Norton Media Library

Chapter 1
Introduction:
The Facts of
Economic
Growth
Charles I. Jones
• There is tremendous variation in
performance across countries of the world

• Some useful data for comparison purposes:

– GDP per capita = GDP/Population

– GDP per worker = GDP/Number of Employed

Sometimes economists will also use:

Output per hour = Output/Total Hours Worked

– Labor Force Participation Rate =

(Employed + Unemployed)/Population
Annual Average Growth Rate, g, for Y from
1960 to 1997 comes from:
( Y1997/Y1960 ) = (1+g)1997-1960

Thus, g = ( Y1997/Y1960 )(1/37) - 1

Or: g = exp{(1/37)( ln(Y1997) - ln(Y1960 ) )} - 1

Rule of 70:
If Y grows at g percent per year, then:
The level of Y Doubles every 70/g years
• t
Table 1.1
• The world income distribution is highly
skewed
– About 2/3 of the world population has at most
20% of the GDP per worker of the US
– About 1/10 of the world population at least
80% of the GDP per worker of the US
• The world income distribution is changing over
time
– The bulk of the world’s population is substantially
richer today than it was in 1960.
– The fraction of people living in poverty has fallen
since 1960.
– A major reason for these changes is the recent
economic growth in China and India
• which together account for 40 percent of the world
population.
• In addition to the level of income rising
over time, the growth rate of world income
has also been rising over time

– Note the ratio or logarithmic scale in the

following graph
• In contrast to world income growth rates
that are accelerating, the growth rate of
income for the US has not been growing at
a faster and faster rate over time
– But the US growth rate has tended to be
positive over time and has been roughly
constant over the long run
Fig. 1.4
Growth Theory
Norton Media Library

Chapter 2
The Solow Model

Charles I. Jones
The Production Function
• A production function describes the way we can
use inputs to produce outputs
– technically speaking, a production function tells us the
maximum amount of output we can produce for every
combination of inputs

• An equation for the production function

Y=AF(K,L)
where :
F is some function of A, K and L,
K=stock of capital,
L=labor
and A is productivity, which accounts for all
other factors that may be involved
• For a production function
Y=AF(K,L)
we typically assume:
– Positive marginal products
• when A rises Y rises
• when K rises Y rises
• when N rises Y rises
– Diminishing marginal products for K and N
• As K gets larger, an additional unit of K causes Y to
increase by a smaller amount
• As N gets larger, an additional unit of N causes Y to
increase by a smaller amount
– But as A rises, Y rises proportionally
• no diminishing marginal product here
• These conditions on marginal products
translate directly into the derivatives on the
production function
– Positive marginal products
• First derivative of F with respect to K is positive
• First derivative of F with respect to L is positive
– Diminishing marginal products
• Second derivative of F with respect to K is negative
• Second derivative of F with respect to L is negative
– It is easy to show that any percentage change in
A has the same percentage effect on Y in this
production function
• The elasticity of Y with respect to A is 1
• We like to compare countries in terms of
output per capita, so divide both sides of the
equation by L
– Assume L is population, not number of
workers, and for now don’t worry about the
distinction
• (Y/L) = (1/L)•F(K,L) = F(K,L)/L
• A common assumption in economics is that
the production function has constant returns
to scale (CRS)
– That means if we double the inputs we will
double the amount of output that we can
produce
– CRS has a certain intuitive appeal
• If we own one factory and build a second factory
that has precisely the same facilities and we employ
identical workers in the second factory, we should
be able to produce twice as much output
• More generally, CRS means that a given
percentage change in every input will change the
amount of output we can produce by precisely that
same percentage
• Assume F(K,L) is a CRS production function
– Suppose z is some percentage change in the factors
– CRS means that if Y=F(K,L) for particular values of K,
L and Y, then zY=F(zK,zL)
• This makes it easy to turn our production function
into a per capita measure
Let z=(1/L):
(Y/L) = F( (K/L), 1 )
• It is convenient to redefine variables in per
capita terms
• y=Y/L output per capita
• k=K/L capital per capita
• And to use f to represent the new function
• y = F(k,1) = f(k)

• A graphical version of the production

function
• Properties of the f(k) production function
– Positive marginal product with respect to k
• Increased k raises the maximum amount of output that can be
produced
– Upward sloping
– Diminishing marginal product of k
• Increased k raises the amount of y that can be produced, but
does so at a diminishing rate
– Concave shape
• These properties derive from our initial production
function, F(K,N), which has positive and
diminishing marginal products for K and for N
separately
– Recall that F(K,N) has constant returns to scale (CRS)
when K and N each change proportionately
• this feature was used to derive the production function in terms
of output per person and capital per person
• A popular example of a production function
is the Cobb-Douglas production function
Y=A•Kα•L1-α with 0 < α < 1
where Y = output, K = capital, L = labor,
and A is productivity.
We need this productivity term in a production
function since there are times when two
economies use essentially the same amount of
the factors of production, K and L, and yet
produced substantially different amounts of Y
• The assumption that α is bounded between zero and
one comes from the marginal product being positive
and diminishing
 Homework
Show that the Cobb-Douglas production
function has:
A. Positive marginal product of capital
B. Positive marginal product of labor
C. Diminishing marginal product of capital
D. Diminishing marginal product of labor
E. Constant returns to scale.
• This production function can be converted
into y as a function of k by dividing it by L
y = Y/L = (1/L) •A•Kα•L1-α
= L-1•A•Kα•L1-α
= A•Kα•L1-α • L-1
= A•Kα•L1-α-1
= A•Kα•L-α
= A•Kα•(1/L)α
= A•(K/L)α
• Therefore, we get y = A•kα
• For the Cobb-Douglas function
– The marginal product of y with respect to k is equal to
the derivative of y with respect to k:
α•A•kα-1
• Assuming A and k are positive, in order for marginal product
of y with respect to k to be positive, α must be positive number
– Using our rules for taking derivatives, the derivative of
this marginal product with respect to k is:
(α-1)•α•A•kα-1-1

