Anda di halaman 1dari 20

NOTES

Extent of Poverty in India


A Different Dimension
K ausik Ga ngopa dh yay, Ka m a l Si ngh

The poverty line in India is usually asso iated "ith a alorie threshold& This alorie threshold approa h suffers from many pro#lems& +n alternative revealed preferen e%#ased approa h has #een provided #y (ensen and )iller& In the () approa h, the staple alorie share reveals "hether a household is alorie deprived& ,e use this approa h to estimate the extent of poverty in India& Though our poverty estimates are very lose to the Tendulkar -ommittee estimates for the ur#an se tor, for the rural se tor our estimates are onsidera#ly lo"er& ,e also find #y our method a remarka#le rise in ur#an poverty #et"een .//0%/1 and .//2%/3&

Indian Institute of Te hnolog y K haragpur&

onsumers! hoi es against in ome "ould ena#le us to dra" a line #et"een onstrained $ and un onstrained $ on% sumers& 'y the () methodology this line divides the poor from others& Estimation of poverty #y the () method% ology re*uires data on disposa#le in ome and a tual food onsumption for a : Introdu tion stimation of the poverty line has #een a mu h de#ated issue in India and expert groups have #een set up from time to time to revie" the methodo% logy for estimation& The on ept of the poverty line "as introdu ed #y a "ork% ing group of the Planning -ommission in :;<. and su#se*uently expanded in :;2; #y a task for e 5Sharma .//06& The most re ent report of the expert ommit% tees "hi h have revie"ed the methodol% ogy of poverty estimation is the one haired #y Suresh 4 Tendulkar, "hi h su#mitted its report in .//;& =et another revie" ommittee "as appointed in ./:. #y the Planning -ommission& +ll these approa hes are #ased on a uniform alo% rie norm for all individuals, "hi h make them sus epti#le to riti ism from dif% ferent *uarters for different reasons $ for example, Su#ramanian 5./::6 and S"aminathan 5./:/6& In spite of different pro#lems em#edded in having a uniform alorie norm for all families, there "as no other alternative methodology avail% a#le for al ulation of the poverty line& In this paper "e ela#orate a ne" methodology, proposed #y (ensen and )iller 5./:/6, to estimate the extent of poverty "ithout use of any uniform alorie threshold for all households& This method% ology proposed #y (ensen and )iller 5hen eforth the () methodology6 is #ased on the theory of hoi e, parti ularly on the assumption that given a "ider hoi e availa#le to a rational individual, she "ill hoose a #etter alternative& Therefore, the hoi e of a onsumer "ith more in ome is prefera#le to the hoi e of a poorer on% sumer #ut not vi e versa& ,e *uantify the desira#ility of a onsumer!s hoi e #y her onsumption of staple foods& + onstrained onsumer "ould onsume more of staple foods ompared to an un onstrained onsumer& Therefore, investigation into

The authors thank 4e#asis )ondal for dis ussions on this study& ,e have also #enefited from the omments of seminar parti ipants at the Indian Statisti al Institute 5Kolkata6 and an anonymous revie"er& ,e thank the Indian Institute of )anagement Ko7hikode for providing us finan ial assistan e 5Small Grant 8esear h Pro9e t6 to undertake this study& The usual dis laimers apply& Kausik Gangopadhyay 5kau sik@iimk.ac.in6 is a mem#er of the fa ulty at the Indian Institute of )anagement Ko7hikode& Kamal Singh is an undergraduate e onomi s student at the

household& @or estimation of poverty in India using this method, "e use the data from the National Sample Survey Offi e 5NSSO6 for .//0%/1 5<:st round6 and .//2%/3 5<0th round6& The details of food onsumption for a household an #e found in these data sets, #ut in omes of households are not availa#le in the NSSO data& ,e have therefore onsidered the expenditure of a household as a su#sti% tute for the in ome of that household& The head ount ratio is a "idely used measure of poverty& Our estimates of the head ount ratio for .//0%/1 are ontrasted "ith the Tendulkar -ommittee estimates& @or the ur#an se tor, #oth

estimates are similar $ the () estimate of .1&3A for all India is mat hed #y the Tendulkar -om% mittee figure of .1&2A& Bo"ever, there is onsidera#le divergen e #et"een the () estimates and the Tendulkar -ommittee estimates for the rural se tor& ,hile the Tendulkar -ommittee estimate for rural poverty is fixed at 0:&3A, its () ounter% part is #et"een ?:&/ and ?<&3A depend% ing on the level of aggregation $ India, or state%"ise& ,e also estimate the extent of poverty "ith .//2%/3 data and find a striking in rease in ur#an poverty to 0/&3A on the fa e of a some"hat un% hanging figure for the rural se tor&

The plan of this arti le is des ri#ed here& Se tion . dis usses the pro#lems "ith the uniform alorie threshold and ho" the () approa h takes are of these pro#lems& Se tion ? elu idates the meth% odology and des ri#es the data& Se tion 0 demonstrates the results of our analysis& Se tion 1 on ludes& . (ensen and )iller )ethodology .&: ,hy 4o ,e 8e*uire a Ne" +pproa hC +verage per apita alorie intake has #een extensively used to assess the extent of

