Chapter 9

Business Cycles

In this chapter we explore the causes of business cycles. Briefly, business cycles are the
recurring fluctuations that occur in real GDP over time. For further descriptions of business
cycles, refer to Barro’s Chapter 9. Here, we concentrate on explaining business cycles. We
begin with an overview of potential explanations. Then we work out a real business cycle
model in detail.

While there are many different theories of business cycles, they share some properties.
There is always a driving force behind economic fluctuations, some sort of shock or distur-
bance that is the original cause of the cycle. In addition, most theories build on a propaga-
tion mechanism that amplifies shocks. Unless the disturbances are already big enough by
themselves to account for the fluctuations, there has to be some propagation mechanism
that translates small, short-lived shocks into large, persistent economic fluctuations.

We will start our search for the cause of business cycles in Section 9.1 by listing a number
of possible shocks and propagation mechanisms. Competing theories of the business cycle
differ in which shocks and mechanisms they emphasize. In Section 9.2 we will concen-
trate on the real business cycle model, which is a straightforward extension of the market-
clearing models that we developed in earlier chapters. Section 9.3 presents simulations for
our real business cycle model and assesses the success of the model in matching real-world
fluctuations.

9.1 Shocks and Propagation Mechanisms

Among the many shocks and disturbances that are present in an economy, only a few
have received special attention in research on business cycles. Here are some of the more
important candidates:
70 Business Cycles

 Technology shocks: Real-world production functions change over time. New tech-
nologies like computers or robots alter the production process and raise overall pro-
ductivity. Sometimes, production facilities break down or do not work as expected,
so productivity falls. This technological change is not always smooth; it often comes
in the form of shocks.
 Weather shocks and natural disasters: Many industries like agriculture or tourism
are weather-dependent. Rainfall and sunshine influence the output of these sectors,
so the weather is a a potential source of fluctuations. This is also true for disasters
like earthquakes or landslides. El Niño is a shock of this kind that received a lot of
attention lately. We can regard these kinds of shocks as one type of technology shock.
Weather changes the production function for wheat, and an earthquake that wiped
out, say, Silicon Valley, would change the production function for computers.
 Monetary shocks: We saw in Chapter 8 on inflation that there are real effects of mon-
etary policy. Therefore random changes to money supply or interest rates are a po-
tential source of fluctuations as well.
 Political shocks: The government influences the economy both directly through gov-
ernment enterprises and indirectly through regulation. Changes in tax laws, antitrust
regulation, government expenditure and so on are a potential source of disruption in
the economy.
 Taste shocks: Finally, it is also conceivable that shifts in preferences cause fluctua-
tions. Fashion and fads change rapidly, and they may cause fluctuations in areas like
the apparel, music, or movie industries.

While the shocks just mentioned are present to some degree in every economy, they are
probably not large enough to serve as a direct explanation of business cycles. For example,
in the United States real GDP fell by 2.8% between October 1981 and 1982. It is hard to
imagine any shock that caused a direct output loss of almost 3% of GDP within only a year,
and if there was one, we would probably be aware of it. It appears more likely that there
are mechanisms present in the economy that amplify shocks and propagate them through
time. Here are some candidates:

 Intertemporal substitution: Shocks that have a negative impact on productivity

lower the marginal return to labor and other factors of production. If marginal prod-
ucts fall, consumer’s might prefer to work less and consume leisure instead. Labor
input would fall, which amplifies the negative impact on output. At the same time,
since consumers prefer a smooth consumption profile they might prefer to lower
savings for some time when a shock hits. On an aggregate level, this leads to lower
investment and a lower capital stock in the future. Therefore a short-lived shock may
have an impact in the future as well.
 Sticky prices: Market economies react to changes with price adjustments. For exam-
ple, a negative productivity shock lowers the marginal product of labor, so that the
9.1 Shocks and Propagation Mechanisms 71

real wage would have to move downward to adjust labor demand and supply. But
if wages are inflexible for some reason, the adjustment cannot take place. The result
is unemployment and an output loss that is larger than the direct effect of the shock.
Similar effects arise if goods prices are sticky.

