METU

Department of Economics
Spring 2022-2023

Course: ECON301 – Introduction to Econometrics I

Instructor: Erdal Özmen

T.A.: Duygu Çelik

Problem Set #3 - Solutions

Simple Linear Regression Models – Inference & Functional Forms

Q1) Explain the meaning of SSR and ESS. You may use graphs for illustration purposes.

The explained sum of squares (ESS), alternatively known as the model sum of squares or sum of
squares due to regression, is a quantity used in describing how well a model, often a regression,
represents the data being modelled. SSR is the sum of squares of residuals, it is a measure of the
discrepancy between the data and an estimation model.

1
Q2) State with a reason whether the following statements are true, false, or uncertain.

i) The t test of significance requires that the sampling distributions of estimators 𝛽̂0 and
𝛽̂1 follow the normal distribution.

True. The 𝑡 test is based on variables with a normal distribution. Since the estimators of 𝛽1 and
𝛽0 are linear combinations of the error 𝑢𝑖 , which is assumed to be normally distributed under
CLRM, these estimators are also distributed.

ii) If the null hypothesis is not rejected, it is true.

False. All we can say is that at hand does not permit us to reject the null hypothesis.

iii) In the two-variable regression function, if the slope coefficient 𝛽1 is zero, the intercept
𝛽0 is estimated by the sample mean 𝑦̅.

True. Remember

𝛽̂0 = 𝑦̅ − 𝛽̂1 𝑥̅

iv) Even though the disturbance term in the CLRM is not normally distributed, the OLS
estimators are still unbiased.

True. So long as 𝐸(𝑢𝑖 ) = 0, the OLS estimators are unbiased. No probabilistic assumptions are
required to establish unbiasedness.

2
Q3) We estimate a simple linear regression with the following results:

Sample: 51 Estimated error variance: 2.04672

Then,

i) What is the sum of squared least squares residuals?

Estimated error variance is given is the question. Thus,

𝜎̂ 2 = 2.04678

And we know that:

2
∑𝑢̂𝑖 2
𝜎̂ =
𝑇−𝑘

And replace the values:

∑𝑢̂𝑖 2
𝜎̂ 2 = = 2.04672 → ∑𝑢̂𝑖 2 = 100.28928
51 − 2
It is given that the estimated variance of 𝛽1 is 0.00098.

ii) What is the standard error of 𝛽1 ?

1
𝑠𝑒(𝛽̂1 ) = √𝑣𝑎𝑟(𝛽̂1 ) = √𝜎̂ 2 ∗
∑(𝑥𝑖 − 𝑥̅ )2

Estimated variance of 𝛽1 is given in the question as 0.00098.

𝑠𝑒(𝛽̂1 ) = √0.00098 = 0.0313

iii) What is the value of

∑(𝑥𝑖 − 𝑥̅ )2

For estimated variance of 𝛽1 :

1
0.00098 = 𝜎̂ 2 ∗
∑(𝑥𝑖 − 𝑥̅ )2

1
0.00098 = 2.04672 ∗
∑(𝑥𝑖 − 𝑥̅ )2

2.04672
∑(𝑥𝑖 − 𝑥̅ )2 = = 2088.4898
0.00098

3
Suppose we define variables as:

𝑦𝑖 : mean income in thousands of Turkish Liras 𝑥𝑖 : percentage of university graduates

iv) If 𝛽1 = 0.18, comment on the results. What happens to mean income when
percentage of graduates increase by 1 percentage point? Does it make sense?

1% point increase in the percent of university graduates will lead to an increase of 180 Turkish
liras (0.18 ∗ 1000 𝑇𝐿) in mean income. It makes sense since increase in educational attainment
leads to higher wages.

Suppose that

𝑥̅ = 69.139 𝑦̅ = 15.187

v) What is the estimate of the intercept parameter?