poverty in India& The approa hes taken #y t"o expert groups, "hi h "ere set up OT E S #y the N Planning -ommission in :;2? 5task for e group6 and in :;;? 5Dakada"ala -ommittee6, onstitute t"o examples& Ender those approa hes, "e spe ify a threshold daily alorie intake per apita $ for example, .,0// and .,:// alories for rural and ur#an populations, respe % tively $ and then de ide a poverty line #asket $ a onsumption #undle $ that ensures the spe ified threshold level of alories for all mem#ers of a household& This #asket, in general, depends on the num#er of mem#ers in a household& The definition of the poverty line, in nominal terms, is the per apita expenditure "hi h ena#les a household to afford that spe ified poverty line #asket& Bouseholds "ith a lo"er in ome are ategorised as alorie%deprived and, hen e, under the poverty line& This approa h has several limitations& @irst, there is, in general, hardly any on% sensus on the su#sisten e alorie threshold itself 54asgupta :;;1F Sved#erg .///6& Se ond, there is also eviden e of hange in alorie norms over time, "hi h makes it even harder to press for a uniform alorie threshold& 4eaton and 4rG7e 5.//;6 demonstrated that per apita alorie onsumption has #een de lining for the last .1 years despite an in rease in the real "ages over the years ausing a do"n"ard shift in the alorie Engel urve& This apparent pu77le is only resolved if the alorie re*uirements hange over time& Third, a#sorption of alories from food items depends on var% ious hara teristi s of an individual, su h as the health status, meta#oli rate and fitness level 5Sved#erg .///H .06& )ost of these hara teristi s, #eing un% o#serva#le and diffi ult to measure, pose a severe limitation on the onver% sion of a alorie threshold into nominal terms& +s an example, a person "ith some stoma h disorder may have to onsume a larger amount of food items ompared to a healthy individual for o#% taining a definite level of alories& If "e use the same alorie threshold for a per% son suffering from a stoma h disorder and a healthy individual, "e either under% estimate the extent of poverty

among the persons disorder or

"ith

stoma h

overestimate the degree of poverty among the healthy individuals& In the prevailing studies, none of these hara teristi s have #een onsid% ered for estimating poverty in India& @ourth, a threshold alorie approa h does not onsider the non%nutritional attri#utes $ for example, taste $ of food items& In pra ti e, a household!s sele % tion of food items not only depends on the alorie ontent #ut is also ontingent on these non%nutritional attri#utes& If "e merely al ulate "hether the house% hold!s in ome is suffi ient to #uy the poverty line #asket, it may not #e suffi% ient& +s these attri#utes have #een om% pletely ignored in the approa hes #ased on the alorie threshold, estimation of poverty #ased on a alorie threshold ap% proa h may give rise to unrelia#le fig% ures& There has also #een a strong indi% ation that there is no tight link #et"een the in ome of a household and alories onsumed #y mem#ers of this household 54eaton and 4rG7e .//;6& The poverty estimates, therefore, ould #e ar#itrary if "e onsider the poverty line #asket and per apita alorie onsumption& .&. (ensen and )iller +pproa h The various short omings of the prevailing approa hes make the ase for a s"it h to a methodology "hi h does not spe ify any parti ular alorie threshold level for estimating poverty& (ensen and )iller 5./:/6 have proposed a novel approa h #ased on revealed preferen e of a house% hold in hoosing its onsumption #asket& + rational onsumer is expe ted not to maximise the alorie onsumption #ut to maximise her utility& Nevertheless, "hen an individual is #elo" her su#sist% en e level of alorie intake, she suffers from various physi al ina#ilities like heada he, di77iness and la k of on en% tration& Su h suffering onstitutes a form of disutility on the part of this on% sumer& Sin e less than the threshold level of onsumption implies disutility for her, she minimises these sufferings #y aug% menting onsumption of food items hav% ing relatively higher alorie ontent per unit of expenditure& The food items, "hi h provides relatively high amount of alorie per unit pri e, are olle tively defined as Istaple foodsJ in this ontext&

+ onsumer near the poverty line has to invest heavily in staple foods to mini% mise the disutility asso iated "ith not NO T E S meeting the re*uired su#sisten e level of alorie intake& In other "ords, for a alorie%deprived person, the marginal utility of an additional alorie is *uite high& 'e ause of this high marginal uti% lity of an additional alorie, a utility maximising onsumer "ho annot af% ford to meet the alorie re*uirement is expe ted to onsume greater amount of staple foods, the heapest availa#le sour es of alories& +s the in ome of a onsumer in reas% es #eyond the poverty line $ "hen she an afford to onsume "ithout #eing onstrained #y the alorifi re*uirement $ her preferen es lend gradually more and more "eight to"ard the non%nutri% tional attri#utes of food items& +s she no longer looks for the heapest availa#le sour e of alories, the marginal utility of an additional alorie de lines and the onsumer su#stitutes to"ards food items "hi h are more expensive sour es of alories& In fa t, 4eaton and 4rG7e 5.//;6 find empiri al eviden e that a ri her onsumer, in general, allo ates her food expenditure differently om% pared to a relatively poorer onsumer, s"it hing from ereal to fattier and s"eeter foods like meat, edi#le oil and sugar& The notion, "hether or not an in% dividual has passed the su#sisten e level, is uno#serva#le from the alories thresh% old perspe tive, #ut her hoi e to s"it h a"ay from staple food to expensive food items reveals that she is #eyond the su#% sisten e level of alorifi re*uirement& 'e ause of this revealed preferen e in the hoi es made #y a onsumer, "e do not need to identify any general alorie threshold for all onsumers or even a parti ular alorie threshold for ea h onsumer& Instead the share of staple alories in the total alorie onsumption "ill determine "hether the onsumer is #eyond the su#sisten e level, also alled the poverty line in this ontext&: It may "ell happen that the staple al% orie share 5hen eforth S-S6 may not rea h the maximum value of ://A even for the poorest on eiva#le onsumer& This is #e ause of the fa t that the ooking te hnology may re*uire use of non%staple

ingredients like oil& In that ase, the hoi es of a onsumer "ill #e on% strained #y te hnology and her S-S has to depart from its maximum on eiva#le value& )oreover, it might also happen that ooking te hnology mandates use of a ertain minimum amount of non% staple ingredients and, therefore, the a tual share of staple food 5S-S6 may fall "hen in ome plummets on the left side of the su#sisten e level& Therefore, a plot of S-S against in ome "ill reveal a on% stant or mildly in reasing part for lo"er values of in ome, su#se*uently follo"ed #y a rapidly de lining part for ompara% tively higher values of in ome 5@igure :, p 236& The point, past "hi h S-S starts de lining rather rapidly, dis loses the position of su#sisten e level or Ipoverty lineJ under () methodology& The () approa h takes are of limita% tions of prevailing approa hes ited #e% fore& @irst, different individuals have dif% ferent alorifi re*uirements and this is usually not onsidered in the threshold alorie approa h& Sin e the () approa h is #ased on revealed preferen e of a on% sumer, it is not re*uired to look into par%

