emolrlcalstudv - shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · the above test...

20
Chapter - 6 EmolrlcalStudv 6.2.4 6.3 Regression on Inflat~on Conclusion 220 222

Upload: others

Post on 13-Jul-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

Chapter - 6

EmolrlcalStudv

6.2.4

6.3

Regression on Inflat~on

Conclusion

220

222

Page 2: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

The choice among alternative functional forms rnvolves a compromise among

several crltena including economic theory, goodness of ht and simpiic~ly

defin!te role c~tes which form is an appropriate one for a grven problem Many

econom~sts, researchers who have studied the impact o f various changes of

monetary instruments employed Regression Stahstics to study the ~mpact of

various monetary changes

~u l t l p le Regresslon Technique 1s used In the study to analyze the impact of

changes In varlous monetary Instruments on the growth of deposits, credlt of all

scheduled commerclal banks, GDP, Wholesale Prlce Index and broad money

There is also a need to ellmlnate the posslblilty of var~abies belng auto correlated

~n which case the results may be lneffic~ent Hence, it Is important to rule out the

presence o f auto correlatron Many statistical tests have been developed for

testing the presence of auto co-relation, but amongst all these tests the one

developed by Durbln & Watson appears to be s~mple and reliable Mult~ple

regresslon results tested uslng the Il2 and F Tests, besides the Durbln &Watson

Regressron relatlon prov~des the relatlonsh~p between the dependent variable 'Y'

and one o r more Independent variable/s (X) Regression relationship proves

estlmate of the total variance In the dependent variable that Is assoclated with

varlance In the Independent varlable/s It prov~des a powerful method of provlng

that there 1s a stat~stically slgniflcant relationship between a dependent vanable

and one or more independent var~ables The regresslon results are generally

used to extrapolate for future periods based on the past behavlor of the

dependent variable to movements of the independent variable/s

The general form of mult~ple regression model is

Where - .- -

Y

a

b P

XI ,

Dependent Variable

Value of y when x equals zero, Called the Y - Intercept

Slope t Change in y resulting from a one unit change in x

Independent varlable/s m,

Page 3: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

o.r 1. u r s r u r L r r r IBSCS ~.onauctea, enunciated here under

I I I F* > F - Infer HI - 1

=t

): ( YI-9 )2 1 -

I ( YI-Y )'

1 ( M- 9 l2 n - k

): ( YI-Y l2 n - I

DW is between 0 and dl, or greater than (4,

dl), reject the null hypothesis and Infer:

that there Is posltlve auto correlatlon

Equatlon Stated Value

Lies between -1 & +I Valuer

nearer to one denote bettel

reliability of the method for forecast

It Indicates a strong co-relatior

between X and Y Posltlve value:

Indicate posltlve correlation and

negatlve values Indicate Inverse

correlatlon Values nearer to zero

Indicate absence of correlation

DW Is between dl and du, ar between 4 - d ~

and 4 - dl the test Is ~nconcluslve

DW is between du and (4- du), accept the

null hypothesis and lnfer that there Is nc

RSSl(k-1) ~ ( V I - Y ) * / ( ~ - - ~(YI.- ?I)'/( n - k ESS/(n-k)

I I I auto carrelat~on I

F* < F - Infer H, -

No relationship

-

DW is between 4 -dl and 4, there 1s

negative auto correlatlon,

Page 4: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

6.1.2 Details of Var~ables used

/ n / Number of observations

X i n

Y

XI n

-

Number of estimated parameters

Fitted value of y = atb,

Independent variablels

Dependent Variable

Mean of the ~ndependent var~ables

Mean of the dependent varlable

- -

a

b

e

S2y,

s2v

Sb

Value of y when x equals zero Called the Y - intercept

Slope +Change in y resulting from a one unit change in x

Residual error that remains after fitting the mod4

The standard error of estimate

Expialned variance t SZyI -

Measure of the amount af sampling error in 'b 'SbC SJ\/ZXI

Page 5: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

6.1.3 rests for Goodness of fir

The cho~ce among alternat~ve functional forms invoives a compromise among

several criterra ~ncluding economic theory, goodness of fit and srmplicily NO defrnlte role rites w h ~ h form 1s an approprrate one for a g~ven problem The

