The Nature and Scope of Econometrics: Confirming Pages
The Nature and Scope of Econometrics: Confirming Pages
The Nature and Scope of Econometrics: Confirming Pages
Confirming Pages
CHAPTER
1
THE NATURE
AND SCOPE
OF ECONOMETRICS
Econometrics, the result of a certain outlook on the role of economics, consists of the
application of mathematical statistics to economic data to lend empirical support to
the models constructed by mathematical economics and to obtain numerical results.2
1
Arthur S. Goldberger, Econometric Theory, Wiley, New York, 1964, p. 1.
2
P. A. Samuelson, T. C. Koopmans, and J. R. N. Stone, Report of the Evaluative Committee for
Econometrica, Econometrica, vol. 22, no. 2, April 1954, pp. 141146.
1
guj75845_ch01.qxd 3/9/09 3:52 PM Page 2
Confirming Pages
2 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 3
may be called upon to develop special methods of analysis to deal with such
errors of measurement.
For students majoring in economics and business there is a pragmatic reason
for studying econometrics. After graduation, in their employment, they may be
called upon to forecast sales, interest rates, and money supply or to estimate de-
mand and supply functions or price elasticities for products. Quite often, econo-
mists appear as expert witnesses before federal and state regulatory agencies on
behalf of their clients or the public at large. Thus, an economist appearing before
a state regulatory commission that controls prices of gas and electricity may be re-
quired to assess the impact of a proposed price increase on the quantity de-
manded of electricity before the commission will approve the price increase. In
situations like this the economist may need to develop a demand function for
electricity for this purpose. Such a demand function may enable the economist to
estimate the price elasticity of demand, that is, the percentage change in the quan-
tity demanded for a percentage change in the price. Knowledge of econometrics
is very helpful in estimating such demand functions.
It is fair to say that econometrics has become an integral part of training in
economics and business.
Confirming Pages
4 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
Collecting Data
For empirical purposes, therefore, we need quantitative information on the two
variables. There are three types of data that are generally available for empirical
analysis.
1. Time series.
2. Cross-sectional.
3. Pooled (a combination of time series and cross-sectional).
Times series data are collected over a period of time, such as the data on
GDP, employment, unemployment, money supply, or government deficits.
Such data may be collected at regular intervalsdaily (e.g., stock prices),
weekly (e.g., money supply), monthly (e.g., the unemployment rate), quarterly
(e.g., GDP), or annually (e.g., government budget). These data may be quanti-
tative in nature (e.g., prices, income, money supply) or qualitative (e.g., male or
female, employed or unemployed, married or unmarried, white or black). As
we will show, qualitative variables, also called dummy or categorical variables,
can be every bit as important as quantitative variables.
Cross-sectional data are data on one or more variables collected at one point
in time, such as the census of population conducted by the U.S. Census Bureau
every 10 years (the most recent was on April 1, 2000); the surveys of consumer
expenditures conducted by the University of Michigan; and the opinion polls
such as those conducted by Gallup, Harris, and other polling organizations.
In pooled data we have elements of both time series and cross-sectional data.
For example, if we collect data on the unemployment rate for 10 countries for a
period of 20 years, the data will constitute an example of pooled datadata on
the unemployment rate for each country for the 20-year period will form time se-
ries data, whereas data on the unemployment rate for the 10 countries for any
single year will be cross-sectional data. In pooled data we will have 200
observations20 annual observations for each of the 10 countries.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 5
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 5
There is a special type of pooled data, panel data, also called longitudinal or
micropanel data, in which the same cross-sectional unit, say, a family or firm, is
surveyed over time. For example, the U.S. Department of Commerce conducts
a census of housing at periodic intervals. At each periodic survey the same
household (or the people living at the same address) is interviewed to find out
if there has been any change in the housing and financial conditions of that
household since the last survey. The panel data that result from repeatedly in-
terviewing the same household at periodic intervals provide very useful infor-
mation on the dynamics of household behavior.
3
We consider here only the aggregate CLFPR and CUNR, but data are available by age, sex, and
ethnic composition.
