World Bank Reprint Series: Number 1,29 Pravin Visaria Demographic Factors and the Distribution of Income: Some Issues Reprinted with permission from Econiomnic and Deinograplhic Clhange: Issues for the 1980s. Proceedings of the Conference (Liege: IUSSI, 1979), vol. 1, pp. 289-320 DEMOGRAPHIC FACTORS AND THE DISTRIBUTION OF INCOME: SOME ISSUES Pravin VISARIA Te i*World Bxnk, D)eveloprinant Reseadrt h Center W&Sa1irqtv)rx D.C. A more equitable distribution of income and wealth has been a major goal of many of the development plans of developing countries as well as the United Nations Declaration proclaiming the Second Development Dec- ade. Partly as a result, most attempts to evaluate the economic devel- opment of developing countries now seek to provide some evidence, usually based on sample surveys (but sometimes on the basis of questions canvassed in the population or economic census), on the distribution of income. A recent comppendium of these data shows one or more distriLbu- tions for 81 countries, with an indication of the coverage (whether thle pertain to the entire country, rural or urban areas, the agriculttural or the non-agricultural sector or a selected metropolis or city) and the population or the unit of aggregation (households, persons, ;workcr-, in- come recipients or the economicallv active population)..L/ The value of / Jain, Shail, .':' za t7 :Lr. 1'.c: A C,pqi7.zt-. 7 of Pfatz Washington, D.C., World Bank, 1975. *The data used for preparing this paper were collected by ilhe iinrO:ni of Economics and Statistics in Gujarat and Maharashtra states of India, the Research Department of the Nepal Rastra Bank, the Department of Census and Statistics in Sri Lanka and the Bureau of Statistics in Taiwan. I am beholden to the Directors of these agencies who have permitted and encouraged me to use their data for the joint ESCAP-IBTW Research Pro- ject on Income Distribution in Asia. The compilation of tables for two Indian states presented in the paper has required the cooperation of a large staff working at the Gujarat Computer Centre in Candhinagar, India, under the leadership of P.B. Buch. Others who have helped in the statistical work include Robert E. Sterrett, Jr., R. Murti Pemmarazu, Shyamalendu Pal, Cynthlia ITwa, Shtail Jain and Pramilla Burger. I am deeply grateful to all of them. I am indebted to Montek S. Ahluwalia, T.N. Srinivasan, Bryan Boulier, V.V. Bhanoji Rao and Oey Astra Meesook for their comments on a prelim- inary draft of this paper, Tie responsibility for all errors is solely mine. The views expressed in this 1;iq.r are those of the author and should not be ascribed to the World "lank or to its affiliatei' institutions. 289 thliS gLWill' volume of information is limited, however, by inadequate attention tL) the quality and comparability of the data for different counitries (as also for some countries over time). The World Bank lhas, therefore, undertaken two fairly large research projects to evaluate the data available from some of the recent household surveys in selected countries of Asina' and Latin America.-V This paper reports some re- sults of an on-going analysis of the data collected in (a) a labor force and consumer expenditure survey undertaken as a part of the annual Nationial Sample Survey (NSS) dluritng October 1972-September 1973 in Gujarat and Maharashtra states of IndiaL3; (b) a household btudget eurvey undertaken during 1973-74 and 1974-75 in 18 "towrns" or develop- nent cenLers of Nepal; (c) a socio-economic survey of Sri Lanka con- ducted hetween November 1969 and October 1970; (d) the income and ex- penidituire surveys of Taiwani conducted between 1968 and '-974. hlere is ";NI'd reason to believe that thie results based on these data have a wider validity. fliE DATA A brief description of the conceptual framework used to collect data on expenditure and/or income and labor force in different survevs is given in Annex 1. For Nepal, we have chosen to use only the expend- itoire data because of unresolved problems in reconciling the data on Ihu OSClld income from different sources. hle Indian NSS does not col- lect data on income, whereas in Sri Lanka and Taiwan the same survevs have coI'L, tted data both on expenditure and income.1' The expenditure data i.ndiicate disparities Lin consumption, whiichi are normally less than the iltoqul.-i itv of income or wealth; but expenditure or consumption is easier to recal 1 or report; and the survey data on them are I ikel. to bo more reliable. Also, oonsumption miglht be more closelv related to what is termed 'permanent income', tak[ng account of the temporal fluc- tuations in the income stream. ti: .KAlKIN CRITERDIO: AnI important first step in the analysis is to establish thie zippro- l'Li.ite criterion fo': r-iikinrg the populat ion (or households) according .\ nnajor criterion used to select the surveys of Asian couintries for thle project was the simultaneous collection of data on both labor force ciiar-icteristics and income or expenditure. This criterion was adopted because of ian evnpli,is on the analvsis of the socio-economic characteristics of households with different levels of income or expenditure. 2/ Thl projects have been tundertaken jointly withi F;SCAP (the Economic a.nd ;o i.lt Coirmission for Asia and the Pacific) and ECT.AJ (the F1o- nomic. to'minissoio for Latin A3nerica), respectively. 3 'Al tib'ughi the survev was conducted in the entire coountry, we have liad acvcess to dlata for only the two states. FIrthler, thue detailed data oin the Pattern of consumption 1xpoiiU1 Lurre are not available. 4+i it hias Inot been possible, howevor, to examine measures of inequ:1lity in expenditure in Taiwan iii this paper, . i5 to their level of income (or expenditure). In a majority of reports on income distribution, the ranking seems to be according to the size of total household income (or expenditure).1/ Only in a few countries like India and Indonesia are the surveyed households ranked according to the per capita expenditure or income (PCE or PCY),2/ The recent United Nations guidelines on the income distribution statistics recommend tabu- latior.s for income distribution mainly in terms of the (pre-tax) total househiold income and only two general tables on a per capita basis for comparstive purposes.,/ Perhaps it is not recognised adequately that the two criteria lead to a significantlv different ranking of the house- holds, althlough this is not evident from the Gini coefficients or other measures of the concentration of income. Table 1 shows three alternative indices of inequality-/ of expendi- ture or income for different countries or states. The indices for Sri Lanka conform to the expectation that inequality of expenditure tinds to be less than that if income. The inequality indices based on the ranking of individuals according to the PCE or PCY of the household are significantly lower than those based on the di...tribution of households according to the total household expenditure (income) (THE or THY). 1/ One can speculate about the reasons for this practice. When the data sources collect information on income on the basis of one (or a few) question(s) in terms of rather broad class intervals, tt is not pos- sible to estimate per capita income with any reasonable precision. If the mid-point of a class interval is taken as the actual income of all households in that class, the hIouseLholds will be ranked essen- tially in terms of their household size (obviously with an inverse relationslhip between the latter and the estimated per capita income). The simple solution is to present the data for households ranked according to their total household income. 2/ In Indonesia, the published tables for the socio-economic survey (SU'SEtNAS) of 1976 relate to only the total household expenditure (although the tables according to per capita expenditure are avail- able). 'Tlis is a departure from the procedure adopted in the report on the 1969-70 survev. 3/ Utnited Nations, Pe)' L.'-z:[ kA.d de"nes o .n t.'IotirsO of thoe Notri- ;' '*. "' ,n(71c-':', , tioizz". +aniR 21 Ac,?,Omulaton of Hotiseholds :C-\- 1Ng M No. 61), Now York, 1977, pp. 58-59. Since the household is generally the production (or consumption) unit and the specific contribution (share) of individual members or workers is difficult to identify or measure, it is only natural that it should alsn be the basis of the reporting of income (or expenditure/consumption). That is particularlv true of the family farms and enterprises which account for a substantial proportion of the population in the rura;l areas of developing countries (and for consumption or expenditure, it is true of all societies). 