statlstrcal method used 1s decided basrng on the problem under swdy The

common agreement on the cr~teflon of goodness of fit is rehed upon R~ The higher the value of R2, the larger the Propoition of the dependent vanable

explained by a set of Independent var~ables

R2 gives the propoftlon of vanation of Y that can be accounted by variation m x The R2 can be calculated as follows

@ l~es between -1 and +I Values nearer to one (1) denoted better

rel~ablllty of the method for forecast It ~ndlcates a strong co-relation

between x and y

6 1 3.2. dd/usted W

The R2 does not adjust the loss of degrees of freedom n - k and n - 1

~onsequently, the unaausted R2 will become larger and larger as more ~arlables

are added t o a model even though the adjusted R2 may be decreasing because of

the (n - k) Thus the unadjusted may mislead one to believe that additom1

varlubles are useFul in modeling /f when in fact they are not The use of

Unadjusted R2 In fllterlng models that Is too comp/eX Thus to ellmlnate this

lacuna, the concept of aojusfed ~ ~ - e f f i o e n t of determ~natlon Is used. Adlusted

co-efficrent of determrnahon compares ~ z y x to 52y, me ad/uskd CO-effrclent of

detenninat~on equals the pmpo&on of the vanance rndependent variable 'Y'that 1s '%(Ylained or ehmlnated through the reiationshlp with the independent "adable

X' The adlusted R2 can be calculared as fOl/Ows'

207

Page 6: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

n - k = I -

Z ( Yi- Y v2

The value of adjusted co-efficfent Of determrnation whfch rs symbcl~cally denoted

hereafter as Adlusted RZ also vary between -1 and tl

The comparison between two or more variable is done throf~gh the F-statistic ( F

stands for the name of Rsher, who Invented thfs stat~stfc) which ~n fact is the

ratio 01' two varfances under study If the two-variance estimates are close to

each other, thefr ratio will approach unity The greater the dlfference between

the two variances, the greater is the valve of F-rabo, which In turn suggests that

the difference between two variances is significant An unb~ased estimator of

common variances may be obtained by poahng together variances of the sub

samples through formula Since the estimates of variances are independent, It

can be shown b a t their ration follows and F-distnbutlon with {(k-1) and (n-k))

degrees of freedom F- Value IS rndicatfve of a dlfference between two variables

that IS not due to chance or The statlshcal test is

RSSl(k-1) Fk-I, 0 k = =

Ufl-Y) (

Where ESSl(n-k) ~(YI.. PI)* 1 ( n - k RSS -- Readual sum of Squares ESS z Explained rum of Squares

The Calculated F- value is compared to values from book appendix F with k-1

degree of freedom ,n the numerator and n-k degrees of freedom In the

denom~nator This F- Value IS used to pehrm the foilowing h~pothesls test!

Ho (Nu// Hypothesis) = There 1s no statf.st/cally significant relatlonshlp b~2tn"Jen Y and X

H1(akrnahve Hypothesfs)'There 1s a statist/cai/y s~gniflcant relationship between Yandx

If F (Calculated) -?= F (Table) inter Ho

1fFcakuiated Is > F k-I, n-k (Table) Infer Hi.

