ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

'How Many Poor in the World?': A Critique of Ravallion's Reply

The EPW special issue (25 October 2008) on the new poverty estimates of the World Bank and the editorial that accompanied it have provoked discussion. This note comments on Martin Ravallion's critique of the EPW editorial.

DISCUSSIONEconomic & Political Weekly EPW january 31, 200967‘How Many Poor in the World?’: A Critique of Ravallion’s Reply S SubramanianI am indebted to Thomas Pogge for helpful comments on an earlier version. The usual disclaimers apply.S Subramanian ( is with the Madras Institute of Development Studies, Chennai.TheEPW special issue (25 October 2008) on the new poverty estimates of the World Bank and the editorial that accompanied it have provoked discussion. This note comments on Martin Ravallion’s critique of theEPW editorial.The EPW editorial “How Many Poor in the World? (25 October 2008) was critiqued by Martin Ravallion in his “Reply” (8 November 2008). This note critiques Ravallion’s own. Ravallion (2008b) has responded to each of four criticisms that have been made, in the EPW editorial, of the World Bank’s exercises in estimating global poverty. What follows is a discussion of thisresponse. As will become apparent to the reader, I find little cause – despite the clarifications provided by Ravallion in his “Reply” – for viewing the World Bank’s approach to measuring world poverty with any less scepticism than the EPW editorial does. On the International Poverty Line Ravallion disagrees with the view that “…‘dollar a day’ (or thereabouts) poverty lines may well be little more than ‘destitu-tion lines’’’. He points out (as the editorial itself indicates) that the World Bank’s international poverty line (IPL) is based on the national poverty lines of a set of poorest countries of the world. Let us set aside the fact that at different points of time there have been differences in the numbers and composition of these poor-est countries, as also in the basis of the relationship between the IPL and country-specific poverty lines. At a more basic level, it is not clear why it is the poorest countries’ poverty standards that must serve as the poverty standard for the world as a whole. There are two kinds of classificatory errors one can make – call these Type-I and Type-II errors, respectively (see also, in this connection, Cornia and Stewart 1995). A Type-I error is one in which a person is wrongly counted as non-poor, and a Type-II error is one in which a person is wrongly counted as poor. One would imagine that a Type-I error is the graver sort of mistake to commit. In a matter as serious as identifying a person’s poverty status, if one must make a mistake, then it would appear to be doing less harm if one were to err (within limits, of course) on the side of generosity than on the side of niggardliness. Thus, ifzi is the national poverty line of the ith ofn countries con-stituting the world, then a procedure aimed at avoiding a Type-I error would pitch theIPL at the level z, where z≡ maxi {zi}, that is, z is the highest of the n countries’ national poverty lines (and is likely to be the poverty norm of the richest of theworld’snations). The World Bank’s procedure, on the other hand, seems to be geared to avoiding a Type-II error, andis close in spirit to pitching theIPL at the level z, where z ≡ mini {zi}, that is, z is the lowest of the n countries’ national poverty lines (and is likely to be the poverty norm of the poorest of the world’s nations). Even a procedure which avoids both extremes would endorse an IPL higher than $1.25 per person per day at 2005PPP: by Ravallion’s own estima-tion, the average of all except the poorest countries (from, presumably, a compila-tion of 75 national poverty lines) is twice as high, at $2.50. The average official poverty line in the United States, Ravallion tells us, is $13 a day. The World Bank’s prescribedIPL of $1.25 a day is less than 10% of the US poverty line. I imagine an American living in, say, California, on $1.25 a day at 2005 purchas-ing power parity (PPP) is entitled to feel s/he is living in circumstances of destitution. It seems unlikely that the World Bank would be disposed to correct the perception of such a hypothetical Californian by coun-selling calm reasonableness to her/him. It is then fair to entertain the expectation that the World Bank should adopt the same stance towards a very real Sierra Leonean or Bangladeshi or Ecuadoran living on $1.25 a day at 2005 PPP. After all, the World Bank is supposed to be a world bank. It is easy enough to dismiss this as virtuous counsel on a matter of manners. It is, however, fundamentally a matter of logic that is involved – as should be apparent if we were to take seriously what Ravallion (2004: 15) himself has said of
DISCUSSIONjanuary 31, 2009 EPW Economic & Political Weekly68the rationale underlying the World Bank’s IPL: “For our global poverty counts, we have but one overriding concern – that two people with the same standard of living, measured by their command over com-modities, be treated the same way no matter where they live.” In light of this, I should think Ravallion really requires no assist-ance in answering his own query, which he raises in his “Reply” (p 78): “… India’s official poverty line is about $1.00 a day using the same PPPs. So if $1.25 marks ‘destitution’ what are we to make of India’s national poverty line?”It is, nevertheless, instructive to dwell a bit on India’s official poverty line. In this connection, it is relevant to consider a particular construction of the dollar-a-day IPL due to Ravallion (2008a): ‘The “$1 a day” line aims to judge poverty in the world as a whole by the standards of what poverty means in poor countries’. This sentiment suggests that what poverty