MAREMAGNUM/CORBIS DOCUMENTARY/

Appendix

GETTY IMAGES
Graphing: A Brief Review
Many economic concepts can be expressed with numbers—the price of bananas,
the quantity of bananas sold, the cost of growing bananas, and so on. Often, these
variables are related to one another: When the price of bananas rises, people buy
fewer bananas. One way to express these relationships is with graphs.
Graphs serve two purposes. First, when developing theories, they offer a visual
way to express ideas that might be less clear if described with equations or words.
Second, when analyzing data, graphs provide a powerful way of finding and inter-
preting patterns. In either case, graphs provide a lens through which a recognizable
forest emerges from a multitude of trees.
Numerical information can be expressed graphically in many ways, just as there
are many ways to express a thought in words. A good writer chooses words that will
make an argument clear, a description pleasing, or a scene dramatic. An effective
economist chooses the type of graph that best suits the purpose at hand.
This appendix discusses how economists use graphs to study the mathematical
relationships among variables. It also points out some of the pitfalls that can arise
when using graphical methods.

Graphs of a Single Variable

Three common graphs appear in Figure A-1. The pie chart in panel (a) shows how
total income in the United States is divided among the sources of income, including

Figure A-1 The pie chart in panel (a) shows how U.S. national income in 2020 was derived from
different sources. The bar graph in panel (b) compares the average income in four countries.
Types of Graphs The time-series graph in panel (c) shows labor productivity in U.S. businesses over time.

Source: U.S. Department of Commerce, The World Bank.

35

22718_Ch02_ptg01.indd 35 14/09/22 10:48 AM

 36 Part I Introduction

compensation of employees, corporate profits, and so on. A slice of the pie represents
each source’s share of the total. The bar graph in panel (b) compares income in four
countries. The height of each bar represents the average income in each country.
The time-series graph in panel (c) traces the rising productivity in the U.S. business
sector over time. The height of the line shows output per hour in each year. You
have probably seen similar graphs in news reports.

Graphs of Two Variables: The Coordinate System

The three graphs in Figure A-1 are useful, but they are limited in how much they can
tell us. These graphs display information about only a single variable. If economists
are looking at the relationships between variables, they may want to display two
variables on a single graph. The coordinate system makes this possible.
Suppose you want to examine the relationship between study time and grade point
average. For each student in a class, you could record a pair of numbers: study hours
per week and grade point average. These numbers could then be placed in parentheses
as an ordered pair and appear as a single point on the graph. Albert E., for instance, is
represented by the ordered pair (25 hours/week, 3.5 GPA), while his “what-me-worry?”
classmate Alfred E. is represented by the ordered pair (5 hours/week, 2.0 GPA).
We can graph these ordered pairs on a two-dimensional grid. The first number
in each ordered pair, called the x-coordinate, tells us the horizontal location of the
point. The second number, called the y-coordinate, tells us the vertical location. The
point with both an x-coordinate and a y-coordinate of zero is called the origin. The
two coordinates in the ordered pair tell us where the point is located in relation to
the origin: x units to the right of the origin and y units above it.
Figure A-2 graphs grade point average against study time for Albert E., Alfred
E., and their classmates. This type of graph is called a scatter plot because it plots
scattered points. Looking at the graph, notice that points farther to the right (indi-
cating more study time) also tend to be higher (indicating a better grade point
average). Because study time and grade point average typically move in the same

Figure A-2
Using the Coordinate System
Grade point average is measured 4.0
on the vertical axis and study time
3.5
on the horizontal axis. Albert E., Albert E.
Alfred E., and their classmates 3.0 (25, 3.5)
are represented by various points. 2.5
The graph shows that students Alfred E.
2.0
who study more tend to get higher (5, 2.0)
grades. 1.5
1.0
0.5

0 5 10 15 20 25 30 35 40

22718_Ch02_ptg01.indd 36 14/09/22 10:48 AM

 Chapter 2 Appendix 37

direction, we say that these two variables have a positive correlation. By contrast,
if we were to graph party time and grades, we would likely find that higher party
time is associated with lower grades. Because these variables typically move in
opposite directions, we say that they have a negative correlation. In either case,
the coordinate system makes the correlation between two variables easy to see.

