I 3 .

2 Port 1-statistlcat Methods

INTRODUCTION
Historically, the th·st. index was constructed in 1764 to compare the I_talian pri~e index in 1750
th
wi~h the price level in 1500. Though originally developed for ~easurmg e _eff_ect of c~ange in
prices. index numbers have today become one of the most widely u~ed statistical devices and
th
there is hardly any field where they are not used. Newspapers_ headlme . e fact th at_prices are
goill.g up or down, that industrial production is r ising or falling, that imports ~re mcreasing
or decreasing, that crimes are rising in a particu lar period compared to th e previous period as
disclosed by index numbe1·s. They are used to feel th e pulse of ~h e economy a nd they hav_e come
to be used as indicators of inflationary or deflationary t endencies. In fact, th~Y are descnbed as
'barometers of economic activity', i.e., if one wants t o get an idea as to w~at is h~ppening to an
economy. he should look to important indices like the index number of rndustnal production,
agricultural production, business activity, etc.
Some prominent definitions of index numbers are given below:

• ''Index numbers are devices for measuring differences in the magnitude of a group of
related variables." - Croxton & Cowden
• "An index number is a statistical measure designed to show changes in a variable
or a group of related variables with respect to time, geographic location or other
11

characteristics. " · ' - Spiegel

• "In its simplest for m, an index number is the ratio of two index numbers expressed as
a per cent. An index number is a statistical measure--a measure designed t o sh ow
changes in one variable or in a group of related variables over time, or with r espect
to geographic location, or in t erms of some other characteristics." - Patterson
• "In its simplest form, an index number is nothing more than a relative number, or
'relative' which expresses the relationship between two figures, wher e one of the
figures is used as a base." - Morris Flamburg
• "Generally speaking, index numbers measure the size or magnitude of some object
at a particular point in time as a percentage of some base or reference object in the
past." - Berenson & L evine
'--
It is clear from the above definitions that an index number is a specialized aver age designed
to measure the change in a group of related variable over a period of time. Thus, wh en we say
that the index number of wholesale prices is 112 for January 2020-21 compare d to J anuary
2019-20, it means there is a n et increase in the prices of wholesale commodities t o t h e extent
of 12 per cent during the year.
Index numbers, in effect, r elate a variable or variables in a given period to t h e same variable
or variables in another period, called the base period. An index, the simplifie d n ame for index
numbers, which is computed from a single variable, is called a univariate ind ex whereas an
index which is constructed from a group of variables is considered a composite index.
For a proper understandin g of the t erm 'index number', the following p oints are worth
considering:
(i) Index numbers are specia!ized .averages: As explaine d in the chapter on measures of
central value, an avera~e 1s a sm gle figure r epresenting a group of fi gures. However, to
obtain an average, ~he item s must ?e
com~arable, for example, t h e average weight of
men, women and children of a certam locality has no me aning at a ll. Furthermore, the
unit of measurement mu st b e the same. for a ll the ite m s · Thus , a n average of th e we1·ght
· lb I
expressed m kg, , etc., ias no m eanmg. Howev er , this is not so with index numbers.
Index numbers are used for purposes of comparison i·n si'tuat·ions w h ere t wo or more
 Index Numbers 13.~-

eries are expressed in different units or the series are composed of dif~er en~ types of
\ems. For example, while constructing a consumer price index, the _var~ous 1~ems are
~ivided into broad heads, namely (i) Food, (ii) Clothing, (iii) Fuel and I:1ghting, (w) House
Rent, and (v) Miscellaneous. These items are expressed in different units: t~u~, under the
h ead 'food), wheat and rice may be quoted per quintal, ghee per kg, etc. Similarly, cloth
may be measured in terms of metres. An average of all these items expressed in different
units is obtained by using the technique of index numbers.
(ii) Index numbers measure the net change in a group of related varia bles: Since index numbers
are essentially averages, they describe in one singl e figure the increase or decrease in a
gr ou p of related variables under study . The grou p of v aria bles may be the prices of a
specified set of commodities, the v olume of product ion in differen t sectors, etc. Thus,
if the consumer price index of working class for Delhi has gone up t o 11 3 in April 2020
compared to May 2019, it m e ans that there is a net increase of 13 per cent in the prices of
commodities includ e d in t h e index. Similarly, if the index of industrial production is 118
in 2020 comp ared to 2019, it means that there is a net increase in industrial production to
the exten t of 18 p er cent. It should be carefully noted that even where an index is showing
a n et incr e ase, it may include some items which have actually decreased in value
others which have remained constant.
(iii) Index numbers measure the effect of changes over a period of time: Index numbers are
most widely use d for measuring changes over a period of time. Thus, we can find out the
net change in agricultural prices from the beginning of First Plan period to the end oftlae
Twelfth Plan period, i.e.,, from 1951 to 2017. Similarly, we can compare the agricultural
production, industrial production, imports, exports, wages, etc., at two different t:iJDPS.
However, it should be noted that index numbers not only measure changes over a
of time but also compare economic conditions of different locations, different in
different cities or different countries. But, since the basic problems are essen ·
same and since most of the important index numbers published by the Gove
private research organisations refer to data collected at different times, we
in this chapter index numbers measuring change relative to time only. Howe
described can be applied to other areas also.

USES OF INDEX NUMBERS

Index numbers are indispensable tools of economic and business analysis. Their
can be best appreciated by the following points:
(i) They h elp in framing suitable policies: Many of the economic and busine
guided by index numbers. For example, while deciding the increase in dear
of the employees, the employers have to depend primarily upon the cost
If wages and s alaries are not adjusted in accordance with the cost of Ii ·
leads to strikes and lock-outs which in turn cause considerable waste
index number s provide some guide post s that one can use in making d
Though index number s are mos t widely used in the evaluation of bu
conditions, there is a large number of other fields also where index nuQl
example, sociologist s may sp e ak of population indices; psychologists
quotients which are essen t ially index numbers comparing a per&
with that of an average for his or her age; health authorities prep
changes in the adequacy of hospital facilities and educational
have devised formulae to measure changes in the effectiveness of
(ii) They reveal trends and tendencies: Since index numbers are
measuring changes over a period of time, the time series so form
~ Part I Statistic-al Motl1ods
nin g ind ex n~1mbo
r s1u rly. For exu mp l e, hy e~Hmi
gt'~lf'l'Hl trt't Hi nft ht'
in
plw
for
nn1
the
1w1
1ns1
1m
8
1 tt ndP
10 Y<' [lrs, W<' r'Llt l sny / h~t. ou r ,mp~,
~r:
rts sho w1~g an
index
of 11~1po1·t~ for lnd yeu r nff r•r ,YC'n r. S 1m1lnrly, by exa mi nin g th<•
tlw y i·isi ng t ~ev:' yea rs, we ca n
up " nrd tNH it'n ry, i.,,.,, Hl'l '
t.h
'tio1 1, hus i,wRs 11 d.ivil y, of<',, for c laa
num ht'r ~ of in< lus1 rin l 1n·o otl< mg th e trend of
h't' nd of prn d rn•1 i nn urn l lnts in ess 11d ,ivit,y. By ex~imi n
con dud (' nho ut 1~w to how much
riy, w e c •nn cfro w ve ry imp or•t n~f , con clu si~ms Ml
tlw plw nnn wn on und t'l' ~lu , irre gu lar_ f?rces, etc.
pln t't' d,w t n 11w l'ffrC '1 of' tH•n son n l if y, f',Yc:lwu l for ce~
chang<' i::- t nik ing ss con cl ,tw ns.
rnm lwr ~ art ' hig hly usd 'u l in sl udy ing f hr• gonerfl l hu s ine
.. Thus. indp:-.: 1 re usef~J l not only
ant in {<> reca stin g fut ure economic activity: I nclcx num ber Frn th
(m ) 'Phcy 01"(" imp ort ~re ~mpr,r tan t
t a nd pn'R t'll1 wo l'ki ngR of our eco nomy, bu t, a lso ey
in :--tud~·ing the pas 1 series
nom ic nct .tvi ly. In dex num ber s the n ar? o~ten use d m ~ me
in fo1\~asting fuh m.' ('co erm tre nd, sea son al var ~atwn
s and _bus ine ss .cycle
l stu d~ of lon g-t
ana lys is. tht ' his tor ica Jc and business
so t hat bus ine ss lea der ma y keep pace wit h cha ngm g econom
dt:,, ·('Jopnl<.'nt. aki ng pur pos es.
luw e bet ter inf ormatio n ava ilab le for decisio n-m
eon dit ion s and ful in deflatjng,
)' use ful in def iati ng: * Ind ex nu mb ers are hig hly ~se
(fr) Ind ex nu mb c "'s a r c l'el for ~ost
ust the ori gin al dat a for pri ce cha nge s, or to adJ ust wa ges
:i.e.,. the y ar e use d to adj . Moreo ver , nom mal
thu s tra nsf orm nom ina l wa ges int o rea l wa ges
ofl ivi ng cha nge s and sal es thr oug h
nsf orm ed int o rea l inc om e and nom ina l sal es int o r eal
inc om e can be tra
app rop ria te ind ex nu mb ers .
ERS
CLASSIFICATIO N OF INDEX NUMB ics and bus ine ss,
cla ssif ied in ter ms of wh at the y me asu re. In econ om
Index n um ber s ma y be ce ind ex is
pri ce; (2) qua nti ty; (3) val ue; and (4) spe cia l pu rpo se. A pri
the cla ssi fic atio ns are (1) the r . A qu ant ity ind ex
It com par es lev els of prices, from one per iod to ano
m ost fre que ntl y use d. Th e val ue ind ex
the qua nti ty or num ber of var iab le cha nge s ove r tim e. h
me as~ es how mu ch
e val ue of a var iab le. In fac t, the val ue index com bin es bot
r upe
me asu res the cha nges in the ati ve ind ex.
ty cha nge s and pre sen ts a mo re inf orm
the pri ce and qua nti l be me ntioned,
ity ind ex num ber s are dis cus sed in det ail . Th e oth ers wil
On ly pri ce an d qu ant nu mb ers
con str uct the m sin ce bot h val ue and spe cia l pu rpo se ind ex
but ~ri .tho ut detail , of how to Since the me tho d of con str uct
ion of var iou s typ es of
pro ble ms in con stru ctio n. is clear,
do no t offer new
sto od if the tec hni que of con str uct ing pri ce ind ex nu mb er
ind ex nu mb er can be un der
the m.
we sha ll dev ote ma jor att ention to
N OF INDEX NUMBERS
PROBLEMS IN THE CONSTRUCTIO st be giv en to the follow ing pro
ble ms :
mb ers , a car efu l tho ugh t mu
Be for e con str ucting ind ex nu

