Вы находитесь на странице: 1из 21

Poverty Centre

INTERNATIONAL
United Nations Development Programme Working Paper number 29 September, 2006

NEW GLOBAL POVERTY COUNTS

Nanak Kakwani International Poverty Centre, United Nations Development Programme

and
Hyun H. Son International Poverty Centre, United Nations Development Programme

Working Paper

Copyright 2006 United Nations Development Programme International Poverty Centre

International Poverty Centre SBS Ed. BNDES,10o andar 70076 900 Brasilia DF Brazil povertycentre@undp-povertycentre.org www.undp.org/povertycentre Telephone +55 61 2105 5000 Fax +55 61 2105 5001

Rights and Permissions All rights reserved. The text and data in this publication may be reproduced as long as the source is cited. Reproductions for commercial purposes are forbidden. The International Poverty Centres Working Papers disseminates the findings of work in progress to encourage the exchange of ideas about development issues. The papers are signed by the authors and should be cited and referred to accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Poverty Centre or the United Nations Development Programme, its Administrator, Directors, or the countries they represent. Working Papers are available online at http://www.undp.org/povertycentre and subscriptions might be requested by email to povertycentre@undp-povertycentre.org

Print ISSN: 1812-108X

NEW GLOBAL POVERTY COU NTS


Nanak Kakwani and Hyun H. Son

ABSTRACT
The m ain obj ective ofthis study is to com pute an internationalpoverty threshol based on the d f ood requirem ent to ensure an adequate cal orie intake f the worl poorest. The study or ds proposes a new m ethodol ogy based on consum er theory to provide a cal oric based international poverty threshol U sing this m ethodol d. ogy, the international poverty l is ine estim ated to be equal to $1.22 in 1993 PPP exchange rates. A ccording to this new yardstick, al ost 1.37 bilion peopl were poor around the worl in 2001. The study al provides gl m l e d so obal estim ates ofhunger, according to which 13.28 percent ofthe worl popul d ation equival to ent 687 m ilion peopl suf ered f l e f rom hunger in 2001.

Keywords:Poverty, Purchasing Power Parity, Gl obalestim ates. JEL Cl assif ication:I D31, O 53, O 57 32,

The authors are gratef to Prof ul essor Peter Townsend f encouraging us to com pl this articl W e woul l to or ete e. d ike acknowl edge hel ulcom m ents and suggestions by Meghnad Desai, Peter Lam bert, Peter Townsend, Sergei Soares and pf Evan Due, which hel ped im prove the f draf ofthis paper. irst t Em ailaddress f correspondence:hyun.son@ undp-povertycentre.org; Tel 55-61-2105 5025; Fax:55-61-2105 5001/5002. or :

I nternationalPoverty Centre

W orking Paper n 29

1 I NTROD U CTI ON
Gl obalestim ates ofpoverty can pl an im portant rol in m onitoring the l ay e eveland change in poverty at the gl oball . I is true that an in-depth anal ofcountry- specif poverty evel t ysis ic prof es is m ore usef f understanding the nature ofpoverty and f devising poverty il ul or or reduction strategies, but gl obalcounts can be used as a powerf device to create pol ul itical awareness in giving priority to poverty reduction, as wel as providing benchm arks f l or individualcountries. The W orl Bank is the m ain institution that produces gl d obalestim ates of poverty. I presented gl t obalpoverty estim ates f the f tim e in its 1990 W orl Devel or irst d opm ent Report. The Report chose a $1 a day poverty l m easured in 1985 purchasing power parity ine, ( PPP) The PPP exchange rates are used because they take into account the l . ocalprices of goods and services that are not traded internationaly. l The $1 a day poverty l was chosen because it was the m ost typicalpoverty l am ong ine ine the l ow-incom e countries f which poverty l or ines were avail e. This poverty l has com e to abl ine be regarded as providing the absol m inim um standard ofl ute iving, bel which m eeting ow basic needs is not f l possibl Based on this poverty l Ravalion and Chen ( uly e. ine, l 1997) estim ated that 1.30 bilion peopl in the worl are poor. l e d Recentl Chen and Ravalion ( y, l 2001)have produced m uch im proved gl obalestim ates of poverty based on an expanded database consisting of297 househol surveys spanning 88 d countries. They have al provided estim ates ofthe change in gl so obalpoverty over tim e. The m ost im portant f eature ofthe new estim ates is that they are based on vastl im proved 1993 y PPP exchange rates f consum ption. These rates are superior to the earl ones in term s of or ier coverage ofcountries and are based on price and consum ption baskets prevail in 1993. ing The W orl Bank determ ined a new poverty l of$1.08 per person per day at 1993 prices, d ine which is the m edian ofthe nationalpoverty l ines in 10 countries:Bangl adesh, China, I ndia, I ndonesia, Nepal Pakistan, Thail , and, Tanzania, Tunisia and Zam bia. I its cal ations ofgl n cul obal poverty, the W orl Bank has now recal ated the num ber ofpoor based on this new poverty d cul threshol which is stil ref d, l erred to as the $1 a day poverty l ( ine Chen and Ravalion, 2001) l . The root idea ofthe poverty l is that it shoul refect the cost ofachieving basic hum an ine d l needs. However, the $1 a day ( $2 a day)poverty l does not refect the cost ofachieving or ine l any kind ofbasic hum an needs. I was chosen m ainl because it was a typicalpoverty l t y ine prevail in the 1980s in a sam pl of33 countries, which incl ing e uded onl 10 l incom e y ow countries. Thus, there is a cl ear-cut need to revise the gl obalpoverty estim ates so that they take som e basic capabil ities into account ( Sen, 1999) . O ne ofthe basic hum an needs is the capabil to be adequatel nourished. A ccording to ity y Lipton ( 1988) access to [ adequate source ofnutrition is a good indicator ofqual ofl e; , an] ity if heal shel education, even m obil which are al refected in nutritionalstatus, al th, ter, ity, l l though not in a l inear or otherwise sim pl way. The m ain obj e ective ofthis study is to com pute an internationalpoverty threshol based on the f d ood requirem ent to ensure adequate cal orie intake f the worl poorest. The study proposes a new m ethodol or ds ogy based on consum er theory to provide cal oric based internationalpoverty threshol I the construction poverty ds. n threshol the study util ds, izes unit record data f 19 l incom e countries ( in A sia and 15 in or ow 4 Af rica) Since the surveys in dif erent countries were conducted in dif erent years, we chose the . f f l atest avail e survey in each country. abl

Nanak Kakwani and Hyun H. Son

The internationalpoverty threshol devel d oped on the basis of19 countries is used to produce gl obalpoverty counts util izing the W orl Bank data base consisting ofover 450 d surveys f about 100 counties covering 93 percent ofthe popul or ation ofl and m iddl ow eincom e countries in the worl d.