4
On this, see Shelly Lundberg, The Added Worker Effect, Journal of Labor Economics, vol. 3,
January 1985, pp. 1137.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 6
Confirming Pages
6 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
through the scatter points and write the relationship between CLFPR and
CUNR by the following simple mathematical model:
CLFPR = B1 + B2 CUNR (1.1)
Equation (1.1) states that CLFPR is linearly related to CUNR. B1 and B2 are known
as the parameters of the linear function.5 B1 is also known as the intercept; it
5
Broadly speaking, a parameter is an unknown quantity that may vary over a certain set of val-
ues. In statistics a probability distribution function (PDF) of a random variable is often character-
ized by its parameters, such as its mean and variance. This topic is discussed in greater detail in
Appendixes A and B.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 7
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 7
67.0
66.5
CLFPR (%)
66.0
65.5
65.0
64.5
64.0
63.5
3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5
CUNR (%)
FIGURE 1-1 Regression plot for civilian labor force participation
rate (%) and civilian unemployment rate (%)
gives the value of CLFPR when CUNR is zero.6 B2 is known as the slope. The
slope measures the rate of change in CLFPR for a unit change in CUNR, or more gen-
erally, the rate of change in the value of the variable on the left-hand side of the
equation for a unit change in the value of the variable on the right-hand side.
The slope coefficient B2 can be positive (if the added-worker effect dominates
the discouraged-worker effect) or negative (if the discouraged-worker effect
dominates the added-worker effect). Figure 1-1 suggests that in the present case
it is negative.
6
In Chapter 2 we give a more precise interpretation of the intercept in the context of regression
analysis.
7
We even tried to fit a parabola to the scatter points given in Fig. 1-1, but the results were not
materially different from the linear specification.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 8
Confirming Pages
8 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
line specified by the mathematical model, Eq. (1.1)? Remember that our data on
labor force and unemployment are nonexperimentally collected. Therefore, as
noted earlier, besides the added- and discouraged-worker hypotheses, there
may be other forces affecting labor force participation decisions. As a result, the
observed relationship between CLFPR and CUNR is likely to be imprecise.
Let us allow for the influence of all other variables affecting CLFPR in a
catchall variable u and write Eq. (1.2) as follows:
where u represents the random error term, or simply the error term.8 We let u
represent all those forces (besides CUNR) that affect CLFPR but are not explic-
itly introduced in the model, as well as purely random forces. As we will see in
Part II, the error term distinguishes econometrics from purely mathematical
economics.
Equation (1.2) is an example of a statistical, or empirical or econometric, model.
More precisely, it is an example of what is known as a linear regression model,
which is a prime subject of this book. In such a model, the variable appearing on
the left-hand side of the equation is called the dependent variable, and the vari-
able on the right-hand side is called the independent, or explanatory, variable.
In linear regression analysis our primary objective is to explain the behavior of
one variable (the dependent variable) in relation to the behavior of one or more
other variables (the explanatory variables), allowing for the fact that the rela-
tionship between them is inexact.
Notice that the econometric model, Eq. (1.2), is derived from the mathemati-
cal model, Eq. (1.1), which shows that mathematical economics and economet-
rics are mutually complementary disciplines. This is clearly reflected in the
definition of econometrics given at the outset.
Before proceeding further, a warning regarding causation is in order. In the
regression model, Eq. (1.2), we have stated that CLFPR is the dependent vari-
able and CUNR is the independent, or explanatory, variable. Does that mean
that the two variables are causally related; that is, is CUNR the cause and CLFPR
the effect? In other words, does regression imply causation? Not necessarily. As
Kendall and Stuart note, A statistical relationship, however strong and how-
ever suggestive, can never establish causal connection: our ideas of causation
must come from outside statistics, ultimately from some theory or other.9 In
our example, it is up to economic theory (e.g., the discouraged-worker hypoth-
esis) to establish the cause-and-effect relationship, if any, between the depen-
dent and explanatory variables. If causality cannot be established, it is better to
call the relationship, Eq. (1.2), a predictive relationship: Given CUNR, can we pre-
dict CLFPR?
