University of Bath

DEPARTMENT OF ECONOMICS

ES50112

Applied Econometrics

January 2020

2 Hours

Answer any 3 questions.

A University Calculator is required for this exam. Statistics tables are supplied with this
exam paper.

PLEASE FILL IN THE DETAILS ON THE FRONT OF YOUR ANSWER

BOOK/COVER AND SIGN IN THE SECTION ON THE RIGHT OF YOUR ANSWER
BOOK/COVER, PEEL AWAY ADHESIVE STRIP AND SEAL.

TAKE CARE TO ENTER THE CORRECT CANDIDATE NUMBER AS DETAILED ON

YOUR DESK LABEL.

DO NOT TURN OVER YOUR QUESTION PAPER UNTIL INSTRUCTED TO BY THE

CHIEF INVIGILATOR
 Answer any THREE questions (marks for each question in brackets)

1a) Outline the main steps involved in building an econometric model, starting with the first
theoretical model step. If when estimated the model is inadequate, how would you overcome
this? [40]

b) A researcher obtained the following ordinary least squares (OLS) estimates for a Bank
using 120 monthly observations.

^
ln Y t =0.72+0.36 ln K t +0.09 ln Lt +0.78 ln Rt −0.07 ln M t
(0.19) (0.12) (0.19) ( 0.20) (0.02)
2 2
R =0.57, R̄ =0.52,F(4,115)=3.68

Where Yt are net loans made to the private sector, Kt and Lt are the bank’s capital and labour
inputs (number of employees) respectively; Dt is regulatory expenditure and Mt is
expenditure on Marketing. Figures in parentheses are standard errors, all variables are in
logarithmic form and RSS is the Residual Sum of Squares.

i) Interpret the coefficients on the variables (ignore the constant). [10]

2 2 2
ii) Explain the difference between the R and adjusted R ( R̄ ) statistics. [15]
iii) Individually using the t-test, test whether each coefficient equals 0, at the 5% level of
significance (ignore the constant) [15]
iv) Using a t-test, is the coefficient on ln Rt equal to 1? [10]
v) What do you conclude about the validity of the above model (assuming all diagnostic
tests are passed)? [10]

2) An investigator estimated the parameters in the model of a firm’s profitability using 60

monthly observations for 2010 to 2015.

ln pt =α 0 + α 1 ln y t +α 2 ln l t + α 3 r t + ut
t

Where pt is profitability, yt is firm output, lt are the levels of bad debts and rt is the interest
rate. This resulted in a residual sum of squares (RSS) of 0.43.

(a) When the two variables measuring bad debts and the interest rate were removed from
the equation the RSS increased to 0.48. Using an F-test, test for the joint significance
of the levels of bad debts and the interest rate. [25]
(b) If the investigator wanted to determine if their profitability was seasonal by including
11 seasonal dummy variables, describe how you could test for this seasonality using a
F test. [20]

( c) The firm Claverton Plc decided to examine the determinants of their output, using a
conventional Cobb-Douglas production function, where output yt depends on capital
 kt and labour lt (All variables in logarithms, standard errors are parentheses), they
produced the following results using 130 observations:

^y =0.59+0.27 log k +0.67 log l

log t t t
(0.42) (0.12) (0.35)
2
R̄ =0.68 , RSS=0.27

They then ran a second regression in which constant returns to scale were assumed to
hold:

log ¿ ¿
(0.16) (0.24)
2
R̄ =0.67 , RSS=0.29

i) Explain what restriction has been applied to the Cobb-Douglas production function
when constant returns to scale are assumed to hold and what the null hypothesis is
going to be in this case? [30]
ii)Use an F-test to determine if constant returns to scale applies in the above case. [25]

(3) a) What is autocorrelation and in the presence of autocorrelation are the estimators
still BLUE? Explain your answer
[25]

b) The following regression was run using monthly data, amounting to 50 observations:

^I t=0.18−0.52 Pt +0.88 Y t (0.08) (0.24 ) (0.39)

2
R =0.58 , DW =1.67 , LM (4 )=7.67

Where It is the demand for Ipods, Pt is the price of the Ipod and Yt is the level of GDP in
the economy as a whole. (standard errors are in parentheses). DW is the Durbin-Watson
statistic, LM the Lagrange Multiplier test.

i) Does the above regression suffer from either first or fourth order autocorrelation?
[20]
ii) Briefly describe the LM test for 4 order autocorrelation
th
[20]

c) Explain what heteroskedasticity is and whether the estimator is BLUE when it is

present ?. [15]

d) What are the main causes of heteroskedasticity and suggest a possible change to a
variable such as GDP leading to heteroskedasticity that can solve the problem. [20]

4)a) What are dummy variables and what are they mainly used for in applied
econometrics? [25]
 b) What is the difference between an intercept dummy variable and a slope dummy
variable? [25]

c ) With an example of your choice, explain what an event study is and assess some of
the potential problems with this approach. [25]

d ) Briefly outline how it is possible to test for a structural break in the data using dummy
variables.
[25]

5) a) Assess the importance of lagged variables in time series econometrics. [20]

(b) Consider the results from the following partial adjustment model (60 observations):

i^ t =0.23+0.57 pt + 0.16i t −1

(0.16) (0.23) (0.05)

R̄2=0.67

Where it is the nominal interest rate in the UK and pt is the inflation rate. (standard errors
are in parentheses)

i) Explain what a partial adjustment model is. [20]

ii) Interpret the coefficient on the lagged value of the dependent variable (it-1).
[15]
iii) Given the above result, what is the value of the long-run relationship
between it and pt? [15]

c ) Explain the concept of stationarity in time series data. With reference to the Dickey-
Fuller test, explain how you would test whether a data series was stationary or not. [30]

6)a) Explain with an example what panel data is. How can it solve the problem
of not enough observations when estimating a model? [25]

b)What is unobserved heterogeneity and how can a panel data model overcome this problem?
[25]

c) What is the difference between a fixed effects and random effects model, suggest a test to
determine which method should be used in the model? [30]

d) Suggest a method for estimating the fixed effects model. What is the difference between
a one-way and a two-way fixed effects model? [20]