OR (α-1)•α•A•kα-2
• Since A, k and α are positive numbers, the only way for
marginal product to be diminishing is if
(α-1) < 0
Which means α < 1
• Graphing the Cobb-Douglas function
 The Basic Solow Model
• Solow won the Noble Prize primarily for
his contributions to Growth Theory
• His model is the foundation of all modern
growth theories
• The simplified Solow model
• Accumulation of physical capital, K, is an
important feature in this model
– Capital is built by investing in capital goods.
– Each dollar of investment, I, builds an additional dollar
of capital. This means the capital stock will increase.
This suggests
ΔK=I
Where ΔK is the change in the capital stock
• Capital increases if we invest
– This simple equation implies the capital stock will
never fall since I can never be negative. But that
implication is not true empirically.
– What is missing is the fact that capital depreciates and
that a lot of investment is done to replenish depreciated
capital. Let D=depreciated capital, and the equation
becomes:
ΔK=I-D
• Capital depreciates by
– wearing out from use or age
– becoming technologically obsolete (i.e. out of date)
• Since we model variables in per capita terms to
compare across countries, divide all variables in
the equation by L
K I D
 
L L L
• Redefine variables so that lower case letters denote
per capita values:
Δk=i-d
• For now assume that L is fixed – not growing
• Assume that investment is a fixed share of
income: =i/y noting that I/Y=i/y and
that 0 <  < 1
• Assume that capital depreciates at a constant rate:
d=δk
• (from D= δK and then dividing by L )
• Assume a general form for a production function:
y=A•f(k)


• Putting all of these assumptions into the equation for
capital accumulation ( Δk=i-d )
We obtain: k    A  f (k)    k
• This equation tells us how the capital stock k is
determined given specific values of , A, δ and
parameters in the production function (e.g. α in a
Cobb-Douglas production function).
• It is convenient to analyze this model graphically (To
make my equation line up with the following graph let
A in the equation be 1, but this is only temporary – it
will be very important to allow for changes in A)
• The graph plots three equations

– Investment per capita:   A  f (k)

– Depreciation per capita: k
– Output per capita: A  f (k)
– The gap between the investment per capita line and
depreciation per capita tells us how much the capital
per capita will change
 
• If the levels of , A, δ and the parameters in the
production function (α if its a Cobb-Douglas production
function) are constant, these 3 curves will be fixed – that is
they will never move
An important implication of these assumptions is that
the economy eventually settles down to a steady state level
of k - and consequently to a steady state level of y as well.
Why?
– If we start out with k>kss depreciation exceeds investment and the
capital stock falls
– If we start out with k<kss investment exceeds depreciation and the
capital stock rises
– When k=kss, depreciation equals investment and the capital stock
does not change
• The tendency for the economy to stay at this point, kss, k reaches that
level, is why the point is called a steady state
– Output is determined by the production function
and the level of capital
• Output follows movements in the capital stock
– At this point, nothing else is changing in the production
function
• When capital settles down to its steady state, output
reaches its steady state, yss, which is determined using
the production function:
yss = Af(kss)
 Homework
• Solve for the steady state values of y and k
in our Solow model, using the Cobb-
Douglas production function.
• The parameters that we have assumed are constant
need not always be constant.
– For example,  may change because:
• Policymakers implement a policy to stimulate more investment
OR
• More savings becomes available to a country, which induces more
investment in that country
– Why? Investment is financed by savings (I=S-CA)

– A or δ also may change for various reasons (we will
consider this later)
• What happens when the investment rate (investment
share of income) increases in our model?

– That is to say, when rises from 1 to  2 which curve or
curves move and what happens as a result?
• When considering how the economy responds to
some change in a parameter or an exogenous
variable, it is useful to assume the economy starts
off in an equilibrium position
– Then we can analyze
• If a change in some parameter (or exogenous variable) pushes
the economy away from equilibrium
– And if so, how this works in the economy and in what way are
variables affected
• If the economy eventually returns to equilibrium
– And if so, how it does this and how is the equilibrium position of
the economy affected
• In the picture the initial position of the economy is

k  k ss1 and y  yss1

• An increase in the investment rate causes
the steady state value of k to increase
– In the graph the new steady state k is kss2
– kss2 is greater than kss1
– So k must rise from kss1 and eventually reaches
kss2
– Once k reaches kss2 it will stop rising and stay
there (until, and unless, something else
happens)
• Similar behavior is found for y
– The initial steady state is yss1.
– As k rises so does y
• This comes from the production function and the fact that k
has a positive marginal product
– k stops rising when it reaches its steady state
• at that point y also reaches its new steady state labeled yss2 in
the graph

– An important implication: A one time increase in the

investment rate will not make the output per capita
grow faster in the long-run since it settles down to a
steady state value in this model
• The previous analysis explained what happens if the
investment rate rises for a country
– But it can just as well describe what happens if two countries have
different investment rates
• Suppose Country 1 has an investment rate of 1
Country 2 has an investment rate of  2 and
 2  1
• Our graphical model predicts that a higher investment share
yields a higher steady state level of output per capita

• What does the cross-country evidence tell us?