Th e Purpose of the Index y dec ide d-w ha t

the pu rpo se of con str uct ing the ind ex mu st be ver y cle arl
At the ver y out set , ind ex is of limit ed and
re and wh y? Th ere is no all- pur pos e ind ex. Ev ery
the ind ex is to me asu ces mu st not
a pri ce ~nde x tha t is int end ed to me asu re con sum er s' pri
~a rtic ula r use. Th us.' r
h an ind ex is int end ed to me asu re the cos t of liv ing of poo
if suc
mc l~~ e wh ole sal e pri ces . An d inc lud e goods ord ina rily u sed by mi
ddl e cla ss and
u.l d be tak e~ not to
fam1li~s, gre at care sho ex wo uld lea d to con fusion
gr? ups . ~a du re t~ dec ide cle arl y the pur pos e of the ind
upp er- mc om e num ber
;rm tful : esu lts. Oth er pro ble ms suc h as t h e bas e yea r, the
a;d wa st~~e of ti~e ~it f ~o th com mo dit ies, etc. , ar e dec ide
d in the lig ht of the
0 com m~ itiehs. toh e h m-,c du e . , ~ prices of theed.
1s bem g con str uct
p urp ose ior w 1c t e m ex
f h .
Kafka: "Index numbers are toda O O
~ t Wld_ely used statistical devices ... They are used
• In the words of Simpson and and they have come 0t b Y ~e t moS
of deflationary tendencies.•
to feel the pulse of the economy e use as Indicators of inflationary
 Index Numbers 13.5
The problem of the scope of the index, i.e.,, the field covered by the index, is bound
up with
the purpose of the index and the data availab le. The data availab le, ur rather the
lack of them,
n1aY necessi tate the modific ation of the purpose .
selection of a Base Period
Whenev er index numbe rs are constru cted, a referen ce is made t o some base
period. The
base period of an index numbe r (also called the referen ce period) is t h e period agains
t which
comparisons are made. It may be a year, a month or a day. The index for base period
is always
taken as 100. Thou gh the selectio n of the base period would primar ily depend upon
the object
of the index, the followi ng points need careful conside ration of base period:
(i) The base period should be a normal one: The period that is selecte d
as base should
be nor mal, i.e., it sh ould be free from abnorm alities like wars, earthqu akes, famine
s,
booms, depres sions, etc. Howev er , at times, it is really difficul t to select a year which
is normal in all r espect s-a year which is normal in one respect may be abnorm
al in
anothe r. To solve this problem , an average of a numbe r of years, 3 or 4 (prefer
ably
coverin g one comple te cycle), may be taken as the base. The process of averag ing
will
reduce the effect of extrem es. Thus, the average of the period from 2017 to 2020 I I : , .
!:
conside red normal wherea s no individ ual year in that span can be conside red I I l l_l -

(ii) The base period should not be too distant in the past: It is desirab le to have ; I --:cl..,,_',.
__
based on a fairly recent period, since compar isons with a familia r set of c i r c
~
are more helpful than compar isons with vaguely :-emem bered conditi ons. For e~~
for decidin g increas e in dearne ss allowan ce at presen t, there is no advant age ·
·
1995 or 2005 as the base; the compar ison should be with the preced ing year or
after which dearne ss allowan ce has not been revised .
(iii) Fixed base or chain base: While selectin g the base, a decisio n has to be
whethe r the base will remain fixed or not, i.e., whethe r we have a fixed
base index. In the fixed base method , the year or the period of ye
other prices are related is consta nt for all times. On the other hand,·
method , the prices of a year are linked with those of the preceding
the fixed year. Natura lly, the chain base method gives a better p
obtaine d by the fixed base method . Howev er, much would depend u
constru cting the index.
Selection of Numbe r of Items
The items include d in an index should be determ ined by the purpose for
constru cted. Every item cannot be include d while constru cting an index nu
has to select a sample . For exampl e, while constru cting a price index, it is
each and every commo dity. Hence, it is necessa ry to decide what com
commo dities should be selecte d in such a manne r that they are rep
habits and custom s of the people for whom the index is meant. Thus, ·
for workin g class, items like scooter s, motor cars, re~_igerator~, cosm
decision must also be made on the numbe r of commo dities to be mclud
we should note that the larger the numbe r of commo dities includ
shall be the index but at the same time the greater shall be the
Purpose of the index will help in decidin g the numbe r of commoditie
index, a larger numbe r of commo dities will have to _be inc~uded a~ co
index as the index numbe r of the prices of foodgra ms or mdustrtal
It is also necess ary to decide the grade or quali_ty of the item inde-.x.
Index numbe rs will give wrong re_sult if at_ one time one ~t and nt
far ns
anothe r time, anothe r set. To avmd confusi on ~bout qualitie
ntified
Possible, only standa rdized or graded items are mcluded IO
after a time lapse.
....__ 13.6 Part I- Statistical Methods

Price Quotations price quotatio ns fo r

. l . . bl Om is to obtain
Af~ pro d't' f
1,eI t 1e commod 1tles 11'n e been selected , the next
th ese commodities. H is •n well known fact that prices of ma~y ~ammo /
1
1
v:ry
\
b;o~
am
~! pl~ce
~nee
to place and even from shop to shop in the same market. H_ JS rp~ac t _ca e ot
ma e of
quotatio ns from all the places where a commodity is deaJI, rn . se ;.c ~on mus 11 :
represe ntative places and persons. These places should be
th0 se ~ JC • are w~ ~- nown for
trading f01· that particular commodity. After the places from where _t ~-r~~e quoh 0 a Ions are to
be obtaine d is d~cided, the next thing is to appoint some person or i~s ~ ~Own \h ~at~
sup~ly
. Great care must be exerc~se se~ a e_ pnce
price quotatio ns as and when required
inaccura cy of price _quotati on~ sup~hed by
reportin g agency is unbiased. In order to check the reliable Journal
· • h gency If there 1s some
an agency, quotatio ns are obtamed from more t an one a . . ·
or magazin e supplyi ng price quotations, then it may be utilised.
st be
In order to ensure uniformity the manner in which prices _are to be _quoted I?u ~lso
decided. There are two method s of quoting prices: (i) money prices, and (1,1,) quantit y price_s. In
4 qumtal
the former case, prices are quoted per unit of commodity, for example , sugar~ ,000 per
~ugar may be quoted
(100 kg) and in the latter case, prices are quoted per unit of mone~. Thus,
ly adopted
as 1/40 kg for one rupee. The former method is free from confusion and is general
while quoting prices.
.
A decision must also be made as to whether the wholesale prices or retail prices are required
Thus, in a consum er price index
The choice would depend upon the purpose of the index.
ities are
the wholesale price will not be represen tative at all. If the prices of certain commod
account
controlled by the governm ent then it is these controlled prices that should be taken into
and not the black market prices which may be much higher.

