QUESTION 1

a) Explain three reasons why we need econometrics?

b) What is meant by a linear regression model?

c) Given the following non linear production model;

i) Transform the above model to a linear model.

ii) What would be the expected sign of the variables' coefficient?

iii) How would you interpret a and p ?

iv) Justify why we should include a stochastic error term in the above model.

d) You are given the scatter diagram as in Figure 1 along with the estimated regression line.

i) What general conclusion can you draw from this diagram?

ii) Based on the scatter diagram, construct the simple regression equation.

QUESTION 2

a) Based on 45 observations, the regression results on cigarette consumption are reported as follows:

where C = cigarette consumption, packs per year

P = real price per pack

Y = real disposable income

i) Do the signs of the coefficients match your priori expectations?

ii) Interpret the estimated partial slope coefficients?

iii) Determine the individual significance at 5% for variable P and Y.

iv) Test the overall significance of the estimated regression at 5%.

b) List three (3) types of specification error and give three (3) reasons why they occur.

QUESTION 3

a) i) What is serial correlation?

ii) Give two (2) reasons why serial correlation exists.

b) Consider the following estimation results estimated by OLS technique with annual data from 1984-2006.

where It is the level of aggregate investment, GDPt is gross domestic product and rt is the interest rate. The values in parentheses are f-statistics.

In order to detect the first order serial correlation, you are required to perform the Durbin-Watson test.

i) State the null and alternative hypotheses for no serial correlation.

ii) Determine the number of observations and number of explanatory variables.

iii) Test the above model for serial correlation at 5% level of significance.

iv) What is your conclusion?

c) Suggest two (2) remedies for serial correlation.

d) State three (3) consequences of serial correlation.

e) "Ordinary least squares is the best procedure for estimating a linear regression model but there are conditions referred to as the classical assumptions. Violation of these assumptions will lead to problems in regression." State three (3) assumptions and the problem related to each assumption.

QUESTION 4

Given below are the output generated from a cross sectional data collected from 4000 fulltime workers with a minimum qualification of a bachelor's degree from an overseas university. The dependent variable is average monthly earnings (AME - in thousands RM) and the independent variables are level of education (Bachelor Degree = 1 and 0 otherwise), gender (Female=1 and 0 if Male), age (Age - in years) and education background (graduated from United States (US), European countries (EU), Australia (AU) and other countries (OC)). Three models were estimated and the results were shown below:

a) Using the regression results in column (1).

i) Interpret the partial slope coefficient for Bachelor Degree.

ii) Do men earn more than women on average? How much more or how much less?

iii) Derive four (4) regression equations for the following:

a) Female with bachelor's degree and female with higher qualification.

b) Male with bachelor's degree and male with higher qualification.

b) Using the regression results in column (2).

i) Is age an important determinant of earnings? Explain.

ii) Ahmad is 30 years old and Siti is 35 years old and both have a bachelor degree. Predict Ahmad's and Siti's monthly earning.

c) Using the regression results in column (3).

i) Does education background appear to be an important determinant of average hourly earnings? Explain.

ii) Why is the independent OC (graduated from other countries) omitted from the regression? What would happen if it was included?

iii) Interpret the slope coefficients for AU and Age.

iv) Muadz is a 28-year-old graduate with a master degree from United States and Madihah is a 30-year-old graduate with a bachelor degree from Australia. Calculate the expected differences in earnings between Muadz and Madihah.

d) The R-squared value is quite low and adding more independent variables to the regression equation may help to increase the value. Suggest three (3) independent variables that you think appropriate.

QUESTION 5

a) Consider the following model:

i) Looking at the regression results above, what symptoms signal multicoliinearity? Comment. 

ii) A regression was run using STUDY as the dependent variable and LIBRARY as the independent variable and the R-squared = 0.856809. Test for the presence of multicoliinearity.

iii) Suggest two ways to detect the existence of multicoliinearity.

iv) Suggest two remedial measures to solve multicoliinearity.

b) Consider the cross-section tax revenue model

To test for heteroscedasticity, the White test was conducted and the results is shown below:

i) Is there a problem of heteroscedasticity? Perform the White test.

Use a= 0.05.

i') Describe two (2) other ways of testing heteroscedasticity