David N.

Weil October 21, 2008

Lecture Notes in Macroeconomics

Growth, Part 2

Endogenous Growth Theory

I have introduced the OLG model here because of it is mathematically very similar to the
Ramsey and Solow models, and because it is a model that is useful in many contexts. But now I
want to forget about the OLG model and ask a more general question about the growth setup that
we have been examining: is it right?

A bit of academic sociology: The Ramsey model dates back to the 1920's. The Solow model
was done in the 1950's. During the 1960's, growth theory was a big part of work on
macroeconomics, and many variations on the basic Solow view were fleshed out (for example,
different ways to model technological change). Then, in the 1970's growth theory petered out,
and macroeconomics focused more on business cycles. What we have covered so far is all stuff
that you might have learned 20 years ago.

Then, starting around 1986, there was a massive increase in the popularity of growth
theory -- and this explosion continues today to the point where it sometimes seems like half the
papers in macro are on growth theory; whole issues of journals are devoted to it, etc.

People identified with the new theory: Paul Romer, Robert Lucas

Reasons for the explosion

1. Theoretical problem with the Solow model (exogenous technological change; Solow
residuals)

2. Availability of new data -- Summers and Heston -- that allowed testing of Solow model.

3. Exhaustion of business cycle research (agreement on many points) combined with long period
w/out recessions.

4. Intellectual changes: Development of new tools by economists that allowed modelling of

endog growth stuff.

5. Realization of relative importance of growth as compared to cycles -- driven home by the

productivity slowdown.

1
Let me first amplify some of these problems

-----------------------------------------------
Solow Residual

One of the potential theoretical problems with the Solow model is that technological growth is
taken to be exogenous. To decide whether this is a problem, we might want to know how big
technological growth is. We have said that in the steady state of the Solow model, total output
grows at the rate n+g -- and since we know the rate of growth of total output (that is, /Y), and we
know n, we can back out g. But what if we are not in the steady state?

Robert Solow provided the answer to this problem. It is important to note that the
technique that he provides is not dependent on the other parts of the Solow model (for example
the exogenous saving rate).

Start with the production function:

Y = A*F(K,L)

Where A is a measure of the state of technology and F() is some CRS function that combines
capital and labor. A is called "total factor productivity."

[Note that another commonly used measure of productivity is Y/L -- that is, output per worker.
While this does get a some measure of economic well being, it does not pretend to measure
technology. One can easily increase Y/L by simply raising the amount of capital in the country,
without necessarily inventing any new methods of production. By contrast, total factor
productivity cannot be changed just by changing the amount of inputs.]

We can totally differentiate to get

We differnetiate w.r.t. time to get

( K , L) + AF ( K , L) L
+ AF ( K , L) K

Y
= AF L K

now divide the left and right sides by the corresponding sides of the production function to get:

2
 Y
A
AFL ( K , L) L
AFK ( K , L) K

= + +
Y A AF ( K , L) AF ( K , L)

We now multiply top and bottom of the second term on the left by L, and do the same with K in
the third term. We use a hat over a variable to indicate its growth rate, that is x
/ x

Y = A+(1- K
)L+

where is just capital's share of output.

This equation shows that there are three reasons that output can grow: increase in capital,
increase in labor, and improvement in technology

Using this equation, we can easily figure out the rate of growth of technology, which is just A
hat.. All of the other pieces of the equation (growth rates of capital, labor, and output, and
capital's share of output) are, in principle, observable.

When we do this exercise, we see two things (see my textbook):

1) technological progress is a big contributor to growth
2) technological progress is variable. There was a big slowdown in the 1970s and 1980s,
then a speed up in the 1990s.

[to be added: conceptual problem with growth accounting. Think about the steady state in the
Solow model. Accounting procedure will attribute only some of growth to technology, while in a
logical sense all of it is due to technology.

Add development accounting here [students: See my textbook for a full treatment]

----------------------------
move this elsewhere? Maybe to the beginning of the growth section? Or to the end of the
course?

====> extend this by adding stuff from Van Wincoop, on the welfare cost of business cycles@
JME, 1994.
===> there is also cc stuff on how it relates to equity premium

3
Lucas' calculation of the relative importance of long- and short-run. (Models of Business Cycles,
Yrjo Jahnsson Lectures, published 1987). Start with the usual discounted utility function

U = E
c

t
1-

1-
t=0

and a stochastic process for consumption that has both a deterministic trend and variation around
the trend:

t
zt
c t = (1+ )(1+ ) 1 2
-
e 2 z

ln( z t ) N(0, 2z )

The term in parentheses is a lognormally distributed multiplicative error with a mean of one.

Assigning appropriate values for the trend growth rate of consumption and the variance of
consumption (both averages of the post WWII US economy), and choosing a value for and ,
one can calculate the expected value of consumption. Lucas then asks two questions. First, what
is the variation in that would be equivalent to eliminating consumption variability entirely?
Answer: .00042 (.04 percent). Second, what would be the variation in that would be
equivalent to raising the growth rate (mu) by one percent. Answer: .17. Conclude: cycles are
unimportant.

What might be wrong with this sort of calculation? One problem is that it ignores
heterogeneity B in a recession it is not that each person has a 5% consumption deline, but that
10% of people have a 50% decline. But note that heterogeneity alone is not enough to convince
us that cycles are bad B after all, in a boom the unemployment rate is low....

Note: habit formation is almost certainly part of the explanation for why people care so
much about cycles relative to growth.

---------------------------------

2. A second problem with the Solow model comes from its predictions about differences in the
marginal product of capital across countries.
Suppose that we have two countries that differ in output by a factor of 5. Assume that
they have the same technology, which is Cobb Douglas with a capital share of alpha.

4
ya k
= 5 = a
yb kb

ka
= 51/
kb

1
MPK a ka1 ka 1
= = =5
MPK b kb1 kb

If = 1/ 3 , a difference of a factor of 5 between the level of output per capita between two
countries should mean that their marginal products of capital differ by a factor of 25. If the
interest rate is
r = f '(k) -
Then, say, for the US the real rate were 5% and depreciation were 5%, then f '(k) in the US
would be 10%, and in a poor country it would be 250% (and the interest rate would be 245%)!!!
This seems pretty much larger than real interest rates that we observe in other countries. Nor do
we observe the flows of capital that would be expected in response to such large differences in
interest rates.

Note that you can get around this problem by assuming that different countries have different
technologies. But you have to believe that technologies are really different to produce the
differences in output observed. [homework question: can calculate how many years behind other
country must be]

3. A third problem with the Solow/Ramsey model that emerged when better data on cross
country growth became available was the convergence problem: Assume that every country
has the same utility function, discount rate, depreciation rate, and production technology.
Suppose that we then look at all of these countries, and notice that they differ in income per
capita. How does the Solow model explain such differences? Clearly, it must be that they
differed in their initial level of capital. For example, if you looked at Germany and Japan after
the war, they had very low levels of capital, and so low output.
What does the Solow/Ramsey model say should happen in such a case? It says that the poor
countries should grow faster than the rich countries (remember that all countries grow because of
technological progress). So if we graphed the growth rate of income per capita against the initial
level of income per capita , we should see a downward sloping line.
In fact, we see nothing of the sort-- rather we see a big mess with no particular slope. This is
the failure of convergence.

So there are three strikes against the Solow model

-exogenous tech change

5
-interest differentials
-convergence

----------------------

Endogenous growth models

[note for future drafts: could make the production function in all of the this or earlier material
y=Ak instead of y=k -- the A is just a constant that does not affect any of the qualitative results.
Doing it this way has the advantage of making it consistent with the endogenous growth section]

We now take up endogenous growth models that solve some of the problems described above.
We start with the simplest of the endogenous growth models, then move on to more complicated
ones.

Start with the Solow growth model, using a Cobb Douglas production function and no
technological change. Y=Ak, where A is a constant measuring the state of technology. The
usual differential equation is:

k
= sAk (n + )k

we can re-write this as:

k

= sAk 1 (n + )
k

As before We can graph the two parts of the right hand side of this equation.

[picture]

The difference between these two is k
/ k , the growth rate of the capital stock. What about the
growth rate of output? To see the relation between these, takes logs and differentiate the
production function:

y
k

=
y k

So the growth rate of output is just proportional to the growth rate of capital.

6
 So what happens to the growth rate of output per capita away from the steady state? If we are
below the steady state, the growth rate of output monotonically decreases as we approach the
steady state.

What happens to the growth of output if we increase saving. in this case, the sAk-1 curve
shifts up, we have a temporary increase in growth, then growth peters out to zero again.

Now consider what happens to this picture as we raise . You should see that the
downward sloping line gets less sloped.

To see the effect of this, imagine starting in a steady state, then raising the saving rate by
the same amount in two economies which have different values for . Initially, the two
economies have similar growth rates of capital. [you can check this by looking back at the
formulas for convergence speed. The bigger is , the longer is the half life of differences.
Specifically, the half life of the difference is .7 / [ (1 )(n + g + ) ] ] -- but in the case where is
low, growth peters out quickly, while in the case where is high, growth lasts a long time, and it
takes longer to get to the new steady state.

Now think about the limiting case, where =1. In this case, growth never stops.

y
k

= = sA (n + )
y k

So here growth is endogenous, in the sense that it is determined by the other parameters in the
model (not exogenously).

This is the simplest endogenous growth model, sometimes called the Ak model, since the
production function is

y = Ak (ie cobb douglas with =1)

Several things to notice about the model:

1). What does it say about interest rate differentials? Well, f '(k) is just A, and so r=A-. So
what should the interest rate differential be between rich and poor countries? Nothing. This is a
big improvement over the huge differentials implied by the Solow model.

2.) Suppose that we looked at data on the determinants of growth rates of countries. What does
this model say about the relation between a country's level of income and its growth rate? It says
that there is no relation! Growth is just a function of the level of saving and of (n+). This

7
seems to be a good property, because as we saw earlier, there is no relation visible when you just
plot growth and initial income.

3) What about saving? The model says that higher saving will be associated with higher growth.

4) n? Model says higher n leads to lower growth. Note in the regular Solow model, n and s
affect level but not steady state growth rate.

In a moment we will talk about this production function, but first lets do the Ramsey version of
the Ak model. The Hamiltonian for the social planner is:

1-
c- t
H =e + (Ak - c - (n+ )k)
1-

Solving through all of the usual steps give the equation of motion for consumption

c
A - - n -
=
c
(alternatively we can just get this from the same c
/ c = (1/ )(r n ) logic as before).

Note a groovy thing about this equation: it doesn't depend on the level of capital in the
economy!

What about the growth rates of k and y?

We know that y
/ y = k
/ k

And we can re-write the differential equation for capital to be

k
c
= A (n + )
k k

so substitute y
/ y for k
/ k , and replace the k with y/A

y
c
= A A (n + )
y y

8
 Now look at this equation for a moment. If y
/ y = c
/ c , then (c/y) will be constant, and so y
/ y
will be constant. If y
/ y > c
/ c , then c/y is falling and y
/ y is rising. In this case, y
/ y will
remain larger than c
/ c , and c/y will fall toward zero. This is clearly not the optimal path.
Similarly, if y
/ y < c
/ c , then c/y will rise continually, eventually going over 1, which is not
c
y
k

feasible. So, = = .
c y k

And all of these growth rates are determined by the preference parameters.

Note that all of this is true all of the time -- not just in a steady state. Put another way, all
of the time is a steady state. There are no "transitional dynamics," as there are in the Ramsey
model. Given some level of output, you just start growing at a rate determined by your
parameters, and keep at this growth rate forever.

So two countries with the same parameters but different starting points would remain the same
percentage difference apart forever.

[picture]

And two countries with different parameters (say different discount rates) would grow at
different rates forever -- and would end up arbitrarily far apart in the long run. So if Japan is
growing faster than the US now, there is no reason why it shouldn't continue to do so once it
passes the US.

