (θ − 1) κc0 + (θ − 1) (κc1 − 1) Ac0

+ ((θ − 1) κc1 Ac3 − γ 1 αc ) (1 − ν) σ 2
+ ((θ − 1) κc1 Ac2 )2 /2 + ((θ − 1) κc1 Ac1 σ λ − γ 1 π cλ )2 /2
+ ((θ − 1) κc1 Ac3 − γ 1 αc )2 σ 2w /2
+ ((θ − 1) κc1 Ac5 )2 σ 2χe /2 + (θ − 1) κc1 Ac5 (1 − vχ ) σ 2χ ,

q1 = γ 21 /2 + ((θ − 1) κc1 Ac3 − γ 1 αc ) v − (θ − 1) Ac3 ,

ζ 1 = γ 21 π 2cχ /2 + (κc1 vχ − 1) (θ − 1) Ac5 ,

and
e = γ − (θ − 1) κc1 Ac4 .
γ1 = γ
h i
(n) (n) (1)
Let pt = ln Pt . Therefore, pt = −rf,t+1 . The price of a risk-free bond that pays
one unit of consumption in n periods:
h . /i
(n) (n−1)
pt = ln Et exp mt+1 + pt+1 .

After much algebra,

(n) " #
pt = pn + 1 + ρ + ... + ρn−1 xt + σ η η t+1 + qn σ 2t − (γ n ρc + (θ − 1) Ac4 ) ∆ct + ζ n σ 2χt ,

with
qn = γ 2n /2 + ((θ − 1) κc1 Ac3 − γ n αc ) v − (θ − 1) Ac3 + vqn−1 ,

ζ n = (γ n π cχ )2 /2 + (κc1 vχ − 1) (θ − 1) Ac5 + vχ ζ n−1 ,

pn = θ log (δ) − γ n µc + γ n αc σ 2 + (θ − 1) κc0 + (θ − 1) (κc1 − 1) A0

+ ((θ − 1) κc1 Ac3 + qn−1 − γ n αc ) (1 − ν) σ 2
"" " ## #2
+ ((θ − 1) κc1 + 1)2 σ 2η /2 + (θ − 1) κc1 Ac1 + 1 + ρ + ... + ρn−2 σ λ − γ n π cλ /2
+ ((θ − 1) κc1 Ac3 + qn−1 − γ n αc )2 σ 2w /2
" # " #2
+ (θ − 1) κc1 Ac5 + ζ n−1 (1 − vχ ) σ 2χ + (θ − 1) κc1 Ac5 + ζ n−1 σ 2χe /2
+p(n−1) ,

49
and the sequence of γ n is given by
" # " #
γ n = γ 1 1 + ρc + ... + ρn−2
c + (1/ ) ρn−1
c + 1 + ρc + ... + ρ n−3
c (θ − 1) Ac4
...
γ 3 = γ 1 (1 + ρc ) + (1/ ) ρ2c + (θ − 1) Ac4
γ 2 = γ 1 + (1/ ) ρc
γ 1 = γ − (θ − 1) κ1 Ac4 .
(n) (n)
Finally, we define the yield on an n-period real zero-coupon bond as yt = − n1 pt .
(1)$
$
The price a one-period nominal bond is pt = −rf,t+1 , with
& '
λt+1
$
rf,t+1 = − log + (1/ ) ρc ∆ct − p1$ − q1 σ 2t − ζ $1 σ 2χt + χt ,
λt
with

p1$ = p1 − µι + σ 2ι /2,
ζ $1 = ζ 1 .

The price of a bond that pays one unit of numeraire in n periods:

. . //
(n)$ $ (n−1)$
pt = ln Et exp mt+1 + pt+1 .

This can be solved to obtain:

(n)$ " # " #
pt = pn$ + 1 + ρ + ... + ρn−1 xt +σ η η t+1 +qn σ 2t −(γ n ρc + (θ − 1) Ac4 ) ∆ct +ζ $n σ 2χt − 1 + ρχ + ... + ρn−1
χ χt

with γ n and qn as defined above and

" #2
ζ $n = 1 + ρχ + ... + ρn−2
X + γ n π cχ /2 + (κc1 vχ − 1) (θ − 1) Ac5 + vχ ζ n−1 ,

pn$ = θ log (δ) − µι − γ n µc + γ n αc σ 2 + (θ − 1) κc0 + (θ − 1) (κc1 − 1) A0

+ ((θ − 1) κc1 Ac3 + qn−1 − γ n αc ) (1 − ν) σ 2
"" " ## #2
+ ((θ − 1) κc1 + 1)2 σ 2η /2 + (θ − 1) κc1 Ac1 + 1 + ρ + ... + ρn−2 σ λ − γ n π cλ /2
+ ((θ − 1) κc1 Ac3 + qn−1 − γ n αc )2 σ 2w /2
. / . /2
$ $
+ (θ − 1) κc1 Ac5 + ζ n−1 (1 − vχ ) σ χ + (θ − 1) κc1 Ac5 + ζ n−1 σ 2χe /2 + σ 2ι /2
2

+p(n−1)$ .
(n)$ (n)$
We define the yield on an n-period nominal zero-coupon bond as yt = − n1 pt .