Curves in the Coordinate System

Students who study more do tend to get higher grades, but other factors also influ-
ence a student’s grades. Previous preparation is an important factor, for instance, as
is talent, attention from teachers, or even eating a good breakfast. A scatter plot like
Figure A-2 does not attempt to isolate the effect that studying has on grades from
the effects of other variables. Often, however, economists prefer looking at how one
variable affects another, holding all other possible variables constant.
To see how this is done, consider one of the most important graphs in econom-
ics: the demand curve. The demand curve traces the effect of a good’s price on the
quantity that consumers want to buy. Before showing a demand curve, however,
consider Table A-1, which shows how the number of novels that Emma buys depends
on her income and on the price of novels. When novels are cheap, Emma buys a
lot of them. As they become more expensive, she instead borrows books from the
library or goes to the movies rather than read. Similarly, at any price, Emma buys
more novels when she has a higher income. That is, when her income increases, she
spends part of the additional income on novels and part on other goods.
We now have three variables—the price of novels, income, and the number of
novels purchased—which is more than can be shown in two dimensions. To put
the information from Table A-1 in graphical form, we need to hold one of the three
variables constant and trace out the relationship between the other two. Because the
demand curve represents the relationship between price and quantity demanded,
we hold Emma’s income constant and show how the number of novels she buys
varies with the price of novels.
Suppose that Emma’s income is $40,000 per year. If we place the number of novels
Emma buys on the x-axis and the price of novels on the y-axis, we can graphically
represent the middle column of Table A-1. When the points that represent these

Table A-1
Novels Purchased by Emma Price For $30,000 For $40,000 For $50,000
Income: Income: Income:
This table shows the number of
novels Emma buys at various $10 2 novels 5 novels 8 novels
incomes and prices. For any 9 6 9 12
given level of income, the data
8 10 13 16
on price and quantity demanded
can be graphed to produce 7 14 17 20
Emma’s demand curve for novels, 6 18 21 24
as shown in Figures A-3 and A-4.
5 22 25 28
Demand curve, D3 Demand curve, D1 Demand curve, D2

22718_Ch02_ptg01.indd 37 14/09/22 10:48 AM

 38 Part I Introduction

Figure A-3
Demand Curve
$11
The line D1 shows how Emma’s (5, $10)
purchases of novels depend on the 10
price of novels when her income is 9 (9, $9)
held constant. Because the price
(13, $8)
and the quantity demanded are 8
negatively related, the demand 7 (17, $7)
curve slopes down.
(21, $6)
6

5 (25, $5)

4 Demand, D1

0 5 10 15 20 25 30

entries from the table—(5 novels, $10), (9 novels, $9), and so on—are connected, they
form a line. This line, pictured in Figure A-3, is known as Emma’s demand curve
for novels; it tells us how many novels Emma buys at any price, holding income
constant. The demand curve slopes down, indicating that a lower price increases
the quantity of novels demanded. Because the quantity of novels demanded and
the price move in opposite directions, we say that the two variables are negatively
related. (Conversely, when two variables move in the same direction, the curve
relating them slopes up, and we say that the variables are positively related.)
Now suppose Emma’s income rises to $50,000 per year. At any price, Emma buys
more novels than she did at her previous income. Just as we earlier drew Emma’s
demand curve for novels using the entries from the middle column of Table A-1,
we now draw a new demand curve using the entries from the right column of the
table. This new demand curve (curve D2) is pictured alongside the old one (curve
D1) in Figure A-4; the new curve is a similar line drawn farther to the right. We
therefore say that Emma’s demand curve for novels shifts to the right when her
income increases. Likewise, if Emma’s income were to fall to $30,000 per year, she
would buy fewer novels at any price, and her demand curve would shift to the left
(to curve D3).
In economics, it is important to distinguish between movements along a curve
and shifts of a curve. As Figure A-3 shows, if Emma earns $40,000 per year and
each novel costs $8, she buys 13 novels per year. If the price of novels falls to $7,
Emma increases her purchases to 17 novels per year. The demand curve, however,
stays fixed in the same place. Emma still buys the same number of novels at each
price, but as the price falls, she moves along her demand curve from left to right.

22718_Ch02_ptg01.indd 38 14/09/22 10:48 AM

 Chapter 2 Appendix 39

Figure A-4
Shifting Demand Curves
$11
The location of Emma’s demand
curve for novels depends on how 10
much income she earns. The more (13, $8)
9
she earns, the more novels she (16, $8)
buys at any price, and the farther 8 When income increases, the
(10, $8)
to the right her demand curve demand curve shifts to
7 the right.
lies. Curve D1 represents Emma’s
original demand curve, based on 6
an income of $40,000 per year. If When income D3
5 decreases, the
her income rises to $50,000 per (income = D1 D2 (income =
4 demand curve $30,000) $50,000)
year, her demand curve shifts to (income =
shifts to the left.
D2. If her income falls to $30,000 $40,000)
3
per year, her demand curve shifts
to D3. 2

