Scribd
Upload
28 views21 pages

SEM205 Econometrics Lecture 3

This document discusses various functional forms of regression models, including double log, log-linear, linear-log, and reciprocal models, explaining their mathematical representations and interpretations. It also covers polynomial regression models, standardized variables, measures of goodness of fit, and regression through the origin. Key concepts such as elasticity, growth rates, and information criteria for model selection are highlighted.

Uploaded by

x52qg7rg2k
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
28 views21 pages

SEM205 Econometrics Lecture 3

This document discusses various functional forms of regression models, including double log, log-linear, linear-log, and reciprocal models, explaining their mathematical representations and interpretations. It also covers polynomial regression models, standardized variables, measures of goodness of fit, and regression through the origin. Key concepts such as elasticity, growth rates, and information criteria for model selection are highlighted.

Uploaded by

x52qg7rg2k
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 21

SEM205 Econometrics

Lecture 3: Functional forms


of regression models
XMUM 2019/4
Associate Professor Sayyed Mahdi Ziaei
E: [email protected]
Functional forms of regression
models
 Double log/constant elasticity models
 Semilog models
 Log-linear/growth models
 Linear-log/Engel expenditure functions
 Reciprocal models
LOG-LINEAR, DOUBLE LOG, OR
CONSTANT ELASTICITY MODELS
 The Cobb-Douglas Production Function:
B2 B3
Qi B1 Li K i
 can be transformed into a linear model by taking natural
logs of both sides:
ln Qi ln B1  B2 ln Li  B3 ln K i
 The slope coefficients can be interpreted as elasticities.
 If (B2 + B3) = 1, we have constant returns to scale.
 If (B2 + B3) > 1, we have increasing returns to scale.
 If (B2 + B3) < 1, we have decreasing returns to scale.
LOG-LIN OR GROWTH MODELS
 The rate of growth of real GDP:
RGDPt RGDP1960 (1  r )t
 can be transformed into a linear model by taking natural
logs of both sides:
ln RGDPt ln RGDP1960  t ln(1  r )
 Letting B1 = ln RGDP1960 and B2 = ln (l+r), this can be rewritten
as:
 ln RGDPt = B1 +B2 t
 B2 is considered a semi-elasticity or an instantaneous growth rate.
 The compound growth rate (r) is equal to (e B2 – 1).
LIN-LOG MODELS
 Lin-log models follow this general form:

Yi B1  B2 ln X i  ui

 Note that B2 is the absolute change in Y responding to a


percentage (or relative) change in X
 If X increases by 100%, predicted Y increases by B2 units
 Used in Engel expenditure functions: “The total expenditure
that is devoted to food tends to increase in arithmetic
progression as total expenditure increases in geometric
proportion.”
RECIPROCAL MODELS
 Lin-log models follow this general form:
1
Yi B1  B2 ( )  ui
Xi
 Note that: 1
B2 ( )
 As X increases indefinitely, the term X i approaches zero and Y approaches
the limiting or asymptotic value B1.
dY 1
 The slope is:  B2 ( 2 )
dX X

 Therefore, if B2 is positive, the slope is negative throughout, and if B2 is negative,


the slope is positive throughout.
POLYNOMIAL REGRESSION MODELS
 The following regression predicting GDP is an example of
a quadratic function, or more generally, a second-degree
polynomial in the variable time:
RGDPt  A1  A2time  A3time 2  ut

 The slope is nonlinear and equal to:


dRGDP
 A2  2 A3time
time
SUMMARY OF FUNCTIONAL FORMS
MODEL FORM SLOPE ELASTICITY
dY dY X
( ) .
dX dX Y
X
Linear Y =B1 + B2 X B2 B2 ( )
Y
Y
Log-linear lnY =B1 + ln X B2 ( ) B2
X

Log-lin lnY =B1 + B2 X B2 (Y ) B2 ( X )

1 1
Lin-log Y B1  B2 ln X B2 ( ) B2 ( )
X Y
1 1 1
Reciprocal Y B1  B2 ( )  B2 ( )  B2 ( )
X X2 XY
STANDARDIZED VARIABLES
 We can avoid the problem of having variables measured in
different units by expressing them in standardized form:
 _
* Yi  Y * Xi  X
Yi  ; Xi 
SY SX
_
 where SY and SX are the sample standard deviationsY
_
and X and are the sample means of Y and X, respectively
 The mean value of a standardized variable is always zero
and its standard deviation value is always 1.
MEASURES OF GOODNESS OF FIT
 R2: Measures the proportion of the variation in the regressand explained by the
regressors. 

 Adjusted R : Denoted as
2 R, 2it takes degrees of freedom into account:
_
n 1
R 2 1  (1  R 2 )
n k
 Akaike’s Information Criterion (AIC): Adds harsher penalty for adding more
variables to the model, defined as:
2k RSS
ln AIC   ln( )
n n

 The model with the lowest AIC is usually chosen.


 Schwarz’s Information Criterion (SIC): Alternative to the AIC criterion, expressed
as:

 The penalty factor here is harsher than that of AIC.


k RSS
ln SIC  ln n  ln( )
n n
REGRESSION THROUGH THE ORIGIN
 Also known as a zero intercept model.
 Example is the well-known capital asset pricing model (CAPM) of
portfolio theory:
( ERi  rf )  i ( ERm  rf )

 where ERi = expected rate of return on security i, ERm = expected rate of return on a
market portfolio, rf = risk-free rate of return, β = the Beta coefficient, a measure of
systematic risk that cannot be eliminated through portfolio diversification.
 Beta coefficient greater than 1: Suggests a volatile security
 Beta coefficient of less than 1: Suggests a defensive security
 Sums of squares and cross-product terms are raw terms here:
n

XY i i 2  e 2
i
b2  i n1 var(b2) = n
2
 X i2 X
i 1
i
2
n 1
i 1

You might also like