2 PU RCH ASI POWER PARI NG TY


The purchasing power parity ( PPP)exchange rates are the essentialingredients ofdeterm ining internationaly com parabl poverty l l e ines. The PPP exchange rates are essentialy the cost of l l iving indices am ong the countries. They alow us to m ake internationalcom parisons ofcosts l ofl iving in dif erent countries. For instance, the W orl Bank has determ ined that $1.08 in 1993 f d PPP ( 1993 prices)dolars is an appropriate internationalpoverty l then util in l ine, izing the 1993 PPP conversion rates, we can estim ate the equival val ofthis poverty l in the countrys ent ue ine l ocalcurrency in 1993 prices. Suppose that we want to estim ate the poverty rate ofthe ith country f which the or househol survey is avail e f 2000, then we need to know the ith countrys poverty l d abl or ine in 2000. This can easil be estim ated using the infation rate that occurred between 1993 y l and 2000. Thus, the estim ation ofgl obalpoverty rates requires both PPP exchange rates and nationalconsum er price indices. The accurate estim ation ofgl obalpoverty rates depends on the accuracy ofboth PPP exchange rates and consum er price indices. I m ust be em phasized that PPP exchange rates were not designed f m aking t or internationalpoverty com parisons; they were m ainl designed f com paring aggregates f y or rom nationalaccounts.1 PPP exchange rates are based on prices ofcom m odities that are not representative ofthe consum ption baskets ofthe poor ( Rao, 2003) More im portantl weights . y, in the PPP baskets ofgoods and services do not adequatel represent the consum ption basket y ofthe poor. The W orl Banks $1 a day poverty l was based on 1985 PPP exchange rates. Recentl d ine y, the Bank has changed the base to 1993. The changing ofthe PPP base can m ake a l of ot dif erence to the poverty l f ines as wel to the poverty rates. I the present study we used the l n 1993 PPP exchange rates. Tabl 1 presents the consum er price indices f the 19 countries in e or 1993 and the survey periods, which are util ized f constructing the poverty threshol This or ds. tabl al presents the 1993 PPP exchange rates. e so

3 H OW D I TH E WORLD BANK ARRI AT TH E $1 A D AY D VE POVERTY LI NE?


The 1990 W orl Devel d opm ent Report presented gl obalestim ates ofpoverty on the basis ofthe $1 a day poverty l m easured using a 1985 PPP exchange rate f the f tim e. The W orl Bank ine or irst d arrived at the $1 a day poverty l based on the country specif nationalpoverty l ine ic ines f a or sam pl of33 countries. These poverty l e ines were obtained f rom a wide range ofsources within and outside the Bank. I is stated that these shoul not be considered as of icial poverty l t d f ines, either ofthe governm ents or ofthe Bank. Many were estim ates f rom independent researchers. Thus, these poverty l ines were evol using widel dif erent m ethodol ved y f ogies.

I nternationalPoverty Centre

W orking Paper n 29

TA BLE 1

Consum er pri i ces, 1993 PPP exchange rates and nati ce ndi onalpoverty l nes i
Countries Survey Year 1998 1998 1998 1996 2000 1998 1994 1998 1997 2001 1996 1997 1996 1999 1998 2000 2000 1998 1996 Consumer price index 1993 73.0 74.4 70.2 70.5 87.2 50.2 90.9 91.8 76.5 48.2 39.7 40.5 36.8 88.5 48.3 159.0 100.0 78.0 82.0 Survey year 186.5 107.1 111.1 103.6 106 223.7 94.7 106.1 121.2 172 144.6 150.2 129.3 122.4 221.6 235.0 168.0 207.0 104.0 1993 PPP exchange rate 56.3 103.4 159.1 142.4 1.3 178.9 339.3 2.4 11.8 530.3 808.0 1.5 11.5 260.0 223.4 12.7 7.0 209.9 9.2 National Poverty line in 1993 PPP dollars 1.21 0.99 1.77 1.81 1.50 2.34 2.26 2.52 1.95 1.11 1.73 1.86 0.76 1.70 1.14 1.42 1.00 1.09 1.11

Burundi Burkina Faso Ivory Coast Cameroon Ethiopia Ghana Guinea Gambia Kenya Madagascar Mozambique Malawi Nigeria Uganda Zambia Bangladesh India Lao PDR Nepal

Source:authors cal ations. cul

Som e countries had m ore than one poverty l For instance, ifthey had separate ine. poverty l ines f urban and ruralareas, then the l or owest poverty l was chosen. The correct ine procedure woul have been to use the weighted average poverty l with a weight proportional d ine to the popul ation ofthe urban and ruralareas. The nationalpoverty l ines so sel ected were constructed around the m id 1980s. Since then, m any countries have revised their l ocalpoverty l ines ( and som e have even changed the m ethodol ogy) This im pl that the $1 a day poverty l constructed on the basis ofm id . ies ine 1980s nationalpoverty l ines m ay not be appl icabl in the new m ilennium . e l The $1 a day l is said to be representative ofthe poverty l ine ines f ound am ong l owincom e countries. I is obvious that this l shoul have been derived f t ine d rom a sam pl ofl e owincom e countries. But the Banks sam pl of33 countries consisted ofonl 10 l e y ow-incom e countries. I incl t uded m any rich industrial ized countries such as Japan, A ustral W est ia, Germ any, Bel gium , Canada and U nited States, which do not have absol poverty. They ute generaly use rel l ative poverty l ines. The W orl Bank attem pted to derive the internationalpoverty l by f d ine itting a crosscountry sem i-l f og unction, rel ating a countrys poverty l with its m ean per capita private ine consum ption, both in 1985 PPP dolars. Since econom etric anal f ed to yiel a reasonabl l ysis ail d e