0 5 10 13 15 16 20 25 30

By contrast, if the price of novels remains fixed at $8 but her income rises to $50,000,
Emma increases her purchases of novels from 13 to 16 per year. Because Emma buys
more novels at each price, her demand curve shifts out, as shown in Figure A-4.
There is a simple way to tell when it is necessary to shift a curve: When a relevant
variable that is not named on either axis changes, the curve shifts. Income is on
neither the x-axis nor the y-axis of the graph, so when Emma’s income changes,
her demand curve shifts. The same is true for any change that affects Emma’s
purchasing habits, with the sole exception of a change in the price of novels. If, for
instance, the public library closes and Emma must buy all the books she wants to
read, she will demand more novels at each price, and her demand curve will shift
to the right. Or, if the price of movies falls and Emma spends more time watching
them and less time reading books, she will demand fewer novels at each price, and
her demand curve will shift to the left. By contrast, when a variable on an axis of
the graph changes, the curve does not shift. We read the change as a movement
along the curve.

Slope
One question we might want to ask about Emma is how much her purchas-
ing habits respond to changes in price. Look at the demand curve pictured in
Figure A-5. If this curve is very steep, Emma buys nearly the same number of
novels whether they are cheap or expensive. If the curve is much flatter, the
number of novels she buys is more sensitive to price changes. To answer ques-
tions about how much one variable responds to changes in another, we can use
the concept of slope.

22718_Ch02_ptg01.indd 39 14/09/22 10:48 AM

 40 Part I Introduction

Figure A-5
Calculating the Slope
of a Line
To calculate the slope of the
demand curve, look at the changes
in the x- and y-coordinates as we
move from the point (13 novels,
$8) to the point (21 novels, $6).
The slope of the line is the ratio
of the change in the y-coordinate
(−2) to the change in the
x-coordinate (+8), which equals
−1/4.

The slope of a line is the ratio of the vertical distance covered to the horizontal
distance covered as we move along the line. This definition is usually written in
mathematical symbols as follows:

Dy
slope 5 ,
Dx
where the Greek letter ∆ (delta) stands for the change in a variable. In other words, the
slope of a line is equal to the “rise” (change in y) divided by the “run” (change in x).
For an upward-sloping line, the slope is a positive number because the changes
in x and y move in the same direction: If x increases, so does y, and if x decreases,
so does y. For a fairly flat upward-sloping line, the slope is a small positive number.
For a steep upward-sloping line, the slope is a large positive number.
For a downward-sloping line, the slope is a negative number because the
changes in x and y move in opposite directions: If x increases, y decreases, and
if x decreases, y increases. For a fairly flat downward-sloping line, the slope is a
small negative number. For a steep downward-sloping line, the slope is a large
negative number.
A horizontal line has a slope of zero because, in this case, the y-variable never
changes. A vertical line is said to have an infinite slope because the y-variable can
take any value without the x-variable changing at all.
What is the slope of Emma’s demand curve for novels? First of all, because the
curve slopes down, we know the slope will be negative. To calculate a numerical
value for the slope, choose two points on the line. With Emma’s income at $40,000,
she buys 13 novels at a price of $8 or 21 novels at a price of $6. When we apply the

22718_Ch02_ptg01.indd 40 14/09/22 10:48 AM

 Chapter 2 Appendix 41

slope formula, we are concerned with the change between these two points. In other
words, we are concerned with the difference between them, which tells us that we
will have to subtract one set of values from the other, as follows:

Dy second y -coordinate 2 first y -coordinate 628 22 21

slope 5 5 5 5 5
Dx second x -coordinate 2 first x -coordinate 21 2 13 8 4

Figure A-5 shows graphically how this calculation works. Try computing the slope
of Emma’s demand curve using two different points. You should get the same
result, − ¼. One of the properties of a straight line is that it has the same slope
everywhere. This is not true of other types of curves, which are steeper in some
places than in others.
The slope of Emma’s demand curve tells us something about how responsive
her purchases are to changes in the price. A small slope (a negative number close
to zero) means that Emma’s demand curve is relatively flat; in this case, she adjusts
the number of novels she buys substantially in response to a price change. A larger
slope (a negative number farther from zero) means that Emma’s demand curve is
relatively steep; in this case, she adjusts the number of novels she buys only slightly
in response to a price change.