This pro#lem is automati ally taken are of in the () approa h, sin e "e dire tly look into the onsumer hoi e, "hi h automati ally im#i#es in ome and pri es in it, irrespe tive of any pri ing index& The () approa h is #ased on a rational onsumer "ith homotheti utility fun % tion implying a proportionate in rease in onsumption of all goods su#9e t to a rising in ome& This kind of onsumption pattern may not al"ays #e o#served& On the ontrary, so ietal norms may play a role in shaping onsumer!s hoi es& )ar9it 5./::6 dis usses su h a frame"ork "ith non%homotheti utility, in "hi h hav% ing lo"er than average in ome in reases the marginal utility of onspi uous on% sumption or onsumption of the status good& + frame"ork of this kind may give rise to the s enario that in spite of in% reasing in ome, a onsumer de reases onsumption of food items for luxury goods& Therefore, poorer onsumers "ould have relatively even higher S-S than in the () e onomi environment&. The () methodology is ro#ust in that respe t& Of ourse, household hoi es may

not #e entirely driven #y rationality in a tuality& +n irrational onsumer may undertake hoi es leading to violation of on lusion dra"n #y the () approa h& ? 4ata and )ethodology ,e use the NSSO survey data from the <:st 8ound, ondu ted in .//0%/1, to ompare the extent of poverty estimated using the () method and other estimates, namely, the Tendulkar -ommittee num% #ers and the Planning -ommission num% #ers #ased on the Dakad"ala approa h& ,e also present the () estimates of pov% erty using NSSO data from the <0th 8ound ondu ted in .//2%/3& In the data, "e have information a#out food items on% sumed #y a household in the last ?/ days pre eding the date of survey along "ith demographi details of the house% hold& ,e al ulate the total amount of alories onsumed #y a household through different ategories of food items, su h as ereals, milk, oil%spi e% sugar, pulse% vegeta#les, fruit%meat,

,e onsider ereals along "ith ereal su#stitutes 5 9a kfruit and tapio a6 as the staple food
et &? ti ularities of different onsumers& Se % ond, the limitation regarding the imper% fe t a#sorption of alories is also taken are of in the () approa h& It has also #een surmised that sin e the effi ien y for a#sorption of alories varies a ross individuals, alories onsumed is a poor measure for alories a#sor#ed& Ender the () approa h, "e look into the on% sumption hoi es made #y an individual, not into the alories onsumed& If an in% dividual!s effi ien y of a#sorption of al% ories is lo" then his onsumption hoi es "ill remain in lined more to"ards staple foods, resulting in a onsumption pat% tern different from that of a healthy individual& So the pro#lem of distin% guishing #et"een a#sorption of alories and onsumption of alories is dealt "ith& Third, there is a re urring ontro% versy related to the pri e indi es #eing used to update the

poverty line using the methodolo gy suggested #y the Dakda"ala -ommittee in :;;?& 4eaton and 4rG7e 5.//;6 re kons that use of the -onsumer Pri e Index for +gri ultura l Da#ourers 5-PI+D6 to revise the rural poverty line results in underesti mation of poverty&

T a b l e 1 : P o v e r t y E s t i m a t e s f o r t h e R u r a l H o u s e h o l d s i n 2 ! " #

l e r + l l % I n d i a f i g u r e a f t e r a g g r e g a t i o n of ?<&3 $ $ $ Planning -ommission estimates .3&? .,/02 11; $


Bead $ $ 4ail +verag Tendulk ar -ut -ommit tee 8at )P- es timates Bou

A6 Intake )P-E Bousehold S-S of E xpenditure

S-S - a lo r ie

(ensen and )iller +ll%India Study 5ur# an s a mp l e 6 .1& 3 :,;<< :,:./ /&1; ., .33 /&<1 (ensen and )iller +ll%India Study 5ur#an%refined sample6 .3&1 :,;;; :,.1; /& 1< .,0?? / &< 0 (ensen and )iller +ll%India figure aggregation of safter tate" is e f i g u re s .<&? $ $ $ $ $ Planning -ommission estimates .1&2 $ $ $ $ $ Tendulkar -ommit tee estimates .1&2 $ $ $ $ $

$: Rural and %rban


Bead ount +verage 4aily +verage +verage -ut%off -ut%off 8atio 5in A6 Intake )P-E S-S Bousehold S-S of E xpenditure - a lo r ie

state %

"ise f igures $ $

(ensen and )iller +ll%India study 5r u r a l sample6 ?:&? .,/:. 2;3 / &< 3 .,?3/ /&2/ (ensen and )iller +ll%India study 5rural%refined sample6 . 3 &; .,/1. ;/0 / &< < . ,? ? 0 /&2/ (ensen and )iller +ll%India figure for rural sample after aggregation of state% "ise figures ?.&? $ $ $ $ $ (ensen and )iller +ll%India study 5ur# an s a mp l e 6 0 /&3 :,;1? :,113 /&12 0,?:1 / & 13 (ensen and )iller +ll%India study 5ur#an%refined sample6 ?< & 3 :,;2; :,223 /&10 0,/03 /& 13
(ensen and )iller +ll%India figure for ur#an after a g g re g a t i o n o f s t a t e % " i s e f i g u re s ?<&1 $ $ $ $ $

0:& 3 $ $ $

of E x p e n di t u r e

-a lo r ie

$ $

(ensen and )iller +ll%India Study 5rural se tor6 ?:&/ .,/1/ 121 /&2/ :,<1/ /&2? (ensen and )iller +ll%India Study 5rural%refined sample6 ??&? .,/10 <?0 /&<3 :,21; /&2: ( e n s e n a n d ) i l

seholds in 2 !" #
Bead ount +verage 4aily +verage +verage -ut%off -ut%off 8atio 5in