Choice of an Average
lar
Since index number s are specialized averages, a decision has to be made as to which particu
be
average (i.e., arithme tic mean, median, mode, geometric mean or harmon ic mean) should
in the
used for constru cting the index. Median, mode and harmon ic mean are almost never used
arithme tic mean and
constru ction of index number s. Basically, a choice has to be made between
ction
geometric mean. Theoretically speaking, geometric mean is the best average in the constru
s we
of index number s because of the following reasons: (i) in the constru ction of index number
ible to major variatio ns
are concerned with ratios of change; (ii) geometric mean is less suscept
number s
as a result of violent fluctuations in the values of the individu al items; and (iii) index
possible.
calculat ed by using this average are reversible and therefore, base shifting is easily
The geometric mean index always satisfies the time reversa l test.
more
Despite theoret~cal justifica~ion_ for favouring geometric mean, arithme tic mean is
tic mean
~opular ly used _while constru ctmg mdex number s. This is for the reason that arithme
Howeve r, wherev er possible in the
~s much more simple to compute tha~ the geometric mean.
note that
In:tehresht of gre~ter accurfac y, tgeozi:ietric mean should be preferre d. It is gratifyi ng to ·
· ]
wit t e growmg use o e1ec romc computers, geometric mean is beco mmg more popu ar m
constru cting index number s.

Selection of Appropriate Weights*

. .
The problem of selectin g suitable weights is quite importa nt and th e s_ame tu1:e qm~e
difficult to decide. The term 'weight' refers to the relative import a~
.
the constru ction of the index. All items are not of equa1 importa ance O the differen t items m
nce and h .t .
. ence, ~ is n~cessa ry
to devise some suitable method whereby the varying import ance Of t h e differen t items 1s taken
,
• •weighting is the term used to describe conscious effort to assign to ea h
ate to its relative importance." c commoct,ty an influence that, in the final result, is
proportion
- Richardson
 Index Numbers 13.7
. to account. This is done by alloc ating weig hts.
Thus , we have broa dly two type s of indic es-
100weighted indic es and weig hted indic es. In the form er case, no speci
~bereas in the latte r case specific weig hts are assig fic weig hts are ass~ gned
ned to vario us item s. It may be p01n ted
out here that_ no in~e x _is unw eigh ted in th~ _strict sen~
e of the term as w~ig hts impl icitly
enter in unw eigh ted indic es beca use we are giV1Ilg equa
l rmpo rtanc e to all the item s and henc e
weights are unity . It is, there fore, nece ssary to adop t
some suita ble meth od of weig hing so that
arbitrary and haph azar d weig hts may not affec t the resu
lts.
There are two meth ods of assig ning weig hts: (i) impl
icit, and (ii) expli cit.
In implicit weig hing , a comm odity or its varie ty is inclu
ded in the inde x a num ber of time s.
Thus , if whea t is to be given in an inde x twice as muc h
weig ht as rice, then two varie ties of whe at
against one of rice may be inclu ded in the serie s. On the
othe r hand , in case of expli cit weig hing ,
some outw ard evide nce of impo rtanc e of the vario us item
s in the inde x is given. Whe n the expli cit
weights are assig ned, the ques tions are: (i) By wha t do
we weig h? and (it) Wha t type of weig ht
do we use?
(i) In orde r to brin g out the economic impo rtanc
e of the comm oditi es invo lved, the weig ht
can be prod uctio n figur es, cons ump tion figur es or distr
ibuti on figur es.
(ii) Weig hts are of two type f: quan tity weig hts
and valu e weig hts. A quan tity weig ht,
symb olise d by q, mean s the amo unt of comm odity prod
uced , distr ibute d, o.. cons ume d
in some time perio d. A valu e weig ht, on the othe r hand
, comb ines price with quan ·
'produced', 'dist ribut ed' or 'cons umed '. The valu e is in
term s of rupe es and is symb
by p x q, wher e p stan ds for the price and q for the quan
tity.
Now, the ques tion is whe ther to choose quan tity weig
hts or valu e weig hts. If the ag
method is used whil e cons truct ing index, then quan
tities are used as weig hts
price time s quan tity will alwa ys give the same units
, name ly, rupe es. On the ot
in aver agin g price relat ives, quan tity figures cann ot
be used . It is for the reas on
multiply perc enta ges by quan tities expr essed in diffe
rent units , we get resu lts
units for exam ple, perc enta ge time s tonn es will give
tonn es and perc enta ge
give kg Such figur es cann ot be used in comp utati on.
But, if perc enta ges are
value figures, whic h are alwa ys expr essed in rupe es,
we gP.t answ er in rupees
statis ticia n will use q as a weig ht in the meth od of
aggr egati ng actu al p ·
P x q as a weig ht in the meth od of aver agin g price
relat ives.*
Anot her prob lem in conn ectio n with weig hts is that
of decid ing whetli
be fixed or fluct uatin g. Sinc e the relat ive impo rtanc
e of the different item s
the same for all time s, it is logic al to vary the weig ht
from time to time. S
give bette r resu lts. How ever, whe n fluct uatin g weig
hts are used , one
in inter preti ng the inde x beca use not only chan ges
in price s but also c
affecting the inda~ .
Selection of an App ropr iate Formula
A large num ber of form ulae have been devi sed for cons
truct ing the ·
often is that of selec ting the most appr opria te form ula.
The choic e oft
not only on the purp ose of the inde x but also on the
data available
sugg ested that an appr opria te inde x is that whic h satis
fies time rev
test. Theo retic ally. Fish er's meth od is cons idere d as
"ideal" for co
However. from a prac tical poin t of view , there are certa
in limita ·
be discu ssed later . As such . no one parti cula r form
ula can be
circu msta nces. On the basis of this knowledge of the
char a.ct.erisl
discr imin ating inve stiga tor will choose those meth ods
which are
• Sometimes. in the absence of actual weights. arbitra
ry magnitu~es may have to be
to use these weights and, thefl'>~re. they should be
used only m the audest tinn
r--_1_3__:•-=-8 Part I- Statistical Methods

is usually the
None of the abon~ problems iR Rim ple Lo so lve in practice and the final index
ble with the
produc t of compromise between theoretical standa rds and the standa rds attaina
given data.

METHODS OF CONSTRUCTING INDEX NUMBERS

numbe rs. Broadly
A large numbe r of formulae have been devised for constructing index
speaki ng, they can be grouped under two heads:
(a) Unwei ghted index numbe rs or indices; and
(b) Weighted Index numbe rs or indices .
the we~g~ted
In the unweig hted indices, weight s are not expres sly assign ed wherea s in
be furthe r divided
indices, weights are assign ed to the various items. Each of these types may
under two heads:
(z) Simple Aggregative Metho d, and
(ii) Simple Average of Price Relatives Method .
The following chart illustra tes the various methods:
Index Numbers

Weighted
Unweighted

Simple Average of Weighted Weighted Average

Simple Aggregative
Relatives Aggregative of Relatives

Unweighted * Index Numbers

Simple Aggregative** Method
d is used to
This is the simple st method of constructing index numbe rs. When this metho
dities in questio n
constr uct a price index, the total of curren t year prices for the various commo
Symbolically:
is divided by the total of base year prices and the quotient is multip lied by 100.