• At this point in the course, we don’t know
how to measure steady state output per
capita
– But we do know how to measure output per
capita in a given year
• Also, we can measure the average
investment share over a period of time
• When we plot actual y in a given year
against the average investment rate
(investment share of GDP) we find there is
a significant positive relationship
• This positive relationship between output-per-
capita and the investment share across countries is
predicted by the Solow model
• But we also saw that a permanent increase in the
investment share WILL NOT lead to persistent
growth in y or k
– Eventually k and y settle down to a steady state and
growth discontinues
• The model still needs something to make y and k
both grow over a long period of time as is
observed in most countries
– In order to get y to grow in the Solow model,
productivity (A) must grow over time
• Some may think y and k could grow forever if the
investment share were to continually increase over time
– This is NOT reasonable for at least 2 reasons:
• First, many of the world’s economies have had output per capita
grow consistently over long periods of time even while their
investment shares have stayed roughly constant
• Second, the investment share, I/Y, can not grow forever
– If it kept rising eventually all a country’s income would go toward
investment leaving nothing for consumption, This would make people
very unhappy, in fact dead, since we all need a minimum amount of
consumption just to survive
– And once again, after this share stops rising a country will eventually
arrive at a steady state and stop growing. There could be no persistent
growth.
STOPPED HERE FOR
MIDTERM #1
• Homework: Suppose that δ falls from δ1 to
δ2 in our model (δ1>δ2 ). Hold all other
parameters as exogenous. Explain what
happens to y and k over time. Do y and k
grow forever or will they eventually arrive
at a steady state? How does the new steady
state compare to the initial steady state?
How would δ have to change in order for y
to grow forever? Is it reasonable to think
that the depreciation rate can continue to
change like this over time?
• Homework: Suppose that A rises from A1 to
A2 in our model (A1<A2 ). Explain what
happens to y and k over time. Do y and k
grow forever or will they eventually arrive
at a steady state? How does the new steady
state compare to the initial steady state?
How would A have to change in order for y
to grow forever? Is it reasonable to think
that productivity can continue to change
like this over time?
• The conclusion from class and the two most recent
homework problems is:
– Rising investment share can not cause continuing growth of y
and k for reasons cited a few pages back
– The only way that changes in δ may cause y and k to grow
indefinitely is if the depreciation rate continues to fall. But the
problem with this hypothesis is that
• Depreciation rates do not exhibit any tendency to fall over time
• The depreciate rate can not fall below zero. This means that eventually
there would come a time when depreciation would stop falling - thus k
and y would eventually stop rising
– The only plausible way for y and k to grow indefinitely is if A
continues to rise
• Unfortunately, Solow assumed productivity
grew at some exogenous rate
• Exogenous means outside of or external to the
system. Hence, Solow’s growth model doesn’t
explain what determines growth in productivity or
the economy
• While Solow’s model does not explain why
productivity grows it has been used by
economists to try to explain cross-country
differences in income per capita. We will
also examine some of these cross-country
differences.
An Analytical Solution to Solow’s Model

• An advantage of the Cobb-Douglas production

function is that it allows us to get an analytical
solution for all variables of interest in the model
• Putting Cobb-Douglas in to our capital accumulation
equation yields

k    A  k    k
• If there is no growth in A, then k settles down to a
steady state: Δk=0
• When Δk=0: 
  A  k    k ss  0
ss

• Solve for the kss: 

  A  k    k ss
ss

k ss   A


k ss 
1 A
k ss 
 1/(1 )
 A 
k ss   
  
 
• Then we can solve for yss: : yss  A  k ss


   A  1/(1 )

y ss  A    
  
 

 /(1 )
 A 
yss  A   
  
 /(1 )  /(1 )
(1 )/(1 )  /(1 )  1/(1 ) 
yss  A A   A  
 
• Our final result is:  /(1 )
1/(1 ) 
yss  A  

• Suppose we have two countries, Country i and
Country j. Each will have its own steady state, yi
and yj and the ratio of the steady states is:
i /(1i )
 i
1/(1i ) 
A i  
yi  i 
  j /(1 j )
yj 1/(1 j )   j 
Aj  

 j 
• The previous equation allow each country to
potentially have a unique value α, δ, A or 
• Suppose the investment share is the only thing that
differs across countries.
– Then Ai=Aj, αi=αj and δi=δj,
• And our general result simplifies as follows:

 i /(1i )
1/(1i )  i 
A    /(1 )
yi
i
 i   i 
  
yj  j 
 j /(1 j )   
1/(1 j )  j
A  
j  j 
 
• Measures are available for
– The investment share for every country, and
– The parameter α
• Economic theory tells us that α is equal to the share of income
that goes to capital, which we estimate as roughly equal to 1/3

 1/ 3 1
• So if α=1/3,  
1  2 / 3 2
– Thus the model predicts the ratio of incomes per capita
will equal: 1/ 2
yi  i 
 
yj   
 j
• Numerical examples
– If  i  .2 (20%) and  j  .05 (5%)
yi
  4
1/ 2
then 2
yj
– If  i  .32 (32%) and  j  .02 (2%)
yi
  16 
1/ 2
then 4
yj
– The ratio of income between richest and poorest countries
is more that 50 today
• This evidence suggests that cross-country variation in investment
rates will not be sufficient to make the Solow model explain cross-
country differences in income per capita
• Lets look at a broad selection of countries to see
how well or poorly the model performs
• We use a graph and a large cross section of
countries that plots