You can also solve for the saving rate explicitly given the c
/ c and y
/ y equations:

c
A - (n+ ) - y
c
= = = A - A( ) -(n+ )= A - A(1- s) - (n+ )= sA - (n+ )
c y y

Notice that there is the same relationship between growth and saving that we saw before.

Continuing to solve gives:

1 1 ( - 1)(n+ ) -
s= [A+( - 1)(n + ) - ] = [1+ ]
A A

9
It is clear that ds/d is negative: if you care about the future less, you will save less.

The derivative of saving with respect to (skipping a few intermediate steps) is

ds n + +
= - -2 [1- ]
d A

This is negative iff A>(n++), which is the condition for consumption (and output) growth to
be positive. The intuition is that, if consumption is being chosen such that it is growing over
time, then an increase in risk aversion makes the marginal utility of consumption lower in the
future, and so makes consumption growth less desirable. So the saving rate will fall. If
consumption was initially falling, then an increase in the coefficient of relative risk aversion will
make the marginal utility of consumption higher in the future, and so will raise the saving rate.
[remember in the original Ramsey model, high meant that the transition from an initial capital
stock to the steady state was slow -- this is the same intuition, but it lasts forever].

Finally, take the derivative of the saving rate with respect to A:

ds 1 ( - 1)(n + ) -
=- [ 2
]
dA A

This can be either positive or negative for reasonable sets of parameters. For example, if n+=,
then it will be zero for =2 and negative for higher values of . Indeed, the higher is , the more
likely it is to be negative: the reason is that for big the consumption smoothing effect
dominates. Note that having ds/dA<0 seems to contradict the spirit of endogenous growth
models (see Carroll, Overland, and Weil for more on this point).

So in general the Ramsey version of this model is just like the Solow version (without the
transitional dynamics that the usual Ramsey model has) -- it's only advantage is that it tells us the
saving rate explicitly as a function of the parameters: Raising or lowers the saving rate.
[Note: we also could have done an OLG version of the AK model.]

so now we have the question: why should be believe in this production function anyway?

How can we reconcile this type of model where =1 with the observation that capital's share of
output is just 1/3?

10
 Two stories:

First story for endogenous growth: externalities.

remember externalities from micro: If you have the apple orchard next to the place where they
keep the bees, etc.

Take the production fn Y = A KL

Now lets imagine that A, the state of technology in the economy, is a function of the amount of
capital.

Why this might be: Well you might think that the more investing that has been done in the
past, the more we would have learned about how to do it right. This is story about "learning by
doing" which traces back to Kenneth Arrow. Well the total amount of investment that has been
done in the past is -- assuming no depreciation -- is just K.

So if we made A a function of capital per person: A = B*(K/L), then the production function
would be

Y = B KL (k) (where k = K/L)

If we assume that +=1, we get constant returns to scale for firms. (note that for a firm, the
economy wide level of k is taken as given). This is good for two reasons: first, it seems to
accord with what we see in the world: that big firms are not much more efficient than small ones.
Second, as you will see in micro, if firms do not have CRS production functions then perfectly
competitive general equilibrium does not exist. This latter problem -- which is a limitation of
theory -- can be gotten around in various ways, but we don't deal with it here.

[A side note, if we had written the externality as a function of total capital, rather than capital per
worker, we would have gotten the aggregate production function:

Y = B K+ L

In this case, there are IRS at the national level -- that is doubling the size of the country more
than doubles output, which seems like a silly story.]

We can write the production function in per capita terms as

y =B k k

11
 So clearly, if +=1, then we have the Ak model again, with A=B.

[I now set B=1, because it is not interesting]

Now one of the things that you learn in micro is that in the presence of externalities, the
competitive equilibrium is not efficient. The reason is that people and firms do not take
externalities into account in making their decisions.

In this case, the firm takes the economy-wide level of k as fixed, and so it's production function
is

yi = kki

or in non-per capita terms: Yi = k Ki Li1-

Notice that from the firm's perspective, labor is productive, and so labor earns a share (1-) of
total output.

The marginal product of capital from the perspective of a firm is

MPK (firm) = ki-1 k

assuming that the economy is made up of identical firms (and so they all have the same
capital/labor ratios)

MPK (firm) = k+-1

While the social planner takes into account the externality of capital, and so sees the marginal
product as

MPK = (+) k+-1

So in the competitive equilibrium there will be too little investment in capital.

So this story produces an endogenous growth model if +=1.

But note that given capital's share of income is 1/3, this means that there are really large
externalities that must be present.

Second story for endogenous growth model: Human Capital

We discussed human capital earlier. Now we look at an endogenous growth model based
on it:

12
Lucas Growth Model

The Lucas article on the reading list goes through a model of growth with human capital.

production fn: Y = K(u*h*L)1-

Where L is labor, h is the amount of human capital per person, and u is the fraction of their time
that people spend working.

or in per capita terms

y = k (uh)(1-)

So production is CRS in capital and this quality and time adjusted labor input.

capital has the usual differential equation: k
= y c (n + )k

But now there is also a production function for human capital. To produce human capital, you
have to take time away from producing goods. So (1-u) is the fraction of work time spent
producing human capital,

h
= h(1-u) or alternatively h = (1 u )
h

This says that human capital is produced by a constant returns to scale production function,
where is just some constant. That is, if we double the amount of input into producing human
capital (h*(1-u)), then we double the amount of human capital produced.

[notice that we treat human capital as not being depreciating, and as not being diluted by the
arrival of new people. Thus in this formulation, human capital is really like knowledge; and
creation of new human capital is really like research and development. Allowing for a -(n+)h
term in the equation would not really change things that much.

Lucas goes on to solve this as a social planner's problem using a Hamiltonian. There are two
state variables (h and k) and two control variables (u and c).

For simplicity, we will do the Solow version of the model, in which u and s(=(y-c)/y)
are taken to be exogenous. Recall that the steady state of the Ramsey model looks just like the
steady state of the Solow model in terms of output and saving being constant, and output being a
function of the saving rate (the only difference between the two models being that for the
Ramsey, we can find the saving rate as an explicit function of the taste parameters). Here, the

13
analogue of the steady state is the balanced growth path (see below). The balanced growth
path of the Lucas model looks just like the balanced growth path of the model presented here: u
and s are constant, and the growth rate of output is a function of them. The only differences
between the true lucas model and the one presented here are in the dynamics off of the balanced
growth path, and that in the Lucas model u and s are explicit functions of the taste parameters.

re-write capital accumulation using s instead of c:

k
= sy - (n+)k

so
-1
k
y -1 1- k 1-
= s - (n+ ) = s k (uh ) - (n + ) = s u - (n+ )
k k h

Now we draw a diagram with bothk k
andh h
on the vertical axis and k/h on the
horizontal axis. Theh h
equation is just a horizontal line. The k
/ k equation slopes downward.

What happens to k/h over time? Where the two curves cross, k and h are growing at the
same rate, and so k/h will remain constant. Call this (k/h)*, the balanced growth ratio of k to h.
For lower levels of k/h, k will be growing faster than h, so k/h will be growing over time. For
k/h bigger, vice versa. So whatever level of k/h you start with, eventually it will move to (k/h)*.

What about the growth rate of output? Starting with the production function, we can take
logs and differentiate:

ln(y) = ln(k) + (1-) ln(u) + (1-) ln(h)

y
k
h

= + (1 )
y k h

So in the long run (along the balanced growth path), the growth rates of output, capital,
and human capital are all equal, and are all determined by the h
/ h equation. The lower is u, the
faster is growth along the balanced growth path.

Unlike the Ak model, there are transitional dynamics, which depend on the initial k/h ratio: if
k/h is initially lower than the balanced growth level, then output will initially grow faster than it
will in the long run.

14
 Do the experiment of lowering u from the steady state. What do time series look like?
The level of y initially falls, but the growth rate jumps (although we dont know whether it jumps
up or down), then continues to rise.

-------------------------
Summary of endogenous growth: there are many ways to go, the key is that you have constant
returns to the accumulable (that is, the non-labor) factors. In the externalities model, the factor is
capital. In the Lucas model, h and k are the factors that can be accumulated, and there are
constant returns to the two of them, that is, if you double h and double k, you double output. By
contrast, in the regular model, there is decreasing returns to capital.

So now we know how to write down models in which growth is endogenous. Question: are
they right?

Last time I said that the three strikes against the solow model

-exogenous tech change

-interest rate differentials
-failure of convergence

The first problem is a theoretical one, and I am not going to be able to fix it. Romer book has
good discussion of endogenous technological change. The problem is that since all of the
countries in the world have the same technology, you can't really test models of technological
change (that is, they don't have any implications for differences between countries).

But we can look a little more closely at the other two.

-- interest rate differentials

Here, we can see that the problem is not too hard to fix if we take a page from the endogenous
growth book and consider human capital. Mankiw, Romer, and Weil examine data from a cross
section of countries,

Recall the version of the Solow model with both human and physical capital that we
examined earlier.

Y = K H (eL)1--

We solved for the steady state level of output per efficiency unit:

15

Y ( + )
y = = (n+ g + ) 1- - s k1- - s h1- -
eL

We can put this back in non-efficiency unit terms. Then taking logs:

Y +
ln = ln( A(0)) + gt ln(n + g + ) + ln( sk ) + ln( sh )
L 1 1 1

One can run this regression, and back out values for and (note that MRW also run
other regressions that produce values for and - these will be discussed later).

MRW conclude that, taking human capital into account, a good approximation to the production
function is

Y = A K H L1--

where ==1/3

16
 In this case, H is supposed to mean schooling (which is how we measure it), not
knowledge.

In this case, half of what we were thinking of as labor's share is really human capital's share.
So if we lump all capital together (that is, if the production functions for human and physical
capital are the same), we can just think of capital's share being 2/3. In this case, as shown on the
homework, the interest rate differential problem goes away. Countries that differ in income by a
factor of 5 differ in MPK by a factor of only sqrt(5).

As for the third strike -- the convergence regressions -- let's examine these a little more closely.

What does the Solow model really say about convergence.

There are two reasons countries can have different output -- different inital condtions or
different values of n and s. When we said that we expected convergence, that is only the case
when countries have the same steady state and different initial conditions. If the reverse were
true -- if they had different steady states, but were all in them already, then everyone would have
the same growth rate no matter what their level of income per capita.

So growth is higher when y is below y*

y
/ y = f ( y* y )

So what happens if you consider a regression of on things like the initial level of income, saving
rate, and population growth?

y
/ y = 0 + 1 s + 2 n + 3 y

s and n determine the steady state, and the higher is the steady state, holding the current level of
income constant, the higher is growth. So 1 is positive and 2 is negative.

Says that holding s and n constant (that is, holding y* constant), higher y leads to lower growth,
so 3 is negative.

The result on 3 is called conditional convergence -- controlling for everything else, coefficient
on income is negative.

What about endogenous growth models? In general these say the same thing for 1 and 2, but
they say that 3 is zero.

17
 Go to the data -- solow wins big.

So conditional converge is not only not a strike against Solow, it is a strike against endogenous
growth.

[add: What is the interpretation of the "convergence" regression when we add other stuff to the
RHS -- black market exchange premium, etc. Are these things good for growth, or good for
level].

Knowledge spillovers in the Lucas setup.

Consider the following two-country version of the Lucas growth model.

yi = output per capita in country i (i = 1,2)

ki = capital per capita in country i
n = population growth (assumed to be the same in both countries)
= depreciation (assumed to be the same in both countries)
s = saving rate (same in both countries)
ui = fraction of their time that people spend working in country i
(1-ui) =fraction of their time that people spend building human capital in country i

Production:

yi = ki (uihi)(1-)

capital accumulation:

k
i = syi (n + )ki

So far all of this is like the model presented in class. For human capital accumulation, we
add a new assumption. Let hl be the level of human capital per capita in the country that has a
higher level of human capital (the "leader"), and hf be the level in the country that has less human
capital (the "follower.")