Pol= -
Ul
xlOO
I:Po
r,p 1 = total of curren t year prices for various commodities
r,pO = total of base year prices for various commodities
The steps
This method of constructing the index is the simple st of all the methods.
require d in compu tation are:
(i) Add the curren t year_ prices for various commodities, i.e., obtain
r,p .
1
(ii) Add the base year prices for the same commo dities, i.e., obtain r,p .
0
(iii) Divide 1P 1 by I.P0 and multiply the quotient by 100.
2019 as base:
Illustrat ion 1: From the following data, construct an Index for 2020 taking

Commodity Price In 2019 m Price In 2020 (")

A 50 70
B 40 60
C 80 90
D 110 120
Solution : E 20 20
that all the values considered in calculating the i d .
• Unweighted means
•• Aggregate means that we add or sum all the values. n ex are of equal importance.
 . - -. - --- -~ -- -- --- - - - -- --
Index Numbers .,,..9
solution:
CONS TRUC TION OF PRICE INDE X

Commodity Price In 2019 Price in 2020

Po P1
A 50 70
B 40 60
C 80 90
D 110 120
E 20 20
'f.P0 =300 'f.P1 =360
360 x 100 = 120
P01 = l:P1 x 100 =
l:Po 300
s included in the
there is a net increase in the prices of commoditie
This means that as compared to 2019 , in 2020,
index to the extent of 20% .
simp le aggr egat ive inde x:
Lim itati ons: The re are two main limi tatio ns of the
s can exer t a big influ ence on the valu e
• The unit s used in the price or quan tity quot ation
of the inde x.
rtan ce of the com mod ities .
• No cons idera tion is give n to the relat ive impo
vari ous term s and thei r price s ue
This inde x is base d on the assu mpti on that the
is diffe rent for diffe rent item s, the
in the sam e unit . If the unit of mea sure men t
let us cons ider the prob lem of cal
produce vast ly dive rgen t resu lts. To illus trate ,
with that of 2019 , and let us incl
index com parin g the 2020 cost of cons truct ion
items, the cost of labo ur and the price of cem ent.
2019
ers t 25
Average hour ly wage to be paid to cons truct ion work
t 230
Price of cem ent per bag
'f.P 1 = 250 and 'f.P O = 230
250
p = xl00 =108 .69% .
01 230
week ly wage . Ta ·
Now, supp ose inste ad of hour ly wage , we take the
we can rewr ite-t he figur es as:
~ 1250
Ave rage week ly wag e
Pric e of cem ent per bag t 230

Now 'f.P1 = 1400 and 'f.P0 = 1250

l 4 00 xl00 = 112%.
P01 = 1250
ly wage , inst
One can see at once that whe n we used woek
nro thus poaaiblo
increased from 108. 69 to 112. Man ipula tions
nd rath er
~uoting per kg rath er than per poun d or per mnu
man ner can
Index whose valu e can be man ipul ated in this
to all the it
. measure. Moreover, equa l impo rtanc e is give n
r gain ea
reasons th3.t the simp le aggr egat ive inde x has neve
13.1 0 Part I- Statisti c a l Metho ds
·nde.x numbe r by taking :
th e 1
Illustra tion 2: For the data given below, calculate
(i) 2012 as the base year;
(ii) 201 9 as the base year; and
(iii) 2012 to 2014 as the base period . Price of Comm odity
--~
of Comm odity
Year
'X'
Year Price
'X' 10
2017
4 9
201 2 2018
5 10
2013 2019
6 11
2014 2020
7
2015
8
2016

Solutio n:
(/) INDEX NUMBE RS TAKIN G 2012 AS THE BASE YEAR
Price of Index Numbe rs
Price of Index Numbe rs Year Comm odity (2012 = 100)
Year Comm odity (2012 = 100) 'X'
'X'
10
10 -x100 =250
100 2017 4
2012 4
9
9 -x 100 =225
~x100 = 125 2018 4
2013 5 4

10 _!Q X 100 = 250

6 ~x100 =150 2019 4
2014 4

2020 11 .!..!x10 0 = 275

2015 7 ~x100 =175 4
4

2016 8 ~x100 =200

4

(//) INDEX NUMBE RS TAKING 2019 AS THE BASE YEAR

Price of Price of
Index Numbe rs Index Numbe rs
Year Comm odity Year Comm odity
(2019 = 100) (2019 = 100)
'X' 'X'

4 10
2012 4 - x 100=4 0 2017 10 -x100 =100
10 10

5 9
2013 5 - x 100=50 2018 9 -x100 =90
10 10

6 10
2014 6 - x 100-60
- 2019 10 -x100 = 100
10 10

7 11
2015 7 x 100 = 70 2020
10 11 -x100 =110
10
8
2016 8 x 100 = 80
10
 (llij INDEX NUMBERS TAKING
Index Numbers ia,11
. 2012 TO 201 4
Z012 to 2014 1s to be taken as base it mean th AS THE BASE PERIOD
wi,en ' s at we hav8 t
0
Average = i_+ 5 +~ _ take an average of 2012, 2013 and 2014.
3 -5
Hence, 2013 will be taken as 100.

- Price of
- -
Year Commodity Index Numbers
Price of
')(' (2013 =100) Year
commodity Index Numbers
- 'X' (2013 = 100)
2012 4 4
5 x100=:8Q 2017
10 10
-x100=200
5
2013 5 5
5 00-100
-x1 - 2018 9
9 -x100=180
5
2014 6 6
-x 100 = 120 2019 10
5 10 -x100=200
5
2015 7 7
-x 100 = 140 2020 11
5 11 -x100=220
5
2016 8 8
-x100=160
5

Simple Average of Price Relatives Method

When this method is used to construct a price index, first of all-price relatives* are
the various items included in the index and then average of these relatives is obt
any one of the measures of central value, i.e.,, arithmetic mean, median, mode, g
or harmonic mean. When arithmetic mean is used for averaging the relatives,
computing the index is

L(¾xlOO)
Po1 = N

where N refers to the number of items (commodities) whose price

averaged.
Although any measure of central value can be 1:1sed to_ obtain the
relatives are generally averaged either by th~ arith~ebc or the
geometric mean is used for averaging the price relatives, the form:
index becomes

nog[P1
Po
x100] ~ og
1p where
log P01 = N or N '

or p 01 = antilog
1(~
l log Pi
~
x100)] _antil
Other measures of central value are not in common use fol
---;-,;,:-. . . ommodity in a given periGd to
P~ce relative is the ratio of the pnce of a smg 1e c d't rice during thl lalH
P8nod or reference period . If Po and Pn denote the comma I Y P
by ~efinition: price relative= Pnl Po . 100
adding generally expressed as percentage by multiplying by ·
 b th aver age of relat ·
13 .] 2 Part I Statistic al Methods 2019 as base y e ives
- for 2020 takin g .
I dex f aver aged relatives. .
, construct an n
Illus trati on 3: From the follo wing data etric mean or . . (?)
(b) geom
meth od usin g (a) arith meti c mea n, and - pr,c e m 2020

Com mod ity Price In 2019 f') 70

50
A 60
40
B 90
80
C 120
110
D 20
20
E

ES
Solu tion : HME TIC MEA N OF PRIC E REL ATIV
(A) INDE X NUM BER S USIN G ARIT Pric e rela tives
Price in 2020 (?)
Com mod ity Pric e in 2019 (?)
P1
~ X 100
Po Po
140.0
70
A 50
150. 0
60
B 40
112. 5
90
C 80
109. 1
120
D 110
100. 0
20 20
E
.I: f'.t x100 611 .6
Po

I: Pi x100
611 ·6 =122. 32
Po N 5
Po1 =
RIC MEA N OF PRIC E REL ATIV ES
(B) INDE X NUM BER USING GEOMET
Price in 2020 Pric e Rela tives log P
Com mod ity Price in 2019
p
Po P1
70 140. 0 2 .1461
A 50
40 60 150. 0 2.17 61
B
80 90 112.5 2.05 12
C
D 110 120 109.1 2.03 78