Nanak Kakwani and Hyun H. Son

internationalpoverty l the Bank then decided to determ ine the poverty l by eyebaling ine, ine l the scatter pl of this equation. This eyebaling m ethod gave rise to the poverty l of ot l ine $31 per m onth, which becam e wel-known as the $1 a day poverty l The Banks cl l ine. aim that $1 a day is representative ofpoverty l ines found am ong l ow-incom e countries has a very weak foundation.2

4 CH ANGI TH E PPP BASE FROM 1985 TO 1993 NG


I the l 1990s, the W orl Bank rel n ate d eased the 1993 PPP exchange rates, which are superior to the earl ones in term s ofcoverage ofcountries and are based on prices and consum ption ier baskets constructed by the 1993 I nternationalCom parison Proj (CP) The new PPP ect I . conversion rates are now avail e f 110 countries, whil the earl Penn W orl Tabl abl or e ier d es provided the PPP rates f onl 60 countries. or y W hil the new and im proved PPP f 1993 are wel st or com e, they al raise som e questions, so m ainl what shoul be the poverty l at 1993 PPP dolars that is equival to the $1 a day y, d ine l ent poverty l at 1985 PPP dolars? The infation rate in the U SA between 1985 and 1993 was ine l l roughl about 50 percent ( y about 5.5 percent per annum ) which m eans that the $1 a day , poverty l in 1985 woul be equival to $1.50 a day in 1993. The W orl Bank did not adopt ine d ent d this poverty l because it was wel above the m edian ofthe 10 l ine l owest poverty l ines in the sam pl of33 countries. The ten countries with the l e owest poverty l ines in 1993 PPP dolars l were Bangl adesh, China, I ndia, I ndonesia, Nepal Pakistan, Tanzania, Thail , and, Tunisia and Zam bia. The m edian poverty l ofthese 10 countries was cal ated to be equalto $1.08 per ine cul person per day. Thus, the W orl Bank adopted $1.08 in 1993 PPP dolars to be its new d l internationalpoverty l which is stil ref ine, l erred to as $1 a day poverty l ine. Many critics have pointed out that the W orl Bank has l d owered the realpoverty l The ine. equival poverty l in 1993 PPP shoul have been $1.50 and not $1.08. The Bank has ent ine d def ended against this alegation by saying that one cannot sim pl adj f infation in the U S l y ust or l between 1985 and 1993 to update the poverty l because there has been a PPP deval ine uation ofpoor countries rel ative to the U S with the switch f rom the 1985 to 1993 based PPP, as a resul ofboth the new price data avail e and better m ethods ofcal ating the PPP exchange t abl cul rates. I this argum ent hol ( or which an adequate expl f ds f anation has not been provided) the , correct procedure woul have been to estim ate the degree ofdeval d uation and then to determ ine the equival poverty l in 1993 PPP. I ent ine nstead, the W orl Bank adopted on an ad d hoc basis the m edian ofthe 10 l owest poverty l ines. Thus, $1.08 in 1993 PPP is not equival ent to the original$1 a day l in 1985 PPP. ine The $1.08 a day is the m edian ofthe poverty l ines ofthe 10 countries with the l owest poverty l ines am ong a sam pl of33 countries. I cannot be regarded as a typicalpoverty l e t ine ofl ow-incom e countries; the countries with the l owest poverty l ines are not necessaril the y countries with the l owest incom es. For instance, Tunisia is a rel ativel rich country with per y capita consum ption of$ 8.00 in 1993 PPP, which is incl uded in these 10 countries. Sim il y, arl I ndonesia and Thail and are not l incom e countries. ow

5 A COM PI LATI OF NEW NATI ON ONAL POVERTY LI NES


A s noted in the previous section, the nationalpoverty l ines com pil by the W orl Bank were ed d constructed around the m id 1980s. Since then m any countries have revised their national

I nternationalPoverty Centre

W orking Paper n 29

poverty l ines ( and som e have even changed the m ethodol ogy) This im pl that the $1 a day . ies poverty l constructed on the basis ofthe m id 1980s l ine ocalpoverty l ines m ay not be appl icabl e in the new m ilennium . I is, theref l t ore, im portant to f an internationalpoverty l that can be ix ine considered representative ofthe poverty l ines f ound am ong l ow-incom e countries in the recent past. To do that, we com pil the nationalpoverty l ed ines f our sam pl of19 l incom e or e ow counties, 15 in Sub-Saharan A f and 4 in A sia. W e obtained the nationalpoverty l rica ines f rom the W orl Banks various poverty assessm ent reports. Most ofthese poverty l d ines were constructed in the m id to l 1990s. W e converted the nationalpoverty l ate ines f rom l ocal currency to PPP dolars using the consum er price indices and PPP exchange rates, given in Tabl 1. l e These l ines are presented in the l col n ofTabl 1. ast um e I is interesting to note that m ost ofthe countries in our sam pl of19 countries in A f t e rica and A sia have poverty l ines that are m uch higher than $1.08 in 1993 PPP. There are onl three y countries, nam el Burkina Faso, I y, ndia and Nigeria, which have l ower poverty l ines than the W orl Bank poverty l Thus, $1.08 a day is not a typicalpoverty l f poor countries. d ine. ine or I our sam pl of19 countries in A f and A sia, the m edian poverty l is cal ated to n e rica ine cul be equalto $1.50 in 1993 PPP. Thus, we woul regard $1.50 to be a typicalpoverty l ofthe d ine l ow-incom e countries. The W orl Banks poverty l of$1.08 appears to be rather l d ine ow. Moreover, the poverty l ines, which the W orl Bank has used, were constructed around the d period in the m id 1980s; our poverty l ines are m ore recent, f rom the 1990s. U sing the typicalpoverty l of$1.50, we com puted the percentage and the num ber of ine poor f the 6 regions based on the W orl Bank cl or d assif ication. These estim ates are ref erred to here as the TPL estim ates. A ccording to the W orl Bank poverty l 1.098 bilion peopl in 2001 d ine, l e l ived in poverty, but the TPL estim ates show that 1.865 bilion peopl l l e ived in poverty in 2001. Thus, there is a wide divergence in gl obalpoverty counts between the W orl Banks estim ates d and our new estim ates based on a m ore typicalpoverty l am ong l ine ow-incom e countries.
TA BLE 2