&i'ure 1: Trend E(hibited by Sta)le *alorie Share

S-S

-ut%off in ome

In omeK,ealth

to al ulate S-S for a household, "hi h is the ratio of alories o#tained from onsumption of staple food and total alories onsumed& Similarly, "e an also al ulate the milk alorie share, oil%spi e% sugar alorie share, pulse%vegeta#le al% orie share and fruit%meat alorie share& In the () methodology, "e need to plot the S-S of a household against its in% ome or "ealth to determine the point of su#sisten e as illustrated in @igure :& NSSO surveys do not re ord the data on the in ome or "ealth of a household& Therefore "e use household expendi% ture as a proxy for in ome&0

,e

examine
Table !: State"+ise Poverty Estimates ,Head-ount Ratio. for 2 !" #
States Er#an Er#an 8ural 8ural 5() 5Tendulkar 5() 5Tendulkar )ethod6 -ommi t tee6 )ethod6 -ommit tee6

the pattern in a plot of S-S against household expendi% ture& In this plot, the point past "hi h S-S starts falling sharply indi ates the point onsumers su#stitute ex% pensive alorie sour es for staple food& This point an #e seen as a partition #et"een the poor and the non%poor& The household ex% penditure orresponding to this point is alled the ut%off household expendi% ture& This ut%off value is the #asis of poverty estimation under () methodol% ogy& On e this point is identified, mem% #ers of a household "ith household ex% penditure #elo" this point are atego% rised as poor& Sin e there are many households in and around a parti ular expenditure level "ith different values of S-S, "e re% *uire estimating the average value of S-S for any parti ular expenditure level& ,e appeal to the kernel method in al% ulating a lo al average of S-S against expenditure of a household& )ore par% ti ularly, "e use the Gaussian kernel fun tion "ith appropriate #and"idth 5Pagan and Ellah :;;;H 23%:126 as the

India, a single nation"ide estimate "ould have #een ade*uate& Other"ise, a separate estimation for ea h state is prefera#le& ,e have arried out #oth for the <:st 8ound& Bo"ever, the <0th 8ound is not a *uin*uennial round of NSSO and therefore due to inade*ua y of data points, state%"ise estimates are not as relia#le as of the <:st 8ound& ,e have done our analysis separately for the rural and ur#an se tors& +long "ith this "e have further segregated the rural and ur#an se tor into t"o su#groups, namely, rural refined and ur#an refined& The refined lasses onsist of house% holds, "hose informant is apa#le and ooperative as identified #y the investi% gator& 'e ause of the nature of the in% formant, data provided #y these house% holds should #e more relia#le& The origi% nal estimates have #een ross% he ked "ith estimates o#tained from this re% fined sample "ith #etter *uality data& 0 8esults and 4is ussion 0&: -onsisten y of () +pproa h The ore idea involved in the () ap% proa h is su#stitution to"ards expensive sour es of alories "ith the in rease in in ome& This #ehaviour should #e revealed in the onsumption hoi es made #y the individuals& One "ay to verify the vera ity of the () approa h is to look at the graph plotting alorie shares of expensive food items 5su h as milk, fruit%meat6 against household expenditure& @igure . 5p 2;6 displays an in reasing milk alorie share against the log of household expendi% ture, "hi h underlines the fa t that milk

+ndhra Pradesh + ssam 'ihar -hhat tisgarh 4elhi Go a Gu9arat Bar yana

.2&1 .;&0 ?<& 3 ?;&/ :?&2 :/&? :1&< .:&:

.?&0 .?&1 .:&3 0?&2 .3&0 :.&; ..& . ./&: ..&0 0 &<

?0&2 1/&? ..&2 0?&3 0 < &; 3& 3 :1&0 ?1&; :;&; .<&2

?.&? ??& < ?<&0 11& 2 11& : :1& < .3&: ?;&: .0& 3 . 1& /

+runa hal Pradesh ?0&:

Bima hal Pr adesh ?1&;

(harkhand Karnat aka Kerala

(ammu and Kashmir .;&2

?2&1 .0&. :1&0

.?& 3 .1&; :3&0

0?&0 ?1&/ . 1& 1

1:&< ?2&1 ./&.

The range of this plot an vary #et"een the mini% mum value of household expenditure and the maximum value& In #oth the ends, the noise element of the plot domi% nates over the deterministi omponent for pau ity of o#servations& ,e make suita#le ad9ustments in the range to on% tain the noise omponent&
underlying kernel fun tion&1 If the ooking te hnology and pri es did not vary mu h a ross regions of
Pun9a# 8a9as than Tamil Nadu :;&2 .:&: .3 & 3 :3&2 .;&2 :;&2 ./&3 0.&3 ??& <

)adhya Pradesh ..&; ?1&: 0:&/ )aharashtra .1&2 .1&< 01&2 )anipur :;&; ?0&1 23&/ )eghalaya ::&< .0&2 :0&/ Tripura )i7oram ?< & . 2 &; . . &; Ettar ? Pradesh Nagaland 0&1 0& ? ?0&/ Ettarakhand Orissa 0?&/ ?2&< 1<&1 Pondi herry ,est 'engal 3&< ;&; ..&; +ll%India

1?&< 02&; ?;&? :/&1 .?&/ :/&/ </&3 ./&3

..&: ?1&3 Sikkim ?2& 1

. .& 1

.1&;

: 3 &0

?:&3

?:&: ?.&< .0&/


?0&/ .<&?

