1 s2.0 S109420251400060X Main PDF

JID:YREDY AID:702 /FLA
[m3G; v1.142; Prn:27/10/2014; 9:35] P.1 (1-21)
Review of Economic Dynamics ()
Contents lists available at ScienceDirect
Review of Economic Dynamics

www.elsevier.com/locate/red
Innovation, product cycle, and asset prices

Ryo Jinnai 1
Texas A&M University, 4228 TAMU, College Station, TX 77843, USA
a r t i c l e
i n f o
Article history:
Received 25 June 2013
Received in revised form 25 September
2014
Available online xxxx
JEL classication:
E30
G12
O30
O41
a b s t r a c t
This paper constructs a simple endogenous growth model featuring the product cycle,
i.e., the transition from monopoly to perfect competition, and studies its implications for
both asset market and business cycle statistics. I nd that the product cycle is a powerful
amplication mechanism; the model incorporating the product cycle is able to generate
nearly twice as large an equity premium as the model without the product cycle and,
as a result, matches the equity premium data. The current paper thereby contributes to
advancing a promising theory on the economic sources of long-run risks, postulating that
innovation and R&D cause long-run uncertainties in economic growth.
2014 Elsevier Inc. All rights reserved.
Keywords:
Innovation
Research and development
Product cycle
Imitation
Recursive preference
Long-run risk
1. Introduction
The literature studying the link between asset market statistics and business cycle variables in versions of the canonical
real business cycle models, exemplied by Boldrin et al. (2001), Jermann (1998), and Tallarini (2000), has experienced
signicant development in recent years. A line of research that has shown great promise incorporates disaster risk (Barro,
2006; Gabaix, 2012; Rietz, 1998) into the production economy (Gourio, 2012). Another research front of interest incorporates
long-run risk (Bansal and Yaron, 2004) into the production economy. Prominent examples along this line of research include
Kaltenbrunner and Lochstoer (2010), who study the long-run consumption risk arising from the households intertemporal
resource allocation, and Ai et al. (2013), Croce (2014), Favilukis and Lin (unpublished results), and Rudebusch and Swanson
(2012), who study the implications of long-run productivity risk. An even more ambitious study is Kung and Schmid (in
press), who link the long-run productivity risk to innovations and thereby identify (instead of assuming) economic sources
of long-run risks in the data. They make a signicant contribution toward solving the equity premium puzzle (Mehra and
Prescott, 1985) by showing that approximately half of the actual size of the equity premium is justied by a highly stylized
endogenous growth model augmented with a recursive utility (Epstein and Zin, 1989).2 This result is remarkable, but the
E-mail address: rjinnai@econmail.tamu.edu.

Fax: +1 979 847 8757.
2
Aghion and Howitt (1997), Grossman and Helpman (1991), and Romer (1990) are prominent contributions to the founding of endogenous growth
theory. See Acemoglu (2008) for a review of the literature.
1
http://dx.doi.org/10.1016/j.red.2014.10.002
1094-2025/ 2014 Elsevier Inc. All rights reserved.
[m3G; v1.142; Prn:27/10/2014; 9:35] P.2 (1-21)
R. Jinnai / Review of Economic Dynamics ()
remaining unexplained half of the equity premium suggests that further elaboration on the basic endogenous growth model
might provide a better understanding of the interesting links between asset market statistics and innovative activities.
This paper proposes such an elaboration, motivated by a simple observation: patents offer only partial protection to
inventors and discoverers. Anecdotal evidence supporting this view is abundant (see Boldrin and Levine, 2008, 2013; Moser,
2013), but a story documented in the graduate-level industrial organization textbook written by Scherer and Ross (1990) is
highly compelling: DuPont, for example, took out hundreds of patents on variants of its nylon synthetic ber technology.
However, even in the directly applicable polyamide molecule family, it left a gap into which Germanys IG Farben moved
with Perlon L, and other companies invented competitive bers using polyester and polyolen molecules. This historical
account is powerful testimony to the fact that the protection provided by the patent system is often weak because there
can be many viable solutions to a technical problem. A survey encompassing 650 executives (Levin et al., 1987) conrms
this view; among eight reasons why the effectiveness of patents was limited, they reported that the ability of competitors
to legally invent around patented inventions was by far the greatest factor. Informal protections that do not rely on patents
are also imperfect, and, in fact, the aforementioned survey suggests that imitating an unpatented product is less costly than
imitating a patented one.3 In sum, imitative research is pervasive. The logical consequence is heterogeneity in the degree of
competition among local markets, which is the feature the current paper adds to the canonical endogenous growth model
to investigate its implications.
I formalize this idea in a highly stylized way so that the deviation from Kung and Schmid (in press) is minimal. Specifically, I adopt the simple product cycle of Futagami and Iwaisako (2007). These authors assume that a product is initially
monopolistically produced and later competitively produced. I assume the same market structure in the current manuscript.
Futagami and Iwaisako also assume that the transition from monopoly to perfect competition is time-dependent, with the
presumption that the duration of monopoly is the same as the duration of a patent.4 I do not follow them in this regard
because I believe that protection from imitation is far from perfect regardless of the method that inventors and discoverers
rely on. Instead, I assume that the transition is stochastic. The stochastic transition is both arguably realistic and, with the
additional assumption that it is i.i.d. among products and across time, convenient in terms of the analysis in that it reduces
the number of state variables.
I show that the product cycle is a powerful amplication mechanism having great potential for improving the models
ability to match asset market statistics. Specically, I show that with this mechanism, an otherwise standard endogenous
growth model generates nearly twice as large an equity premium as the canonical endogenous growth model without the
product cycle. In other words, given that Kung and Schmid (in press) demonstrate that the canonical model is able to justify
approximately half of the actual size of the equity premium, augmenting the product cycle allows the model to nearly
perfectly match the equity premium.
Two factors are important for this result. The rst is the effect of the product cycle on economic uctuations. Specically,
because product entries are not constant across time and imitation is a slow process, the ratio of monopolistic products
to competitive products is endogenous. In particular, it is procyclical at high frequencies because innovative activities are
procyclical and new products are more likely to be monopolistically produced. In contrast, the ratio of monopolistic products
to competitive products is counter-cyclical at medium frequencies because productivity gains from competition become a
more prominent mechanism for a longer time span. Both of these effects raise the equity premium, as the former makes
the prot share procyclical and the equity more risky and the latter increases the persistence of economic uctuations.
The second factor that contributes to a high equity premium in the model incorporating the product cycle is its effect
on the steady-state asset sizes. Specically, the size of intangible capital is smaller in the model with the product cycle than
in the model without the product cycle due to the shorter duration of the monopoly. Because the aggregate stock market
return is a value-weighted return of tangible and intangible capital, this change in weight makes the tangible capital return
the more important determinant of equity premium, and because tangible capital return is higher than intangible capital
return in the simulation, it contributes to a high equity premium.
A number of recent papers examine the link between technological progress and asset prices. Prominent examples are
Garleanu et al. (unpublished results) and Garleanu et al. (2012), and Pastor and Veronesi (2009). These papers introduce
differences in the relative importance of intermediate goods in nal goods production, among the productivities of investment options, and in productivities between new and old technologies respectively. The current paper, in contrast, does not
introduce any built-in cross-sectional heterogeneity in the physical attributes of products, instead introducing heterogeneity
in the modes of competition in the local markets. Iraola and Santos (unpublished results) and Comin et al. (unpublished results) study the effects of technical advancements on asset markets. They introduce a slow adoption process after invention,
which strongly amplies the stock market volatility, but do not consider the imitation of adopted products. They focus on
medium-term business cycles, the measurement of aggregate productivity, and the timings of responses of macroeconomic
variables but put less emphasis on long-run risks.
The remainder of this paper is organized as follows. Section 2 presents the model economy. Except for the possibility
that monopolists may lose market exclusivities, it is identical to the Kung and Schmid model. Section 3 inspects the product
cycle in a controlled environment. Specically, I compare the two models, with and without the product cycle, keeping the
3
4
A consistent result is found in Manseld et al. (1981).