– actual data yi/yj along with

1/ 2
 i 
– predicted yi/yj which equals:  
 j 
(this prediction comes from assuming countries only
differ in terms of the investment share and α=1/3 for all
countries)
• If this simple version of Solow’s model is correct,
all the data points will be on a straight line with a
slope of 1 that goes through the origin
– The model is perfectly correct if the only thing that
differs across countries is the investment share
• We don’t expect any macroeconomic theory to be perfectly
correct because macroeconomies are rather complicated social
systems
• How far the data points are from this line yields some
indication of how well or how badly the simple version of
Solow’s model actually works
• So putting that specific line on the previous graph
…
 – The US is, by construction, on the line
• Output per capita in US relative to output per capita in the US is
always equal to 1
– Nothing is learned here, but this helps us draw the line
• What do we learn from this picture?
– First, there is a positive relationship between the investment
rate and income per capita, as is predicted by the model
• Countries with higher investment rates have higher GDP per capita
• Of course, we already saw this in a previous graph
– Second, the Solow model with only the investment rate
differing across countries does NOT do a good job of
explaining the observed differences in output-per-capita
• Ireland is the only other country on the line (the US is also, but it is by
construction - incomes are measured relative to the US level)
– Different investment rates seem to explain the difference in output-per-
capita between Ireland and the US
• Except for Luxemburg which is well above the line, all other countries
are below the line, and most of them are well below the line
• Interpretation: For nearly all of the countries, the model predicts that
output-per-capita should be much higher in countries than is actually
seen when we examine the cross-country data
• So the model in its current form does not look
very promising
– It doesn’t explain why countries grow for an extended
period of time
– It doesn’t explain most of the variation in income per
capita that we observed across countries
• But don’t give up totally on Solow
– We can reexamine the model under alternative
assumptions
• The key conclusion from the last graph: Cross-
country variation in investment rates, by itself, is
unable to explain some of the most important
features in the data
• Our current simplistic version of Solow’s model
does have an interesting property:
– The farther a country operates below its steady state,
the faster capital-per-worker and output-per-worker
will grow
– And similarly, the farther a country operates above its
steady state, the slower capital-per-worker and output-
per-worker will grow
• that property turns out to be a general property of
the Solow model under all kinds of conditions
• It is also a feature of many other growth theories
• To see this point, return to the capital accumulation
equation 
k    A  k    k
• Divide both side of the equation by k
k 1
 Ak 
k
k
• Define the growth rate of k as: k̂ 
k
1
• Then k̂    A  k 
• We can plot the function that determines growth in k. It is
constructed from two terms
1
1. The function of k: Ak
Which is
• A declining function of k
(negative first derivative with respect to k,
which stems from diminishing marginal product of k)
• A convex function of k
(positive second derivative with respect to k)

2. the depreciation rate

(a straight line since it is not a function of k)
• The difference between the two terms gives us the growth rate
of k
• First notice that the graph provides another way of
seeing why kss is a steady state
– k rises when k<kss and k falls when k>kss
• This graph also shows that the farther k is below kss,
the larger is the growth rate of k
• And similarly, the farther k is above kss the slower
is the growth rate of k
– in this case, growth of k is negative when k is above kss
• Output per capita is determined by the production function,
and so it behaves like capital per capita
– Recall the Cobb Douglas production function
• 
y  Ak
– It turns out that if y is determined by this production function, one can
show that the growth rate of y is equal to the growth rate of A plus α
time the growth rate of k:
ˆ  kˆ
ŷ  A
– The farther y is below yss, the larger is the growth rate of y
– The farther y is above yss the slower is the growth rate of y
• in this case, when productivity growth is zero, the growth of y is negative
when y is above yss
 We can derive 3 important conclusions from this
analysis (these can also be are found in a more general
Solow model that would allow more than just the
investment share to vary)

1. If two countries have the same levels of productivity,

depreciation rate and investment rate and they both use
the same production function, but one country starts out
at a lower level of k and y, that country will grow at a
faster rate (until both countries get to the steady state)
– This result is a direct implication of the last graph
– This illustrates a concept known as convergence – holding all
parameters the same across economies, poorer economies will
grow faster than richer economies
2. If two countries start at the same level of y, but one
of the countries has a higher investment rate, then
that country will grow at a faster rate
– The country with higher  has higher steady state levels
for k and y.
• If both countries start at the same levels of k and y, the one that
has farther to go (higher steady state levels of k and y) must grow
at a faster rate to make it to these higher levels of k and y (until
both countries get to their steady states)
– In the following graph, country 2 has a higher investment
rate than country 1:
 2  1
and so country 2 is growing faster (vertical gap between
points b and a) than country 1 (which is in steady state)
Growth rate of k
 2  1
b
c
δ
a
1
2  A  k
1
1  A  k
k
ka kc
3. When a country raises  it will begin to grow at a
faster rate
– The higher investment rate causes k to grow faster which
of course causes y to also grow at a faster rate
– This is seen in the previous graph, by assuming a
1 from 2 to
country’s growth rate rises

Hence, the growth rate for k is initially equal to the

vertical gap between points b and a in the previous graph
– However, as k rises the growth rate begins to slow (the
vertical gap decreases) until the economy reaches the new
steady state (kc) at which point growth stops
• Note: We still are assuming no productivity growth
 Savings and Investment
• Savings and investment are intimately connected
by an important national income accounting
identity
I = S - CA
(I/Y) = (S/Y) - (CA/Y)
• If CA (the current account) is equal to zero (must
be zero for a closed economy) then savings is
equal to investment. Furthermore,
(I/Y) = (S/Y)
investment rate = savings rate
• But in general CA is not small and so we need the more
general result:
(I/Y) = (S/Y) - (CA/Y)
• Feldstein and Horioka (1980) showed a very close
cross-country relationship between the investment rate
(I/Y) and the savings rate (S/Y)
– This is called the Feldstein and Horioka puzzle because:
• International capital markets should funnel savings toward the highest
rate of return, adjusted for risk
– The fact that people do not invest that much in these high yielding
foreign assets implies a level of risk aversion that is unbelievably high
• Hence, the positive relationship between investment
shares and savings rate implies a positive relationship
between savings rates and income-per-capita, which is
very evident in aggregate data
• Using our definition of national savings
S=Spvt+Sgovt
• In the equation from before yields:

I Spvt +Sgovt CA
= -
Y Y Y

OR I Spvt Sgovt CA
=  -
Y Y Y Y

Thus for the domestic investment rate to increase, at least one of

the following must occur:
• An increase in the private savings rate;
• An increase in the government savings rate;
• An increase in the rate at which net foreign savings flows in (the net
inflow of foreign savings is equal to –CA)
• The empirical evidence relating savings
rates to income per capita provides some
support for Solow’s model – But other
interpretations are possible
– Frequently, alternative structural explanations
for a set of empirical evidence exist, and many
times an alternative works in the opposite way
to the first theory
How does population growth affect
variables in the Solow model?
– For a fixed level of capital (K), a higher population
(L) reduces capital per person (k) and so output per
person (y) would also be smaller
• This effect of population on capital is known as capital
dilution (As L gets larger a given stock of capital must
be shared by more workers)
– for output per capita to reach a steady state, capital
per person must also reach a steady state
• Thus if population grows, capital must grow at the same
rate as population growth to maintain a steady state
capital per person
• We need to modify the capital accumlation
equation to allow for population growth
since we derived the earlier version
assuming there was no population growth
• First, we will show that the earlier capital
acummulation equation doesn’t work
properly and then we will repair it
• Recall our capital per person accumulation equation
(assuming fixed population)

k    A  f (k)    k
• We know that by dividing both sides of this
equation by k we get an expression for the growth
rate of k

k
 k̂    A  f (k) / k  
k
• Suppose that the investment rate and the depreciation rate are
equal to zero. In that case the last equation says that k grows at
the rate zero. If k growth is zero then k is at some constant
value.
• Recall an even earlier equation: ΔK=I-D
– If the investment rate = 0, investment = 0
– If the depreciation rate = 0, depreciation = 0
– Thus ΔK=0, which means K grows at a rate of zero (divide both side of
the equation by K, to get ΔK/K


• Here’s the problem: k=K/L
– If K is NOT growing while L is growing then k must be shrinking ---- it
can NOT reach a positive steady state level
• We need to fix our Δk/k equation

• The fix is simple and intuitive: If δ and are zero then K must
also be zero. And, if L grows at a constant rate, n, k will be
shrinking at that rate, that is k will grow at the rate –n
• This suggests that the fix obtains by subtracting the population
growth rate from the right side of the previous growth rate of k
equation
 We change the capital accumulation
equation in specific ways
• It is convenient to change capital accumulation from a
difference equation to a differential equation (this
means changing from discrete time to continuous time
model)
• From calculus we know that:
K dK
as t  0, lim 
t dt
• Dividing by some finite number doesn’t change that,
so: 1 K 1 dK
as t  0, lim 
K t K dt
• Thus our original capital accumulation equation
becomes:
dK
 F(A, K, L)  K
dt
• We, also, will modify our Cobb Douglas production
function to have labor augmenting productivity
 1
F(A, K, L)  K (AL)
• Labor augmenting means that in the production
function A multiples L.
• AL is sometimes called “effective labor input”.
– It adjusts labor effort by the productivity of labor
 Putting this all together yields our
new capital accumulation equation
•
dK  1
 K (AL)  K
dt
• But we want to put things in terms of output per capita and
given constant returns to K and L that means when the
production function is divided by L that will make of output
per capita a function of productivity and capital per capita.
• Thus we would like to put our capital accumulation equation
in terms of capital per capita.
 We will assume for convenience:
Population (or the work force) grows
at a constant rate
 1  dL
  n
 L  dt
• Where n is a constant equal to the growth rate
• This is a differential equation that is easy to solve (if
you know something about differential equations)
• The solution is: L(t)=Loent
• This is the equation for the level of L if it grows at a
constant rate
 How are capital growth and capital
per capita growth related?
• First, recall that k = K/L
• We can take the logarithm of the last equation
log(k)  log(K)  log(L)
• Then take time derivative of this equation
d d d
log(k)  log(K)  log(L)
dt dt dt
• And finally obtain
 1  dk  1  dK  1  dL
     
 k  dt  K  dt  L  dt
• Thus, we can take the production capital accumulation equation from before
divide by K, and use the last equation plus the assumption that labor grows at a
constant rate to get:
 1  dk  1 
 
 k  dt
  
K
 K 
(AL)1
 K   n

• Or simplifying we obtain:

