Introduction

I.1 What is econometrics?

In a literal interpretation, econometrics means 'economic measurement'. Although the measurement
It is an important part of econometrics, its scope is much broader, as shown by the se-
following quotes:

Econometrics, resulting from a certain perspective on the role of economics, consists of the application
from mathematical statistics to economic data to provide empirical support for the models formulated by
mathematical economics and obtaining numerical results.1

Econometrics can be defined as the quantitative analysis of economic phenomena that occur.
based on the parallel development of theory and observations and with the use of inference methods
suitable ones.2

Econometrics can be defined as the social science in which the tools of economic theory, of
mathematics and statistical inference are applied to the analysis of economic phenomena.3

Econometrics relates to the empirical determination of economic laws.4

The art of the econometrician is to find the set of sufficiently specific and realistic hypotheses.
tasks that allow you to make the best use of the data you have.5

The econometricians [...] are a positive aid in the attempt to dispel the negative public image of
economics (whether quantitative or not) as a subject in which empty boxes are opened assuming
existence of can openers to reveal contents that ten economists will interpret in 11 ways
different.6

The method of econometric research essentially aims at the conjunction of economic theory with
concrete decisions, using the theory and technique of statistical inference as a bridge.7

1TinTner, Gerhard. Methodology of mathematical economics and econometrics. Chicago: The University of Chicago.
go Press, 1968. p. 74.
2SamUelSon, P.a.; KooPmanS, T.C.; STone, J. r. n. Report of the evaluative committee for Econometrica. Economé-

[Link]. 1954, v. 22, n. 2, p. 141-146.

3Goldberger, Arthur S. Econometric Theory. New York: John Wiley & Sons, 1964. p. 1.
4Theil, H. Principles of econometrics. New York: John Wiley & Sons, 1971. p. 1.
5Malinvaud, E. Statistical Methods of Econometrics. Chicago: Rand McNally, 1966. p. 514.
6darnell,adrianC.; evanS, J. [Link] limits of [Link], inglaterra: edward elgarPublishing,

1990. p. 54.
7haavelmo,T.“Theprobability approachin econometrics”. Suplemento daEconometrica.1944. v. 12, prefácio

[Link].

25
26ecobasic nomenclature

I.2 Why a separate discipline?

As the definitions presented suggest, econometrics is an amalgam of economic theory,
mathematical economics, economic statistics, and mathematical statistics. However, the subject deserves to be
studied independently for the following reasons.
Economic theory makes statements or hypotheses primarily of a qualitative nature. For example-
the microeconomic theory states that, all else being equal, a reduction in price
an increase in the price of a commodity should result in an increase in the quantity demanded for that commodity.
Thus, economic theory posits a negative or inverse relationship between price and quantity of-
dispatch of a merchandise. But the theory itself does not offer any quantitative measure of the
relationship between the two variables; it does not inform us how much the quantity will increase or decrease in
consequence of a certain variation in the price of the merchandise. It is up to the econometrician to provide
such numerical estimates. In other words, the econometrician provides practical content to the greater
part of economic theory.
The main concern of mathematical economics is to express economic theory in a way that...
theme (equations) without considering whether the theory can be measured or verified empirically. A
Econometrics, as already mentioned, is primarily concerned with the verification of economic theory.
As we will see, the econometrician often uses the mathematical equations formulated by
mathematical economist, but applies them in a way that can be tested in practice. And this conversion
Transforming mathematical equations into econometric equations requires a lot of ingenuity and skill.
Economic statistics primarily seeks the collection, processing, and presentation of data.
economic data in the form of graphs and tables. This is the task of the economic statistician. He is the main
responsible for collecting data on gross national product (GNP), employment, unemployment, the
prices etc. The collected information constitutes the raw data of the econometric work. But the tra-
the work of the economic statistician does not go beyond this; its focus is not to use data to test theories
economic. Of course, if it were, he would become an econometrician.
Although mathematical statistics provides many of the tools used in its activity, the
economists in general need special methods due to the specific nature of most of
economic data, that is, because they are data that were not generated through controlled experiments
sides. The econometrician, like the meteorologist, generally depends on data that cannot be
directly controlled. As Spanos correctly observes:
In econometrics, those who model often encounter data from observations in
oposição aos [Link] tem duas implicações importantes para a modelagem empírica
in econometrics. First, those who model must master skills that are very different from those required
to the analysis of experimental data [...]. Secondly, the difference between those who collect data and those who analyze it.
requires that those who model are deeply familiar with the nature and structure of the data
in question.8