4/ See Annex II for a brief outline of the thlree measures. We have chosen to deal withi declines of houser.olds although it is possible to delimit deciles so that eachi decile has an equal proportion of population. The points raised in this paper would hold even if we were to deal with deciles of population, although in that case the decile limits would naLurallv he dLfFeronr. 291 TABLE 1 Indic" of Ineuelicy vith Alternative Laking Criteria RmAoing Criterion' for Unit of Cin1 Ku.ets Etcropy C.oer/State/Ar.e Houjehold. Agbr.S.tloa Coefficient Index Neseure 1. GUJARAT 1972-73 itural Iadividual 0.259 0.194 0.110 HHZ ou.ehold 0.320 0.246 0.169 m / Adul-Cquivalnt 0.262 0.196 0.113 Urban c nadividual 0.261 0.193 0.118 oX Household 0.304 0.225 0.148 AIX a/ Adult-Equivalmnc 0.260 0.193 0.118 2. HIaAAAS8IA 1972-73 ra ?CZ Individua1 0.272 0.201 0.126 t1 H-ouaMhld 0.354 0.263 0.204 AL a/ Adul. -Equivalent 0.273 0.202 0.126 Urb*a rZcz Individual 0.331 0.250 0.177 TE Household 0.371 0.280 0.213 AL/ Adult-Equivelent 0.317 0.247 0.173 3. 2-PAL LIlaeen T.-. ICE Individual 0.312 0.235 0.164 1973-74 TE Household 0.388 0.291 0.24A *A/ Adult-Equivlat 0.305 0.230 0.157 S*ve Toa PCE Individual 0.303 0.229 0.154 197475 mounehold 0.367 0.277 0.219 ALE Adult-Equive.lac 0.294 0.222 0.146 4. SRI LAM PM Tnd4vidual 0.309 0.231 0.156 1969-70 :E Househld 0.330 0.247 0,175 ALE j Adulr-Equivalent 0.312 0.233 0.159 AKE I4 Au Ic-Equivlanc 0.298 0.222 0,146 rcE Inividual 0.364 0.273 0.218 Houaehold 0.381 0.287 0.235 ALT Adult-Equialent 0.366 0.274 0.220 AZT 4/ Adult-Equivale:n 0.350 0.261 0.204 5. TAX2A5 1968 rcr Individu-l 0.321 0.240 0.173 MEr .oushold 0.312 0.267 0.175 ALEY Adl-Equivalent 0.306 0.227 0.159 E Adult-Equivalent 0.319 0.234 0.171 1974 CE' Individual 0.287 0.2.! 0.139 E' uenld 0.301 0.. 23 0.151 AE' b/ Adult-Equivalent 0.294 0.U19 U1.146 AET * Adult-quivalent 0.30q 0.229 0.161 A?CZ: Per capita epnditure of the hou.hold THY. Total expenditure of the houshold ALE: Household expenditure per equivalent dult PCTE: Fer capita incom of the household THE: Total household income AEY: Household intm - per adult equivalent .1/ With veights ised in India by the NSS. bI With veights used by the Bureau of Statistics, Taiwan. Table 2 shows the proportions of households (and the proportion of total population in these households) falling in the same decile and the same quintile according to the two criteria. Less than 20 percent of the households in Gujarat, Maharashtra, Nepal, and Sri Lanka fall in the identical decile according to the PCE and THE criteria. In Sri 292 Lanka and Taiwan, i.t h the ralkinig in termiis of 1PCY and 'TiY, I etweoll t:ti J., percent of the li,ua;.holds ii 1 I in the same decile Tht o *I.us sit i cation of howeihoi.ds ini b roader o rotiipsi of quintiles naturallyv itnreoaso the ,: 'c iienl ti:', in ideLntieal quintiles ar:^ordoilig to the two criteria.,,, -. the t -,4.t group. of households for poverty alovloi,- tiol rams w)o'l.Ld lho I ' rLf lepending on whether the r.ank ini.. ig ini terras of totaL or ; capita consumption or income. ; ' :-r . fact or for tI e dil'I'0E-011CL' in the ranilng ol hou sehe, Ids acrording t ) the two criteria is the di,fferent rolat Lenshlll between (a) IioU.etlhel4i size aild (b) total or per capita expenditure or income. V,s shilo>wn in Table 3, thie iverage. household size rises withi the dec'lte in t,.rms of --.- T " ,- - and it falls witlh decile in terms of l'1JIWY. Ilie di-itrifbitions t5ft idS in terms of THE or TIY overstate the lIt :r of inequality hease they overlook the fact that the low total i II h or experndLtitrel i. associaetd with a lower average household SI ' inter-deLi r -C' fir 1 n hIouselhold size are generall'y s-maller w-;ith the p:er .: : i in' than withl the THE or TIY Pai.Y tlha 1,i not the case, ho-ever, fo.r Ti:-:i1i in 1974 wher;a itnter-dectile variation7l '1. L'' A : z ic crLattr witlh the rnk Ling in t.ermsi P. e1r apt:: v.rI in 5 r!:ome -3 iab i Ai ad . la 2 ccn5z:x III attempt to shed some more liglt on tol chawng- in t r '-itu,holds V by shlowing (i) THE, (ii) PiT; ani the av -i. isul i.le age groulpj ot the IthousEhodld ti. . n' o soon tbllt tilh .iJe'.p;t househldsiiw toiiuY .o Itz t ': oi . ,ernt of hous11e1holds are ooloani'lk dt ''''5 Wt;s n; S to TnH or TIY and also PCIE or PlY. i; tU' hat it -r a 1Y poor 'iouseholds would cover hot woco ' I' it 0' sI Y pooor hou.s,eholds and between 41 . pterLe't (f t op n' the i TCY poor househtolds. i boca n C. hi ea ' . 11 orh- an average size of thie liouz:eho lId-: ill tt. a . rM; fit,7f T/I11TY . The PP(I:/PCY poor Ihouto-+- hoK'i it coldl5 ' - u s. o >'lo iproportion of Tl1Y/THY poor ,oioteh his tI bweL t . t 7of tlth populat ion of T}1'E/ Ill' poor 1X tSimon Kullet s tTits !. in hiis recent paper en "cti of the A:e ri1irC i;.7fl of Income: An xpilor.i .i'rv *~ ~ ~ ~ ~ ~ ~ ~ ~~~~~Vl 25 ""' . 1.,;...w 2.w\t r Pi L-< , ;:. lw) , lalt7 nsi up To'"' 't"' tbn .average size of lioiu-ieliolds and tie , t'le in ter.t; ot il Io'ta X.'pditUr or income is not moInlo- in all. MI . . ens to he the case wiLh iL.al expoenditur.^ .e de. di' t. '< ",ti fr Taiwait in I %(8), 4it tf it 'n of the liead of the househiold is not alway;s m' ' t v w'hat the rk.-p,inden ut:; report. However, thle, - flo:;(i; of t ho tiata oll the relationship to the household heid i'i .',''.!' '. lo '.1 available dlata l'apes for G;u jarat., * .1dL rot'atilva and 1i41 i atz,l l do not. expil .i it-l' idtL1i fv thle household hod nil Other wotd; t1w ho i- t ion on ttht re lationship to tile. head ot h.I-l ha"'; not b-een t .i. no rr ed to the tapes. Ilo-wever, "w oazvr been - s'! ri tl tsa persons tlie their serial number c'.IInil to sue were the heads, t' houeholds TABLE 2 PERCENTAGE OF HOUSEHOLDS IN THE SAMF. DECILE OR QUINTELE ACCORDING TO THE ALTERNATIVE CRITERIA OF PER CAPITA AND TOTAL HOUSEHOLD EXPENDITURC: (OR INCOME) Percentage of Percˇ.ntage of Households in the Households in the Country/State/Area Same Decile Same Quintile 1. GUJARAT 1972-73 Rural 17.4 (16.5) 32.9 (32.0) Urban 17.7 '(16.1) 31.3 (29.0) 2. MAHARASHTRA 1972-73 Rural 15.9 (15.2) 32.1 (31.3) Urban 15.3 (14.7) 26.0 (25,4) 3. NEPAL Eleven Towns 1973-74 14.7 (13.5) 31.3 (29.7) Seven Towns 1974-75 14.5 (12.8) 29.4 (28.1) 4. SRI LANKA 1969-70 19.5 (18.3) 35.9 (34.7) 1969-70* 23.2 (21.9) 39.2 (38.1) 5. TAIWAN 1968* 20.2 (18.3) 36,0 (34.3) 1974* 28.3 (22.7) 43.4 (37.8) *I)uciles of income. Figiires in parentheses show the percentage of total population accounted for by thie hotiseholds in the same decile or quintile. 294 TABLE 3 Average Size of Households in Different Deciles with Alternative Ranking Criteria Rural Rural Urban Urban Nepal Gujarat laharashtra Gujarat Maharashtra 11 Towns 7 Towns Sri Lanka Taiwan Declle of louseholds (1972-73) (1972-73) (1972-73) (1972-73) (1973-74) (1974-75) (1969-70>' (1968)* (1974)* (A) Households Ranked by Per Capita Expenditure (or Income) 1 6.8 6.3 6,7 6.6 6.4 6.3 7.3 7.7 7.1 2 6.6 6.0 6,4 6.3 6.3 6.3 6.9 6.9 6.5 3 6.2 5.6 6.3 6.1 6.1 5.9 6.4 6.6 6.1 4 6.2 5.5 6,0 5.8 5.8 5.9 6.3 6.4 5.8 5. 