Beruflich Dokumente
Kultur Dokumente
Department of Economics
Econometrics I
Do.Dr. zgr mer Ersin
PART 1:
Introduction, Basic Definitions
and Concepts
23.09.2017
Source:
Stock and Watson, Introduction to Econometrics,
3rd Edition.
Extra Material::Gujarati, Basic Econometrics, 4.Ed.
RESEARCH AREAS IN ECONOMETRICS
2
What is Econometrics?
Y 1 2X 0 2 1
Y: Consumption (Dependent variable)
X: Income (Independent variable, also called
explanatory variable)
1 and 2 are parameters. They are also called intercept
and slope, respectively.
Econ 102 Review: The Keynesian Theory of Consumption
An Aggregate
Consumption Function
The aggregate
consumption function
shows the level of
aggregate consumption at
each level of aggregate
income.
The upward slope
indicates that higher
levels of income lead to
higher levels of
consumption spending.
7 of 39
The Keynesian Theory of Consumption
C
marginal propensity to consume slope of consumption function
Y
SYC
8 of 39
The Keynesian Theory of Consumption
C
marginal propensity to consume slope of consumption function
Y
SYC
9 of 39
In the mathematical model, the slope
parameter is called the marginal propensity to
consume (MPC).
Y 1 2X 0 2 1
3. Defining the Econometric Model of
Consumption
Mathematical models are deterministic. The changes between
variables are determined without any errors or deviations.
However, in social sciences, we observe stochastic relations
between variables.
Y 1 2X u
TUIK www.tuik.gov.tr
TCMB www.tcmb.gov.tr
DPT www.dpt.gov.tr
DTM www.dtm.gov.tr
OECD www.oecd.org
Eurostat www.eurostat.org
World Bank www.worldbank.org
IMF www.imf.org
5. Estimation of an Econometric Model
8,000
7,000
6,000
T KET IM
5,000
4,000
3,000
2,000
1,000
2,000 4,000 6,000 8,000 10,000 12,000
GELIR
Yt - 299.5913 0.7218 * X t
Tablo I-1
6. Hypothesis Tests
By accepting that the estimated parameters are good
approximations of the theoretical parameters, we have to define a
method of measurement or criteria to evaluate the fit of
parameters.
We face several questions:
In reality and according to the data at hand, can we accept that mpc was below 1 as
Keynes stated?
Is mpc below 1?
Is mpc different than zero? If this is the case, what can we say about the Keynesian
model?
We can test if mpc is below 1.
We can follow statistical inference methods.
Now, its your turn. What are the hypotheses that we should write down? Is it a t test?
7. Prediction
Predict the average consumption for the year 2006.
GDP in the year 2006 is 11319.4 billion dollars. By plugging this value, we
obtain the following result:
Y2006 - 299.5913 0.7218 * (11319.4)
7870.7516
In reality, in the year 2006, the average consumption is 7870.7516 billion
dollars. The predicted value is 174 billion dollars below the actual consumption.
These types of errors will be discussed later on.
8. Using the Model for Political Decisions
By using the Keynesian consumption equation we
estimated, we can evaluate the multiplier in Turkish
economy.
1) cross-sectional data:
ndividuals, families, firms, cities, industries
Countries etc. for a specific time.
8
Time series data:
9
Time series data: min wage, u, gnp
10
GSMH
12,000
10,000
8,000
6,000
4,000
2,000
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005
0
10000
20000
30000
40000
50000
60000
70000
80000
11-02-87
21-04-1988
19-10-1988
2447712
2447596
23-03-1990
26-09-1990
20-03-1991
23-09-1991
16-03-1992
2448936
2449203
2449270
2449415
31-08-1994
17-02-1995
18-08-1995
2450298
2450335
2450571
2450577
27-01-1998
29-07-1998
27-01-1999
27-07-1999
2451606
2451552
IMKB
29-01-2001
26-07-2001
24-01-2002
18-07-2002
13-01-2003
2452951
2453250
2453194
2453431
27-06-2005
19-12-2005
15-06-2006
ISE100 Index, Daily.
2453960
2454197
26-11-2007
22-05-2008
14-11-2008
14-05-2009
2454994
2455261
26-10-2010
25-04-2011
Sample Period: 11.02.1987-25.04.2011
IMKB
Cross sectional data: Cross section data consists of
variables taking different values with respect to
different observations in a given time. This means,
we do not have time dimension for this type of
data.
Example: Car sales in 1990 in USA in different
states.
Weight and height data of economics students in
Beykent University this semester.
Tablo 1-1
Panel data: A data type having both time and cross
section dimensions. If we record the car sales
volume for different states in America not only for
1990, but keep recording for 1990-2011 period, we
have larger amount of observations in our sample.
