Scribd
Upload
20 views

Exercise 11 Answers

This document contains the answers to an econometrics exercise. It compares pooled OLS, random effects, and fixed effects models. For the first dependent variable (lnY), the Breusch-Pagan test indicates random effects, the Hausman test is inconclusive, so the random effects model is chosen. For the second dependent variable, the Breusch-Pagan and Hausman tests both indicate fixed effects is best. A 1% increase in income increases gasoline consumption by 0.6623%, while a 1% increase in price decreases it by 0.3217% and cars per capita decreases it by 0.6405%.

Uploaded by

emy anisha
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
20 views

Exercise 11 Answers

This document contains the answers to an econometrics exercise. It compares pooled OLS, random effects, and fixed effects models. For the first dependent variable (lnY), the Breusch-Pagan test indicates random effects, the Hausman test is inconclusive, so the random effects model is chosen. For the second dependent variable, the Breusch-Pagan and Hausman tests both indicate fixed effects is best. A 1% increase in income increases gasoline consumption by 0.6623%, while a 1% increase in price decreases it by 0.3217% and cars per capita decreases it by 0.6405%.

Uploaded by

emy anisha
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

EIA2014 ECONOMETRIC II / EIE3003 ECONOMETRIC ANALYSIS

Exercise 11 answers

1. -

Dependent variable: ln Y ¿
Pooled OLS Random effect Fixed effect
Constant 2.2863 -0.4135 3.3829
ln X 1¿ ¿ 0.9126*** 0.3556* -0.0455
ln X 2¿ ¿ -1.6899** -0.0481 -0.3762
ln X 3¿ ¿ 0.0779 0.1668** 0.2640**
Breusch-Pagan 2
χ =¿ 50.1112 - -
LM test p-value = 0.0000
2
Hausman test - χ =¿ 1.3497 -
p-value = 0.7174
Wald test^ - - F = 96.9207
p-value = 0.0000
*, ** and *** indicate the respective 10%, 5% and 1% significance level.
^ testing the null hypothesis of common intercept.

Breusch-Pagan LM test: H 0 : σ 2λ =0 vs H 1 : σ 2λ ≠ 0

p-value = 0.0000 < 0.05. The test result suggests that there is random effect.

Hausman test: H 0 :Cov ( λ i , X ¿ )=0 (no correlation between λ i and X ¿ ) - RE

H 1 :Cov ( λi , X ¿ ) ≠ 0 (correlation between λ i and X ¿ ) – FE

p-value = 0.7174 > 0.05. The test result suggests that there is no correlation
between λ i and ln X ¿ which favours random effect model.

Wald-test of common intercept: H 0 :α 2=α 3 =…=α 6 =0 - Pooled model

H 1 : at least one α j ≠ 0 – FE

p-value = 0.0000 < 0.05. The result suggests that individual effects are
different from one another.

Based on the results of three tests, we decide to use random effect model.

1
2. a) Individual heterogeneity can be controlled.

b) –
Dependent variable: ln Y ¿
Pooled OLS Random effect Fixed effect

Constant 2.391326*** 1.996698*** 2.402670***


ln X 1¿ ¿ 0.889962*** 0.554986*** 0.662250***
ln X 2¿ ¿ -0.891798*** -0.420389*** -0.321702***
ln X 3¿ ¿ -0.763373*** -0.606840*** -0.640483***

Breusch-Pagan 2
χ =¿ 1465.552 - -
LM test p-value = 0.0000
Hausman test - 2
χ =¿ 26.495054 -
p-value = 0.0000
Wald test^ - - F = 83.96080
p-value = 0.0000
*, ** and *** indicate the respective 10%, 5% and 1% significance level.
^ testing the null hypothesis of common intercept.

(i) Breusch-Pagan LM test: H 0 :σ 2λ =0 vs H 1 : σ 2λ ≠ 0

p-value = 0.0000 < 0.05. The test result suggests that there is random effect.

Hausman test: H 0 : Cov ( λ i , X ¿ )=0 (no correlation between λ i and X ¿ ) - RE


H 1 : Cov ( λi , X ¿ ) ≠ 0 (correlation between λ i and X ¿ ) – FE

p-value = 0.0000 < 0.05. The test result suggests that there is correlation
between λ i and ln X ¿ which favours fixed effect model.

Wald-test of common intercept: H 0 :α 2=α 3 =…=α 18=0 - Pooled model

H 1 : at least one α j ≠ 0 – FE

p-value = 0.0000 < 0.05. The result suggests that individual effects are
different from one another.

Therefore, we should use fixed effect model.

(ii) 1% increase in real income per capita will increase the gasoline
consumption by 0.6623%.

2
1% increase in real gasoline price will decrease the gasoline consumption by
0.3217%.

1% increase in numbers of cars per capita will decrease the gasoline


consumption by 0.6405%.

You might also like