The Solow Model

Prof. Lutz Hendricks

September 17, 2009

L. Hendricks () Solow Model September 17, 2009 1 / 51

The Solow Model: Motivation

In the production model capital is exogenous.

We learn how much capital matters, but we cannot learn why some
countries lack capital.
We need a model with capital accumulation (investment, saving).
That also answers the question: Does capital drive long-run growth?

L. Hendricks () Solow Model September 17, 2009 2 / 51

The Model

We add just one piece to the production model:

an equation that describes how capital is accumulated over time
through saving.
Time is discrete: t = 1, 2, 3, ...
Production function (cf Production model):

Yt = F (Kt , Lt ) = ĀKtα L1t α

(1)

Resource constraint (cf National Income Accounting):

Yt = Ct + It + Gt + NXt (2)

L. Hendricks () Solow Model September 17, 2009 3 / 51

The Model

The new piece: Capital accumulation obeys

Kt +1 = Kt + It dKt (3)

Or

∆Kt = Kt + 1 Kt (4)
= I dKt (5)

d is the depreciation rate (7% to 10% in the data).

L. Hendricks () Solow Model September 17, 2009 4 / 51

A capital accumulation example

Capital accumulation runs out of steam - with constant It the growth of K

slows down.

L. Hendricks () Solow Model September 17, 2009 5 / 51

Choices

People make two fundamental choices (in macro!):

1 How much to save / consume.
2 How much to work.
Work: we assume L is …xed.
Consumption / saving: We assume that people save a …xed fraction
of income:
Ct = (1 s ) Yt (6)

Equivalently:
It = sYt (7)

L. Hendricks () Solow Model September 17, 2009 6 / 51

Model Summary

The rental prices of K and L are the same as in the production model (w
and r ).

L. Hendricks () Solow Model September 17, 2009 7 / 51

Solving the model

Even this simple model cannot be "solved" algebraically.

That is, we cannot write the endogenous variables as functions of the
parameters.
This is almost never possible in dynamic models.
Dynamic means: there are many time periods.
All interesting macro models are dynamic.
What we can do is
1 graph the model and trace out qualitatively what happens over time.
2 solve the model for the long-run values of the endogenous variables
(e.g. Kt as t ! ∞).

L. Hendricks () Solow Model September 17, 2009 8 / 51

The Solow Diagram

We condense the model into a single equation in K .

It will be a dynamic equation that tells us Kt +1 as a function of Kt .
Then we graph the equation.

L. Hendricks () Solow Model September 17, 2009 9 / 51

The Solow Diagram

Start from
Kt +1 = It + (1 d ) Kt (8)

Substitute in

It = sYt
Yt = ĀKtα L1t α

to get
Kt +1 = s ĀKtα L1t α
+ (1 d ) Kt (9)
or
∆Kt = s ĀKtα L1t α
dKt (10)

L. Hendricks () Solow Model September 17, 2009 10 / 51

The Solow Diagram

L. Hendricks () Solow Model September 17, 2009 11 / 51

Steady state

De…nition
A steady state is an equilibrium in which all (per capita) variables are
constant over time.
Figure 5.1 shows:
1 The Solow model has exactly one steady state K
Actually: K = 0 is another steady state, which we ignore.
2 The economy converges to the steady state for any initial K0 .
3 For the economy to grow, it must be on the transition path to the
steady state.

L. Hendricks () Solow Model September 17, 2009 12 / 51

Steady state

Why does the economy settle in a steady state instead of growing

forever?
The reason is diminishing returns to capital in production.
Recall ∆K = sY dK .
Save a constant amount each period (constant I = sY ).
It contributes less and less to output (falling marginal product of K -
falling ∆Y ).
It adds the same amount to depreciation.

L. Hendricks () Solow Model September 17, 2009 13 / 51

Steady state

Why does the economy settle in a steady state instead of growing

forever?
The reason is diminishing returns to capital in production.
Recall ∆K = sY dK .
Save a constant amount each period (constant I = sY ).
It contributes less and less to output (falling marginal product of K -
falling ∆Y ).
It adds the same amount to depreciation.
Note: this is not quite right!
With constant s we have rising I over time.
And if ∆K falls over time, so does the additional depreciation.

L. Hendricks () Solow Model September 17, 2009 13 / 51

The dynamics of output and consumption

L. Hendricks () Solow Model September 17, 2009 14 / 51

Solving for the steady state

Set ∆Kt = 0 to obtain

sY = s ĀK α L1 α
= dK (11)

or
1/(1 α)
k = K /L = [s Ā/d ] (12)
with 1/ (1 α) = 3/2.

L. Hendricks () Solow Model September 17, 2009 15 / 51

Solving for the steady state

Steady state output:

y = Y /L (13)
= Ā (K /L)α (14)
α/(1 α)
= Ā [s Ā/d ] (15)
α/(1 α) 1/(1 α)
= [s/d ] Ā (16)

L. Hendricks () Solow Model September 17, 2009 16 / 51

Reality check

A key prediction of the model: sY = dK or K /Y = s/d.

Countries with higher saving rates have higher capital output ratios.