E 20 20 100. 0 2.00 00

~ log P = 10.4 112

P 01 = Antilog[I:'; P]= Antilog[10·!112]= Antilog 2.0822 =120 9

hm etic mea n ·
ric mea n hav e bot h bee n use d, the arit
Alt hou gh arit hm etic mea n and geo met e eco nom ists ,
pre ferr ed bec aus e it is eas ier to com put e and mu ch bet ter kno wn. Som
is ofte n rem e
ewo rth, h~v ~ pre ferr ed to use the med ian whi ch is not affe cted by ext
not abl y F. Y. Edg y sma ll num ber of
nt only whe n an ind ex is bas ed on a ver
val ues . Sin ce the arg um ent is imp orta d in actu al
es, it gen era lly doe s not carr y muc h wei ght and the me dia n is seld om use
com mo diti
pra ctic e.
 Index Numbers 13.13
. . prepare Index Numbers of price for three years wil• h average price as base from the data given below:
u tratton 4·
111 5
RATE PER RUPEE COMMODITIES
A B C
1st year 10 kg 4 kg 3 kg
2nd year 9 kg 1
3- kg 3 kg
2
3rd year 9 kg 3 kg 1
2- kg
2

.
SOiution: Convert first the prices into rupees for 40 kg then determ ine average price . .
mpute relatives. and with this average price as
l)aSe co [Base: Average Price (in rupees) =100]
Average
Unit price= 100 First Year Second Year Third Year
(Base)
Price p Price p Price p
A 40 kg 4.3 4.0 93 4.4 102 4.4 102
B 40 kg 11.6 10.0 86 11.4 98 13.3 115
C 40 kg 14.2 13.3 94 13.3 94 16.0 113
Total of Relatives 273 294 330
Average of Relatives 91 98 110

Note: P indicates Price Relatives

For 1st year, 10 kg of commodity A costs t 1.
Hence, 40 kg of commodity A will cost J_ x 40 = t 4
10
Similarly, other prices are obtained .
4·3 is the average of 4, 4.4 and 4.4, i.e., the respective prices of the three years.

Merits: This method has the followin 6 two advantages over the previous meth ·
• Extreme items do not influence the index. Equal importance is given to
• The index is not influenced by the units in which prices are quoted
level of individual prices. Relatives are pure numbers and are the
the original units. Consequently, index number computed by the rel
be the same regardless of the way in which prices are quoted. This sim
relatives is said to meet what is called the units test.*
Limitations: Despite these merits, this method is not very satisfactory b
• Difficulty is faced with regard to the selection of an appropriate a
arithmetic mean is considered as questionable sometimes because •
The use of geometric mean involves difficulties of computation. 0
never used while constructing index numbers.
• The relatives are assumed to have equal importance. This is
~eighing system that is highly objectionable since economica
important than others.

Weighted Index Numbers

The so-called unweighted index numbers discussed above are U
the ~rm. They assign equal importance to ~~l the items include
are in reality weighted, weights being imphc~t rather than expl
of unweighted indices, it is possible to get different results by
13.1.4 Port l ~ Stotistical McthlX.1~
• , • of' t.hc• cnsc•s. Construction of useful ind
1
(or tlw ,m,wighh'd in1h-:\) i~ fnr from n•llh A!tr. 111 mosh, . moclity n wPight in accordance _ex
· • f" t t Aign to c•ur com with
. . l'0\'s l.'l'Q\lll'l'~ a 1·1m~1'1onA 1' for o ns.
mnn .
ti t thr 111drx is supp osed to descrjbe.
•
it~ 1mport~m1'1' in tlll' totnl plwnomrnon u, ·
W1,4;hh'd. t1Hl('X numht'l'R nrr oft.wo types:
• Weighted AggrcgntIYc Indices, and
• Weighted Awrage of Relatives
JJeighted Aggregative Indices · t with the f un dament a1difference that weights
·
These indices are of the simple aggregative y~e . thods of assigning weights d
· d h · · · luded m various me an
are ass1gne to t e various items me . · dex numbers have been devised 0f
consequently. a large number of formulae for constructing m
which some of the more important ones are:
• Laspeyres method
• Paasche method
• Dorbish and Bowley's method
• Fisher's ideal method
• 1farshall-Edgeworth method, and
• Kelly's method.
All these methods are named after the persons who have suggested them. . .
(i) Laspeyres Method:* The Laspeyres Price I~d.ex i_s a weighted a~gregate pnce mdex,
where the weights are determined by quantities m the base period. The formula for
constructing the index is:
p = 'f.pl_qO X 100
01 'f.poqo
Steps:
• Multiply the current year prices of various commodities with base year weights and
obtain r.p 1q0.
• Multiply the base year prices of various commodities with base year weights and obtain

LPoQo ·
• Divide r.p q by r.p q and multiply the quotient by 100. This gives us the price index.
1 0 0 0
Laspeyres Index attempts to answer the question: ''What is the change in aggregate value
of the base period list of goods when valued at given period prices?" This index is very
widely used in practical work.
The primary disadvantage of the Laspeyres method is that it does not take into
consideration the consumption pattern. The Laspeyres index has an upward bias. When
prices increase, there is a tendency to reduce the consumption of higher priced items.
Hence, by using base year weights, too much weight will be given to those items which
have increased in price the most. Similarly, when prices decline, consumers shift their
purchases to those items which decline the most. By using base period weights, too little
weight is given to those items which decrease most in price, again overstating the index.
(ii) !'aas~he's Met~od:** The Paas~he price index is a weighted aggregate price index
m which the weights are determmed by quantities in the given year. The formula for
constructing the index is:
P, : l:PJ ql X 100
01 1:poqi

• This method was devised by Laspeyres In 1871 for detennlnlng price Increases used for cal .1 • n
•• This is atterGerma~ s~tlst1c1an Hermann Paaschewhofirstused It in 1874. He Is knownfor.:latlng the rate of ~flali~~1desa
calculation of the pnce index. Paasche studied economics ' agriculture ' statistics and PhllosophYPaasche Index
at University Halle, PGennany·
ofwhich
 Index Numbers
--------1

..,,: .
• Multi ply curre nt year
.
pnces
.
of vario us comm oditie s with curre nt year wejgh ts and obtai n
tp1q, .
curre nt year weigh ts and ob-
• Multi ply the base year price s of vario us comm oditie s with
tain 'f.poQ1
• Divide lp1q1 by lpoq1 and multi ply the quoti ent by 100.
of the given perio d
In gener al, this formu la answ ers the quest ion: "Wha t would be the value
list of goods when value d at base- perio d prices ?"
that revise d weigh ts, or
The diffic ulty in comp uting the Paasc he index in practi ce is
g to the data collec tion expen se
quantities, mus~ be comp ~ted each yea~ or each period , addin
m the prepa ration of the index . For this reaso n, the Paasc
he index is not used frequ ently in
practice where the numb er of comm oditie s is large.
index meas ures chang e in
Comp ariso n of Lasp eyres and Paas che Meth ods: Laspe yres
a "fixed mark et baske t" of goods and servic es. The same
quant ities are used in each period .
curre nt consu mptio n.
The Paasc he index conti nually updat es the quant ities to the levels of comp uted from the
value s
These two appro aches tend to produ ce oppos ite extrem es in index
often prefe rred to Paasc he's for
same data. From a pract ical point of view, Laspe yres index is
the base year quant ities
the simpl e reaso n that in Laspe yres index , weigh ts (q 0) are
use of Paasc he index
not chang e from one perio d to the next. On the other hand, the
dered and in most '-Gi-l'f l
the contin uous use of new quan tity weigh ts for each period consi
index numb ers are co
weights are diffic ult and expen sive to obtain . In most count ries,
by using Laspe yres formu la.
that the forme r ·
An intere sting prope rty of Laspe yres and Paasc he indice s is
latter tends to un
expec ted to overe stima te or to leave an upwa rd bias where as the
is usual ly a red
i.e., shows a down ward bias. When the prices increa se, there
most prono unced .
consu mptio n of those items for which the increa se has been the
t to the prices that
using base year quant ities, we will be giving too much weigh
large . When
the most and the nume rator of the Laspe yres index will be too
have declin ed
consu mers often shift their prefe rence to those items which
by using base perio d weigh ts in the nume rator of the Laspe yres
index,
and the mun
sufficient weigh t to the prices that have gone down the most
when th
too large. Simil arly, becau se peopl e tend to spend less on goods
ces an index which ten
the use of the Paasc he or curre nt weigh ing produ
argum en
the rise in prices , i.e., it has a down ward bias. But, the above
Laspe yres index must neces sarily be large r than Paasc he's.
year all!
Unles s drasti c chang es have taken place betwe en the base
difference betwe en Laspe yres and Paasc he's will gener ally be
s~all
we1g
as a satisf actory meas ure. In practi ce, howe ver, ~he ~~se year
ty. The Paa
remai ns the most popul ar for reaso ns of its practi cabili
const ructed when up-to -date data for the weigh ts are availa ble.
of a given year can be comp ared only with the base yea_r. For exam
and P 2 0 = 145 · Then , p 18 and p 19 are using differ ent weigh ts and
fi
other. If these indice s had been obtain ed by the Laspe yres or
becau se in that case, the weigh ts are the same base year wei
Practice, the Paasc he formu la is usual ly not used and the Laa
PDpular for reaso ns of its practi cabili ty.
(iii) Dorbi sh and Bowle y 's Metho d: Dorbi sh and Bowle y
mean of the two indice s (Lasp eyres and Paasc he) JD!
13. 16 I - Statistical Methods
Part _
ula
acc oun t the infl uen ce of both the peri ods,
i.e., cur ren t as wel l as bas e peri ods . The form
. the inde x is: P :.... L+P
fi con stru ctin
or g O1 2
L = Las pey res Ind ex and P - Paa sche index
whe re

