Lecture Notes 1: Solow Growth Model

Zhiwei Xu ([email protected])

Solow model (Solow, 1959) is the starting point of most of the dynamic macroeconomic theories.
It introduces dynamics and transitions into the aggregate economy. We first list several key
assumptions regarding the model’s setup. To be consistent with future lectures, we assume that
the time is discrete.

1 Model Setup
1. Production Technology
The firm produces final goods by combining capital Kt and labor Lt . The production function
is assumed to take the form
Yt = Ktα (At Lt )1−α , (1)

where At is the neutral technology process, α ∈ (0, 1) is the capital share. The production
2 2
function is concave in K and L: ∂∂KY2t < 0 and ∂∂LY2t < 0. This assumption implies that the
t t
marginal product of capital (MPK) is decreasing in K and the marginal product of labor
(MPL) is decreasing in L. The concavity of production function ensures the uniqueness of
balance growth path and the global convergency.

2. Labor Supply
The labor input Lt equals the population (full employment) and grows at a constant rate
n > 0, that is
Lt = (1 + n) Lt−1 . (2)

3. Technology Progress
The neutral technology At grows at a constant rate g > 0 :

At = (1 + g) At−1 . (3)

Note that Solow model introduces exogenous growth by assuming the population and the
technology are growing over time.

1
 4. Consumption and Saving
Households are representative. The saving, St , and consumption, Ct , decisions are static:

St = sYt , (4)

Ct = (1 − s) Yt , (5)

where s is the constant saving rate. Note that the above decisions may not necessarily
optimal for the household. Hence, the Solow model does not consider the micro-level opti-
mization decisions. This assumption will be relaxed in the Ramsey model in future lectures.

5. Capital Accumulation
To make the model be dynamic, we need to introduce the dynamic process of capital accu-
mulation:
Kt+1 = (1 − δ) Kt + It , (6)

where Kt is the beginning-of-period capital stock. That is, Kt is predetermined at period

t − 1, therefore Kt is a state variable.

6. Capital Market Clearing

It = St . (7)

In the closed economy, aggregate investment is always equal to aggregate savings. However,
in the open economy, two variables may not necessarily be the same. The difference between
saving and investment is the current account balance.

2 Dynamics
2.1 Dynamic System
The full system is given by equations (1) to (7). From (4) to (7), we can simplify the system into
one difference equation
Kt+1 = (1 − δ) Kt + s (Kt )α (At Lt )1−α . (8)

To remove the non-stationary trend in the last equation, we need to detrend each variable. Divide
the both sides of (8) by At Lt , and define new variable xt ≡ AXt Lt t . We can rewrite (8) as

kt+1 (1 + g) (1 + n) = (1 − δ) kt + sktα . (9)

2
Rearranging the terms yields

(1 + g) (1 + n) (kt+1 − kt ) = sf (kt ) − (δ + g + n + g × n) kt . (10)

where f (k) = k α . For simplicity, we ignore the terms with the coefficient g × n. We then obtain

(1 + g + n) (kt+1 − kt ) = sf (kt ) − (δ + g + n) kt . (11)

 
The above difference equation determines the process of the effective capital stock kt or AKt Lt t .
Before we discuss the dynamics implied by (11), we first look at the stationary growth in the
long run, where ∆kt = 0 or kt+1 = kt .

2.2 Balanced growth path (BGP):

BGP is the path on which all the variables of the model grow at a constant rate n + g, and the
detrended variables (e.g. kt ) stay constant. On the balance growth path, kt is constant: kt = k ∗ ,
for all t. It can be shown that under the Cobb-Douglas technology, there exists a unique k ∗ to
solve the equation
s (k ∗ )α − (δ + g + n) k ∗ = 0, (12)

or  1
 1−α
∗ s
k = . (13)
δ+g+n
On the balance growth path, the capital stock, Kt , is given by

KtB = k ∗ At Lt . (14)

Plugging KtB into (1) and (5), balance growth output, consumption and saving, are given by

YtB = (k ∗ )α At Lt , (15)
StB = s (k ∗ )α At Lt , (16)
∗ α
CtB = (1 − s) (k ) At Lt . (17)

The BGP indicates that long-run growth does not depend on the initial state and capital accumu-
lation. The long-run growth rate is only determined by the growth of technology g and population
n.

