AN INTRODUCTION TO

ECONOMETRICS
Oxbridge Economics; Mo Tanweer
[email protected]
Econometrics
Econometrics is concerned with the tasks of developing and
applying quantitative or statistical methods to the study of
economic principles
Economics + Statistics = Econometrics
What is econometrics?
Econometrics is the use of statistical techniques to analyse
economic data and compare with economic theory
What makes Econometrics different to Statistics?
Economic data tends to be observational and more complex
One of its aims is to give empirical content to economic theory
Econometrics means economic measurement
Economic theories tend to be qualitative in nature
Examples of econometrics
As economists, we would like to understand the
relationship between economic variables.
Is human capital a fundamental cause of growth?
Do improvements in educational spending by governments
improve academic performance?
Can increases in the price of oil lead to reductions in
national income?
Does an increase in national savings lead to an increase in
investment?
If a government decreases the duration of time it offers
unemployment benefit, does this lead to a lower
unemployment rate?
Do fertility decisions in Pakistan exhibit socio-economic
conforming behaviour?
Regression Analysis
When we consider the nature and form of a
relationship between any two or more variables, the
analysis is referred to as regression analysis.
Theoretical econometrics:
considers questions about the statistical properties of
estimators and tests,
Applied econometrics:
is concerned with the application of econometric
methods to assess economic theories
Descriptive
Forecasting
Causal
- How does the stock market and interest rate move together?
- Will we have a recession next year?
- If we raise the minimum wage, would unemployment
soar?
Data types
- A data set containing observations on a single phenomenon observed
over multiple time periods is called time series
Time-series
Cross-section
Panel / Longtitudinal / TSCS
- Two-dimensional data
- A data set containing observations on multiple phenomena observed
over multiple time periods is called panel data
- A data set containing observations on multiple phenomena observed at
a single point in time is called cross-sectional
- In cross-sectional data sets, the values of the data points have meaning,
but the ordering of the data points does not.
- In time series data, both the values and the ordering of the data points
have meaning
E.g. Incomes for a country
Time-series
Forecasting
Chronology
matters
Interdependency
issue
Seasonal data
Time-series
Forecasting
Chronology
matters
Interdependency
issue
Seasonal data
Cross-sectional
Panel/Longitudinal
The PSID had collected information on more than 70,000
individuals spanning as many as 4 decades of their lives.
measures economic, social, and health factors over the life course and across
generations
The Panel Study of Income Dynamics
Do fertility decisions in Pakistan exhibit socio-economic
conforming behaviour?
Write down a list of data variables
you would like in your model
Statistical Packages
EViews
SPSS
Microfit
Stata
Excel
PcGive
Minitab
Shazam
Correlation vs Causation
Correlation is not necessarily causation.
Post hoc ergo propter hoc / Cum hoc ergo propter hoc
Correlation vs Causation
A occurs in correlation with B.
Therefore, A causes B.
This is a logical fallacy
because there are at least
four other possibilities:
B may be the cause of A
some unknown third
factor C is actually the
cause of both A and B
B may be the cause of A
at the same time as A is
the cause of B
coincidence
Be careful what you infer from your
statistical analyses.
Be sure your relationship makes sense.
Econometrics in practice
The relationship between Income and Consumption.
Economic theory:
Keynes postulated a positive relationship between
consumption and income
General Theory of Employment, Interest and Money.
The fundamental psychological law is that men
[women] are disposed, as a rule and on average, to
increase their consumption as their income increases, but
not as much as the increase in their income
Statistics :
Lets get the data and look at the causal relationship
Income and Consumption
Data on Personal Consumption Expenditure
And Gross Domestic Product;1982-1996)
all in 1992 billions of dollars
C GDP
1982 3081.5 4620.3
1983 3240.6 4803.7
1984 3407.6 5140.1
1985 3566.5 5323.5
1986 3708.7 5487.7
1987 3822.3 5649.5
1988 3972.7 5865.2
1989 4064.6 6062
1990 4132.2 6136.3
1991 4105.8 6079.4
1992 4219.8 6244.4
1993 4343.6 6389.6
1994 4486 6610.7
1995 4595.3 6742.1
1996 4714.1 6928.4
Casual observation suggests that
the relationship between income
and consumption is positive in
that consumption rises as income
rises.
However, we want to analyse
more formally if a relationship
exists.
We use regression analysis to
look at this relationship formally.
3000
3500
4000
4500
5000
5500
6000
6500
7000
7500
3000 3200 3400 3600 3800 4000 4200 4400 4600 4800
Consumption vs Income
Income
Income and Consumption
Disposable Income is the independent variable