I.3 The econometric methodology

How do econometricians analyze an economic problem? What methodology do they use?
But although there are several schools of thought on econometric methodology, here I present...
we are traditionalclassical, which still dominates research in economics and other social sciences
and behavioral.9

8SPanoS,aris. Probability. Theory and statistical inference: econometric modeling with observational data. kingdom
United: Cambridge University Press, 1999. p. 21.
9For an enlightening discussion, although advanced, of econometric methods, see Hendry, David F. Dynamic

[Link] York: Oxford University Press, [Link] also SPanoS, aris op. cit.
 introduction27

In general terms, the traditional econometric methodology follows these steps:

1. Presentation of the theory or hypothesis.
2. Specification of the mathematical model of the theory.
3. Specification of the statistical or econometric model.
4. Data collection.
5. Estimation of the parameters of the econometric model.
6. Hypothesis testing.
7. Projection or forecast.
8. Use of the model for control or policy purposes.

To illustrate these steps, let's take a look at the well-known Keynesian consumption theory.

1. exposition of the theory or hypothesis

Keynes affirmed:

The fundamental psychological law [...] is that men [women] are willing, as a rule and in
average, increasing its consumption as its income increases, but not in the same proportion as the au-
income increase.10

In summary, Keynes postulated that the marginal propensity to consume (MPC), the rate of change
the consumption change for a change of one unit (say, one dollar) of income is greater than zero, but
less than 1.

2. specification of the mathematical model of the theory

Although Keynes postulated a positive relationship between consumption and income, he did not specify the
exact form of the functional relationship between the two variables. To simplify, an economist mated
One could suggest the following form for the Keynesian consumption function:

YDØ1CØ2X 0 <Ø2< 1 (i.3.1)

in which consumption expenses and yield1 e Ø2known as the parameters of

modelo, são, respectivamente, ointerceptoe o coeficienteangular.
The angular coefficient, Ø2, with a PMC. Figure 1.1 shows the geometric representation of
Equation (I.3.1). This equation, which specifies that consumption is linearly related to income, is
an example of a mathematical model of the relationship between consumption and income and is known as a function
consumption in economy. The model is just a set of mathematical equations. If the model
there is only one equation, as presented, it is called a uniequational model, while
if there is more than one equation it will be called a multiple equations model (which we will see
but further ahead).
In Equation (I.3.1), the variable that appears on the left side of the equality sign is called
The dependent variable and the variable(s) on the right side are called variable(s)
independent(s) or explanatory. Thus, in the Keynesian consumption function, consumption (expenditure) is the
the dependent variable is income and the explanatory variable is.

10KeYneS, [Link] general theory of employment, interest and [Link] York:harcourtbraceJova-

novich, 1936. p. 96.
 introduction29

Figure i.2 Y
Econometric model
of the consumption function
Keynesian.

X
Income

in the period 1960-2005. In the table, the variable Y corresponds to personal consumption expenditures
(DCP) aggregated (that is, for the economy as a whole) and the variable X to the gross domestic product
gross (GDP), an indicator of aggregate income, both measured in terms of billions of dollars
de 2000. Portanto, os dados são apresentados em termos “reais”, isto é, foram medidos a pre-
constant cycles (from 2000). The data is graphically represented in Figure I.3 (compare
with Figure I.2). For now, let's set aside the line drawn on the graph.