Or
ulae for
her 's 'Ide al' met hod : Prof . Irvi ng Fish er* has give n a num ber of form
(iv) Fis Fish er's
e he call s one as the 'ide al' inde x. The
con stru ctin g inde x num ber and of thes
Idea l Ind ex is give n by the form ula:
Po i= J:.Pi_qo x J:.Pi_qi x lOO or P 01 = JLx P
I:.poqo I:.poq1
n of the
t Fish er's Idea l Ind ex is the geo met ric mea
It will be clea r from the abo ve form ula tha rica l form ulae
the Fish er's met hod , we ave rage geo met
Las pey res and Paa sch e indi ces. Thu s, in
tha t err in opp osit e dire ctio ns.
ause of the foll owi ng reas ons :
The abo ve form ula is kno wn as 'Ide al' bec be the best
n whi ch is theo reti call y con side red to
(i) It is bas ed on the geo met ric mea
.
ave rage for con stru ctin g inde x num bers
ren t yea r as wel l as bas e yea r pric es and qua ntit ies.
(ii) It take s into acco unt both cur sug ges ted by
test as well as the fact or reve rsal test as
(iii) It sati sfie s both the tim e reve rsal
Fish er.* t emb ody the
(iv) It is free from bias . The two
form ulae (La spey res and Paa sch e's) tha
the idea l form ula, cros sed geo met rica lly, i.e., by
opp osin g type and wei ght bias es are, in cell atio n
no bias . The resu lt is the com plet e can
an ave ragi ng proc ess tha t of itse lf has
reve rsal and fact or reve rsal test s.
of bias es of the kind s reve aled by time The
pute bec ause it is exc essi vely labo riou s.
It is not, how eve r, a prac tica l inde x to com le. In
t of the inde x, are not read ily ava ilab
data , par ticu larl y for the Paa sch e seg men exac t,
rely upo n sim ple, alth oug h per hap s less
prac tice , stat istic ians will con tinu e to
inde x num ber form ulae . as
rsh all- Edg ewo rth Met hod . In this met hod also , both the cur ren t yea r as well
(u) Ma ex is:
red. The form ula for con stru ctin g the Ind
bas e yea r pric es and qua ntit ies are con side
Poi = I:. Pi (qo +q1) xlOO
I:. Po(qu +q1)

on ope ning the brac kets

Po i= J:.Pi_qo+I:.Pi_q1 xlOO
L Poqo + L Poq]
resu lts
, givi ng a very clos e app rox ima tion to the
It is a sim ple, read ily con stru cted mea sure
obt aine d by the idea l form ula. ula for con-
Kel ley' s klet hod : Tru man Lee Kell ey** has sug gest ed the foll owi ng form
(vi)
stru ctin g inde x num bers :
- J:.Pi_q
P 01 - - - x lOO
'J:.poq
the base
may refe r to som e peri od, not nec essa rily
Her e, wei ghts are the qua ntit ies whi ch as wei ghts .
cur ren t rea r. Thu s, the ave rage qua ntit y of t-:1"_0 or mor e yea rs may be use d
yea r or ghts , the
qua ntit ies of two yea rs is use d as wei
If in the Kel ley s form ula the ave rage of the
form ula bec ome s: .
. l
ican economist statistician inventor and
• Irving Fishe r (186 7-19 47) was an Amer in the introdu~tion of statlstlcal method ~o,cia campa,~ner.studi.es.
ntial olog1cal
.. Kelley (188 4-19 61) was highly influe in o psych
 '' ,.,.., u

Index Numbers 13.17

P 01 = ; Pi_q x 100; where q = qo + ql

-'-'Poq 2
. il rly the average of the quantities of three or more years can be used as weights. This
~:dais know~ as fixed weight ~ggrega~ive index and is currently _in great fa:'our in ~he
me t uction of index, number series. An important advantage of this formula is that hke
co:s \res index it does not ~e~and yearly cha_nges in the ~eights. ~oreover'. t~e base ~eriod can
La h 1 ged without n.ecessitatin g correspondi
ng change m the weights. This is very important
be c ate the construction of appropriate quantity weights for a general purpose index usually
tteca~ses a considerable amount of work. Weights can thus be kept constant until new census
. h . d
or other survey) data become avai1 a-ble to revise t e m ex. ·
~uir

mustration 5: Construct index numbers of price from the following data by applying:

1. Laspeyres' method 2. Paasche's method

3. Bowley's method 4. Fisher's Ideal method, and
s. Marshall-Edgeworth method.
2019 2020
Commodity
Price Quantity Price Quantity

A 20 8 40 6
B 50 10 60 5
C 40 15 50 15
D 20 20 20 25

Solution:
CALCULATION OF VARIOUS tNmCES

2019 2020
Qty.' Price Qty. P1Qo PoQo P1q1
Commodity Price
qo P1 q1
Po
40 6 320 160
A 20 8
60 5 600 500
B 50 10
50 15 750 600
C 40 15
20 25 400 400
D 20 20
LP1 qo LPoQo
= 2070 = 1660

p = r.p1qo x 1oo; where r,P1qo =2070 and l:A,Qo = 1

1. Laspeyre's Method: 01 'l:.poqo

p = 2070 x 1oo = 124.70.

01 1660

p = r,p1q1 x 100; where r,P1q1 = 1790, l'-Po"1

2. Paasche's Method: 01 r,poq1

~ x 100 = 121 .77

Po1 = 1470
r, P1qo
r, P1q1 2070 + 1780
r, Poqo + r, Poq1 x 100 = 1660 21470
3. Bow/ey's Method: Po1 = 2
7
1.25 + 1.22 ,<1 00 = 2 .4 x100
= 2 2
 13.18 Port I-Statis1ical Methods
11 - - - - - - - - - -
L+P 124.7+ 121 .77 - ~ = 123.24
Po1 = - 2-= 2 - 2
or
----==
2070 1790
Po1 = l: P,Qo x ;PoP1qq1 x100 = 1660 x 1470 x100
4. Fisher's Ideal Method: l: Poqo L, 1

= ✓1.525 X 100 = 1.235 X 100 = 123.5

l:r>,(qo+q1)x100 l:r>,qo+'1:.P,q1 x100

5. Marshall-Edgeworth Method: Po1 = r, Po (qo + q1) r. Poqo + '1:. Poq1

= 2070+1790 x 1oo = 3860 x 100 = 123.32.