6.0 5.8 5,8 5.4 5.7 5.2 6.0 6.3 5.4 6 6.1 5.3 5.5 5.1 5.6 5.4 5.6 5.8 5.4 7 5.5 5.2 5,0 4.4 5.0 5.0 5.7 5.9 4.9 8 5.5 4.9 4,2 4.0 5.0 4.7 5.0 5.0 4.5 9 5.0 4.5 3.9 3.2 4.5 4.2 5.0 4.3 4.3 10 4.3 4.1 3.2 2.7 3.5 2.a 4.6 3.6 3.5 All 5.8 5.3 5.3 4.9 5.4 5.2 5.9 5.8 5.3 (B) HJouseholds Ranked by Total Expenditure (or Income) 1 2.4 2.2 1.9 1.7 2.2 2.4 3.6 3.0 3.4 2 4.0 3.5 3.5 2.8 3.4 3.2 4.5 4.8 4.6 3 4.8 4.1 4.3 3.8 4.1 3.8 5.0 5.1 5.0 4 5.1 4.7 4.8 4.5 4.2 4.3 5.5 5.5 5.2 5 5.7 5.1 5.5 5.0 5.0 4.8 5.9 6.2 5.4 6 6.3 5.4 6.0 5.5 5.4 5.3 6.1 6.0 5.6 7 6.6 5.9 6.1 5.9 5.9 5.8 6.4 6.4 5.7 8 7.0 6.4 6.5 6.2 6.a 6.1 6.7 6.7 6.0 9 7.5 7.1 6.8 6.6 7.4 7,2 7.2 6,9 6.1 10 8.9 8.8 7.7 7.4 9.4 8.9 7.7 7.6 6.6 All 5.8 5.3 5.3 4.9 5.4 5,2 5.9 5.8 5.3 Sample Hou.eholds 5,560 5,314 3,528 11,103 4,393 2,254 9,664 2,776 5,256 *Ranking in terms of per capita or total household income. ), to rise witlh the age of the head, at least tp to the age group 40-44; it levels off or declines for older heads. For many of the households headed by persons aged less than 30 or 35, total expenditure is likely to be below the average; but their per capita expenditure tends to be above the average. The decile of these households is altered when the ranking criterion changes from THE to PCE. A similar reclassification is likely to occur if the rankings were in terms of income rather than expenditure. The PCE or PCY ranking has an advantaly if the difforrn1ts in in- come l.vels are to be related to welfare- 'Insofar as both oarnings or income and the size of the household take an inverted f-shape with respect to the age of the household head, the liie cycle effects would probably be less important in difftrentials in PCE or PCY than in THE or THY. DIFFERENTIALS IN AGE COMPOSITION OF THE HOUSEHOLD [JEMBERS Since the size of the household varies with the age or the stage of the life cycle of the household head, it is logical to presume that the age composition of the population in deciles with a higher average household size might be significantly different from the average and that of the upper deciles. Table 4 shows the age-dependenev ratios (defined as the number of persons aged 0-14 and 60 and over -- or 65 and over -- per 1000 persons in the working ages of 15-59 -- or 15-64 --)/ for different deciles in te ms of both the per capita and total household expenditure of income.3J With a ranking in terms of PCE or PCY, the age-dependency ratio varies inversely with decile, virtually monotonically. The variations are caused mainly by differences in the proportien of children aged 0-14, who account for a majoriLt of the populatl,n in dependent ages. In rural areas of Gujarat and Maha- rashtra, the dependency ratios of the bottom and the top deciles of 1/ There may be some economies of scale in consumption which are over- looked in the per capita ranking. However, these economies are un- likely to exist with respect to all connodities and will seldom be so important that an increase in the size of the household would not raise the consumption requirements. 2/ We have chosen age 60 as the upper limit of working ages in Gujarat, Maharashtra, Nepal and Sri Lanka because the public sector employees usually retire before reaching that age. In Taiwan, on tht. other hand, the retirement age is around 65 years, and, therefore, ages 15-64 are taken as the working ages. 3/ The well-known errors of age reporting -- preference for ages ending in 5 or 0 and, to a lesser extent, 2 and 8 -- do not affect our sur- vey data; but the smoothing of age data that is commonly used in the analysis of census data seems hazardous with the rel,zti'c.;. small number of sample households or persons in eachl decile. The broad age groups help to mitigate the influence of these orrors. Besides, the differentials follow a consistent pattern and arur so large thlat the errors cannot be a major contributOry facttr. 296 tAZLrl 4 Age DOeVeoacty Uce .e by D0case in Tferm of Per Capita "A Total IZpeo4itr. or Iarono of the Household DcenelA M.SETU NUEAL z eI DRu1r. a BU1 Urban Rural Urbem ElA"-. Town* Saeu Towas SRI w of PcXlC?/* (1972-73) (1972-73) (1973-74) (1974-75) _1_969-70 (1968) (1974) 1 1185 110 1304 1203 1279 4235 1319 1303 1173 2 1;229 980 1203 1037 1135 1123 1138 1101 996 3 [ 90 r1100 f5 979 1073 1033 1124 877 4 915 914 886 976 88o 5 (907 s08 (960 733 787 815 846 842 761 6 779 812 749 745 650 7 58S6 (618 (901 C555 669 819 707 724 570 628 653 575 597 499 9 707 8-49 810 409 577 592 608 510 439 10 657 388 776 331 417 352 506 404 305 AL. 922 790 1002 762 321 855 831 846 715 Docila in Terse of TEI/TRT* 1 637 684 742 476 615 653 1005 815 870 2 814 715 774 595 750 656 931 94 93Z 3 928 759 938 732 900 708 919 917 923 4 958 781 910 839 775 771 931 1004 892 5 973 845 960 883 9Z1 844 941 1014 862 6 990 921 1018 827 906 945 886 906 804 7 993 834 1038 862 865 920 815 866 674 8 935 815 1076 807 825 942 777 758 621 9 967 792 1056 7" 853 833 723 752 557 10 863 680 1077 633 746 864 632 656 472 A"L 922 790 988 75b 821 855 83L 846 725 * PCX: Per capita expendture of the housewold. PCT: Per capita incoe of che houseold. T: Total expenditure of the household. MT. Total inco.e of the household. ,/Dec1t based on i me dat. Notes: (1) Dependenco ratios for Cujarat, laharaslhtra, Nepal and Sri Lanka show persons aged 0-14 arnd 60 and over pcer 1000 persons aged 1,5-59. (2) Dependency' ratios for Taiwan show persons aged 0-14 and 65 and over per 100( persons aged 15-64. (3) Dependency ratios for all deciles LogE tliUr "or rural and urban har. tshitra differ slighitly because some sample households were excluded durinig cdliting aft r tihe first runs in terms of PCE decile. ,,9-;, l.ouiehol1ds differ by less than a factor of 2:1; but in other areas, die iatio of the two ratios approaclhes (or even exceeds) 3:1. Witli a ranking of houseliolds in terms of total household expondId- ture or income, the inter-decile range of variation in depenletincy ratios is always smaller than that witlh the PCE or PCY ranking . 'T1ereL 1s nio clear relationship between the TIIE or THY decile and t:he depend- enev ratio. In Sri Lanka, the dependency ratio falls with the 'ThIY docile; in rural Maharashtra, it tenrds to rise, In otther data sets, the dtependenicy ratios seem to take soinelthing like an inverted U.' or V shape t,ith THY decile, with varying degrees of approximition. oset differ- ences in tlle relationslhip between the dependency ratios anid thie PCE(PCY) or Tl['1I'1T(Y) deciles reflect clhangus' in the age composition td the heads olt households fal LinF in various deciles when the ranking criterion is al t red . Ii\ 1I o'i oF All 1i r k.tr L VALENTS AND ALTERNATI VE IND CES itY L AL Given the distinct age-related differences in the capacity to work .r earn as well as what Kuznets calls "tlhe volume of vo|d.s required to '1!'i1` whatever mav be considered acceptable or warranted needs",1/ - -t,' *7.ei in per capita income or consumption do not seem 0ood indi- ,ator.: of differences in welfare. SoMe form of standardization of aiousehold . mbers inta adult equivalents seems necessary. The FAt, the Wi0, as also some countries lhave aittempted1 to identifv the sex- . -.p.i ti: differentials in caloric requirements, but no norms have bcen len l ted I Witlh Ul-ect to the requirements tof other goods. The reial;i.on,-"il, between expenditure oin food andl othier coiunodities aind services, estimated from consumption expenditure data can i irdlyv be Laken as a . ,d index of the "needs" because they are obvioulIv in- tlueneed iv the oxi:sLing level and distribution of income, wealth alnd i.lrvin. customs among diffure.nt soeio-economic groups. Ihowever, t:i Bureau of Statistui. in Taiwan has been using some weiglhts to esti- Mate adult equivalents, which orovide a starting point. For India, some '*t'ights for the conversion of individuals into adtult consumption units lhave been estimated on the basis of caloric requiroments. ThieLse - ,itL are shown in Annex IVt. he Indian weighits have been used to ostimate adult equivalerts for 1 uajrant, Maharashtra and Nevpal. For rt. Lanka and baiwan, both the Indian and the Taiwanese weights have een usetd. As shown in Table 2 of Annex IV, the correlation between tIO numiber of members and thie number of adult equivalents in a hlouse- %old is very high, a little higher when the Indian weights are used tlha .xit1 the Taiwan weits ,.d Ic 5 sihows thte proportions of households that fall in the same decilt or quintile ac.'ording to alternative rankings in terms of per c ap1ita aiid per adL Lt equivalent expenditure or income. Witlh the Indian weights, betuween 58 to 73 percent of the househiolds (accounting for about 6I to 76 percent of the populatiotn) fall irn the identical decile : ling to thte two criteria; the use of Talvu1in we ighI1t , lowers these ;t cent lg,Usi :n;ikit icantlyv. The weiglits cliosen to e;.LL11n.itlv aduIlt otl iv- ienits cakni Indeed in ll IWlce the ranking F l househlolds. 17; 4i m:,nu T. oi 7 , p* 30 TABLE 5 Percentage of Households in the Same Decile or Quintile According to the Alternative Criteria of Per Capita and Per Adult Equivalent Expenditure or Income Variable Percentage of Percentage of for Weiglhts Households in the Hlouseholds in the Country/StatelArea Ranking Used* Same Decile Same Quinctle 1. GUJARAT, 1972-73 Rural Expenditure Indian 62.4 (65.1) 80.1 (82.6) Urban Expenditure Indian 59.3 (62.1) 78.6 (80.6) 2. MAHARASHTRA, 1972- 73 Rural Expenditure Indian 57.6 (60.9) 78.0 (80.7) Urban Expenditure Indian 63.4 (68.5) 81.5 (84.6) 3. NEPAL Eleven Towns 1973-74 Expenditure Indian 64.3 (67.5) 82.1 (84.3) Seven Towns 1974-75 Expenditure Indian 62.7 (66.3) 80.7 (82.8) 4. SRI LANKA, 1969-70 Income Indian 68.6 (70.7) 85.0 (86.4) Income Taiwan 51.5 (55.0) 75.6 (77.6) Expenditure Indian. 