1
. Methods and models employed,
. Formulation of the null and alternative hypotheses,
. Hypothesis tests,
. Inference drawn, and
. Explanation of economic intuition of the relationship(s) estimated.
Data file:
Data set for this project is provided in an Eviews file “project-2010.wf1”. Download this
data. This file contains data on:
gdp = real GDP (Output) of a country, say a Sample Country.
ks = real physical capital stock
emp = labour force employed
br = stock of Domestic knowledge (R&D) (i.e., total stock of domestic technology)
sf = stock of foreign knowledge (i.e., total knowledge stock of Sample country’s trading partners)
All variables are in natural logarithms. Data sample is 1960 to 2008. Data frequency is annual.

Question 1. Tabulate descriptive statistics (mean, median, maximum, minimum and standard deviation) of the dataset and describe them. Plot all five series and comment on their time pofile. (Marks 5).
Question 2. Estimate the following production function by OLS Estimator:
qt=α0 +β1kst+β2empt+β3brt+β4sft+et(1)

Where q = gdp and rest of the variables are also as defined above. Also compute (i) the second and the third order Breusch-Godfrey LM tests of residual serial correlation, (ii) Breush-Pagan test of hetroscedasticity, (iii) White’s test of heteroscedasticity and (iv)
RESET test of functional form. Comment on the overall (OLS) results of above production function as well as the diagnostic tests computed in (i) – (iv). Marks (15).

Question 3. Does the model (1) suffer from residual serial orrelation and heteroscedasticity? If so, first address the problem of auto-correlation by estimating either an AR(1) or an AR(2) specification of the above model (equation (1)). Please note in Eviews, estimating AR(1) and AR(2) specifications are straightforward and they are
equivalent to the Cochrane-Orcutt corrections for the residual autocorrelation. Your final specification (model), whether AR(1) or AR(2), must be based on the requirement that the model passes the second order residual serial correlation. Report the results and
relevant tests of your preferred specification and interpret the results. Test if your model still shows the heteroscedasticity and report the results. You must comment on all the results. (Marks 15).
--
Question 4. Conduct CHOW test of structural break treating 1990 as the date of break. Report and interpret the CHOW test result. (Marks 5).

Question 5. Conduct unit root (DF/ADF) tests on each of the variables of equation (1) above. You must conduct unit root tests sequentially and report all the results in a Table. Using these results, explain clearly whether the variables in equation (1) are I(0), I(1) or
I(2). (Marks 15).

Question 6. Specify and concisely explain the two steps of Engle-Granger co-integration test. Following this method, estimate, test and report if equation (1) forms a cointegrating relationship. Interpret the co-integrating parameters. Plot the error correction
term (ECT) and comment. (Marks 10)
Question 7. Briefly discuss the differences between Engle-Granger and the Dynamic OLS (DOLS) estimators of a co-integrating relationship. Specify and estimate a first order DOLS co-integrating regression of equation (1) and report the results. Test for the second order residual autocorrelation on this DOLS and assess whether you need to opt for a DGLS (Dynamic GLS) estimator. Depending on the residual serial correlations, estimate a suitable (either AR(1) or AR(2)) DGLS co-integrating regression such that your preferred (final) specification passes residual autocorrelation tests. Conduct
significance tests on co-integrating parameters through Wald test and interpret the results. Plot the ECT (error correction term) derived from the DGLS model of your choice and comment. (Marks 15).

Question 8. Compare and contrast the co-integration results obtained from the Engle-Granger approach and your preferred DGLS model. Explain with illustration the concept of an error-correction model. (Marks 10).

Question 9. Specify two regression equations for Granger causality tests between output growth and domestic R&D growth in a multivariate framework using all the variables of equation (1). Conduct short run and long run Granger causality tests and report and interpret the results. (Marks 10).