Percentage and num ber of poor by regi ons i 2001 n


Regions East Asia and Pacific East Europe and Central Asia Latin America and the Caribbean Middle East and North Africa South Asia Sub-Saharan Africa Total World Bank (i) 14.89 3.46 9.96 2.35 31.89 46.38 21.27 TPL (ii) 28.54 8.63 15.68 9.04 56.58 61.82 36.12 World Bank 271.12 16.34 52.21 6.95 439.23 312.57 1098.42 TPL 519.66 40.76 82.19 26.74 779.29 416.63 1865.26 Percentage of poor Number of poor (million)

Source:authors cal ations. cul ( W orl Bank estim ates based on $1.08 per day at 1993 PPP. i) d ( Estim ates based on a typicalpoverty l am ong l ii) ine ow-incom e countries in the l 1990s. ate

Nanak Kakwani and Hyun H. Son

6 PROPOSED M ETH OD OLOGY BASED ON CONSU M ER TH EORY


A poverty l specif the societys m inim um standard ofl ine ies iving to which everybody in that society shoul be entitl Def d ed. ining a m inim um standard ofl iving is not easy. A n obvious way to def it is in term s ofthe util an average person derives by consum ing a given bundl of ine ity e goods and services. Suppose an internationalm inim um standard ofl iving to which everybody shoul be entitl exists and, which is m easured by a util l d ed ity eveldenoted by u*. A nyone * ied f ixed f al or l whose actualenj oym ent ofutil is l than u is identif as poor. I u* is f ity ess countries, then such estim ated poverty counts wil obviousl be internationaly com parabl l y l e. W e m ay def a util f ine ity unction as:
u =u q f qn , r n

( 1)

where q f and q n are the quantity vectors off ood and non-f ood item s ofconsum ption, respectivel r is the average cal y; orie requirem ent ofa country and n is a m easure ofsom e other basic needs in a country. I a country with extrem e cl ate, n m ay be the additional n im non-f ood requirem ent such as expenditure on extra cl othing and heating. evelf al countries, the f or l ood basket and I is evident f t rom ( that ifu is f 1) ixed at the u* l non-f ood baskets, denoted by q f and q n , respectivel wil be dif erent f dif erent countries y, l f or f because the countries dif er with respect to their basic needs as m easured by r and n. I al f f l countries have the sam e basic needs, then there wil be a singl poverty l which wil be l e ine, l internationaly com parabl l e. The util f ity unction given in ( cannot be directl estim ated f 1) y rom househol surveys. d However, we can m ake som e j udgm ent about its m agnitude indirectl I this study, we have y. n devel oped a m odelbased on consum er theory, which hel us to arrive at internationaly ps l com parabl poverty l e ines. W e present a detail discussion ofthe m odelin the A ppendix. ed W ith the hel ofour m odel we have dem onstrated that the cal p , orie cost in f ood PPP dolars is a m onotonicaly increasing f l l unction ofthe util peopl enj This resul im pl ity e oy. t ies Lem m a 1 ( proofofwhich is given in the A ppendix) the . Lemma 1:I the peopl in tw o countries have the sam e cal costin food PPP dolars,then f e orie l they w il enj the sam e l of util l oy evel ity. This l m a has an im portant im pl em ication. I tels us that we can obtain internationaly t l l com parabl f e ood poverty l ines ifwe determ ine the f ood poverty l in each country by using ine a constant cal orie cost in f ood PPP dolars across al countries. Thus, the internationaly l l l com parabl f e ood poverty l wil be obtained by m ul ying a countrys cal ine l tipl orie requirem ent r by the cal orie cost ( ccost*)in f ood PPP dolars. The f l ood poverty l ines so obtained f each or country wil im pl the sam e util l l y ity evelenj oyed by the peopl in dif erent countries. e f O ur m odelpresented in the A ppendix al dem onstrates that we can obtain the non-f so ood poverty l ines by util izing the f ood and totalexpenditure f unctions derived f rom consum er theory. The non-food poverty l ines estim ated in this way woul be com parabl across d e countries because they wil provide the sam e l l evelofutil to individual at the poverty ity s l ines but l iving in dif erent countries. The totalinternationalpoverty l f ines wil be obtained l by sum m ing the f ood and non-f ood poverty l ines. W e describe this procedure using Figure 1.

I nternationalPoverty Centre

W orking Paper n 29

I Figure 1, the horizontalaxis represents util l n ity eveland the verticalaxis represents expenditures. Figure 1 depicts the f ood and totalexpenditure f unctions, which tel us how m uch l expenditure f ood and non-f ood, respectivel wil be required in order f the consum ers to enj y, l or oy a given l evelofutil These expenditure f ity. unctions are m onotonicaly increasing in the util l l ity evel u. C is the point that corresponds to the f ood poverty l on the f ine ood expenditure f unction. Corresponding to point C, we obtain B on the x-axis, which gives the util l ity evelu* that is im pl icit in the f ood poverty l A t this point, consum ers wil enj the m inim um basic standard of ine. l oy l iving, which wil m eet their cal l orie requirem ents. Corresponding to point B on the x-axis, we obtain point D on the totalexpenditure f unction, which gives BD as the totalpoverty l that is consistent with the util l ine ity evelu*. O bviousl then, CD wil be the non-f y l ood poverty l The non-f ine. ood poverty l so obtained ine wil be consistent with standard consum er theory. Note that this m ethod ofcal ating the l cul non-f ood poverty l is equival to Ravalions ( ine ent l 1998)upper poverty l which he derived ine, using a dif erent m ethodol f ogy. Ravalion ( l 1998)al suggested estim ating the non-f so ood poverty l using the idea that ine ifa persons totalincom e is j enough to reach the f ust ood threshol anything that a person d, spends on non-f ood item s wil be considered as basic non-f l ood needs. A ccording to this idea, the non-f ood poverty l is the househol non-f ine ds ood expenditure at which the househol ds totalexpenditure is equalto the f ood poverty l A t this point, the househol incom e is j ine. ds ust suf icient to buy onl the nutritionaly adequate f f y l ood basket so that any expenditure a househol incurs on non-f d ood wil be absol y essential l utel . I the f n igure, E is the point at which the totalexpenditure is equalto the f ood poverty l ine. A t this point, FE wil be the non-f l ood poverty l which wil al ine, l ways be l than CD. The ess non-food poverty l corresponds to the util l ine ity evel_u*, whereas the f ood poverty l ine * ood and non-f ood poverty l ines do not im pl y corresponds to the util l ity evelu . Thus, the f the sam e l evelofconsum er util Thus, we say that Ravalions m ethod is inconsistent with ity. l standard util theory. W e recom m end using CD as the non-f ity ood poverty l ine.
FI RE 1 GU