We assume that human capital production in the leader country works exactly as in Lucas'
model:

h
l = (1-ul)hl

18
 Human capital is produced in the follower country by two methods: first, there is the
same production as in the leader country. But second, human capital (which in this model is the
same as "knowledge") spills over from the leader country to the follower country. The amount of
this transfer depends on the difference in their levels of knowledge.

h
f = (1-uf)hf + (hl - hf) >0

Assume that the two countries have the same saving rate, but u1<u2. In other words, in
country 2 people spend a larger fraction of their time working (and a smaller fraction producing
human capital) than in country 1.

A) Describe the steady state of the model. Solve for each country's growth rate in steady state.

Obviously country 2 will be the technology follower. So we can solve for country 1
exactly as if it were alone in the world. Now the key question is what do the country=s growth
rates of h look like? Answer: they must be the same: if not, then either country 2 would catch up
or fall infinitely far behind.
(Figure: put h1 / h2 on the horizontal axis, and the two h dot lines on the vertical)

B) Solve for the relative level of human capital per person in the two countries in steady state.
How do the parameters phi and affect the ratio of h1/h2.

Set the two h
/ h equations equal. Then .....

C) (hard) Solve for the relative level of consumption in the two countries in steady state. How
(and why) do the parameters phi and affect the relative level of consumption.

Extending this idea:

Basu, Feyrer, and Weil. Instead of doing the Solow version of this model, we can do
the Ramsey version, in which we take u as a decision variable. (For simplicity we don=t think
about changing s; in fact, in the paper, we take s as given.)

First, think about the one-country case. Suppose that there is a social planner who cares
about discounted consumption. Then clearly if he has a higher value of , there will be less
R&D, slower growth, but higher initial consumption.

19
 Now, consider the case of a world with two countries. Suppose, for simplicity, that R&D
in one country is fixed. Suppose that I am a social planner, and I was considering doing exactly
the same amount of R&D as the other guy is already doing. My optimal path will now be to do
_less_, and let him be the leader.

It is even possible that I might have wanted to grow faster than him, but if his is fixed, I
will let him be the leader.
(All of this gets into dynamic games, which are too hard for me.... )

Models of Technological Progress

[Note: we do not cover these this year, since Peter Howitt himself is teaching in the second
semester]

Aghion and Howitt

(This is the Abasic model@ from their book)

Production Setup

Production function for final output

y = Ax

where x is an intermediate good. Output is produced by perfectly competitive firms, that take x
as their only input. Good x is in turn produced by a monopolist, who uses labor as his only input.
The production function for x is that one unit of labor produces one unit of the intermediate
good. Ths production function for the intermediate good never changes.

Good x is specific to a technology that is represented by A. The monopolist holds the

patent and faces no competition.

The wage at time t is wt

We can solve for the monopolist=s optimal price, production, and output.

20
 Since the final goods sector is competitive and takes only x as an input, the price of x, px ,
is just its marginal product

px = Ax-1

The monopolist takes this as the demand curve that he faces. His marginal cost is wt. So his
profit, , is

= px x - w x

We can solve this by substituting in the demand curve, and we get the optimal price

px = w /

We can also solve for optimal quantity of x

1
2 A 1-
x=
w

Production of x is the same as the labor used in producing x. Defining (small omega)
as the ratio w/A, we have

x = x ()

which is a decreasing function.

Finally, profit is

-
1 2 w 1-
t = A - 1 1- 0
A

defining (small omega) as the ratio w/A, we have

t = At ( )

21
where is a decreasing function (in the Aghion and Howitt book this function has a tilde on top
of it, but I can=t get my word processor to do that.)

Innovations

We index innovations by t B so note that t is not time. You can think of t as indexing the
generation of technology (or the Aepoch@) that is currently in place.

At is the technology after innovation t. New innovations raise the level of A by a factor :

At+1 = At

Thus the form of technological progress in this economy is new generations of this intermediate
goods and associated new values of A.

An innovation gives a monopoly right to the owner of the innovation until the next innovation.

Labor Market

Workers can produce the intermediate good or do R&D (not that no labor is used in
production of the final good.) Innovations are produced by labor alone.
The labor force is fixed at L. Let n be the number of workers doing R&D. Thus:

L=n+x

Technology Production

The arrival rate for innovations is

n B that is, this is the probability that a new innovation will arrive.

where n is the number of workers doing R&D, and lambda is the probability of a worker coming
up with an innovation. (Note that there is no possibility of simultaneous innovations, so
doubling the workers doubles the rate of arrival of new innovations.)

Firms that are doing R&D use only labor. They are also risk neutral, so their spending on
research has the simple arbitrage condition

wt = t Vt+1

22
 We take the interest rate r as exogenous and constant. What is the value of holding an
innovation? Call t+1 the profits that are earned by a firm that holds patent t+1. These are
enjoyed for as long as you hold the monopoly. Since t+1 nt+1 is the rate at which new
innovations will arrive, this gets factored into the value

t+1
V t+1 =
r + n t+1

(Could derive this starting with the PDV integral)

Solution

We can take the arbitrage condition, substitute in for Vt+1 and for t+1 to get

At+1 ( t+1 )
wt =
r + nt+1

Dividing both sides of this equation by At gives

( t+1 )
t = (1)
r + nt+1

we can also rearrange the labor market equation to be

x( t ) = L = nt (2)

Notice that in the first of these equations, t depends on future values of and n B the reason is
that the value of doing R&D today depends on how long you expect to hold onto your monopoly
in the future.

We examine only steady states, where and n are constant (Aghion and Howitt consider
out of steady state stuff, I think.) So equations 1 and 2 become two equations in two unknowns.
We can graph them, with n on the horizontal axis and on the vertical. There is always an
intersection at some value of n less than L. This is the equilibrium.

Notice that in equilibrium n will be constant, but growth will still be stochastic, since the
time to a new innovation is stochastic.

23
Comparative statics:
B increase L: this will shift the curve for (2) outward: higher value of n
B lower R: This will shift the curve for (1) upward: higher value for n (and )

Technological Progress Through the Creation of New Intermediate Goods

Here is a second model of technological progress: in this case, progress takes place
through the creation of new intermediate goods. Model is that of Grossman and Helpman. This
presentation based very heavily on Barro and Sala-i-Martin.

Setup of the model:

- no capital
- constant quantity of labor: L
- Y is output, produced with labor and intermediate goods, X
- N is the number of intermediate goods.

Intermediate goods are not capital -- they are measured as

flows (produced using output, used in producing output).

i indexes firms
j indexes intermediate goods (j=1...N)

production function for a firm producing final output:

N
Y i L i ( X i, j ) = L i X i,1 + L i X i,2 + ...
1- 1- 1-
=
j=1

note:
-- use of each intermediate good does not affect the productivity of the others.
-- there is decreasing marginal productivity of each intermedate good [picture]
-- So: the firm will use all intermediate goods in the same quantity.

Call Xi the quantity of each intermediate good used by firm i

24
 1- 1- -1
Y i = L i N X i = L i (N X i ) N

note that NXi is the total input of intermediate goods.

so there is CRS to inputs of L and intermediate goods (ie L and NXi). But raising N
(holding NXi constant -- that is, not changing the total quantity of intermediate goods) raises
output. The reason can be seen by thinking about the productivity of a given intermediate good
in the firm's production function: Bigger N allows the firm to use less of each one, and this raises
the average product for the part of production that used that intermediate good.

We assume that the firms that produce final output (Y) using labor and intermediate
goods are perfect competitors -- they have CRS and will earn zero profits (that is, euler's theorem
will hold).

What are their demands for inputs?

N
profit i = Y i - w L i - P j X i, j
j=1

d profit i
= L 1-i X i, j-1 - P j = 0
d X i, j

so the firm's demand is

1
1-
X i, j = L i
Pj

and economy-wide demand is

1 1
1-
L
1-
X j = X i, j = L i =
i Pj i Pj

Technological change in this model takes place via the expansion in the number of
intermediate goods.

25
 This is supposed to be a metaphor for technological progress more generally -- in real life
some technological progress takes place via the creation of new final goods, and some takes
place through the replacement of old intermediate goods with new ones, and some takes place
through other improvements in the production process. The point is that we want a way to model
technological progress where inventors earn back the cost of doing inventing, and this model of
intermediate goods supplies such a model.

So: inventors create new intermediate goods in order to reap future profits -- otherwise
they wouldn't do it. Note the special nature of creating new technologies (as opposed to creating
goods or services): technologies are "non-rival." If I figure out a new way to do some task, your
using it does not diminish my ability to use it. The non-rivalry of technology means that there is
a big social benefit to producing new technologies, but it also means that there is a potential
incentive problem -- in the absence of well-developed intellectual property rights, there will be
no incentive to develop new technologies. Patent protection is an example of how we set up
rules to give incentives for technology development.

So in our model, we assume that the inventor of a new intermediate good gets a perpetual
monopoly on the right to produce it in perpetuity.

The cost of producing a new invention is assumed to be a constant, eta.

Note two ways in which this assumption might be made more realistic. First, it might be
better to assume that there is some stochastic element in the process of inventing stuff. Second,
it might be reasonable to argue that there aren't constant returns to scale in invention -- if you
double R&D spending you don't double results. (See Aghion and Howitt for a model that fixes at
least problem number 1 -- maybe number 2 also).

What is the value of having invented a new good?

Pj is the price. We assume that the cost of producing one unit of the intermediate good is
one. so

V(t) = ( P j - 1) X j e -r(v-t) dv
t

where we are assuming that the interest rate is constant -- see Barro and Sala-i-Martin for the
demonstration that it is.

The monopolist chooses X and P to maximize profit, subject to the demand curve derived
earlier.

26
 1
1- -1
profit = ( P j - 1) X j = ( P j - 1)L = ( P j - 1)L 1/1- P 1-j
Pj

1
-1
d profit -1 - 2-
= L 1- P 1-j + ( P j - 1) P j1- = 0
d Pj 1-

which solves to Pj = 1/

so Xj = L 2/(1-)

Note that the price is the same for all j, and similarly Xj will be the same for all j (but
only for those that exist at a given time).

1 2
V(t) = ( P j - 1) X j e -r(v-t) dv = ( - 1)L 1- e -r(v-t)dv
t
t

1 2 1
V(t) = - 1 L 1-
r

We assume that there is free entry (and thus zero profits) in getting into the business of
being an inventor/monopolist. That is, once you get an invention, it will earn you monopoly
rents -- but the PDV of these rents will be exactly equal to the cost of getting into the business,
that is, to eta. So setting V(t)=eta, we solve to get:

L 1- 2
r = 1-

So the cost of invention (and the other technological parameters) pin down the interest
rate.

What is going on here? There is no capital, so the only "investment" that you can make is
creating a new intermediate good; the value of this is determined by the technology of inventing.

27
 Note that we are assuming that some inventing is being done -- another way to put this
point is that there is infinitely elastic demand for borrowing at the interest rate given above. Of
course, if the interest rate is higher than this, then no inventing will be being done, and so this
equation for the interest rate will not hold.

Note that smaller eta (inventing is easier) implies rate of return is higher.

Households

We assume that households have CRRA preferences, with at the coefficient of relative
risk aversion, and as the time discount rate. So we know the rate of consumption growth
c
1
= (r - )
c

so given r,

c
1 L 1- 2
= 1- -
c

Note that if this led to negative consumption growth, there would be a problem. Since
there is no capital stock to run down, consumption can't fall. So we assume the parameters are
such that consumption growth is positive.

Aggregate output is

Y = Y i = L 1-i N X i
i i

[What is going on with this next line??? fix it.]