does, or should, mean in poor countries is credibly reflected in the national poverty linesof these countries. It is not self-evidently obvious that such is indeed the case, nor – if this is an intended interpretation – that the national poverty lines of poor countries invariably reflect perceptions of poverty that are a product of demo-craticconsultation or informed public rea-soning. (Such an interpretation certainly seems to be stretching it a bit when the poverty lines in question have been pre-scribed by the World Bank itself.) In fact, it is well known that – since the head-countratio of poverty is a non-decreasing function of the poverty line – most govern-ments, especially the governments of poorer nations, would be reluctant to set the poverty line at realistic (leave alone generous) levels. India’s is a particularly salient case in point.Back in the early 1960s, the Perspective Planning Division of the Planning Com-mission put out an important document titled “Perspective of Development: 1961-1976” (a reprint of which is available in Srinivasan and Bardhan 1974). The “Per-spective” makes what amounts to a can-did admission of the extent of poverty which India could “afford” at the time. It was calculated that the balanced diet recommended by the Nutrition Advisory Committee, supplemented by a modest allowancefor non-food requirements, yielded a poverty line of Rs 35 per person permonthat 1960-61 prices. A further calculationrevealed that this poverty line would plunge some 80% of the Indian population into poverty. Apparently, wiser – or at any rate, more sober, and cautious, and politic – counsels prevailed: a Work-ing Group set up by the Planning Commis-sion recommended in July 1962 that the poverty line should be pitched at Rs 20 per person per month at 1960-61 prices.Subsequently, in the mid-1980s, the poverty line for 1973-74 was fixed at that level of consumption expenditure at which a calorific norm of 2,400 (respectively, 2,100) kilocalories per person per day in rural (respectively, urban) areas was ob-served to be realised. This yielded a rural (respectively, urban) poverty line of (roughly) Rs 49 (respectively, Rs 56) per person per month for rural (respectively, urban) India at 1973-74 prices. The official practice has been to “update” these pov-erty lines in subsequent years through employment of suitable consumer price indices. It is no longer news that these poverty lines have, over time, displayed a steady “calorific drift”, that is, at the officially stipulated poverty lines, the calorific consumption, over time, has steadily fallen further short of the original 2,400 and 2,100 kilocalories norms: the “anchoring” of the poverty line in a nutri-tional standard is therefore a moot point. There is a vast literature on this subject, and more generally on the conceptual and methodological inadequacies of the Plan-ning Commission’s procedure of identify-ing the poverty line, a literature which, as it happens, also has implications for an understanding of the debate on the World Bank’s poverty estimates. A restricted sample of the India-related literature would include Nayyar (1991), Suryanarayana (1996, 2000), Panda and Rath (1999), Mehta and Venkatraman(2000),Meen-akshi and Vishwanathan (2003), Patnaik (2004, 2007), Ray and Lancaster (2005), Subramanian (2005), and Reddy (2007). I cannot insist, but I believe it would not beilliberal on my part to suggest that an acquaintance with some of this literature might actually be helpful in acquiring an appreciation of the reliability of official Indian estimates of poverty. The official poverty line for urban Tamil Nadu in 2004-05 was (roughly) Rs 547 per person per month, or (without allowing for any economies of scale), roughly Rs 2,200 per four-member household. I happen to know that, in my own city of Chennai, a modest low-income two-room flat with attached bathroom and kitchen would have commanded a rent of at least around Rs 1,500 per month in 2004-05. A modest meal at the local tea-shop would have cost Rs 10 per head; assuming two such meals per head over the entire day to be adequate nutritional intake, the monthly cost of food for a four-member household would be Rs 2,400. Let us, quite arbitrarily and drastically, cut this figure in half, to allow for the fact that home-cooked food is cheaper: the monthly cost of food would then be Rs 1,200, and the monthly require-ment for meeting the needs of shelter and food together would work out to Rs 2,700. This – if the poverty line in 2004-05 were pitched at the official level of Rs 2,200 per month for the household – would have left a good deal less than nothing to spendon education, clothing, and trans-port, not to mention other exotica, like films or an occasional drink. Of course, a fractured arm, an episode of leptospirosis, any other incidence of morbidity or sick-ness, would have had to be treated as an unaffordableluxury, if not an outright sinful indulgence. Using Ravallion’s own (2008b) conver-sion rates, anIPL of $1.25 in 2005 PPP would correspond to a rural (respectively, urban) all-India poverty line for a four-member household of Rs 1,709 (respec-tively, Rs 2,584), which is more – but not vastly more – generous than the Planning Commission’s allowances. When we deviate a little from such stringency, we discover what the National Commission for Enter-prises in the Unorganised Sector (2008) discovered: namely, that – effectively – if we employed a national poverty line of Rs2,400 per month for a four-member household (corresponding to Rs 20 per person per day) across the board for India in 2004-05, then the proportion of the population in poverty would be 77%. This is rather more serious than the headcount ratios corresponding to the Indian Planning Commission’s norms (27.5%) or the World Bank’s norm (42%).