5. estimation of the parameters of the econometric model

Now that we have the data, our next task is to estimate the parameters of the consumption function.
The numerical estimation of the parameters provides empirical content to the consumption function.
the mechanism for estimating the parameters will be examined in Chapter 3. For now, note that
The statistical technique of regression analysis is the main tool for obtaining estimates.
Applying this technique to the data in Table I.1, we obtain the following estimates of Ø.1Ø2, e
specifically, — 299.5913 and 0.7218. Therefore, the estimated consumption function is:

OtD °299,5913
Y C 0.7218Xt (i.3.3)

The circumflex accent over the Y indicates that it is an estimate.11Figure I.3 shows the
estimated consumption function (that is, the regression line).
As shown in Figure I.3, the regression line fits well to the data, in the sense that
The points on the graph that represent the data are very close to the regression line. A
The figure shows us that, for the period 1960-2005, the angular coefficient (aPMC) was almost
0.72, indicating that, in the sampled period, an increase of one dollar in real income led to
average, an increase of about 72 cents in actual consumption expenses.12They say
the relationship between consumption and income is inaccurate; as is clear in Figure I.3, neither
all the data points are exactly on the regression line. In simple terms,
we can say that, according to our data, average consumption expenses are increasing
about 70 cents for each real increase of one dollar in real income.

11The use of a circumflex accent over a variable or parameter indicates, by convention, that it is a
estimated value.
12For now, don't worry about how these values were obtained. As we will show in Chapter 3,

These estimates were obtained through the statistical method of least squares. Also, for when-
So, don't worry about the negative value of the intercept.
30basic econometrics

Table i.1 yes DCP(Y) Pib(X)

Consumption expenses 1960 1597.4 2501.8
people (Y) and product
gross domestic product (X) 1961 1630.3 2560.0
1960-2005 in billions 1962 1711.1 2715.2
from 2000 dollars 1963 1781.6 2834,0
Source: Economics Report of 1964 1888.4 2998.6
the President, 2007, Table
B-2, p, 230.
1965 2007,7 3191.1
1966 2121.8 3399.1
1967 2185.0 3484,6
1968 2310.5 3652.7
1969 2396.4 3765.4
1970 2451.9 3771.9
1971 2545.5 3898.6
1972 2701.3 4105,0
1973 2833,8 4341.5
1974 2812.3 4319.6
1975 2876,9 4311.2
1976 3035.5 4540.9
1977 3164.1 4750.5
1978 3303.1 5015.0
1979 3383,4 5173,4
1980 3374.1 5161.7
1981 3422.2 5291.7
1982 3470.3 5189,3
1983 3668.6 5423.8
1984 3863.3 5813,6
1985 4064.0 6053.7
1986 4228,9 6263.6
1987 4369.8 6475.1
1988 4546.9 6742,7
1989 4675.0 6981.4
1990 4770.3 7112,5
1991 4778.4 7100.5
1992 4934.8 7336.6
1993 5099.8 7532.7
1994 5290,7 7835.5
1995 5433.5 8031.7
1996 5619,4 8328,9
1997 5831.8 8703.5
1998 6125,8 9066.9
1999 6438.6 9470.3
2000 6739.4 9817.0
2001 6910.4 9890.7
2002 7099.3 10048.8
2003 7295.3 10301,0
2004 7577.1 10703,5
2005 7841.2 11048.6
 introduction31

Figure i.3 8000

Consumer spending
people (Y) in relation
to GDP (X), 1960- 7000
2005 in billions
2000 dollars.
6000

5000

4000

3000

2000

1000
2000 4000 6000 8000 10,000 12,000
GDP (X)

6. Tested hypotheses
Assuming that the fitted model is a reasonably good approximation of reality, it is
I need to develop appropriate criteria to verify if the estimates obtained, let's say, in the Equation-
Section (I.3.3) is in accordance with the expectations of the theory being tested. According to economists-
"positives" like Milton Friedman, a theory or hypothesis that cannot be verified with evidence
empirical evidence may not be admissible as part of scientific research.13
As previously noted, Keynes expected that the marginal propensity to consume would be positive, but lower than
1. In our example, the PMC is around 0.72. However, before we accept this value as
a confirmation of Keynesian consumption theory, we need to ask ourselves if this estimate is
sufficiently below unity to convince us that it is not a result due to chance
or a peculiarity of the data we use. In other words, 0.72 is statistically lower
What 1? If so, it will support Keynes' theory.
Such confirmation or refutation of economic theories based on sample evidence is foundational
it is in a branch of statistical theory known as statistical inference (hypothesis testing).
In the course of the book, we will see how this process is conducted in practice.