Page 7: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

One of the basfc assumptions of the linear regression mode/ is tllat the

disturbances (or error) terms are random and follows a normal dlstnbution with

mean 0 and vaflance 4 Thrs implies that the covariance of t/>ese terms are

Zero For a model wtth f lo rmai l~ distrfb~ited disturbances, th,s imp/ies [hat ail

disturbances are independent I n time senes dala, this means serla/

Independence for the drsturbance terms, while for the cross-sechon data tt means

that we are assumrng the disturbance value that 1s drawn for any one unit 1s

uninfluenced by the values drawn for other units However, there are situat~ons rn which the assumptions of serial tndependence In disturbance term may not

exists There could be situations where we make an incorrect assumption of t i ~ e

form of the relatton between the variables For example, suppose we specrfy a

Irner relabon between Y and X, when in fact the true relation ts, say a polynomiai

of order three Even though the dtsturbance term in the actual relation may be

free of auto correlation, the semi-drsturbance term associated wrth the linear

relation contains terms ~n X2 and X3 If there is any serial correlatron In the X

values, that IS, correlation between observations that are adjacent in time order

or any other order o f data coiiection, then we wtll have serial correlation ~n the

composite disturbance term in general, we inciude only certain important

var~abies m the specrfied reiat~on so that the disturbance term represents the

mfluence of omrtted varfabies 1 Durbln and G, Watson developed the test o f

general kind concern~ng auto correlation In 1951 Thls test uses what Is usualiy

referred to as the Durbin-Watson d-stal~strc, and Is based on the sum of the

squared differences in successive value of the estimated disturbance terms

If we have positwe auto correlation, the successive values of the disturbance

terms W I I I tend t o close to another 1.e , posltive value of the et would most likely be followed by another posltlve value et+' Th~s suggests that terms in the

numerator OF d-stabstlc will be relatively small. We would, therefore, expea

POSltlve auto correlation to result In small value for d. Convenely, negative auto

Correlat~on tends to generate large difference between S U C C ~ S S I V ~ value of et,

The slgnal for this type of auto correlation is an unusually large value of d We Compare the calculated value of cj wlth the theoretical value of d; with (n-k)

209

Page 8: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

degrees of treedom (K being the number of explanatov variables rncludlng

term) The theoretical value of d is the value wh~ch do would assume

""11 hypothesis Were true I e , if there is no auto correlation

There are two 5tatlstlcs dSsoclated with the DW, dl and du, whlcll are lower and

upper values used to make Inferences The number of observabons, n, the number of Independent variables determines the DW table statistic, (which equals

k-1) The calculated value of DW varies between 0 and 4 with value of 2

denoting no auto correlation and values Of 0 and 4 denoting perfect posttve and

negative auto CO-relatlons, respectively The conclusions associated with

different values of the DW static, the following Inferences resulted when the

b DW is between 0 and dl, or greater than (&dl), reject the null hypothesis and

~nfers that there Is positlve auto to-relat~on

p OW IS between dl and du, or between 4-du and 4 - dl the test is inconclusive

OW is between do and (4- du), accept the null hypothesis and mfer that there

is no auto co-relatlon

b DW is between 4 -dl and 4, there is negative auto co-relation

The above test is widely used in econometric appllcatlons A great advantage of

thls test being that It is based on the estimated residuals Because of thls

advantage, it IS a common practice to report the Ourbin-Watson d.statlst1c along

with other summary statistic/s such as-R', adjusted R2, F-test etc

6.2. Analysis of Data

Many a Changes took place in the monetary instruments during the past 10 years

of econarnlc reforms With an 0bjectiVe of analping the relationship between the

changes in monetary instruments and their Impact on various economic

parameters, collected Quarterly data of monetary aggregates and m0neiaW

instruments as furnished in Table 6 1 The sample period used Is from the first

quarter of 1991 through the first quarter of 2001, so that 41 observations are

available for the estimation The multiple regression model was developed and

With the help of the data and the results were obtained

The estimated regression model with the help of the Computer (pH slat

Programming) The package also selected the best sets of the dependent and

independent variables where found to be In good flt, The best sets selected adjusted R', basing on the results were again tested uslng other tests Such as R 1