L. Hendricks () Solow Model September 17, 2009 17 / 51

Why does Y/L di¤er across countries?

Our static production model answered: K /L and Ā di¤er:

y = Ā k α (17)

The Solow model gives a similar answer: s and Ā di¤er:

y = Ā1/(1 α)
[s/d ]α/(1 α)
(18)

This is the same answer in disguise: higher s means higher K /L.

L. Hendricks () Solow Model September 17, 2009 18 / 51

Why does Y/L di¤er across countries?

But the answer is quantitatively di¤erent:

Production model: Double Ā ! double y .
Solow model: Double Ā ! raise y by 21 /(1 α) = 23 /2 = 2.8.
In the Solow model, the contribution of Ā to output gaps is larger -
why?
Draw a picture...

L. Hendricks () Solow Model September 17, 2009 19 / 51

How important is K/L for cross-country Y/L gaps?

Our previous answer was: K /L accounts for a factor near 4.

In the Solow model:
3/2 1/2
yUS ĀUS sUS
= (19)
ypoor Āpoor spoor
32 = 16 2 (20)

Why factor 2 for saving rates?

Because s (or K /Y ) di¤ers by a factor near 4.

L. Hendricks () Solow Model September 17, 2009 20 / 51

How important is K/L for cross-country Y/L gaps?

This is a central and robust result:

Capital accumulation accounts for only a small fraction of
cross-country income gaps.

L. Hendricks () Solow Model September 17, 2009 21 / 51

Long-run Growth

There is no long-run growth in the Solow model.

The reason is diminishing returns to capital.
This is perhaps the central lesson from the model:

Fact
Capital accumulation is not a source of long-run economic growth.

How does this square with the fact that politicians always talk about
promoting growth through investment?

L. Hendricks () Solow Model September 17, 2009 22 / 51

Shocking the model

We want to understand how the economy responds to changes in

certain parameters.
This is useful for understanding short-term growth experiences of
countries.
We will come back to that when we talk about business cycles, too.

L. Hendricks () Solow Model September 17, 2009 23 / 51

Higher investment rate

L. Hendricks () Solow Model September 17, 2009 24 / 51

Higher investment rate

L. Hendricks () Solow Model September 17, 2009 25 / 51

Higher investment rate

L. Hendricks () Solow Model September 17, 2009 26 / 51

Higher investment rate
Reality check

The model says: more investment (or lower consumption) generates

a period of faster growth.
Isn’t everybody saying: the U.S. is in a recession (slow growth)
because consumption is too low?
How does the contradiction get resolved?
Where is the e¤ect of lower consumption demand in the Solow model?
Where is the demand side anyway?

L. Hendricks () Solow Model September 17, 2009 27 / 51

Higher depreciation rate

L. Hendricks () Solow Model September 17, 2009 28 / 51

Higher depreciation rate

L. Hendricks () Solow Model September 17, 2009 29 / 51

Higher depreciation rate

L. Hendricks () Solow Model September 17, 2009 30 / 51

The Principle of Transition Dynamics

Fact
In the Solow model, the farther away the economy is from its steady state,
the faster it grows (or shrinks)

Why is this true?

Look at the examples of rising s or rising d - the growth rate of Y
rises / falls over time.
But we should derive this algebraically ...

L. Hendricks () Solow Model September 17, 2009 31 / 51

The Principle of Transition Dynamics
Proof

The claim: the growth rate of K (∆Kt /Kt ) increases with the gap
between current and steady state capital (K /Kt ).
Proof:
The law of motion for K is
∆Kt Yt
=s d (21)
Kt Kt

Sub in the production function:

∆Kt 1 1 α
= s ĀK α L d (22)
Kt

L. Hendricks () Solow Model September 17, 2009 32 / 51

The Principle of Transition Dynamics
Proof

Sub in K = (K /K ) K :

∆Kt 1 1
= s Ā (K /K )α (K ) α L1 α
d (23)
Kt

Everything on the RHS (incl. K ) just depends on parameters, except

for K /K .
Plot that ...

L. Hendricks () Solow Model September 17, 2009 33 / 51

Empirical Evidence

Countries that were poorer in 1960 grew faster after 1960.

We could do the same with K /L instead of Y /L on the horizontal axis.

L. Hendricks () Solow Model September 17, 2009 34 / 51

Empirical Evidence

Should we conclude that transitional growth explains cross-country

di¤erences in output growth?
No!
Figure 5.8 only shows OECD countries - mostly rich Western
European countries + North America.

L. Hendricks () Solow Model September 17, 2009 35 / 51

Empirical Evidence

In a broader set of countries, those that are poor initially do not grow
faster.
L. Hendricks () Solow Model September 17, 2009 36 / 51
Empirical Evidence

But …gure 5.9 is the wrong experiment!

The Solow model does not say: "poor countries grow faster"
It says: "countries that are poor relative to their steady states grow
faster."
That is true in the data.
Figure 5.9 tells us: much of the growth since 1960 is not of the
transitional dynamics type - not explained by the Solow model.

L. Hendricks () Solow Model September 17, 2009 37 / 51

Convergence Clubs?