Table #: S)earman/s Ran0 *orrelation *oeffi-ient for the 12 and Tendul0ar *ommittee Estimates for States 5.//0%/16
Er#an 8ural

:; )a9or States

/&?;

)arginally not signif i ant at 1A 5p %value /&/16


Er#an Poor 8ural Poor

/&22
Er#an Non%Poor

Signif i ant at :A
8ural Non%Poor

Table 3: Avera'es of Sele-t Statisti-s for Poor and Non"Poor 5.//0%/16


S-S Per apita daily alorie intake )onthly per apita expenditure B o u s e h o l d e x p e n d i t u re /&1< :,23< 20; :, 113 /&2: :,3<0 0.2 :, :0? /&1; :,;2. :,:?0 1, < 02 /&2/ .,/.; 13. ?, 0 0:

+ll States 5p%value /&:36 :A

/&:2

Not signifi ant at 1A /&21 Signifi ant at

Table 4: Avera'es of Sele-t Statisti-s for Poor and Non"Poor 5.//2%/36

is onsumed proportionately more #y the ri h ompared to the poor& The alorie share from fruit and meat also reveals a some"hat similar trend& + lot of Indians do not onsume meat and, therefore, the in reasing trend of meat alorie share is some"hat dampened #y this fa tor& One interesting o#servation is that the alorie share of oil%spi es% sugar is almost on% stant or mildly in reasing at all levels of household expenditure& This points to the te hnologi al onstraint in ooking, "hi h for es people to use some expen% sive food items like oil and spi es, irre% spe tive of their household expenditure& This is very "ell refle ted in almost flat or the very mildly in reasing oil%spi e%

in .//0%/1 turns out to #e 8s .,.33 and it leads us to a figure of .1&3A as the extent of poverty in ur#an India& Poverty esti% mates for year .//2%/3 stand at ?:&?A and 0/&3A in the rural and ur#an se tors "ith ut%off household expenditures of 8s .,?3/ and 8s 0,?:1, respe tively& In the se ond pro edure, an all%India

is done "ith sample households restri ted to the refined samples 5"ith apa#le and ooperative informants6 se tor%"ise $ rural refined and ur#an refined, respe tively& Poverty estimates using this pro edure stand at ??&?A and
study< .3&1A, "ith 8s :,21; and 8s .,0?? as ut%offs for household expenditures in rural%refined and ur#an%refined sample,

respe tively for .//0%/1& @or .//2%/3, the poverty estimates stand at .3&;A and ?<&3A "ith 8s .,??0 and 8s 0,/02 as ut%off household expenditures for rural% refined and ur#an%refined samples, re% spe tively& In the third pro edure, "e restri t our sample of households to a parti ular state, se tor% "ise& ,e estimate the extent of poverty, state%"ise,2 through the

head ount ratio follo"ing the same methodology of plotting S-S against household expenditure and spotting the ut%off household expenditure& Su#se% *uently, "e aggregate state%"ise estimates "ith appropriate "eights to o#tain the all%India poverty estimate& @or .//0% /1, this aggregated poverty estimate for
sugar alorie share& @rom @igure ., it is evident that there has #een an in rease of :&23A in the oil%spi e% sugar alorie share on moving from point + to point ', "hereas for the orrespon ding move% ment $ from point - to point 4 $ on the milk alorie share urve, the in rease is of a magnitud e of
&i'ure 2: Plot of 2il05 Oil"S)i-e" Su'ar5 &ruit"2eat *alorie Share a'ainst Household E()enditure
/&:3

1 & 0 & ,

+ 5:< :., /&:: 216 4 513 ;<, /&/ 3.. 6 5:< :., /&/ .; <;6

' 513;<, /&:?1?6

a l o r

Staple
Oil%Spi e%Sugar @r uit s% ) ea t

S-S against household expenditu / re for the rural and ur#an se tors respe tivel y for .//0% /1 5@igures ? and 06 and .//2% /3 5@igures 1 and <, p 3/6& +long "ith S-S, "e also plot the shares of other atego% ries of food itemsH fruit% meat, milk, oil% spi e% sugar, and pulse% vegeta#le s& ,e al ulate the poverty estimates using three different pro edur es& @irst, an all% India studyH "e plot S-S against house% hold expendit ure for all househol ds in our sample,
:// / .2/ / 2.; /

:<:. 13;<

O i l %
Bo ehold Exp dit 5Do rith S a

ts% )e at -al orie Sha re

&i'ure !: *alorie Shares a'ainst Household E()enditure for All"7ndia %rban Sam)le 5.//0%/16

/&2 /&< /&1 /&0 /&?

Staple -alo

S p i :/3/ e % S u g a r 0// :<1/

&i'ure 6: *alorie Shares a'ainst Household E()enditure for All"7ndia Rural Sam)le 5.//0%/16
/&3 /&2 /&< /&1 /&0 /&? /&. /&: /

.;:< 232?&.
Bo use hol d Ex pe ndi tur e 5Do gar ith mi S ale 6

P u l s

a l o

ulate the household expenditu re at "hi h the lo al average of S-S attains the maximum& ,e define this as the ut% off expenditu re& -ut%off household ex% penditure for the rural sample in .//0% /1 is found to #e 8s :,<1/& ,e al ulate the extent of poverty in rural India through the

head ount onsid%

ratio

#y

ering the proportio n of the populatio n "ith a lo"er household expenditu re ompared to the riti al value& The figure stands at ?:A in .//0%/1& -ut%off house% hold expendit ure for the ur#an sample
; / /

Oil%Spi e%Sugar

Pulse% Megeta#le )ilk -alorie Share @ruits%)eat -alorie Share

a l o r

.0?/ <1<: :22:0&2 Bousehold Expenditure 5Dogarithmi S ale6

3/

E onomi > Politi al ,eekly


NOTES

f e #rua ry ;, ./:?