C
o
n
s
u
m
p
t
i
o
n

i
s

t
h
e

d
e
p
e
n
d
e
n
t
v
a
r
i
a
b
l
e

)
In regression analysis an
important issue is the direction
of causation between
variables.
Consumption is the dependent
variable (Y)
Disposable Income is the
independent variable (X)
This confirms the positive
relationship but we need to be
more rigorous in our analysis
Let us assume that the relationship between consumption and
income takes the form of the Keynesian consumption function:
Where is between 0-1
The Theory: Keynes
Let us assume that the relationship between consumption and
income takes the form of the Keynesian consumption function:
and are known as the PARAMETERS of the
model (a.k.a the intercept and slope coefficients)
Y is the dependent variable, in this case, consumption
X is the independent variable, in this case, income
is a measure of what?
Terminology
Dependent variable
Explained variable
Predictand
Regressand
Endogenous
Controlled variable
Independent variable
Explanatory variable
Predictor
Regressor
Exogenous
Control variable
Econometrics
is the POPULATION regression equation
The actual consumption Y of a household will not
always equal its expected value E(Y).
Actual consumption of a household may be
disturbed from its expected value by any one of
innumerable factors, and we shall therefore write
actual consumption as:
The disturbance (u) (or, e, error term) represents
the effect on household consumption of all
variables other than income.
Econometrics
The population regression equation is unknown
to any investigator, and remains unknown.
Therefore have to fit a straight line to the
scatter points. This line can then be regarded
as an estimate of the population equation.
The fitted line we write as:
The sample regression equation represents a
straight line with intercept alpha-hat and
slope beta-hat.
Y-hat is known as the predicted value (or
estimate)of Y
X Y

X Y

Y
X
Y
Econometrics
Y
X
X Y

YY
Residual
u X Y

So the residuals (or error terms) are simply the differences
between the actual and estimated Y values.
So we minimise the sum of residuals Sum of U-hat
i
= SUM(Y
i
Y-hat
i
)
making it as small as possible
u Y Y

Y Y u

X B Y u
1

Econometrics
Y
X
X Y

YY
Residual
So we use the LEAST SQUARES METHOD
Sum of (u-hat)
2
= sum (y yhat)
2
= Sum (y -hat
1
hat-X)
2
By squaring, you give more weights to
the bigger errors.
So now it is not possible for the Sum of the u-hats
(the error terms) to be small even if u is widely
spread.
But this wont work e.g. +10, -10; +2, -2 = sum of errors = 0.
= OLS
OBVIOUSLY each time we change and
1
(hats),
we will change the u-hat.
So we need to pick / find a and
1
-hat such
that the u-hat is as small as possible.
OLS
When we fit a sample regression line to a
scatter of points, it makes sense to select a
line (that is, choose values for alpha-hat
and beta-hat) such that the residuals given
by that result are in some sense small.
The most popular and best known way of
ensuring this is to choose alpha-hat and
beta-hat so as to minimise the sum of the
squares of the residuals.
This method of estimating the parameters
alpha and beta is known as the method of
ordinary least squares (OLS).
Provided that a whole series of further
assumptions are valid, the OLS method can
be shown to provide good estimators of
alpha and beta.
Computer software can compute OLS
regressions automatically. The results may
be reported as:
Y
X
X Y

YY
Residual
X Y 812 . 0 71 . 30

Measures of closeness of Fit
The sample regression equation fits the
scatter fairly closely
But fairly closely is a vague expression,
and it is often convenient to have a precise
summary statistic (that is, a single number) by
which we can assess and compare the
closeness of fit of different scatters and
different sample lines.
Y
X
Y
X
R
2
= 0.86
R
2
= 0.58
We ask: What proportion of the variation in
the consumption among our sample of
households can we attribute to the variation
in their incomes?
We define the coefficient of determination
(R
2
) as the proportion of the sample
variation in Y that can be attributed to the
sample variation in X.
The closer the coefficient is to one, the
better the line will fit the points.
Testing for significance
OLS regression will always try to fit a line
through the points
However we want to be able to test if our
coefficient is significantly different from zero.
Does household
disposable income have
a STATISTICALLY
SIGNIFICANT effect on
household consumption?
To do this we calculate a t-value:
t = /s.e.
That is, the t-value is the coefficient value
divided by its standard error (a measure of
variance)
This value can be looked up in statistical
tables. However, as a rule of thumb a variable
is significant if the t-value is greater than 2.
Modelling
Economic Theory
Mathematical model of theory
Econometric model of theory
Data
Estimation of econometric model
Hypothesis Testing
Forecasting or prediction Using the model for policy purposes
Mis-specification
Once you have an economic theory, collected the data, you still have to
come up with a mathematical model to test. Getting the mathematical
model is CRUCIAL
E.g. Data on Demand for Apples
From microeconomic theory, it is known that the demand for a commodity
generally depends on the real income of the consumer, the real price of
the commodity, and the real prices of competing or complementary
commodities.
Specification of the model what shape will this demand curve take?
what variables do I need to put into this specification?
E.g. real income; real price of apples; real price of oranges; real price of
bananas
Do I model it:
Demand for Apples = +
1
(Price of Apples) +
2
(Price of oranges) +

3
(Real income)
Demand for Apples = +
1
Ln(Price of Apples) +
2
(Price of
oranges+Price of bananas) +
3
(Real Income)
2
A word to the Econometrician
It should be clear that model building is an art as well
as a science: The Ten Commandments of Applied
Econometrics (Feldstein)
1) Thou shalt use common sense and economic theory
2) Thou shalt ask the right questions (i.e. put prevalence before mathematical elegance)
3) Thou shalt know the context (do not perform ignorant statistical analysis)
4) Thou shalt inspect the data
5) Thou shalt not worship complexity
6) Thou shalt look long and hard at the results
7) Thou shalt beware the costs of data mining
8) Thou shalt not confuse statistical significance with practical signifiance
9) Thou shalt use common sense and economic theory
10) Thou shalt not underestimate the task of data collection
AN INTRODUCTION TO
ECONOMETRICS
Oxbridge Economics; Mo Tanweer
[email protected]