7. Projection or forecast
If the chosen model does not refute the hypothesis or theory considered, we can use it to
predict the future value(s) of the dependent variable Y, based on the
known or expected future value(s) of the predictive variable X, or explanatory variable.

13seeFriedman, Milton. 'The methodology of positive economics.' Essays in Positive Economics. Chicago: University
of Chicago Press, 1953.
32basic econometrics

For illustration purposes, suppose we want to predict average consumption expenses for
2006. O valor do PIB nesse ano foi de $ 11319,4 bilhão.14Placing the GDP value on the right side
From Equation (I.3.3), we obtain:

O2006D °299,5913C0,7218(11319,4)
Y (i.3.4)
D7870,7516

about $7.870 billion. Thus, given the value of GDP, the average projected consumption expenditures
is about $7.870 billion. The value of these expenses effectively recorded in 2006 was
$8.044 billion. Therefore, the estimated model (I.3.3) underestimated real consumption expenditures by
about $174 billion. We can say that the forecast error is about $174 billion, which is approximately-
about 1.5% of the GDP value recorded in 2006. When we examine more closely the mo-
In the next chapters on linear regression, we will try to verify if such an error is 'small.'
or "great". But the important thing now is to observe that these prediction errors are inevitable, given the
statistical nature of our analysis.
There is another use for the model estimated in Equation (I.3.3). Suppose the president decides to
for a reduction in the income tax rate. What would be the effect of this policy on income and,
therefore, about consumption expenses and, finally, about employment?
Assume that, as a result of the proposed change, investment expenses increase.
What would be the effect on the economy? As macroeconomic theory shows, the change in income
What follows, let's say, from the variation of a dollar in investment expenses is given by the multiple
framework of income, which is defined as:

1 (i.3.5)
MH
1°PMC

If we use the MPC of 0.72 obtained in Equation (I.3.3), this multiplier will be about
MD3,57. That is, an increase (decrease) of one dollar in investment will ultimately lead to an increase
(reduction) of more than three times in income; note that the multiplier takes time to produce its effect.
A critical value in these calculations is the PMC, since the multiplier depends on it. And this estimate
PMC can be obtained through regression models such as that of Equation (I.3.3). The estimates
The quantitative measures of the PMC provide valuable information for the formulation of economic policy.
mica. By knowing the PMC, we can predict the future course of income, consumer spending, and
employment after a change in the government's fiscal policy.

8. use of the model for control or policy purposes

Suppose we have estimated the consumption function from Equation (I.3.3). Also suppose that the
the government believes that consumption expenditures are about $8.750 billion (in 2000 dollars)
they will maintain the unemployment rate at its current level of about 4.2% (in early 2006). What level
Does the income guarantee the desired amount (target) of consumption expenses?
If the results of the regression of Equation (I.3.3) seem reasonable, an arithmetic calculation yes-
please show that:

8.750D °299.5913C0.7218 (GDP2006 ) (i.3.6)

which gives approximately $1.2537. In other words, a income level of about $1.2537 billion, given
A PMC of about 0.72 will generate an expense of about $8.750 billion.

14the dodCP data and the PIB for 2006 were available, but we deliberately set them aside to
illustrate the topic examined in this section. As we will see in subsequent chapters, it is a good idea to save
a part of the data to verify how the fitted model predicts the observations that are out of the sample.
 introduction33

As these calculations suggest, an estimated model can be used for control purposes or
policy formulation. With an appropriate combination of fiscal and monetary policies, the government
cannot manage the control variable X to generate the desired level of the target variable Y.
Figure I.4 summarizes the anatomy of classical econometric modeling.