8
Instatistical lingo, the random error term is known as the stochastic error term.
9
M. G. Kendall and A. Stuart, The Advanced Theory of Statistics, Charles Griffin Publishers, New
York, 1961, vol. 2, Chap. 26, p. 279.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 9
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 9
Note that we have put the symbol on CLFPR (read as CLFPR hat) to remind
us that Eq. (1.3) is an estimate of Eq. (1.2). The estimated regression line is shown
in Figure 1-1, along with the actual data points.
As Eq. (1.3) shows, the estimated value of B1 is L 69.5 and that of B2 is L 0.58,
where the symbol L means approximately. Thus, if the unemployment rate
goes up by one unit (i.e., one percentage point), ceteris paribus, CLFPR is ex-
pected to decrease on the average by about 0.58 percentage points; that is, as eco-
nomic conditions worsen, on average, there is a net decrease in the labor force
participation rate of about 0.58 percentage points, perhaps suggesting that the
discouraged-worker effect dominates. We say on the average because the
presence of the error term u, as noted earlier, is likely to make the relationship
somewhat imprecise. This is vividly seen in Figure 1-1 where the points not on
the estimated regression line are the actual participation rates and the (vertical)
distance between them and the points on the regression line are the estimated
us. As we will see in Chapter 2, the estimated us are called residuals. In short,
the estimated regression line, Eq. (1.3), gives the relationship between average
CLFPR and CUNR; that is, on average how CLFPR responds to a unit change in
CUNR. The value of about 69.5 suggests that the average value of CLFPR will
be about 69.5 percent if the CUNR is zero; that is, about 69.5 percent of the civil-
ian working-age population will participate in the labor force if there is full
employment (i.e., zero unemployment).10
10
This is, however, a mechanical interpretation of the intercept. We will see in Chapter 2 how to
interpret the intercept term meaningfully in a given context.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 10
Confirming Pages
10 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
be an important decision variable. In the short run at least, a higher wage may at-
tract more workers to the labor market, other things remaining the same (ceteris
paribus). To see its importance, in Table 1-1 we have also given data on real average
hourly earnings (AHE82), where real earnings are measured in 1982 dollars. To
take into account the influence of AHE82, we now consider the following model:
These results are interesting because both the slope coefficients are negative.
The negative coefficient of CUNR suggests that, ceteris paribus (i.e., holding the
influence of AHE82 constant), a one-percentage-point increase in the unem-
ployment rate leads, on average, to about a 0.64-percentage-point decrease in
CLFPR, perhaps once again supporting the discouraged-worker hypothesis. On
the other hand, holding the influence of CUNR constant, an increase in real
average hourly earnings of one dollar, on average, leads to about a 1.44 percentage-
point decline in CLFPR.11 Does the negative coefficient for AHE82 make eco-
nomic sense? Would one not expect a positive coefficientthe higher the hourly
11
As we will discuss in Chapter 4, the coefficients of CUNR and AHE82 given in Eq. (1.5) are
known as partial regression coefficients. In that chapter we will discuss the precise meaning of partial
regression coefficients.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 11
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 11
earnings, the higher the attraction of the labor market? However, one could
justify the negative coefficient by recalling the twin concepts of microeconomics,
namely, the income effect and the substitution effect.12
Which model do we choose, Eq. (1.3) or Eq. (1.5)? Since Eq. (1.5) encompasses
Eq. (1.3) and since it adds an additional dimension (earnings) to the analysis, we
may choose Eq. (1.5). After all, Eq. (1.2) was based implicitly on the assumption
that variables other than the unemployment rate were held constant. But where
do we stop? For example, labor force participation may also depend on family
wealth, number of children under age 6 (this is especially critical for married
women thinking of joining the labor market), availability of day-care centers for
young children, religious beliefs, availability of welfare benefits, unemploy-
ment insurance, and so on. Even if data on these variables are available, we may
not want to introduce them all in the model because the purpose of developing
an econometric model is not to capture total reality, but just its salient features.