64.6 (66.7) 81.6 (82.8) Expenditure Taiwan 47.4 (50.8) 71.2 (73.1) 5. TAIWAN, 1968=' Income Indian 63.4 (60.8) 81.7 (78.4) Income Taiwan 49.5 (52.5) 73.2 (75.0) TAIWAN, 1974 Income Indian 60.7 (61.9) 79.9 (80.7) Income Taiwan 44.5 (47.2) 68.3 (70.2) *The weights are shown in Table in Annex III. Note: Figures in parentheses show the percentage of total population accounted for by the households in the same decile or quintile. 299 The indices of inequality in expenditure or income per adult equiv- alent (AEE or AEY) lhave been shown in Table 1, to facilitate comparison with measures of inequality in terms of other criteria. on the whole, the inequality indices based on AEE or AEY are not much different from those based on per capita ranking. The use of adult equivaleiits estimated with the Taiwan weights lowers the inequality indices to a slightly greater extent than those with the Indian weights. Pending further experimentation with different weights, one can conclude that although the ranking of housenolds in terms of per capita expenditure or income does not seem conceptually ideal, the resulting measures of inequality are not very dif- ferent from those in terms of adult equivalents. IMPLICATIONS OF DIFFERENTIALS IN AGE DEPENDENCY BY PER CAPITA DECILES As noted earlier, the steady inverse relationship between the age- dependency ratio and the PCE/PCY decile of household, observed in all the data sets, might reflect differdnces in the stage of the life cycle of household heads. With the passage of time, there would be some mobility and a different group or cohort of household heads might come to fall in the bottom decile. Ideally, we need longitudinal or panel studies which would trace the movement of the sample households in an initial survey and also include a sample of the new households formed since the initial survey was conducted. An alternative is to compare the decile distributions of households headed by persons from specific birth cohorts in surveys conducted at different time intervals; but the validity of inferences based on this approach would (depend on the com- parability of concepts and definitions based on this approach would depend on the comparability of concepts and definitions in the various surveys as well as the stability of the headship rates (i.e. the ratio of household heads to the total population of a given sex-age group). Pending further exploration of the life cycle related issues, some implications of the differences in the age composition of population in different deciles of households may be noted. First, the poor house- holds face a constraint in te-ms of the proportion of persons in working ages or potential earners; it would persist even with full employment. Hotvever, insofar as the incidence of unemployment and underemployment is greater among the poor,Il programs to relieve unemployment can help the poor households in a very tangible manner. Secondly, the effect of differences in age composition on the labor force participation rates of population in different deciles deserves careful consideration. The non-uniformity of labor force definitions in different surveys (outlined in Annex I) would probably not affect the relationship between decile and the participation rates. Table 6 shows participation rates for ages 10 and over or 15 and over.Z! The rates 1/ Althoughl the measurement of unemployment is not based on uniform cri- teria in different surveys, the incidence of unemployiment varies inver:;ely with PCE or PCY decile, i.e. is greater among the poor, in four of the five areas studied in this paper. For Nepal, data on unemploy)ment are not available. 2/ An analysis of the determinants of labor force participation, taking into account the characteristics of individuial sample households, is yet to be undertaken. 300 TP'7LE 6 Labor Force Participatiou Races of Pooulacton Ages 10 and Over (or 135 ;d Over) by Sax and Docile of Houaeholds in Ter of Per Capita Expanditure or tncnm CUJ-RAAT ASRT NA TAI$Ab Eleewn Seven Rural Urban IuraJ Urban Toian TCQW Docile (1972-73) -(1972-7)) (97374) (1974-75) 1969-70* (1968)0 (1974)5 1 77.6 63.3 72.2 63.3 76.8 84.3 78.6 90.6 87.8 2 76.7 65.4 74.0 64.7 76.0 80.4 81.8 92.2 86.4 3 77.2 j66.3 76.5 f65.3 75.8 79.0 82.8 90.2 84.1 4 \ 73.2 77.1 84.0 86.5 85.6 5 175.7 63.5 r77.6 J64.9 74.4 76.8 83.1 85.1 86.1 4 'i 67.6 73.9 82.5 84.5 78.4 7 73.0 8.9 '77.0 69.6 72.9 74.3 83.0 83.8 79.2 a K 68.4 72.9 82.6 79.2 76.6 § 73.3 66.8 79.2 73.6 69.6 68.4 83.9 80.2 74.7 10 76.2 71.2 75.2 75.1 74,7 76.6 80.0 78.7 81.5 AlU 75.6 66.2 76.2 67.1 72.9 76.3 82.3 85.2 82.1 IEALES I. 62.2 24.2 60.0 27.9 42.2 70.7 27.9 61.6 64.6 2 :8.5 18.4 65.0 22.9 44.1 63.7 28.8 61.1 58.8 3 56.3 J16.7 '63.8 17.6 40.1 65.7 29.7 56.3 52.9 4 38.0 56.3 31:2 50.9 45.5 5 0.6 dj1.3 t62.6 12.9 36.4 52.1 26.9 43.8 43.1 i 30.0 52.9 26.5 43.5 44.7 7 ,;8.2 J1.3 Z.1 [1.3 29.3 53.1 27.2 41.2 41.1 8 a N 26.2 52.6 25.5 42.9 39.5 S 45.6 11.1 34.2 4.0 24.3 32.9 21.8 35.3 40.2 10 40.4 12.0 5l.1 18.0 27.0 30.2 29.1 33.5 41.7 A13 52.0 14.9 61.1 17.1 34.4 55.4 27.5 47.8 47.7 FESONS 1 70.1 43.3 62.2 45.4 58.7 77.2 .51.7 75.3 75.9 2 §7.8 42.0 69.6 44.2 59.9 72.0 54.9 76.0 72.3 3 66.9 42.1 '69.5 '42.3 57.4 72.2 55.6 73.0 68.2 4 3. 55.4 66.7 56.9 68.2 64.6 5 63.3 38.3 70.0 41.0 55.5 64.5 55.0 64.0 64.1 6 48.9 63.7 54.6 64.1 61.6 7 1.0 1.5 69.8 3.5 51.2 63.9 55.8 61.6 59.9 a 47.8 63.5 55.1 61.2 58.4 9.5 .4 6.8 8.7 48.8 54.0 54.0 58.9 57.9 10 58.9 45.0 64.4 51.6 55.5 61.5 55.6 58.1 62.4 Al 64.1 41.4 68.7 43,9 54.0 66.3 55.0 66.3 64.8 For porxona aged 15 and o-r. aJ Tor persons ased 13-64. 7 Decile, of incom. 301 for males do not show any consistent relationship with PCE or PCY decile and, not unexpectedlv, the differences are small. The female rates, on the other hand, vary inversely with decile and, except in Sri Lanka, the decile differentials are sizeable. The "crude" participation rates (for all ages, i.e. including children in the denominator) show a positive relationship between tOhe male rates and the PCE/PCY decile, which is partly offset by the inverse relationship between the female rates and decile. For both sexes to- gether, the participation rates show a rise with decile in urban areas of Gujarat and Maharashtra and in Sri Lanka. In other cases, the decile difference in crude female participation rates or at least the partici- pation rates for both sexes together are relatively small and .-low little consistent pattern. On the whole, the poor households try to overcome the hi3h hurden of young dependents through greater labor force participation bv their females.l/ But the decile differences in labor force participation of both sexes together (or for that matter, in the incidence of unemplov- ment) are much smaller than those in the levels of consumption or in- come and are unlikely to explain much of the variance. To anticipate the results of some multivariate analysis in progress, the main vari- ables explaining th. differentials in PCE or PCY are likelv to he dif- ferences in househiold size, educational attainment (or human c.-ittal) and the associated variations in the composition or the structure of employment in terms of occupation and/or industrv and the status or class of workers. Some of these characteristics are characteristics of intdividuals or workers and not o-f households. WhRien there is more than one worker in a hiousehold, one can take account of tLle liiriic t-A'r- istics of the main earner, if the data permit his identificatioti;<.. 1/ The conventional wisdom suggests that the poor attempt Lo mitigate the burden of dependency by having their children participate in economic activity; but with the growth of population and labouir force over the past two decades or so, the direct contribution of these child-workers to the income or earnings of the family tends to be quite limited, at least in the densely populated countries. Perlhaps that factor and the rise in school enrollment ratios, rather than the biases of enumerators and interviewers, explain the very low labour force participation rates repcorted in our sur- veys for ages 5-9 and 10-14 and shown below. Of course, the clhildren might contribute to family earnings by taking care of siblings and permitting women to work. Guj arat Maharashtra Nepal Taiwan Rural Urban Rural Urban 11 Towns 7 Towns Sri Lanka 1968 1974 Males 5-9 1.0 0.3 1,2 0.1 3.2 3.6 0.3 0.1 -- males 10-14 19.4 4.5 20.1 5.0 16.0 25.7 6.3 12.3 1.6 Females 5-9 1.2 0.2 1.1 0.1 4.4 3.7 0.4 0.3 0.1 Females 10-14 20.3 1.5 22.7 2.8 14.3 35.0 3.5 20.1 3.9 Persons 5-9 1.1 0.2 1.1 0.1 3.8 3.7 0.4 0.2 -- Persons 10-14 19.9 3.1 21.4 4.0 15.2 30.1 4.9 16.1 2.7 2/ Irn households with self-tmplozyed or unpaid familv worklers, it is virtually impossible to estimate the varnings of individtuals. 