1660+1470 3130
. data, cons tru ct a price index number of the group of four commod ities by using the
Illustration 6: From the followmg
appropriate fonnula:
-----r-- --------
Base Year Current Year ----
-i---::: -:-=~=
Commodity Expenditure Price per unit Expenditure
Price per unit
(?) (~)

A 2 40 5 75
B 4 16 8 40
C 10 2 24
D 5 25 10 60
[M.Com. Kumaun Univ., PEI, B.Com. (H), Jamia Mi/lia Univ., 2012)

Solution: Since we are given the base year and current year prices and expenditure Fisher's, Ideal Formula will be
most appropriate.
CONSTRUCTION OF PRICE INDEX

Commodity Po qo P1 q1 P1 qo Poqo P1 q1 Po q1
A 2 20 5 15 100 40 75 30
B 4 4 8 5 32 16 40 20
C 1 10 2 12 20 10 24 12
D 5 5 10 6 50 25 60 30
rp1 qo rpoqo !'.p1 q1 LPo q1
-202 91 199 =92

Po1 = /"E.Piqo x ¼q1 x100

v
f./Joqo f./Joq1

= /202 199
V91x92 x100= 2.1912x100= 219.12
ruustration 7: Calculate the weighted price index fro th ~ .
m e ,ollowmg data:
Materials Units Q
required uantity Price(?)
required
Cement 2019 2020
100 lb. 5001b.
Timber 5.0 8.0
c.ft.
2,000 C.ft
Steel sheets 9.5 14.2
cwt.
50 cwt.
Bricks per '000 34.0 42.0
20,000
12 0
Solution: Since the weights here are fixed weights n Ith · 24.0
e er relating to curre
nt Year nor relating to base year. we app,y
 Index Numbers

, ...8 thod for computing index.

l(elleY 5 "'
CALCULATION OF WEIGHTED PRICE INDEX

Materials required Units Quantity Price~

2019 2020
-
q Po P1 Po q P1 q
Cement 100 lb. 500 Ib 5.0 8.0 25 40

Timber c.ft. 2,000 c.ft. 9.5 14.2 19,000 28,400

Steel sheets cwt. 50 cwt. 34.0 42.0 1,700 2,100

Bricks per '000 20,000 12.0 24.0 240 480

r, Poq r, P1 q
=20,965 =31 ,020
- 'f.p1q
= 31,020 and 'f.Poq = 20,965
...,.,oq 0; 'f.P1q
Po1 - ~x'l'0

_ 31,020
Po 1 - , x 100 = 147·96
20 965

Weigl,ted Avera ge ofRelat ives

were not co
In the weighted aggre gative m~thod discussed above price relatives
te weighted av;
However, like unwei ~hted relativ e method, it is also possible to compu
etic mean or the
relatives. For purpo ses of averag ing we may use either the arithm
relatives ind
mean. The steps in the compu tation of the weighted arithmetic mean of
are as follows:
(i) Express each item of the period for which the index numb er
is being c
percentage of the same item in the base period.
(ii) Multiply the percen tages as obtained in step (i) for each item
by the w ·
been assigned to that item.
d out
(iii) Add the result s obtain ed from the several multiplications carrie
ts
(iv) Divide the sum obtain ed in step (iii) by the sum of the weigh
index number. Symbolically,
P01 = LPV where P = Price relative
LV '
V = Value weights, i.e., PoQo
Instead of using arithm etic mean, the geometric mean may be used
fl
The weighted geometric mean of relatives is computed in the same m
geometric mean of relativ es index numb er except that weights are
them to the logarithms of the relatives. When this method is used,
the index is:

p 01 = Antilog
. {1:V.1: log
V
P}
where p = Pi xlOO
Po
V = Value Weights, i.e., Poqo fore
__ !~~ 2 0 Port l ~ statistlcal Methods

Steps:
• Obtain percentage relatives for each item. .
• Find the logarithm of each percentage relative found in step (i).
• Multiply the logarithms by the weights assigned.
• Add the results obtained. in step (iii).
• Divide the total obtained in step (iv) by the sum of the weights.
• Find the antilogarithm of the quotient obtained in step (v).
Thia is the weighted geometric mean of relatives index number.
The following example will illustrate the steps.
Illustration 8: From the foUowing data, compute price index by applying weighted average of price relative method
using:
(a) arithmetic mean, and
(b) geometric mean.
P0 (t} · qo P 1 (f)
Commodity
30 20 kg 40
Sugar
Flour 15 40 kg 16

Milk 10 10 It. 15
Solution:
-
(a) INDEX NUMBER USING WEIGHTED ARITHMETIC MEAN OF PRICE RELATIVES

Poqo Pi X 100
Po
Commodity qo V p PV
Po P1
40
Sugar t 30 20 kg f 40 600 x100 80,000
30

16
Flour t15 4J kg n6 600 x100 64,000
15

10 It. ns 15
Milk f 10 100 -x100 15,000
10
r.v= 1,300 r. PV 1,59,000

p = r.PV = 1,59,000 _
01 r.v 1•300 -122.31
This means that there has been a 22.31 per cent increase in prices Over th e base level.
(b) INDEX NUMBER USING GEOMETRIC MEAN OF PRICE RELATIVES

Commod1ty Po qo P1 V p logP VlogP

Sugar t 30 20 kg t40 600 133.3 2.1249 1274.94
Flour ns 40kg t 16 600 106.7 2.0282 1216.92
Milk t 10 10 It. f 15 100 150.0 2.1761 217.61
r. V= 1,300 r. Vlog P
- 2709.47
tVlo
Po1 AnUlog[ gP]:::Antil [2709.47] .
. . r.v og 1300 ::: Antilog2.084=121.3
The result obtained by applying the Laspeyres method w Id
arithmetic mean of price relatives method (as shown below~:u come out to be the same as obtained by weighted
 Index Numbers 13.21
PRICE INDEX BY LASPEYRE$
METHOD
commodity Po qo
P1 P1%
- sugar ~ 30 20 kg Pa%
~ 40
Flour ~15 800 600
40 kg { 16
~10 640 600
Milk 10 It.
- ~15 150 100
r.p1q0 = 1590 t p0 q0 = 1300
L/)1Q 1 590
Po1 - o 100
r.Poqo x == 1300 x 100 == 122.31

Th e answe1·
· · b
is the same as that obtained by weighted ar'th
th · h d 1 me t·ic mean of pr1ce
. .
hod Tlns 1s ecause e we1g te average . relative
of price relative meth d b h .
met ·
aggregative meth o d (gr~-n
· b L
Y aspeyres) as follows: o can e c anged to s1mple
I: -· xpoqo
Po = r.Piqo
L.JJoqo r.poqo
When the transf?rmed formula _(weighted average of price relative) is the same as that for the
weighted aggregativ~ of ~ctual prices, the question arises as to why do we treat it as a separate
method of constructrng index numbers. The transformation is based upon two conditions not
necessarily always present:
1. that the arithmetic mean is being used; and
2. that base year values are used as the weight_s .
If these two conditions are not present, then the original formula is different and
susceptible to transformation in this way. Moreover, sometimes, index numbers are con
by combining existing index numbers. Since the data in this case index numbers are ·
the form of relatives, the untransformed formula involving rela-tives must be used,.
transformed formula is therefore ngt applicable.
,"
It may be advisable to use the weighted average price relative method for
reason inherent in the average of relatives method. This is because computing
price relatives requires us to compute the price relatives for each single com 11 I 11 '.
series of price relatives and these series can be stu~e~ and analysed in them
Dlustration 9: By using the average of the quantities of two y~ars as weights, compute a price i

Commodity Quantity Price in 2019 Price

2019 2020 (t)
A 10 16 20
B 9 7 25
C 2Q 24 40

Solution:
COMPUTATION o;;: PRtCE INDEX USING AVERAGE OF THE ~O-YEAR QUA
Quantity Quantity (qo ;q,) Price in Price In

Commodity
in 2019 in 2020
q1
q 2019
Po
2020
p
qo
A 10 16 13 20 25
B 9 7 8 25 28
C 20 24 22 40 40

r.p1q x 100,
Applying Kelly's method , we have Po1 = r.poq
I ; .i.

1, 429 x 100 = 106.64

Po1 = 1,340
 Index Numbers 13.29
Hence, time reversal test is satisfied by the given data.
Factor Reversal Test: Factor reversal test Is satisfied when :

Po1 >< 001 = :tp,q,

l:/Joqo
tq,Po l:q1P1
--X--= --X-
1914 2172
l:qoPo l:qop 1 1774 2015

p 01 x 0 01 = 2015 x 2172 x 1914x2172 = 2172

1774 1914 1774 2015 1774
r.p,q1 . 2172
¼qo ts also equal to . Hence, factor reversal test is satisfied by the given data.
1774

THE CHAIN INDEX NUMBERS

In the fixed base method discussed so far, the base remains the same throughout the series of
the index. This method, though convenient, has certain limitations. As time elapses conditions
which were once important become less significant and it becomes more difficult to compare
accurately present conditions with those of a remote period. New items may have to be
included and old ones may have to be deleted in order to make the index more representative.
In such cases, it may be desirable to use the chain base method. When this method is used, the
comparisons are not made with a fixed base; -r ather the base changes from year to year. For
example, for 2020, 2019 will be the base; for 2019, 2018 will be the base, and so on. If, however,
it is desired to associate these relatives to a common base, the results may be chained to obtain
chain indices. Thus in its simplest form, the chain index is one in which the figures for each
year (or sub-period thereof) are first expressed as percentages of the preceding year. These
percentages are then chained together by successive multiplication to form a chain index.