50
8.4. Appendix D

This Appendix contains a set of tables that provide versions of the results discussed in the
main text that use ex-post, instead of ex-ante versions of the real interest rate.

51
 Table 4 (appendix)

Parameter Estimates and Standard Errors

Models estimated with ex-post real bond yields

Parameter Benchmark,model Extended,model Parameter Benchmark,model Extended,model

γ 1.7005 2.2228 ρ 0.99143 0.99108

(0.017871) (0.22671) (0.00047649) (0.00059552)

ψ 1.4623 2.3876 σλ 0.00054228 0.00040323

,(0.043679) (0.62451) ,(2.0143e?05) (2.6299e?05)

δ 0.99784 0.9978 Implied value of θ ?2.2158 ?2.1041

,(0.0021963) ,(0.00013075) (0.099733) ,(0.50544)

σc 0.0068554 0.0046005 ν 0.00 0.99673

(0.00015674) ,(0.00046367) ,(0.00057417)

µ 0.0015234 0.0009117 σω 0.00 1.70E?06

,(5.2956e?05) (5.7211e?05) ,(2.6272e?07)

πcλ 0.00 ?0.0018895 αc 0.00 ?119.3912

(0.00029618) (28.7094)

σd 0.017153 0.025564 αd 0.00 ?145.5293

(0.00041145) (0.52335) (68.6766)

πdc 0.0015658 ?1.3161 ρc 0.00 0.090348

(0.00041998) (0.80412) (0.045439)

πdλ 0.00 ?0.010369 ρd 0.00 0.33438

(0.0054912) (0.39606)

ση 0.00 0.0082606
(0.00043981)
 Table(5((appendix)

Models estimated with ex-post real bond yields

Data((constrained,( Data(
Benchmark(
Moments(Matched(in(Estimation ex4post(bond( (unconstrained,(ex4 Extended(model
model
yields) post(bond(yields)

Average(growth(rate(of( 1.4941 2.2197 1.84 1.20

of(consumption (0.30395) (0.22413)

Average(growth(rate(of(dividends 1.4941 0.12538 1.84 1.20

(0.30395) (0.65732)

Standard(deviation(of(the(growth 2.1865 2.1624 2.36 2.72

rate(of(consumption (0.34494) ((0.31345)

Standard(deviation(of(the(growth 6.914 6.5181 5.89 6.95

rate(of(dividends ((1.0409) ((1.1219)

Contemporaneous(correlation 0.2215 0.23722 0.09 0.21

between(consumption(growth(and ((0.1006) (0.10649)
dividend(growth

First4order(serial(correlation 0.2001 0.1584 40.01 0.43

of(consumption (0.1475) (0.1240)

Correlation(between(stock(returns 40.1555 40.2607 0.00 40.04

(0.1070)
(and(one4period(lagged(consumption(growth (0.0999)

Correlation(between(stock(returns 0.4950 0.5093 0.01 0.22

(and(one4period(lead(consumption(growth(0.0670) (0.0490)

Average(return(to(equities 7.8252 6.3948 6.25 3.87

(1.74) (1.8804)

Standard(deviation(of(return(to 17.2537 17.6831 15.93 16.67

equities (1.3783) (1.461)

Average(risk4free(rate 0.19347 0.10476 0.19 0.19

(0.84341) (0.84457)

Standard(deviation(of(the(risk4free 3.4656 3.5054 3.99 4.23

rate (0.82769) (0.83489)

First4order(serial(correlation(of(the 0.59242 0.59736 0.90 0.49

risk4free(rate (0.082348) (0.078284)

Equity(premium 7.6318 6.29 6.06 3.68

(1.758) (1.8812)

Average(price4dividend(ratio 3.3761 3.3775 3.12 3.30

(0.14887) (0.14887)

Standard(deviaton(of(price4 0.47324 0.44502 0.28 0.40

dividend(ratio (0.074827) (0.081116)

First4order(serial(correlation(of 0.95041 0.93257 0.84 0.89

price(dividend(ratio (0.034499) (0.042679)

Contemporaneous(correlation 0.20139 0.21274 40.95 40.24

between(risk(free(rate(and (0.12648)( (0.12982)(
price(dividend(ratio
 Table*6*(appendix)

