SVKM'S NMIMS

SARLA ANIL MODI SCHOOL OF ECONOMICS

Academic Year: 2019-2020
Program: B.Sc.Economics Year: Ill Semester : V
Subject: Econometrics -11 Batch: 2017-20
Date: 13th January, 2020 Time: 01:00 pin to 03:00 pin (2 hrs.)
Marks: 50 No. of p;ges:i ,
Re - Examination

Instructions:

1. Each question carries equal marks (10 marks).

2. Attempt any 5 (five) out of 6 (six) questions.
3. Figures to the right indicate maximum marks..
4. Answer all sub-parts of a question together.
5. Answer all the qHestions that require writing R-scripts together in one `.R' ffle.
6. Use comment lines sufficiently in R-script to highlight question number and
interpretations wherever necessary.
7. Name of the a.R' file should be specffied as per instructions given at the time of
examination.
8. Answer each question on a fresh page.

1. Consider the following models A and 8 estimated from the same dataset
containing information of 5 0 students who appeared for GRE :

A: 8HE£ = 275.465 + 34.486 coczcfr£ + 22.762mci€fr{

a: GTE[ = 268.856 + 22.143 cocicfr£ + 15.582777cifh£ + 16.356 coachj. mciffri

where, GREf is the marks obtained, coachf and 77iczfhf are binary variables taking

a value of 1 if the student took additional coacbing c,lasses and studied

mathematics in higher secondary respectively.

Based on the above models assuming that all coefficients are significant at
conventional levels of significance, answer the following question:

(a) Calculate the average marks obtained by students who:

(i) took additional coaching and studied maths in nigher seconday,

flEffl
 (ii) studied maths in higher secondary tut didn't take additional coaching, 1
(iii) took additional coaching but didn't study maths in higher secondary and 1
(iv) didr't take additional coa`ching and didn't study maths in higher secondary. 1

What do the differential intercept coefficients in Model A signify?

Calculate the difference in the average marks obtained by the students who:

(i)didnottckeadditionalcoachingbutstudiedmathsinhighersecondaryandthe
students who took coaching and studied maths in higher secondary.

(ii) took additional coaching but did not study' maths higher secondary and the
students who took coaching and studied maths in higher secondary.

(d) Briefty describe the utility ofModel A and B thereby describing the situations in
which they are used respectively.

2. (a) How do Logit and Probit models improve over Linear Probabifity Model in
estimating models with binary dependant variables?

a) Given the following estimation result from a Logit Model:

iusttffincei =. -1.715 + 0,1969 r€f{re£ + 0,00230 frfrincomei

-0,14596 cig€£ + 0.11426 gdreyeci7i£ + 0.5786 mc!rrzgd£

where,i7Lsffittnc€brefirefand7rLCFTi,ediapebinaryvariablesthattakesavalue
of 1 (one) if the person has health insurance, is retired and married respectively;
0 (zero) otherwise. Further, frfa£7ico7r.eb czgef and ecfucygGr{ denote the

household in.c.cme in thousaiids of rupees, age and number of years spent in

education respe ctively. Based on the above output calculate the odds ratio with
respect to each regressor in the model and interpret the same. S

(c)Refertothedatagiveninprobit.csvwhichcontains915observationsonthefollowing
5 (five) variables:
j7%6Jz.Jfr @z.7?OWJ: 1 if the student has published three or more articles, 0 otherwise;
ge72der/cfearczc}e7'J:genderofthestudent(male/female);77£arz.edrchor¢cfeJ~J:yesif
the student is maried, no otherwise; fe.dg /73%REe7~z.c/: number of children less than 6
yearsold;and77!e7?for/7?3t77Se7.i.cJ:nunberofarticlespublishedbythestudent'smentor.

Makenecessaydatatransfomationsandrunaprobitmodeltoestimatetheprobability
ofpublishingthreeormorearticlesasa.froctiontheotherfourvariablesandcalculate
the pseudo R2 for the estimated model.

ENH
3. (a) How do errors in measuring the regressand and regressors affect estimates of

coeffroients associated with regressors and their standard error? Explain with

suitable derivations.

@) Explain the method of Kyock transformation for est.imating an infinite order

distributed lag model. What are its main features?

4; (a) Derive the mean and variance ofa random walk process without drift and without
trend, thereby showing that it is a non-stationary process.