vol x lv i i i no <

EP8

India stands at ?<&3A and .<&?A for rural andNur#an OT ES se tors, respe tively& 0&? -omparison of the () @igures "ith the Tendulkar Estimates ,e ompare the poverty estimates of our study "ith estimates of the Plan% ning -ommission and the Tendulkar -ommittee& @or the rural and ur#an se tors in .//0%/1, the results have #een presented in Ta#les : and . 5p 226 respe tively& @or .//2%/3, head ount ratios for estimating poverty are ta#u% lated in Ta#le ? 5p 226& 8egardless of methodology, "e al"ays find that the extent of poverty is higher in the rural se tor ompared to the ur#an se tor in .//0%/1& )oreover, the variation #e% t"een different estimates on the extent of poverty in the ur#an se tor is rather lo" "ith various estimates ranging ap% proximately #et"een .1A and .3A& Bo"ever, the extent of poverty in the rural se tor is more ontroversial& The Planning -ommission reported the head% ount ratio of rural poverty at .3&?A& This estimate of rural poverty seemed unrealisti & The Tendulkar -ommittee 5Planning -ommission .//;6 ela#o% rated on various issues "hi h led to the underestimation of poverty in the rural se tor, su h as the use of the outdated poverty line #asket that "as dra"n up in :;2?%20, the -onsumer Pri e Index for +gri ultural Da#ourers 5-PI%+D6 under% estimating the inflation and no expendi% ture #eing attri#uted to the onsump% tion of health and edu ational servi es& The Tendulkar -ommittee re tified these issues "orking "ith NSS data and "ith their revised methodology, they estimated the rural poverty figures at

ur#an se tors, respe tively $ defined #y the () approa h& There#y, "e estimate that the monthly per apita expenditure around the ut%off expenditure stands at 8s 0/: and 8s 1?/ for rural and ur#an, respe tively& ,e have ta#ulated () pov% erty estimates, state%"ise, vis%N%vis their Tendulkar -ommittee ounterparts in Ta#le 0 5p 236& +part from some small states like Goa, )i7oram, Nagland, et , #oth estimates do not differ too "idely& In some states, the head ount ratio is larger #y the Tendulkar -ommittee methodology, "hereas the reverse is also o#served in many states&
/&3

The extent of differen e #et"een t"o sets of estimates ould also #e measured #y the ro#ustness of the ranks of entities under these t"o sets of estimates& Spear% man!s rank orrelation oeffi ient is a standard measure to find ro#ustness of ranks for t"o sets of estimates& ,e om% pute the Spearman!s rank orrelation o% effi ient #et"een the () estimates and the Tendulkar -ommittee estimates for states 5reported in Ta#le 06, for #oth rural and ur#an se tors& Ta#le 1 5p 236 reports this statisti for all states as "ell as for :; ma% 9or states&3

,e test the

null hypothesis of
7ero rank orrelation using this statisti s&

&i'ure #: *alorie Shares a'ainst Household E()enditure for All"7ndia Rural Sam)le 5.//2%/36
Staple -alorie Share 5S-S6 /&2 /&< /&1 /&0 /&? /&. /&: / <// :<./ .?3/ Oil%Spi e%Sugar Pulse%Megeta#le )ilk -alorie Share @ruits%)eat -alorie Share

a l o r
0?20 ::3/;&3

Bousehold Expenditure 5Dogarithmi S ale6

&i'ure 3: *alorie Shares a'ainst Household E()enditure for All"7ndia %rban Sam)le 5.//2% /36
/&2 Staple -alorie Share 5S-S6 /&< /&1 /&0 /&? a l o r Oil%Spi e%Sugar Pulse%Megeta#le )ilk -alorie Share

@ruits%)eat

-alorie Share
/&. /&: / 1// 0?:1

.<12.&/1

:?1/

?<01

;30:&1

Bousehold Expenditure 5Dogarithmi S ale6

0:&3A $ an upper #ound for the extent of rural poverty& () estimates of rural pov% erty are in #et"een the values pro9e ted #y the Tendulkar -ommittee and the Planning -ommission& + omparison of the poverty limits us% ing t"o approa hes is a pertinent *ues% tion& @or .//0%/1, the Tendulakar -om%

&i'ure 4: Distribution of Rural Households by S*S


0&1 =ear .//0%/1 0 ?&1 ? .&1 .

mittee de ided 8s 002 and 8s 12; as pov% erty lines for rural and ur#an se tors, respe tively& ,e find the lo al NO T E S averages for the household expenditure 5and house% hold si7e6 around the ut% off expenditure $ 8s :,<1/ and 8s .,.33, for rural and

:&1 : /&1 / /&. /&? /&0

=ear .//2%/3

/&1 /&< /&2 Staple -alorie Share %%%%%%%L

/&3

/&;

&i'ure $: Distribution of %rban Households by S*S


?&1 ?

.//2%/3 as ?:&?A 5Ta#le ?6, "hi h is *uite similar to the estimate of ?:&/A 5Ta#le :6 in .//0%/1& Nevertheless, there is a massi . & ve 9ump . in the extent of : ur#an & povert : y in .//2% / /3 & "hen / ompa red to the orres pondi ng figure s in .//0% /1& Going #y the all% India study, ur#an pov% erty has #een augm ented #y a "hop ping :1A $ from .1&3A 5Ta#le .6 to 0/&3 A 5Ta#le ?6& Eve n if .//2% /3 is not a *uin* uenni al round, this

m a

/&/1 /&?1 /&<1 /&;1

/&:1 /&01 /&21

/&.1 /&11 /&31 Sta Sha

&i'ure 9: Distribution of Rural Households by Their Daily Per *a)ita *alorie 7nta0e
:&3 :&< :&0

: & :

