Problem Set 3

Simple Linear Regression Model

N. Juaneda, A. Sansó, V. Troster and [Link]

1. We know that Syx = 150, Sx2 = 100, Sy2 = 300, ȳ = 5, x̄ = 15, n = 30, and we wish to
estimate the simple linear regression model (SLRM) yi = β0 + β1 xi + ei .

a) Calculate the point and the interval estimates of β0 and β1 .

b) What is the coefficient of determination?

c) What is the variance of the residuals?

d) Are the estimated coefficients statistically significant? Use α = 5%.

e) Obtain the point and interval (at 95% confidence level) prediction for y0 when x0 = x̄.

2. Engel’s law is an observation in economics stating that as income rises, the proportion of
income spent on food falls, even if absolute expenditure on food rises. Consider the following
random sample of 10 households:

where wi is the proportion of income spent on food and ln Ii is the logarithm of income.

a) Write down the equation of the econometric model of interest.

b) Generate a scatterplot for both variables.

1
c) The parameters of the simple linear regression model were estimated by Ordinary Least
Squares (OLS). Discuss the estimation results below. Does the coefficient of ln Ii have
the expected sign?

b i = 5.145 − 0.5063 lnI

w
(0.648) (0.0706)
2
n = 10 R = 0.865 ln I = 9.17 Var(ln I) = 0.05122 S = 0.048
(standard errors in parentheses)

d) Evaluate the goodness of fit of the model.

e) Test the significance of the explanatory variable at the α = 5% significance level.

f) Calculate the 95% confidence interval for the marginal effect of the logarithm of income
(ln Ii ) on the proportion of income spent on food (wi ).

g) If we wish to test Engel’s law, how should we specify the null and alternative hypothesis?
Test it at the 5% significance level. Can we conclude that the Engel’s law can be verified
(that indeed, as income rises, the proportion of income spent on food falls)?

h) Calculate the point and interval (at the 95% confidence level) estimates of:

1. The proportion of income spent on food (y0 ) for a household with income of 10000 e.
2. The expected proportion of income spent on food (E(y0 )) for a household with income
of 10000 e.

i) Compare the calculated intervals for both cases above.

3. Consider the following linear regression model yi = β0 +β1 xi +ei , that has been estimated
using 30 observations. Given that x0 = x̄ = 10, the following interval prediction has been
obtained:

Pr(3 < y0 < 7) = 0.95.

Also, we know that Syx = 15 and Sx2 = 20.

a) What are the estimated values of β0 and β1 , obtained by the OLS?

b) Test individually the significance of β0 and β1 at the α = 5% significance level.

c) Calculate the coefficient of determination.

d) Test whether β1 = 1. Use α = 5%.

4. The capital asset pricing model (CAPM) describes the relationship between the expected
return for asset j at time t, rjt , and the market portfolio return, rmt . The relationship is
given by

E(rjt ) = r0 + βj E(rmt − r0 ),

2
where r0 is the return of the risk-free asset. The slope of the regression, βj , is called the
systematic risk coefficient. In case the asset is risk-free, the βj = 0, and if the asset
has the same return as the market, then βj = 1. The returns are calculated as: rjt ≈
ln(Pjt ) − ln(Pjt−1 ). The following is the estimation output that has been obtained using
the closing prices of the shares of Telefónica and IBEX35 index from December 7, 2000 till
December 7, 2001.

a) What is the value of the systematic risk coefficient beta for the Telefónica shares?

b) Test if β1 = 1. Use α = 5%.

c) What is the estimated return of the risk-free asset?

5. (Exam January 2021) Using data from a sample of 32 observations, we want to estimate
a simple linear regression model to explain the time each individual spends sleeping. For this
purpose, we have the following
P32 information on twoPvariables: SLEEPi (in weekly minutes)
32 2
and AGEi (in years): i=1 SLEEPi = 97700, i=1 AGEi = 1186, SSLEEP = 118005,
2
SAGE = 55.43, and COV ARIAN CE(SLEEP, AGE) = −1627.8.

a) Specify the regression model that you think is appropriate and estimate it. Once esti-
mated, interpret the values of the slope and the constant, taking into account the units
in which each variable is expressed.

b) Calculate the value of the coefficient of determination and interpret it.

6. (Exam Frebruary 2023) We want to estimate a regression model where EXP HEALT Hi
is the expenditure on health insurance (in euros) made by the resident population, and
IN COM Ei is the total income of the population (in euros):

EXP HEALT Hi = β̂0 + β̂1 IN COM Ei + ei .

Given the observations of a sample of 51 regions, the regression model above was esti-
mated, obtaining the following results:

3
a) Estimate the variance of the residuals and the standard deviations of β̂0 and β̂1 . Justify
your answers.

b) Make an interval estimate of β1 at the 99% confidence level.

Solutions
1. a) b0 = −17.5, b1 = 1.5. CI(β0 ; 95%) = [−23.54; −11.46], CI(β1 ; 95%) = [1.17; 1.83]; b)
R2 = 0.75; c) S 2 = 80.357; d) Yes, we individally reject each H0 : βj = 0, for j = 0, 1, at
the 5% significance level; e)ŷ0 = 5, CI(y0 ; 95%) = [−13.657; 23.657].

2. a) wi = β0 + β1 ln Ii + ei ; c) Yes, it has the expected sign according to the Engel’s Law;

e) We reject H0 : β1 = 0 at the 5% significance level; f) CI(β1 ; 95%) = [−0.609; −0.343];
g) Engel’s Law: H0 : β1 ≥ 0 vs. HA : β1 < 0. We reject H0 at the 5% significance level;
therefore, the Engel’s Law is verified; h) ŵ0 = 0.4815, CI(w0 ; 95%) = [0.361; 0.602]; (ii)
E(wˆ 0 ) = 0.4815, CI(E(w0 ); 95%) = [0.433; 0.529].

3. a) b0 = −2.5, b1 = 0.75; b) We reject H0 : β1 = 0 at 5% significance level; c) R2 = 0.929;

d) We reject H0 : β1 = 1 at 5% significance level.

4. a) b1 = 1.41059; b) We reject H0 : β1 = 1 at the 5% significance level; c) −0.0000420963.

5. a) SLEEPi = β0 +β1 AGEi +ei . Estimated model: SLEEPi = 4141.53−29.37AGEi +ei .

Interpretation: each additional year implies 29.37 fewer weekly minutes of sleep, whereas
individuals at birth sleep on average 4141.53 weekly minutes (E(SLEEPi |AGEi = 0) =
b0 ); b) R2 = 0.405: 40% of the total variation in weekly minutes of sleep is explained by
age.

6. a) S 2 = 3.035, Sβ̂0 = 0.32 y Sβ̂1 = 0.002; b) IC(β1 ; 99%) = [0.137; 0.147].