D eterm i nati of Non-food Poverty Li on ne

Total expenditure function

D Food expenditure function

E C Food poverty line O F A B

* _u

* u

Utility

Nanak Kakwani and Hyun H. Son

7 ESTI ATES OF A NU TRI ON BASED I M TI NTERNATI ONAL FOOD POVERTY LI NE


7.1 CA LO RI REQ U I E REMENTS
W e use the basic needs approach to construct the f ood poverty l This approach requires ine. cal oric requirem ent ( needs)ofindividual I m ust be em phasized that these requirem ents or s. t depend on severalf actors such as age, sex, body weight, cl atic conditions and activity l s. im evel The U nited Nations Food and A gricul turalO rganization ( O )is m ainl concerned with the FA y m easurem ent ofunder nutrition in the worl so it provides cal d, orie requirem ents f dif erent or f countries in the worl W e com pil the average cal d. ed orie requirem ent ( individualper day) per f our sam pl of19 countries. The f col n in Tabl 4 presents the cal or e irst um e orie requirem ents per person per day.

7.2 CA LO RI CO ST A ND THE FO O D PO VERTY LI C NE


The f ood poverty l ofa househol can be determ ined by m ul ying the cal ine d tipl orie requirem ent ofthe househol by the cal d orie cost. I was dem onstrated in Section 2 that the cal t orie cost has a m onotonic rel ationship with the util l ; the higher the util l , the greater the cal ity evel ity evel orie cost. This m eans that ifwe know the cal orie cost ofthe poor, we woul al know the util d so ity l evelenj oyed by the poor. How do we f out the cal ind orie cost ofthe poor when we do not know who the poor are? W e used the f l olowing procedure: First, we want to determ ine how the cal orie cost varies with the standard ofl iving m easured by the househol per capita consum ption. To do so, we divided the popul ds ation into f quintil by ranking the househol by their per capita f ive es ds inalconsum ption. The f inal consum ption is def ined as a sum ofal net cash and in-kind expenditures. W e constructed a l f ood basket f each quintil or e. Househol surveys generaly provide us with expenditures and quantities ofeach f d l ood item consum ed by househol W e determ ined the average expenditure and average quantity ds. ofeach f ood item in each ofthe f quintil 3 Thus, we constructed f f ive es. ive ood baskets ( one f or each quintil . W e were al abl to acquire the quantity to cal e) so e orie conversions, which provided us with the totalcal ories obtained f rom these baskets. Since we know the totalf ood expenditure f each quintil and the totalcal or e ories obtained f rom the f ood basket f each quintil we or e, coul cal ate the cal d cul orie cost f each quintil by dividing the totalf or e ood expenditure by the totalcal ories. The cal orie costs f Bangl or adesh are presented in Tabl 3. e
TA BLE 3

Cal c cost per 1,000 ki o cal es i Bangl ori l ori n adesh


Quintiles Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Source:authors cal ations. cul Takka in 2000 7.62 9.78 11.47 13.81 19.15 1993 PPP dollars 0.41 0.52 0.61 0.74 1.02

10

I nternationalPoverty Centre

W orking Paper n 29

I Bangl n adesh, peopl bel e onging to the f quintil spend 7.62 Takka on f irst e ood in order to obtain 1,000 cal ories. The cal orie requirem ent f Bangl or adesh is 2,080 cal ories per person per day, so the average f ood poverty l f Bangl ine or adesh woul be 7.62 2080 /1000 , which is d equalto 15.85 Takka per person per day. This is the f ood poverty l f Bangl ine or adesh. I is noted that the cal t orie cost increases as we go f rom a l ower quintil to a higher e quintil This resul hol f al countries. This supports the concl e. t ds or l usion f rom our m odelthat the cal orie cost is an increasing f unction ofthe standard ofl iving ( easured by the per capita m totalconsum ption)4 . I order to m ake a com parison ofcal n orie costs across countries, we need to convert the cal orie costs in l ocalcurrency to PPP dolars. This is easil done util l y izing the CPIand PPP exchange rates given in Tabl 1. The cal e oric cost in Bangl adesh f the peopl in the f or e irst quintil is 41 cents in 1993 PPP dolars. This gives the f e l ood poverty l f Bangl ine or adesh in 1993 PPP dolars as equalto 0.41 2080 /1000 , which is 85 cents per person per day. I these l n cal ations, we have m ade a j cul udgm ent that the f quintil is a reasonabl ref irst e e erence group. I is essentialto m ake such a j t udgm ent about the ref erence group. U sing the f quintil as a irst e ref erence group im pl that the f ies ood poverty l of85 cents per person per day is the ine absol m inim um f ute ood requirem ent in Bangl adesh bel which m eeting the basic ow nutritionalrequirem ent is not possibl e.
TA BLE 4

New nutri on based poverty l ne at 1993 PPP dol ars ti i l


Countries Burundi Burkina Faso Ivory Coast Cameroon Ethiopia Ghana Guinea Gambia Kenya Madagascar Mozambique Malawi Nigeria Uganda Zambia Bangladesh India Lao PDR Nepal Median Calorie requirement* 2,070 2,080 2,170 2,170 2,000 2,170 2,130 2,160 2,140 2,110 2,200 2,080 2,120 2,050 2,110 2,080 2,140 2,000 2,120 2,120 Poverty line in 1993 PPP dollars Food 0.85 0.85 0.89 0.89 0.82 0.89 0.87 0.89 0.88 0.87 0.90 0.85 0.87 0.84 0.87 0.85 0.88 0.82 0.87 0.87 Non-food 0.20 0.50 0.74 0.74 0.39 0.52 0.68 0.34 0.30 0.21 0.35 0.30 0.27 0.63 0.27 0.38 0.41 0.28 0.34 0.35 Total 1.05 1.35 1.63 1.63 1.21 1.41 1.56 1.22 1.18 1.07 1.26 1.16 1.14 1.47 1.14 1.23 1.29 1.10 1.21 1.22

Source:authors cal ations. cul Note:* Per capita per day.