With this assumption, the authors analyze the welfare-maximizing patent policy in an endogenous growth model.
[m3G; v1.142; Prn:27/10/2014; 9:35] P.3 (1-21)
parameters set at the same values across models as much as possible. I show that the model including the product cycle
generates nearly twice as large an equity premium as the model without the product cycle. Section 4 presents the main
results. After calibrating the parameters in the model with the product cycle following Kung and Schmid (in press), I show
that the calibrated model is able to match the equity premium in the data. I also show that this improvement is not at the
expense of business cycle moments. Specically, I show that the model is consistent with the data in the following dimensions: long-horizon autocorrelations of consumption growth, standard deviations of key macroeconomic variables detrended
with a bandpass lter at both high and medium frequencies, autocorrelations of the same set of variables, and contemporaneous correlations with the output of the same set of variables. Section 5 discusses the models other implications, and
Section 6 concludes.
2. Model
2.1. Household
The representative household has EpsteinZin preferences dened over consumption:
U t = (1 )C t
1 1
+ E t U t +1 1
1
where is the coecient of relative risk aversion, is the elasticity of intertemporal substitution, and 11/ . The
household supplies labor inelastically and maximizes utility by participating in nancial markets. Specically, the household
can take position Zt in the stock market, which pays an aggregate dividend Dt , and B t in the bond market. Accordingly, the
budget constraint of the household becomes
C t + Qt Zt +1 + B t +1 = W t L + (Qt +Dt )Zt + R t B t

where Qt is the stock price, R t is the gross risk-free rate, W t is the wage, and L denotes hours worked. The stochastic
discount factor is given by

M t +1 =
C t +1
1
E t [U t +1 ] 1
Ct
U t +1
where the second term captures preferences concerning long-run growth prospects. The agent dislikes shocks to long-run
expected growth rates if > 1 .
2.2. Final-goods rm
A representative rm uses capital K t , labor L t , and a composite of intermediate goods G t to produce the nal good
according to the production technology
Y t = K t (t L t )1
1
Gt
(1)
where the composite G t is dened as

Gt =
X i,t di
i It
X i ,t is an intermediate good i It , where It is the set of intermediate goods in use at time t; is the capital share; is
the intermediate goods share; and is the parameter affecting the elasticity of substitution between intermediate goods,
which is given by /( 1). t is the productivity shock following a stationary Markov process:
t = eat
and
at = at 1 + t ,
t i .i .d. N (0, 1)
The rms objective is to maximize the shareholders value. Taking the stochastic discount factor M t as given, this problem can be formally stated as
max
{ I t , L t , K t +1 , X i ,t }t 0, i It
E0

t =0

Mt D t
[m3G; v1.142; Prn:27/10/2014; 9:35] P.4 (1-21)
where D t is the rms dividends
D t = Y t I t W t Lt
P i ,t X i ,t di
i It
I t is capital investment, and P i ,t is the price per unit of intermediate good i, which the nal-good rm takes as given.
Investment is subject to capital adjustment costs, so the capital stock evolves as

K t +1 = (1 ) K t +
It
Kt
Kt
(2)
where is the depreciation rate of capital and () is the capital adjustment cost function. () is specied as in Jermann
(1998):
It
Kt
1 1
1 1
It
Kt
+ 2
The parameter represents the elasticity of the investment rate. The parameters 1 and 2 are set such that there are no
adjustment costs in the deterministic steady state.
Solving the cost minimization problem of purchasing intermediate goods leads to a downward-sloping demand function:

X i ,t =
P i ,t
1
Gt
P G ,t
(3)
where P G ,t is the price index dened as

P G ,t =
1
1
P i ,t di
(4)
i It
and the total expenditure on intermediate goods is given by

P i ,t X i ,t di = P G ,t G t
i It
Let qt denote the shadow value of capital. The rms rst-order optimality conditions are given by
qt =
(5)
t
W t = (1 )(1 )
q t = E t M t +1
P G ,t =
Yt
(6)
Lt
(1 )
Y t +1
K t +1
I t +1
K t +1
Yt

+ qt +1 t +1 + qt +1 (1 )
(7)
(8)
Gt
where t ( KI t ) and t ( KI t ). (5) equates the shadow value of capital with the cost of installing capital. (6) equates
t
t
wage rates with the marginal product of labor. (7) states that the shadow price of capital is the present discounted value of
the marginal product of capital, the marginal reduction of adjustment costs, and the resale value. (8) equates the price of
the intermediate-goods composite with the marginal product of the intermediate-goods composite.
2.3. Intermediate-goods rms and product cycle
I assume that each intermediate good is produced by an atomistic producer or producers depending on the goods
location in the life cycle. Specically, following Futagami and Iwaisako (2007), I assume that a product is initially monopolistically produced and later competitively produced. Production is round-about, implying that the marginal cost of
producing one intermediate good is unity. A monopolistic rm chooses price P tM to maximize the prots:

P tM 1
t max P tM 1
Gt
P tM
P G ,t
[m3G; v1.142; Prn:27/10/2014; 9:35] P.5 (1-21)
Solving this problem leads to the optimal markup pricing,
P tM =
(9)
The producer of a competitive product sets the price
P tC
at the marginal cost because competition drives prots to zero:
P tC = 1
(10)
Let N tM denote the mass of monopolistic products in period t and N tC denote the mass of competitive products in
period t. Their transition rules are given by
N tM+1 = N tE + N tM
N tC+1 = (1 ) N tM + N tC
(11)
where N tE denotes the mass of newly invented products in period t. The underlying assumptions are the following: (i) products are monopolistic when they are invented, (ii) a fraction (1 ) of randomly chosen monopolistic products evolve into
competitive products in every period, and (iii) as in Bilbiie et al. (2012), there is both a time-to-build lag and a death shock;
i.e., products invented in period t start producing in period t + 1, and a fraction (1 ) of both randomly chosen monopolistic and randomly chosen competitive products become permanently unavailable to the economy. Let N t denote the mass
of total products, i.e., the measure of It . Because N t = N tM + N tC is the accounting identity, its transition rule is given by
N t +1 = N tE + N t
Notice that the environment considered by Kung and Schmid (in press) is nested in the current framework as a special case
in which the parameter is set at unity and hence products are always monopolistic, i.e., N t = N tM .
The value of the exclusive production right of a monopolistic product is given by the present discounted value of the
current and future monopoly prots:
V t = t + E t [ M t +1 V t +1 ]
(12)
Competitive products, on the other hand, are valueless in the stock market because they do not generate any monopoly
prots. However, this does not mean that they are useless in the society. On the contrary, competitive products are more
valuable for consumers because their productions are more ecient.
2.4. R&D sector
Innovators develop new products for intermediate goods used in the production of nal output. They do so by conducting research and development (R&D) using the nal good as input. The exclusive production rights of newly developed
products can be sold to intermediate-goods producers. This market is competitive. Households can directly invest in R&D.
New products are created with the technology given by
N tE = t S t
where S t denotes R&D expenditures and t represents the productivity of the R&D sector that is taken as exogenous by the
R&D sector. Following Kung and Schmid (in press), I assume that this technology coecient involves a congestion externality
effect capturing decreasing returns to scale in the innovation sector
t =
Nt
1
S t ( N t )
where > 0 is a scale parameter and [0, 1] is the elasticity of new intermediate goods with respect to R&D.5 Because
there is free entry into the R&D sector, the following break-even condition must hold:
= E t [ M t +1 V t +1 ]
which equates the costs of inventing a product with the present discounted value of the revenues from selling it.
Kung and Schmid (in press) follow Comin and Gertler (2006) for the specication of the R&D productivity.
(13)
[m3G; v1.142; Prn:27/10/2014; 9:35] P.6 (1-21)
2.5. Equilibrium
The competitive equilibrium is dened in a standard way. This section discusses its properties. Using (4), (9), (10), and
the symmetry of products in each group, we nd that the price index of the intermediate-goods composite is given by
P G ,t = N t
N tM
Nt
N tC
Nt
( ) 1
(14)
Using (1), (8), and (14), we nd that the nal-goods production is given by
M
1
1
1
Nt
N tC
1
Yt =
K t (t L )
Nt
+
( )
Nt
Nt
Following Kung and Schmid (in press), I make the parameter restriction + 1 = 1. An advantage of this assumption is
that it leads to a production function that resembles the standard neo-classical function with labor-augmenting technology,
Y t = K t ( Z t L )1
where the equilibrium productivity measure is given by
t N t
Zt = A
N tM
Nt
N tC
Nt
( ) 1
(1 )(1 )
( )
and A
> 0 is a constant. The equilibrium productivity process thus contains a component driven by three
factors: (i) the exogenous forcing process, t ; (ii) an endogenous trend component reecting the accumulation of intermediate goods, N t ; and (iii) an endogenous cyclical component reecting variations in the product composition, N tM / N t and
N tC / N t . The second factor is a well-known variety effect: the expansion of product varieties allows more ecient use of
labor and capital in nal-goods production because it lowers the real price of the intermediate-goods composite. The third
factor captures productivity gains from competition: an increase in the share of competitive products leads to a productivity
gain because competitive products are produced more eciently. Although transitory, the endogenous variation in product
composition is a powerful amplication mechanism with important implications for business cycles and asset prices.
Now, I discuss the national income accounts. Because production is roundabout, I dene the aggregate net-value-added
output Yt as

Yt Y t N tM X tM + N tC X tC
where X tM denotes the production of a monopolistically produced intermediate good and X tC denotes the production of a
competitively produced intermediate good. The resource constraint of the economy is given by
Y t = C t + I t + N tM X tM + N tC X tC + S t
Rearranging this expression, we nd an identity
Yt = C t + I t + S t
(15)
This equation states that the aggregate net-value-added output is the sum of consumption, tangible investment, and intangible investment.
An alternative approach to the aggregate net-value-added output is from income. Using (6) and (8), we nd
Y t = W t L + (1 )Y t + P G ,t G t
Notice that the second term on the right-hand side of the equation is the compensation for capital service. Because the third
term in the right-hand side, the total revenues of intermediate-goods rms, is further decomposed into costs and prots,
we nd
Yt = W t L + (1 )Y t + N tM t
(16)
This equation states that the aggregate net-value-added output is the sum of compensation for labor service, compensation
for capital service, and monopoly prots.
Interestingly, income shares are not constant in the current economy. Specically, using (3), (8), (9), (10), and (14), we
nd that the aggregate net-value-added output and gross nal-good production are related as follows:
M
1

M
1
Nt
NC
Nt
NC
Yt = 1
+ t ( ) 1
+ t ( ) 1
Yt
Nt
Nt
Nt
Nt
[m3G; v1.142; Prn:27/10/2014; 9:35] P.7 (1-21)
Table 1
Model summary.
Y t = K t ( Z t L )1
Production function
Preference

