ECS3706 Exam Pack-1 (1) (1) Ecs4863
ECS3706 Exam Pack-1 (1) (1) Ecs4863
ECS3706 Exam Pack-1 (1) (1) Ecs4863
1
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
ECONOMETRICS STUDY PACK
MAY/JUNE 2016
Question 1
(ii) usually population size may be too large or unknown therefore inference is made
on the sample drawn from the population using the estimated regression equation which
is the statistical technique that attempts to explain movements in one dependent
variable given independent variables
2
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
∑ . ESS is the variation of Y around its mean which is explained by the
equation it is measured by ∑ . RSS is the variation of Y around its mean
which is not explained by the equation, measured by ∑ .
TSS= ESS+RSS
(c) is the co-efficient of determination which measures the goodness of fit of the
whole equation. The problem with is that whenever an independent variable is added
it will increase even when it is an irrelevant variable. measures the percentage of the
variation of Y around its mean that is explained by the regression equation, but adjusted
for the degrees of freedom. Usually a higher fit does not guarantee a good estimation
because the fit can be only measure used.
Question 2
(iii) The which determines the goodness of fit of the whole equation
(iv) Bias
3
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
(b) Sequential specification involves the initial specification of an equation and
estimates it, then adding other independent variables until a plausible equation is found
while data mining involves estimating a variety of alternative specifications before that
best equation has been chosen.
(c) (i) Specification bias occurs when an independent variable is omitted which is
relevant in the equation.
(iii) To avoid specification the four important specification criteria must be followed
(refer to question 2 (a).)
Question 3
(a) The stochastic error term is added to the regression equation because it helps
capture the minor variations in Y which are not captured by the Xs. It is virtually
4
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
impossible to avoid some sort of measurement error hence the stochastic error
captures that. The underlying theory may differ from the chosen functional form
hence the need for an error term.
(b) OLS means ordinary least squares, it is a technique that calculates the Bs so as
to minimize the sum of p squared residuals which is calculated as .
(c) (i) Review the literature and develop the theoretical model
Question 4
5
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
(a) They are 2800 households in the sample which is given by ∑
= 22.6-0.755(11.2)
= 14.144
Y = 14.14+0.755X
6
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
B0 is the estimated constant which could be called the y-intercept. While B1 is the
estimated slope coefficient of the equation.
=1 -
= 1-
=0.90.
The explanatory variable are a good fit because 40% of the variation in Y is explained
by the Xs.
Section B
7
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
Question B1
(a). When it comes to the interest rate using the rule of 2 we fail to reject the null
hypothesis although the test statistic carries the correct negative sign but interest rates
when coming to the determination of money supply they are statistically insignificant.
This could be because both money supply and interest rates are the proprieties
(instruments) of monetary policy with the authorities being able to control the supply of
money which is done through the accommodation policy while on the other hand the
authorities control interest rate via the repo rate, therefore the direct influence of one
over the other may be limited.
Using the rule of 2 on the test statistic of GDP we reject the null hypothesis. Plus the
test carries the correct expected sign. This means GDP is statistically significant when it
comes to explaining the changes in money supply. Theoretically changes in GDP would
refer to changes in money supply via inflation effect plus the changes in economic
growth. The overall goodness of fit as measured by the seems to be good enough to
infer that it is a good equation.
(b) The equation seems to carry an irrelevant variable which is interest rate in the
determination of money supply.
(c) Yes it changes the explanations to the answer above. A simple correlation coefficient
of -0.89 shows a strong negative relationship between GDP and interest rates. This
means multicollinearity exists with a variance inflation factor exceeding S that means
the multicollinearity is severe
8
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
We could drop the redundant variable in this case interest rate since it is statistically
insignificant, we could increase the sample size to see if multicollinearity fades away
Question B2
(a) Bpc is the coefficient which estimates the substitute for keep; it implies that a R1
increase in the price of chicken results in a 1.4kg decrease in the quantity of beef.
Although this is theoretical incorrect showing signs of statistically insignificance.
9
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
(b) T-value is calculated as = = -2.21
(d) This regression equation has been incorrectly specified when it comes to PC
therefore PC should be dropped or its coefficient specified. Multicollinearity is present in
the independent variables with both PS and PC explaining almost the same outlook.
Question 3
10
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
(a)The above equation is a double log. The values of Y and the Xs have been logged so
as to draw a meaning-full inference out of them because naturally the values of Y and X
are relatively too large.
11
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
This econometric problem carries multicollinearity and misspecification of equations
(d) Yes they are problems related to Y which have been caused by multicolinearity
between the log of average income and the log of current income. Again serial
correlation exists between the log of the rate of interest and the log of average income.
12
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
MAY/JUNE 2015
Question 1
(a) (i) Zero mean error means the mean is normally distributed and most of its variables
are located to zero
(ii) Homoscedastic errors means the variance of the error terms are constant over a
wide variety of samples.