Nanak Kakwani and Hyun H. Son

11

I the construction off n ood poverty l ines f dif erent countries, our m odelsuggested or f that we shoul use the sam e cal d orie cost in PPP dolars f al countries. This requirem ent wil l or l l ensure that the m inim um standard ofl iving rem ains the sam e f al countries, which im pl or l ies that the f ood poverty l wil be internationaly com parabl ine l l e. I our obj f ective is to m ake an internationalcom parison ofpoverty rates, then we shoul d use a cal orie cost that is typicalofthe whol worl This woul require cal ating the cal e d. d cul orie cost f al countries in the worl and cal ating the m edian cal or l d cul orie cost. This task, cl y, is earl beyond the scope ofthe present study. I our view, this is al very unnecessary. W e can use n so the cal orie cost ofa typicall ow-incom e country. I our study, we sel n ected Bangl adesh as an anchor country. This sel ection was partl m otivated by the f that we did not have detail y act ed f ood baskets f other l or ow-incom e tropicalcountries. Thus, the f ood poverty l ines in 1993 PPP dolars f the 19 countries presented in Tabl 4 were com puted by m ul ying the cal l or e tipl oric requirem ents ofeach country by the cal oric cost of41 cents per 1,000 cal ories.

8 ESTI ATES OF GLOBAL H U NGER M


W e m ay def a househol to be suf ering f ine d f rom hunger ifit does not have enough incom e to be abl to m eet its basic f e ood needs. Since the f ood poverty l determ ines the m inim um ine basic f ood needs ofthe househol we can say that the househol suf ers f d, d f rom hunger ifits per capita totalconsum ption is l than the f ess ood poverty l Further, we assum e that ifa househol ine. d suf ers f f rom hunger, then al its m em bers al suf er f l so f rom extrem e poverty. U nder these assum ptions, we can cal ate the percentage ofpopul cul ation that suf ers f f rom hunger. The m edian food poverty l is 87 cents per person per day ( 1993 PPP dolars) W e ine in l . appl this poverty l to m easure the percentage ofpeopl suffering f ied ine e rom hunger. The estim ates are presented in Tabl 5. I is noted that 13.28 percent ofthe worl popul e t d ation ( 685.78 m ilion peopl suff l e) ered f rom hunger in 2001. The m aj ority ofpeopl suffering f e rom hunger are residing in South A sia and A f rica.
TA BLE 5

Percentage and num ber of peopl sufferi from hunger i 2001 e ng n


Regions East Asia and Pacific East Europe and Central Asia Latin America and the Caribbean Middle East and North Africa South Asia Sub-Saharan Africa Total Source:authors cal ations. cul Percentage of people (%) 7.99 1.58 6.4 0.93 18.02 36.85 13.28 Number of people (million) 145.48 7.46 33.55 2.75 248.19 248.34 685.78

9 NON-FOOD POVERTY LI NE
I Section 6, we dem onstrated that ifwe know the f n ood poverty l we can util consum er ine, ize theory to estim ate the non-f ood poverty l which takes account ofcountry specif non-f ine, ic ood basic needs.

12

I nternationalPoverty Centre

W orking Paper n 29

The basic idea ofconsum er theory is that we determ ine the non-food poverty at the point where per capita househol food consum ption is equalto the food poverty l W e d ine. have util ized the folowing non-param etric approach to cal ate the non-food poverty l l cul ine 5 for each country. ( i) First, cal ate the ratio ofa househol per capita f cul ds ood expenditure to the f ood poverty l m ul ied by 100. This ratio wil be equalto 100 when the househol ine tipl l ds per capita f ood expenditure is equalto the househol per capita f ds ood poverty l ine.

( ii) A rrange the househol in ascending order according to the f ds ood poverty l ratio ine ( ( )using the househol survey data. in i) d ( iii) Sel the househol whose f ect ds ood-poverty l ratio l between 95 and 105.6 ine ies ( iv) Cal ate the average non-f cul ood poverty l f the individual bel ine or s onging to these househol ds.

The above procedure provided the non-f ood poverty l f each country in l ine or ocal currency. To m ake internationalcom parisons ofthe non-f ood poverty l ines, we needed to convert them to PPP dolars. This was easil done using the consum er price indices and PPP l y exchange rates, which are given in Tabl 1. The non-f e ood poverty l ines in 1993 PPP dolars are l presented in Tabl 4. e

10 TOTAL POVERTY LI AND POVERTY RATES NE


The totalpoverty l ines f 19 countries were obtained by adding up the respective f or ood and non-f ood poverty l ines in 1993 PPP dolars. The m edian poverty l am ong 19 l l ine ow- incom e countries is cal ated as equalto $1.22 (per person per day)varying f cul rom $1.05 in Burundi to $1.63 in the I vory Coast.
TA BLE 6

Percentage and num ber of poor usi nutri on based poverty l ne, 2001 ng ti i
Regions East Asia and Pacific East Europe and Central Asia Latin America and the Caribbean Middle East and North Africa South Asia Sub-Saharan Africa Total Source:authors cal ations. cul Percentage of people (%) 19.23 4.96 11.58 4.09 41.13 52.27 26.43 Number of people (million) 350.14 23.42 60.70 12.10 566.50 352.26 1,365.13

W e have util ized this new internationalyardstick to com pute gl obalpoverty counts. These estim ates are presented in Tabl 6. I is noted that 26.43 percent ofthe worl popul e t d ation ( 1.365 bilion peopl l l e) ived in absol poverty in 2001. The m aj ute ority ofpoor are l iving in South A sia.