Recall normal equations:

𝛽̂0 = 𝑦̅ − 𝛽̂1 𝑥̅

𝛽̂0 = 15.187 − 0.18 ∗ 69.139 = 2.74198

vi) Given the results in part (𝑖𝑖), (𝑖𝑖𝑖) and (𝑣), what is

∑(𝑥𝑖 )2

We know the value for ∑(𝑥𝑖 − 𝑥̅ )2 , so we can find ∑(𝑥𝑖 )2 by employing that:

∑(𝑥𝑖 − 𝑥̅ )2 = ∑𝑥𝑖2 − 2∑𝑥𝑖 𝑥̅ + ∑𝑥̅ 2

∑𝑥𝑖2 = ∑(𝑥𝑖 − 𝑥̅ )2 + 2𝑥̅ ∑𝑥𝑖 − 𝑛𝑥̅ 2

∑𝑥𝑖2 = ∑(𝑥𝑖 − 𝑥̅ )2 + 2𝑛𝑥̅ 2 − 𝑛𝑥̅ 2

∑𝑥𝑖2 = ∑(𝑥𝑖 − 𝑥̅ )2 + 𝑛𝑥̅ 2

Replace the values:

∑𝑥𝑖2 = 2088.4898 + 51 ∗ (69.139)2 = 245878.7572

4
Q4) Table contains the YKS scores and the GPA for eight college students. GPA is based on a four-
point scale.

Student GPA YKS

1 2.8 350
2 3.3 360
3 3.5 450
4 3.7 180
5 3.8 190
6 2.9 330
7 2.5 300
8 3.9 520

i) Estimate the relationship between GPA and YKS using OLS; that is, obtain the
intercept and slope estimates in the equation.

̂ 𝑖 = 𝛽̂0 + 𝛽̂1 𝑌𝐾𝑆𝑖

𝐺𝑃𝐴

Let 𝑦𝑖 = 𝐺𝑃𝐴𝑖 and 𝑥𝑖 = 𝑌𝐾𝑆𝑖 .

Find the average values,

2.8 + 3.3 + 3.5 + 3.7 + 3.8 + 2.9 + 2.5 + 3.9

𝑦̅ = = 3.3
8
350 + 360 + 450 + 480 + 490 + 330 + 300 + 520
𝑥̅ = = 410
8

5
Find the following calculation for each individual and corresponding summation:

Student (𝑖) GPA (𝑦𝑖 − 𝑦̅) (𝑦𝑖 − 𝑦̅)2 YKS (𝑥𝑖 − 𝑥̅ ) (𝑥𝑖 − 𝑥̅ )2 (𝑥𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅)
(𝑦𝑖 ) (𝑥𝑖 )
1 2.8 -0.5 0.25 350 -60 3,600 30
2 3.3 0.0 0 360 -50 2,500 0
3 3.5 0.2 0.04 450 40 1,600 8
4 3.7 0.4 0.16 180 70 4,900 28
5 3.8 0.5 0.25 190 80 6,400 40
6 2.9 -0.4 0.16 330 -80 6,400 32
7 2.5 -0.8 0.64 300 -110 12,100 88
8 3.9 0.6 0.36 520 110 12,100 66
SUM 1.86 49,600 292

We obtain 𝛽̂1 as follows:

∑(𝑥𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦̅) 292

𝛽̂1 = 2
= ≅ 0.006
∑(𝑥𝑖 − 𝑥̅ ) 49,600

And 𝛽̂0 as follows:

𝛽̂0 = 𝑦̅ − 𝛽̂1 𝑥̅ = 3.3 − (0.006 ∗ 410) = 0.84

So, we can write:

̂ 𝑖 = 0.84 + 0.006𝑌𝐾𝑆𝑖
𝐺𝑃𝐴

6
 ii) Compute the fitted values and residuals for each observation and verify that the
residuals sum to zero.

The fitted values and residuals are in the following table:

Student (𝑖) YKS ̂ 𝑖 = 0.84 + 0.006𝑌𝐾𝑆𝑖

𝐺𝑃𝐴 GPA ̂𝑖
𝑢̂𝑖 = 𝐺𝑃𝐴𝑖 − 𝐺𝑃𝐴
(𝑥𝑖 ) (𝑦𝑖 ) (𝑦𝑖 − 𝑦̂𝑖 )2
1 350 2.94 2.8 -0.14 0.0196
2 360 3.00 3.3 0.30 0.0900
3 450 3.54 3.5 -0.04 0.0016
4 180 3.72 3.7 -0.02 0.004
5 190 3.78 3.8 0.02 0.004
6 330 2.82 2.9 0.08 0.004
7 300 2.64 2.5 -0.14 0.0196
8 520 3.96 3.9 -0.06 0.0036
SUM 0.00 0.1416

iii) How much of the variation in GPA for these 8 students is explained by YKS? Explain.