1
NM
+ 1 Ntt ( ) 1
1

1
1
U t = (1 )C t + E t U t +1
Stochastic discount factor
M t +1 =
Investment optimality
q t
Capital price
qt = E t M t +1
Capital accumulation
K t +1 = (1 ) K t +
t N t
Zt = A
Equilibrium productivity
NtM
Nt
R&D optimality
It
Kt
C t +1 1 E t [U t1+1 ] 1 1
Ct
=1
U t +1
(1 ) YKtt++11 + qt +1 (1 )
It
Kt
Monopolistic products
N tM+1 = t S t + N tM
N tM
Nt
( ) 1

1
Net-value-added output
1
t = NS tt

N M
Yt = 1 Ntt + 1
Resource constraint
Yt = C t + I t + S t
Stock market value
Qt = Qtd + Qtic
Tangible capital value
Qtd = E t [ M t +1 ( (1 )Y t +1 I t +1 + Qtd+1 )]
Intangible capital value
Qtic = E t [ M t +1 (t +1 N tM+1 S t +1 + Qtic+1 )]
Productivity in R&D sector
I t +1
K t +1
Kt
Total products
Prots
+ qt +1
= E t [ M t +1 V t +1 ]
V t = t + E t [ M t +1 V t +1 ]
1 Y t NtM
t =
Nt Nt + 1
N t +1 = t S t + N t
Product value
I t +1
K t +1
N tM
Nt

N M
( ) 1 Ntt + 1
N tM
Nt
( ) 1
1
Yt
In addition, the aggregate dividends and the gross nal-good production are related as follows:
N tM t
1
N tM
Nt
N tM
Nt
N tC
Nt
) 1
Yt
Therefore, combining the two equations above, we nd
N tM t
Yt
1
(17)
1 + ( ) 1 (1 )( N tC / N tM )
The remaining income is divided between capital and labor in constant shares. (17) states that factor shares are affected by
product composition; specically, the prot share increases as the share of monopolistic products increases. If the product
entries are procyclical, this mechanism makes the prot share also procyclical. The total number of products, however, is
indifferent to factor shares given that the composition of products is constant.
Finally, I dene some nancial variables. Following Kung and Schmid (in press), I assume that the stock market value
includes all production sectors: the nal-goods sector, the intermediate-goods sector, and the research and development
sector. The aggregate dividend then becomes
Dt = (1 )Y t I t + N tM t S t
The stock market value is dened as the discounted sum of future aggregate dividends:
Qt E t

M t +1+i Dt +1+i
i =0
Tangible capital value is dened as the discounted sum of future aggregate dividends associated with physical capital:

Qtd
Et
M t +1 + i
(1 )Y t +1+i I t +1+i
i =0
Intangible capital value is dened as the discounted sum of future aggregate dividends associated with R&D:
Qtic
Et

i =0
M t +1+i N tM+1+i t +1+i
The returns are dened accordingly.
S t +1 + i
[m3G; v1.142; Prn:27/10/2014; 9:35] P.8 (1-21)
Table 2
Calibration I.
Parameter
Description
Subjective discount factor
Elasticity of intertemporal substitution
Risk aversion
Intermediate goods share
Gross markup
Capital share
Autocorrelation of
Survival rate of intermediate goods
Elasticity of new intermediate goods wrt R&D
Depreciation rate of capital stock
Volatility of productivity shock
Elasticity of capital investment share
Probability of maintaining monopoly
Scale parameter
t et
Trend growth rate specied as Z t = A
ENDO
EXO
HET
HOM
0.984
1.85
10
0.5
1.65
0.35
0.95
0.9625
0.83
0.02
1.75%
0.70
0.944
0.665
0.984
1.85
10
0.5
1.65
0.35
0.95
0.9625
0.83
0.02
1.75%
0.70
1
0.332
0.984
1.85
10
0.35
0.95
0.02
0.97%
0.70
1.90%
Note: This table reports the parameter values used in Section 3, which are taken from Kung and Schmid (in press) except for the probability of maintaining
the monopoly and the scale parameter in the endogenous growth model with the product cycle.
The model economy is summarized in Table 1. The eighteen equations in the table jointly determine the equilibrium
dynamics of eighteen endogenous variables, Y t , K t , Z t , Yt , C t , I t , S t , U t , M t , qt , V t , t , N t , N tM , t , Qt , Qtd , and Qtic . I study
the quantitative implications of this model using the perturbation method. Specically, I de-trend the variables that have a
trend component and then study the local dynamics of the de-trended system around its non-stochastic steady state, with
the system approximated by a second-order Taylor expansion of a logarithmic expression. The computations are performed
using Dynare.6
3. Preliminary results
I rst examine the role of the product cycle by comparing the quantitative implications of two otherwise identical endogenous growth models with and without the product cycle. They differ only in the probability of maintaining monopoly
and the scale parameter . The probability of maintaining monopoly is set to unity in one model and to 0.944 in the other.
The latter value was chosen based on the empirical evidence of Manseld et al. (1981), who reported that approximately 60%
of patented innovations are imitated within 4 years.7 The scale parameter was individually chosen in each model to match
the balanced growth evidence.8 The other parameters are set to the same values across both models, which are taken from
Kung and Schmid (in press) and summarized in Table 2. The headings HET and HOM are used to identify the two endogenous growth models because the products are heterogeneous in the modes of competition in the model with the product cycle and homogeneous in the model without the product cycle. Following Kung and Schmid (in press), I also consider a nested
standard real business cycle model with exogenous growth, whose parameter values are also taken from Kung and Schmid
(in press) and are reported in Table 2. This model is used as a benchmark to evaluate the endogenous growth models.
Table 3 reports the asset market statistics, in which the second column contains the data.9 As is well known, in the
real world the risk-free rate is low and smooth, and the equity premium is high and volatile. The fth column reports
the simulated moments from the standard real business cycle model with exogenous growth.10 Clearly, this model cannot
account for the actual asset market statistics. The mean risk-free rate observed in the model is approximately 2.6%, which
is six-fold larger than the actual mean risk-free rate; the mean equity premium observed in the model is nearly zero, while
the actual mean equity premium is 5%; and both the risk-free rate and equity premium are smoother in the model than in
the real world.
The fourth column shows that, consistent with Kung and Schmid (in press), replacing exogenous growth with endogenous
growth greatly improves the models ability to match asset market statistics, even if the product cycle is absent. Relative
to the standard real business cycle model with exogenous growth, the mean risk-free rate is nearly halved to 1.3%; the
mean equity premium is elevated from zero to 2.5%; and the volatilities of both the risk-free rate and the equity premium
http://www.cpremap.cnrs.fr/dynare.
These authors analyze data obtained from rms in the chemical, drug, electronics, and machinery industries concerning the cost and time of legally
imitating 48 product innovations.
8
An annual growth rate of 1.90% is targeted.
9
I construct the risk-free rate using the one-month T-Bill rate divided by a trailing 12-month average of ination, following Bansal and Yaron (2004).
Equity return is measured by the value-weighted NYSE return. The sample period is 1930:012008:12. The return data are taken from CRSP, and ination
is calculated using the CPI taken from BLS.
10
The model-implied moments are calculated and reported based on the simulation. I generate one thousand articial data sets of the sample size
corresponding to the actual data and report the median, 5th, and 95th percentiles of the moments across simulations. The same practice applies to the rest
of the paper.
7
[m3G; v1.142; Prn:27/10/2014; 9:35] P.9 (1-21)
Table 3
Asset pricing implications.
Data
First moments
E [r f ]
r ]
E [rm
f
0.42%
5.01%
E [rd r f ]
r ]
E [r ic
f
ENDO
EXO
HET
HOM
0.71%
(0.09%, 1.29%)
1.31%
(0.81%, 1.75%)
4.86%
(3.70%, 6.05%)
2.47%
0.05%
(1.78%, 3.21%)
(0.34%, 0.46%)
2.61%
(2.50%, 2.74%)
5.13%
3.66%
0.05%
(3.91%, 6.38%)
(2.59%, 4.76%)
(0.34%, 0.46%)
1.62%
0.78%
(0.66%, 0.90%)
0.29%
0.20%
(0.20%, 0.44%)
(0.14%, 0.31%)
0.03%
(0.01%, 0.05%)
(1.31%, 1.90%)
Second moments
r f
rm r f
1.13%
18.81%
rd r f
ric r f
6.48%
4.10%
(6.04%, 6.91%)
(3.80%, 4.39%)
2.22%
(2.07%, 2.37%)
6.83%
6.14%
(5.72%, 6.54%)
(2.07%, 2.37%)
(6.36%, 7.28%)
2.02%
(1.87%, 2.17%)
1.20%
(1.11%, 1.28%)
2.22%
Notes: This table reports asset-pricing implications. The rst and second moments of the risk-free rate, equity premium, and risk premiums in terms of the
tangible capital and intangible capital are reported. The summary statistics are annualized. The risk premiums in the models are levered following Boldrin
et al. (2001).
are signicantly amplied. The asset market statistics in this model are fully comparable with the data, which is a truly
signicant achievement. However, the actual size of the equity premium is not yet fully rationalized.
The third column shows that adding the product cycle further improves the models ability to match the asset market
statistics. Relative to the endogenous growth model without the product cycle, the mean risk-free rate is nearly halved to
0.7%; the equity premium is nearly doubled to 4.9%; and the volatilities of both the risk-free rate and the equity premium
are amplied by approximately 50%. Although this model is not yet carefully calibrated, it is notable that the model nearly
perfectly matches the rst moments of the data for both the risk-free rate and equity premium, suggesting that the product
cycle has the potential to improve the models ability to match asset market statistics in a fully calibrated version as well.
The seventh and eighth rows of the table report a breakdown of the aggregate stock market return. They show that the
large equity premium in the model with the product cycle is attributed to two factors. First, excess returns on individual
assets are elevated by the introduction of the product cycle. This implication must be related to the uctuations of the
stochastic discount factor and asset returns because the covariance between them is the source of excess return.11 Second,
the aggregate stock market return is more similar to the tangible capital return in the model with the product cycle than in
the model without the product cycle. This implication must be related to the sizes of assets in the steady state because the
aggregate stock market return is the value-weighted return of tangible and intangible capital.
To understand the rst point, I plot the impulse response functions of the stochastic discount factor, risk-free rate, and
excess returns on tangible capital and intangible capital for each of the three models in Fig. 1. The top panel shows that
the stochastic discount factor is most volatile in the endogenous growth model with the product cycle, second most volatile
in the endogenous growth model without the product cycle, and least volatile in the standard real business cycle model
with exogenous growth. Because all three models share the same recursive utility function, this result indicates that both
endogenous growth and the product cycle amplify the effects of a productivity shock on consumption growth, continuation
utility value, or both. The second panel plots the risk-free rate. The response is largest in the endogenous growth model
with the product cycle, second largest in the endogenous growth model without the product cycle, and smallest in the
standard real business cycle model with exogenous growth. Because the most important determinant of risk-free rate is
expected consumption growth, this result indicates that both endogenous growth and the product cycle amplify the effects
of a productivity shock on expected consumption growth. Strong persistence in the responses of the risk-free rate in the
endogenous growth models implies that the responses in the expected consumption growth are also persistent in these
models. The existence of this long-run consumption risk contributes to volatility of the continuation utility value in these
models. The bottom two panels plot the impulse response functions of excess returns on both tangible capital and intangible
capital. As in the top two panels, the responses are most volatile in the endogenous growth model with the product cycle,
11
Because E t [ M t +1 (1 + r f ,t )] = 1 and E t [ M t +1 (1 + rd,t +1 )] = 1, we have E t [ M t +1 (rd,t +1 r f ,t )] = 0. Using the denition of covariance and ignoring the
risk-free rate, i.e., 1 + r f ,t 1, we obtain
E t [rd,t +1 r f ,t ] = Covt ( M t +1 , rd,t +1 r f ,t )