Steps in Constructing a Chain Index

In constructing a chain index, following steps are desirable:
(i) Express the figures for each year as percentages of the preceding year. The results so
obtained are called link relatives.
(ii) Chain together these percentages by successive multiplication to form a chain index.
Chain index of any year is the average link relative of that year multiplied by chain
index of previous year divided by 100.
Chain Index for current year
= Average link relative of current year x Chain index of previous year
100
The link relatives* obtained in step (i) facilitate comparison from one year to another, i.e.,
between closely situated periods in which the q's are a process of chaining binary comparisons
facilitate long-term comparisons. ~
Chain relatives differ from fixed-base relatives in computation. Chain relatives are computed
from link relatives whereas fixed base relatives are computed directly from the original data.
The results obtained by the two different methods should be the same, but they may differ from
each other slightly due to rounding off of decimal places. Since the process of computing chain
relatives is quite complicated and the results are same as the fixed base relatives obtained from
the original data, chain relatives should be used when the original data are not available but
the link relatives are.

Current year's Pi-ice

• Link Relative = . . x 100
Previous year's price
13.30 Part I stattsttcal Methods

Illustration 14: From the following data of the wholesale prices of wheat for ten years, conStruct index numbers
taking (a) 2011 as base, and (b) by chain base method:

Year Price of Wheat Year Price of Wheat

(f per 10 kg) (~ per 10 kg)

2011 50 2016 78

2012 60 2017 82

2013 62 2018 84
2014 65 2019 88

2015 70 2020 90

Solution:
(a) CONSTRUCTIO N OF INDEX NUMBERS TAKING 2011 AS BASE

Price of Index Number Price of Index Number

Year Year
Wheat (2011 = 100) Wheat (2011 = 180)
78
2011 50 100 2016 78 - x 100 =1 56
50

60 82
2012 60 x100=120 2017 82 x 100 =1 64
50 50

62 84
2013 62 x100 = 124 2018 84 x100 =1 68
50 50

65 88
2014 65 x100=130 2019 88 x 100 = 176
50 50

2015 70 ?O x100=140 2026 90 go x100 = 180

50 50

This means that from 2011 to 201'2, there is a 20 per cent increase; from 2011 to 2013, there is a 24 per cent increase;
from 2011 to 2014 there is a.30 per cent increase. lfwe..are interested in finding out increase from 2011 to 2842., from
2012 to 2013 and from 2013 to 2014, we shall have to compute the chainiodices.

(b) CONSTRUCTIO N OF CHAW INDICES

Year Price of Wheat Link-Relatives Chain Indices

(2911 = 100)
2011 50 100.00 tOO
2012 60 60 120 x100
x100 = 120.00 = 120
50 100

2013 62 62 103.33 x120

x100=103.33 = 124
60 100

2014 65 65 104.84x124
-x100=104
62 .84 =130
100
2015 70 70 107.69x130
x100=107.69 =140
65 100
2016 78 78 111.43 x 140
x100 = 111.43 =156
70 100
2017 82 82
x100=105.13 105.13x156
78 = 164
100
Contd....

J
 lnrJex Numbers 13.31

Year Chain Indices

Price of Wheat Link Rolatlves (2011 e 100)

2018 84 102.44x164 = 168

84 x100 102.44
82 100

2019 88 104.76x168 -= 176

88 x100=104.76
84 100

2020 90 102.27x176
90 x100 = 102.27 - 180
88 100

Note. The chain indices obtained in (b) above with 2011 = 100 are the same as the fixed base indices obtained in
(a) above. In fact chain index figure will always be equal to fixed base index figure if there is only one series.
Illustration 15: Compute the chain index number with 2016 prices as base from the following table giving the average
wholesale prices of the commodities A, Band C for the year 2016 to 2020: ·

AVERAGE WHOLESALE PRICE (IN ~)

Commodity 2016 2017 2018 2019 2020

A 20 16 28 35 21
B 25 30 24 36 45
C 20 25 30 24 30

Solution:
COMPUTATION OF CHAIN INDICES

Link Relatives based on preceding year

Commodity
2016 2017 2018 2019 2020
A 100 28 35 21
~x100=80 x 100 = 175 x100=125 -x100=60
20 16 28 35

B 100 30 24 36 45
x100 =1 20 x100=800 -x100=150 -x100=125
25 30 24 36

C 100 25 30 -24
-x100=125 -x1 00=120 - x 100=80
20 25 30

Total of Link 375 355

300 325
Relatives
Average of 125 118.33
100 108.31
Link Relatives
1· ..
Chain Index 108.33 x100 = _ 125x108.33 _ 118.33x135.41 _
100 108 33 135 41 160 23
(2016 = 100) 100 100 100

Conversion of Chain Index to Fixed Index

At times, it may be desired to convert the chain base index numbers into fixed base Index
numbers. In such a case, the following procedure is followed:
1. For the first year, the fixed base index will be taken the same as the chain base index.
However, if the index numbers are to be constructod by taking first year as the base in
that case the index for the first year is taken as 100.
2. For calculating the indices for other years, the following formula is used:
Current year's C.B.I. x Previous year's F.B.I.
Current year's F.B.l. = 100
F.B.l. = Fixed base index number; C.B.l. = Chain base index number
- l3.32 f'ort l~ stotlstical Methods

llluStratlon 16: f-rom the chain base index numbers given bolow, prepare fixed base 1nd8x numbers:
2019 2020
2016 2017 2018
120 105 95
80 110
(B.Com., Madras Univ.)
Solution:
COMPUTATION OF FIXED BASE INDEX NUMBERS

Chain ba;, Index number Fixed base Index n u ~

Year
80 80
2016
110 >< 80 88.00
2017 110 100
120x88 = 105.60
2018 120 100
105 x105.6 110.88
2019 105 100
95 x110.88 105.34
2020 95 100

Merits of the Chain Base Method

The merits of this method are enumerated here:
1. The chain base method has a great significance in practice be~ause i~ economic _and
business data we are more often concerned with making comparisons with the previous
period and no~ with any distant past. The link relatives obtained by chain hase method
serve this purpose.
2. Chain base method permits the introduction of new commodities and the detection of
old ones without necessitating either the recalculation of entire series or other drastic
changes. Because of this flexibility, chain index is used in many types of indices such as
the consumer price index and the wholesale price index.
3. Weights can be adjusted as frequently as possible. This flexibility is of great significance
in many types of index numbers.
4. Index numbers calculated by the chain base method are free to a greater extent from
seasonal variations than these obtained by the other method.

Limitations of the Chain Index

The main limitation of the chain index is that while the percentages of p-revious year figures
give accurate comparisons of year-to-year changes., the . long-range comparisons of chained
percentages are not strictly valid. However, when the index number user wishes to make year-
to-year comparisons, as is so often done by the businessman, the percentages of the preceding
year provide a flexible and useful tool.*
BASE SHIFTING, SPLICING AND DEFLATING THE INDEX NUMBERS
Base Shifting
For a variety of reasons, it frequently becomes necessary to change the reference base of an
index number series from one time period to another without returning to the original raw data
and recomputing the entire series. This change of reference base period is usually referred to
as "shifting the base". There are two important reasons for shifting the base:
l. The previous base has become too old and is almost useless for purposes of comparison.
By shifting the base, it is possible to state the series in terms of a more recent time period.
* Croxton and Cowden: Applied General Statistics.
 Splicing
Sometimes, an index number series is available for a period of time, and then undergoes
substantial revision including a shift in the reference period.
For example, the weights of an index number may become out of date and we may construct
another index with new weight. Thus, two indices would result. At times, it may be necessary to
conv:ert these two indices into a continuous series. The procedure employed for this conversion
is called splicing.
The process of splicing is very simple and is akin to that used in shifting the base. It is
expressed in the form of a formula as follows:
Index number of current year x Old Index of New Base Year
Spliced index number. =
JOO
The following example would illustrate the procedure:
Assume that a price index number series was revised by inclusion of certain new products,
exclusion of some old products and change in the definition of some other products. In the
following table is shown such an old series on a reference base of 2015 and a revised series on
a base of 2017.