F-test and DW Statistic Regression 1s also done wkh lndlvldual variables* As

none of the variables was recommenQed for conslderatlon in the star

Page 9: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

sets, the F-test and DW statistic have not incorporated in the analysls However,

the study found that some of the Independent var~ables have slgnlficant

relatlonsh~p with dependent variable and others have less s~gn~Rcant reiatlonslllp

explalnlng the dependent var~ables

The significance of the relationship between dependent and independent variable

Is tested uslng R2 Turning to the statistical slgnlficance of the estimated co-

efflc~ents, the slgn~ficance of co-efficient Of Independent variable 1s tested with F-

test The auto corelatlon between the Independent variable and Dependent

1s tested uslng Durbln Watson (DW) Test

T ~ I S study made an attempt a t finding out the case and effect relationship among

the monetary aggregates, economic growth and banklng variable Using multiple regression, aggregate deposits, gross credlt, gross domestic product, whole sale

prlce index and broad money and monetary instruments such as deposits Interest

rates, Cash Reserve Ratlo, Statutory Llqu~d~ty Ratlo and ~nterest rates on

advances as dependent variables as well as independent variables by trylng out

varlous permutations and combinations The bas~c oblective was to find out the

core variables among the monetary Instruments and monetary aggregates for

bringing about the desired level of growth In aggregate deposits, bank credit etc , The results o f the study have been tested, as already stated, with R2, adjusted

R2, F-test and DW S t a t ~ s t ~ c The results lnd~cated prevalence of auto correlation

whenever only two vanables were taken for finding out cause effect relationship

between them However, for a comblnatlon of variables, the study found that

auto correlation IS not prevalent in many of the cases

211

Page 10: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

Source Statistical Data on Indian Economy by RBI, Various issues

6.2 1 Aggregate Deposits

As Per the results, the dependent variable Aggregate Deposits is not being

Influenced by any s~ngle independent variable A comblnatlon of two or more

variables only was influencing the Aggregate Deposits, Obtalned the f ~ l l o ~ l n g

results of the regression analysis with Aggregate Deposits as Dependent variable

Page 11: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

P Test and DW Statistic

Any one of the individual Independent variables taken for the study is not

influencing the Aggregate Deposits. The aggregate deposits are Influenced by all

the four independent variable as a group vlz, Broad Money, Whole sale Prlce

Index, GDP at Factor Cast and Deposit Rates, Uslng the RZ, which a t O 99889 IS

very very nearer to 1, tests the slgnlflcance. It shows that the enllre explanatory

variable accounted for 99.89 per cent of the varlatlons In ths Aggregate deposits 21 3

Page 12: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

Turning to the statlstlcal s~gnlflcance of the estimated co-eff~c~ent, the s ~ y n ~ f i c a n c ~

of CO-effic~ent of independent variable ts tested with F Test The calculated F

(8790 13) IS far higher than the table value or F at 2 61 and hence the Ho null

hypothests ts rejected In favour of H1 This means that the regression co-eff~clent

1% s~gn~ f~can t The results are tested for any auto correlat~on between the

dependent and Independent variables uslng Durbin Watson (DW) Test Since the calculated DW at 1 928 a between du (1 29) and 4 -du (2 28) and hence

accepted the null hypothesis and that there IS no auto correlation between these

var~ables Finally after testing the results, It Is concluded that all the four

Independent var~ables as 8 group are slgnlflcantly Influencing the dependent

var~able

of all the four parameters, Broad Money and GDP at factor both put together Is

hav~ng scgntficant influence on the Aggregate Deposits, because the R' and

adjusted R2 are the are very close to one for two parameters put together than

any other cornb~natton and the standard error Is the lowest among all other

comb~nat~ons S~nce the calculated value of F (17976 7) Is far higher than the

table value of F (3 231, the null hypothesis Ho Is re~ected in favor of H l and

hence there IS stgn~ficant relatlon The auto co-relation Is tesled using Llie DW

Stat~stlc The calculated value of the DW Statlst~c (1 877) Is between the du and

4- du, accepted the null hypothesis and found that there Is no auto correlat~on

Of the other comb~natlons, Broad Money, GDP at factor and Deposlt Rates put

together 1s having slgnlflcant Influence on the Aggregate Deposits, because the R~