The same is true for the excess return on intangible capital.
10
[m3G; v1.142; Prn:27/10/2014; 9:35] P.10 (1-21)
Fig. 1. Impulse response functions of nancial variables. This gure shows quarterly log-deviations from the steady state for the endogenous growth model
with the product cycle (solid line), the endogenous growth model without the product cycle (dashed line), and the standard real business cycle model with
exogenous growth (dotted line). A one-standard-deviation productivity shock hits the economy in the rst quarter. All deviations are multiplied by 100.
second most volatile in the endogenous growth model without the product cycle, and least volatile in the standard real
business cycle model with exogenous growth. Because a negative covariance between the stochastic discount factor and
asset return is the source of a risk premium, the responses of excess returns combined with the responses of the stochastic
discount factor mechanically explain why excess returns on these assets are largest in the endogenous growth model with
the product cycle, second largest in the endogenous growth model without the product cycle, and smallest in the standard
real business cycle model with exogenous growth.
To connect the responses of nancial variables to the responses of quantity variables, I plot the impulse response functions of the equilibrium productivity measure, aggregate net-value-added output, consumption, and investment in tangible
capital in the rst two rows of Fig. 2. In the standard real business cycle model with exogenous growth, these variables
initially respond positively to a productivity shock but quickly change courses and then slowly revert to the steady-state
growth path. In the endogenous growth models, however, initial jumps are followed by continuous growth, which eventually
leads these variables to a new steady-state growth path at a permanently higher level. The growth expectation stimulates
investment in tangible capital, which causes a large appreciation of the shadow value of capital, as seen in the left panel of
the third row.
In the right panel of the third row, I plot the impulse response functions of R&D spending in the endogenous growth
models. They are positive in both models, but the responses are larger in the model with the product cycle than in the
model without the product cycle. This quantitative difference explains why the growth rates of the standard macroeconomic
variables plotted in the rst two rows of Fig. 2 are generally higher in the model with the product cycle than in the model
without the product cycle. In the left panel of the fourth row, I plot the impulse response functions of the value of a
monopolistic product. Being consistent with the impulse response functions of R&D spending, the value of a monopolistic
product appreciates to a larger degree in the model with the product cycle than in the model without the product cycle.
The key to understand this difference is product composition. Specically, in the model with the product cycle, the ratio of
monopolistic products to competitive products increases after a productivity shock due to a high product entry rate. This
change in the product demographics alters the division of national income in favor of monopolistic products, as I plot in the
right panel of the fourth row. In other words, the product cycle creates additional room to accommodate more products by
squeezing both labor and capital shares. This mechanism amplies the responses of R&D spending and raises the value of a
monopolistic product.
The change in the demography of products also explains why the equilibrium productivity measure in the model with
the product cycle grows relatively slowly in the rst few years after the shock. In fact, the initial responses of the equilib-
[m3G; v1.142; Prn:27/10/2014; 9:35] P.11 (1-21)
11
Fig. 2. Impulse response functions of real variables. This gure shows the impulse response functions for the endogenous growth model with the product
cycle (solid line), the endogenous growth model without the product cycle (dashed line), and the standard real business cycle model with exogenous
growth (dotted line). I rst calculate impulse response functions of the stationary variables (C t / N t , N t +1 / N t , and so on); then, I translate them into
impulse response functions of level variables. Responses to a one-standard-deviation productivity shock are plotted. All deviations are multiplied by 100.
Table 4
Sizes of assets.
Variable
Description
HET
HOM
E [Qd /(4 Y )]
Tangible capital/annual GDP
262.53%
(243.47%, 280.48%)
216.56%
(204.98%, 227.49%)
E [Qic /(4 Y )]
Intangible capital/annual GDP
25.12%
(24.00%, 25.93%)
169.47%
(164.49%, 175.48%)
Notes: This table reports the sizes of tangible capital and intangible capital relative to the net-value-added output in the endogenous growth models. The
sizes of assets are measured by the market values.
rium productivity measure are nearly identical across the endogenous growth models, although the responses of intangible
investment are larger in the model with the product cycle than in the model without the product cycle. This is because
new products are monopolistic, and monopolistic products only weakly contribute to the equilibrium productivity measure.
There is, however, a reversal at longer times. That is, the invented products eventually become the targets of imitation,
causing a subsequent acceleration in the growth rate of the equilibrium productivity measure.
Now I discuss the second point: the effects of the product cycle on the sizes of assets. Table 4 shows that the size of the
tangible capital relative to the aggregate net-value-added output is 20% larger in the model with the product cycle than in
12
[m3G; v1.142; Prn:27/10/2014; 9:35] P.12 (1-21)
Table 5
Calibration II.
Parameter
Description
HET
HOM
Subjective discount factor

