pag es.

]
[Th is que stio n pap er con tain s 4 prin ted

/. You r Rol l N 0 ..... ..... ..... ..... ..... .. .

/.
Sr. No. of Que stio n Papei• 681 2

. '•-.. '1'2481402- " . .

Uni que Pap er Cod e

Nam t of the Pap er Basic Econometrics ·

-6 .2N ame of the Cou rse B.A . (H) Bus ines s Eco nom ics 202 3
(LOCF)

Sem este r IV

3 Hou rs
Duration

Maximum Mar ks 75

Ins truc tion s for Can did ates

I. Wrife you r Roll No. on the top immedia

tely on receipt of this question paper.
s.
stion l and any five othe r ques tion
:_ Q ... ~'i>lion I i:. co111vui:;(•ry. Ai.te mpt Qut:

the foll~~!BB.~.t~_tem ents are true or

Q.i, Provide reas ons and stat e wheth_~r ';

false.

(i) = /31 + f3 2Xi + Ut, if P2 is multiplied by

ln the reg ress ion model In Yt
100, we obtain the growth rate esti mat e of
Y~. -r ~
\
\
I

(ii} Adjusted R2 is always lesser than una dju

sted R2• fcJ .s e j T.-..U- \
2..
the sam ple reg ress or values ,- \
• (iii) Ceteris par ibu s, the mor e disp erse d are ~
anc e of its coefficient.
in a line ar regression, the gre ater is the vari
than the cho sen level.of
(iv) If p•value of a test statistic is gre ater
esis at the a level of
significance a, then we reject the null hyp oth
significance.
t= ~

P.T O.
 2

residuals are
(v) In th~ regression model Ye = P1 + P2Xc + u,, if the OLS
patte rn is obse rved , then it
plotted against time (t), and a distinct
indicates the presence of heteroscedasticity. F c:...!s
c:... . (5 x 5)

s of rupe es (Y)
· A firm's five-year data of profits earn ed in thou sand
is given below,
and amo unt of a good produced in million units (X)

2010 · 2011 2012 2013 2014

Year
Y (in Rs. '000) 60 90 80 70 so
6 5 4 3 2.
X (in million units)
.,._ .
Use the data to: ~ +
of Yon X 5 '-1
(i) Estimate the equation of the linea r regre ssion line
(ii) Calculate the residual sum of squa res for the regre 1
ssion . It c; = 9 '-f 0
(iii) What proportic;m of the variation in Yis expl
ained by X? R =.16
slop e coefficient 1 · _ • ., S
t _(iv) Test at the 5% level of significance whe ther the
is grea ter than 2. b -=- 0 · ~ 1" 8 d.P 'h.o-\ 'f""C-1 ~ He
1' Calculate the 90% confidence interval for the mea n pred(3,1,
GJ icted
-cc - 2- ..'>

1,3,2)
value of Ywhen X = 4.5 ',t

data collected
Q.3. (a) The following regression model was estimated using
from 34 retail stores:
?t = 5837 .53 - 53;217Xz{+ 3.613X3,
se = (628 .151 ) (6;853) · (C.6852)
RSS =1947 2.33 TSS = 52093.55
(in Rs.) at shop
wher e Yi is ~e monthly sales of a good; X2 t is the price
thou sand
i; and X3t is the "in store" promotional expe nditu re (in
rupe es) at store i.

al slope coefficient of X3 - 2. . 6 2.. \

(i) Inter pret the estim ated parti 0 0
(ii) Calculate the values of R and adju sted R2.
2 R.1,.. = o. b l G2 R.. -

(iii) Test the model for overall goodness of fit at a = 0.01. F c =

& 5 . 'l6 1 ::
5
.

popu latio n • .J: -+ J .2.; 3 I ) 0 ' 0 1

(b) List the st_eps requ ired to test for the norm ality of the
. (1,3,3,3)
distu rban ce term.

to the Savings
Q.4. Cons ider the resu lts of the regre ssion model appl ied ·
(Y) and Inco me (X) data for the two perio ds:
to 1988 -89. ·
Period-I: 1989 -90t o 1995 -96a ndPe riod -II: 1974 -75
P, = -300 4.95 9 - 25595.554Dr + 0.249Xc + 0.075DcXt
(-0.9 10) (-2.945_) . (13.752) (3.454) R2 = 0.9951
t=
whe re D = {1 f~>r obse rvati ons from Perio d I
0 for obse rvati ons from Perio d II
 1
3