and adjusted R2 are the are very dose to one for these three parameters put

together than any other cornblnat~on and the standard error is lower Since the

calculated value o f F (11934 26) Is far hlgher than the table value of F (2 84), the

null hypothes~s Ho Is rejected In favor of H I and hence there Is slgniflcant

relatlon The auto correlation Is tested uslng the DW Statlstlc. The calculated

value of the DW Statlstlc (1 87668) Is between the du and 4- du, accepted the

null hypothesis and found that there Is no auto correlation,

Individually, the R2 and adlusted R2are maximum, Standard Error Is m l n l m ~ m for

broad money out of the four lndlvldual parameters followed by GDP and WPI

The Statist~cs also found that the changes that took place duflng the past 10

years, In the Interest rates do not have any slgnlflcant relatlonshlp on the

Aggregate Deposits movements as the R2 at 0 09985 and adjusted 17~ at 0,07616

and standard error a t 215392. The lnslgnlficant. relatlonshlp Is also proved with

calculated F value (4.2154) bolng only very marginally hlgher than the table

value (4.08) The result of the DW Statistic (0.360485) Is between 0 and dl 214

Page 13: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

(1 44) provlng that there is a posltlve auto correlation between deposit rates anc

Aggregate Deposlh Finally concluded that the growth in Aggregate deposrts I!

interest rate ineiastlc and it has very Iim~tedlno influence in bririging aboul

changes in the deposit levels The changes in the Aggregate Deposlts also liave

an Impact on the movement of Deposlt Rates

The empirical resuits obtalned with the use of statistical methods can also be

collaborated with the help of theoret~cal understanding of the study and also

applying the commonsense economic arguments Deposits In a society are

determined by the state of economy In general and not by any parlicular

parameter All the Individual parameters may Influence very marginally I f Broad

Money Increases, it 1s natural that it 1s likely to effect deposlts, because the

economy 1s monetised and hence any changes in the flow of money to the society

will naturally increases the tendency to deposit more Similarly where there are

changes in lnflatlon, it 1s not possible that ail the money dlvetted to the deposlts

However, depend~ng on the level o f prices the quantum of Aggregate Deposits

may be affected L~kewise, i t is not necessary that the changes In deposlt rates

totally reflected i n the changes In aggregate deposits, Just because of Increase in

deposlt rate, people cannot be Induced to deposit thelr entlre money in deposlts

and the same way, because of fall in deposit rates, all bank depos~ts will not be

w~thdrawn Depos~tors' behav~or would depend upon other Interest rates and

market returns on various competrng instruments as well Similarly, the Impact of

growth In the GDP on depends will again depend upon the savings habit of the

sector which contributes most for the growth in GDP Hence, ail these

parameters are Interdependent and not a single monetary Instrument completely

affects changes in aggregate deposits

6.2.2. Gross Credit

As per the results, the dependent varlable Gross Credit Is also not being

influenced by any single independent variable. A comblnatton of two or more variables tends to ~nfluence the movement in credit Obtalned the following

results of the regression analysis wcth credit as Dependent varlable

Page 14: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being
Page 15: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

Any one of the rndlvldual Independent vartables taken for the study is not

lnfluenclng the Gross Credit All the seven Independent varlables as a group vIz,

Broad Money, Whole Sale Price Index, GDP at Factor Cost, Deposlt Rates, CRR,

SLR and lending rates Influence the Gross Credlt Uslng the Rz, whtch 1s (0 gg ~ l ) ,

tests the slgnlf~cance and adjusted R2 (0 9952) Is slgn~ficantly greater rrom Zero

and nearer to 1 It shows that all the explanatory vartable accounted for 99 61

per cent of the varlatlons In Gross Credit Turning to the statlstlcal significance of

the estlrnated co-efficient, the significance of co-efficient of lndependent variable