Question 2
(a) October November 2015 Question (a) and May June Question 1 (a)
(c) The theorem focuses on the classical assumption I and VI with 4 concepts paid
attention to:
Unbiased
Minimum variance
Consistent
Normal distribution
Question 3
(a)
(b) V I F Cooks at the extent to which a given explanatory variable can be explained by
all the other explanatory variables in the equation
Question 4
(a)
(c)
14
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
We introduce m-1 dummies with the other dummy being the slope dummy
Section B
Question B1
(c) The equation might contain multicollinearity because GDP and GDP R(-1) may be
one and the same thing but on of 0.979 implies that 97.9% of changes in Y is
explained by the independent variables which implies an overole good fit of the model.
Question B2
(b)
15
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
critical)
Degrees of Freedom KV2 n-k-1=2.045
D1 Unsure 0.007147/0.099133=0.721 Do not reject
Ho (t-test<t-
critical)
D2 Unsure 0.004917/0.135639=0.036 Do not reject
Ho (t-test<t-
critical)
Question B3
16
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
Double log: log (GDP)=0.63+0.682(log(M3))
Slope:
Slope:
17
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
Semi log (logX): GDP=356214+47254 (log(M3))
Slope=
18
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
19
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
OCTOBER/NOVEMBER 2015
Question 1
(b) Usually the population size is too large or unknown therefore estimating the true may
be difficult. We use the estimate to draw inference about the true linear regression
equation. The true is represented by B while the estimated is represented by B. B-B=0
which means the estimated B is an unbiased estimator of the true B.
(c) r measures the strength of the relationship between two variables. It runs from -1 to
1 with -1 showing a strong negative relationship and +1 showing a strong positive
relationship. measures the goodness of fit of the whole equation.
Question 2
(a) Discrete random variable has a countable number of possible values. Each outcome
can only take one possible number as the outcome. When throwing a dice you can only
get a number between 1 and 6, one cannot get 2.5. The probability of getting any
number when throwing a dice randomly P(X) = , there is a one in six chances that the
(b)
Stochastic error term is the value which captures the variations in Y not
explained by the changes in X
May June 2016 Question # (a)
October November 2015 Question 1 (b)
Question 3
20
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
(a) A linear functional relationship between 2 or more independent variables that is so
strong that it can significantly affect estimation of the coefficients of the variables. It
could be detected using high simple correlation coefficients and high variance inflation
factors (V I Fs)
(b) May June 2016 Question 1 (b) (iii) and (c) combined
(c) Irrelevant variables lead to (i) No bias (ii) then increase the variance. Both for the
dependent and independent variables
(d) Histograms are used to read statistical outcome of the data. Usually most economic
data are normally distributed the histogram would be well shaped. However discrete
random variables which have an equal probability usually follow a uniform distribution
and the histogram will be equal throughout.
Question 4
(a) Pure serial correlation occurs when classical assumption IV which assumes
uncorrelated observations of the error term is violated in a correctly specified equation.
Impure social correlation is caused by a specification error such as an omitted variable
or an incorrect function form.
21
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
The y-intercept us -0.1668 while the slope is 0.523.
Section B
Question B1
(a) A 1% increase in total household income will lead to a 0.036 kilowatt decrease in
electricity consumption by households. A 1% increase in the average price of electricity
leads to a decrease in electricity consumption by 10%.
(b)
22
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
(c) Reject Ho if t-test > t- critical. Do not reject Ho if t-test ≤ t- critical. T test = 245. T
critical = 2.35. Reject Ho because t-test>t-critical.
(d) Electricity has been log alongside income on total number of residence. Advantage
is that correct inference can be drawn from the logged numbers since the absolute
original values may be too large to draw inference.
Question B2
(a) Positive serial correlation may be present because total population of area (pop) and
total number of household resident within area are SD are related.
Ho! p≤0
Ha! p>0
DW-d= 2.08
dl= 1.49
du= 1.77
Qt=Bo+BaAt+BbBt+Ɛ** where the Ɛ** contains the omitted variable which could then be
written as Ɛ**= Ɛi-BcCt
(c)
Question B3
(a) They is slope dummy variable and the coefficient dummy variables
(b) The first quarter dummy variable serves as the benchmark dummy variable which is
the y-intercept
(c) Yes it makes sense because during the last quarter of the year it is the coldest
season in the USA therefore little or no sales are made
(d) Frig= 456.2 + 2.7 (500) + 2.42.5 (0) + 325.1 (1) – 86.1(0)
The production function shows the percentage change in output brought about by
production change in capital input, technology and technology input.
The advantage is that absolute output maybe large to draw inference therefore it is
important to log it
Log (Y)= log (a) + b log (X) where percentage change in Y is caused by percentage
change in X
(g) measures well variables of linear form when transformed variables may have a
small which means poor fit however this cold ne am incorrect indication.
24
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za
REED TUTORIALS Page 23
FOR FURTHER QUERIES CONTACT
PHILLIP MANGAVA
FOUNDER & CEO
PHONE O787912965
EMAIL info@reedtutorials.co.za
WEB www.reedtutorials.co.za
TWITTER @reedtutorials1
25
Reed Tutorials (pty) Ltd
www.reedtutorials.co.za