(i) What is the base/control category for this regression? Pd 2-

(ii) ls such a model called and ANOVA model or an ANCOVA model? AN Co vf\
(iii) Interpret the meaning ofthe coefficient value 0.075? ~ ~ D'"' X t · .
(iv) Test whether the model suggests the presence of structural ~J. ~ f.>--\ ".,It;, 6
change in the two periods. · ·
(v) Write out the regression equation for Period-II, (1,1,2,4,2)

Q.S. In order to check for the presence of multicollinearity in the model

Yt =P1 + P2 X2i + p3X3t + u,, an auxiliary regression was conducted
and the result is:
X2t = 2.456 + 0.7952X3t
·se = (0.56) (0.0598) R2 = 0.90

(i) Compute the variance inflation factor. Do you find evidence of

multicollinearity? VI~~ \0 · "lu
(ii) Are the estimates of p2 and P3 biased? Are these ineffi~~ \AfhY ~~..J GL, ~ N 1.--iP/J.J
or why not. ~ / ""° ~,
(iii) Without conducting the auxiliary regression, how could one
detect the presence of multicollinearity?
(iv) Describe two remedies to deal with multicollinearity and their
limitations. · (2,2,2,4)

Q.6. In a regression of average wages (W) on the nurriber of employees (N)

for c.1 random sampi~ of3G Hrms, the following resui,,.) were:obtained:
Regre~sion 1: () v ~~ _ ~ ·
r'1> IL ~\e /,- .. .,s'
W[ = 7.5 + 0.009N, .,,- 11 L .~
t= (16.10) 2
R = 0.9 JJ_ . "°"' t-1
Regression 2: ~ ~ _ . t \J~
Wz . 1 ~~ ~~cl)
N, = 0.008 + 7.8 Nt o.o -o~ -> . ~-,
t = (14.43) (76.58) R2 = 0.99 <:.A
~!'" - ).-
(i) How would you interpret the two regressions? ~ ~ . -:J.. 6 ~l.
(ii) What might be the reason for transforming Regression 1 into ~ i v ~ tl\, ..P- ol-
Regression 2? What assumption has been made about the error 6'.., \I/
variance in going from Regression 1 to Regression 2? 1- ~
(iii) Can you relate the slopes and intercepts o°f the two models? Can N f ~ \.
y_ou compare the R2 of the two mod~ls? Give· reasons. ~J 1.- ~ · c,..-r··'.A--<
(1v) How would you check for heteroscedasticity, test its existence
and then remedy it? (use Regression 1 as the reference) (2,2,2,4)

j
~
~/'~.· . C\i"~~.,.J
~ ~ '~ ~

;c ~~~
.

()9~ 1/,
~ b.~¢.\ v\t>
V P.T.O.
 4

Q.7. The results of a logarithmic regression of demand for food on price

and personal disposable income is given as:
log Qt = 2.34 - 0.31 log Pc + 0.45 lc,g Ye ,±j).65 log.Qc~.1 .
se = (0.05) (0.20) (0.14)
R2 = 0.90 d = 1.8
where Q = food co·nsurription per capita; P =food price; Y =real per
capita disposable income • - Jv,ocl
(i) Interpret the value of the coefficient 0.31. e,l~7 -b P~ ' " ~ c...::> +.
(ii) Just by looking at the estimated re_gression, do you suspect serial c....,-r-a J~
correlation in it? Why or why not. 'I ~ ·· -1--o .
• (iii) Which test do you use to confirm your suspicion and ~hy7 / Y
.· (iv) Outline the steps of the above mentioned test and°provlde a/ y
conclusion on the basis of your calculations.
Q )
,,,,~

: .,..,_ -~~Lf ~
(v) State the conditions under which you may choose differ~nt tests
for serial correlation. (1,2,2,~,3)

(100)
 ::.. ,-~,. . ,
~..,,.
. . '"4.'~".T....::.~ ~ ,--~~~~~i·• ~ ·· .
__

f . .,

~-
Question paper 68lZ
oOS to
1 2481402 (Basic Economet . Question 2
rics)
p>(flJ Bus ness Economics 2022 _
· 1
23 [3,1,1,3,2}
y X x2 xv yz Y"
1 60 6 36 360
U" U"2
-tlon [5 marku 90 3600 78 ·18 324
5 25 450
80 8100 74 16 256
4 16 320
give 1 mark~ if no explanation is provided 70 3
6400 70 10 100
9 210
zero marks if explanation is wro · so 2 4
4900 66 4 16
100 2500 62
4 marks if explanation is correctn:nd . -12 144
not~wrong. answer1s 350 20 90 1440 25500 0 840
(i) v"fRUE: d(ln Yt) = 2.!!!:i _ . (i)
350 =Sa +20b & 1440 =20a + 90b
dx Yt dx - P2, Regression equation: ? = 54 + 4X
Hence p~ is ~e estimate of the rate of
proportional change (or the continuously (ii) RSS = 840
compounded rate of growth) in y with
respect to X. (iii) ESS =160; TSS =1000
Mul~plying by 100 provides the growth R2 = 0.16 or 16% of the variation is
rate m percentage terms. explained.