IS tested wlth F Test The calculated F (1170 3599) 1s far hlgher than the table

value of F (2 34) and hence the i io null hypothesls Is rejected In favour of HI

~ h l s means that the regression co-efflclent IS significant The results are tested

for any auto correlatlon between the dependent and lndependent varlables using

Durbln Watson (DW) Test Slnce the calculated DW (1 1721) IS below dl (1 23)

reject null hypothesls and conclude that there IS auto correlatlo~l among these

variables When tested the other cornblnat~ons,/best sets, it Is statlsttcally proved

that there 1s auto correlatlon in most of these comblnatlons, as well i t means

that the reverse correlation IS possible among all the varlables Hence, we cannot

justlfy the tests, though they satisfy all other tests

It has stat~stlcally proved that the changes that took place durlng the past LO

years, In the Interest rates do not have any slgnlficant relatlonshlp on the Gross

Credlt movements The lnslgnlflcant relatlon IS also proved w i th adjusted

R2 belng 0.49380, whlch IS nearer t o zero and far below 1. Hence,

concluded tha t t h e growth in gross c red~ t is interest lnelastlc and It has

very I ~m l t ed /no Influence in brlnglng about changes i n the credlt levels.

The changes In the gross credlt also have an Impact on Llle movement of lnterest

Rates

Any one of the monetaw policy Instruments Is not able to Influence lndlvldually

the gross credlt All the four monetary pollcy Instruments effected during the

past ten years are influencing gross credlt but only as a comblnatlon The

slgnlflcance 1s tested by uslng the R2, which Is (0 9364751 slgnlflcantly greater

than Zero and nearer to One. The significance of co-efficient tested uslns F test

reveals that the calculated F value (128,9931) Is far hlgher than the Table vallie

and hence there 1s a slgnlflcant relatlonshlp between the dependent and independent varlaMes But the DW test reveals that there Is a posltlve auto

correlation between the dependent and Independent variables as the calculated

Value (0 520789) Is below the Table value dl (1 29)

Page 16: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

The empirical results obtarned from the use of stat~stical methods can also bo

collaborated with the help of theoretlcai understandlng of the study and also

apply~ng the common sense economic arguments Credlt deployment IS

determined by many a parameters of the economy, and not by any single

parameter It is not necessary that the amount released to the public by way of

reductran i n CRR and SLR wrll not go to credit alone Practically, the reduction in

lending rates will not induce people to ava~l addltlonal credit from banks

Requirements of credit largely depend upon many other factors ralher than the

lendlng rates alone Llkewlse increase In wholesale prrce Index will not Increase

the credit i n the same ratlo Sometimes it may lead to fall In the credit also

People may not come forward to avail fresh loans just because the prices have

gone up Of all the four Monetary Policy instruments selected for study CRR is

Influencing the most on the gross credlt followed by SLR All the variables are

hav~ng the auto correlation deplctlng that there Is converse relationship among

the variables

- XiX2X3X4

X2 X3 X4

6 2 3 Regression on GDP

TO amplify the existence of auto-corelation, regresslon equation was f~tted for

GDP as depedent variable and four Independent variable vlz., M3, WPI, Credlt and Deposlts as shown In the following Table The study found that all the four

Independent variables chosen are influencing GDP (dependent variable) - aggregate deposits are Influenclng GDP wholesale price Index, gross credit and

broad money

~ - -

- ... D W stat ist7 - -- .. DU ~-;~I?.uF ~ e s u t

1 72 2 71 __1_28 0520789

0 936475972 0 8809285Ol 0 76624831 -. - -~ 0 5067882871

- 0 08688002 F - Test and Durb~n Watson Statistic

_ P 929216083 0 877795041 0 760096949T6399 0 493809031 0 062850547

-. -- - - 30635 29113 - 402530813-

15568 81924 i&3f-' 111470 2937

T?llsl?~_de!. . _ Ye_s

N o . NO

. _ ------ .NO -- . . NO

Page 17: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

I Regress~on Results for GDP with Broad Money, WPI, Gross Cred~t and Dcpos~ts

The significance is proved by adjusted R2 which Is at 0 9941 and the calculated F