(D) For the series Yl and Y2 specified in time.tat, check for stationarity using
Dickey-Fuller or Augmented Dickey-Fuller test by taking into account the
correct functional form of the underlying data generating process.

5. (a) What do you mean by "sinultaneous eqution bias"? How does it affect the least-
square estimators? Explain with a suitable example.

a) Considerthefollowingmodels: ' .

Model A: -AGDpt = yo + yiAAf£ + y2AMf_1 + tti£

. Model 8: AGDP£ = Go + GiAEf + G2AEt_1 + tt2t

where, AGDpf, AMf and AEf is the percent change in Gross Doinestic Product,

Money Supply and goverrment expenditure respectively.

We estimate the following equations on lines of Davidson-M.acKinnon J Test:

Equation 1: A65Tpf = Cro + CriAMf + cr2AM€_1 + Cr3fi?b£

Equafion2`. ArffBfpt = Po + PLAEt + P2AEt+ + P3ffiat

 where, fiEaf and /ffbt are the fitted values derived from model A and 8
respectively.

Estimated values :
Equation 1 Equation 2
Coefficient Estimate StandardError Coefficient Estinate StandardError

•0.294
fro 0.027 Po 0.182 0.019

Crl 0.485 0.085 P1. 0.354 0.077

Cr2 0.026 0.012 P2 0.016 0.007
Cr3 0.384 0.234 P3 0.539 0.164

j32 = 0.746 Ill = 294 j32 - 0.878 7t2 -294

Check and report the significance of all the coefficients at 5 percent level of
significance-Use the above information to perform a Davidson-Mackirmon J-test
andchoosetheappropriatemodelfromthespecifiedmodelsAand8thatisbetter
in explaining the growhi in GDP (AGDpt). Also provide suitable reasoning for
the sane.

6. (a) Using the data in probit.csv, and the model estimated in question 2 (c), calculate
the Likelihood Ratio qR) sta:tistic for testing for joint significance of the model
fit along with its associated p-value and interpret the result.

@) Calculate the average marginal effects with res.peat to each regressor in the al]ove
Case.

(c) Using the data vocab included in the package ccz7:Dc#cz, estimate a suitable linear
model to check for difference in the vocabulary test score (vocczb#Ja7)/) across

sex of the respondent (sex). hteapret the results of the regression model and plot
the boxplot indicating the vocabulary test score across sex of the resfiondent.

(d) Estimate the above model after controlling for their years of education
(education) and interpret the difference in results thus obtained.
 SVKM'S NMIMS
SARLA ANIL MODI SCHOOL 0F ECONOMICS
Academic Year= 2018-2019
Year: Ill Semester: V
Program: B.Sc.Economics
Batch: 2017-20
Subject: Econometrics -11 ;
Time: 11:00 am to 01=00 pin {2 hr-s.)
Date: 4th january, 2019 f
No.of pages:4
Marks: 50
7€-` Re Examination

Instructions:

1. Attempt any five questions.

2. Figures to the right indicate maximum marks.
3. Answer each question on a fresh page.
4. Provide figures wherever necessary.
5. Create only one R file for all data based questions.

1. Consider the model depicting relatiol|between log of real GSDP (rGSDP), inter-state
imports plus exports as a % of GSDP (trade) and strike dummy (which takes the value
1 if the state experienced.strike leading to more than 1 day of work loss and zero
otherwise) for 30 states. The result obtained is given below. (Values in parenthesis
indicate standard errors)
lr\(rFaslDPL) = -0.28 + 0.06tradet -0.14strikei -0.008tradei * strikei

(0.08) (0.003) (0.04) (0.0005)

(i) Interpret the coefficients and plot the regression lines. (5)

(ii) Suppose you wish to test whether strike has any relation with ln(rGSDP) or not.
Write down the null and alternative hypothesis for the sane and the rejection
rule. (2)

(iii) Find the difference in expected growth rates of the below two states:
Statel : trade variable increases from 2% to 4% and the state did not experience
any strike during the year.