(ii) FALSE: when k = 1, adjusted and (iv) Ho: P2 =2; Ha: P2 > 2
unadjusted R2 are the same.
se ~ = 5.29; t =~
= ✓"to = 0' 378
I
illl .. S.29
t3;0.os = 2.353; ltl < 2.35
TJ\UE: Since ii2 = 1 - !!!!.. . ~ and
~ TSS n-k Unable to reject HO.
1 n1,l > 1 for usual models. Therefore P2 is not statistically greater
-r.:.- than 2 at 5% level of significance.

I
_f.: qZ 2
. (iii) F1J.SE: Var(P2)
.~.
= -:E x,z = nux2
_u__ Hence an (v) give 2 marks to all who attempt 0.2
in1rease in the dispersion of X values
wif uld increase the variance of X and Yo = 72; t3; o.os = 2.353;
t\•retore _decrease the variance of the
se Cy;O) -_ <1 2 -
(X _X)2)
(1 + __;, _ 0 -
slttpe coefficient.
)/; n rx?
;r,~

(iv) FALSE: p value ,is the probability of Type

1 0.25) .
I "1-ror committed if the null hypothesis = 240 ( 5+ 10 = 7;348
is rejected. a is the permitted level of
Type I error. Therefore if p > a, null Confidence limits are 72 ± 17 .29
hypothesis cannot be rejected. Confidence interval (54.71, 89.29)

(v) FALSE: A pattern in the ui values with Question 3 [1,3,3,31

time shows that Ut values ~re related to
those in the past or the future. This is (a)
indicative of autocorrelation. (i) 3.613 is the average increase in Y due to
OR a 1 unit increase in X3 when X2 is
Heteroscedasticity occurs when the constant.
variance of the Ut terms is related to the
values of a regressor. So the pattern can (ii) R2 =1- 19472.33
52093.55
= 0.6262
ind_icate heteroscedasticity only if the
u,
dispersion of values changes over time . iP = l- (1- R2) G:) = 0.6021
and when time is a regressor of the
model. (iii) Ho: /12 = /13 = 0 or H0 : R2 = 0
Ha: at least one slope coeff is not zero
= 32621.22 31 = 25.967
Fst at 19472.33 · 2
Fcrtt = F2,31;0.01 = 5.39
 · t H. o·
Since Fstat > Fcritlcat, reJe~ Ques tion 5
r2,2,2,11
The mod el is statis tical ly signi fican t at
1
the 1% level. (i) VIF = - -2 = -1 = 10.
1-Tz,3 1-0.9
A VIF grea ter than 5 is cons idere d
(b) Use the Jarqu e-Be ra test (1 mark )
evid ence of mult icoll inear ity.
Step s (2 mark s) .
1. Run the regre ssion and estim ate the il
(ii) p;_ and p'; are both unbi ased and
term s.
2. Calculate the coefficients of skew ness mini mum varia nce estim ates of the
popu latio n valu es.
(S) and kurto sis (K) of the erro r term s
52 (K-3) 2] Reas on is that none of the assu mpti ons
3. Calculate JB statis tic n [ + ~ of CLRM are viola ted and there fore the
6
4_Set up the null hypo thesi s that u is Gaus s Mark ov theo rem rules that the
normally distr ibute d again st the estim ates are mini mum varia nce in the
alter nativ e hypo thesi s that is is not class of all linea r and unbi ased
normally distr ibute d. estim ators .
S. Find the critical value x2 z:a at a.
(iii) R2 valu e is close to ± 1 whil e t valu es of
Question 4 [1.1,2.4,21 the regr esso r coeff icien ts are
insig nific ant; .
(i) The base categ ory is OR
Period II (1974-75 to 1988 -89) the sign of any of the regr esso r ~•
coeff icien ts is wron g are indic atior ls of
(ii) Since there are both quan titati ve and mult icoll inear ity.
qualitative varia bles, the mode l is
ANCOVA. (iv) any 2 of the following three. 1 mark for
the remedy and 1 mar k for its limitiition
(iii) 0.075 is the addit ional incre?.se in
savings in Periu d i for on~ unit incre ase
Drop ping a varia ble wou ld remo vJ the
in income comparP.d with the base
prob lem of muit 1coll inear ity but could
Period II.
lead to a more serio us prob lem of a
(iv) give 2 marks [or each test Hy_/?_othesis, t spec ifica tion error , whe rein the
stat, t critical and decision be given 0.5 estim ated coeff icien ts may be biase d
each. with wron g stan dard error s.