L (L )
2-
1-
=N i 1- = N L 1-
i

soY /Y =N /N

note that we are fudging and making N continuous. We could have done this from the beginning
by writing the firm-level production function as

28
 N

Y i = L i ( X i(j) ) dj
1-

Now we just need to show that the growth rates of N and Y have to be equal to the
growth rate of c.

From the economy-wide budget constraint we know that consumption has to be equal to
output less spending on intermediate goods and spending on new inventions:

C = Y - NX - N

divide all of the terms in this by Y

C N N
N N Y

= 1 - X - = 1 - X -
Y Y Y Y Y Y

where for the last step we have multiplied and divided by N, and substituted /Y = /N.

We know from above that N/Y is constant; we also know that X (the quantity of each
intermediate good used) is constant.

C N

- - x +1
Y
=
Y Y
Y N

Y

So suppose that the growth rate of output is not equal to the growth rate of consumption
at one point in time, it will diverge from it permanently (ie if oiutput growth is higher than
consumption growth, then C/Y will fall, and output growth will rise). Eventually, this will
violate some implicit transversality condition.

Extensions:

note that we have size effects, which are not present empirically. This suggests two things:

-- 1) we should think of a model where technology crosses borders

29
 2) for the world, might be right that there are size effects, but it depedns on how cost of new
technologies changes as you try to develop them more quickly. (ie question of constant returns to
R&D , raised above).

Note also that we haven't said anything about optimality -- it turns out that there is too
little invention (and the prices of intermediate goods are too high) compared to the social
planner's outcome.

[End of Material Not Covered when Howitt is Teaching]

--------------------------------------------------------
Malthus: do simple graphical version.

Lucas' version of the Malthus/Ricardo population model

Simple demographic structure. Individuals live in dynasties. There is one economically

active period, in which individuals work, consume, and raise children. For simple version of the
model, we assume that all members of a generation are identical.

ct = consumption per person

nt = children per person

Production:

X= quantity of land
L = number of families = number of workers

in the simple version of the model there is no capital. Land is the only productive
resource.

production is CRS: Y = F(X,L) = A X L1-

define x = X/L (land per worker)

y = output per worker

y = A x

We assume that land is not owned privately, so it earns no return. Individuals earn their
average product -- that is, y. (Can think of this as a pre-enclosure grazing economy, or a hunting
and gathering society, or fishing, or any other common resource. Note -- this characterization of

30
people in agriculture earning their average (not marginal) product crops up in a lot of
development literature. Later in the paper Lucas considers the case where land is privately
owned).

children cost k each to raise (no worry about integer constraints). [Think about in a more
advanced model: probably not a good idea to make the cost of children fixed in terms of goods.
The real cost of children is in _time_, which means that as the wage rises, the cost of children
goes up].

So the budget constraint is

c + kn = y

What about utility:

The general formulation is to give people utility over their own consumption, the number
of kids, and utility per kid. We specialize this to be additively separable.

u t = W( c t , n t ,u t+1 ) = ln( c t ) + ln( n t ) + u t+1

Notice that in this version of the model there is nothing that the parent can do to affect the
utility of the kids. (That is, the parent can't give the child land (which is not owned) or capital
(which does not exist). The fact that the kids utility enters the parent's utility in a separable
fashion means that it does not affect the marginal utility of anything else, and so it will not have
any effect on the parent's decisions. So it is not really doing anything in this version of the
model.

We can maximize utility subject to the budget constraint, and get the following first order
condition:

nt =
ct k

We can actually learn a lot just from this. We know that in the steady state, x (which is
land per worker) will be constant. So L must be constant, and so in steady state, n=1. So we can
solve for the steady state level of consumption directly:

k
c ss =

Interesting things to note about this: steady state consumption doesn't depend on X, A, ,

31
etc. It just depends on preference parameters and the cost of kids. The more that people value
children, and the less that people value their own consumption, the lower is steady state
consumption. The more expensive are children, the higher is steady state consumption.

What about the model's dynamics?

the evolution of land per worker (x) is given by

X X x
x t+1 = = = t
L t +1 n t Lt n t

Combining the FOC from utility maximization with the budget constraint, we can solve
for nt in terms of yt, and then substitute in the production function

y t A xt
nt = =
k + k +

Substituting this into the equation for the evolution of x:

1- k +
x t +1 = x t
A

We can graph xt+1 against xt, and include a 45 degree line, to see that this leads to a single
stable equilibrium.

We can also use this equation to solve for the steady state values of x, and therefore of y

1/
k +
x ss =
A

+
y ss = k

Comparing this with steady state consumption, we can see that the steady state ratio of
consumption to output is /( + ) , which corresponds to their weight in the relevant part of the
cobb-douglass utility function.

32
 Note that the equation for steady state output says that it is higher, the more expensive are
kids.

So, like consumption, output per worker does not depend of the level of technology (A),
or the amount of land. What about the total number of people, L?

Using the equation for xss

-1

1/ +
L ss = XA k

So the number of people is proportional to the amount of land -- no big surprise. Further,
if there is better technology, you get more people per unit of land (that is, higher population
density), but you get no higher per capita output or consumption. This is the Malthusian deal:
population grows to run resources down.

In the rest of the paper, Lucas goes on to consider more sophisticated versions of the
model -- with capital and private property, for example.

Does this characterize the world?

Lucas argues that it is a good characterization of the world until recently: Prior to
roughly 1700, countries seemed to be in Malthusian steady states. Income per capita was roughly
constant at the subsistence level. Better technologies meant more people, but not higher living
standards.

Then around 1750 (in the west), something changed: income per capita started rising, and
the gains in income did not immediately result in higher population.

Phases of Economic Growth in Europe

Pop- GDP GDP

ula- per
tion capita

Agrarianism 0.1 0.0 0.1

500-1500

33
Advancing Agrarianism 0.2 0.1 0.3
1500-1700

Merchant Capitalism 0.4 0.2 0.6

1700-1820

Capitalism 0.9 1.6 2.5

1820-

Source: Angus Maddison, Phases of Capitalist Development

Explaining what changed is really hard. It is easy to have a Malthusian model with stable
population -- but once you get away from the resource constraint, it is very hard to have a model
that produces constant population....

Notice that in the Malthusian regime, we can read the progress of technology directly
from the population density. So the more than doubling of popualtion in Europe over the period
500-1500 means that there was a lot of technological improvement, even though living standards
did not change.

Also explains why in, say, 1950, China had the same income per capita as Africa, but
much higher technology -- it also had less land per person.

Kremer's Model of Population and Technology

We now turn to a related model that integrates population and technology change. The
setup of the model is simple. Output is produced with land and labor -- no capital (capital could
easily be incorporated, and it would not make a difference as long as the production function was
such that there was decreasing return to labor and capital while land was held fixed -- in other
words as long as there was constant returns to land, labor, and capital together).

Y = X ( AL)1

It is assumed that the growth rate of A (knowledge) is proportional to the size of the
population. The argument here is that people are equally likely to come up with new productive
ideas, so that the more people you have, the more ideas you will come up with. [it is also
assumed that the productivity of each new idea is proportional to the current level of
productivity]

34
A
/ A = BL

finally, there is a Malthusian model of population. Population adjusts to keep output per person
at the subsistence level (this adjustment is not modeled, but you can think of a Lucas-style model
in the background).
Y /L= y= y

Given the assumption that income is at subsistence and the production function, we can
solve for the level of population as a function of the state of technology and the quantity of land:

yL = X [AL ] 1-

1/
1 1-
L= A X
y

So once again population is proportional to land, and we can read the level of technology
off from the level of population density.

Since y and X are constant, the growth rate of population is related to the growth rate of
technology:

L
1- A
1-
= = BL
L A
So the growth rate of population is proportional to the level of population. That is, the
growth rate of population is increasing over time.

This seems like a pretty silly prediction, but when he looked at data it fit pretty well.
Specifically, he plots population against the growth rate of population from 1 million BC to the
present, and it fits as an almost straight line (see figure in Romer book). It only falls apart in the
last few decades.

Of course we know from the discussion of Lucas above that the Malthusian model stops
fitting in the developed countries a while ago...

He also cites evidence about the relative growth rates of different isolated populations...

To be added: incorporating resources in the production function and estimating share and
elasticity of substitution.

35
 Problems

1) Consider the Constant Elasticity of Substitution (CES) production function:

Y = [aK- + L-]-(1/) where > -1

A. Show that this production function has constant returns to scale.

B. Calculate the marginal products of capital and labor.

C. Calculate the elasticity of substitution of capital for labor.

D. Calculate the fraction of total output that goes to capital. For what values of the elasticity of
substitution is it the case that as the capital/labor ratio rises, the fraction of total output that goes
to capital rises? For what values does it fall? For what values is it constant?

2) In a country, the production function is y=k1/2 , where y is output per worker and k is
capital per worker. Suppose that in 1990, k is equal to 400. The fraction of output saved is 50%.
The depreciation rate is 5%. Assume that there is no population growth or technological change.

A. Is the country at its steady state level of output per capita, above the steady state, or below the
steady state? Show how you reached your conclusion.

B. Now suppose that there is a large immigration to the country. The population of the country
quadruples. The new immigrants do not bring any capital with them. Following the
immigration, is output per capita above, below, or at its steady state level?

3) Consider a Solow model, where the production function is

Y = A KL1-

Population grows at rate n and capital depreciates at rate . There is no technological change.

There are two countries (both of which are closed to capital flows) that are both at their steady
states. They have the same levels of y, n, and . The value of A in country 2 is twice as large as

36
in country 1. What is the ratio of their saving rates?

4) (old final exam question) Consider a country with production function (in per capita
terms) y=k (0<<1). Population grows at rate n and capital depreciates at rate . There is no
technological change. Consumption is equal to a constant fraction, c, of output. In addition,
every period a payment in the amount p per capita must be made to the foreign power which
provides protection for this country. All output that is not consumed or paid to the foreign
country is invested.

A. Write down the differential equation governing the evolution of the per capita stock of
capital in this country.

B. Draw the usual Solow model diagram for this country. Is there more than one equilibrium
level of the capital stock and of output? If so, identify all of the equilibria. Indicate on your
diagram which equilibrium the economy will move to depending on its initial stock of capital per
worker.

5) Consider two countries, A and B, which are described by the Solow model. In both
countries, the production function (in per capita terms) is y=k.5, and the rate of depreciation () is
.02. There is also no technological change. In A, the saving rate (s) is .2 and the population
growth rate (n) is zero. In B, the saving rate is .4 and the population growth rate is .02.

Both countries are initially observed to have income per capita of 5 at time zero. Draw a
graph with time on the horizontal axis and income per capita on the vertical axis showing how
the level of income per capita in the two countries will evolve over time. Explain.

6) A country described by the Solow model has a production function y=k.5, and some rates
of population growth (n) and depreciation (). There is no technological change. Currently, the
saving rate in the country is .6, and the country is at the steady state level of output per capita.
Two courses of action are proposed: One is to lower the saving rate to .5, and the other is to
lower the saving rate to .4. Graph the paths of consumption per capita over time following the
introduction of the new policy, and show how these levels of consumption compare to the level
of consumption per capita in the old steady state. Explain what is going on.

7) Consider a Solow model with positive rates of population growth, depreciation, and
technological change. Imagine that a country is in steady state, and that suddenly its rate of
technological change increases. Describe how output per efficiency unit and the growth rate of
output per capita evolve.

8) Assume a Solow model in which two countries have the same savings rates and the the
same values of n, , and g, but differ in the state of technology.

37
 ym= Amkm yn= Ankn

Output in country m is five time higher than in country n. Both countries are in steady state.

A. By what factor do the two countries' capital stocks differ?

B. By what factor do the two countries' values of A differ?
C. By what factor do the two countries' MPK differ?

9) A country has the following production function:

Y = A(bK)(cL)1- 0<<1

the growth rate of A is gA, the growth rate of b is gb, and the growth rate of c is gc. Population
growth is zero. Calculate the growth rate of output per capita in the steady state of a Solow
model.

