III Economic Growth (continued)

D0 Endogenous Growth: Lucas’ (1988) Model

1 Introduction

• The problem of economic development: the problem of account-

ing for the observed pattern, across countries and across time, in
levels and rates of growth per capita income.
• Per capita income levels and growth rates are diverse. “I do not
see how one can look at figures like these without seeing them as
representing possibilities. Is there some action a government of
India could take that would lead the Indian economy to grow like
Indonesia’s or Egypt’s? If so, what, exactly? If not, what is it
about the ‘nature of India’ that makes it so? The consequences
for human welfare involved in questions like these are simply
staggering: Once one starts to think about them, it is hard to
think about anything else.”
• I prefer to use the term ‘theory’ in a very narrow sense, to refer
to an explicit dynamic system, something that can be put on a
computer and run.

58
2 Neoclassical Growth Theory: Review

• Preferences:
 
1−σ
−ρt  c − 1
Z ∞
e   N dt, (1)
0 1−σ
where N = N0eλt.
• Technologies:

Y = AK β N 1−β , (2)

where Ȧ/A = µ > 0.

• Market Clearing Condition (Resources Constraint):

Y = N c + K̇, (3)

• Social Planner’s Problem: The social planner’s problem is to

choose K to maximize (1) subject to (2) and (3), where K(0), N (0)
and A(0) are given and K(∞) = K∞ ≥ 0 is free. The current-
value Hamiltonian is
c1−σ − 1
H=N + θ(Y − N c),
1−σ
where Y is given by (2). The first-order conditions are

N c−σ − N θ = 0, (4)

Y − N c = K̇, (5)

θβY /K = ρθ − θ̇, (6)

lim Kθe−ρt = 0, lim He−ρt = 0, lim θe−ρt ≥ 0, lim K ≥ 0.

t→∞ t→∞ t→∞ t→∞

59
 Solving the first-order conditions:
θ ċ
(4) ⇒ c−σ = θ ⇒ = −σ (7)
θ c
θ
(6) ⇒ = ρ − βY /K (8)
θ
ċ βY /K − ρ
(7) and (8) ⇒ g ≡ = (9)
c σ
K̇ Y N c σg + β N c
(5) ⇒ = − = − (10)
K K K β K

Since N c/K is constant (due to K̇/K and (σg + β)/β are con-
stant), we have
K̇ Ṅ ċ
= + =λ+g (11)
K N c
(9) ⇒ βY /K = σg + ρ ⇒ βAK β−1N 1−β = σg + ρ (12)

Differentiating (12) with respect t gives

Ȧ K̇ Ṅ
+ (β − 1) + (1 − β) = 0
A K N
⇒ µ + (β − 1)(g + λ) + (1 − β)λ = 0
µ
⇒g= .
1−λ
• Growth effects vs. level effects: The parameters (σ, ρ) do not
have growth effects, that is, they do not affect the growth rate

60
 g. But they have level effects, that is, they affect the income
level. The level effects can be seen from the savings rate
S K̇ K̇/K (λ + g)β
s= = = = .
Y Y Y /K σg + ρ
where s depends on the parameters (σ, ρ).
• Optimality of decentralized equilibrium: In this model, the de-
centralized equilibrium is Pareto optimal.
• The US economy [Based on Denison’s (1961) study (1909-1957)]:
λ = 1.3%, g + λ = 2.4% or 2.9% (take the average = 2.7),
β = 0.25 and s = 10%. These parameters give g = 1.4% and
µ = 1.05%. The savings rate equation implies that ρ and σ
satisfy
ρ + 0.014σ = 0.067.
Note that either output growth is underpredicted or capital growth
overpredicted.

3 Neoclassical Growth Theory: Assessment

• Under the assumption of no factor mobility, the neoclassical

growth theory is unable to account for the diversity in income
levels and growth rates we observed. When factor mobility is
permitted, the neoclassical theory predicts a stronger tendency
to income equality and equality in growth rates.
• Variations in the parameter (ρ, σ, λ, β) and initial technology
levels A(0) cannot explain why the observed income levels and
growth rates are so diverse.
61
 • Off-steady-state consideration also has difficulty explaining the
large differences in income and growth:
gyt = βgkt + µ.
• Variations in ‘technology’ has the potential to account for large
differences in income levels and growth rates.
• The role of human capital: How does human capital accumula-
tion affect the level and growth rate of income?