value (1653 75) 15 far higher than the table value 2 61 and the posilive auto

correlation between GDP and its independent variables Is establlslied by DW

Statrst~cs (0 478315) whlch IS below the table value DL (1 3) When the

stat~stical results are collaborated with Lheoret~cal understanding study and also

applylng the common sense economic arguments, it is true that GDP cannot be

lnfluenced by any in single independent variable and there will be an aulo-

correlat~on between the GDP and independent variables GDP Is be~ng dcc~ded by

all the other monetary aggregates and the all-monetary aggregates are be~ng

equally lnfluenced by GDP

F Test and D W Statistic

XlX2X3 XlX2X3X4

- Regression Resu lb for GDP wlth CRR SLR Deposlt & Lending rates

F - Test Table1 Result 2 841 2208 94 2 611 1653 75

Durbin Watson Test d l / du 14-dl 1 4-du I Result 1 31 1661 2 661 2 341 0 363148 1 31 1721 2 711 2 281 0 478315

X3 = 'SLR X4 :: Lendlng Rates Y = 'GDP X I = 'Deposit Rates

XZ = 'CRR

Page 18: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

F Test and D W S t a t ~ s t ~ c 1

of all the monetary policy Instruments, only deposlt rates with an adjusted R' of

0 90having a slgnlflcant Influence on GDP R2 of all the other three is far below

one Out o f the combination of 811 the four monetary policy Instruments, CRR and

SLR are able to lnfluence more the GDP with an adjusted R2 of 0 9666 and the

calculated F value (565 547) is far h~gher than the table value The posltve auto

correlatlon among GDP, CRR and SLR 1s proved calculated DW stat~stc (0 9725)

ly~ng below the table value dl (1 39)

6 2 4 Regresslon o n In f la t ion (WPI)

Just as In case o f common sense knowledge the wholesale prlce Index I S belng

Influenced by all factors in the economy Stat~st~cally also it 1s proved that a

combinabon of Independent variables only influence the wholesale prlce Index and

not any slngle parameter The R2 of all the best subsets is very close to one and

calculated F value IS far higher than the table value The DW statistics for all tlie

cornb~nat~ons IS far below the Table value dl conflrmlng that there Is a poslt~ve

auto correlatlon

Y =WPI X i =M3 X2 = GDP X3 = CUR X4 = SLR

Page 19: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

Broad money is being Influenced by a comblnatlon of CUR, SLR and deposlt rates

rather than the comblnatlon of all the four (le , all the above three plus lendlng

rates) The adjusted R2 of all these three put together Is 0 9499 and the

calculated F value (247 806) 1s far higher than the table value 2 84 There IS an

auto correlation between the broad money and the comblnatlon of CRR, SLR

depos~t rates with calculated DW (0 7845) Is lying below the table value dl

(1 11) A comblnatlon of all the four dependent varlables registers p0~ilIVe auto

Co-relation Among the dependent varlables, only CRR Is havlng an adjusted R2

of 0 8913 which IS nearer to one and for all other var~ables it IS far below one,

lnd~catlng that no single Instrument is havlng slgn~ficant relation with broad

money movement

Page 20: EmolrlcalStudv - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/39078/14... · The above test is widely used in econometric appllcatlons A great advantage of thls test being

Durlng the past ten years of economlc reforms, many changes took place In the

Implementatlon of monetary policy Instruments All the monetaly policy changes

put together only Influenced the monetary aggregates af the economy None of

the monetary pol~cy instrument ind~vidualiy affected any one of tlie monetary

aqgregates It is statistically proved with empirical results that a comblnatlon of

the monetary policy ~nstruments are only influencing the economy