One hypothesis one often reads:

Countries that share similar "fundamentals" (e.g. OECD countries)
converge to similar steady states.
For them the principle of transition dynamics explains post-war growth.
Other countries have di¤erent fundamentals, so the poorer ones need
not grow faster.
This is the wrong explanation for post-war growth.

L. Hendricks () Solow Model September 17, 2009 38 / 51

Convergence Clubs?

The Solow model makes a quantitative prediction about growth rates.

If you work out the numbers, countries should converge fairly quickly
to their steady states (perhaps within 20 years).
Then they all should grow at almost the same rates.

Fact
The Solow model cannot explain why countries grow at di¤erent rates for
long periods of time.

Post-war growth may look like Solow convergence to a common

steady state, but it is not.
Post-war growth is driven by growth in TFP, not by growth in K /L.

L. Hendricks () Solow Model September 17, 2009 39 / 51

Exercise
Take a spreadsheet.
Fix parameters at plausible empirical values: α = 1/3, d = 0.08,
s = 0.2.
Compute the steady state.
Fix K0 at some multiple of K .
Compute the transition path for Kt by iterating over
Kt +1 = sYt + (1 d ) Kt .
Plot the growth rate of Yt against time.
You should see that growth is very high initially, if K0 is small. But
growth slows dramatically very quickly.
Now plot the growth rate of Yt against over 40 years against initial
Yt - this is the model analogue of …gure 5.8.
You should see that the model relationship is much ‡atter than the
data relationship.
The model fails to explain large variation in 40 year average growth
rates.
L. Hendricks () Solow Model September 17, 2009 40 / 51
Simulating the Solow Model

Example: Start an economy from 1/10th of steady state k .

0.11

0.1

0.09

0.08

0.07
Y/ L growth

0.06

0.05

0.04

0.03

0.02

0.01
0 20 40 60 80 100 120 140
Year

L. Hendricks () Solow Model September 17, 2009 41 / 51

Simulating the Solow Model

Growth varies less with y /y than in the data.

0.04

0.035
Avg growth over 35 years

0.03

0.025

0.02

0.015
0.4 0.5 0.6 0.7 0.8 0.9 1
Y/L relative to steady state

This suggests that part of growth is not transitional in the data.

L. Hendricks () Solow Model September 17, 2009 42 / 51

Simulating the Solow Model
Speed of convergence

0.9

0.8

0.7

0.6

0.5

0.4
K/KSS
0.3 Y/YSS

0.2

0.1
0 20 40 60 80 100 120 140
Year

Convergence is too fast.

In the data, the "half-life" is about 30 years – 10 years in the model.
Convergence is even faster when the saving rate is endogenous.

L. Hendricks () Solow Model September 17, 2009 43 / 51

Did we just invalidate the Principle of Transition
Dynamics?

No, we did not.

Countries grows faster when their capital stocks are low.
But this does not account for the observed di¤erences in long-run (40
year) growth rates across countries.
It does account for growth rates at business cycle frequencies (we
come back to that later).

L. Hendricks () Solow Model September 17, 2009 44 / 51

The Tigers

There are a few countries that sustained growth by capital

accumulation for a long period of time.
How?
It cannot work with a constant saving rate s - the Solow model shows
this.
Such countries must have saving rates that rise over time.
Examples are: South Korea, Singapore, Hong-Kong.

L. Hendricks () Solow Model September 17, 2009 45 / 51

The Tigers

L. Hendricks () Solow Model September 17, 2009 46 / 51

Exercise: Rising saving rate

Simulate the Solow model with a saving rate that rises from 10% to
40% (Singapore).
Start the model in steady state: K0 = K .
Show that the growth rate of y stays positive for a long time.
You could now compare that growth path with data for Singapore
and convice yourself that a large share of Singapore’s spectacular
growth since 1960 is indeed due to capital accumulation (as shown by
Alwyn Young).

L. Hendricks () Solow Model September 17, 2009 47 / 51

Exercise: The Saving Rate Depends on Income

Consider an alternative version of the Solow model.

The saving rate depends on income.
What happens?

L. Hendricks () Solow Model September 17, 2009 48 / 51

Conclusion: Is the Solow Model Useful?

As a model of growth or large cross-country income di¤erences, the

model is a failure.
But its failure contains important insights:
1 Capital does not drive growth.
2 Capital does not drive large fractions of cross-country income gaps.
Both …ndings are surprising - and often not understood in the policy
debate.

L. Hendricks () Solow Model September 17, 2009 49 / 51

Conclusion: Is the Solow Model Useful?

But the main signi…cance of the Solow model itself is as a building

block for macro models.
We always have to keep track of how capital is accumulated.
A Solow block is therefore part of virtually every model.
The same logic extends to other accumulated factors: human capital,
knowledge capital, organization capital.
The Solow transition dynamics is an important piece for
understanding business cycle dynamics.

L. Hendricks () Solow Model September 17, 2009 50 / 51

Reading

Jones, Macroeconomics, ch. 5.

L. Hendricks () Solow Model September 17, 2009 51 / 51