4 Human Capital and Growth

• Preferences: the same as (1).

• Technologies:
(i) Final good:
Y = AK β (N hu)1−β hγa
where A > 0 (a constant), 0 < β < 1 and γ ≥ 0.
(ii) Human capital:
ḣ = δh(1 − u)
where δ > 0.
• Market clearing condition (Resources constraint)
Y = N c + K̇.
• Decentralized equilibrium:
 
1−σ
−ρt  c − 1
Z ∞
Max e   N dt
0 1−σ
62
 subject to K̇ = AK β (N hu)1−β hγa

ḣ = δh(1 − u).

The current-value Hamiltonian

c1−σ − 1
H=N + θ1(Y − N c) + θ2δ(1 − u)h.
1−σ
The first-order conditions are

N c−σ − N θ1 = 0, (13)

θ1(1 − β)Y /u − θ2δh = 0 (14)

θ1βY /K = ρθ1 − θ̇1, (15)

θ2(1 − β)Y /h + θ2δ(1 − u) = ρθ2 − θ̇2, (16)

Y − N c = K̇, (17)

δh(1 − u) = ḣ, (18)

lim He−ρt = 0, lim θie−ρt ≥ 0, i = 1, 2,

t→∞ t→∞

lim xi ≥ 0, lim xiθie−ρt = 0, xi = K, h.

t→∞ t→∞

• Steady State: g = ċ/c, v = ḣ/h and u = constant. Now we find

the steady state growth rates. First, we find two expressions for
θ̇1/θ1.

θ̇1 ċ
(13) ⇒ = −σ = −σg (19)
θ1 c

63
 θ̇1 Y
(15) ⇒ =ρ−β (20)
θ1 K

(19) and (20) ⇒ ρ+σg = βY /K = βK β−1(N hu)1−β hγa .(21)

Differentiating (21)with respect t gives (note that ha = h in

equilibrium)

0 = (β − 1)(λ + g) + (1 − β)(λ + v) + γv

0 = g(β − 1) + (1 − β + γ)v. (22)

Second, we find two expressions for θ̇2/θ2.

θ1 δhu θ̇1 θ̇2 ḣ Ẏ

(14) ⇒ = ⇒ − = −
θ2 (1 − β)Y θ1 θ2 h Y

which, along with θ̇1/θ1 = −σg, implies

θ̇2
= (β − σ)g − (β − γ)v + λ. (23)
θ2

θ̇2 θ1 (1 − β)Y
(16) ⇒ =ρ− − δ(1 − u) = ρ − δ. (24)
θ2 θ2 h
(23) and (24) ⇒ (β − σ)g − (β − γ)v + λ = ρ − δ. (25)

Solving (22) and (25) gives

(λ + δ − ρ)(1 − β + γ)
g=
σ(1 − β + γ) − γ

(λ + δ − ρ)(1 − β)
v=
σ(1 − β + γ) − γ

64
• Social planner’s problem: Replacing (18) by

θ1(1 − β + γ)Y /h + θ2δ(1 − u) = ρθ2 − θ̇2

and following exactly the same solution procedure as that for the
decentralized economy, we have
   
1−β 1 − β + γ ∗
v ∗ = δ − (ρ − λ) /σ and g ∗ =  v .
1−β+γ 1−β
Note that both v and v ∗ should not exceed δ, leading to the
following restriction on the parameters
  
1 − β  ρ − λ
σ ≥1− .
1−β+γ δ
• Decentralized equilibrium is not Pareto optimal if γ > 0 : v∗ >
v. Note that if γ = 0, then v = v ∗ = g = g∗ = (λ + δ − ρ)/σ.
• The model’s ability to fit the US data? Progress has been made
in explaining cross-country differences in income levels, but more
things need to be done to account for differences in growth rates.

65
5 Learning-By-Doing and Comparative Advantage

This model emphasizes the importance of on-the-job-training in the

formation of human capital.
• Assume that there are two consumption goods, c1 and c2 and hu-
man capital is the only input in production (no physical capital).
Consumption good is is produced according to

ci = hiuiN, i = 1, 2, (26)

where hi is human capital specialized to the production of good

i and u1(respectively, 1 − u1) is the fraction of the labor force
used in producing good 1 (respectively, good 2).
• Specialized human capital hi is accumulated according to

ḣi = δiuihi, i = 1, 2, (27)

where it is assumed that good 1 is a high-technology good (i.e.,

δ1 > δ2) and that the effects of hi are entirely external in the
sense that production and human capital accumulation for each
good depend on the average human capital level in that industry.
• Since there is no physical capital and human capital is external,
the representative consumer’s optimization problem is a static
problem. Assume that the consumer’s utility function is given
by

U (c1, c2) = [α1c−ρ −ρ −1/ρ

1 + α2 c2 ] , (28)

where αi > 0, α1 + α2 = 1, ρ > −1 and σ ≡ 1/(1 + ρ) is the

elasticity of substitution between the two goods.