 Frictions in financial sector: Even small shocks can force the firms the are hit di-
rectly into bankruptcy. This will affect other firms and banks that lent money to the
now bankrupt firms. Often additional firms have to declare bankruptcy, and some-
times even banks fail. Bank failures affect all creditors and debtors and therefore can
have large economic consequences. Serious economic crises are often accompanied
and amplified by series of bank failures. Examples are the great depression and the
current Asian crisis.

Business cycle models can be broadly subdivided into two categories. Some theories re-
gard cycles as a failure of the economic system. Because of frictions or imperfections of
the market mechanism, the economy experiences depressions and fails to achieve the effi-
cient level of output and employment. Models of this kind often rely on financial frictions,
sticky prices, or other adjustment failures as the propagation mechanism. Both technology
shocks and monetary shocks are considered to be important sources of fluctuations. The
Keynesian model of output determination1 falls into this category.

On the other hand, there is a class of models that regards business cycles as the optimal re-
action of the economy to unavoidable shocks. Shocks are propagated through intertempo-
ral substitution within an efficient market mechanism. Technology shocks are considered
to be the main cause of economic fluctuations. Models of this kind are often referred to as
real business cycle models.2

We can be fairly certain that there is some truth to both views of economic fluctuations.
Major economic breakdowns like the great depression or the recent Asian crisis appear
to be closely connected to disruptions in the financial sector. Bank failures and financial
instability played an important role in both cases.

On the other hand, most business cycles are far less severe than the great depression or the
Asian crisis. In the entire post-war history of the United States and the Western European
countries there is not a single depression that caused an output loss similar to the one suf-
fered between 1929 and 1933. The question is whether normal business cycles are caused
by the same kind of frictions that caused the great depression. The Keynesian model with
its emphasis on slow adjustments and sticky prices supports this view. Real business cycle
theorists argue that breakdowns like the great depression are a phenomenon distinct from
usual business cycles, and that usual cycles can be explained as the optimal reaction of an
efficient market system to economic shocks.

1 SeeBarro, Chapter 20.

2 The term derives from the fact that shocks in real business cycle theory are real, as opposed to monetary, and
that sluggish nominal adjustment does not play a role as a propagation mechanism.
72 Business Cycles

In this chapter, we will primarily look for explanations for normal-scale business cycles,
like those experienced in the United States since World War II. How can we determine
whether such cycles are small-scale failures of the economic system rather than simply the
markets’ efficient reactions to shocks? A natural way to answer this question is to build a
number of model economies that include alternative propagation mechanisms, expose the
model economies to shocks, and see whether the outcomes look like real-world business
cycles. This is exactly the road that has been taken by real business cycle theorists. They
have taken standard equilibrium models as a point of departure and exposed them to pro-
ductivity shocks. As it turns out, models of this kind are quite successful at explaining
real-world business cycles. We will now take a closer look at such a real business cycle
model.

9.2 A Real Business Cycle Model

Real business cycle models are straightforward extensions of equilibrium models of the
kind that we use throughout this course. In most cases, the models feature infinitely lived
consumers, and business cycles are generated by random disturbances to production pos-
sibilities. Unfortunately, solving that kind of model is difficult. Often no explicit solution is
available, so numerical approximations have to be used. To keep the presentation tractable,
in this chapter we will use a simpler framework in which people live for two periods only.
The model does not fit the facts as well as a full-scale real business cycle model, but it
serves its purpose as a simple illustration of the main ideas of real business cycle theory.