An Economic Model
Example 1: An Economic Model of Crime
y = f(x1,x2,x3,x4,x5,x6,x7)
6
Econometric Model
u: error term, disturbance term, the total amount of effect of the variables not included in
the model.
0, 1, 2, 3 : parameters that show the degree of the relation, the strength of the
relation, sign of the relation between the explanatory variables and the dependent
variable.
7
POOLED CROSS SECTIONAL DATA
Data set has both cross-sectional and time
series characteristics. Example:
Cross sectional data for 10 years
Here, the sampling is gathered so that each
Cross sectional sample doesnt have the
identical units (different families, different firms
etc).
12
Pooled cross sections
13
PANEL DATA (LONGITUDINAL DATA)
14
PANEL DATA
15
Causality and Ceteris Paribus
n Econometrics
The reason of economic modelling is to put forth the
Relationship between one variable with the other.
This relationship can be considered as causality. But does not
show causality necessarily. It is better to put this as multiple correlation
Ceteris paribus: holding others constant.
Consider the demand function. By holding tastes, income
wealth, prices of other goods constant, we say:
Ceteris paribus, if price increases by one unit, the demand
for chocolates decrease by 0.5 unit.
16
Economics 102 Review:
Government Purchases (G), Net Taxes (T), and Disposable
Income (Yd)
AE C I G
36 of 63
Fiscal Policy at Work: Multiplier Effects
The Government Spending Multiplier
1
government spending multiplier
MPS
37 of 63
Planned Investment and the Interest Rate
Planned Aggregate Expenditure and the Interest Rate
0.04
0.02
gr_cons
gr_gdp
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 79 81 83 85 87 89 91 93
-0.02
-0.04
-0.06
Model 1. What can we say about the
Consumption Regression in Japan?
Figure 2: Now, we expand our model. We have GDP,
Investments, Government Expenditures, Consumption in
Japan. 1981.Q3-2003.Q3 (All variables are quarterly. All
variables are given as yearly growth rates)
.20
.15
.10
.05
.00
-.05
-.10
-.15
-.20
82 84 86 88 90 92 94 96 98 00 02
GR_GOV GR_INV
GR_GDP GR_CONS
Model 2: GDP Model
We have Investments, Government Expenditures,
Consumption in Japan.
By using these variables, how can we predict the GDP?
.06
.1
.04
.0
.02
-.1
.00
-.02 -.2
82 84 86 88 90 92 94 96 98 00 02 82 84 86 88 90 92 94 96 98 00 02
GR_GDP GR_CONS
.08 .08
.06 .06
.04
.04
.02
.02
.00
.00
-.02
-.02 -.04
-.04 -.06
82 84 86 88 90 92 94 96 98 00 02 82 84 86 88 90 92 94 96 98 00 02
Model is estimated with E-views 5.0
package program:
Example Questions
By looking at the model estimated in the previous slide for Japan, answer
the following questions.
a. Write down the economic and the econometric model.
b. Test the hypothesis that the parameter of GR_INV is no different than zero at 5% significance
level. Interpret your result.
c. Calculate 99% confidence interval for the parameter of GR_GOV is no different than zero at 5%
significance level. Interpret your result.
d. In the econometric model, comment: What is the impact of a 1% increase in the investment
growth rate on the GDP growth rate.
e. If the government expenditure growth rate increases by 1% this year, what is the impact on GDP
growth rate?
f. Assume that the Consumption growth rate increases by 1%. What is the impact on the GDP
growth rate?
g. predict the GDP growth rate by assuming that investment growt rate is 5%, C growth rate is %5,
G growth rate is 5%.
h. Assume that G growth rate is recorded as -20% whereas, other variables grew at the rates given
in part (g). Now, solve question E again. Predict the new GDP growth rate.
i. Are there any missing variable or variables in the model you evaluated? Accordingly, what type
of changes are required on both the Economic and the Econometric Models?
j. Investigate: What is the R-square? In the model, R-square is calculated as 0.94. What does this
mean?
The same model is now estimated for Turkey for the
period of 1999Q1-2011Q3. Estimation results are
given below. Solve the same questions for the model
estimated for Turkey.
The figure below shows the two variables, d4cons
and d4y; which are the yearly growth rates of
consumption and income. Note the positive
correlation between the two variables.
.6
.5
.4
.3
.2
.1
.0
-.1
98 99 00 01 02 03 04 05 06 07 08 09 10 11
D4CONS D4Y
The figure below shows the two variables, d4inv and
d4y; which are the yearly growth rates of investment
and income. Note the positive correlation between
the two variables.
.6
.5
.4
.3
.2
.1
.0
-.1
-.2
-.3
98 99 00 01 02 03 04 05 06 07 08 09 10 11
D4INV D4Y