 1  dk 1 1 1 1

   K (AL)  (   n)  k A  (  n)
 k  dt
• Note that the resulting equation:
 1  dk 1 1
   k A  (  n)
 k  dt
• Is the same as our previous equation for growth rate of k, except that we have an
additional term, -n. We need that sort of adjustment for population growth, to
make the analysis right in terms of k.
• Thus we will see that there is a steady state for this
model too, and the condition is the same as before:
dk
0
dt
• It comes essentially from the same graph, assuming all
the parameters, including productivity, stay fixed.
• In this case, the steady state is the solution for kss to:
1 1
0  k ss A  (  n)
• Which has the following solution:
1/(1 )
  
k ss  A  
n 
• To prove that the model reaches a steady state, first
multiply the growth of k equation by k:
dk  1
 k A  (  n)k
dt
• And note that from our previous Cobb-Douglas
production function for Y, it is easy to show that:
 1
yk A
• Thus we can graph the dynamic equation in k, assuming
all parameters (including A) are fixed
(note that δ=d in the graph, both refer to the
depreciation rate)
• We can add the production function to
answer questions about what happens to
output per capita and consumption per
capita
– Output per capita comes directly from the
production function
– Consumption per capita comes from: y – i,
which is the gap between the production
function and the investment line
 We can use this model (or any
model) to do experiments
• The first experiment examines the effects of
an increased investment rate
• The model predicts that countries with
higher investment rates will tend to have
higher levels of output per capita (and
higher levels of capital per capita)
• There is some empirical support for the
hypothesis that countries with higher
investment rates have higher output per
capita
• A second experiment examines the
relationship between population growth rate
and output per capita
• The model predicts that countries with higher
population growth rates will tend to have lower
values for output per capita (and lower capital per
capita)
– We could also examine the effects of an increase in the
depreciation rate, using the previous graph. The results
are the same, both qualitatively and quantitatively. But
we don’t expect there to be great differences in
depreciation rate across countries.
• There is some empirical support for the hypothesis
that countries with higher population growth rates
have lower output per capita
• We can take the previous capital accumulation equation
and divide by k:
1 dk
 k 1A1  (  n)
k dt
• This is the growth equation for k, similar to one we
derived before when there was no population growth
– Compared to the earlier graph, the only differences are that
we used a differential equation and we allowed for population
growth
– It looks very similar and has similar implications
 This model maintains the key conclusions we mentioned before
in a model without population growth
- Convergence
- Transition dynamics

If two countries have the same levels of productivity,

depreciation rate, investment rate and NOW POPULATION
GROWTH, they both use the same production function, but one
country starts out at a lower level of k and y, that country will
grow at a faster rate (until both countries get to the steady state)

Another way of looking at this graph: The farther below (above)

its steady state the faster (slower) an economy will grow
• You can also show that:
– If two countries start at the same level of y, but one of
the countries has a higher investment rate, then that
country will grow at a faster rate initially
– When a country raises  it will begin to grow at a faster
rate (though not forever because it will again reach a
steady state in the long run)
• This version of Solow’s model maintains a key
limitation: There is no growth in k or y in the long-run
– That is the implication of these two variables settling down
to a steady state
– You might conjecture that a continual decline in the
population growth rate could be a plausible explanation for
persistent increases in output per capita. But it is not.
• Eventually n would have to become negative for this declining
population growth rate story to explain continual growth in y. And
once population growth becomes negative the level of population is
headed to zero
• Also, many countries experience growth in output per capita with no
change in population growth rate
• Now we will allow for productivity growth
• If productivity grows at a constant rate, g, then, we can write this
as the following differential equation
1 dA
g
A dt
• As before, we can solve this sort of differential equation for the
level of A: A(t)=Aoegt

• Productivity growth occurs whenever there is a new idea about a

product innovation or a new idea about how to produce a product
more efficiently (e,g, producing the same goods at a lower cost
or producing more goods at no additional cost)
• We will see that productivity growth is the only thing that yields
a plausible explanation for growth in y, k or any other per capita
variables
• It would be nice if we could put the model in
terms of variables that do not grow. Then we
might be able to obtain variables that achieve
steady states.
• We saw previously that by dividing by labor
achieved steady state when that was the only
thing growing. Since we now have labor and
productivity growing, we can try to divide by
both to see if that will give us a transformation of
variables that achieves steady state.
• Recall our newest production function
 1
Y  F(A, K, L)  K (AL)
• If we divide by AL we get:
 1  
Y K (AL) K  K 
  
 
AL AL (AL)  AL 

Y K
• Define y
% and k%
AL AL
%
yk
% 
• Then the production function becomes
• these transformed variables are defined as:

• output per effective labor unit: %

y
• capital per effective labor unit: k%
• where AL measures effective labor units, being the
quantity of labor adjusted for productivity

%the capital technology ratio

• Jones likes to call k
• Now, return to our capital accumulation equation
dK
 Y  K
dt
• First lets use a new notation that is common for
growth modeling and common in mathematics.

&
• For any variable Z, define: Z 
dZ
dt
& dK
• That means: K 
dt
• Using this definition in the capital accumulation
equation &
K  Y  K
• Then dividing by K yields: &
K Y
 
K K
Y
Y AL %
y %1
• Note that   k
K K %
k
AL
• And so we obtain:
&
K %1
 k  
K
• We need an expression for the growth rate of capital in
terms of capital per effective labor unit.
• Given that: k% K
AL
• We know we can take logarithms:
%  log(K)  log(L)  log(A)
log(k)
• Then take time derivative of this equation
d % d d d
log(k)  log(K)  log(L)  log(A)
dt dt dt dt
• And finally obtain
 
1 %
dk  1  dK  1  dL  1  dA
 %      
 k  dt  K  dt  L  dt  A  dt
• And the last equation becomes:
&  &
k
% K
   ng
%
k K
  