In the model world there is a sequence of overlapping generations. Each period a new gen-
eration of consumers is born, and each consumer lives for two periods. We will sometimes
refer to the periods as years, and for simplicity we assume that exactly one consumer is
born each year. People work in the first period when they are young. In the second period
they are retired and live on savings. Throughout the model, superscripts refer to the year
when a person was born, while subscripts refer to the actual year. For example, ctt is the
period-t consumption of a consumer who was born in year t, so such a consumer would be
young in period t. Similarly, ctt+1 is the consumption of the same consumer in period t + 1,
when he is old. The consumers do not care about leisure. A consumer born in year t has
the following utility function:

u(ctt ; ctt+1 ) = ln(ctt) + ln(ctt+1 ):

We could introduce a discount factor, but for simplicity we assume that the consumers
value both periods equally. Note that at each point of time there are exactly two people
around: one who was just born and is young, and another who was born the year before
and is now retired. In each period the young person supplies one unit of labor and receives
wage income wt . The labor supply is fixed, since consumers do not care about leisure. The
wage income can be used as savings kt and as consumption ctt . The budget constraint of a
9.2 A Real Business Cycle Model 73

young worker is:

ctt + kt = wt ;
i.e., consumption plus savings equals income from labor. In period t + 1 the consumer born
in t is old and retired. The old consumer lends his savings kt to the firm. The firm uses the
savings as capital and pays return rt+1 to the old consumer. A fraction Æ of the capital wears
out while being used for production and is not returned to the consumer. Æ is a number
between zero and one, and is referred to as the depreciation rate. The budget constraint for
the retirement period is:

ctt+1 = (1 Æ + rt+1 )kt ;

i.e., consumption equals the return from savings.

The household born in period t maximizes utility subject to the budget constraints, and
takes prices as given:
 t ) + ln(ct ) ; subject to:
max ln( ct t+1
ctt ;ctt+1 ;kt
ctt + kt = wt ; and:
ctt+1 = (1 Æ + rt+1 )kt :
We can use the constraints to eliminate consumption and write this as:

max fln(wt kt ) + ln((1 Æ + rt+1 )kt )g :

kt
This is similar to the problem of the consumer in the two-period credit market economy
that we discussed in Section 3.2. From here on we will drop the practice of denoting opti-
mal choices by superscripted stars, since the notation is already complicated as it is. The
first-order condition with respect to kt is:

1 1 Æ + rt+1
0= + :
wt kt (1 Æ + rt+1 )kt
Solving this for kt yields:
w
(9.1) kt = t :
2
Thus, regardless of the future return on capital, the young consumer will save half of his
labor income. Again, this derives from the fact that wealth and substitution effects cancel
under logarithmic preferences. This feature is is helpful in our setup. Since there will
be productivity shocks in our economy and rt+1 depends on such shocks, the consumer
might not know rt+1 in advance. Normally we would have to account for this uncertainty
explicitly, which is relatively hard to do. In the case of logarithmic utility, the consumer
does not care about rt+1 anyway, so we do not have to account for uncertainty.
74 Business Cycles

Apart from the consumers, the economy contains a single competitive firm that produces
output using capital kt 1 and labor lt . Labor is supplied by the young consumer, while the
supply of capital derives from the savings of the old consumer.The rental rate for capital is
rt , and the real wage is denoted wt . The production function has constant returns to scale
and is of the Cobb-Douglas form:

f (lt; kt 1) = At lt kt1 1 :

Here  is a constant between zero and one, while At is a productivity parameter. At is the
source of shocks in this economy. We will assume that At is subject to random variations
and trace out how the economy reacts to changes in At . The profit-maximization problem
of the firm in year t is:
n o
max At lt kt1 1 wt lt rt kt 1 :
lt ;kt 1

The first-order conditions with respect to lt and kt 1 are:

(FOC lt ) At lt 1 kt1 1 wt = 0; and:

(FOC kt 1) At (1 )lt kt 1 rt = 0:
Using the fact that the young worker supplies exactly one unit of labor, lt = 1, we can
use these first-order conditions to solve for the wage and return on capital as a function of
capital kt 1 :

(9.2) wt = At kt1 1 ; and:

(9.3) rt = At (1 )kt 1 :
Since the production function has constant returns, the firm does not make any profits in
equilibrium. We could verify that by plugging our results for wt and rt back into the firm’s
problem. Note that the wage is proportional to the productivity parameter At . Since At
is the source of shocks, we can conclude that wages are procyclical: when At receives a
positive shock, wages go up. Empirical evidence suggests that wages in the real world are
procyclical as well.