= e

/ & / & / & / & /

an explanati on& It ould also #e noted that there is no su h in rease in the or% respondin g estimate of rural poverty& The o#vious suspe t is high episodes of inflation during .//3, espe ially in food items& There is eviden e that Ifood prod% u t sJ at ego ris ed un der Im an ufa tu red pro du tsJ $ "h i h are def init ely o n% su me d mo re

i n . /

Year

0 < 3 :/

1 2 ; :: Dog p

&i'ure 1 : Distribution of %rban Households by Their Daily Per *a)ita *alorie 7nta0e
:&3 :&< :&0

of poor "orkers, often migrants, #e ame unemploye dF so laying%off of these "ork% ers may have resulted in higher poverty in the ur#an se tor& 0&1 )oveme nt of S-S and -alorie P e r a p i t a o n s u m p t i o n ,e not e the dif fer en es #et "e en t" o gro ups H one #el o"

t h

.
0/ < 3 :/

: & : / & / & / & / & /

= e

+ eptan e of the null hypothesis implies that the t"o sets of estimates, the () esti% mates and the Tendulkar -ommittee esti% mates, are not related& @or the rural se % tor, the ranks of different states are *uite ro#ust irrespe tiv e of the methodol ogy or sample of states hosen& @or the ur#an se tor, the ro#ustness of ranks is moder% ately less& It may #e related to the in iden e

to note that the Do mean S-S da for poor peo% ple is less than the ut%off S-S value, of greater "hi h may variation #e in interprete population d as a at% tri#utes pie e of in the eviden e ur#an regarding se tor& the presen e 0&0 of an Poverty in reasing Estimate part in the s in #eginning .//0%/1 of S-S and urve& +s .//2%/3 mentioned +s far as #efore, this in% rural reasing poverty is is on erned, part there is #e ause of the virtually ooking no hange in .//2% te hnology onstraint /3 over "hi h .//0%/1& In the all% for es peo% ple to use India study, "e some expensive have food items al ulated oil, the head like ount ratio spi es, irrespe tiv in e of their Year h o u s e h o l d e x p e n d i

1 2 ; ::

t u 3.

e &

f e #rua ry ;, ./:? vol x lv i i i no < EP8 E onomi > Politi al ,eekly

,e plot the kernel density fun tion $ a non%parametri "ay to plot the fre*uen% y distri#ution $ for S-S& @igure 2 5p 3/6 sho"s the distri#ution of S-S for rural households for the years of .//0%/1 and .//2%/3, respe tively& @or a point in the x%axis, the area #eneath the urve and to the left of a point gives us the proportion of people "ho are #elo" that parti ular value of S-S& @igure 3 5p 3:6 presents the same for ur#an households& Though the distri#ution of S-S for rural households stays roughly the same in #oth years, for the ur#an households there has #een a left"ard shift in the urve "hi h sho"s that over the years people are su#stitut% ing a"ay from staple foods& This is inter% esting onsidering that ur#an poverty has gone up in this time& It definitely im% plies a hange in food ha#its altogether so that even after onsumption of less ereal, overall there is an in rease in poverty as measured #y the pattern of S-S& This is also onfirmed #y mu h lo"er ut%off value for S-S in .//2%/3 ompared to .//0%/1& Our plots depi t the kernel density distri#ution of daily per apita alorie intake for rural and ur#an se tors $ @ig% ures ; and :/ 5p 3:6, respe tively& There has also #een a left"ard shift over the years in per apita daily alorie on% sumption& This de line in alorie intake is, indeed, onsistent "ith the findings of 4eaton and 4re7e 5.//;6& Bo"ever, "hen this is a ompanied #y an in rease in poverty estimates for the ur#an se % tor, it poses a on ern of inade*uate al% orie intake for a large population& 1 8emarks -on luding

either& 'y these t"o ounts, the () methodology provides a useful alter% native "ay to meaningfully estimate the extent of poverty in India&
No t e s
1

"eight of /&2 to ea h adult mem#er ex ept the first, to "hom "e assign the full "eight of :, and /&1 to ea h person under the age of :3& ,e use household expenditure per adult using this e*uivalen e s ale& ,e end up "ith head ount ratios of .&/A and 0&1A for rural and ur#an households, respe tively& The formula for the kernel average is given #yH E5S-SOExpenditure6 P iP:n

surd results 54ev .//16 and exaggera% tion of poverty figures is no solution

+ numeri al example may elu idate the s e%

55Expenditurei $

,hat makes this study interesting is #road ompara#ility of head ount ratio estimates using the () approa h to the Tendulkar -ommittee figures& 'y no means is this a oin iden e #e ause of the signifi ant ompara#ility #et"een t"o sets of figures for all ma9or states& The less populated states suffer from la k of data and that ould #e one of the reasons of divergen e #et"een t"o sets of estimates en ountered& In this on% text, "e highlight the fa t that alorie norm estimation often gives rise to a#%

