a framework for analyzing the impact of business cycles on endogenous growth
TRANSCRIPT
Introduction Model Results Conclusions
A Framework for Analyzing the Impactof Business Cycles on Endogenous Growth
Marcin Bielecki
University of WarsawFaculty of Economic Sciences
PhD Students’ Seminar
March 16, 2015
Introduction Model Results Conclusions
Motivation
Business cycles literature typically employs a variant of theso-called neoclassical growth model, where the trend growth isassumed to be exogenous and constant over time
Endogenous growth literature typically abstracts from theshort-term fluctuations and focuses on the balanced growthpath results or the transition dynamics
My research program aims to fill the gap in the literature byemploying a single framework to analyze both business cycleand growth phenomena and seek links between the two
Introduction Model Results Conclusions
Literature review
The seminal paper of Aghion and Howitt (1992) rekindledinterest in Schumpeterian-type endogenous growth theory
Klette and Kortum (JPE 2004) develop a very rich andpowerful model of product innovation performed byheterogeneous firms
Acemoglu et al. (NBER 2013) provide further refinements byincluding firms’ heterogeneity with respect to their innovativecapacity
Bilbiie et al. (JPE 2012) use the closed economy Melitz modelto relate endogenous firm entry decisions to the business cycle
Introduction Model Results Conclusions
Households
Standard CRRA utility function
Ut =
∞∑τ=0
βτC1−θt+τ − 1
1− θ (1)
No storage technology – the entire output is consumed
Ct = Yt (2)
Constant inelastic labor supply with a fraction s of workerssupplying skilled labor Ls and a fraction 1− s supplyingunskilled labor Lu
Lt = L (3)
Lst = sL (4)
Lut = (1− s)L (5)
Introduction Model Results Conclusions
Final Goods Producer
A perfectly competitive representative firm producing finalgoods output from mass Mt of intermediate goods yit,where σ denotes the elasticity of substitution between varieties
Yt =
[∫ Mt
0yσ−1σ
it di
] σσ−1
(6)
There exists an associated price index P of the final good,where pit denotes the price of i-th variety
Pt =
[∫ Mt
0p1−σit di
] 11−σ
(7)
The resulting demand function for an intermediate good
yit = YtPσt p−σit (8)
Introduction Model Results Conclusions
Intermediate Goods Producers
There exists a mass Mt ∈ (0, 1) of active intermediate goodsproducing establishmentsEach period an establishment hires f units of skilled labor andgains access to the following production technology
yit = ztqit`it (9)
where zt is an aggregate productivity shock,qit is an establishment-specific quality parameter,and `it denotes units of employed unskilled labor
An establishment’s (nominal) marginal cost is proportional tothe (nominal) unskilled wage wt and inversely proportional tozt and qit, and the optimal pricing rule is in fact a constantmark-up applied to the marginal cost
pit =σ
σ − 1︸ ︷︷ ︸mark-up
wtztqit︸︷︷︸
marginal cost
(10)
Introduction Model Results Conclusions
Intermediate Goods Producers
An establishment’s operating profit can be expressed as follows
πoit =(σ − 1)σ−1
σσYtP
σt z
σ−1t w1−σ
t qσ−1it − wst f (11)
where wst is the skilled labor wage
Establishments can improve their goods’ quality by investingin R&D activities. The resulting improvements are bestthought of as process (rather than product) innovations
R&D costs are expressed as follows
cR&D (wst , xit) = wst · x (qit, Qt, αit) (12)
where xit denotes the demand for R&D capable labor, whichdepends on the establishment’s quality, the aggregate qualityindex of all intermediates Qt and the success probability αit
Introduction Model Results Conclusions
Success probability function
R&D intensity x0
1
Succ
ess
prob
abili
tyα
Introduction Model Results Conclusions
R&D Decision
The success probability αit is a function of R&D productivity
parameter a and the adjusted R&D intensityxit
(qit/Qt)σ−1
(similar approach to Ericson and Pakes (RES 1995))
α (xit, qit, Qt) =a xit(qit/Qt)
σ−1
1 + a xit(qit/Qt)
σ−1
(13)
The idea behind the adjustment is that if an establishment issignificantly more productive than the others, it has hardertime generating new ideas, whereas the ones that are lessproductive can imitate the successful establishments
At this point I introduce a new variable φit ≡ (qit/Qt)σ−1, so
that
α (xit, φit) =axit/φit
1 + axit/φit(14)
Introduction Model Results Conclusions
R&D Decision
The success probability function can be inverted to yield thedemand function for R&D capable labor
α (xit, φit) =axit/φit
1 + axit/φit(15)