To close the model, we have to specify the market-clearing constraints for goods, labor, and
capital. At time t the constraint for clearing the goods market is:

ctt + ctt 1
+ kt = At lt kt1 1 + (1 Æ)kt 1 :
On the left hand side are goods that are used: consumption ctt of the currently young
consumer, consumption ctt 1 of the retired consumer who was born in t 1, and savings
kt of the young consumer. On the right hand side are all goods that are available: current
production and what is left of the capital stock after depreciation.

The constraint for clearing the labor market is lt = 1, since young consumers always supply
one unit of labor. To clear the capital market clearing we require that capital supplied by
9.2 A Real Business Cycle Model 75

the old consumer be equal to the capital demanded by the firm. To save on notation, we
use the same symbol kt 1 both for capital supplied and demanded. Therefore the market-
clearing for the capital market is already incorporated into the model and does not need to
be written down explicitly.

In summary, the economy is described by: the consumer’s problem, the firm’s problem,
market-clearing conditions, and a random sequence of productivity parameters fAt g1 t=1 .
We assume that in the very first period there is already an old person around, who some-
how fell from the sky and is endowed with some capital k0 .

Given a sequence of productivity parameters fAt g1 t=1 , an equilibrium for this economy is
an allocation fctt ; ctt 1 ; kt 1 ; lt g1
t=1 and a set of prices frt ; wt g1
t=1 such that:

 Given prices, the allocation fctt; ctt 1 ; kt 1 ; lt g1

t=1 gives the optimal choices by con-
sumers and firms; and

 All markets clear.

We now have all pieces together that are needed to analyze business cycles in this economy.
When we combine the optimal choice of savings of the young consumer (9.1) with the
expression for the wage rate in equation (9.2), we get:

1
(9.4) kt = At kt1 1 :
2
This equation shows how a shock is propagated through time in this economy. Shocks
to At have a direct influence on kt , the capital that is going to be used for production
in the next period. This implies that a shock that hits today will lead to lower output
in the future as well. The cause of this is that the young consumer divides his income
equally between consumption and savings. By lowering savings in response to a shock,
the consumer smoothes consumption. It is optimal for the consumer to distribute the effect
of a shock among both periods of his life. Therefore a single shock can cause a cycle that
extends over a number of periods.

Next, we want to look at how aggregate consumption and investment react to a shock. In
the real world, aggregate investment is much more volatile than aggregate consumption
(see Barro’s Figure 1.10). We want to check whether this is also true in our model. First, we
need to define what is meant by aggregate consumption and investment. We can rearrange
the market-clearing constraint for the goods market to get:

ctt + ctt 1
+ kt (1 Æ)kt 1 = At lt kt1 1 :

On the right-hand side is output in year t, which we are going to call Yt . Output is the
sum of aggregate consumption and investment. Aggregate consumption Ct is the sum of
the consumption of the old and the young person, while aggregate investment It is the
76 Business Cycles

difference between the capital stock in the next period and the undepreciated capital in
this period3 :

ct + ct 1
+ kt (1 Æ)kt = At lt kt1 1 :
|t {z t }
1
| {z } | {z }
Ct + It = Yt :
Consumption can be computed as the difference between output and investment. Using
equation (9.4) for kt yields:

Ct = Yt It = At kt1 1 + (1 Æ)kt 1 kt
1
= At kt1 1 + (1 Æ)kt 1 At kt1 1
  2
1
(9.5) = 1  At kt1 1 + (1 Æ)kt 1 :
2