*
Ho : Pz = 0 again st Ha: Pz 0,
tcrit = t1a,o.01 = 2.878
Tran sform ing the mod el to a log mod el
may work in some case s but carri es the
ltstatl = 2.945 > tcrttical risk of an erro r relat ing to a wron g
Reject H0 • Pz is statistically different funct ional form .
from zero at the 1 % level.
A ratio trans form ation with a regre ssor
*
Ho: p4 = 0 again st Ha: P4 0, tcrtt =
t1a,o.01 = 2.878
or with a comb inati on of regre ssors may
work but it carri es the risk of creat ing
ltstatl = 3.454 > tcritical heter osce ~asti city.
Reject H0 • P4 is statistically different from
Question 6
zero at the 1% level. [2,2.2,4]
.I We conclude that there is evidence of (i) Regr essio n 1 is the simp le 2 varia ble
struc tural change both due to inter cept linea r regre ssion wher e 7.5 is the
I
I and due to slope. avera ge wage for a firm with zero
I empl oyee s and 0.009 is the incre ase in

I~ (v) Equa tion for Perio d II is:

fr = -300 4.95 9 + 0.249Xt
avera ge wage for ever y addit ional
employee.
Regression 2 is a recip rocal mode l wher e
I 0.008 is the limit ing value of W /N as N
beco mes indef inite ly large . And w/N

I
 c0mes indefin itely large as N tends to
4. Conclude autocor relation If lhl > Za/2•

In order to deal with heteros cedasti city

OR
/I It has been assume d that the varianc e of
Breusch Godfrey Test
u is propor tional to N2 i.e.u?= N/ <12.
1. ~o: Pi = P2 = ... = Pp = 0
2· Fmd Ui from the given regression
(Iii) Multiplying Regres sion 2 by N provide s a 3 . Run an auxiliary regression:
result very close to Regres sion 1,
flt = a1 + a2Pt + a 3 Yt + a 4 Qt-i
Consta nt and the coeffic ient values are
interch anged in the two regress ions.
+ P1 at-1 + ... + Pp at-p
and calculate R2 •
2
4. Find the BG test statistic (n - p )R
R2 of the two models cannot be
5. If statistic Is greater than x n-p,a re,·ect
2
compar ed,sinc e the regress and variabl e
is differe nt an.d therefo re the TSS will be Ho and conclude autocorrelation
differen t.
(v) Durbin Watson d test is used to test for
first order autocor relation but cannot be
(iv) We could use the graphic al method of
used if there is (i) missing data, (ii)
plotting the square of 1lt against the X
lagged regress and as regressor;, (iii) lack
values to check for heteros cedasti city. of intercep t term, (iv) non stochastic
Run the Park's test, or the Goldfeld regressors, (v) non normal error terms.
Quand t test or White's General Test to
test it. BG is a more general test that may be
Remed y the problem using the used to test for higher order
transfo rmation in Regress ion 2 or use autocor relation and is less restrictive.
White's heteros cedasti city consist ent
standa rd errors. optional
Durbin h test may be used instead of d
Questio n 7 [1.2,2,2,3] test iflagged regress and is included as
regress or
(i) - Q.31 is the elastici ty of per capita food
cci'psumption with respect to food price.
Or,
pQr capita food consum ption on an
aiFrage falls by 0.31 percent when food
p~jce rises by 1 percent .

(ii) We could expect serial correla tion since

it is a time series model. Also food prices
and demand may exhibif cobweb
phenom enon. Further demand in one
period is related to demand in a previou s
period.

(iii) We could use the Durbin h test or the BG

test. Since a lagged regress and is a
regress or, the Durbin Watson d test
would be inappro priate.

(iv) Durbin h test

= 1 - 5!:.2 = 1 - 0.9 = 0.1
1. Calculate jJ
2. h = p ,----n -- = 0 1 , - - ; -
1-nfvar(.d;) · ✓~
3. For hypoth esis H0 : p = 0; Ha: p *0
h is Z distribu ted for large n