Correlation*Between*Stock*Returns*and*Per*Capita
Growth*Rates*of*Consumption*and*Dividends

Model*estimated*with*ex9post*bond*returns

Consumption'growth

Data* Data* Benchmark*

Extended*model
(constrained) (unconstrained) model
Moments

19year*correlation*between 90.07 90.05 0.04 90.14

equity*returns*and (0.11) *(0.12)
consumption*growth

59year*correlation*between 90.01 90.05 0.05 0.01

equity*returns*and (0.16) (0.13)
consumption*growth

109year*correlation*between 90.08 90.21 0.05 0.18

equity*returns*and *(0.30) (0.21)
consumption*growth

Dividend'growth

19year*correlation*between 0.08 0.05 0.37 90.11

equity*returns*and *(0.11) *(0.11)
dividend*growth

59year*correlation*between 0.22 0.29 0.35 0.07

equity*returns*and *(0.14) *(0.14)
dividend*growth

109year*correlation*between 0.51 0.59 0.41 0.26

equity*returns*and *(0.22) (0.15)
dividend*growth
 Table)7)(apendix)

Predictability)of)Excess)Returns)by)Price1dividend)Ratio)at)Various)Horizons

Model)estimated)with)ex1post)real)bond)yields

Benchmark*model

Model) Model) Model) Model)

Data (median) (plim) Data (median) (plim)
Slope)coefficient R1square (%)of)values)larger)than
R1square)in)data)
1)year 10.10 10.05 0.008 0.08 0.01 0.0003
)(0.03) (0.02)

3)years 10.27 10.14 0.027 0.14 0.03 0.001

(0.07) (0.08)

5)years 10.41 10.21 0.034 0.26 0.04 0.001

)(0.12) (0.04)

Extended*model

1)year 10.10 10.06 10.03 0.08 0.02 0.01

)(0.03) (0.05)

3)years 10.27 10.18 10.08 0.14 0.06 0.02

(0.07) (0.19)

5)years 10.41 10.26 10.13 0.26 0.09 0.04

)(0.12) (0.12)
 Table-8-(appendix)

Correlation-of-price)dividend-ratio-and-ex)post-risk)free-rate

Sample Full)sample Pre-1983 Post-1983

Canada 1952)2014 0.02 0.55 )0.31

(0.14) (0.06) (0.25)

France 1932)2014 )0.56 )0.67 0.07

(0.17) (0.13) (0.17)

Germany 1968)2014 )0.11 )0.10 )0.20

(0.08) (0.13) (0.20)

Italy 1940)2014 )0.54 )0.63 0.35

(0.14) (0.08) (0.21)

Japan 1960)2014 0.39 )0.01 0.66

(0.16) (0.20) (0.08)

U.K. 1934)2014 0.32 0.27 )0.25

(0.11) (0.06) (0.22)

US 1939)2011 0.21274 0.172 )0.44

(0.12982)- (0.148) -(0.21)
 Table*9*(appendix)

Term*Structure*of*Bond*Yields

Model*estimated*with*ex-post*yields

Data*(ex-post* Data*(ex-ante* Benchmark* Extended*

Moments real*yields) real*yields) model model

Average'yield

Long-term*bonds 1.32 2.90 5.30 2.53

(sample*1929-1992) (1.10) (0.84)

Intermediate-term*bond 1.39 1.93 2.25 1.53

(sample*1929-2007) (0.91) (0.99)

Short-term*bond 0.42 0.46 0.19 0.19

(sample*1929-2011) (0.80) (0.78)

Return*to*equity*minus*long-term*bond*yield 4.16 2.54 0.95 1.34

(sample*1929-1992) (2.39) (2.09)

Standard'deviation'of'yield

Long-term*bonds 3.02 2.59 1.73 1.39

(sample*1929-1992) (0.65) (0.52)

Intermediate-term*bond 3.29 3.14 3.19 2.49

(sample*1929-2007) (0.53) (0.57)

Short-term*bond 3.88 3.85 3.99 4.23

(sample*1929-2011) (0.77) (0.77)

Return*to*equity*minus*long-term*bond*yield 20.2 20.09 15.59 16.53

(sample*1929-1992) (2.47) (1.96)
 Table$11$(appendix)

Regressions$of$Excess$Stock$Returns$on$Long(term$Bond$Yields$in$Excess$of$Short$Rate

Model$estimated$with$ex(post$real$yields

Data$1929(2011 Benchmark$model Extended$model

Long%term*government*bonds*(20*years)

R(square 0.02 0.12 0.04

Slope 0.60 3.40 0.97

(0.96)

Intermediate*term*government*bonds*(5*years)

R(square 0.002 0.07 0.03

Slope (0.30 3.30 0.97

(1.10)