x (φit, αit) =1
a
(αit
1− αit
)φit (16)
The operating profit of an establishment can be rewrittenusing the relative productivity variable φit as follows
πoit =PtYtσMt
φit − wst f (17)
The total profit equation is a linear function in φit
πit =PtYtσMt
φit − wst f︸ ︷︷ ︸operating profit
−wst1
a
(αit
1− αit
)φit (18)
Introduction Model Results Conclusions
Value Function
The profit function
πit =
[PtYtσMt
− wsta
(αit
1− αit
)]φit − wst f (19)
The value function
Vt (φit) = maxαit∈[0,1]
{πit (φit)
Pt+ max {0,Et [Λt,t+1Vt+1 (φi,t+1|φit, αit)]}
}(20)
where Λt,t+1 = β(Yt+1
Yt
)−θ(1− δ) is the stochastic discount factor
with δ denoting the incumbent exit probability and
φi,t+1 =
ιφitηt
with probability αit
φitηt
with probability 1− αit(21)
where ι is the incremental innovation step size and ηt is the rate ofgrowth of the aggregate quality index
Introduction Model Results Conclusions
Value Function
I drop the subscript i since an establishment’s solutiondepends only on its relative quality variable φ. Also, I usethe ′ notation to denote the t+ 1 period’s variables
V (φ, Y,M, ωs) = maxα∈[0,1]
Y
{[1
σM− ωs
a
(α
1− α
)]φ− ωsf
}(22)
+ max{
0,E[β (Y ′/Y )
−θ(1− δ)V
(φ′, Y ′,M ′, (ωs)
′ |φ, Y,M, ωs, α)]}
where ωs ≡ ws/Y
Introduction Model Results Conclusions
Balanced Growth Path
Along the BGP the value function is linear (γ ≡ Y ′/Y )
V (φ) = maxα∈[0,1]
Y
{[1
σM− ωs
a
(α
1− α
)]φ− ωsf
}(23)
+ max
0,E
βγ1−θ (1− δ)︸ ︷︷ ︸ϑ
V(φ′|φ, α
)
Introduction Model Results Conclusions
R&D Intensity (Partial Equilibrium)
α∗ =
aσMωs − 1−ϑ
ϑηι−η
1 + aσMωs
(24)
R&D intensity α∗ is the larger:
the lower is the number of active establishments M
the higher is the innovative step size ι
the closer to 1 is the discounting factor ϑ
the lower is the high skilled wage relative to output ωs
the higher is the unit R&D productivity a
the lower is the aggregate quality index growth rate η(good for numerical stability)
Introduction Model Results Conclusions
Entry and General Equilibrium
A prospective entrant solves the following problem (the R&Dcost function is the same as for the φ = 1 incumbent)
VE = maxαE∈[0,1]
{αEβγ
1−θEV − ws[fE +
1
aE
(αE
1− αE
)]}(25)
Free Entry Condition ensures that VE = 0
A successful entrant has a (1− δexo)M chance of replacingan incumbent and 1− (1− δexo)M chance of starting a newproduct line, with δexo being the ‘pure’ exogenous exitprobability. The resulting incumbent survival probability is
(1− δ) = (1− δexo)[
1−M1− (1− δexo)M
](26)
Constant Mass of Firms
Labor Market Clears
Introduction Model Results Conclusions
General Equilibrium – Numerical Solution Procedure
1 Compute αE
2 Guess M and ωs
3 Compute δ
4 Jointly determine α and η in a loop
5 Update M and ωs and iterate steps 2-5 until convergence
Introduction Model Results Conclusions
Entry Costs and Fixed Costs vs Growth
Entry cost fE
0.0
0.5
1.0
1.5
2.0
Fixed co
stf
1.0
1.2
1.4
1.6
1.8
2.0
2.2
Gro
wth
rateγ
+1.0
02
0.001
0.002
0.003
0.004
0.005
Introduction Model Results Conclusions
Entry Costs and Fixed Costs vs Active Establishments
Entry cost fE0.0
0.5
1.0
1.5
2.0
Fixed co
stf
1.0
1.2
1.4
1.6
1.8
2.0
2.2
Mas
sof
acti
vepr
oduc
tlin
esM
0.04
0.05
0.06
0.07
0.08
0.09
0.10
Introduction Model Results Conclusions
Entry Costs and Fixed Costs vs Welfare
Entry cost fE
0.0
0.5
1.0
1.5
2.0
Fixed co
stf
1.0
1.2
1.4
1.6
1.8
2.0
2.2
Uti
lityU
−0.612
−0.610
−0.608
−0.606
−0.604
−0.602
−0.600
Introduction Model Results Conclusions
Growth and Business Cycles (Work In Progress)
V (φ, Y,M, ωs) = maxα∈[0,1]
Y
{[1
σM− ωs
a
(α
1− α
)]φ− ωsf
}+ max
{0,E
[β (Y ′/Y )
−θ(1− δ)V
(φ′, Y ′,M ′, (ωs)
′ |φ, Y,M, ωs, α)]}
Y = M1
σ−1ZQL
lnZ ′ = ρ lnZ + εZ
Along the business cycle, a change in Y can be caused by change inM , Q or Z. Thus, if changes in Z influence M or Q, then largevariation in Y may not require large variation in Z if theamplification mechanism is strong enough.
Introduction Model Results Conclusions
Preview of Future Results
Impulse response functions for output under the RBC model (red)and this model (blue)
0 5 10 15 20 25 30 35 40-0.5
0
0.5
1
Introduction Model Results Conclusions
Preview of Future Results
The persistence and amplitude of exogenous shocks has animpact on the average growth rate of an economy
Entry costs have a decisive role in the behavior of entry alongthe business cycle
Fixed costs impact mainly Balanced Growth Path behavior
Hypothesis: given the nature of the exogenous shocks, thereis a welfare-optimal combination of f and fE
Introduction Model Results Conclusions
Thank you for your attention