Scribd
Upload
70 views

Assignment 1

This document provides an assignment problem on estimating a linear regression model relating public debt (y) to GDP (x1) and budget surplus (x2). It shows how to estimate the model using OLS in STATA, interpret the coefficient signs and significance. It also demonstrates how to calculate fitted values for given regressors, and perform hypothesis tests on the coefficients. Specifically, it tests whether at least one coefficient is different than zero, and whether the surplus coefficient equals -3500.

Uploaded by

jroflcopter
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
70 views

Assignment 1

This document provides an assignment problem on estimating a linear regression model relating public debt (y) to GDP (x1) and budget surplus (x2). It shows how to estimate the model using OLS in STATA, interpret the coefficient signs and significance. It also demonstrates how to calculate fitted values for given regressors, and perform hypothesis tests on the coefficients. Specifically, it tests whether at least one coefficient is different than zero, and whether the surplus coefficient equals -3500.

Uploaded by

jroflcopter
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 4

ECON2206: ASSIGNMENT 1

PROBLEM 1 (THEORY)

10 MARKS

Show E ( y ix i)= 0 + 1 x i

a)

(3 marks)

From the error term assumption (zero conditional mean) we know that:

Then

yi
E( y ix i )=E
= E(

)=

E ( ui|x i )=E ( ui ) =0

E( 0 + 1 x i +ui )

0 +E ( 1 x i)+E (ui)
0+ 1 xi +0
0+ 1 xi

E ( y i|x i )= 0 + 1 x i asrequired
b)

Show Prob ( y i=1|x i )= 0 + 1 x i

(4 marks)

Y is a binary variable which only takes on the values of 1 and 0. Therefore this is a case of binomial
probability. Then y=1 (individual uses public transport to go to work) is depicted as success and y=0
(individual does not use public transport to go to work) as failure, we have:

E ( y i|x i) =1 Prob ( y i=1|x i ) +0 Prob ( y i=0|x i )


Prob ( y i=1|x i )
Therefore, using part (a)

Prob ( y i=1|x i )= 0 + 1 xi asrequired


This depicts the linear probability model which is estimated by OLS.

c)

Show the point ( x , y ) is alwaysonthe OLSregressionline y= ^ 0+ ^1 x

(3 marks)

The sample regression model is given by:

SSR y i ^0 ^1 x i

y i= ^ 0 + ^ 1 x i +ui

i=1

^ 0

Rearranging the equation

ui= y i ^0 ^ 1 x i

=0

Using the chain rule:


(1)
n

2 ( y i ^ 0 ^ 1 x i ) =0

The sum of squared residuals (SSR) equation


is given by

i=1

u^ i2

i=1

i=1

y in ^o ^ 1 x i=0

i=1

Substituting equation (1)

Divide the equation by n

y i ^ 0 ^ 1 x i

yi

Produces the equation

i=1

y ^ o ^1 x=0

To find the estimated parameters we need to

SSR ( ^ 0 , ^ 1)

xi
^
^
n o 1 i=1 =0
n
n

i=1

minimise

. The first order

condition to show the point

( x, y )

Rearranging

y= ^ o + ^ 1 x

is always

on the OLS regression line is given by:


Therefore the point

SSR ( ^ 0 , ^ 1 )
=0
^

( x, y )

is always on the

OLS regression line.

PROBLEM 2 (STATA)

a)

Use STATA to estimate the model by OLS. Interpret the signs of coefficients and
the significance. (2 marks)

Estimating model by OLS


Using the data in Data_assignment_1.dta and applying the command line regress debt gdp sur
provides the following OLS regression line:

^ =3443.654 +4.473415 gdp 3564.548 sur


debt
t
t
t
2

where n = 92, R2=0.0969


Figure 1: Regression table

Interpreting Signs & Significance


B0 is the intercept which means that a country with no budget surplus and no gdp would have
a predicted public debt of 3443.654.

The coefficient of gdp (B1) is positive which indicates that for every unit increase in gdp,

^
debt

will increase by 4.473415 assuming all else is constant. This impact is significantly

lower than the impact of a unit increase in surplus on debt levels.

The coefficient of sur (B2) is negative. The model demonstrates that for every unit increase in
sur,

^
debt

will decrease by 3564.548 assuming all else is constant. This indicates that one

unit increase in budget surplus will completely offset public debt if gdp is 0. Despite the large
coefficient of budget surplus, the model still predicts a range of 2460.705 to 3451.303 in
public debt. This allows an inference that gdp has much larger values relative to sur.

b)

Use STATA to calculate the fitted values (predicted values) for the following
regressors values. (4 marks)

Firstly, entering the command predict debthat provides the predicted values for all data entries.
Secondly, using the display command calculates the values of the predicted public debt (given gdp
and sur values) shown in the screenshot.
Figure 2: Predicted Values

This information is summarised in the


table below

GDP

SUR

108.547

0.366381

2623.247

92.913

0.282695

2851.612

90.213

0.398573

2782.936

134.792

0.220599

3260.299

^
Debt

c) Use STATA to test the null hypothesis H0: 1= 2=0 against the alternative H1: At
least 10 or 20. Show the F-statistic and the p-value of the test. How many
degrees of freedom does the test have? (2 marks)
Using the command line test gdp sur to test for H0 produces the following statistics.
Figure 3: Hypothesis Testing 1

F statistic = 4.78

P-value = 0.0107
The test has 2 model (numerator) df and 89
(denominator) residual df.
Therefore we reject the null hypothesis with a 95%
confidence level (=0.05) as the p-value of 0.0107 is lower than alpha.

d) Use STATA to test the null hypothesis H0: 2=-3500 against the alternative H1: 23500. Show the statistics and the p-value of the test. (2 marks)
Entering the command test sur=-3500 produces the following values:
Figure 3: Hypothesis Testing 2

F-statistic = 0.00
T-statistic = -0.0317774
P-value = 0.9747
Df=91 (t test)
Therefore we fail to reject the null hypothesis with a 95%
confidence level (=0.05) as the p-value of 0.9747 is much
higher than alpha.
The t statistic is calculated by the general formula:

t=

Figure 4: T Statistic

^ 0
SE( ^ )

You might also like