Expenditure6Kh6 Q S-S i K iP: K55Expenditurei nario& Suppose, a onsumer an have t"o items $ ri e and hi ken& @or simpli ity, "e assume that one kilogram of ri e ontains ?,?// alo% ries and osts 8s ?/ implying that ri e ontains ::/ alorie per rupee& On the other hand, one kilogram of hi ken ontains :,/// alories and osts 8s ://& -onse*uently, hi ken on% tains :/ alories per rupee& 'e ause ri e on% tains more alories per unit ost, it is a staple food& Det us onsider t"o onsumersH one of them is poor and the other ri h, earnings re% spe tively 8s ?/ and 8s 2/ per day& 'oth of them "ould like to take hi ken, taste"ise, #ut if their daily alorie intake falls short of :,2//, they feel #ody in onvenien es making it man% datory for them to onsume at least :,2// alo% ries& The ri h onsumer may spend 8s </ for hi ken 5 alories </Q:/ P <//6 and 8s :/ for ri e 5 alories :/Q::/ P :,://6 and there#y satisfy her #iologi al re*uirements& The staple alorie share 5S-S6 for her is <0&2A 5P:,://K:,2//6& The poor onsumer may on% sume 8s :< for hi ken 5 alories :<Q:/ P :</6 and 8s :0 for ri e 5 alories :0Q::/ P :,10/6 and there#y satisfy his #iologi al re*uirements of :,2// alories& Bis S-S, ;/&<A 5P:,10/K:,2//6, is remarka#ly more ompared to the ri h on% sumer& Therefore, "e an differentiate #et% "een a ri h person and a poor person #y om% paring S-S of their onsumption #undles& I use another numeri al example, similar to the one mentioned in previous note 5Note :6 to il% lustrate this point& Det us suppose, the ri h per% son earns 8s 22 per day and the poor person 8s ??& In addition to onsumption of food 5ri e and hi ken together6, there is also a luxury good& Det us say the ri h person spends one% eleventh part of her earnings for the luxury good& Therefore, she is left "ith a spending of 8s 2/ for food& Ber onsumption of ri e and hi ken as "ell as her S-S "ill #e identi al to the previous numeri al example& The poor per% son, #y the frame"ork of )ar9it 5./::6, has a #igger marginal utility for the luxury good ompared to the ri h person and therefore, "ill spend proportionately more for luxury good& In other "ords, he "ill spend more than one%elev% enth of his in ome, 8s ? for luxury goods& Therefore, his expenditure on food "ill #e less than 8s ?/& ,ith less money #eing invested on food, he "ill #e for ed to use even more of staple food leading to an augment in S-S over the previous numeri al example& In sum, the differen e #et"een the ri h and the poor, S-S%"ise, is enhan ed in the frame"ork of )ar9it 5./::6& The alorifi ontents of food items are mostly taken from Gopalan et al 5:;2:6& ,e have also onsulted the Enited States department of ag% ri ulture data #ase a essed at httpHKK"""& nal&usda&govKfni Kfood ompKsear hK 5last a % essed :1 (uly ./::6& ,e have also appealed to per apita expendi% ture and adult e*uivalent ad9usted per apita expenditure as alternative proxies for in ome, #ut the poverty estimates are unrealisti ally lo"& @or rural and ur#an se tors in .//0%/1, the poverty ut%off values of )P-E are found to #e 8s ./? and .?., respe tively& This is e*uiva% lent to .&/A and :&/A of poverty in head ount terms, "hi h is surely una epta#le& ,e have "orked "ith an adult e*uivalent s ale giving a

$ Expenditure6Kh6, "here K5R6 is a kernel fun % tion 5Gaussian6, h is the #and"idth, n is the num#er of data points and Expenditurei

8 e fe r e n e s
4asgupta, Partha 5:;;16H An Inquiry into Well- Being and Destitution 5DondonH Oxford Eniver% sity Press6& 4eaton, +ngus and (ean 4rG7e 5.//;6H I@ood and Nutrition in IndiaH @a ts and InterpretationsJ, Economic & Political Weekly, 00526H0.%<1& 4ev, )ahendra S 5.//16H I-alorie Norms and PovertyJ, Economic & Political Weekly, 0/536H 23;%;.& Ghosh, (ayati 5.//;6H IGlo#al -risis and the Indian E onomyJ in Global inancial !risis" Im#act on India$s Poor% &ome Initial Pers#ecti'es, Enited Nations 4evelopment Programme India, vie"ed on .< 4e em#er ./:: 5htt pHKKdata& undp&org&inK@inan ial-risisK@inal@-P&pdf 6& Gopalan, -, ' M 8ama Sastri and S 'alasu#rama% nian 5:;2:6H (utriti'e )alue o* Indian oods 5Bydera#adH National Institute of Nutrition, Indian -oun il of )edi al 8esear h6& (ensen, 8o#ert and Nolan )iller 5./:/6H I+ 8e% vealed Preferen e +pproa h to

and S-S are

o#servation& i

the

values

for

the

ith

<

2 3

+nother popular kernel fun tion, namely, Epane hnikov kernel, yields similar plots and similar estimates& @or the sake of #revity, "e desist from pu#lishing them in the paper& The plots of rural refined and ur#an refined samples are *uite similar to the plots o#tained for rural and ur#an se tors& These plots are availa#le from the authors on re*uest& The plots of S-S against household expendi% ture, state"ise and se tor%"ise, are availa#le from the authors on re*uest& These :; ma9or states are +ndhra Pradesh, +ssam, 'ihar, -hhattisgarh, Gu9arat, Baryana, Bima hal Pradesh, (harkhand, Karnataka, Kerala, )adhya Pradesh, )aharashtra, Orissa, Pun9a#, 8a9asthan, Tamil Nadu, Ettar Pradesh, Ettarakhand and ,est 'engal&

)easuring Bunger and EndernutritionJ, N'E8 ,orking Paper No :<111& )ar9it, Sugata 5./::6H I-onfli ting )easures of Poverty and Inade*uate Saving #y the Poor STI The 8ole of Status 4riven Etility @un tionJ, 4is ussion Papers Series 0.0, S hool of E o% nomi s, Eniversity of Uueensland, +ustralia& Pagan, +drian and +man Ellah 5:;;;6H (on#arametric Econometrics 5-am#ridgeH -am#ridge Eniversity Press6& Planning -ommission 5.//;6H +e#ort o* t,e E-#ert Grou# to +e'ie. t,e /et,odology *or Estimation o* Po'erty, Government of India, Ne" 4elhi& Sharma, Savita 5.//06H IPoverty Estimates in IndiaH Some Key IssuesJ, E84 ,orking Paper No 1:, +sian 4evelopment 'ank, )anila& Sved#erg, Peter 5.///6H Po'erty and 0ndernutrition 5DondonH Oxford Eniversity Press6& Sthanumoorthy, 8 5.//36H INature of -urrent In% flation in @ood Pri esJ, Economic & Political Weekly, 0? 50/6H :2%.:& Su#ramanian, S 5./::6H IThe Poverty DineH Getting It ,rong +gainJ, Economic & Political Weekly% 0< 5036H ?2%0.& S"aminathan, )adhura 5./:/6H IThe Ne" Poverty DineH + )ethodology 4eeply @la"edJ, Indian 1ournal o* 2uman De'elo#ment, 05:6H :.:%.1&

Anda mungkin juga menyukai