3
 Figure 1: Phase diagram for the dynamics of kt

2.3 Dynamics of kt and yt

The concavity of production technology ensures the global convergency of kt . To proof this,
suppose kt < k ∗ , the right hand side of (11) implies kt+1 − kt > 0. Thus kt will monotonically
increase until kt = k ∗ . For the case of kt > k ∗ , similarly we have kt+1 < kt , and kt will monotonically
decrease until kt = k ∗ . Therefore, for any initial value of kt , it will eventually converge to the
steady-state k ∗ . The phase diagram below illustrates the dynamics of kt .
To exactly see how the capital kt transits from the initial state to the steady-state k ∗ , we do
the following simulation. First, we set the values of deep parameters as

s = 0.4, n = 2%, g = 5%, α = 0.5, δ = 0.1.

We consider two scenarios: (1) k0 = 0.8k ∗ ; (2) k0 = 1.2k ∗ . Figure 2 shows that capital kt
monotonically converges to its steady state k ∗ .
Suppose that poor country and rich country have the same economic structure (same parameter
values), Figure 2 shows that the poor country (low initial value) grows fast at the beginning,

4
 Figure 2: Transition paths of capital for different initial values (k0 = 0.8k ∗ , 1.2k ∗ )

5.8

5.6

5.4

5.2
t
k

4.8

4.6

4.4
0 20 40 60 80 100 120
*
k0=0.8k

6.5
t

6
k

5.5

5
0 20 40 60 80 100 120
*
k0=1.2k

Notes: This figure plots the transition dynamics for the capital stock kt . The vertical axis is the level of capital
kt . The horizontal axis is the time period. The dashed line indicates the steady-state level of capital k ∗ .

5
 Figure 3: Transition path of output (k0 = 0.8k ∗ )

2.5

2.4

2.3

2.2

2.1
20 40 60 80 100 120
Output yt

0.01

0.008

0.006

0.004

0.002

0
20 40 60 80 100 120
Growth of yt

Notes: This figure plots the transition dynamics for the output yt when the initial capital is k0 = 0.8k ∗ . The
vertical axis is the output level yt or growth rate ∆ log(yt ). The horizontal axis is the time period. The dashed
line indicates the steady state.

6
afterward the growth rate declines. Eventually, the poor country catches up with the rich country
and converges to a unique steady state. The aggregate output (non-detrended) Yt will be on the
balanced growth path.

2.4 The effect of saving rate (s) on the long-run growth

To see how the saving rate affects the steady state k ∗ , we take logs of both sides of (13) and take
derivative w.r.t s,
∂ log k ∗ 1
= . (18)
∂ log s 1−α
For the steady-state output, we have

∂ log y ∗ α
= . (19)
∂ log s 1−α

A higher capital share implies a larger effect of saving rate on output. Since the saving rate has
a positive effect on capital k ∗ , a rise in capital intensity in the economy implies that raising the
saving rate will increase the output more.
As the consumption is given by c∗ = (1 − s) (k ∗ )α , we further have

∂ log c∗ s α
=− + . (20)
∂ log s 1−s 1−α
∗
Golden Rule: the consumption c∗ achieves the optimal level, which implies that ∂∂log c
log s
= 0 or
s = α.
To give a numerical example, Figure 4 computes the dynamic path for the transition from the
old steady state (s = 0.4) to the new steady state with high saving rate (s = 0.5).