10) Consider a Solow model in which the rate of growth of population is endogenous.
Specifically, the rate of population growth is nh when y<y* and is nl when y$y*. nl<nh. The rest
of the model is standard. Production is Cobb-Douglas: y = k. Capital depreciates at some rate
. There is no technological change. The saving rate is exogenous.

A. Draw pictures showing the different possible configurations of equilibria (ie a single, globally
stable equilibrium vs multiple equilibria) in the model.

B. For what range of values of the saving rate will it be the case that the model displays
multiple equilibria? For what values will there be a single equilibrium?

11.5) [midterm 2006] Consider a Solow economy with no technological progress. The
economy is closed to international capital flows. There is some subsistence constraint that
people must meet prior to doing any savings. Specifically, saving is equal to a constant fraction
of income above the subsistence level ysub.

A) Write down the equation of motion for capital

B) Draw all the possible configurations of the model. Are there multiple steady states, one
steady state, no steady states? For each configuration, discuss the stability of the steady state(s)
that exist.

C) You observe that in this county, at time zero, capital exogenously jumps (say the World Bank
decides to build them a dam) from k0 to k1 and the growth rate of output per person jumps
upwards. Then at time 1, you observe another exogenous increase in capital, from k1 to k2, but
this time the growth rate of output per person falls. Describe how this pattern could arise in this
model.

38
11) Consider two countries described by the Solow model.

They both have the same saving rates, population growth rates, and depreciation rates.
There is no technological change. In each country the production function is (in per capita terms):

y = k 0<<1

In country 1 there is a proportional tax at rate . The proceeds of the tax are thrown into
the sea, and do not have any effect on the economy. Saving is done out of post-tax income.

In country 2, there is a constant per-capita tax (ie a head tax) of per person. As in
country one, the proceeds of the tax are thrown into the sea. Saving is done out of post-tax
income.

The per-person tax in country 2, , is set to be equal to taxes collected per person in
country 1 when country 1 is in steady state.

A. Solve for .

B. Describe the dynamics of the two economies (how they evolve from different initial levels of
capital, etc.). Show that there different possible configurations of steady states in country two
relative to country one.

C. Show how the specific configuration of steady states in country 2 relative to country one
depends on the parameters. Derive the conditions under which the different possible cases will
appear.

12) Consider a Solow model in which the production function (in per capita terms) is

y=k1/3h1/3,

where k is physical capital per worker and h is human capital per worker. Physical capital is
accumulated according to

k
= sk y - (n+)k

and human capital is accumulated according to

h
= sh y - (n+)h

where n is population growth, is depreciation, and sk and sh are the rates of saving in physical

39
and human capital.

Two countries have the same levels of sk, but differ by factor of two in their levels of sh.
By what factor will their steady state levels of output differ?

40
13) Country one is exactly described by the standard Solow model (with no technological
progress):

y = k

k
= sy (n + )k

The values of the parameters are: = .5, (n+) = .1, s = .1

Country 2 is the same as country 1, except that the saving rate is a function of the capital stock:

1
s = s0
1+ k
Where s0 = .2.

A. Show that the two countries have the same steady state levels of output. Intuitively, which
country should move more rapidly toward its steady state?

B. Linearize around the steady state to get an expression for output growth of the form:
y

= - ( ln(y) - ln( y* ) )
y

How does the value of compare in the two cases?

14) s=.2; n=.02; g=.02; =.03; = 1/3. Use the linearized form of the simple Solow model
to calculate how quickly output per capita should be growing in a country with income per
efficiency unit equal to half its steady state level. Do the same for a country with income per
efficiency unit equal to one quarter of its steady state level. Calculate the time that the country
will take to get from one quarter its steady state level of output per efficiency unit to one half its
steady state level.

To test the quality of this linearization: use a calculator, computer, etc. to figure out the
actual rate of output growth for the two countries examined in question 1. Finally (this requires a
little more work), calculate the time that the country will take to get from one quarter to one half
the steady state level of income per efficiency unit.

16) Consider a world in which there are two countries with equal sized labor forces. The

41
countries have the same production function. In per-worker terms:

y = k 0<<1

there is no technological progress. The depreciation rate is zero. In each country, factors are
paid their marginal products. The growth rates of the labor force in the two countries are the
same and are greater than zero:

n1 = n2 = n >0

Capital flows freely between the two countries to equalize the marginal product of
capital.

In each country, savings takes place only out of capital income. In other words, all
wages earned by residents of the country are consumed. Of payments to capital owned by people
who live in country i, a fraction si is saved and (1-si) is consumed. Notice that this saving rate
applies to income paid to capital owned by people who live in country i, whether or not that
capital is located in country i.

The saving rate out of capital income is higher for people in country 2 than country 1, in
other words,

s2 > s1

Define a1 and a2 as the levels of assets per worker in countries 1 and 2, and k1 and k2 as
the levels of capital per worker in each country.

Solve for the steady state levels of capital per worker and assets per worker in each
country.

17) Consider a Ramsey model, with depreciation , population growth n, time discount rate
, and production function (in per capita terms) f(k) = k.5. Solve for the steady state level of
consumption per capita in terms of the three parameters.

17.5) (Midterm exam, 2003) Consider a Ramsey model with a social planner, where instead
of time going to infinity, it is know from time zero that the world will end at time T. You should
assume that the capital stock at time zero is well below the steady state that would apply if time
were infinite. You should also assume that T is a long way in the future. Draw the usual
diagram for the Ramsey model, including the stable arm that would be appropriate if time went
to infinity. On this diagram, show the paths of consumption and capital stock. Also draw graphs
of consumption and capital with time along the horizontal axis. Explain how you got your
results.

42
18) Consider two economies described by the Ramsey model, which are the same in every
respect (ie population growth rate; time discount rate, etc.) except their discount rates: People in
country A discount the future more that people in country B. Assume that both countries start off
the same capital stock k0 which is below either coutnry's steady state level. Which country will
have higher initial consumption? Is it possible for the two countries' stable arms to cross?

18.5) [midterm exam, 2005] Consider a Ramsey model in which the social planner weighs
future consumption according to the number of people who will be alive in the following
manner. The social planner wants to maximize

V = U (c(t )) N (t ) e - t dt
0

where 0 < < 1 . Population grows exogenously at rate n>0. The rest of the problem is
standard. There is no technological progress. The deprecation rate is zero. The production
function is

y = k

and the differential equation for capital stock is

k
= f (k ) c nk .

A) What assumption regarding the values of n, , and is required to ensure that the social
planners utility is finite?

B) Making this assumption, solve for the steady state level of capital per worker.

19) (midterm, 2001) Consider two economies described by the Ramsey model. The
economies have the same production functions and rates of population growth. Both of them are
also closed to capital flows from the rest of the world.

The countries differ in the preferences of their social planners, that is, in , the time
discount rate, and , the coefficient of relative risk aversion (both social planners have CRRA
utility.) We know that 1 > 2 . We do not know how their values of compare (but we will
figure it out.)

43
 The time-zero values of capital per worker, k0, are equal in the two countries. Also, by
coincidence, the optimal values of time-zero consumption chosen by the social planners, c0, are
also equal.

Figure out which country has a higher value of . Also draw a graph with time on the
horizontal axis and consumption on the vertical axis, and show how the time paths of
consumption in the two countries compare.

19.5) [midterm exam, 2005] A country has production function

Y = Ak

Time is continuous. Population is constant. There is some depreciation rate > 0 .

Consumption decisions are made by a social planner who has instantaneous utility of the CARA
form:
1
u ( c ) = e c

The social planner has a time discount rate > 0 . You should assume that A > .

Reminder: the continuous time FOC for an individual with CARA utility facing an interest rate r
is

1
c
= (r - )

A. Write down the differential equations for k
and c
(not k
/ k or c
/ c ).

B. Draw a diagram with k on the horizontal axis and c on the vertical axis. Draw arrows
indicating the dynamics of c and k.

C. Suppose that there is some initial capital stock k(0). Add to the diagram from part (B)
some representative paths the economy would follow based on different choices of initial
consumption. Indicate which path the social planner will choose.

D. Suppose that there are two countries run by social planners with the same preferences.
The two countries have the same value of A. The countries differ only in their initial capital
y
stocks. Suppose that at time zero, 1 = 2. Toward what value will the ratio of incomes in the
y2
two countries asymptote?

44
E. Solve for initial consumption, c(0), as a function of initial capital stock, k(0).

20) Consider a Ramsey model with government spending. The differential equation for the
evolution of capital is

k = f(k) - c - nk - G

There are two levels of government spending, Gh > Gl. Spending alternates between
these two levels in a fixed cycle. Use the phase diagram to describe how consumption and
capital will behave once the economy has been in this pattern for a long time. Draw time series
pictures of how the interest rate behaves over the political cycle. Consider the limiting cases as
the period of alternation becomes very short or as it becomes very long. (hint: thinking about
very long periods of alternation is the easiest way to approach the problem).

21) (final, 2001) Consider a typical Ramsey model in which the social planner
maximizes the discounted integral of future per-capita utility. Population grows are rate n. The
time discount rate is . The equation for the accumulation of capital is

k
= f(k) - c - nk

The production function is

y = A k

Prior to time t, the value of A is constant and equal to A1, and the economy is in steady
state. At time t, it is announced that at some future date, time s, the value of A will jump from
A1 to A2, where A1>A2. After this, A will remain constant.

Consider the different possible relations between the position of the stable arm associated
with the new steady state and the initial steady state (there are three possibilities.) For each of
these cases, draw a diagram with time on the horizontal axis and the marginal product of capital
on the vertical axis, and show how MPK will change over time. In each case, show where r is
rising, falling, jumping, etc. Also show how the level of r in the new steady state compares to the
initial value.

22) [core exam, 2003] In a certain country, consumption and investment decisions are made by
a social planner who maximizes

45


ln(c) e
- t
dt
0

where c is consumption per capita, and > 0 is the time discount rate. The population (which is
equal to the labor force) grows exogenously at rate n>0.

There are two kinds of capital used in production. The quantities per worker are denoted
k1 and k2. Once capital has been created, it cannot be turned back into output. Similarly, one
type of capital cannot be turned into the other. Call i1 the per worker quantity of investment in
type 1 capital and i2 the per worker quantity of investment in type 2 capital.

The social planner=s budget constraint is

y = c + i1 + i2

The per-worker production function is

y = k1 k2 <2

The deprecation rate is zero.

A. Solve for the steady state levels of c, k1, and k2. [Note, there are some slightly ugly algebraic
expressions that crop up here and in part C. You do not have to simplify too much, as long as I
can see what you doing.]

Now, suppose that the economy is in steady state when suddenly an asteroid collides with
the earth, destroying half of the type 2 capital. None of the type 1 capital is destroyed.

B.Draw a picture with the quantity of k1 on the horizontal axis and the quantity of k2 on the
vertical axis. Draw a point indicating the steady state. Now sketch on the figure the time path of
(k1 , k2) following the collision with the asteroid. Also describe (in words) how investment
behaves during this transition.

C.Solve for the growth rate of consumption immediately following the collision with the
asteroid.

23) A country is initially closed to the international capital market. Production and
consumption are described by the Ramsey model. The interest rate is ra (a for autarky).

46
 The world interest rate, r*, is greater than the closed-economy interest rate: r*>ra. The
economy is small, so that once it opens to the world economy it will be a small open economy.

Assume that the opening to the world economy is a surprise. Prior to it, the economy is
in steady state.

Describe the time paths of consumption, output, and net foreign assets at the time of the
opening and subsequently. Be sure to indicate which variables do or do not jump at the time that
the economy is opened up. [Note: figuring out what happens to consumption initially may not be
possible, but say what you can about it]

23.25) [midterm, 2006] Consider the country of Microland, which is described by the Ramsey
model. There is no technological change. Population grows at rate n>0. The deprecation rate is
zero. Consumption and investment decisions are made by a social planner with a time discount
rate theta>0. The production function has the usual properties.