3102 or else the characteristies of the houselhold head may be assumed to reflect those of otlher .vmKbers as well. It is important, hoWever, thllait inequality is studied in terms of the per capita income or expenditur,e so that the gredLer burd]cn of dependents in bottom deciles of t2oiIo.1dv is not overlooked, as '.:jrtnif the income distribuition is e:;arinmr] in terms of only thie earners or the economically active population. TIE KUZNETS CURVE The discussion ss' far lias M,icliih ted the difft rence.; in some of the characteristics of households ranked -iccording to their 1PCF. (or PCY) and THE (or 'ITIY). M'ucii of thie current discussion about r1Ia[c' in income distribution in the process of economic growth is conducted on the basis of ranking of households in terms of total incomne. Of course, the original formulation hy Simon Kliznots was in terms of "the .n,;l' ity in the size distribution of tottal income" of households.!' Te- same is true of the recent cross-sectional regression analysis by AhluwaLia of the variations in the income shares of bottom 40 percent associated with the per capita GNP of 40 developing countries or a total of h6 countries because most ,L- the income distribution data shoew the rank "in of households b THY.I As shlown e:trlier in Table 1, the indices of inequalitv based on the distribution of individuals ranked accor, . to the. PCY of their households would be smaller than those based oni the distribution of lioiseliolds ac,cording to their THY. However, the difter- ences are small and the r ikin4, of countries according to the levl el. income inequality seems aiiiikelv to ch;ini:,o. If the Kuznets; eirve can Tho isumed to lhold even when 1ino,1.L! itv is measured in terax; of `C6, thie earlier discussion of thei s character-- istics of households in terms Of size, cu eomposition of the popeuL- tion and labor force I' .1 'tion provides some basis for f por n I about the poss iile 1cnt rii'utoiic o f the proces,s of demograph I ri tr_1' it to the roiaL:,Li*' between .W it, and economic growth. Let us assune thlat the process ef d-semographic transition t an idealized courbc, with Ca) an initial condition of almost univtV r.;. high levels of mortality and ferti1i.tv 0wi t th low rates of natural !i, crease); (b) a phiase of declinii- mortil ir;v uniaccompanied by I- rr litv decline (and therefore a rise in the rato of natural inc-areask.); (c) a third phase in which Kr: .it'; declines with a time lag (initiallv eintv offsetting the decline in irtality but eventually Icidiiv, to a dt- cline in the rate of naturlij in ervase); and (d) thl final phase which both mortal i .v nid f ertilitv arc owy more or less thirougheo r the society and the rate ntiztural inerease is determined essentially bhv fluctuations in fetrlit't. K "f- rentials in fertilitv, or more .; ir- tantlv in the rate of nratural. increase 1 income class, whLich are, mentioned as likely to cent rihute to the iZiiot.. curveA,! in' arto I Sinion Kuzn't., , ''i nt. t-. lve Aspects of the FEconiomic Growth of Nations: VIII Distribution oif Incolme bv Size", . ; 'I'if u t'L.l ';., ', . XI Nc. 2, P.11 L II, Januarv 1% 3, p. i'7. 2/ Montek S. V\h1 'Xi'l 11i. I' in..qna l iv, ov 1 coil Ihvc I nL';u '' , r 1. .3. 1976, pp. 307-3'1 3/ Ahluilwalia, -.cnt t_l1ly duiring the second and third phases of the process of demo- V,1iC tranisition. However, some other links in the chain can also be identified. It seems to me reasonable to presume that in the initial situation before the start of mortality declines, differences between income .:rups in the household size or the age composition of household members wuld be relatively small except for those associated with the age of thie household hieads.11 As the transition begins, the better-off or the richl would be the first to enjoy the benefits of mortality decline and .ub.q1lUuntly they would take the lead in fertility conitrol while the poor also b.egin experiencing a decline in mortality and an incerease in tih number of surviving children. The result would be a widening of the differnLtials in household size. Since the age composition of thle pop- uLation (of a group) responds quickly to the decline in fertilitv (more than to the decline in mortality), the rich would also begin to have a 'maller dcpeudLnc% burden. the differentials in the age-dependency burden wouild also widen as a result. If social custom requires evervone above a certain age to take up some work, tliere would be a simultaneous .lninig of the differentials in labor force participation rates, and, ;tArofore, in per capita output of the rich and the poor hiouseholds. Insoifar as the rich, whose dependency burden would be smaller than that ot hie poor, send their children to school to develop their skills and abilities, differences in productivity would also contribute to an in- erease in inequality. As the process of demogJrap1hic transition com- p.etes the idealized third phase noted above, tlhe difforentiall; in household size, agt dependency burden and labor force 1p.arLicilpLtioll rates would once igain tend to shrirnk. Of course, a narrowi.ngs of the iLneome-related differentials in school enrollment rates and educational attainment tf workers (that has occurred in most developed couintries) would also lead to a decline in income inequality. )bviously, the reasoning outlined above abstracts from several other factors contributing to disparities in income and consumption. Also, it in difficult to find any empirical evidence to test the valid- itv of the speculation just outlined. Data on household size in the now developed countries for the pre-transition period seem to be virtually non-existent. However, there is strong evidence indicating an increase during the past two to three decades in the average household size in several Latin American and Asian developing countries..2/ In the pre- .ontly developed countries on the other hand, the secular trend in I7 lilt differentials in income in this situation may be a result of dift:erential endowmencs of labor skills or property. * ! ire. )probably, the relatively rapid decline in mortality unaccom- panied by fert-ilitv decline has been the main contributory factor. tLIlmn IcNEnn1:'.- towards urbanization and a rise in the female partici- pation rates would be expected to have the effect of lowerinr thie .nvor-ot'e family size. See: United Nations, JD4't-,",t. vimf.' and ("ono- wfr?U of i . O t art : o Tvtealr. - N.-. . ivarori, of F5o. X ; ,o7tn fntora.o,tien ;;+ ,:.q ?j , '.. i,'.o>rr. bo *i Th Jz'.t,w.,, Vol 1, New York, 1973, p. 3!47. 304 average household size during the period for which data are available has been generally downward, although in England and Wales during 1811-1851 and 1961-1891 and in Japan during 1925-1950, there was some increase in average household size.l/ Even with a declining or constant average household size, the range of variation or variance of household size could rise. In the United States, for example, while the average size of a household was 5.79 in 1790 and 4.76 in 1900, its standard deviation was 2.82 and 3.01, respec- tively. By 1950, however, the average size had declined to 3.52 with a standard deviation of 2.70.2/ In Taiwan, between 1968 and 1974, the decline in the average household size from 5.8 to 5.3 (shown in Table 3) has been accompanied by a decrease in its standard deviation from 2.74 to 2.26. More research into the variations in household size may provide some interestinf further evidence. While it may be possible to study the time-trends in the disper- sion of the size of households in different countries, there is little hope of studying the age composition of members in different deciles of households. Even for recent dates, special tabulations of the pri- mary data are necessary for such work. It seems to me quite likely, however, that differences in the age composition of the population and the proportions of workers (or earners) in households falling in differ- ent deciles would increase with the increased variations in household size, and would contribute towards an increase in the level of inequal- ity during the process of demographic transition. These differentials would probably narrow as the low fertility becomes universal. The attainment of the final phase of demographic transition with lowq levels of both fertility and mortality is usually associated with a substantial spread of literacy and education, as well as high levels of urbanization and industrialization. In other words, the demographic transition normally occurs simultaneously with substantial changes in the sectoral occupational and status distribution of workers and the narrowing of the inter-sectoral differentials in productivity, which have been emphasized by Kuznets. While it is not possible to quantify with any precision the contribution of the process of demographic transition to the observed inverted U-shaped relationship between in- equality and economic growth, these phenomena seem to be inter-related to some extent and perhaps some simulations would yield interesting results. 