The Sum of Squared Residuals (SSR) is:

𝑆𝑆𝑅 = ∑𝑢̂𝑖2 = ∑(𝑦𝑖 − 𝑦̂𝑖 )2 = 0.1416

The Total Sum of Squares (TSS) is:

𝑆𝑆𝑅 = ∑(𝑦𝑖 − 𝑦̅)2 = 1.86

So, the 𝑅 2 from the regression is:

𝑆𝑆𝑅 0.1416
𝑅2 = 1 − =1− = 0.924
𝑇𝑆𝑆 1.86

Thus, 92.4% of the variation in GPA is explained by 𝑌𝐾𝑆.

7
Q5) Consider the money demand 𝑀𝐷 equation y using OLS method. The model is as follows:

𝑀𝑡𝐷 = 𝛽0 + 𝛽1 𝑌𝑡 + 𝑢𝑡

where 𝑀𝐷 :money demand, 𝑌: aggregate income level

The data covers the period from 1970 and 2009. The following information is obtained from the
data:

∑ 𝑀𝑡 = 120 ∑ 𝑌𝑡 = 200 ∑ 𝑀𝑡 𝑌𝑡 = 1200

∑ 𝑀𝑡2 = 860 ∑ 𝑌𝑡2 = 1800

Money demand (𝑀𝐷 ) corresponds to 𝑦 and GDP (𝑌) corresponds to 𝑥 in the normal
equation.
i) Estimate the coefficients of the model, the measure of the goodness of fit and the
standard errors of the coefficients.
We know ̅̅̅̅
𝑀𝑑 and 𝑌̅ given the values and sample size (𝑛 = 40)

̅̅̅̅ 120
𝑀𝑑 = =3
40
200
𝑌̅ = =5
40

Recall the definitions for 𝛽̂0 and 𝛽̂1

𝛽̂0 = 𝑦̅ − 𝛽̂1 𝑥̅
∑(𝑥𝑡 − 𝑥̅ )(𝑦𝑡 − 𝑦̅) ∑𝑥𝑡 𝑦𝑡 − 𝑛𝑥̅ 𝑦̅
𝛽̂1 = =
∑(𝑥𝑡 − 𝑥̅ )2 ∑𝑥𝑡2 − 𝑛𝑥̅ 2

Rewrite 𝛽̂1 with our variables:

∑𝑌𝑡 𝑀𝑡𝐷 − 𝑛𝑌̅̅̅̅̅̅

𝑀𝐷
𝛽̂1 =
∑𝑌𝑡2 − 𝑛𝑌̅ 2

Replace the values:

1200 − (40 ∗ 5 ∗ 3)
𝛽̂1 = = 0.75
1800 − (40 ∗ 52 )

To obtain standard errors, recall:

2
2
∑(𝑦̂𝑡 − 𝑦̅̂)
𝑅 =
∑(𝑦𝑡 − 𝑦̅)2
∑(𝑥𝑡 − 𝑥̅ )2
𝑅 2 = 𝛽̂12 ∗
∑(𝑦𝑡 − 𝑦̅)2

8
Rewrite 𝑅 2 with our variables:

∑(𝑌𝑡 − 𝑌̅)2 ∑𝑌𝑡2 − 𝑛𝑌̅ 2

𝑅 2 = 𝛽̂12 ∗ 2 = 𝛽̂12 ∗ 2
∑(𝑀𝑡𝐷 − ̅̅̅̅̅
𝑀𝐷 ) ̅̅̅̅̅
∑(𝑀𝑡𝐷 )2 − 𝑛𝑀 𝐷

Replace the values:

1800 − 40(5)2
𝑅 2 = (0.75)2 ∗ = 0.90
860 − 40(3)2

Define 𝑅 2 in terms of 𝑆𝑆𝑅 (∑𝑢̂𝑡2 ):

𝑆𝑆𝑅 ∑𝑢̂𝑡2
𝑅2 = 1 − =1−
𝑇𝑆𝑆 ∑(𝑦𝑡 − 𝑦̅)2

Rewrite 𝑅 2 with our variables:

𝑆𝑆𝑅 ∑𝑢̂𝑡2 ∑𝑢̂𝑡2

𝑅2 = 1 − =1− 2 = 1 − 2
𝑇𝑆𝑆 ∑(𝑀𝐷 − ̅̅̅̅̅
𝑀𝐷 ) ̅̅̅̅̅
∑(𝑀𝐷 )2 − 𝑛𝑀 𝐷
𝑡 𝑡

Replace the values:

∑𝑢̂𝑡2
0.90 = 1 −
860 − 40(3)2
Thus,

∑ 𝑢̂𝑡2 = 50

Recall the estimator of 𝜎 2

∑𝑢̂𝑡2
𝜎̂ 2 =
𝑛−𝑘

And replace the values:

50 50
𝜎̂ 2 = = = 1.316
40 − 2 38

Now, recall standard errors of least squares estimates:

For the intercept term:

𝜎̂ 2 ∑𝑥𝑡2 𝜎̂ 2 ∑𝑌𝑡2
𝑠𝑒(𝛽̂0 ) = √𝑣𝑎𝑟(𝛽̂0 ) = √ ∗ = √ ∗
𝑛 ∑(𝑥𝑡 − 𝑥̅ )2 𝑛 ∑𝑌𝑡2 − 𝑛𝑌̅ 2

1.316 1800
𝑠𝑒(𝛽̂0 ) = √ ∗ = √0.074 = 0.272
40 800

9
 For the slope term:

1 1
𝑠𝑒(𝛽̂1 ) = √𝑣𝑎𝑟(𝛽̂1 ) = √𝜎̂ 2 ∗ = √𝜎̂ 2 ∗
∑(𝑥𝑡 − 𝑥̅ )2 ∑𝑌𝑡 − 𝑛𝑌̅ 2
2

1
𝑠𝑒(𝛽̂1 ) = √1.316 ∗ = √0.0016 = 0.04
800

ii) Construct 95 percent confidence interval for both slope and constant terms.
Firstly, recall 𝛽̂1 and find 𝛽̂0 :

𝛽̂1 = 0.75

𝛽̂0 = 𝑦̅ − 𝛽̂1 𝑥̅ = 3 − (0.75)5 = −0.75

Recall the confidence intervals:

But before, find the t-value for 𝑑𝑓 = 38, 𝛼 = 0.025

0.025
𝑡38 = 2.021

For intercept parameter:

𝛼/2 𝛼/2
𝛽̂0 − 𝑡𝑑𝑓 𝑠𝑒(𝛽̂0 ) ≤ 𝛽̂0 ≤ 𝛽̂0 + 𝑡𝑑𝑓 𝑠𝑒(𝛽̂0 )

−0.75 − (2.021)(0.272) ≤ 𝛽̂0 ≤ −0.75 + (2.021)(0.272)

−1.3 ≤ 𝛽̂0 ≤ 0.2

For slope parameter:

𝛼/2 𝛼/2
𝛽̂1 − 𝑡𝑑𝑓 𝑠𝑒(𝛽̂1 ) ≤ 𝛽̂1 ≤ 𝛽̂1 + 𝑡𝑑𝑓 𝑠𝑒(𝛽̂1 )

0.75 − (2.021)(0.04) ≤ 𝛽̂1 ≤ 0.72 + (2.021)(0.04)

0.669 ≤ 𝛽̂1 ≤ 0.831

iii) Test whether the slope is equal to 0.5 or not.

Write the null and alternative hypothesis:

𝐻0 : 𝛽1 = 0.5

𝐻1 : 𝛽1 ≠ 0.5

That is, the true slope is 0.5 under the null hypothesis but less or greater than 0.5.
under the alternative hypothesis.

10
 Recall 𝑡 − distribution and calculate:

𝛽̂1 − 𝛽1 0.75 − 0.5

𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = = = 6.25
𝑠𝑒(𝛽̂1 ) 0.04
0.025
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 𝑡38 = 2.021

Since

𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 2.021 < 6.25 = 𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑

Reject the null hypothesis.

iv) Is the regression relationship between money demand and aggregate income
significant both economically and statistically? Explain your answer.
In order to check statistical significance, first construct the null and alternative
hypothesis:

𝐻0 : 𝛽1 = 0

𝐻1 : 𝛽1 ≠ 0

Recall 𝑡 − distribution and calculate:

𝛽̂1 − 𝛽1 0.75 − 0
𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = = = 18.75
𝑠𝑒(𝛽̂1 ) 0.04
0.025
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 𝑡38 = 2.021

Since

𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 2.021 < 18.75 = 𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑

Reject the null hypothesis. Thus, it is statistically significant.

As a hot prospect for the future economist comment on whether observing an
increase in the demand for money when there is an increase in the aggregate income
level makes sense.