Aggregate investment can be computed as output minus aggregate consumption. Using

equation (9.5) for aggregate consumption yields:
 
1
It = Yt Ct = At kt1 1 1  At kt1 1 (1 Æ)kt 1
2
1
(9.6) = At kt1 1 (1 Æ)kt 1 :
2
We are interested in how Ct and It react to changes in the technology parameter At . We
will look at relative changes first. The elasticity of a variable x with respect to another
variable y is defined the percentage change in x in response to a one percent increase in y .
Mathematically, elasticities can be computed as @x y
@y x . Using this formula, the elasticity of
consumption with respect to At is:


@Ct At 1 1
 At kt1 1
=
2
< 1;
@At Ct 1 1
2  At kt1 1 + (1 Æ)kt 1

and for investment we get:

@It At 1
At kt1 1
= 2
> 1:
@At It 1
2 At kt 1 1
(1 Æ)kt 1

It turns out that the relative change in investment is larger. A one-percent increase in At
leads to an increase of more than one percent in investment and less than one percent in
consumption. Investment is more volatile in response to technology shocks, just as real-
world investment is. Of course, to compare the exact size of the effects we would have to
specify the parameters, like  and Æ , and to measure the other variables, like kt .
3 More precisely, I in the model is gross investment, which includes replacement of depreciated capital. The net
t
difference between capital tomorrow and today kt kt 1 is referred to as net investment.
9.3 Simulations 77

If we look at absolute changes instead of relative changes, the results are less satisfactory.
The absolute change is higher in consumption than in investment, while in the real world it
is the other way around. This failure of the model derives from the fact that people are too
short-lived. In real business cycle models, the smaller variations in consumption relative
to investment result from consumers trying to smooth their consumption. In our model,
the possibilities for smoothing are rather limited. The old person has no more time left
and therefore cannot smooth at all, while the young person has only one more year to go.
Therefore a comparatively large fraction of the shock shows up in consumption. In more-
advanced real business cycle models with infinitely lived consumers, the absolute changes
in consumption are much smaller than the absolute changes in investment.

9.3 Simulations

We can get an even better impression of the business cycle in our model by simulating the
economy. This means that we specify all parameters, start at some initial capital stock, and
generate a series of random shocks. We can use the solutions to the model to compute
consumption, investment, output, and the capital stock in the economy for any number of
periods. Then we can compare the results to real-world business cycles.

There are only two parameters to be specified in the model,  and Æ . Our choices are
 = :7 and Æ = :05. The choice for  matches the labor share in the economy to real
world data4 , while the value for Æ is an estimate of the actual average depreciation rate
in an industrialized economy. The initial capital stock k1 was set to .22. The productivity
parameter was generated by:

At = Ā + t :
Here Ā is the average level of productivity, while the t are random shocks. We set Ā =
1. The t where generated by a computer to be independent over time and uniformly
distributed on the interval [ :1; :1]. Thus the shocks can change productivity by up to ten
percent upward or downward.

Figure 9.1 shows the reactions to a single productivity shock of five percent. That is, in
the first period At is equal to its average, A1 = 1. In the second period the shock hits,
A2 = 1:05. From then on, At is back to one and stays there. We can see that even this
single shock has an impact that can be felt for a long period of time. Figure 9.1 shows
the absolute deviations of consumption, investment, and capital from their average values.
It takes about eight periods until all variables are back to their average. In the second
period, when the shock takes place, both consumption and investment are up. In period
3 the capital stock is higher because of the higher investment in period 2. At the same
time, investment falls. Consumption is higher than average because the capital stock is
4 The labor share in an economy is defined to be total wages as a fraction of output. See Chapter 11 to see why
 is equal to the labor share.
78 Business Cycles

higher, even though productivity is back to normal again. From then on, all variables
slowly return to their average values. Note that from period 4 on no one is alive anymore
who was present when the shock took place. The higher investment in the period of the
shock has increased the capital stock, and the effects of that can be felt for a long time.
Thus even a single shock has long-run effects, and investment goes through a full cycle in
response to this shock.