• And plugging this result into:
&
K
 k% 1

K
&
%
• Yields: k %1  (g  n  )
%  k
k
• Which can be written as:
& %
% %
k  k  (g  n  )k
 Homework
• Solve for the steady state values of capital and output per
effective labor unit in our model with population growth
and productivity growth, assuming a steady state occurs.
(This means solve for the values of endogenous variables
in terms of parameters and exogenous variables)
• Calculate the steady state ratio of K to Y (called the
aggregate capital to output ratio). Is it the same as the
ratio of k to y or is it different?
• On a ratio scale, graph steady state values of K and Y
(hint the ratio scale and the logarithm scale are the
same). How fast are these variables each growing?
• On the same ratio scale, graph steady state values of k
and y. How fast are each of these variables growing?
 Does this model achieve a steady
state?
• Yes
• The two lines are fixed – that is they stay in one position
– for fixed values of parameters of n, g, δ, α and
 

• There is only one positive steady state value for

k%
• And starting from any positive level of
k%
the dynamics will push the economy in the
 direction
of this steady state.
• Now we can examine an important question
that we’ve asked before with these models:
What happens to the economy when the
investment rate increases?
• An increase in the investment rate shifts the
investment per effective labor unit line up
• Capital per effective labor unit begins to rise since
investment exceeds depreciation, with both of
these now measured per effective labor unit
• Finally, the economy settles down to a new steady
state level of capital per effective labor unit
• We can also analyze this model using the growth rate
of capital per effective labor unit form of the
equation

&
%
k %1
%  k  (g  n   )
k
• Graphing this equation yields the following which
illustrates how an increase in the investment rate
causes fast growth of capital per effective labor unit
initially,
• But, as the economy gets closer and closer to the
steady state, the growth rate of capital per effective
labor unit slows down
• And using the last equation it is easy to solve for
the steady state value of capital:
1/(1 )
%   
k ss   
gn
• And steady state output per capita comes from
putting the previous solution into the production
function:  /(1 )
  
yss  
% 
gn
• This model yields the same predictions as
before:
– The farther an economy is below its steady
state the faster it grows
– The farther an economy is above its steady state
the slower it grows
We now examine the growth of
output per capita
• We can get output per capita from the definition of
output per effective labor unit. Since
Y
y
%
AL
• Multiplying both sides by A gives: If we divide by
AL we get: Y
y %
yA
L
y& y& A&
%
• This means that:  
y %
y A
• And from our production function, in per effective
labor units:
yk
% %
• We know that:

&
% &
%
y k
%
%
y k
• So finally,

& &
%
y& A k
 %
y A k
• Using the constant growth assumption for A, we get:
&
%
y& k
g
y %
k
• This equation tells us that:
– In the steady state y grows at the rate of g
– When the economy is below its steady state, y grows at a rate greater
than g
– And when the economy is operating above its steady state y is grows
at a rate less than g

• Thus, when the investment rate went up, y grows faster than g
for a while, but eventually the economy returns to steady state
and y growth rate returns to g
How does output per capita behave
over time?
• Recall that: yt  %
ytAt
• Using our expression for A, we obtain: :

yt  %
y t A 0e gt

• This equations tells us how y behaves over time in

this model:
– Changes in output per effective labor unit shift the level of
y, but don’t affect its long-run trend which is given by the
trend in A
– Changes in A0 will also shift the level of y, but not its trend
line
• This idea inspired what became known as Real Business Cycle
models (the first generation of those models)
• Thus we see that an increase in the investment rate
will cause the level of y to increase, but not affect its
long-run growth rate
• The following graph illustrates precisely what would
happen in this model, plotting how the log of y
changes over time:

log y t  log(y
%t A 0 )  gt
 Once again the model predicts
convergence and transition dynamics
• The economy converges to a steady state
for fixed values of the parameters
– In the steady state real per capita variables grow
at the rate g
• The farther below/above steady state the
faster/slower the economy will grow.
• Suppose there are two economies named
– InitiallyBehind (IB)
– InitiallyAhead (IA)
• We know that IB will grow faster in this
model
– Transitional dynamics are the driving force
behind convergence
How well do the data support the
convergence hypothesis?
We have convergence when we look
at the sample of countries for which
we have long time series

• For some countries we can obtain

reasonably reliable estimates of GDP and
population dating back at least as far as the
late 1800s
We also find convergence when we
look at relatively well-developed
countries in the postwar period
– This is a broader selection of countries, as well
as a more recent and more reliable sample of
data for the countries
However, when we broaden the set
of countries to include all countries
in the postwar period, there is
absence of convergence

– Many of the countries that are much poorer

than the developed countries of the world are
growing slower than those developed countries
Adding Human Capital to the model
• It seems plausible that human capital
differences across countries may provide a
reasonable explanation for why countries
have different performance levels
– Specifically, a low level of human capital may
explain why many poor countries show no
evidence of converging
– We also want to allow for differences in
population growth and investment rate across
countries
• Assume that a person’s human capital augments
their labor input (labor measured in terms of hours
or people)
– In that case, human capital enters the production
function like productivity
– We multiply labor input, L, by human capital per
person, h
• This change will lead us to redefine output per
effective labor unit as:
Y
y
%
hLA
• But, after that redefinition our analysis remains
essentially the same
• So we can plot the predicted ratio of GDP
per capita versus the actual GDP per capita,
holding A to be the same for each country
• This is a standard problem with Solow type
models: They tend to over-predict how much
income per capita countries will have, relative to
US income per capita when we fix productivity in
all countries to be the same
• So what if we allow productivity to differ across
countries? We can use the equation to measure
productivity in each country
• When you allow productivity to be different
across countries, and use the Solow model’s
steady state to calculate it, you find
– extremely wide variation in productivity in the
world
– Productivity is highly correlated with income
per capita
– richer countries are more productive than
poorer countries
• In fact, if you were to decompose cross
country output per capita differences into
productivity differences and differences
explained by all the other factors in the
Solow model, productivity differences
account for most of the cross country
differences in y
 Conditional Convergence
• Our model allows each country to have its own
steady state depending on its own investment rate,
population growth rate and human capital.
• Conditional convergence is the idea that countries
converge to their own steady state, and how far
they are away from their own steady state is what
determines how fast they are growing
– We are assuming that g is the same for all countries
 Conditional convergence holds up
fairly well in the data
%
y1960
• Growth rate since 1960 is plotted against
%
y*
 Growth Accounting
• is is a technique for decomposing economic
growth into contributions from various sources
• More specifically, growth accounting decomposes
output per capita growth into contributions from
the various factors (e.g. capital per capita growth)
and also from productivity growth
– This last component is sometimes called
• the Solow residual or
• the contribution from multifactor productivity

