Test again the model specification using ramsey reset test

Advanced Econometrics (NEKN31)
Fall 2014

Computer Exercise One
Cross-sectional Data

You are allowed to copy figures from Eviews into your document, but you are not allowed to directly copy tables and regression outputs from Eviews. Eviews gives you a lot of information, which you never use. Therefore, create your own tables in your word processor that summarizes the information you find important and that you have used in your analysis.

You should upload your lab report as a PDF file to Live@Lund before the deadline, otherwise it will not be graded and 0 points will be given automatically. Finally, do not forget to include your name and e-mail address on the front cover. Furthermore, do not forget to include the names of all group members!

where is total costs for firm and , and are price of labour, capital and fuels faced by firm respectively. The degree of returns to scale is , whereas constant returns to scale implies when .

Taking the logarithm of Eq. (2) yields

2

3. Data

You can find the data in the file data_lab1.csv. The cross-section data set is taken from Nerlove (1963) and it contains the observations on 145 firms in the year 1955. The description of the variables are provided in the table below

a.Estimate the total cost function in Eq. (4) with OLS.

b.Test the model specification using (i) Ramsey RESET test, (ii) redundant variables test and (iii) omitted variables test. When you test for omitted variables, test for non-linear effects by adding ( ), ( ), ( ) and ( ). What do you conclude?

e.If you find statistical evidence for heteroskedasity in disturbances, use the following methods to correct the model. For the first three methods, test if the heteroskedacity has been removed from the model. Which method do you prefer? Motivate your choice.

1)Assume that ( ) ( ) and estimate the model using weighted least squares (WLS) estimator.

g.Test for normality using Jarque-Bera test. What do you conclude?

h.Once you have obtained a model that does not seriously violate Gauss-Markov assumptions, test
1)If the parameter estimates agree with economic theory or if for

( ) with group-specific dummy variable. Finally, test if the estimated parameters for ( ) are the same across the groups. What is your conclusion?

5. Help

5.2 White standard errors

. (7) Given that the model in Eq. (6) is correct, the model in Eq. (7) has homoskedastic residuals. In our case, assume that
( ) ( ) (8) which implies that ( ). If we use the weights ( ) to transform our model, we obtain homoskedastic errors. In Eviews we estimate the weighted least squares

To estimate a regression model using FGLS in part e.2. we assume that
( ) { ( )} (9) which implies that { ( )}. However, now depends on unknown parameter which could be consistently estimated using OLS residuals . Note that ( ) { ( )} (10) Taking the logarithm of Eq. (10) yields
( ) (11) Or equivalently
( ) (12) where and ( ). Hence, the FGLS estimator for our case could be easily obtained in two step procedure.

Step 1. Save the estimated residuals and estimate Eq. (12) using OLS.

and compute ̂ { ̂ ( ) ̂ ( )}.

5.4 Creating dummy variables for part i.

• for and 0 otherwise

• for and 0 otherwise

for !g=1 to 5

series dum_{!g}=@recode(@obsid>29*(!g-1) and @obsid<=29*!g,1,0)