Response to a Single Shock

0.025

0.02
Deviation from Trend

0.015 Capital
Consumption
0.01
Investment
0.005

0
1 2 3 4 5 6 7 8
-0.005

-0.01

Periods

Figure 9.1: Response to a Five-Percent Productivity Shock

Figure 9.2 shows the same information as Figure 9.1, but variables are divided by their
mean so that we can see the relative changes. Investment is by far the most volatile series.
Compared to investment, the changes in capital and consumption are hardly visible.

By looking at a single shock, we were able to examine the propagation mechanism in iso-
lation and to get an impression of the relative volatility of consumption and investment.
But if we want to compare the model outcomes to real-world business cycles, we need
to generate a whole series of shocks. Figure 9.3 shows such a simulation for our model
economy. The combined effects of many shocks cause an outcome that looks similar to
real-world business cycles. There are booms and depressions, the cycles vary in length
within a certain interval, and investment is more volatile than consumption.

Our simple business cycle model is quite successful in emulating a number of business-
cycle facts. Shape, length, and amplitude of business cycles are comparable to real-world
data, investment is relatively more volatile than consumption, and the wage is procyclical.
More-advanced real business cycle models are even better in matching the facts. By in-
troducing variable labor supply we can generate procyclical employment. Using infinitely
lived consumers would get the absolute changes in consumption and investment right.
9.3 Simulations 79

Response to a Single Shock

2.5

2
Relative Deviation

1.5
Capital
Consumption
Investment
1

0.5

0
1 2 3 4 5 6 7 8
Periods

Figure 9.2: Relative Changes in Response to a Five-Percent Pro-

ductivity Shock

State-of-the-art real business cycle models match most business cycle facts, and when fed
with measured productivity shocks, they generate cycles that explain about 70% of the size
of actual business cycles.

This success has led some researchers to the conclusion that business cycles are exactly
what standard economic theory predicts. In the presence of shocks to production possi-
bilities, optimal adjustments of households and firms within an efficient market system
generate just the pattern of fluctuations that is observed in the real world. From this per-
spective, business cycles are no miracle at all. We would be surprised if there were no
business cycles!

Even though technology shocks combined with efficient markets appear to provide a con-
vincing explanation for business cycles, it cannot be ruled out that other shocks or propa-
gation mechanisms also play a role. After all, real business cycle theory does not account
for 100% of the amplitude of actual business cycles, so there have to be other factors as well.
Other types of shocks can be analyzed within the real business cycle framework. There are
also a number of models that emphasize other propagation mechanisms. The Keynesian
model of output determination is the most prominent example5 but models that combine
monetary shocks with frictions in the financial sector have also received a lot of attention
lately. However, so far none of these models matches the ability of the real business cycle
model to mimic actual economic fluctuations.

5 See Barro’s Chapter 20 for a discussion of that model. For a critique of the real business cycle model from

a Keynesian perspective, try the following article: Mankiw, N. Gregory. 1989. “Real Business Cycles: A New
Keynesian Perspective”. Journal of Economic Perspectives 3(3): 79-90.
80 Business Cycles

Many Shocks

2.5

2
Relative Deviation

Capital
1.5
Consumption
Investment
1

0.5

0
1 2 3 4 5 6 7 8 9 10 11 12 13
Periods

Figure 9.3: Capital, Consumption, and Investment with Many

Shocks

Variable Definition
ctt Consumption of generation t when young
ctt+1 Consumption of generation t when old
u(
) Utility function
kt Capital saved in t and used in t + 1
lt Labor
wt Wage
rt Rental rate of capital
f (
) Production parameter
At Productivity parameter
 Parameter in the production function
Æ Depreciation rate
Ct Aggregate consumption
It Aggregate investment
Yt Aggregate output
 Random shock

Table 9.1: Notation for Chapter 9

Exercises 81

Exercises

The following exercises make up a project that can be done in groups or individually.

