Scribd
Upload
25 views7 pages

Assignment 2

The document contains the analysis of an econometrics assignment. It includes: 1) Testing a consumption model for autocorrelation using the Durbin-Watson and Breusch-Godfrey tests, finding evidence of autocorrelation. Possible causes are discussed. 2) Estimating a growth model including capital stock and a dummy variable for apartheid. The effect of independence is calculated by substituting values into the equation. 3) Discussing a productivity model including lags, defining the impact and total effects of changes in an explanatory variable on the dependent variable.

Uploaded by

Gerdus Claassen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
25 views7 pages

Assignment 2

The document contains the analysis of an econometrics assignment. It includes: 1) Testing a consumption model for autocorrelation using the Durbin-Watson and Breusch-Godfrey tests, finding evidence of autocorrelation. Possible causes are discussed. 2) Estimating a growth model including capital stock and a dummy variable for apartheid. The effect of independence is calculated by substituting values into the equation. 3) Discussing a productivity model including lags, defining the impact and total effects of changes in an explanatory variable on the dependent variable.

Uploaded by

Gerdus Claassen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

ECOH 617

Assignment 2
13 April 2023
G.Claassen (34663495)
Question 1

The estimated equation is LCONS = c + B1LINC

1.1 a)

b.) The Durbin-Watson Test

H0: No Autocorrelation

H1: Positive Autocorrelation

The Durbin-Watson Statistic is 0.360976

From the Durbin-Watson Table we can obtain dL and dU for k=1 and 60 observations and a=0.05 as
1.549 and 1.616 respectively

Because the Durbin-Watson Statistic is between 0 and dL, we can reject H0: No Autocorrelation and
conclude that there is positive autocorrelation at 95% confidence.
Because the Durbin-Watson test only assumes that the model has first-order autocorrelation, we use
the Breusch-Godfrey test to test for higher orders of autocorrelation.

The Breusch-Godfrey LM Test

H0: No Autocorrelation

H1: Autocorrelation

As the order of autocorrelation is arbitrary, the Breusch-Godfrey LM-test is conducted for 2 lags and
4 lags.

With a 95% level of confidence, we reject H0: No Autocorrelation in both situations since the Prob
Chi in both cases is 0.05, suggesting that the model contains autocorrelation of a higher order.

We may conclude that the model does exhibit autocorrelation and hence violates a key tenet of the
CLRM based on the data in examples a and b.

1.2)

a.) The Cobweb phenomena

The data may exhibit autocorrelation due to lag responses. When considering agricultural items, we
can see that the choice to plant a crop in a certain period of time (t) is impacted by the price of the
commodity in period (t-1) as well as the amount supplied in that period. Moreover, the rain from
period t-1 rather than the present rain often effects the crop of period t. (indicating lagged reactions)
If we have such lag responses in our data, autocorrelation will highlight this as seen in our error
terms.

b.) Data Manipulation

In empirical analysis, raw data is frequently modified. In time-series regression, for example,
quarterly data is frequently produced from monthly data by adding three months' observations and
dividing the sum by three. Averaging data smoothness data by neutralizing monthly data oscillations.
This smoothness might result in a consistent pattern in error terms, generating autocorrelation.

c.) Model Specification

Autocorrelation can develop as a result of model specification mistakes such as missed explanatory
variables or the incorrect functional form.
Omitted Variables: If you neglected a specific explanatory variable that should have been included,
there will be a consistent fraction of the variance in your dependent variable that will not be
explained, as evidenced by patterns in your error terms.

If you estimate a linear relationship when it should be a quadratic relationship, your residuals will
show a clear pattern because you are now explaining a small part of your regression and a larger part
is not being explained by the explanatory variables, which will be captured in the error terms, which
will graphically show an increase in error terms over time and thus autocorrelation.

Question 2

The Estimated equation is lrgdpt = B0 + B1lcst + B2Dem + ut

2.1) For each one percent increase in the Capital Stock, The South African Real GDP will increase with
0.957210%, it therefore has a positive effect on Real GDP.

2.2) Dummy variables can be used to correct for structural breakdowns, with the dummy variable
indicating the years before to the structural break with 0 and the years after the structural break with
1. A structural break in econometrics can occur when there is a war, a dramatic shift in government
policy, or some similarly abrupt event. Apartheid was a significant change in government policy, thus
we include it as a structural break by using a dummy variable.

2.3) For South Africa’s independence we use “1” for the dummy variable, if we measure the GDP of a
post-Apartheid time period in South Africa, we will substitute the equation as follows:

LRGDP = -0.429815 + 0.957210LCS – 0.216758(1)

LRGDP = -0.646573 + 0.957210LCS

The Real GDP will be 0.0216758 percent lower than it would’ve been in a pre-democracy year. And if
the Capital stock of the country is zero, the Real GDP will be -0.646573
Question 3
3.2 a) If we rewrite the equations in terms of a’s and B’s we will get:

Yt = a + B0Xt + B1Xt-1 + B2Xt-2 + B3Xt-3 + B4Xt-4

The coefficient B0 is the weight attached to the current X (Xt) given the by ΔYt/ΔXt - It shows the
average change in Yt when Xt changes by one utit. B0 is the impact multiplier.

The impact effect will thus be that for one unit increase in training per year, productivity will increase
with 0.236 units cetris paribus.

b) The total effect is given by the sum of the effects on all periods or the Long Run effect: Σ𝛽𝑖𝑝=4𝑖=0=
𝛽0+ 𝛽1+ 𝛽2+ 𝛽3+ 𝛽4 Σ𝛽𝑖𝑝=4𝑖=0= 0.236+0.366+0.581+0.324+0.145=1.652

You might also like