2.5 Short-run dynamics

The above discussions are related to the long-run dynamics. I.e., we study the dynamic path of
the economy transits from the initial state to the steady state. Now, we discuss the short-run
dynamics around the steady state. We first linearize the equation (11) around the steady state
k∗ :

∗ ∗ αs (k ∗ )α−1 − (δ + g + n)
(kt+1 − k ) − (kt − k ) = (kt − k ∗ )
1+g+n
(1 − α) (δ + g + n)
= − (kt − k ∗ ) . (21)
1+g+n

7
 Figure 4: Transition path to the steady state of high saving rate

9
8
7
6
5
20 40 60 80 100 120
Capital kt

2.5

2
20 40 60 80 100 120
Output yt

0.02

0.01

−0.01
20 40 60 80 100 120
Growth of y
t

1.6

1.4

1.2

1
20 40 60 80 100 120
Consumption c
t

Notes: This figure plots the transition dynamics for the economy when the saving rate increases from 0.4 to 0.5.
The vertical axis is for the real variables. The horizontal axis is the time period. The dashed line indicates the new
steady state.

8
The second line is due to the relationship s (k ∗ )α = (δ + g + n) k ∗ . Define the convergent speed λ
as (1−α)(δ+g+n)
1+g+n
. Equation (21) implies

kt − k ∗ = (k0 − k ∗ ) (1 − λ)t . (22)

Last equation describes the evolution of capital around the steady state. Regarding the output,
similarly we have
yt − y ∗ = (y0 − y ∗ ) (1 − λ)t . (23)

Given our previous calibration, λ is equal to 0.08, implying that it takes approximate 8 years
ln 0.5
(' ln(1−λ) ) to get the halfway of the balance-growth-path value.

2.6 Further Discussion

The Solow model is a good framework to understand economic growth. One natural question is:
can we directly apply the Solow model for understanding the Chinese Economy? To answer this
question, we need to think more deeply about the crucial assumptions made in the model. One
important prediction it makes is that a high saving rate may induce a high level of income in the
steady-state because one country can accumulate a high level of capital, see equation (13). This
argument can be applied to advanced countries such as the United States, but may not be true
for developing countries like China. The key assumption in the model is that there are no market
frictions. In particular, the model assumes a perfect financial market, which is reflected by two
features in the model: (i) the total savings in the household side can be sufficiently channeled to
firm’s investment, i.e., St = It ; (ii) firms invest all of their capital to production purpose, i.e., It
fully contributes to the production capital without the efficiency loss. However, if one country
has less developed financial market, i.e., the total savings cannot be efficiently allocated to the
production sector, i.e., the feature (i) fails, and the non-financial firms do not invest in production
capital, i.e., the feature (ii) fails, then we cannot conclude that a high saving rate is good for one
country’s economic growth.
For the Chinese economy, the financial market is underdeveloped. The savings cannot be
efficiently allocated to the domestic production side (credit misallocation), resulting in a severe
credit shortage for small and private business, and a large imbalance in the financial account and
current account (huge foreign reserve and trade surplus).
Figure 5 plots the saving rate (defined as the gross domestic saving as percentage of GDP)
and the investment rate (defined as the gross domestic capital formation as percentage of GDP).
It shows that the saving rate in China is constantly above the investment rate. The gap of these
two series reflects the capital outflow (lending to foreign countries) or current account surplus.
Also the non-financial firms invest a lot in the nonproduction (or financial) assets, for instance,

9
 Figure 5: Saving and investment rate in China

52 10
Invest. rate Net saving
50 Saving rate Current account
48

46
5
44

42

40
0
38

36

34

32 −5
1980 1990 2000 2010 2020 1980 1990 2000 2010

real estate assets. These investment behaviors may crowd out resources allocated to production
sector and dampen the GDP growth. Figure 6 shows that an upward trend in the share of financial
assets of the firm’s total assets (the solid line) is associated with a housing boom (the dashed line).
The figure reveals that the strong demand for investing in real estate markets from the firm side
may largely contribute to the surge in house prices. The data also imply that a higher return on
real estate investment may boost the firm’s financial investment and lead to an expansion in the
housing sector. The correlation between the share of financial assets and real GDP is -0.43, and
the correlation between the share of financial assets and real house prices is 0.41.

