LAB4 - Panel Data
1. Data – T4.1 Compensation and Productivity
a. Build a line graphs and scatterplot for variables 𝑋 and 𝑌!
b. Estimate this equation
𝑌𝑡 = 𝛽̂0 + 𝛽̂1 𝑋𝑡 + 𝑢̂𝑡
where
𝑌 = index of real compensation per hour
𝑋 = index of output per hour
Interpret your result!
c. Create the residual series for regression on point a. Create a line graph for the
residual series
d. Identify autocorrelation through graph – create a scatterplot for 𝑢̂𝑡 and 𝑢̂𝑡−1 .
What does the scatterplot suggest regarding the relationship amongst the
regression residuals?
e. Identify autocorrelation through run test
f. Identify autocorrelation through Durbin-Watson test
g. As identified from point a., the series for 𝑋 and 𝑌 exhibits time trend. It might
be the case that the issue of autocorrelation occurs because of model
misspecification (the model is not capable in explaining the time trend). Thus,
try to add time trend to the regression model:
𝑌𝑡 = 𝛽̂0 + 𝛽̂1 𝑋𝑡 + 𝛽̂2 𝑡𝑖𝑚𝑒𝑡 + 𝑢̂𝑡
Estimate this equation. Does the new regression still suffer from
autocorrelation?
h. Try an alternative specification as stated below:
𝑌𝑡 = 𝛽̂0 + 𝛽̂1 𝑋𝑡 + 𝛽̂2 𝑋𝑡2 + 𝑢̂𝑡
Interpret the regression result! Does the new specification suffer from
autocorrelation?
i. Assuming our standard regression model suffers autocorrelation, use
Cochrane-Orcutt Iterative procedure to deal with the issue. Estimate the
coefficient of covariance! Obtain the estimates of the GLS. Interpret your
result.
2. Data – T4.2 Investment
a. Estimate the common effect model:
𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖𝑡 = 𝛽̂0 + 𝛽̂1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑡 + 𝛽̂2 𝑉𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛𝑖𝑡 + 𝜀̂𝑖𝑡
b. Estimate the fixed effect model:
𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖𝑡 = 𝛽̂0𝑖 + 𝛽̂1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑡 + 𝛽̂2 𝑉𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛𝑖𝑡 + 𝜀̂𝑖𝑡
c. Which model is better suited to explain investment value? Write your
arguments.
d. Estimate the random effect model:
𝐼𝑛𝑣𝑒𝑠𝑡𝑚𝑒𝑛𝑡𝑖𝑡 = 𝛽0̅ + 𝛽̂1 𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖𝑡 + 𝛽̂2 𝑉𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛𝑖𝑡 + 𝑣𝑖 + 𝜀̂𝑖𝑡
e. Considering the random effect model, now which model is better suited to
explain investment value? Write your arguments.
3. Data – T4.3 Unemployment
a. Consider the model:
𝑌𝑖𝑡 = 𝛽1 + 𝛽2 𝑋𝑖𝑡 + 𝑢𝑖𝑡
A priori, what is the expected relationship between Y and X? Why?
b. Estimate the model for each country.
c. Estimate the model by pooling all the 60 observations.
d. Estimate the fixed effects model.
e. Estimate the random effects model.
f. Which is a better model, FEM or REM? Justify your answer.