If we decide to include every conceivable variable in the regression model, the
model will be so unwieldy that it will be of little practical use. The model ulti-
mately chosen should be a reasonably good replica of the underlying reality. In
Chapter 7, we will discuss this question further and find out how one can go
about developing a model.
12
Consult any standard textbook on microeconomics. One intuitive justification of this result is
as follows. Suppose both spouses are in the labor force and the earnings of one spouse rise substan-
tially. This may prompt the other spouse to withdraw from the labor force without substantially
affecting the family income.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 12
Confirming Pages
12 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
Step Example
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 13
13
Some of the proofs and derivations are presented in our Basic Econometrics, 5th ed., McGraw-
Hill, New York, 2009.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 14
Confirming Pages
14 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
QUESTIONS
1.1. Suppose a local government decides to increase the tax rate on residential prop-
erties under its jurisdiction. What will be the effect of this on the prices of resi-
dential houses? Follow the eight-step procedure discussed in the text to answer
this question.
1.2. How do you perceive the role of econometrics in decision making in business
and economics?
1.3. Suppose you are an economic adviser to the Chairman of the Federal Reserve
Board (the Fed), and he asks you whether it is advisable to increase the money
supply to bolster the economy. What factors would you take into account in
your advice? How would you use econometrics in your advice?
1.4. To reduce the dependence on foreign oil supplies, the government is thinking
of increasing the federal taxes on gasoline. Suppose the Ford Motor Company
has hired you to assess the impact of the tax increase on the demand for its cars.
How would you go about advising the company?
1.5. Suppose the president of the United States is thinking of imposing tariffs on im-
ported steel to protect the interests of the domestic steel industry. As an economic
adviser to the president, what would be your recommendations? How would you
set up an econometric study to assess the consequences of imposing the tariff?
PROBLEMS
1.6. Table 1-2 gives data on the Consumer Price Index (CPI), S&P 500 stock index,
and three-month Treasury bill rate for the United States for the years 19802007.
a. Plot these data with time on the horizontal axis and the three variables
on the vertical axis. If you prefer, you may use a separate figure for each variable.
b. What relationships do you expect to find between the CPI and the S&P index
and between the CPI and the three-month Treasury bill rate? Why?
c. For each variable, eyeball a regression line from the scattergram.
TABLE 1-2 CONSUMER PRICE INDEX (CPI, 19821984 = 100), STANDARD AND POORS COMPOSITE
INDEX (S&P 500, 19411943 = 100), AND THREE-MONTH TREASURY BILL RATE (3-m T BILL, %)
Year CPI S&P 500 3-m T bill Year CPI S&P 500 3-m T bill
Source: Economic Report of the President, 2008, Tables B-60, B-95, B-96, and B-74, respectively.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 15
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 15
1.7. Table 1-3 gives you data on the exchange rate between the U.K. pound and the
U.S. dollar (number of U.K. pounds per U.S. dollar) as well as the consumer
price indexes in the two countries for the period 19852007.
a. Plot the exchange rate (ER) and the two consumer price indexes against time,
measured in years.
b. Divide the U.S. CPI by the U.K. CPI and call it the relative price ratio (RPR).
c. Plot ER against RPR.
d. Visually sketch a regression line through the scatterpoints.
1.8. Table 1-4 on the textbook Web site contains data on 1247 cars from 2008.14 Is
there a strong relationship between a cars MPG (miles per gallon) and the
number of cylinders it has?
a. Create a scatterplot of the combined MPG for the vehicles based on the num-
ber of cylinders.
b. Sketch a straight line that seems to fit the data.
c. What type of relationship is indicated by the plot?
14
Data were collected from the United States Department of Energy Web site at http://www.
fueleconomy.gov/.
guj75845_ch01.qxd 3/9/09 3:52 PM Page 16
Confirming Pages
16 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS
Confirming Pages
CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS 17
Confirming Pages
18 CHAPTER ONE: THE NATURE AND SCOPE OF ECONOMETRICS