Elasticity of intertemporal substitution
Risk aversion
Intermediate-goods share
Gross markup
Capital share
Autocorrelation of
Survival rate of intermediate good
Elasticity of new intermediate goods wrt R&D
Depreciation rate of capital stock
Volatility of productivity shock
Elasticity of capital investment share
Probability of maintaining monopoly
Scale parameter
0.974
1.85
10
0.5
1.65
0.35
0.978
0.9625
0.83
0.02
1.27%
0.70
0.944
0.684
0.984
1.85
10
0.5
1.65
0.35
0.95
0.9625
0.83
0.02
1.75%
0.70
1
0.332
Notes: This table reports the parameter values used to simulate the endogenous growth model with the product cycle in Section 4. The fourth column
reproduces the parameter values calibrated by Kung and Schmid (in press).
the model without the product cycle, and the size of intangible capital to the aggregate net-value-added output is 85% lower
in the model with the product cycle than in the model without the product cycle. Because the aggregate stock market return
is a value-weighted return of tangible capital and intangible capital, this result explains why the aggregate stock market
return is closer to the tangible capital return in the model with the product cycle than in the model without the product
cycle. Tangible capital stock relative to aggregate net-value-added output is larger in the model with the product cycle than
in the model without the product cycle for at least two reasons. First, everything else being equal, the household in the
model with the product cycle has a stronger incentive to save because the product cycle amplies economic uctuations.
Second, everything else being equal, the product cycle makes exclusive production rights a less ecient saving tool by
shortening the duration of the monopoly. Both of these factors contribute to the large tangible capital stock in the model
with the product cycle.
4. Quantitative results with recalibration
In the previous section, I use the parameter values calibrated by Kung and Schmid (in press) to simulate the endogenous growth model with the product cycle. Although it is convenient to examine the role of the product cycle, borrowing
parameter values calibrated for a different model is problematic for making a quantitative assessment of the model. In this
section, I calibrate the parameter values for the endogenous growth model with the product cycle and report its quantitative
implications.
I follow Kung and Schmid for the calibration strategy. They handpicked parameter values for those parameters that are
standard in the literature and for those whose values can be set using direct empirical evidence. I set the same set of parameters at the same values. Kung and Schmid then set the following ve parameters based on simulation: the autocorrelation
of technology shock , the standard deviation of technology shock , the adjustment cost parameter , the scale parameter , and the subjective time discount factor . The empirical moments they attempt to match are the autocorrelation of
R&D intensity, i.e., S / N; the long-run output growth volatility12 ; the relative volatility of long-run consumption growth to
output growth; the growth rate along the balanced growth path; and the average risk-free rate. Having set the probability
of maintaining monopoly at the same value as in the previous section,13 I set these ve parameters in the same way.
I report the calibrated parameter values for the model with the product cycle in the third column of Table 5 and
reproduce the parameter values for the model without the product cycle calibrated by Kung and Schmid (in press) in the
fourth column of the same table for comparison. Notice that the volatility of the productivity shock is set at a smaller
value in the model with the product cycle than in the model without the product cycle, and the persistence of exogenous
productivity shocks is set at a larger value in the model with the product cycle than in the model without the product
cycle. The reasons for these changes are clear when comparing Fig. 2 and Fig. 3, which plot the impulse response functions
before and after recalibration, respectively. Before recalibration (Fig. 2), the volatilities of the output at medium frequencies
are larger in the model with the product cycle than in the model without the product cycle. After recalibration (Fig. 3),
they become comparable because the volatility of the productivity shock in the model with the product cycle is reduced.
Similarly, before recalibration (Fig. 2), the persistence of responses of intangible investment is weaker in the model with
the product cycle than in the model without the product cycle, but after recalibration (Fig. 3), they become similar because
the persistence of the exogenous productivity shock is reinforced in the model with the product cycle. Table 6 reports both
empirical and simulated moments relevant to the calibration. Overall, the simulated moments from the model with the
12
Kung and Schmid identify long-run components with frequencies of 100 to 200 quarters.
Calibrating to match the standard deviation of the rm entry relative to the standard deviation of the value added output at the business cycle
frequencies leads to a nearly identical value of the parameter.
13
[m3G; v1.142; Prn:27/10/2014; 9:35] P.13 (1-21)
13
Fig. 3. Impulse response functions in the recalibrated model. This gure shows the impulse response functions for the recalibrated, endogenous growth
model with the product cycle (solid line) and the endogenous growth model without the product cycle (dashed line). I rst calculate impulse response
functions of the stationary variables (C t / N t , N t +1 / N t , and so on); then, I translate them into impulse response functions of level variables. Responses to a
one-standard-deviation productivity shock are plotted. All deviations are multiplied by 100.
Table 6
Moments targeted in calibration.
Variable
Description
Data
HET
HOM
AC 1( S / N )
Autocorrelation of R&D intensity
0.93
0.91
(0.81, 0.97)
0.91
(0.81, 0.96)
LRgdp
Long-run output growth volatility
0.24%
0.24%
0.24%
(0.11%, 0.43%)
(0.11%, 0.43%)
LRc
Long-run consumption growth volatility
0.28%
0.23%
(0.10%, 0.40%)
0.22%
(0.10%, 0.39%)
E [ gdp ]
Average growth rate
1.90%
1.90%
(0.28%, 3.92%)
(0.74%, 2.98%)
1.25%
(0.17%, 2.29%)
1.31%
(0.81%, 1.75%)
E [r f ]
Average risk free rate
0.42%
1.89%
Notes: This table reports the empirical and simulated moments relevant to the calibration. Except for the mean risk-free rate, which is calculated by the
author, the data are taken from Kung and Schmid (in press).
[m3G; v1.142; Prn:27/10/2014; 9:35] P.14 (1-21)
14
Table 7
Asset-pricing implications.
Data
HET
HOM
0.42%
1.25%
(0.17%, 2.29%)
1.31%
(0.81%, 1.75%)
5.01%
4.95%
(4.01%, 5.97%)
2.47%
(1.78%, 3.21%)
E [rd r f ]
5.27%
(4.26%, 6.35%)
3.66%
(2.59%, 4.76%)
r ]
E [r ic
f
1.60%
(1.42%, 1.74%)
0.78%
(0.66%, 0.90%)
First moments
E [r f ]
r ]
E [rm
f
Second moments
r f
rm r f
1.13%
18.81%
0.27%
0.20%
(0.18%, 0.44%)
(0.14%, 0.31%)
5.50%
4.10%
(5.08%, 5.95%)
(3.80%, 4.39%)
rd r f
5.86%
(5.41%, 6.33%)
6.14%
(5.72%, 6.54%)
ric r f
1.53%
(1.41%, 1.67%)
(1.11%, 1.28%)
1.20%
Notes: This table reports the asset-pricing implications. The rst and second moments of the risk-free rate, equity premium, and risk premiums
in terms of the tangible capital and intangible capital are reported. The summary statistics are annualized. The risk premiums in the models
are levered following Boldrin et al. (2001).
product cycle match the empirical moments well. Both the long-run consumption growth volatility and the average risk-free
rate in the model with the product cycle are slightly different from the empirical moments, but they closely match the
simulated moments from the model without the product cycle. The main results do not change even if parameters are reset
so that these two simulated moments from the model with the product cycle are even closer to the data.
In Table 7, I report the asset-pricing implications. The second column reports the data, the third column reports the
simulated moments from the model with the product cycle, and the fourth column reproduces the simulated moments
from the model without the product cycle in the previous section for comparison. The mean equity premium in the model
with the product cycle is twice as large as that in the model without the product cycle; therefore, the model with the
product cycle is able to match the mean equity premium in the data. A breakdown of the aggregate stock market return
shows that the same two factors from the previous section contribute to this result: excess returns on individual assets are
amplied by the introduction of the product cycle and the aggregate stock market return is more similar to the tangible
capital return in the model with the product cycle than in the model without the product cycle. The volatilities of both
the risk-free rate and equity premium are larger in size and closer to the data in the model with the product cycle than
in the model without the product cycle. However, the difference between models in the volatility of the equity premium is
approximately 35%, while the difference between models in the mean equity premium is nearly 100%. This is problematic
for the Sharpe ratio; while it is already larger than the data in the model without the product cycle, the discrepancy from
the data is worsened in the model with the product cycle. A large Sharpe ratio implies that the stochastic discount factor
plays an important role in increasing the mean equity premium.
Fig. 4 reports the long-horizon autocorrelations of consumption growth. The left and right panels plot the simulated
moments from the models with and without the product cycle, respectively. Consumption growth is more persistent in the
model with the product cycle than in the model without the product cycle. This result explains why the stochastic discount
factor plays a more important role in increasing the equity premium in the model with the product cycle than in the model
without the product cycle. The two models, however, share important similarities in terms of the sample autocorrelations in
the data, which are plotted for comparison in each panel. First, they closely match the rst autocorrelation of the consumption growth in the data. Second, the second and third autocorrelations are negative in the data and positive in the models;
moreover, the empirical moments are outside the 90% probability bands at these data points. Third, the autocorrelations
with lags larger than ve are all within the 90% probability bands. In short, despite some quantitative differences, the two
models are broadly consistent with the data.
The next three Tables 8, 9, and 10 report the business cycle moments of both the actual data and the articially generated data from the model economies at high and medium frequencies. Comin and Gertler (2006) is a pioneering study
investigating these moments, showing that a version of the endogenous growth model performs well when matching a battery of business cycle statistics at both high and medium frequencies. Because our models are basically a simplied version
of theirs, this exercise is not intended to show any improvement on their study; instead, it is conducted to identify any
cost or benet associated with the introduction of the product cycle in terms of conventionally measured business cycle
statistics. The chance of the success is not very obvious because the targets of the calibration are not directly connected to
the moments in question.
[m3G; v1.142; Prn:27/10/2014; 9:35] P.15 (1-21)
15
Fig. 4. Autocorrelations of consumption growth. This gure reports the long-horizon autocorrelations of consumption growth. The data are taken from Kung
and Schmid (in press), who report sample autocorrelations for the period 19532008. Solid lines represent the median across 1000 model simulations of
the same sample size, and the shaded areas represent the 90% probability bands, respectively.
Table 8
Standard deviations (in percentages).
Variable
High frequency (232)