1/ Op-cit., pp. 341-342. 2/ Louis Winnick, A.rioaenz n i,, ucz and Its Use: ThIzc Demand for SheLter : New York, John Wiley & Sons, Inc., 1957, p. 79. For calcu- lating the standard deviation from the size distribution of house- lholds shown by Winnick, it has been necessary to estimate the aver- .lge size of households with 10 or more persons by taking into account the reported mean household size of the entire population. Note that the territorial coverage of the United States had changed substan- tially between 1790 and 1860. Tlierefore, the estimates are believed to be only crude indicators of the likely tendencv. 305 THE PROSPECTS AHEAD The Kuznets curve as well as the differentials in fertility that are an integral part of the process of demographic transition suggest that ceteris paribus, the inequality of income and consumption in the developing countries would increase in the decade or so ahead, when according to the United Nations projections, th- birth rate (per 1000 population) in the LDC's is expected to decline from about 39.0 during 1970--75 to 34.9 during 1980-85 and 29.9 during 1990-95.L/ However, I believe there is nothing inevitable about the Kuznets curve and with a proper mix of policies, it would be possible to translate the aspira- tions for greater equality into reality. As pointed out by Dudley Kirk, the time span required to complete the process of demographic transition appears to have shortened in sonie of the countries that have lowered their fertility since the 19501S.2/ Further, the official family planning programs do hold promise of low- ering the threshold of development when fertility behavior of even the poor would begin to change. One factor that strengthens such optimism is the substantial nar- rowing of the differentials in school-enrollment ratios for ages 5-14 in different PCE/PCY deciles, reZativeZy to the differentials in liter- acy and educational attainment of the older age cohorts, evident in our data for Gujarat, MAaharashtra, Nepal and Sri Lanka.3/ Of course, Guj- arat and Maharashtra are among the more developed states of India and our data for Nepal pertain to only 18 development certers or towns. Further, even now differentials in school enrollment rates are not neg- ligible and their elimination should be the goal of public policy. IHowever, the data point to a significant narrowing of differentials, which could help both in shortening the time-lag betwqeen mortality and fertility declines and in reducing the differentials in earnings re- lated to differences in schooling and education. In addition, of course, land reforms, progressive taxation, and public expenditure policies designed to maintain and raise the consump- tion standards of the poor and to give them equaliLy of opportunity, 1/ The average annual rate of natural increase, however, would fall from 2.5 during 1970-75 to 2.4 during 1980-85 and 2.1 during 1990-95. See: United Nations, The VEt..rW,'itS and c;vwls of P. u.a- tion Trenda, op.cit., pp. 564-567. These projections were made in 1968. 2/ I)udley Kirk, "A New Demographic Transition?" in: Iq'7 o'-r:i. ItiOn i0rowth: OLnscql4encc'S Zd .' - r'i 7 iy7tioiw, presented by a SLudy (;omeitLee of the National Academy of Sciences, Baltimore and London: 'lTe Jolhns Htopkins Press, 1971, pp. 123-147. 3/ In Taiwan, over 85 percent of the population aged 15 and over in 1974 wqas literate and almost 80 percent had completed prtmarv School. 306 sometimes summed up under the "basic needs" oriented straLegy of devel- opment, would also help to achieve a reduction of disparities in con- sumption and income. Since the demographic processes operate over rela- tively long time spans, these oLhor policies would be more potent and critical over the next decade. 307 ANNEX I CONCEPTUAL FRAMEWORK OF THE DATA ON (i) EXPENDITURE AND/OR INCOME AND (ii) LABOR FORCE IN THE SURVEYS SELECTED FOR ANALYSIS Given below is a summary of the important procedures and rules adopted in different surveys for recording expenditure, income and labor force data. INDIA Expenditure 1. Consumer expenditure in cash and kind (and also quantities) were re- corded for items purchased or consumed during the 30 days preceding the day of interview (including value of domestic consumption out of home-grown stock; rent and taxes -- but excluding income-tax). 2. Expenditure relating to productive enterprise was excluded. 3. For clothing and footwear, items purchased during the reference period but not brought into maiden use were excluded. 4. Expenditure on the purchase or repair of durable goods for domestic use was included. 5. Value was imputed for non-purchased consumption. (a) for commodities received in exchange, at average local retail price. (b) for consumption out of home-grown stock, at ex-farm or ex-factory price. (c) for consumption out of transfer receipts, i.e., borrowing, gifts, charities, etc., at average local retail price. (d) for consumption out of free collection on the basis of retail prices in a market neavest to the sample village. 6. Consumptiort on ceremornial and other occasions was included, with the appropriate information. 7. No value was imputed for owner-occupied housing. 308 Labor Force/Employment (a) Usual activity (determined on the basis of the "normal working pat- tern ... over a long period in the past") was recorded for every respondent. For persons pursuing more than one activity, that whici accounted for more labor time was recorded as the usual activity. (b) Current activity or the activity of the week preceding the date of interview was recorded for every respondent. (c) For persons in the labor force during the reference week, the sur- vey recorded (i) occupation; (ii) industry; (iii) class of worker; (iv) days worked during the last week; (v) days of unemployment; and (vi) days outside the labor force. (d) For persons usually in the labor force, the survey recorded: (i) usual occupation; (ii) usual industry; (iii) class of worker; (iv) preference for location of work and factors determining availabil- ity for additional work. Notes: (1) Available data relate to the "monthly per capita consumer expenditure" (MPCE) of the household. (2) The participation rates shown in Table 6 are based on the usual activity data; but they are not significantly differ- ent from the rates based on the reference period of one week. NEPAL Income 1. Earnings during "last month" ("total" earnings for persons with more than one occupation), including "pay in kind" were recorded for employees. 2. "Net profit last month" was to be recorded for an employer or own- account worker. (However, the information was obtained for only about 47 percent of such workers in the eleven towns and 35 percent of those in the seven towns.) 3. An attempt was made to record the value of drops and other goods received as rental payments (presumably for "last" month only). 4. Changes in assets and liabilities were recorded. 5. Data were collected on the ownership of 28 "durable" goods for use in constructing an index of socio-economic status. E.xpenditrure (a) Expenditure on non-food consumption during the "past calendar month" was recorded. 