66
 5.1 Autarky Equilibrium

• Let (1, q) be the equilibrium prices of (c1, c2). The price of c2

must equal to the marginal of substitution in consumption, i.e.,
 
U2 α2  c2 −(1+ρ)
q= = , (29)
U1 α1 c1
, which gives
 
c2  α2 σ −σ
= q . (30)
c1 α1
Profit maximization and (26) imply
c2 u2h2
= . (31)
c1 u1h1
Combining (30) and (31), along with u1 + u2 = 1, yields
     
1 − u1  α2 σ  h2 σ−1  α2 σ 1−σ
= = q , (32)
u1 α1 h1 α1
leading to
  σ 
−1
α 2 1−σ
u1 = 1 +   q  − δ1.
α1
Then from q = h1/h2, we have
 σ  −1  
q̇ ḣ1 ḣ2 α 2 1−σ
= − = (δ1 + δ2) 1 +   q  − δ2. (33)
q h1 h2 α1
There are three cases: σ > 1, σ = 1 and σ < 1.
• Suppose that the two goods are good substitutes (i.e., σ > 1),
· µ ¶ ¸
α2 σ 1−σ −1
then 1 + α1 q is an increasing function of q. Letting q ∗
67
 be the solution to q̇ = 0, we have: (i) If q(0) > q ∗, then q̇ > 0,
that the economy converges to specialization in c1; (ii) if q(0) <
q ∗, then q̇ < 0, that the economy converges to specialization in
c2; (iii) if q(0) = q ∗, then q̇ = 0, that the economy dose not
converges to specialization.
• The decentralized equilibrium is not efficient because learning-
by-doing effects are external.

5.2 Free-Trade Equilibrium

• All countries are assumed to be small, prices in all countries will

equal world prices (1, p) and each country will take p as given.
• All countries specialize in producing one good depending on their
endowments (h1, h2). Those with h1/h2 > p (condition 1) pro-
duce only good 1 (so their h2 endowments are fixed) while those
with h1/h2 < p (condition 2) produce only good 2 (so their h1
endowments remains constant). As a result, the world supply of
good i is given by
X
ci = hi,
condition i

which gives
P
c2 h2
= Pcondition 2 .
c1 condition 1 h1
• Since all counties have identical homothetic preferences, world
relative demand remains the same as in the autarky case, i.e.,
 
c2  α2 σ −σ
= p
c1 α1
68
• Dynamics of p: From
  P
c2  α1 σ −σ h2
= p = Pcondition 2 ,
c1 α1 condition 1 h1
we have
ṗ δ1 − δ2
= > 0, (34)
p σ
if producers do not switch from one good to the other.
• Possibility of switching production: (i) Good 2 producers do
not switch because if h1(0)/h2(0) < p(0), then h1(t)/h2(t) <
p(t), ∀t > 0; (ii) good 1 producers switch if h1(0)/h2(0) > p(0)
and h1(t)/h2(t) < p(t), ∀t > t∗. This can occur only if
ṗ/p > ḣ1/h1 which is equivalent to σ < 1 − δ2/δ1.
• Growth rates real output: Assume that σ ≥ 1 − δ2/δ1, then (34)
holds true. As a result, good 1 producers’s growth rate is

g1 = δ1

and good 2 producers’ growth rate is

δ1 − δ2
g2 = δ2 + ṗ/p = δ2 + .
σ
We can easily see that g1 > g2 if σ > 1. That is, if the two goods
are good substitutes, then countries producing the high-learning
good (good 1) have a higher growth rate.
• The model generates endogenous growth. Different countries
have different growth rates.
• Important factors the model does not capture: composition of
demand and introduction of new goods.
69
• Strategies for economic development: import substitution (de-
pending on initial comparative advantage) and export promotion
(through taxes and subsidies).

70