Exercise 9.1 (Moderate)

As the word “cycle” indicates, for a long time economists thought of business cycles as reg-
ular, recurrent events. The length and severity of business cycles was thought to be mostly
constant. For example, the typical length of one full cycle (from boom through recession
back to boom) was supposed to be between four and seven years. In this question you will
examine the actual business cycles of a country of your choice and examine whether they
seem to follow a regular pattern.

The first thing to do is to get the necessary data. Business cycles are roughly defined as
deviations of real GDP from trend. Therefore you will need to acquire data on real GDP for
some country. A good source is the Penn World Tables, a set of standardized measures of
economic activity for most countries in the world. You can access the World Tables through
a website at the University of Toronto. The address is:

http://arcadia.chass.utoronto.ca/pwt/

Once you are there, select “Alphabetical List of Topics”, then “Real GDP per capita in con-
stant dollars using chain index”, then click on the country of your choice (not the United
States), then use the “Submit Query” button to get the data. Load the data into a spread-
sheet, and you are ready to go.

The first step is to compute the trend component of GDP. Good methods for computing
the trend of a time series require a relatively high amount of complicated computations.
Therefore we will offer you an ad hoc, quick-and-dirty method of computing the trend.
Once we get to the business cycles, it turns out that this method works sufficiently well
for our purposes. We will use GDPt to denote real GDP at time t. The computation of the
trend proceeds in steps:

 Compute the growth rate of GDP for each year. In terms of your spreadsheet, let us
assume that column A is year and column B is real GDP. The first year is in row one.
Now you can put the growth rates into column C. Put the growth rate from year 1 to
2 into cell C1, and so on.

 From now on, we are going to apply a method called exponential smoothing to get
smooth versions of our data. Assume you want to get a smooth version of a times
series xt . Let us call the smooth version x̂t . Basically, the x̂t are computed as a forecast
based on past observations of xt . The first x̂t is set equal to the first xt : x̂1 = x1 .
From then on, the forecasts for the next period are computed as an average of the last
forecast and the actual value: x̂t+1 =  x̂t + (1  )xt , where  is a number between
82 Business Cycles

zero and one. If you plug this formula recursively into itself, you will see that each
x̂t is a weighted average of past xt .
Let us now put a smooth growth rate into column D. Since x̂1 = x1 , the first smooth
value is equal to the original value: D1=C1. For the next value, we apply the smooth-
ing formula. We recommend that you set  to .5: D2=.5*D1+.5*C1. In the same way,
you can get the other smoothed growth rates. For future reference, We will call the
smooth growth rates ĝt .

 In the next step, we are going to apply the same method to real GDP, but additionally
we will use the smooth growth rates we just computed. This smooth real GDP is
the trend we are looking for, and we will place it in column E. As before, in the first
year the smooth version is identical to the original one: Trend1 = GDP1 , thus E1=B1.
From then on, we get the trend in the next period by averaging between the trend
and the actual value (as before), but also applying the smooth growth rates we just
computed. If we do not do that, our trend will always underestimate GDP. From year
two on the formula is therefore:

Trendt+1 = (1 + ĝt )(0:5)Trendt + (0:5)GDPt :

In terms of the spreadsheet, this translates into E2=(1+D1)*(.5*E1+.5*B1), and so on.

This completes the computation of the trend. Plot a graph of GDP and its trend. If the
trend does not follow GDP closely, something is wrong. (Document your work, providing
spreadsheet formulas, etc.)

Exercise 9.2 (Moderate)

Now we want to see the cyclical component of GDP. This is simply the difference between
GDP and its trend. Because we are interested in relative changes, as opposed to absolute
changes, it is better to use log-differences instead of absolute differences. Compute the
cyclical component as ln(GDP) ln(Trend). Plot the cyclical component. You will see
the business cycles for which we have been looking. (Document your work, providing
spreadsheet formulas, etc.)