Data
Yt
2.34
1.12
N tE
TFPt
Qt
5.10
5.84
1.36
9.15
Data
HOM
1.01
1.39
(1.18, 1.66)
0.78
(0.66, 0.93)
3.27
0.83
(0.70, 1.00)
0.88
(0.75, 1.05)
0.57
(0.47, 0.70)
It
EXO
HET
(0.85, 1.20)
Ct
Medium frequency (32200)
ENDO
1.49
1.56
0.56
(1.26, 1.78)
(1.32, 1.86)
(0.48, 0.67)
2.51
(2.09, 3.03)
1.96
(1.64, 2.32)
1.01
1.40
0.78
(0.85, 1.20)
(1.18, 1.66)
(0.66, 0.93)
2.05
1.52
(1.29, 1.82)
0.82
(0.69, 0.98)
(1.74, 2.46)
ENDO
EXO
HET
HOM
3.19
(1.79, 5.39)
(2.08, 5.73)
1.52
(0.97, 2.41)
2.67
2.86
(1.53, 5.02)
2.83
(1.58, 4.91)
1.71
(1.09, 2.71)
6.08
3.67
(2.23, 6.03)
3.59
(2.21, 5.87)
1.13
(0.71, 1.78)
5.22
4.25
(2.68, 6.87)
(3.33, 8.46)
2.15
2.85
(1.66, 4.80)
3.23
(1.98, 5.22)
1.48
(0.94, 2.33)
23.00
4.98
(3.05, 8.13)
3.78
(2.28, 6.26)
1.57
(1.00, 2.48)
10.20
3.47
Notes: This table reports the standard deviations of the aggregate net-value-added output, consumption, tangible investment, rm entry, TFP, and stock
market value at high and medium frequencies extracted by a band-pass lter.
Table 8 reports the standard deviations of the aggregate net-value-added output, consumption, tangible investment, rm
entry, TFP, and stock market value.14 I report both empirical moments and the simulated moments from the endogenous
growth model with the product cycle, the simulated moments from the endogenous growth model without the product
cycle and the simulated moments from the standard real business cycle model with the exogenous growth considered in
the previous section. The band-pass lter of Christiano and Fitzgerald (2003) was applied to extract the high-frequency
(between 2 and 32 quarters) and medium-frequency components (between 32 and 200 quarters). The table shows that, in
general, the two endogenous growth models are able to generate empirically reasonable volatility sizes at both high and
medium frequencies, while the standard real business cycle model with exogenous growth is not. Standard deviations at
medium frequencies are particularly well replicated by the endogenous growth models; for three out of the six variables
reported in the table, the empirical moments are within the 90 percent probability bands. The exceptions to this are the
tangible investment, rm entry, and stock market value, the volatilities of which are smaller in the models than in the
14
Aggregate net-value-added output is measured as the production in the nonfarm business sector, consumption is measured as personal consumption
expenditures on service and nondurable goods, and tangible investment is measured as personal consumption expenditures on durable goods and xed
private investment; these data are taken from the BEA. The stock market value is measured as the total market value of the NYSE taken from the CRSP.
The aforementioned series are divided by both the nonfarm business sector implicit price deator, taken from the BEA, and the civilian, non-institutional
population over 16 years of age, taken from the BLS. TFP data are taken from Fernald (unpublished results). Firm entry data are generously shared by Marc
Melitz, measured as new incorporations minus failures. Net rm entry data are commonly used in the literature (Bilbiie et al., 2012; Lewis and Poilly,
2012). The sample period is 1948.Q1 to 1998.Q4.
[m3G; v1.142; Prn:27/10/2014; 9:35] P.16 (1-21)
16
Table 9
First autocorrelations.
Variable

Data
ENDO
EXO
HET
HOM
Data
ENDO
EXO
HET
HOM
Yt
0.84
0.68
(0.45, 0.83)
0.67
(0.45, 0.82)
0.67
(0.45, 0.82)
0.99
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
Ct
0.84
0.70
(0.48, 0.85)
0.68
(0.46, 0.83)
0.67
(0.45, 0.82)
0.99
0.99
(0.99, 0.99)
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
It
0.85
0.67
(0.44, 0.82)
0.67
(0.44, 0.82)
0.67
(0.44, 0.82)
0.99
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
N tE
0.78
0.67
(0.44, 0.82)
0.67
(0.44, 0.82)
0.99
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
TFPt
0.78
0.67
(0.45, 0.82)
0.67
(0.44, 0.82)
0.67
(0.45, 0.82)
0.99
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
Qt
0.75
0.67
(0.45, 0.82)
0.67
(0.45, 0.82)
0.67
(0.45, 0.82)
0.99
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
0.99
(0.98, 0.99)
Notes: This table reports the rst autocorrelations of the aggregate net-value-added output, consumption, tangible investment, rm entry, TFP, and stock
market value at high and medium frequencies extracted by a band-pass lter.
Table 10
Correlations with output.
Variable

Data
Ct
It
N tE
TFPt
Qt
0.78
0.82
0.38
0.83
0.43
ENDO
EXO
Data
HET
HOM
0.98
(0.96, 0.99)
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.81
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.70
0.99
(0.98, 0.99)
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.85
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
0.57
ENDO
HOM
0.97
0.97
(0.84, 0.99)
0.99
(0.99, 0.99)
0.99
(0.98, 0.99)
0.99
(0.99, 0.99)
0.99
(0.92, 0.99)
0.99
(0.93, 0.99)
0.99
(0.97, 0.99)
0.99
(0.99, 0.99)
0.99
(0.99, 0.99)
(0.85, 0.99)
0.98
(0.90, 0.99)
0.35
EXO
HET
0.93
(0.49, 0.99)
0.99
(0.92, 0.99)
0.97
(0.72, 0.99)
Notes: This table reports the contemporaneous correlations between the aggregate net-value-added output and the following variables at high and medium
frequencies extracted by a band-pass lter: consumption, tangible investment, rm entry, TFP, and stock market value.
data. Tangible investment in each of the three models is relatively stable because the adjustment cost is strong and labor
supply is inelastic. Firm entry is stable in the model economies as well, while a slightly higher volatility is observed in
the endogenous growth model with the product cycle than in the endogenous growth model without the product cycle.
The model economies cannot justify the actual stock market volatility either. However, because replication of this moment
is a well-known challenge (see, e.g., Cochrane, 2008), the ability of the endogenous growth model with the product cycle
to generate approximately one-fth of the actual stock markets volatility at both high and medium frequencies might be
considered a strength of the model.
Table 9 reports the rst autocorrelations, showing that all three models are able to replicate the persistence in the cyclical components of the data. Table 10 reports contemporaneous correlations with an aggregate net-value-added output. In
the data, consumption, tangible investment, TFP, stock market value and rm entry positively correlate with the aggregate
net-value-added output at high frequencies.15 All three models are able to replicate these patterns qualitatively, but the
correlations implied by the model are stronger than those seen in the data, which is not very surprising given the simple
structures of the models. Firm entry is, somewhat surprisingly, counter-cyclical at medium frequencies in the actual data,
though the endogenous growth models in the current manuscript cannot account for this. Despite such differences, I conclude that, in general, the endogenous growth models both with and without the product cycle are able to replicate the
basic patterns of the business cycle moments at both high and medium frequencies.
15
While the standard TFP measure is strongly procyclical with the output, the puried (i.e., utilization adjusted) TFP measure is not at high frequencies.
See Basu et al. (2006).
[m3G; v1.142; Prn:27/10/2014; 9:35] P.17 (1-21)
17
Table 11
Growth enhancing effect and static eciency effect.
Growth rate
Value-added output
Consumption
1.0
0.975
0.95
0.925
0.9
3.88%
0.64
0.42
2.01%
0.82
0.70
0.52%
1.0
1.0
0.64%
1.17
1.32
1.51%
1.31
1.65
Notes: This table reports selected statistics in the non-stochastic steady states in otherwise identical economies with different levels of monopoly protections. The fourth row reports quarterly growth rates. The fth and sixth rows report value-added outputs and consumptions under the assumption that
the economies are not only on the steady state growth paths but also have the same underlying technology levels measured by the mass of total products
available to the economy at the time of comparison. Value-added output and consumption are normalized using an economy in which the probability of
maintaining monopoly is 95 percent per quarter as a benchmark.
5. Discussion
5.1. Probability of maintaining monopoly
Given its critical role, it will be instructive to discuss the general macroeconomic implications of the probability of
maintaining monopoly. In this section, I experiment with different values of with all of the other parameters xed, namely,
at 0.98 and at 0.88; the other parameter values are reported in Table 2. A small assures that the continuation utility
value is properly dened throughout the exercise, which can be an issue when the non-stochastic steady state growth
rate is high. Table 11 reports selected statistics of the non-stochastic steady states in otherwise identical economies with
different values of . It shows that as Lerner (1934) rst notes, monopoly protection has a tradeoff. On the one hand, as
the fourth row of the table illustrates, longer monopoly durations are associated with a higher growth rate of the economy.
This is what Futagami and Iwaisako (2007) call the growth enhancing effect and is present because larger post-innovation
prots (in the present discounted value sense) stimulate innovative activities. On the other hand, a monopoly leads to an
inecient use of economic resources at any given point in time. This is what Futagami and Iwaisako call the static eciency
effect. I numerically assess this effect in the fth and sixth rows of the table by comparing the value-added outputs and
consumptions under the assumption that the economies are not only on the steady state growth paths but also have the
same technology levels (i.e., the same levels of N t ) at the time of comparison. It is clear that stronger monopoly protections
depress both the output and consumption. With respect to the consumption, there is a fourfold difference between the
perfect monopoly protection case and weak monopoly protection case, in which the probability of the monopoly expiring
is 10 percent per quarter. Based on this insight, Futagami and Iwaisako present a formal argument that an innite patent
length may not always maximize welfare.
Fig. 5 plots impulse response functions in economies with different values of , showing that a large value of amplies
the responses of the equilibrium productivity and other important macroeconomic variables to a positive productivity shock.
These are growth enhancing effects in the stochastic context, i.e., stronger monopoly protection encourages innovation
after a productivity shock. An interesting exception to this is the prot share, which uctuates more when the probability
of maintaining monopoly is low. This is because weak protections make a pool of monopolistic products (stock) more
comparable in size to the product entries (ow), making both the pool and share of monopolistic products more sensitive
to a change in the product entries. Notice that this exibility in the intensive margin amplies short-run responses of both
a monopolistic products value and rm entry rate.
Table 12 reports the asset pricing implications for different values of . The mean risk-free rate increases with the
probability of maintaining monopoly because it increases the steady state growth rate of the economy. The risk premium
also increases because a large value of amplies the volatilities of both the stochastic discount factor and the tangible
capital return. Because a large also implies a large R&D expenditure, these results are roughly consistent with Chan
et al. (2001) and Lin (2012), who documented that high R&D-intensive rms earn higher average stock returns than low
R&D-intensive rms. The same pattern is observed across industries as well.16
Notice, however, an important difference between the current and previous sections. Namely, in the current section,
all of the parameters, except for , are xed, while in the previous section, several important parameters are separately
calibrated across the models to target the average growth rate of the economy as well as both the long-run output growth
volatility and long-run consumption growth volatility. The calibration, therefore, largely equalizes the degrees to which the
aforementioned growth enhancing effects work at the aggregate level, allowing us to focus on the effect of endogenous
uctuations in the product composition.
16
I compare returns for 10 industries that are large R&D performers (BEAs research and development satellite account) and/or reported in surveys
that patents were important (Hall et al., 2014; Manseld, 1986), i.e., the aircraft, automobile and truck, chemical, communication, computer, computer
software, electrical equipment, medical equipment, petroleum and natural gas, and pharmaceutical product industries, with returns for 10 industries that
have opposite general characteristics, i.e., business services, business supplies, personal services, printing and publishing, railroad equipment, real estate,
recreation, shipbuilding, steel works etc., textiles, and wholesale. The data are taken from Kenneth Frenchs website. The average value weighted returns in
the past 30 years are clearly different between the two groups.
18
[m3G; v1.142; Prn:27/10/2014; 9:35] P.18 (1-21)
Fig. 5. Dynamics with different probabilities of maintaining monopoly. This gure shows the impulse response functions for otherwise identical endogenous growth models with different probabilities of maintaining monopoly, . I rst calculate the impulse response functions of the stationary variables
(C t / N t , N t +1 / N t , and so on); then, I translate them into impulse response functions of level variables. Responses are plotted to a one-standard-deviation
productivity shock. All deviations are multiplied by 100.
5.2. Conditional predictability