model selection
When a government agency (for example, the Department of Commerce of the United States)
(two) collects data as presented in Table I.1, does not necessarily have an economic theory.
any mica in mind. So how do we know that the data really confirms the theory of
Keynesian consumption? It would be because the Keynesian consumption function (the regression line) of the figure
Is I.3 extremely close to the available data? Is it possible that another model (theory) of the con-
Does it fit the data equally well? For example, Milton Friedman developed a model
of consumption, called the permanent income hypothesis.15Robert Hall also formulated a model-
the consumption, known as the permanent income hypothesis in the life cycle.16Any of these
Could models, or both, also serve for those in Table I.1?
In summary, the question that the researcher faces in practice is how to choose between the diffe-
rent hypotheses or models for a given phenomenon, such as the consumption-income relationship. As Miller
argument

No encounter with the data is a step towards authentic confirmation unless the hi-
It is better to handle data than some natural rival [...]. What strengthens a hypothesis, in this
it is the victory that, at the same time, is the defeat for another plausible hypothesis.17

How, then, to choose between the various competing models or hypotheses? Is it worth having in
mind the advice of Clive Granger:18

figure i.4 Economic theory

Anatomy of
modeling
Mathematical model of the theory
econometric.

Econometric models of the theory

Data

Estimation of the econometric model

Hypothesis testing

Projection or forecast

Using the model for purposes

of control or of policy

15Friedman,milton.A theory of consumption [Link]:n. J.: Princeton University Press, 1957.

16Hall, R. "Stochastic implications of the life cycle permanent income hypothesis: theory and evidence." Journal
of Political Economy, 1978. v. 86, p. 971-987.
17Miller, R. W. Fact and method: explanation, confirmation, and reality in the natural and social sciences.

ceton, n.J.: Princeton University Press, 1978. p. 176.

18GranGer, Clive W. [Link] modeling in [Link] Unido: Cambridge University Press, 1999. p. 58.
34basic econometrics

I would like to suggest that, in the future, you ask the following questions when you are presented with a new one.
theory or empirical model:

(i) What is the purpose of this? What economic decision will it contribute to? and;
(ii) Is there any evidence that allows me to assess its quality in comparison to theories or models
alternatives?
I think that if due attention is given to these questions, economic research and discussion
will be strengthened.

As we progress through the book, we will encounter several hypotheses that compete to explain
the various economic phenomena. For example, economics students are well acquainted with the concept
to the production function, which is basically a relationship between output and inputs (capital and labor)
In literature, two of the most well-known are the Cobb-Douglas functions and elasticity.
of constant substitution. We will need to discover, due to the production data and of insu-
but if any of them better reflects the data.
The classic eight-step econometric method presented earlier is neutral in the sense of
what can be used to test any of these rival hypotheses.
It is possible to formulate a methodology that is sufficiently comprehensive to include hypotheses.
Competitors? This is a complex and controversial topic that will be discussed in Chapter 13, later.
that we have acquired sufficient theoretical knowledge.

I.4 Types of Econometrics

As suggested by the classificatory structure of Figure I.5, econometrics can be divided into two
broad categories: theoretical econometrics and applied econometrics. In each category we can
address the discipline according to classical and Bayesian traditions. In this book, we emphasize the classic
Sica. As for the Bayesian approach, the reader can refer to the references indicated at the end of
chapter.
Theoretical econometrics deals with the development of adequate methods to measure the relationships
economic actions specified in the econometric models. In this regard, econometrics
depends heavily on mathematical statistics. For example, one of the most used methods in this
The book is the method of least squares. Theoretical econometrics must make clear the hypotheses of this.
method, its properties and what happens to them when one or more hypotheses of the method
they are not met.
In applied econometrics, we use the tools of theoretical econometrics to study one or
but special fields of economics and business, such as the production function, the investment function
mento, the functions of supply and demand, portfolio theory, etc.
This book largely deals with the formulation of econometric methods, their assumptions,
uses, limitations. These methods are illustrated with examples from various areas of the economy and of
businesses. However, this is not a book on applied econometrics, in the sense of delving into it.
particularly in any of the fields of economic application. This task is up to the works of
specifically written for this purpose. At the end of the book, we will offer some bibliographic references.
graphs.