Exercise 9.3 (Easy)

Now we will examine the cycles more closely. Define “peak” by a year when the cyclical
component is higher than in the two preceding and following years. Define “cycle” as the
time between two peaks. How many cycles do you observe? What is the average length
of the cycle? How long do the shortest and the longest cycles last? Do the cycles look
similar in terms of severity (amplitude), duration, and general shape? (Document your
work, providing spreadsheet formulas, etc.)

Exercise 9.4 (Moderate)

Having seen a real cycle, the next step is to create one in a model world. It turns out that
doing so is relatively hard in a model with infinitely lived agents. There we have to deal
with uncertainty, which is fun to do, but it is not that easy as far as the math is concerned.
Exercises 83

Therefore our model world will have people living for only one period. In fact, there is just
one person each period, but this person has a child that is around in the next period, and
so on. The person, let us call her Jill, cares about consumption ct and the bequest of capital
kt+1 she makes to her child, also named Jill. The utility function is:
ln(ct ) + A ln(kt+1 );

where A > 0 is a parameter. Jill uses the capital she got from her mother to produce
consumption ct and investment it , according to the resource constraint:
p
ct + it = Bkt + t ;
where B > 0 is a parameter, and t a random shock to the production function. The shock
takes different values in different periods. Jill knows t once she is born, so for her it is just
a constant. The capital that is left to Jill the daughter is determined by:

kt+1 = (1 Æ)kt + it ;
where the parameter Æ , the depreciation rate, is a number between zero and one. This just
means that capital tomorrow is what is left over today after depreciation, plus investment.

Compute Jill’s decision of consumption and investment as a function of the parameters kt ,

and t .

Exercise 9.5 (Moderate)

If we want to examine the behavior of this model relative to the real world, the next step
would be to set the parameters in a way that matches certain features of the real world.
Since that is a complicated task, we will give some values to you. B is a scale parameter
and does not affect the qualitative behavior of the model. Therefore we set it to B = :1.
Æ is the depreciation rate, for which a realistic value is Æ = :05. A determines the relative
size of ct and kt in equilibrium. A rough approximation is A = 4. Using these parameters,
compare the reactions of ct and it to changes in t . (Use calculus.)

Exercise 9.6 (Moderate)

In the last step, you will simulate business cycles in the model economy. All you need to
know is the capital k1 at the beginning of time and the random shocks t . As a starting capi-
tal, use k1 = 3:7. You can generate the random shocks with the random number generator in
your spreadsheet. In Excel, just type “=RAND()”, and you will get a uniformly distributed
random variable between zero and one. Generate 50 such random numbers, and use your
formulas for ct and it and the equation for capital in the next period, kt+1 = (1 Æ )kt + it ,
to simulate the economy. Plot consumption and investment (on a single graph). How does
the volatility of the two series compare? Plot a graph of GDP, that is, consumption plus in-
vestment. How do the business cycles you see compare with the ones you found in the real
world? You don’t need to compute the length of each cycle, but try to make some concrete
comparisons.
84 Business Cycles

Exercise 9.7 (Easy)

Read the following article: Plosser, Charles. 1989. “Understanding Real Business Cycles”.
Journal of Economic Perspectives 3(3): 51-78. Plosser is one of the pioneers of real business
cycle theory. What you have done in the previous exercises is very similar to what Plosser
does in his article. His economy is a little more realistic, and he gets his shocks from the
real world, instead of having the computer draw random numbers, but the basic idea is
the same.

Describe the real business cycle research program in no more than two paragraphs. What
question is the theory trying to answer? What is the approach to answering the question?

Exercise 9.8 (Moderate)

What does Plosser’s model imply for government policy? Specifically, can the government
influence the economy, and is government intervention called for?