An important contribution of Kung and Schmid (in press) to the long-run risk literature is that they not only construct
a model economy in which time-varying growth prospects in consumption arise with an endogenous mechanism but also
present a direct empirical support for it. Namely, they document that R&D intensity, a variable they dene as R&D investment divided by the R&D stock, predicts consumption growth in the data as well as in the model. Because the current paper
emphasizes the importance of time-varying factor shares in the aggregate income as an additional amplication mechanism
associated with innovations, I empirically document this new channel following their study.
The rst panel of Table 13 reports the results from projecting log future prot shares in GDP for horizons of 1, 3, and 5
years on log R&D intensity.17,18 The slopes are positive in the data, and at the horizon of 5 years the slope is statistically
signicant and R 2 is economically sizable (0.15). I found consistent results in the model, by dening R&D intensity as S t / N t ,
The regression equation is yt +k = + xt + v t ,t +k where yt +k denotes a generic regressand and xt denotes a generic regressor.
Prot share is measured as the corporate prots after tax with inventory valuation and capital consumption adjustments divided by GDP. R&D intensity
is measured by private business R&D investment divided by private business R&D stock, and is observed from 1959 to 2006. The data are taken from BEA.
17
18
[m3G; v1.142; Prn:27/10/2014; 9:35] P.19 (1-21)
19
Table 12
Asset-pricing implications.
0.975
0.95
0.925
10.79%
(9.28%, 12.25%)
8.32%
(7.30%, 9.24%)
6.13%
(5.59%, 6.59%)
3.38%
(2.64%, 4.18%)
2.06%
(1.28%, 2.87%)
1.35%
(0.62%, 2.08%)
E [rd r f ]
4.16%
(3.23%, 5.16%)
2.24%
(1.39%, 3.11%)
1.39%
(0.64%, 2.15%)
r ]
E [r ic
f
1.03%
(0.84%, 1.19%)
0.75%
(0.48%, 0.99%)
0.63%
(0.28%, 0.95%)
r f
0.34%
(0.22%, 0.58%)
0.26%
(0.18%, 0.42%)
0.19%
(0.14%, 0.29%)
rm r f
4.36%
(3.99%, 4.78%)
4.40%
(4.09%, 4.71%)
4.03%
(3.76%, 4.30%)
rd r f
5.40%
(4.99%, 5.88%)
4.77%
(4.44%, 5.11%)
4.15%
(3.87%, 4.42%)
ric r f
1.24%
(1.14%, 1.34%)
1.55%
(1.43%, 1.66%)
1.83%
(1.70%, 1.96%)
First moments
E [r f ]
r ]
E [rm
f
Second moments
Notes: This table reports the asset-pricing implications in economies that are otherwise identical with different probabilities of maintaining monopoly. The
rst and second moments of the risk-free rate, equity premium, and risk premiums in terms of the tangible capital and intangible capital are reported. The
summary statistics are annualized. The risk premiums are levered following Boldrin et al. (2001).
Table 13
Predictability of prot shares, consumption growth, and excess returns.
Horizon
(years)
Data
HET
S.E.
R2
Prot share forecasts with R&D intensity

1
1.10
3
1.98
5
4.21
1.11
1.54
1.30
0.01
0.04
0.15
Consumption growth forecasts with R&D intensity

1
0.21
3
0.49
5
0.60
0.11
0.27
0.35
Excess return forecasts with pricedividend ratio