The economy of Microland is small relative to the rest of the world. The rest of the world
is also described by the Ramsey model, with the same production function, population growth
rate, and social planner preferences as Microland. The rest of the world is in steady state.

[Note: you can talk your way through these answers; you dont have to do any math at all]

A) Suppose that up to time t, Microland has been closed to the world economy, and also
expected to remain closed. At time t, Microland opens up to capital flows from the rest
of the world. Also, assume that just prior to time t, the level of capital per worker in
Microland was below the steady state level.

Graph the paths of GDP per capita and consumption per capita and in Microland. Also show
how consumption per capita in Microland compares to consumption per capita in the rest of
the world, and also to what the path of consumption would have looked like if the economy
had not opened to the rest of the world. On your figure, you should note the important
features of the paths, and explain where they come from.

B) Suppose that the setup is the same as in part A, but that at time t the economy is not
actually opened up. Instead, at time t, it is announced that at some later date, s, the
economy will open to capital flows from the rest of the world. Sketch the paths of GDP
per capita and consumption in Microland. Show how the path of consumption compares
to that in part (A), and also show how the path of consumption compares to what would
have happened if the economy had not opened to the rest of the world. Show how the
path of GDP per capita compares to that in part (A) and to what would have happened if
the economy did not open to foreign capital flows.

As in part (A), you should note the important features of the paths you sketch out, and

47
 explain where they come from.

23.5) [midterm exam, 2004] Consider a simple OLG model in which people supply one unit of
labor in each period of their lives. They are born and die with zero assets. Assume production is
Cobb-Douglas with = 2. Also assume that capital fully depreciates after it is used, so that
old people consume only the marginal product of their capital, not the capital itself. People have
log utility with a time discount rate of zero. Population is constant.

Solve for the steady state level of output per worker. (Remember, both old and young
people work).

23.75) [final exam, 2005] Consider the following OLG economy. There is no population
growth or technological change. Individuals work only in the first period of life. Individuals
have Leontief lifetime utility of the form

u = min(c1 , c2 )

Firms have per-worker production function

y = k

1
Where < . Factors are paid their marginal products. The depreciation rate is > 0.
2

Solve for the steady state level of capital per worker.

24) (old midterm question) Consider an open economy populated by OLG people. There is
no depreciation, time discounting, technological change, or population growth.

Utility is: U = log(c1) + log(c2)

The world interest rate, r* is taken as exogenous by the country. People can either borrow or
lend to the rest of world at interest rate r*.

The production function is f(k) = k

Calculate the steady state levels of capital, the wage, saving, and wealth in the country as a
function of r*. What determines whether, in steady state, the country has positive or negative net

48
exports?

24.5) [midterm exam, 2005] Consider two economies described by the standard OLG model.
There is a standard production function and no population growth. In country 1, people have
time discount rate 1 and log utility. In country 2, people have time discount rate 2 and CRRA
utility with coefficient of relative risk aversion > 1 . The two countries have the same steady
state levels of capital and output per person.

Draw a diagram with kt on the horizontal axis, kt+1 on the vertical axis, and a 45-degree
line. On this diagram, show how the functions kt +1 = (kt ) for the two countries relate to each
other. Which country will move more rapidly toward its steady state. Explain fully how you
know what the picture looks like.

25) [core exam, 2001] Consider the following Overlapping Generations economy: People
live for two periods. The lifetime utility of an individual born at time t is given by

Ut = ln(c1,t - x) + ln( c2,t+1) x>0

where c1,t is consumption in the first period of life of a person born at time t, and c2,t+1 is
consumption of the same person in the second period of life. x is an exogenous constant.

Other than having x in the utility function in the first period of life, the model is
completely standard. People work in the first period of life, inelastically supplying one unit of
labor. They do not work in the second period of life. The savings of the current elderly
generation make up the capital stock used in production. Population size is constant. Factors are
paid their marginal products. The rate of depreciation is zero. The production function, in per
worker terms, is

y= k

A. Draw a diagram showing how capital evolves from period to period (i.e. a diagram with
kt on the horizontal axis and kt+1 on the vertical axis.) Indicate any possible steady states and
discuss their stability. Show how the number of steady states and their stability is affected by the
size of x.

B. Now suppose that the value of x is endogenous. In particular, assume that x at time t (which
applies only to the young people at time t) is based on the consumption of old people at time t:

xt = c2, t

49
Discuss the conditions under which a stable steady state will exist in this economy.

25.5) [midterm 2006] Consider an economy made up of overlapping generations. Individuals

live for two periods and supply labor only when young. They have the following lifetime utility
function

U = ln(c1 ) + ln(c2 )

Individuals choose how much to save for period 2 after they receive their period 1 wages. The
money they save is locked in long-term saving and cannot be accessed until period 2. There is a
government welfare program that gives individuals in the first period of life (but not the second)
a guaranteed minimum consumption level of c . So if the money left over after individuals have
put aside their long term saving is less c , then the government will give them enough money to
consume c in that period. (Unlike real-life welfare programs, the government will give money
to people even if those people have put aside lots of money to consume in the second period of
life.) We dont worry about where the government gets the money to support this transfer
scheme.

The production function in per-worker terms is

y = k 1/ 2

The growth rate of population is zero. Factors are paid their marginal products. Capital fully
depreciates after it is used in production. This means that the elderly consume the earnings (i.e.
the marginal product) of the capital they saved when young, but not the capital itself.

Our goal is to analyze the dynamics of the capital stock, using the usual diagram with kt on the
horizontal axis and kt+1 on the vertical axis. But we will build up to this in steps.

A. [5 points] Describe informally the saving behavior implied by the utility function.

B. [5 points] Calculate the level of the first period wage that corresponds to the critical
point in your description in part A.

C. [5 points] Find the level of capital corresponding to the first period wage you found in part
B.

D. [5 points] Draw the graph of kt+1 as a function of kt.

E. [5 points] For what range of values of c is there not a stable steady state in which the saving
rate is constant.

50
26) Consider an OLG model in which there is capital mobility. There are two countries in the
world, with equal populations. There is no population growth, technological progress, or
depreciation. The production functions in the two countries are the same: Y = K.5L.5. In each
country, people work in the first period of life only, and consume in both the first and second
periods of life. Their utility functions are:

Country 1: U = (1-) ln(cy) + ln(co)

Country 2: U = (1-) ln(cy) + ln(co)

where cy is consumption when young, co is consumption when old, and 1>>>0.

A. Solve for the steady-state level of GDP per worker in each country.

B. Solve for the steady-state level of GNP per worker in each country.

27) Consider the following variation on the OLG model presented in class. Individuals live
for two periods. In the first period of life they work, supplying one unit of labor. They do not do
any consumption during the first period of life. In the second period of life, they do not work,
and they consume their savings from the first period along with any accrued interest. The
economy is closed.

The production function, in per-worker terms, is y=k. The depreciation rate is zero. The
population grows at rate n -- that is, each generation is (1+n) times as big as the one that came
before.

A.Solve for the steady-state level of capital per worker.

B. Assume that is equal to one-third. For what value of n will this economy be at the Golden
Rule level of the capital stock?

27.5) [midterm exam, 2004] Consider an OLG model in which people live for two periods. In
the first period, they work, raise children, and consume. In the second period, they consume.
Their utility functions are

V = ln(c1) + ln(c2) + ln(n)

where n is the number of children. The price of children (in terms of the consumption good) is
one.

51
 The production function is y = k , where k is capital per worker and y is output per
worker. The deprecation rate is zero. Factors are paid their marginal products.

Note: the number of children per worker in this model is n, not (1+n), which is how we
usually denote it.

A) [15 points] Solve for the steady state level of output.

B)[15 points] Assume = 2. Consider two countries: one is described by the model above; in
the other one, n is fixed at 1/6. Describe (and graph) the difference equation relating kt+1
to kt in both countries. If both countries started off at half of their steady state level of
output per worker, which would move more rapidly to the steady state? Explain why.

28) Consider a variation of the OLG model in which people may die at the end of the first
period. At the beginning of the second period of their lives, people die with probability (1-p) and
live with probability p. The do not know during the first period of their lives whether they will
die or not.

There is no time discounting. Their expected utility (which they maximize) is

E(U) = ln(c1) + p*ln(c2) ,

where c2 is what they will consume in period 2, if they live that long.
The production function is y=k, and there is no population growth.
Assume that when people die their wealth is distributed among the remaining members of their
generation. Assume that there are a sufficiently large number of people in each generation so
that there is no uncertainty about the size of bequests that survivors will recieve. Also assume
that interest is earned by the capital belonging to the deceased before this capital is divided up
among the survivors.

A. Write down the single period and lifetime budget constraints of an individual. Call the
amount that she receives as a bequest b.

B. Solve for the individual's optimal saving in period 1 as a function of b, r, and w.

C. Solve for b as a function of the amount of capital in the second period.

D. Put everything together into a difference equation for K.

E. How does a decrease in p (that is, an increase in the probability of premature death) affect the

52
steady-state capital stock?

29) (Core exam, 2001-02) Consider the following OLG model. People live for two periods.
They are born and die with zero assets. They work only in the first period of their lives. The
capital stock in any period is composed of the savings of the older generation.

People have the following lifetime utility functions:

U = ln(c1) + ln(c2)

The production function is

y = k,

where y is output per worker and k is capital per worker. Workers are paid their marginal
products.

Population grows at rate n, so that the number of young in period t+1 is (1+n) times the
number of young people in period t.

Capital fully depreciates after it is used in production. This means that the elderly
consume the earnings from the capital they own, but not the capital itself.

The economy is in steady state.

A) Solve for the steady state level of capital per worker.

B) Solve for the steady state levels of first and second period consumption.

C) Solve for the steady state level of lifetime utility.

D) Assume that = 2. Is steady state lifetime utility higher in a country where population
growth is positive (i.e. n>0) or negative (i.e. n<0)? Indicate how you know this.

29.5) (Midterm exam, 2003) Consider an OLG model in which people live for two periods.
They work in the first period and consume in both periods. The size of the population is
constant. People have log utility with a time discount rate of zero.

The production function is

Y = K (eL)1- ,

where e is a measure of the number of efficiency units of labor per worker. Define capital and

53
output per efficiency unit of labor as follows:

kt = Kt / et L
yt = Yt / et L

Technological progress takes the following form. The value of e does not change from an
odd period to an even period. But from an even period to an odd period, e increases by a factor
of (1+ ). So, for example,

e1 = e0 (1 + )
e2 = e1
e3 = e2 (1 + ) , etc.

Write down the difference equations relating k in an odd period to k in the previous even
period, and similarly k in an even period to k in the previous odd period. Using these two
equations, analyze the dynamics of the economy. Draw a diagram showing how k will evolve
from some initial condition (you may want to draw several of these, to show what can happen in
different cases). Also solve algebraically for the Asteady state@ values of k in odd and even
periods.

30) (core exam: 1996) Consider the following small open economy OLG model.

The world interest rate is >0. There is no depreciation or population growth.

A country is perfectly open to the world capital market. The production function in the
country is

Yt = Kt.5(etLt).5

Where Yt is output at time t, Kt is capital stock, Lt is labor supply, and et is a measure of

efficiency units per worker.

e increases at rate g: et+1 = (1+g) et g>0

Every period, a new generation is born. People live two periods: in the first they work and
consume. In the second they do not work and they consume their savings. There are no bequests
or intergenerational transfers. People have lifetime utility functions:

V = ln(C1) + ln(C2)

where C1 is consumption in the first period of life and C2 is consumption in the second
period of life.

Derive an expression showing whether, in steady state, this country will have positive,

54
negative, or zero net foreign assets.

31) (Final 2001) Consider the following OLG economy. People live for two periods. They
work only when they are young, and consume in both periods. The economy is closed. The
utility function is

U = ln (c1) + ln(c2).