309 (b) For food expenditure, each household was visited for "six consecutive days" to report the ptirchases of food items and the use of home- produced food items. On thle sixth day, food consumption was re- ported for the "previous two days". (It is not clear wlhetlher the procedure implied that the consumption of the fiftlh day was re- ported twice.) The relevant schledule was devised to make a "seven-day record". (c) Consumption of lhome-produced goods was vaLued at retail market prices. (d) Rental value of owner-occupied and free 11o0s LoI; was estimated and recorded. Labor Force (a) No explicit reference period was specified; essentially the giinful worker approach was followed with an instruction to record "the occupation even though the person is not employed at the moment". However, for retired persons (those who have given up the occupation with no intention of working again), the current situation was to be recorded. Ulnpaid family workers ware counted as "emploved" if they worked "15 hours or more during the week". (b) Provision was made to record the eMph'.mL`Tt staLtus (i.e. class of worker of the employed and also unemphowmvtt) for everY one with an occupation. A person witlh an occupaltioIn but "not wo'-V.i,,i be- cause he has no job, is sick, is a seasonal worker, or for ativ other reason" was to be classified as uLnemp lonved. (Iln f;wt., lhow- ever, we lhave not found anyone coded as uinemployed in the seven towns surveved durinig 1974-75 and onlv 44 uneniploWed, i.e. 0.5 per- cent of the labor force in the 1I towns surveved during 1973-74.) (c) The employed have been classified according to a one-digit indus- trial classification and a three-digit occupational classification. "Usual" hours of work were recorded for most workers. Notes: (1) Tlie reported expenditure on marri..ip's, ftnerals and festi- vals and government taxes has been excluded because (nlv a few households had reported such expenditure. The inclusion of these items tended to distort the mean levels of expenditure. (2) The seven tonns surveved durinig 1974-75 have been kept dis- tinct from the eleven towns surveyed dturinig 1973-74 because of significant inflation during the intervening period and the non-availabilirv of any suitable priet index numbers for an adjustment. 310 SRI LANKA Income 1. Income was defined as "receipts in cash or kind from work, property, transfers and other sources that contribute to the individual's or individual earne-'s spending power." 2. The households whose income was subject to considerable seasonal fluctuations were asked to report "average monthly income based on the total income over the year preceding the Stirvev". For others, the reference period was "last month". 3. Value was imputed for "home-grown" consumption, free goods and ser- vices and owner-occupied housing. The latter was valued at "pre- vailing market prices", adjusted for maintenance costs and taxes. 4. A distinct record was made of income from: (a) Wages, salaries and related receipts; (b) Profits from business enterprise or farm; (c) Rent, dividends, and interest; (d) Pension, remittances, and cash allowances; (e) Other periodic cash receipts; and (f) Non-monetary income. Expenditure (a) For food, drink and tobacco, data were collected through daii'. oisits on seven consecutive days (blown up to get monthly estimates). (b) For non-durable goods and services, data were collected for the calendar month immediately preceding the field survey. (c) For consumer durables, data were collected for the "preceding year" and then averaged to get monthly estimates. (d) No value was imputed for free ration of rice. (If the value of free rice ration - two pounds per week per person - is taken into account oii the basis of the average price paid for rice purchased from the market, the indices of inequality would be lower.) (e) "Whenever possible", a responsible member of the household was asked to maintain a daily record of consumption. Labor Force/Employment (a) For "regular" workers, the reference period was one month preceding the survey. (b) For persons in seasonal occupations, the "usual status" was recorded. 311 (c) Persons working temporarily or casually on a contractual basis were classified as employed if they had worked for at least 10 days during the month preceding the survey. (d) The unpaid family workers in household enterprises were classi- fied as employed if they had worked for at least 10 days during the month or the season preceding the survey. (e) The unemployed were defined as persons aged 15 to 55, who were not classified as employed in terms of (a) to (d), and were eel isZn. work (excluding persons "mainly engaged" in household work and students). TAIWIAN Income 1. Income during the previous year was recorded with a detailed accounting of all receipts (both in cash and kind), consumption, changes in fixed assets such as land, house, or other construction and also in financial assets and liabilities. 2. Personal income was recorded for all wage or salary earners. 3. Net farm or business income was recorded for the whole household. Labor Force (a) No explicit reference period was used for the questions on occu- pation; but a distinction is made between major and secondary occupations. (b) Data on the major and secondary occupations of the employed are available in terms of; (i) 12 industrial divisions; (ii) 12 occupational divisions; (iii) whether employed in (1) public corporations, (2) other government sector, or (3) the non-government sector; and (iv) class of worker, (c) Unemployed persons have been identified, but there is no information on the duration of their unemployment. ANNEX II INDICES OF INEQUALITY The indices of inequality shown in Table 1 were calculated by using the computer programs developed at the World Bank for the compilation 312 of income distribution data for different countries.I/ These programs fit a Lorenz curve to the observed data on the basis of a mathematical function proposed by Kakwani and Podderl/ and derive the income (or ex- penditure) shares of 20 five-percentile groups of population, which are used for some of the inequality measures. The Gini :oenif7cient is a ratio of (a) the area on a graph that lies between the Lorenz curve and the egalitarian line (or line of per- fect equality, which forms a 45-degree angle with both the x- and y-axes) to (b) the area of the entire triangle formed by the egalitarian line and the x- and y-axes. As a measure of income concentration, the Gini coefficient ranges from 0 to 1 -- the larger the coefficient, the greater the inequality. Our Gini coefficients are calculated directly from the estimated parameters of the Lorenz curve. As derived by Kakwani and Podder, the Gini coefficient is given by G = 2fo ar (Fi - i) di = 2a(T) )+a+6 B(l+a, 1+f) where B(l+a, 1+8) is the Beta function which has been widely tabulated. The Kuznets index is estimated by calculating the absolute mean de- viation of income (expenditure) shares of each five-percentile group of the population from five percent. The arithmetic average of the 20 absolute deviations ranges from zero in the case of perfect equality to 9.5 in the case of maximum inequality. Division by 9.5 is necessary to standardize the measure to a range of zero to one. This gives: D = I d| 20 9.5 where d is the absolute deviation of the income share of each 5 per- centile group from 5 percent. 1/ Shail Jain, Size Distrinbution of Income, A ConpiZation of Data, Washington, D.C.: The World Bank, 1975, pp. xii-xv. 2/ N.C. Kakwani and'N. Podder, "Efficient Estimation of the Lorenz Curve and Associated Inequality Measures from Grouped Observations", Fcor0o2nmetz-ca, Vol. 44, No. 1, pp. 137-148. 313 The following entropy measure, or standard7zed E index, reported in the tables follows a suggestion by Marfels,i! which can be written: E* = 1 - antilog H(y)/n where H(y) is the entropy measure defined as follows; H(y) v i log (l/yd - -Ey log Yi where yi are income shazes of the it individual. H(y) is at a max- imum when all income shares are equal, that is, when Yi = I/n for all i . In this case H(y) = log n and the inequality measure E* = 0 . Conversely, H(y) is at a minimum when all income accrues to one in- dividual. In this case, H(y) = 0 and E* = 1 - l/n . Thus, E* varies from 0 to 1 - 1/n . E* can be standardized by dividing it by (n - 1)/n to give an index of inequality E ranging from zero in the case of maximum equality to unity in the case of maximum inequality. ANNEX III Tables 1 and 2 below illustrate the changes in the relative posi- tion of households with heads of different five-year age groups, when the ranking criterion of total household expenditure is replaced by per capita expenditure. The data examined in these tables relate to Gujarat and Maharashtra states of India and the 18 towns or develop- ment centers of Nepal. A similar pattern is expected if the income data were available for these states/countries and when they are tabulated for Sri Lanka and Taiwan. 