• And is sometimes also called our measure of ignorance since it

is the part of growth that we can’t explain with measurable
changes in factors
• A simplistic example of how to do a
decomposition calculation
• In practice economists have tried to further
decompose growth into additional
components such as
– Labor composition
• This is how labor may be shifting from production
to R&D, or vice versa
– Also, the capital per worker contribution can be
separated into IT (information technology)
capital and other components
Table 2.1
• Growth accounting teaches us that the big changes
in the growth rate of output per capita (or per
worker) are largely the result of changes in the
growth rate of productivity
• From this accounting we observe the productivity
growth slow-down that plagued all countries
starting in the early 1970s
– This was particularly problematic for most of the
developed countries
• And we can also see that strong productivity
growth returned starting in the mid 1990s for the
US.
– The US benefited more from this productivity growth
than most other developed countries
 Why did productivity growth slow
down in the early 1970s?
• Historically, the slow-down has been attributed to
various sources
• One that looked particularly promising was the
rapid rise in oil price shocks in the 1970s
– In out model this amounts to a reduction in A0, or better
yet, a persistent decline in A0 which would lead to a
persistent slow down in the growth rate of output per
capita
– The problem: In the 1980s there was no resurgence of
rapid productivity growth as oil prices fell dramatically
• A variety of other explanations have also been
offered . For example:
– Running out of ideas
– Natural resource depletion
– Excessive regulatory burden on firms
– Reduced labor ability and/or reduced labor force
work ethic
• No single explanation seems a totally
satisfactory by itself
• Each one may provide a partial explanation
• But none of them, or all taken together, seem to
be a good candidate for explaining why
productivity growth picked up in the mid-1990s
• An explanation for both why the productivity growth slowed
and why it then increased in 1990s: Technological revolution
• The revolution is thought to be the expanding use of computers
and information technology. The basic idea:
– Computers (and later, the Web) represent a general purpose technology
that can broadly affect how we produce and the types of goods we make
– Any new technology requires time to learn how to utilize it properly
– This takes labor time away from production causes output to fall relative
to the amount of labor input
• A larger share of labor is being devoted to learning rather than production
• Economic historians, such as Paul David, have examined how
the introduction of new general purpose technologies (e.g. steam
engines, electricity, etc.) in earlier periods coincided with a
decline in output per capita that lasted for some period of time,
but eventually in all those cases more rapid growth returned
• This story can explain the resurgence of productivity growth in
the 90s, By then, people understood the new technologies well
enough to make significant and frequent improvements in
productivity. Thus caused output per capita to start growing at a
faster rate
 Endogenous Growth Models
• Endogenous growth usually make the
growth rate of productivity endogenous
• However, some models generate growth by
having constant returns to scale for capital
in the production function
– This can be generated from learning by doing,
for example.
– However, CRS for capital yields implications
that do not appear valid in the data
• So I’m going to ignore these models from now on
Romer’s model of Endogenous
Growth
• The effectiveness of workers depends on
how many ideas there are in the economy
– Assumption: The more ideas there are, the more
ways ideas can be combined to create new ideas
OR the more ideas can be refined/extended to
create new ideas
• How are world population, world
population growth and world output growth
related?
• World population has been increasing since
the dawn of man (with some notable
exceptions like the Dark Ages)
– But while the population was rising the
population growth was at a very small rate,
until relatively recently in human history
• Around 300 years ago the rate of output
growth started to increase and that
coincides with the increasing world
population growth rate
• There is some evidence that the number of
new ideas has been growing rapidly,
– Suppose new ideas are measured by the number
of patents
– US patents have increased dramatically over
time
– But the growth rate of ideas has not been rising
• Trust me – it may not be evident from this graph
• However, the number of scientists has risen
dramatically over time
– This implies that if the model is correct, the
growth rates for productivity and output per
capita should have been rising during this time
period
• Romer’s model predicts that this rise in the
number of scientists and engineers will lead
to faster productivity growth
• There is no clear relationship between the
number of scientists and either the pace of
output growth in the developed world or the
pace at which ideas are created.
• Since 1970 the dramatic increase in
scientists and engineers in the developed
world has not cause these countries to grow
at noticeably faster rates

• This has suggested to Jones and others that

we modify our growth model along certain
plausible lines
• Therefore, this version of the endogenous growth
model predicts that higher growth rates of
productivity and output will coincide with a higher
population growth rate.
• This is what the evidence shows, in contrast to the
first endogenous growth model which predicted
that higher level of population would lead to
higher growth rates in productivity and output.
• This implication of the model will be tested in
your (expected) lifetime.
– World population growth will slow down substantially
within about 50 years – maybe go to 0%.
– This model predicts productivity growth will slow
dramatically, maybe even stop growing
• Schumpeterian models of endogenous
growth