There is no population growth. The production function (in per worker terms) is

y = A k

where = 2. A is not constant. Rather, A takes the value Ao in odd periods and Ae in even
periods, where

Ao > Ae

The depreciation rate is 100%. That is, capital fully depreciates after use.

A) [15 points] Solve for steady state level of output in even period and in odd periods. In
which periods is output higher?

B) [10 points] Who has higher utility: workers born in odd periods or workers born in even
periods?

32) (old core exam question) Consider the following variation on the Solow model. There is
no technological change or population growth. Output is produced using capital and labor.
Capital depreciates at rate . Production is Cobb-Douglas. There is, however, a productive
externality that takes effect when the level of capital per worker is above some level k*. Thus the
production function is:

y = k if k<k*
Ak if k$k*

where A>1 and 0<<1

Suppose that the world is composed of many countries, but that all of these countries have
closed economies. Assume that countries have one of three saving rates, sl (low), sm (middle), or
sh (high). The following is true about these saving rates:

sl < sm < sh

sl A k* < k*

55
 sm k* < k* < sm A k*

k* < sh k*

Each country also has initial capital stock per worker k0. Assume that there is wide variation in
initial capital stocks, and further that there is no relation between a country's saving rate and its
initial capital stock.
Analyze how the time paths of the capital stocks of the different countries will depend on their
saving rates and initial capital stocks. Pay particular attention to the questions of "convergence"
(that is, to what extent will countries converge to the same steady state) and multiple equilibria.
Under what conditions will countries with the same saving rates end up at the same steady state?
When will countries with the same initial capital stock end up at the same steady state?

32.5) Consider an economy described by the Solow model. The production function is

Y = AK1/3 L2/3

A grows exogenously at a rate of 2% per year. L grows exogenously at a rate of 1% per

year. The economy is closed, and the fraction of output saved (and invested) is 20%. The
depreciation rate is 1% per year.

At time zero, the level of output is 40 and the capital stock is 100.

At time zero, an economist conducts a Agrowth accounting@ exercise, by starting with the
production function, taking logs, and differentiating with respect to time.

Y
A
1 K
2 L

= + + 0
Y A 3K 3L

The economist interprets the first term on the right hand side of this equation as Agrowth
due to technology,@ the second as Agrowth due to capital accumulation,@ and the third as Agrowth
due to population increase.@

A.What fraction of total growth at time zero is due to capital accumulation, using this method?

B.Suppose that instead of the calculation in part A, the researcher reasoned as follows: In this
model, growth of output can be due to three things:
(i) technological progress
(ii) population growth
(iii) accumulation of capital beyond what is due to technological progress and population

56
 growth (that is, accumulation of capital due to movement of the economy toward
its steady state).

The economist figures out the part of capital growth due to cause (i) and decides to count
growth in output resulting from this sort of capital growth as part of Agrowth due to technological
progress@ rather than Agrowth due to capital accumulation.@ Similarly, she decides to count
output growth that results form capital growth due to cause (ii) as part of Agrowth due to
population increase@ rather than Agrowth due to capital accumulation.@ She counts only growth in
output due to growth in capital of type (iii) as part of Agrowth due to capital accumulation.@
Using this approach, what fraction of growth at time zero is due to capital accumulation?

33) (old midterm question) Consider the following variation on the Solow model. Suppose
that the true production function (in per capita terms) is:

y = A1k + A2k

There is no exogenous technological change. Population grows at rate n and capital depreciates
at rate .

Assume that all countries in the world have the same values for A1 and A2, and that they all
have the same saving rate. Countries differ only in their initial level of capital per person.

Discuss the extent to which countries with different initial levels of the capital stock will or
will not converge over time. Distinguish, if appropriate, between different special cases based on
the values of the parameters.

34) (old core exam question) Consider an augmented Solow-style model which includes
human capital. Let k be the level of physical capital per worker, h be the level of human capital
per worker, and y be the level of output per worker. Both human and physical capital depreciate
at rate . There is no population growth. Let sk be the fraction of output invested in physical
capital, and sh be the fraction of output invested in human capital. The production function is

y = Akk + Ahh ,

where Ak and Ah are two constants. Assume that skAk < and that
shAh < .

Analyze the dynamics of this model using a phase diagram in h-k space. Describe the
paths of h and k from different initial positions. Under what condition does the model produce
"endogenous growth," and what happens when this condition is not met?

35) Consider the following model. Output is produced according to the production function
y=Ak, where y is output per capita and k is capital per capita. There is no technological progress.

57
Population grows at rate n. A constant fraction, s, of output is saved.

Capital depreciates in an unusual fashion: every period, d units of capital per person
depreciate.

Analyze the dynamics of this economy. Describe the conditions under which there will
be steady states, endogenous growth, etc. Calculate the long-run rate of growth, if there is one.

36) A country has production function y=Ak, where y is output per person and k is capital per
person. Capital depreciates at rate >0. There is no population growth. Consumption is chosen
by a social planner who maximizes

c(t )1 -

U = e- t dt
0
1-

where c(t) is consumption at time t, >0 is the discount rate and >0. The capital stock at time 0
is k(0). Solve for the optimal value for initial consumption, c(0).

36.25) [midterm, 2006] A social planner makes all the consumption decisions in a country.
The only source of output in the country is wheat. Each grain of wheat harvested can either be
eaten or planted. If it is planted, there is one percent chance that it will grow into a plant the next
year, and if it does so, it will produce 104 grains of wheat. (The labor and land required to grow
the wheat are not constraints on how much can be produced.) You should assume that the law of
large numbers holds, so that there is no uncertainty about the number of grains that will be
harvested.

The social planner cares about utility per capita. She discounts the future at a rate of 1%
per year, and has log utility. Population grows at 1% per year.

In year t, the stock of wheat is equal to Wt. Solve for the optimal consumption in year t.

Note: even though this problem is phrased in discrete time, you will find it much easier to solve
in continuous time.

36.5) (final exam, 2003) Consumption in an economy is determined by a Ramsey-style social

planner who has log utility and time discount rate = 0.05. The rates of depreciation and
population growth are both zero. Time is continuous. The production function is

y = Ak

From time zero to time 1, the value of A is 0.05. At time 1, A will remain at 0.05 with

58
probability 50% and will rise to 0.10 with probability 50%.

What can you say about the growth rate of output between time zero and time 1? You
don=t have to give exact initial value or time path. Just say whether output is rising, falling, or
constant, and explain why. [Note: this problem is harder than it looks.]

36.75) [final exam, 2008] Consider two economies. They have the same production function
(in per worker terms):

y = Ak

Population growth and deprecation are both zero. Time is continuous. In both countries, the
A
= .
social planner has the same time discount rate > 0. Finally, assume 2

In country one, the social planner has log utility. In country two, the social planner has
2
= .
CRRA utility, with coefficient of relative risk aversion 3

The two countries start at time zero with the same level of capital per worker. After how
many years will the levels of consumption in the two countries be equal?

37) Consider the following OLG model. Individuals live for two periods. There is a new
generation born every period.

In the first period, individuals supply one unit of labor inelastically. The do not consume
in the first period. In the second period, they do not work, and consume their first period savings
along with accumulated interest. Individuals are born and die with zero assets. There is no
intergenerational altruism and no population growth.

Production is carried by identical small firms. Firm i has production function:

Yi = B Ki Li(1-)

where Ki and Li and the quantities of labor and capital used by firm i. Firms pay factors their
marginal products.

The aggregate productivity coefficient (B), is determined by the aggregate capital stock

B = A K(1-),

59
where K (without a subscript) is the total capital stock in the economy. Firms do not take the
effect of their capital stock on B into account in their production decisions.

Solve for (and describe) the steady-state of the model. If there is either a steady-state
level of output or a steady state growth rate of output, solve for it. If not, explain why not.

38) (Final exam, 2008) A country is described by the version of the Lucas model presented in
class:

y = k (uh)(1-)

h

= (1 u )
h

k
= sy k

The country is on a balanced growth path. Now suppose that there is an increase in u. For
convenience, consider an infinitesimal increase in u, so you can just take derivatives.

What are the conditions regarding the parameters u, , s, , and such that the growth
rate of output immediately following the rise in u will be the same as it was along the old
balanced growth path?

Your solution should not be in terms of k, h, or y.

39) Consider a variation on the Lucas model analyzed in class.

Production (in per capita terms) is

y = k.5h.5

where k and h are per capita levels of physical and human capital.

Physical capital is accumulated according to

k
= sk y - k

60
 Human capital is accumulated according to

h
= sh y - h

Further, it just so happens that sk = sh =

A. What is the steady state growth rate of output?

B. Suppose that a country is in steady state, when suddenly 3/4 of its physical capital (but none
of its human capital) is destroyed. Will the growth rate of output immediately after the shock be
higher, lower, or the same as the growth rate in steady state?

40) Three countries, A, B, and C, are all described by the version of the Lucas model
presented in class:

y = k (uh)(1-)

h
/ h = phi*(1-u)

k
= sy - (n+)k

Initially, all three countries have the same levels of s, u, h, and k. In year t, dictators seize power
in countries A and B and engage in policy experiments. Country A raises the rate of saving to
some higher level, holding u constant. Country B reduces u to some lower level, holding s
constant. Country C keeps both s and u constant. After 10 years, the dictators are overthrown,
and s and u are returned to their original levels in countries A and B.
Draw a graph showing the time paths of the growth rates of output in the three countries
during and after the period of policy experimentation. Distinguish between different special
cases based on the parameters, if appropriate. Draw a second graph showing the time paths of
the log of output in the three countries.

40.5) [midterm exam, 2004] Consider a variation of the Lucas growth model. The production
function is

y = (uk k) (uh h)(1-)

Where uk is the fraction of physical capital that is used in producing output and similarly uh is the
fraction of human capital that is used in producing output.

The accumulation equations for k and h are

61
k
= (1-uk)k - (n+ ) k

h
= (1-uh)h - (n+ ) h

A) solve for the steady state growth rate of output

B) Suppose that = .5, uk = uh = .5, (n+ = .05), = 0.1. A country is in steady state.
Now suppose that uk falls to .125, while uh remains constant. How many years (approximately)
will it take until output regains the level it would have had if there had been no change is uk?
[Helpful hint: ln(2) . 0.70].

41) (core exam: 1996) Consider the following hybrid of the Solow model and the Lucas model
with human capital.

Output is produced with physical capital, human capital, and labor according to the
production function (written in per-worker terms):

y = k [(1-u)h]1-

where y is output per worker, h is human capital per worker, k is physical capital per
worker, and (1-u) is the fraction of their time that workers spend producing output

There is no population growth. Physical capital accumulates according to

k
= sy - k

where the saving rate, s, is exogenous and fixed. When workers are not producing
output, they are producing human capital (there is no leisure). The only input required to
produce human capital is time (unlike the Lucas model, where human capital itself is one of the
inputs to human capital production). Human capital also depreciates at the same rate () as
physical capital. Thus, the evolution of human capital per capita is given by:

h
= u - h

Assume that s = .

62
A. Describe the steady state of the model. Is it one with a constant level of output or with a
constant growth rate of output?

B. Solve for the level of u that maximizes either the level or the growth rate of output
(depending on which is constant in steady state).

C. Let u* be the level of u that you solved for in part B. Suppose that the economy is in steady
state with u<u*. Suddenly, u jumps up to u*. Use a phase diagram in (h,k) space to analyze the
behavior of h and k. Draw time series pictures showing how h, k, behave during the transition to
the new steady state.