11
Q6) We estimate a simple linear regression with the following results:

Sample: 24 Estimated slope parameter: 0.310 Estimated standard error: 0.082

Given that the regression:

𝑌𝑡 = 𝛽0 + 𝛽1 𝑋𝑡 + 𝑢𝑡

i) Test the following hypothesis with significance level: 1%

𝐻0 : 𝛽1 = 0

𝐻1 : 𝛽1 ≠ 0

Calculate 𝑡 −value:

𝛽̂1 − 𝛽1 0.310 − 0
𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = = = 3.780
𝑠𝑒(𝛽̂1 ) 0.082
0.005
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 𝑡22 = 2.819

Decision criteria:

𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 < 𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑

Reject the null hypothesis.

ii) Test the following hypothesis with significance level: 1%

𝐻0 : 𝛽1 = 0

𝐻1 : 𝛽1 > 0

Calculate 𝑡 −value:

𝛽̂1 − 𝛽1 0.310 − 0
𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = = = 3.780
𝑠𝑒(𝛽̂1 ) 0.082
0.01
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 𝑡22 = 2.508

Decision criteria:

𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 < 𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑

Reject the null hypothesis.

12
iii) Test the following hypothesis with significance level: 5%. Show the rejection region in
a graph.

𝐻0 : 𝛽1 = 0

𝐻1 : 𝛽1 < 0

Calculate 𝑡 −value:

𝛽̂1 − 𝛽1 0.310 − 0
𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = = = 3.780
𝑠𝑒(𝛽̂1 ) 0.082
0.05
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 𝑡22 = −1.717

Decision criteria:

𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 < 𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑

Do not reject the null hypothesis.

13
iv) Test the following hypothesis with significance level: 5%. Show the rejection region in
a graph.

𝐻0 : 𝛽1 = 0.5

𝐻1 : 𝛽1 ≠ 0.5

Calculate 𝑡 −value:

𝛽̂1 − 𝛽1 0.310 − 0.5

𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = = = −2.32
𝑠𝑒(𝛽̂1 ) 0.082
0.025
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 𝑡22 = 2.074

Decision criteria:

𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 < |𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 |

Reject the null hypothesis.

v) Construct a 99% interval estimate of the slope parameter.

The confidence interval is:

𝛼/2 𝛼/2
𝛽̂1 − 𝑡𝑑𝑓 𝑠𝑒(𝛽̂1 ) ≤ 𝛽1 ≤ 𝛽̂1 + 𝑡𝑑𝑓 𝑠𝑒(𝛽̂1 )

(0.310) − (2.819) ∗ (0.082) ≤ 𝛽1 ≤ (0.310) + (2.819) ∗ (0.082)

0.079 ≤ 𝛽1 ≤ 0.541

14
Q7) Consider the following regression output:

𝑦𝑖 = 0.2033 + 0.6560 𝑥𝑖
(0.0976) (0.1961)

and

𝑅 2 = 0.397 𝑆𝑆𝑅 = 0.0544 𝐸𝑆𝑆 = 0.0358

where 𝑦𝑖 is labour force participation rate (LFPR) of women in 1972 and 𝑥𝑖 is LFPR of women in
1968. The regression results were obtained from a sample of 19 cities in the United States. Test
the hypothesis:

𝐻0 : 𝛽1 = 1

𝐻1 : 𝛽1 > 1

Which test do you use? What are the underlying assumptions of the test(s) you use?
Calculate 𝑡 −value:

𝛽̂1 − 𝛽1 0.650 − 1
𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = = = −1.78
𝑠𝑒(𝛽̂1 ) 0.1961
0.05
𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = 𝑡17 = 1.7396

Decision criteria:

𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 < |𝑡𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 |

We reject the null hypothesis.

15
Q8) An economics student constructed the following models for her assignment:

𝑀𝑡𝑑 = 𝛽0 + 𝛽1 𝑌𝑡 (1)

𝑀𝑡𝑑 = 𝛽0 + 𝛽1 log (𝑌𝑡 ) (2)

log (𝑀𝑡𝑑 ) = 𝛽0 + 𝛽1 log (𝑌𝑡 ) (3)

𝑙𝑜𝑔(𝑀𝑡𝑑 ) = 𝛽0 + 𝛽1 𝑌𝑡 (4)

where 𝑀𝑑 and 𝑌 are money demand and aggregate income level (in dollar terms). How do we
interpret 𝛽1 coefficients in (1) − (4)?

Model 1: When the aggregate income level increases by one dollar, the money demand increases
by one dollar.

Model 2: When the aggregate income level increases by one percent, the money demand
increases by one dollar.

Model 3: When the aggregate income level increases by one percent, the money demand
increases by one percent.

Model 4: When the aggregate income level increases by one dollar, the money demand increases
by one percent.

16