1
0.08
3
0.37
5
0.66
0.07
0.16
0.21
S.E.
R2
0.42
0.41
0.33
0.02
0.03
0.06
0.84
0.80
0.53
0.07
0.08
0.07
0.09
0.22
0.31
0.02
0.05
0.09
0.45
0.40
0.35
0.02
0.19
0.37
0.16
0.13
0.11
0.10
0.10
0.10
0.03
0.02
0.02
Notes: This table presents evidence of the prediction of future prot shares and future consumption growth rates by the R&D intensity (the rst panel and
the second panel) and the prediction of future excess returns by the pricedividend ratios (the third panel) in the data and the endogenous growth model
with the product cycle (HET) for horizons (k) of 1, 3, and 5 years. The regressions are estimated via OLS with NeweyWest standard errors using k 1 lags
with overlapping annual observations. The estimates from the model regression are averaged across 1000 simulations that are equivalent in length to the
data sample. The excess return predictability data are taken from Bansal and Yaron (2004).
following Kung and Schmid, and the prot share as N tM t /Yt . The result was robust with other specications as well.19 The
second panel reports that R&D intensity predicts consumption growth both in the data and in the model. Together, these
results give empirical support for mechanisms emphasized by the current manuscript; that is, innovations have implications
on both consumption growth and factor shares.
Finally, I explored whether the model economy could replicate a well-known empirical feature in nance, namely, that
the current dividend yield predicts a rise in future expected returns. I ran the same regressions as in Bansal and Yaron (2004)
using the simulated data, dening the pricedividend ratio as Qt /Dt . The third panel of Table 13 reports that there is little
evidence that the model captures the positive relationship between the expected returns and dividend yields. While the
model has endogenous movements in factor shares, they are probably too weak to explain the excess-return predictability.
19
As a robustness check, I use the dividend share as the regressand and the growth rate of the R&D stock and log rm entry as the regressor.
20
[m3G; v1.142; Prn:27/10/2014; 9:35] P.20 (1-21)
6. Conclusion
This paper studies an endogenous growth model incorporating the product cycle. I nd that imitation is a powerful
amplication mechanism with which the canonical endogenous growth model is able to match the equity premium nearly
perfectly. I also show that the modied model is largely consistent with the data on a battery of business cycle statistics.
The current paper thereby supports a promising theory on the economic sources of long-run risks originally proposed by
Kung and Schmid (in press): innovation and R&D cause long-run uncertainties in economic growth.
The product cycle may have broader implications on topics that are not considered in this paper. Jinnai (2014), for
example, discusses its implication for the news shock literature (Barsky and Sims, 2011; Beaudry and Portier, 2004, 2006;
Jaimovich and Rebelo, 2009). Specically, I show that an unexpected improvement in sector-specic productivity in the
research and development sector can be seen as productivity news. In addition, I show that although theoretical studies in
the literature generally nd it dicult to replicate a stock market boom in response to a news shock, with the product cycle,
the simple endogenous variety model of Bilbiie et al. (2012) is able to replicate this pattern. Other cyclical implications of
the product cycle have yet to be fully explored. Stylized modeling of both imitation and product exit should be elaborated.
Cross-sectional heterogeneity is an especially important consideration because, in reality, some lines of business are natural
monopolies whereas others enjoy very small monopoly power. Different industries have different degrees of innovation, too.
These considerations are ignored by the current manuscript, but investigating them will be an interesting future research
topic. Such efforts, I believe, will lead to a better understanding of productivity, business cycles, and asset markets.
Acknowledgments
I thank Nobu Kiyotaki for his advice and encouragement. I thank Marc Melitz for generously sharing his data. I am
also grateful to Roland Benabou, Saroj Bhattarai, Francesco Bianchi, Toni Braun, Kenichi Fukushima, Jean-Francois Kagy,
Amy Glass, Tatsuro Iwaisako, Dennis Jansen, Urban Jermann (editor), Dirk Kruger, Per Krusell, Alisdair McKay, Marc Melitz,
Tamas Papp, Woong Yong Park, Bruce Preston, Ricardo Reis, Esteban Rossi-Hansberg, Felipe Schwartzman, Chris Sims, Satoru
Shimizu, Takuo Sugaya, Jade Vichyanond, Yuichiro Waki, Steve Wiggins, Anastasia Zervou, and an anonymous referee for
their comments. A part of this research was supported by JSPS KAKENHI Grant No. 24330094.
References
Acemoglu, D., 2008. Introduction to Modern Economic Growth. Princeton University Press, Princeton, NJ.
Aghion, P., Howitt, P., 1997. Endogenous Growth Theory. MIT Press, Cambridge, MA.
Ai, H., Croce, M.M., Li, K., 2013. Toward a quantitative general equilibrium asset pricing model with intangible capital. The Review of Financial Studies 26,
491530.
Bansal, R., Yaron, A., 2004. Risks for the long run: a potential resolution of asset pricing puzzles. The Journal of Finance 59, 14811509.
Barro, R.J., 2006. Rare disasters and asset markets in the twentieth century. The Quarterly Journal of Economics 121, 823866.
Barsky, R.B., Sims, E.R., 2011. News shocks and business cycles. Journal of Monetary Economics 58, 273289.
Basu, S., Fernald, J.G., Kimball, M.S., 2006. Are technology improvements contractionary? The American Economic Review 96, 14181448.
Beaudry, P., Portier, F., 2004. An exploration into Pigous theory of cycles. Journal of Monetary Economics 51, 11831216.
Beaudry, P., Portier, F., 2006. Stock prices, news, and economic uctuations. The American Economic Review 96, 12931307.
Bilbiie, F.O., Ghironi, F., Melitz, M.J., 2012. Endogenous entry, product variety, and business cycles. Journal of Political Economy 120, 304345.
Boldrin, M., Christiano, L.J., Fisher, J.D.M., 2001. Habit persistence, asset returns, and the business cycle. The American Economic Review 91, 149166.
Boldrin, M., Levine, D.K., 2008. Against Intellectual Monopoly. Cambridge University Press, Cambridge, UK.
Boldrin, M., Levine, D.K., 2013. The case against patents. The Journal of Economic Perspectives 27, 322.
Chan, L.K.C., Lakonishok, J., Sougiannis, T., 2001. The stock market valuation of research and development expenditures. The Journal of Finance 56,
24312456.
Christiano, L.J., Fitzgerald, T.J., 2003. The band pass lter. International Economic Review 44, 435465.
Cochrane, J.H., 2008. Financial markets and the real economy. In: Mehra, R. (Ed.), Handbook of the Equity Risk Premium. Elsevier, Amsterdam, The Netherlands, pp. 237330.
Comin, D., Gertler, M., 2006. Medium-term business cycles. The American Economic Review 96, 523551.
Comin, D.A., Gertler, M., Santacreu, A.M., unpublished results. Technology innovation and diffusion as sources of output and asset price uctuations. National
Bureau of Economic Research Working Paper No. 15029.
Croce, M.M., 2014. Long-run productivity risk: a new hope for production-based asset pricing? Journal of Monetary Economics 66, 1331.
Epstein, L.G., Zin, S.E., 1989. Substitution, risk aversion, and the temporal behavior of consumption and asset returns: a theoretical framework. Econometrica 57, 937969.
Favilukis, J., Lin, X., unpublished results. Wage rigidity: a solution to several asset pricing puzzles. Fisher College of Business Working Paper No. 2012-03-016.
Fernald, J., unpublished results. A quarterly, utilization-adjusted series on total factor productivity. Federal Reserve Bank of San Francisco Working Paper
2012-19.
Futagami, K., Iwaisako, T., 2007. Dynamic analysis of patent policy in an endogenous growth model. Journal of Economic Theory 132, 306334.
Gabaix, X., 2012. Variable rare disasters: an exactly solved framework for ten puzzles in macro-nance. The Quarterly Journal of Economics 127, 645700.
Garleanu, N., Kogan, L., Panageas, S., unpublished results. The demographics of innovation and asset returns. National Bureau of Economic Research Working
Paper No. 15457.
Garleanu, N., Panageas, S., Yu, J., 2012. Technological growth and asset pricing. The Journal of Finance 67, 12651292.
Gourio, F., 2012. Disaster risk and business cycles. The American Economic Review 102, 27342766.
Grossman, G.M., Helpman, E., 1991. Innovation and Growth. MIT Press, Cambridge, MA.
Hall, B., Helmers, C., Rogers, M., Sena, V., 2014. The choice between formal and informal intellectual property: a review. Journal of Economic Literature 52,
375423.
Iraola, M.A., Santos, M.S., unpublished results. Long-term asset price volatility and macroeconomic uctuations. University of Miami.
[m3G; v1.142; Prn:27/10/2014; 9:35] P.21 (1-21)
21
Jaimovich, N., Rebelo, S., 2009. Can news about the future drive the business cycle? The American Economic Review 99, 10971118.
Jermann, U.J., 1998. Asset pricing in production economies. Journal of Monetary Economics 41, 257275.
Jinnai, R., 2014. R&D shocks and news shocks. Journal of Money, Credit, and Banking 46, 14571478.
Kaltenbrunner, G., Lochstoer, L.A., 2010. Long-run risk through consumption smoothing. The Review of Financial Studies 23, 31903224.
Kung, H., Schmid, L., in press. Innovation, growth, and asset prices. The Journal of Finance.
Lerner, A.P., 1934. The concept of monopoly and the measurement of monopoly power. Review of Economic Studies 1, 157175.
Levin, R.C., Klevorick, A.K., Nelson, R.R., Winter, S.G., 1987. Appropriating the returns from industrial research and development. Brookings Papers on Economic Activity 1987, 783831.
Lewis, V., Poilly, C., 2012. Firm entry, markups and the monetary transmission mechanism. Journal of Monetary Economics 59, 670685.
Lin, X., 2012. Endogenous technological progress and the cross-section of stock returns. Journal of Financial Economics 103, 411427.
Manseld, E., 1986. Patents and innovation: an empirical study. Management Science 32, 173181.
Manseld, E., Schwartz, M., Wagner, S., 1981. Imitation costs and patents: an empirical study. The Economic Journal 91, 907918.
Mehra, R., Prescott, E.C., 1985. The equity risk premium: a puzzle. Journal of Monetary Economics 15, 145161.
Moser, P., 2013. Patents and innovation: evidence from economic history. The Journal of Economic Perspectives 27, 2344.
Pastor, L., Veronesi, P., 2009. Technological revolutions and stock prices. The American Economic Review 99, 14511483.
Rietz, T.A., 1998. The equity risk premium a solution. Journal of Monetary Economics 22, 117131.
Romer, P.M., 1990. Endogenous technological change. Journal of Political Economy 98, S71S102.
Rudebusch, G.D., Swanson, E.T., 2012. The bond premium in a DSGE model with long-run real and nominal risks. American Economic Journal: Macroeconomics 4, 105143.
Scherer, F., Ross, D., 1990. Industrial Market Structure and Economic Performance, third ed. Houghton Miin Company, Boston, MA.
Tallarini Jr., T.D., 2000. Risk-sensitive real business cycles. Journal of Monetary Economics 45, 507532.

1 s2.0 S109420251400060X Main PDF

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

1 s2.0 S109420251400060X Main PDF

Hochgeladen von

Copyright:

Verfügbare Formate

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.1 (1-21)

Review of Economic Dynamics ()

Contents lists available at ScienceDirect

Review of Economic Dynamics

Innovation, product cycle, and asset prices

E-mail address: rjinnai@econmail.tamu.edu.

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.2 (1-21)

R. Jinnai / Review of Economic Dynamics ()

A consistent result is found in Manseld et al. (1981).

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.3 (1-21)

R. Jinnai / Review of Economic Dynamics ()

C t + Qt Zt +1 + B t +1 = W t L + (Qt +Dt )Zt + R t B t

where the composite G t is dened as

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.4 (1-21)

R. Jinnai / Review of Economic Dynamics ()

where D t is the rms dividends

where P G ,t is the price index dened as

and the total expenditure on intermediate goods is given by

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.5 (1-21)

R. Jinnai / Review of Economic Dynamics ()

Solving this problem leads to the optimal markup pricing,

The producer of a competitive product sets the price

at the marginal cost because competition drives prots to zero:

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.6 (1-21)

R. Jinnai / Review of Economic Dynamics ()

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.7 (1-21)

R. Jinnai / Review of Economic Dynamics ()

Stochastic discount factor

Stock market value

Tangible capital value

Intangible capital value

Qtic = E t [ M t +1 (t +1 N tM+1 S t +1 + Qtic+1 )]

Productivity in R&D sector

Therefore, combining the two equations above, we nd

M t +1+i N tM+1+i t +1+i

The returns are dened accordingly.

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.8 (1-21)

R. Jinnai / Review of Economic Dynamics ()

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.9 (1-21)

R. Jinnai / Review of Economic Dynamics ()

E t [rd,t +1 r f ,t ] = Covt ( M t +1 , rd,t +1 r f ,t )

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.10 (1-21)

R. Jinnai / Review of Economic Dynamics ()

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.11 (1-21)

R. Jinnai / Review of Economic Dynamics ()

Tangible capital/annual GDP

Intangible capital/annual GDP

JID:YREDY AID:702 /FLA

[m3G; v1.142; Prn:27/10/2014; 9:35] P.12 (1-21)

R. Jinnai / Review of Economic Dynamics ()

Subjective discount factor