Page 1 of 4
 State2- trade variable increases from 3% to 6% but the state did experience
multiple strikes during the year.
Comment on your findings.
(3)
2.

a. Consider the Affairs data in AER package in R that provides infidelity data based
on a cross-section data from a survey conducted by Psychology Today in 1969.
Using the dataset answer the following:

(i) Regress education on gender dummy and test for the presence ofheteroscedastic
residuals-using White's test. Report the heteroscedasticity consistent standard
errors for the coefficients. (3) i
.i,
(ii) Also, check for the nomality of the regression residuals using Jarque-Bera test.
Mention the null and alternative hypotheses. (2) -i
b. A study was conducted for a sample of 1 196 individuals to understand the factors
thataffectsmokinghabits.ThedependentvariableSmokerisabinaryvariable.The
results obtained by the researcher from the Linear Probability Model are given
below.

Smoke7ii = 1.12 -0.004Age! -0.02Edttc{ + 1,03 x 10-6J7icome{ -0.01PCJGS79i

+fti
where Smoker = 1 for smoker and zero otherwise, Age is age in years, Educ is number
of years of schooling, Income is the family income, PCIGS79 is price of cigarettes in
1979. `

(i) Interpret the coefficients in the model. (3)

(ii) For Educ = 12, Income = 1000 and PCIGS79 = 1, at what age(s) will the

probabilities exceed the boundaries? (2)

3. Consider the Affairs data in AER package in R that provides infidelity data based on
across-sectiondatafromasurveyconductedbyPsychologyTodayin1969.Usingthe

dataset answer the following:

(i) Define a binary variable ¢jLT¢!.r which takes the value one if the individual had
one or more extra-marital affairs and zero otherwise. (I)

(ii) Estimate the log odds of having an ¢jLT¢J.r as a function ofge#der, ¢ge,%.#g
and j/e¢rs"¢rrz.ed. Interpret the regression output, edso, provide the odds
interpretation.

(iii) Find the pseudo R-squared of the model.

Page 2 of 4
 (iv) Find and interpret the marginal effects at the average. (3)

4. Consider the Affairs data in AER package in R that provides infidelity data based on
a cross-section data from a survey conducted by Psychology Today in 1969. Using the
dataset answer the following:

(i) Using a two sample t-test can we say that years of education differ for males
and females in the sample? (2)

(ii) Using F-test can we say that variation in wages also differ across male and
female workers? (2)

(iii) Regress education on gender dummy and interpret the regression output.
Comment on the relation between the coefficients to the sample mean values in
RE
(i) above. (3)

(iv) Now also regress log(education) on gender and compare it to the model in (iii)
above using AIC values. Which model is better and why? Explain. (3)

a. Consider the adaptive expectations model given by

y£ = cro + crixc + cr3yt-1 + t7t

where t7t = (ttc - (1 -y)ttt_1) and "a has mean zero, variance a.2 and tt£'s are not
serially correlated. y is the coefficient of expectation. Are the error terms t7c and

ut_1 correlated? Justify your ansvirer. (2)

-\
b. The relationship between per capita personal consumption expenditure (PPCE) and

per capita personal disposable income (PPDI) is .given by the model below
PPCEt = -2101 + 1.12PPDlt + 0.05PPDlt_1 -0.16PPDlt_2 + tit

(i) If the income of individuals increases by €100 in time t due to a one time aid

provided by the government to all iridividuals of the country. Find the

immediate impact on consumption expenditure, edso find the effect on
consumption expenditure in the subsequent periods. (2)

(ii) If the income of individuals increases by €100 from time t, find the immediate
impact on consumption expenditure, also find the total effect on consumption
expenditure in the subsequent periods. (2)

Page 3 Of 4
c. Are the mean and variance for random walk with drift and trend constant? (4)

a. Consider the EustockMarkets data in R containing daily closing prices of major

European stock indices: Germany DAX (Ibis), Switzerland SMI, France CAC, and
UK FTSE. The data are sampled in business time, i.e., weekends and holidays are

omitted.

(i) Estimate a Koyck type ofmodel where the dependent variable is log(FTSE) and
explanatory variables are log(DAX) and one period lag of log(FTSE). Interpret
the regression out. (3)

(ii) Find the short run and long run effect on change in closing price ofDAX on the
closing price ofFTSE for the model estimated in (i) above. (2)

b. Consider the system of simultaneous equations given below.

y|t = C¥o + C¥|y2C + C¥2X|t + u|t

Y2t = Po + PiYit + P2X2t + u2t

(i) Derive the reduced form coefficients. Based on the structure of these
coefficients, comment on the aspect of identification. (3)

(ii) If we drop x2 from the system, will the aspect of identification be affected?
Justify your answer. (2)

Page 4 of 4