Nanak Kakwani and Hyun H. Son

13

11 CONCLU D I REM ARK S NG


Every society has its own views on what constitutes a m inim um standard ofl iving. Strictl y speaking, we shoul not be abl to m ake cross-country com parisons ofpoverty rates since d e it woul be virtualy im possibl to agree on a com m on poverty basket that is acceptabl in d l e e every country. I spite ofthis, gl n obalestim ates do pl an im portant rol in m onitoring the l ay e eveland change in poverty around the worl They can be used as a powerf toolto heighten publ d. ul ic awareness about the need to f ight poverty and achieve the Milennium Devel l opm ent Goal s. The ef orts m ade by the W orl Bank to produce gl f d obalpoverty counts based on an internationaly com parabl threshol m ust be appl l e d auded. U nf ortunatel the Bank has paid y, l e attention to im proving the m ethodol ittl ogy f constructing such a threshol I this paper, or d. n we have dem onstrated how shaky the W orl Bank m ethodol d ogy has been. The national poverty l ines com pil f the purpose ofdeterm ining the originalinternationalpoverty l ed or ine were constructed around the m id-1980s. Many countries have revised them and som e have even changed the m ethodol ogy for their cal ation. I is, therefore, im portant to f an cul t ix internationalpoverty l that is representative ofthe poverty l ine ines ofl ow-incom e countries in the recent past. I this study, we have attem pted to address this deficiency by com pil the national n ing poverty l ines for a sam pl of19 l e ow-incom e countries, 15 in Sub-Saharan A f rica and 4 in A sia. W e arrived at a poverty l of$1.50 com pared to the W orl Banks figure of$1.08. U sing this ine d poverty l we estim ated that there were nearl 1.9 bilion peopl l ine, y l e iving in poverty around the worl in 2001. This f d igure is considerabl higher than the 1.1 bilion poor reported by the y l Bank. O ur estim ates are based on a m ore typicalpoverty l prevail in the m id 1990s ine ing am ong the l ow-incom e countries. However, the m ain obj ective ofthis study has been to com pute the internationalpoverty threshol based on the f d ood requirem ent to ensure adequate cal orie intake f the worl or ds poorest. The study proposes a new m ethodol ogy based on consum er theory to provide a cal oric based internationalpoverty threshol U sing this m ethodol d. ogy, we com puted the internationalpoverty l as being equalto $1.22. A ccording to this new yardstick, al ost ine m 1.37 bilion peopl were poor around the worl in 2001. l e d I this study, we have al provided gl n so obalestim ates ofhunger. A person is def ined as suf ering f f rom hunger ifhe or she does not have enough incom e to be abl to m eet his or her e basic f ood needs. A ccording to this def inition, 687 m ilion peopl are suf ering f l e f rom hunger. These estim ates wil cl y be hel ulto internationaldonor agencies that are concerned l earl pf about achieving the Milennium Devel l opm ent Goalofreducing worl hunger. d O ur cal ations cl y suggest that an ef ort m ust be m ade to im prove the m ethodol cul earl f ogy f estim ating the num ber and percentage ofpoor peopl around the worl I f the W orl or e d. n act, d Banks poverty counts are in need ofserious adj ustm ent ifthey are to m ore accuratel refect y l the situation ofthe worl poorest in the new m ilennium . ds l

14

I nternationalPoverty Centre

W orking Paper n 29

APPEND I :PROPOSED M OD EL X
This appendix presents the m odelthat hel to arrive at internationaly com parabl poverty l ps l e ines. Suppose p f and p n are the price vectors off ood and non-f ood item s ofconsum ption, respectivel then, using the conventionaltreatm ent ofconsum er choice, we m axim ize util y; ity f unction u= u
q f qn , r n

( 1) A

subj to budget constraint ect


p f q f + pn qn x

( 2) A

where x is the totalexpenditure or incom e that is avail e to the consum er. I is assum ed that abl t al countries have the sam e util as ( 1)but dif erent r and n. l ity A f This m axim ization procedure yiel the f ds ood and non-f ood dem and f unctions as
q f = rg f x, rp f , np n

( 3) A

and
q n = ng n x, rp f , np n

( 4) A

respectivel These equations are the Marshalian dem and f y. l unctions ( Marshal, 1930) l . Substituting ( 3)and ( 4)into ( 1)yiel the expenditure f A A A ds unction
x = e u , rp f , np n

( 5) A

which is the m inim um cost ofbuying the individualutil u at given f ity ood and non-f ood prices. Further, substituting ( 5)into ( 3)and ( 4)yiel the Hicksian f A A A ds ood and non-f ood dem and equations ( Hicks, 1957) :
q f = rg f u , rp f , np n

) )

( 6) A

and
q n = ng n u , rp f , np n

( 7) A

respectivel y. The f ood and non-f ood poverty l ines are then obtained by substituting u = u* in ( 6)and A ( 7) respectivel as A , y,

Nanak Kakwani and Hyun H. Son

15

F = p f q f = rp f g f u*, rp f , np n

) )

( 8) A

and
NF = p n q n = np n g n u * , rp f , np n

( 9) A

Equations ( 8)and ( 9)give the f A A ood and non-f ood poverty l ines at the point where individual ofthe country concerned enj the sam e l s oy evelutil u*. These l ity ines are internationaly com parabl because they m aintain the sam e util l l e ity evelacross countries. The olowing sol ution is proposed. probl is:How do we determ ine u*? The f l em The f ood poverty l shoul satisf the requirem ent that cal ine d y orie intake is equalto cal orie requirem ent. Suppose c is the vector that converts a f ood quantity vector qf into cal ories. c is f ixed f each country depending on the kind off or ood the countys popul ation is consum ing. c qf is the num ber ofcal ories that are obtained f rom the f ood basket qf , which shoul be d equalto the cal orie requirem ent r. Thus, using ( 6) we obtain A ,
c.g f ( u * , rp f , np n ) = 1

( 10) A

I this equation is to hol f al exogenousl determ ined val ofr, pf , n and pn, then the f d or l y ues * f unction g f ( u , rp f , np n ) shoul not contain rpf and npn as its argum ents, that is, it shoul d d * depend onl on u . The f y ood poverty l in ( 8)wil then be given by ine A l
F = p f q f = rp f g f u *

( )

( 11) A

Since the f ood poverty l can al be written as the product ofcal ine so orie requirem ent and cal orie cost ( which is the expenditure on f ood per cal orie) ( 11)im m ediatel gives the cal , A y orie cost f unction as
c cos t = p f g f u *

( )

( 12) A

Thus, the cal orie cost ofa country depends on two f actors, f ood prices and the util l ity evelu*. The cal orie cost in ( 12)is in l A ocalcurrency ofthe country. The PPP converts prices in l ocalcurrency ofa country to prices in U S dolars. Suppose l p is the vector ofinternationalf ood prices, the PPP conversion rate k is given by
* f

p f = kp * f
which on substituting in ( 12)gives A
c cos t * = p*f g f ( u * )

( 13) A

( 14) A

where c cos t * = c cos t / k is the cal orie cost in f ood PPP dolars. Since p *f is the international l * f ood price ( which is sam e f al countries)and g f u is a m onotonicaly increasing f or l l unction rom ( 14)the cal A orie cost in f ood PPP dolars is a m onotonicaly increasing f l l unction of ofu*, f the util peopl enj This proves Lem m a 1 given in Section 2. ity e oy.