1/ Christian Marfels, "Absolute and Relative Measures of Concentration Reconsidered", Kyklos, Vol. 24 (1971), fasc. 4, pp. 753-66. 314 *. -- . - ->>w ... -+ u. cef o -w o c1 - . '. ' ' I --I - I ^ ,C-CCO' --CC,, I - O -; C 0 w a a * -C, C C - -.CCa OOCaC,- vJsG- G- -C4 t- 0 0> o- -a a0 a a. -C 0 - - i aCaM 00 M0 C CO C C -" C' *> C Jw- -- . .,- C- -a-,C-C- a-a - - !',)Ca-7 !-~C,C-C w wlo [ -I TABLE 2 SRMAL.: .E CUTTIA A2M TO'. :1J.UST1LD EYi': I-'2.C (1R NEPi'&LI RlPEES) RY rC. :RC,-MP OF ]H E i4.A7I W F F'.',73-7' .':.0 (F.) S .:::l T. *- I'.A-iS A_Qp of Nv:zber Average Average Average Index of Index cf Hcuaschold of Hou-'ez.old RouSOahld Per Capita Elo. d P-r Capita F(ad - 1 '. Size t u I ; - ; , r,.Frrt i ture Ex enditure F,.pendit.zc. (A) EZeven , 1 3-, 4 All 4L93 5.4 500.0 92.9 100.0 100.0 less than 20 62 3.5 334.6 95.8 66.9 103.1 20-24 228 3.4 418.7 122.9 83.8 132.3 25-29 436 4.0 406.0 101.2 81,2 108.9 30-34 595 4.8 451.0 94.5 90.2 101.7 35-39 646 5.4 487.5 90.5 97.5 97.4 40-44 599 5.9 528.5 90.1 105.7 97.0 45-49 474 5.9 513.9 86.7 102.8 93.3 50-54 4 69 6.0 591.2 98.8 118.2 106.4 55-59 292 6.2. 580.5 93.6 116.1 100.8 60+ 602 6.3 545.2 86.5 109.0 93.1 (B) f, .>n !':~.. i7z-75 All 2254 5.2 583.6 113.1 100.0 100.0 less than 23 76 2.8 397.2 143.1 68.1 126.6 20-24 203 3.3 475.3 142.0 81.5 125.6 25-29 258 4.0 483.9 121.2 82.9 107.2 30-34 321 4.9 608.1 125.1 104.2 110.7 35. 39 309 5.4 544.6 101.6 93.3 90.0 40-44 2',3 6.1 644.4 106.1 110.4 93.8 45-49 24t 6.1 651.6 106.2 111.6 94.0 53--54 19l) 5.9 617.5 104.9 105.8 92.8 55=59 11' 6.G 676.3 213.6 115.Q 103.5 60 2 5.9 651.4 11).7 111.6 97.9 ANNEX IV ESTIMATION OF ADULT Z9UIVALENTS Table 1 below shows the two sets of weights or coefficients for different sex-age groups used to convert individuals into adult equiv- alents. The basis for the weights used by the Bureau of Statistics of Taiwan is not known. The Indian coefficients were calculated on the basis of the recommendations of a Committee constituted by the W4HO and the FAO. They were used to standardize the estimates of caloric con- sumption from the data on quantities of different food items consumed by sample households during the month preceding the date of survey, during July 1971-June 1972. The weight for males aged 13-15 is lower than that for ages 10-12 or 16-19 because of similar differences in the FAO estimates of daily caloric requirements. The weights do not take account of the increased caloric needs of women during pregnancy and lactation. The estimation of equivalent adults in Maharashtra has required some arbitrary assumptions because the specific ages of children aged less than five years were not recorded on the tape. An average weight of 0.53 was estimated on the basis of the detailed age composition of the age group 0-4 reported by the 1961 Census of Maharashtra. This weight was applied to all children below five years of age in every household. To some extent, the usefulness of fine age-specific weights for estimating equivalent adults is limited by the errors in age reporting which are particularly large in single-year age distributions. However, as shown below in Table 2, the correlation between the number of per- sons and the number of adult equivalents in households has been ex- ceedingly high -- a little higher when the Indian weights were used than when the Taiwan weights were applied. 317 TALT,F, 1 Weights for Estimating Equivalent Adult Consumers India, 1971-72 Taiwan, 1974 Age Group Males Females Age Group Males Females 0 0.43 0.43 0-1 0.3 0.3 1-3 0.54 0.54 2-4 0.4 0.4 4-6 0.72 0.72 5-7 0.5 0.5 7-9 0.87 0.87 8-10 0.7 0.7 10-12 1.03 0.93 11-14 0.8 0.8 13-15 0.97 0.80 15-20 0.9 0.9 16-19 1.02 0.75 21 and 1.0 0.9 over 20-39 1.00 0.71 40-49 0.95 0.68 50-59 0.90 0.64 50-69 0.80 0.51 70+ 0.70 0.50 Sources: (1) For the Taiwan data, figures supplied by the Bureau of Statistics, Taipei, Taiwan. (2) The National Sample Survey, Draft Report No. 258/10, Calorie and Protein Content of Food Items Consumed Per Diem Per Consumer Unit: All India, Rural, 26th Round, July 1971-June 1972, mimeo. 318 TABLE 2 Measures of Association between Household Size and the Number of Adult Equivalents in Households for Some Data Sets Household Size and Household Size and Adult Equivalents Adult Equivalents (Indian Weights) (Taiwanese Weights) Correlation Coefficient of Correlation Coefficient of State or Couatrv Coefficient Determination Coefficient Determination GUJARAT, 1972-73 Rural 0.99 0.97 -- Urban 0.99 0.98 MAHARASHTRA, 1972-73 Rural 0.99 0.97 -- -- Urban 0.99 0.98 -- - SRI LANKA, 1969-70 0.99 0.97 0.96 0.92 TAIWAN 1968 0.98 0.97 0.97 0.94 1974 0.98 0.96 0.96 0.91 NOTE: All figures are independently rounded. 319 F,icteurs denvgraphiques et rdpartition des revenus: Etude de certnines questions R6sumd L'un des objectifs de ce document est de mettre en lumiare les consucluenccs qu'a Vapplication de l'un ou l'autre des deux criteres possibles de classement par raiigs lorsqu'on veut calculer les inegali- tks de revenus et identifier les populations pauvres au profit duquel peuvent eLre conqus des programmes spgciaux d'aide ou de secours. S'il est vrai que le mdnage est l'unit6 logique pour le recensement du reve- nu (ou des depenses), la r6partition des mgnages en fonction du revenu de chacun d'eux n'est pas un indicateur valable du degre d'in6galitg qui existe entre eux, car la taille des m4nages n'est pas la meme selon le ddcile out ils se situent. L'autre solution, qui est &videmment plus simpie, consisce a classer les menages ou la population d'apres le re- venu ou les depenses par personne (du meanage). La proportion de men- ages et de population que l'on retrouve dans le xndme decile du fait de l'application de ces deux criteres est extrgmemont faible, et de ce fait, les groupes de population pauvre ainsi identifids pour l'etab- lissement de programmes d'aide seront tres diff8rents selon le critere utilisd. La taille moynine du mdnage diminue d'un d6cile h l'autre lorsqu'on prend comme critLre le revenu ou les dl3pensuj par personne; elle aug- ment *r contre lorsque le critere est le revenu ou les d6penses totales du mn6!liiC '-; diff4renc.t s tiennent ani fait que la taille moyenne du ::en fotion de clu du clief de famille jusqu'a 40 a 44 ans envi rov. En oultre, ..;o! de m,'n_iFvs dont les ethefs de famille soknt jeune. se riiWrouvent dans les dMles inf6rieurs si l'on utilise le cri'tZ.ro des tlpenses totales du mdnage et dans les deciles superieurs si. i'on utilise le crit4re du revenu ou des ddpenses par personne. On note des difLfreonces importantes dans la r6partition par age der Tlnr':one.n dont se composent les mrnnages situ6s dans diffdrents d6- ciles et range; selo.- le revonu (ou les dk'-pensos) par personne. De ce fait, le revenu (ou les depenses) par iquivalent adulte senoble gtre un cr7 '- pPlus valable; to'utef-ois, les estimations des in6galit4s qui re.;:itnt de Vtapplication de ce deonier critere ne sont pas sensible- ment difforents de celles qui se fondent sur la r9partition des menages d'a4v.r>.; le revenu (ou les d5pe-,s-s) par personne, du moins pour les deux s4ries de poelde3r.if-i'ri cholsies pour convertir les individus en u'.'; '.-i ' .ie ts, adultes. La d.llXi;me partie du document est un essai de caractere assez *&Wt lti r, *iur la "courbe de Kuznets"t, qtle l'auteur explique par le fa t que les dcarts entre les composantes de taille et d'age des mgn- ai's gi;nentent au cours d'une phase de "transition dmograpmIique" oe le rvthme de croissance de la population augmente en raison du decalage entre labaissement de la mortalit6 et de la f6conndit.. 11 reste, pour vkerifier cette }lyut ;, tl rtunir les preuves empiriques qui devront {n;c r sur des donndes eit.;rLu r st certainw modeles de simula- tion. I.'auteur s'efforce IV.xl1, n.:ent de presenter certaines hypothEses siur leS i:euAdanc.Ps futuret, et pro'1'able- de la r'pcartitinn des revenus (tvu des d6penseo.) dans les pays en d;vevoppement au cours de la ..'