3 Extensions
3.1 Natural Resources and Land
We extend the benchmark Solow model by augmenting production function with two additional
inputs: natural resources Rt and land Tt . Specifically, we assume that the production function is

Yt = Ktα Rtβ Ttγ (At Lt )1−α−β−γ , (24)

where α > 0, β > 0, γ > 0, α + β + γ < 1. The process of natural resource Rt is assumed to follow

Rt = (1 − b) Rt−1 , (25)

10
 Figure 6: Share of financial assets in firm side and aggregate economy in China

level level
25 2.5
24

8.5
22
2

20 20 8.4

1.5
18
8.3

16
15 1 8.2
2011 2012 2013 2014 2015 2016 2011 2012 2013 2014 2015 2016

HP filtered HP filtered
2 1 2 0.1

1 0.05

0 0 0 0

−1 −0.05

−2 −1 −2 −0.1
2011 2012 2013 2014 2015 2016 2011 2012 2013 2014 2015 2016

Share of fin. asset Share of fin. asset

Real GDP House prices

and the land is assumed to be constant, i.e., ∆Tt = 0. On the balance growth path, it can be
shown that capital and output have the same growth rate. This is because (8) implies

Kt+1 Yt
= (1 − δ) + s . (26)
Kt Kt

On BGP, Yt /Kt is constant or the growth rates satisfy gY = gK .

To compute the growth rate of output on BGP, we take logs of both sides of (24) and get

gY = αgY − βb + (1 − α − β − γ) (g + n) , (27)

or
−β 1−α−β−γ
gY = b+ (g + n) . (28)
1−α 1−α

11
Define g̃Y = g + n the growth rate where capital and labor are the only inputs. Since 1−α−β−γ
1−α
<1
β
and − 1−α b < 0, the growth rate of output in the model with natural resource gY is less than that
in baseline model g̃Y . That is, the natural resources (limited supply) reduce the BGP growth rate.
The gap between gY and g̃Y is called the “growth drag”.

3.2 Growth Trap: Production with Fixed Cost

So far there is no friction in the Solow model. We now introduce one friction—fixed production
cost. You may see that with this minor extension, the dynamics of the model will change a lot.
Suppose that each period, in order to make production the producer has to pay a fixed amount
of cost ξ. In reality, this cost may be due to the rental rate of land, loan payment, etc. The
production function for yt (Yt /At /Lt ) takes the form of

yt = max {ktα − ξ, 0} . (29)

The difference equation (11) now becomes

(1 + g + n) (kt+1 − kt ) = s max {ktα − ξ, 0} − (δ + g + n) kt . (30)

It can be shown that there exist two k such that

s max {k α − ξ, 0} = (δ + g + n) k. (31)

Denote them as k ∗ and k ∗∗ , where k ∗ < k ∗∗ . As shown in Figure 7, k ∗∗ is stable in the sense
that for any capital kt > k ∗ , kt would eventually converge to k ∗∗ . While k ∗ is not stable because
for any kt < k ∗ , kt would converge to the zero point.

3.3 Application: Middle Income Trap

The middle income trap is an economic development situation, where a country which attains a
certain income (due to given advantages) will get stuck at that level. Solow model with fixed
production cost may capture the main idea. Consider two dynamic systems

(1 + g + n) (kt+1 − kt ) = sz l ktα − (δ + g + n) kt , (32)

(1 + g + n) (kt+1 − kt ) = s max z h ktα − ξ, 0 − (δ + g + n) kt ,

(33)

where z l < z h . The economy starts with a low level K will trap at the K mid equilibrium. To pull
out the economy from middle income trap, the government need to either reduce the friction ξ or
raise the aggregate capital K. Figure 8 provides a graphical illustration.

12
 Figure 7: Growth Trap

Figure 8: Middle Income Trap

13