figure i.5 Econometrics

Categories of
econometrics.
Theoretical Applied

Classical Bayesian Classic Bayesian

 introduction35

I.5 Mathematical and statistical prerequisites

Although this book is written at an elementary level, the author assumes that the reader is...
already familiar with the basic concepts of statistical estimation and hypothesis testing. The
Appendix A provides an overview and non-technical explanation of basic statistical concepts employed
This is for those who wish to recycle their knowledge. Regarding mathematics,
It is desirable to have an understanding of differential calculus, although it is not essential. Despite the
Most econometrics books aimed at graduate studies widely use algebra.
matrix, I want to make it clear that it is not necessary to benefit from this book. I am
completely convinced that the fundamental ideas of econometrics can be trans-
mitigated without resorting to matrix algebra. However, for students more inclined towards mathematics,
Appendix C presents a summary of the basic theory of regression in matrix notation and Appendix B
provides a summary of the main results of matrix algebra.

I.6 The role of the computer

Regression analysis, the main tool of econometrics, is unthinkable today without
the computer and access to some statistical software or statistical packages. (Believe me,
I come from the generation raised with the slide rule!) Fortunately, several excellent packages for regression -
They are available in the market, both for large computers and for microcomputers.
computers, and the list grows every day. Software like ET, LIMDEP, SHAZAM, MICRO TSP,
MINITAB, EVIEWS, SAS, SPSS, STATA, Microfit, PcGive e BMD provide for the majority of the techniques
concepts and econometric tests examined in this book.
Throughout the book, the reader will be invited from time to time to conduct Monte Carlo experiments.
with the help of one or more statistical packages. Monte Carlo experiments are 'fun' and
will allow the reader to appreciate the properties of various statistical methods used here. The deta-
The results of the Monte Carlo experiments will be discussed at the appropriate time.

I.7 Suggestions for additional readings

The topic of econometric methodology is vast and controversial. For those interested, I suggest the books
next:

MARCHI, Neil de; GILBERT, Christopher. (Eds.). History and methodology of econometrics. Nova
York:OxfordUniversityPress,[Link]
econometric methodology and extensively examines the British approach to econometrics and its
relationship with time series, that is, data collected over time.
CHAREMZA, Wojciech W.; DEADMAN, Derek F. New directions in econometric practice: gen-
eral to specific modelling, cointegration and vector autogression.2. ed. Hants, England: Edward Elgar
Publishing Ltd., 1997. The authors criticize the traditional approach of econometrics and present
a detailed exposition of the new approaches to the econometric method.
DARNELL, Adrian C.; EVANS, J. Lynne. The limits of econometrics. Hants, England: Edward
Elgar Publishers Ltd., 1990. This book offers a fairly balanced examination of the various approaches
econometric methodologies, with a renewed fidelity to the traditional method.
MORGAN,[Link]:CambridgeUniversityPress,1990.A
author offers an excellent historical perspective of econometric theory and practice, with a
in-depth examination of the initial contributions of Haavelmo (winner of the Nobel Prize in Economics in 1990)
to econometrics. In the same spirit, the book by David F. Hendry and Mary S. Morgan, The foundation of
econometric analysis, United Kingdom: Cambridge University Press, 1995, brings together a selection of tex-
seminal texts to show the evolution of econometric ideas over time.
36basic econometrics

COLANDER, David; BRENNER, Reuven. (Eds.). Educating economists. Ann Arbor, Michigan:
University of Michigan Press, 1992. The book presents a critical, sometimes agnostic, view of
education and economic practice.
For those interested in Bayesian statistics and econometrics, the following books are very
useful: DEY, John H. Data in doubt. England: Basil Blackwell Ltd., Oxford University Press, 1985;
Peter,[Link]:[Link]:OxfordUniversityPress,1989;andPORIER,
Dale [Link] statistics and econometrics:a comparative [Link], Massachusetts:
MIT Press, 1995. ZELLER, Arnold. An introduction to Bayesian inference in econometrics. New York:
John Wiley & Sons, 1971, this is an advanced reference book. Another advanced reference book is
Palgrave Handbook of Econometrics: Volume 1: Econometric Theory, edited by Terence C. Mills
e Kerry Patterson, New York: Palgrave Macmillan, 2007.