( )

16

I nternationalPoverty Centre

W orking Paper n 29

Lemma 1:I the peopl in tw o countries have the sam e cal costin food PPP dolars,then f e orie l they w il enj the sam e l of util l oy evel ity. A s pointed out in Section 2, Lem m a 1 tels us that we can obtain internationaly l l com parabl f e ood poverty l ines ifwe determ ine the f ood poverty l in each country by ine using a constant cal orie cost in f ood PPP dolars across al countries. Thus, the internationaly l l l * com parabl food poverty l ( e ine denoting by F )wil be obtained by m ul ying a countrys l tipl ood PPP dolars. l cal orie requirem ent r by the cal orie cost ( ccost*)in f Given F*, which is in PPP dolars and the f l ood PPP exchange rates, we can determ ine the f ood poverty l f each country in l ine or ocalcurrency, which we denote by F. Substituting F into ity evelenj oyed by peopl at the f e ood poverty l ine ( 11) we can sol f u*, which is the util l A , ve or and is f ixed across countries. Further, substituting u* into ( 9) we can sol f the non-f A , ve or ood poverty l NF, which wil be the non-f ine, l ood poverty l in l ine ocalcurrency. U sing the non-f ood PPP conversions, we obtain the non-food poverty l in internationaldolars. The non-food ine l poverty l ines estim ated this way woul be com parabl across countries because they provide d e the sam e l evelofutil to individual at the poverty l ity s ines but l iving in dif erent countries. The f totalinternationalpoverty l ines are obtained by sum m ing the f ood and non-f ood poverty l ines.

Nanak Kakwani and Hyun H. Son

17

REFERENCES
Chen, S. and M. Ravalion ( l 2001) How Did the W orl Poorest Fare in the 1990s Review of , ds I ncom e and W eal Series 47, No. 3, Sept. pp. 283-300. th, Chen, S., Datt, G. and M. Ravalion ( l 1994) I Poverty I , s ncreasing or Decreasing in the Devel oping W orl Review of I d ncom e and W eal Series 40, pp. 359-76. th, Deaton, A . ( 2001) Counting the W orl Poor:Probl s and Possibl Sol , ds em e utions W orl Bank d Research O bserver, Vol 16, No. 2, pp. 125-147. . Hicks, J. R. ( 1946) Val and Capital Cl , ue , arendon Press:O xf ord. Lipton, M. (1988), The poor and the poorest, Discussion Paper No. 25, W ashington D.C.: W orl Bank. d Marshal, A . ( l 1930) Principl of Econom ics, 8th ed., Macm ilan & Co., London. , es l Rao, P. ( 2003) I , ntegration ofCPIand PPP:Methodol ogicalissues, f easibil and ity recom m endations, Paper presented at the Joint W orl Bank-O ECD Sem inar on Purchasing d Power Parities, W ashington, D.C. Ravalion, M. ( l 1998) Poverty Lines in Theory and Practice, LSMS W orking Paper No. 133, The , W orl Bank, W ashington, D.C. d Ravalion, M. ( l 2001) Com m ents on Counting the W orl Poor by A ngus Deaton, W orl Bank , ds d Research O bserver, Vol 16, No. 2, pp. 149-156. . Ravalion, M. and S. Chen ( l 1997) W hat Can New Survey Data Tel U s about Recent , l Distribution and Poverty, W orl Bank Econom ic Review , Vol 1 ( , pp. 357-82. d . 2) Ravalion, M., Datt, G. and D. van de W ale ( l l 1991) Q uantif , ying A bsol Poverty in the W orl ute d, Review of I ncom e and W eal 37, pp. 345-61. th, Sen, A . ( 1999) D evel , opm entas a Freedom , O xf U niversity Press:O xf ord ord. Srinivasan, T. N. ( 2001) Com m ents on Counting the W orl Poor by A ngus Deaton, W orl , ds d Bank Research O bserver, Vol 16, No. 2, pp.157-168. . W orl Bank ( d 1990) W orl Devel , d opm ent Report 1990, O xf U niversity Press:New York. ord .

NOTES
1. There are two m ain l itations to PPP. Firstl as PPP com parisons weigh prices by the share in worl consum ption, im y, d worl dwide rel ative prices cl y track rel osel ative prices in the U nited States, W estern Europe, and Japan. U sing such rich country prices wil inf ate poor country incom es ( l l and vice versa) A s a resul this underestim ates worl poverty. Thus, if . t, d onl f y ood prices are used, which our m ethodol ogy requires, this wil reduce the bias. l 2. Defenders of a 1$ a day poverty l argue that one of the m ost im portant m erits of this approach is sim pl ine icity. The argum ent is that if the def inition of poverty l ines becom es too com pl then the debate on poverty wil revol ex, l ve around technical aspects of how to def the poor and not on the m easures and pol ine icies necessary to l t them f if rom poverty. O n the other hand, pol ef orts f icy f ocused on the wrong target, though sim pl m ay be sel-def e, f eating. 3. Q uintil were constructed using househol per capita consum ption expenditure. es d 4. A n intuitive interpretation of this resul is that richer househol consum e their cal t ds ories f rom m ore expensive food stuf s, f exam pl substituting m eat f potatoes. The richer househol m ay al be paying higher prices f the sam e f or e or ds so or f ood item s because they generaly buy higher qual f l ity ood item s ( or exam pl a superior qual ofrice) f e, ity . 5. W e are suggesting a non-param etric approach. To im pl ent this m ethodol em ogy, we woul require unit record data on d househol expenditure surveys. O ne can al use a regression m odel( proposed by Ravalion 1998) W e bel d so as l . ieve that the non-param etric approach is m ore robust. 6. W e shoul sel househol whose per capita f d ect ds ood expenditure is equal to the f ood poverty l ine, which m eans we shoul sel the househol at the point where the househol food-poverty l ratio is equal to 100. Since it is d ect ds ds ine im possibl to cal ate this ratio for a specific point, it is reasonabl to sel a range in the neighbourhood of 100. e cul e ect W e sel ected a range ofthe f ood-poverty l ratios l ine ying between 95 and 105.

International Poverty Centre SBS Ed. BNDES,10o andar 70076 900 Brasilia DF Brazil povertycentre@undp-povertycentre.org www.undp.org/povertycentre Telephone +55 61 2105 5000 Fax +55 61 2105 5001

Вам также может понравиться