42) [midterm exam 2002] Consider the problem of a social planner choosing the fraction of
the labor force to be devoted to R&D and the fraction that will work. The population is constant
and is normalized to one. The social planner has the standard discounted utility function

V = e - t ln(c(t)) dt 0
0

There is no physical capital in the model. Consumption is equal to production, which is

given by the production function

y=hu

where h is the level of human capital and u is the fraction of their time that workers spend
producing output. The growth rate of human capital is given by

h

= (1- u) 0
h

The level of human capital at time zero is equal to one. At time zero, the social planner
must choose a value of u that will be constant forever.

Finally, we assume that < .

Your task is to solve for the optimal level of u.

Big hint: To solve this problem, you do not need to think about first order conditions or dynamic
optimization or things like that. Rather, you can just write out the present discounted value of
utility as a function of the choice of u and maximize.

Essential helpful fact:

63

1
te
- t
dt =
0 2

(To get this fact you have to do a bunch of limits and other junk B it is not something that I would
expect you to know off the top of your head.)

43) [final exam 2002] This question extends the previous problem to deal with the issue of
technology spillovers. In that problem, you considered the problem of a social planner
choosing the fraction of the labor force to be devoted to R&D and the fraction that will
work. The population is constant and is normalized to one. The social planner has the
standard discounted utility function

V = e - t ln(c(t)) dt
0

There is no physical capital in the model. Consumption is equal to production, which is

given by the production function

y=hu

where h is the level of human capital and u is the fraction of their time that workers spend
producing output. The growth rate of human capital is given by

h

= (1- u)
h

The level of human capital at time zero is equal to one. At time zero, the social planner
must choose a value of u that will be constant forever.

Finally, we assumed that < .

Your task was to solve for the optimal level of u.

Now consider a model where there are technological spillovers from a leader country to a
follower country. Unlike the version of the model that we considered in class, the effect of the
spillover does not depend on how large the gap in technology is B a follower benefits as long as it
is even slightly behind the leader. Specifically, let the growth rate of technology in the leader
country be

64
 h

= L (1- u L )
h L

while in the follower it is

h

= F (1- u F )
h F

where < L < F

There are two countries, labeled 1 and 2. At time zero, the two countries have the same
level of h0, which for convenience we set equal to one.

In country 1, the value of u is fixed at some level u1, and will not be affected by the action
of country 2. The social planner in country 2 has the same infinite horizon utility function we
introduced above. She is able to choose some level of u2. However, the social planner in country
2 can only choose a single value of u2 that will hold for all points in time starting at time zero.
She cannot choose a path where u2 changes over time.

Finally, assume that the level of u in country 1 is set at the level that the social planner in
country 2 would have chosen if there were no technology spillovers between countries.

A. [10 points] Draw a diagram showing the growth rate of output in country 2 as a function of
u2. Calculate the value of u2 at which there is a the key inflection point.

B.[10 points] Draw a diagram showing the present discounted value of utility in country 2 as a
function of the choice of u2. Label the optimal choice.

44) Consider an economy in which there are two kinds of capital: private capital, consisting
of machines, buildings, etc., and government capital, consisting of highways, dams, etc. The
production function is:
y = k.5 z.5

where y is the per capita level of output, k is the per capita level of private capital, and z is the
per capita level of government capital. Both private capital and government capital depreciate at
rate . There is no population growth or technological change.

The government collects a constant fraction, , of output, which it spends on building

government capital. Of the remaining output, a constant fraction, s, is invested in building
private capital.

65
 Show that k/z converges to a steady state along the balanced growth path. Find the
growth rate of the economy when k/z is at this steady state level. Find the level of taxes, ,
which maximizes the steady-state growth rate of output.

45) (old core exam question) Consider a country with production function:

Y = K.5 H.5

where K is the stock of physical capital and H is the stock of human capital. There is no
depreciation, population growth, or technological change. Physical capital is accumulated
according to:

K = sk Y

and human capital is accumulated according to:

H = sh Y

A country is growing at a constant, positive rate along its balanced growth path. At a
point in time, the country's saving rate in physical capital (sk ) doubles, while it's saving rate in
human capital (sh) remains constant.

[Note: you do not have to answer part A in order to answer part B]

A. What will be the effect of this change on the growth rate of output along the new balanced
growth path? That is, by what factor will the growth rate of output along the new balanced
growth path differ from that along the initial balanced growth path?

B. What will be the instantaneous effect of this change on the growth rate of output? That is, by
what factor will the growth rate of output change instantaneously?

46) (Midterm exam, 2001) Consider the Aghion and Howitt model of technology creation.
Suppose that there are only a fixed number of possible new inventions. Call this number T.
Once the Tth new technology has been invented, there will never be another new technology to
replace it. Further, suppose that patents last forever, so that the firm which invents the Tth new
technology will have a monopoly on producing the intermediate good forever. Obviously, once
the Tth new technology has been invented, the labor force devoted to R&D will be zero. Now,
consider how much labor will be devoted to R&D during the epoch of the T-1 technology, the T-
2 technology, etc. To be more concrete, let us compare R&D labor force in these periods to
R&D labor force that would be observed in the steady state if there were no limit to the number
of new technologies. How would the R&D labor force in epoch T-1 compare to the steady state
of the regular model? How about the R&D labor force in epoch T-2? T-3? T-4? Etc.

66
 I do not want you to go through the whole model to answer this question. Based on the
logic of the model it should not be too hard to see the answer.

47) Consider the following model with endogenous technology and population.

Population growth depends on the difference between current income per capita, y, and
(which is some sort of subsistence level):

n = (y y)

Output is produced with labor (L) and land (R).

Y = (AR) L1-

Technology growth depends on output per capita:

A
/ A = y 0< <

Analyze the model's dynamics and steady state(s). Solve for any steady state values of
output per capita.

48) A researcher examines cross country data. She fits the following equation for growth
rates:

= .01 - .02 ln(y) + .03 ln(s) + .04 DEM

Where is the growth rate of income per capita, s is the saving rate, and DEM is a dummy
variable that takes the value 1 if the country is a democracy and zero if it is not.

Assume that the world is properly described by the Solow model (with no human capital),
but that in addition, the form of government affects the production function.

Calculate the steady state difference in income per capita between a democracy and a
non-democracy which have the same saving rate.

48.5) [midterm exam, 2004] Consider the partial equilibrium problem of an individual deciding
how many children to have. He is endowed with one unit of time, which he can spend
working or raising children. His wage per unit time, w, is exogenous. The cost of raising

67
 children is z units of time per child [he is able to have non-integer numbers of children].
His utility function is

V = ln( c - c*) + ln(n)

Where c* is the subsistence level of consumption (we assume that w> c* ).

Trace out and describe as fully as possible the relationship between the wage and fertility.

48.75) [Midterm exam, 2006] As discussed in class, it is not clear whether the social planner
should care about total utility or utility per capita. Suppose that as a compromise, the social
planner cares about both. Specifically, let the instantaneous utility function of the social planner
be

U = ln(c) + (1 ) ln( Lc) 0 < <1

Where L is total population and c is consumption per capita. Time is continuous. The social
planner discounts the future at rate > 0 .

In this economy, output is produced solely by land and labor. The quantity of land, X, is fixed.
The production function is:

Y = L X 1 0 < <1

Consumption is equal to output.

The population at time zero is L(0). The social planner can set the growth rate of population at
any level between 10% per year and -10% per year. Solve for the social planners optimal rate of
population growth.

Note: you do not need to use fancy math to solve this problem.

49) (midterm exam, 2001) In our analysis of the Lucas model of population, we assumed
that the Aprice@ of children was in terms of goods. Suppose instead that the price of children is in
terms of time. Specifically, suppose that raising one child takes z units of time.

An economy is composed of individuals who are adults for one period. In that period
they work, consume, and raise children. (Children do not do anything in the model other than use
up time from their parents.) They are endowed with one unit of time, which is used for either
working or child-raising (there is no leisure). The number of children per person is n. The
budget constraint is thus

68
 c = w(1-zn)

where w is the wage.

People get utility from consumption and the number of children:

U = ln(c) + ln(n)

However, there is also a minimum level of consumption,c , below which c cannot fall (think of
this as being the level of consumption necessary for subsistence.) That is, c $c .

A) Solve for the optimal levels of fertility and consumption as functions of the wage. Draw a
diagram with n on the vertical axis and c on the horizontal axis, and show how changing the
wage affects the budget constraint and the optimal choices of fertility and consumption. If there
are any critical levels of wage where the behavior changes, solve for them.

B) Now, suppose that we embed this model of population in a very simple growth model.
Production uses land and labor. The quantity of land in the economy, X, is fixed. The
production function is

Y = (L(1- zn) )1/2 X 1/2

where L is the number of adults. We assume that land earns no return, so workers are paid their
average product.

Assume that z< 2 .

What are the steady state levels of consumption and fertility? Explain how you know this. You
should be able to figure this out very easily, without doing a lot of math.

50) [midterm exam, 2005] Consider a country described by the following model. The
production function is

y = k (uh)1 x1

Where k is capital per worker, h is human capital per worker, u is the fraction of their time that
workers spend producing output, and x is land per worker. Population grows at rate n and capital
depreciates at rate .

Capital accumulation is given by the standard equation

69
 k
= sy (n + )k

Human capital accumulates according to the equation

h
= h(1 u ) nh

(Notice that I have added a term for dilution of human capital per worker due to population
growth.)

The fraction of time spent working (u) and the saving rate (s) are taken as exogenous.

k
Analyze this model in a diagram with on the horizontal axis and the three measures
xh
k
h
x

, , and all measured on the vertical axis. Indicate the steady state and briefly describe its
k h x
stability properties.

50.5) (Final exam, 2008) Consider an OLG economy where output is produced with land,
labor, and capital. The production function is

Y = K L X 1

Where X is the total quantity of land.

People work, consume, and raise children in the first period of life. The live off their
capital income and consume in the second period of life. All factors are paid their marginal
products. All the land in the economy is owned by an absentee landlord who receives the rent
from land, does his consumption, and has children outside the country. Thus the children of the
landlord are not counted in the economy (you can think of the rent on land as simply being
thrown into the sea). The deprecation rate is zero. There is no technological progress.

The utility function of individuals is

U = ln(c1) + ln(c2) + ln(n)

Where n is the number of children. The price of children in terms of the output good is one.

Solve for the steady state level of GDP per worker.

50.75) (Core exam, 2008) A country has an aggregate production function

70
 Y = K L X 1

Where X is land. Population is equal to the labor force. Population growth is given by the
equation

L

= 0 + 1 y
L

Where y is GDP per capita, 0 < 0 , and 1 > 0 . The saving rate, s, is fixed. Capital depreciates
at rate .

Suppose that a country is in steady state. Suddenly half of the people are killed. The quantities
of capital and land are not affected. Find the growth rate of output per capita immediately
following the incident. Your answer should be in terms of the parameters of the problem
( 0 , 1 , , , ), not in terms of y or other endogenous variables.

51) (Core exam, 2007) An economy has production function

Y = K (eL) X 1 ,

where L is the labor force, e is efficiency units per worker, K is capital, and X is the quantity of
land. Assume > 0 , > 0 , and + < 1 . The labor force grows at rate n. Efficiency units
per worker grow at rate g. The quantity of land is fixed. The economy is closed to capital flows.
A constant fraction s of output is invested in new capital. The depreciation rate is > 0.

Note #1: because the production function is CRS in capital, labor, and land taken together, it is
DRS in capital and labor. That means that in this economy there will not be a steady state in
which K/(eL) or Y/(eL) is constant. So applying the trick of re-normalizing K and Y by dividing
by (eL) will just get you in trouble.

Note #2: What I call below a steady state will be more familiar to some as a balanced growth
path.

A) Using the production function and the differential equation for capital accumulation, derive a
differential equation showing how the K/Y ratio behaves over time. Show that the ratio K/Y
moves toward a steady state value.

B) Using the production function, and the fact from part (A) that K/Y has a steady state ratio,
solve for the steady state growth rate of output per worker.

71
C) Assume that land earns its marginal product. What is the steady state growth rate total
earnings of land?

72