Cause and Effect

Economists often use graphs to advance an argument about how the economy
works. In other words, they use graphs to argue about how one set of events causes
another set of events. With a graph like the demand curve, there is no doubt about
the cause and effect. Because we are varying price and holding all other variables
constant, we know that changes in the price of novels cause changes in the quantity
Emma demands. Remember, however, that our demand curve came from a hypo-
thetical example. When graphing data from the real world, it is often more difficult
to establish how one variable affects another.
The first problem is that it is difficult to hold everything else constant when
studying the relationship between two variables. If we are not able to hold other
variables constant, we might decide that one variable on our graph is causing chang-
es in the other variable when those changes are actually being caused by a third
omitted variable not pictured on the graph. Even if we have identified the correct
two variables to look at, we might run into a second problem—reverse causality.
In other words, we might decide that A causes B when, in fact, B causes A. The
COURTESY OF RANDALL MUNROE/XKCD.COM

22718_Ch02_ptg01.indd 41 14/09/22 10:48 AM

 42 Part I Introduction

omitted-variable and reverse-causality traps require us to proceed with caution

when using graphs to draw conclusions about causes and effects.

Omitted Variables To see how omitting a variable can lead to a deceptive graph,
consider an example. Imagine that the government, spurred by public concern about
the large number of deaths from cancer, commissions an exhaustive study from
Big Brother Statistical Services, Inc. Big Brother examines many of the items found
in people’s homes to see which of them are associated with the risk of cancer. Big
Brother reports a strong relationship between two variables: the number of cigarette
lighters that a household owns and the probability that someone in the household
will develop cancer. Figure A-6 shows this relationship.
What should we make of this result? Big Brother advises a quick policy response.
It recommends that the government discourage the ownership of cigarette light-
ers by taxing their sale. It also recommends that the government require warning
labels: “Big Brother has determined that this lighter is dangerous to your health.”
In judging the validity of Big Brother’s analysis, one question is key: Has Big
Brother held constant every relevant variable except the one under consideration?
If the answer is no, the results are suspect. An easy explanation for Figure A-6 is
that people who own more cigarette lighters are more likely to smoke cigarettes
and that cigarettes, not lighters, cause cancer. If Figure A-6 does not hold constant
the amount of smoking—and it doesn’t because Big Brother never looked at that
variable—it does not tell us the true effect of owning a cigarette lighter.
This story illustrates an important principle: When you see a graph used to sup-
port an argument about cause and effect, it is important to ask whether the move-
ments of an omitted variable could explain the results you see.

Reverse Causality Economists can also make mistakes about causality by misread-
ing its direction. To see how this is possible, suppose the Association of American
Anarchists commissions a study of crime in America and arrives at Figure A-7,
which plots the number of violent crimes per thousand people in major cities against
the number of police officers per thousand people. The Anarchists note the curve’s
upward slope and argue that because police increase rather than decrease the amount
of urban violence, law enforcement should be abolished.
Figure A-7, however, does not prove the Anarchists’ point. The graph simply
shows that more dangerous cities have more police officers. The explanation may
be that more dangerous cities hire more police. In other words, rather than police

Figure A-6 Risk of

Graph with an Omitted Variable Cancer

The upward-sloping curve shows

that members of households with
more cigarette lighters are more
likely to develop cancer. Yet we
should not conclude that ownership
of lighters causes cancer because
the graph does not take into 0 Number of Lighters in House
account the number of cigarettes
smoked.

22718_Ch02_ptg01.indd 42 14/09/22 10:48 AM

 Chapter 2 Appendix 43

Figure A-7
Graph Suggesting Reverse
Causality
The upward-sloping curve
shows that cities with a higher
concentration of police are more
dangerous. Yet the graph does not
tell us whether police cause crime
or crime-plagued cities hire more 0
police.

causing crime, crime may cause police. We could avoid the danger of reverse cau-
sality by running a controlled experiment. In this case, we would randomly assign
different numbers of police to different cities and then examine the correlation
between police and crime. Without such an experiment, establishing the direction
of causality is difficult at best.
It might seem that we could determine the direction of causality by examining
which variable moves first. If crime increases and then the police force expands, we
reach one conclusion. If the police force expands and then crime increases, we reach
the other conclusion. This approach, however, is also flawed: Often, people change
their behavior not in response to a change in their present conditions but in response
to a change in their expectations about future conditions. A city that expects a major
crime wave in the future, for instance, might hire more police now. This problem
is easier to see in the case of babies and minivans. Couples often buy a minivan
in anticipation of the birth of a child. The minivan comes before the baby, but we
wouldn’t want to conclude that the sale of minivans causes the population to grow!
There is no complete set of rules that says when it is appropriate to draw causal
conclusions from graphs. Yet just keeping in mind that cigarette lighters don’t cause
cancer (omitted variable) and that minivans don’t cause larger families (reverse
causality) will keep you from falling for many faulty economic arguments.

22718_Ch02_ptg01.indd 43 14/09/22 10:48 AM