DISCUSSIONEconomic & Political Weekly EPW january 31, 200969Finally, in speaking of “destitution lines”, one is not committed, in a literal-minded way, to what Ravallion calls “a life-threateningly low level of living” (p 78) note that this is one dictionary definition of “destitution”: Merriam-Webster Online refers to destitution as “such ex-treme want as to threaten life unless re-lieved.” However, there are alternative ways of conceptualising destitution, as pointed out by Devereux (2002). In one such conceptualisation, which Devereux (op cit: 10) attributes to Craig and Potter (2003), “…lack of access to affordable, effective healthcare is dreaded, not just as a source of ‘ill health’, important though that is, but as a source of vulnerability and, ultimately, destitution”. Such a con-ceptualisation, in view of the discussion, in the preceding paragraph, of official Indian and World Bank poverty lines in relation to the requirements of meeting fairly basic needs, renders the use of the term “destitution lines” a not extravagant description even by the Merriam-Webster definition of the word in question. The term acquires even greater plausibility with reference to the construal by Dasgupta (1993: viii) of destitution as “an extreme condition of ill-being”, or by the American Heritage Dictionary of the English Language as “a deprivation or lack; a deficiency”. In any event, the issue does not turn on a semantic dissection of the word “destitu-tion”, so much as on the socio-economic import of the notion, read in context. In the end, the issue reduces to advancing the claim of a simple but serious point: in fixing poverty lines – an exercise which requires us to assess the needs of the poor – let us, please, not be excessively stoical and Spartan on behalf of the poor. PPP CalculationsRavallion is right to point out that the latest World Bank estimates of global poverty have not been totally impervious to the possible merits of what one may call “PPPs for the poor” (or PPPPs, for short). From what one can understand, the conversion rates employed continue to bePPPs which, apparently on the basis of some applied sensitivity analysis, have been judged to be not unacceptable proxies for PPPPs. Overall, one senses that Ravallion is in agreement with the notion thatPPPPs, if faithfully implemented, would constitute an improvement on PPPs. Having said this, it may be recalled that theEPW editorial had drawn reference, following the criticism made by Reddy and Pogge (2008), to two problems of cross-section comparability occasioned by the use of PPPs, which may be summarised as, respectively, the “irrel-evant countries problem” and the “irrele-vant commodities problem”. The use of PPPPs in place of PPPs would, in some measure, address the “irrelevant com-modities problem”: as such, it would at best (and subject to proper implementa-tion) be a partial response to the issue of cross-sectional comparability, geared to-ward mitigation, rather than elimination, of a difficulty. More fundamentally, given that the use of price indices is seldom unproblematic, there is a case for employing a defensible procedure of poverty comparisons that does not rely on the employment of price indices at all. This focuses directly on the issue of the “common standard” according to which poverty comparisons are most meaningfully effected. Reddy and Pogge (2008), Reddy (2004), and Subramanian (2005, 2007) all point to the conceptual soundness of locating the “common stand-ard” in an identified set of basic human capabilities to function (as opposed to al-ternative proposals that have been made of locating the common standard in an identified commodity bundle or an identi-fied level of “real income”). As Sen (1983) has pointed out, deprivation is “absolute” in the space of functionings but may be “relative” (given the possibility of inter-personal variations in the ability to convert resources into functionings) in the space of resources or incomes. For any given region at any given point of time, the poverty line can be taken to be the mone-tary resources that will be needed to achieve some minimallyacceptablelevel of functionings deemed necessary to avoid capability deprivation. This could be Rsx in India and $y in theUS (per appropri-ately defined unit of time). If the world consisted only of these two countries, the global headcount of poverty would be the sum of the Indian population with in-comes less thanRsx and the US popula-tion withincomes less than $y. The proce-dure, as one can see, does not require theuse of any price indices. I am not suggesting this is a practically easy route to identification, but this fact does not entail accepting that practically feasible solutions which are conceptually unsound constitute a satisfactory resolution of the problem. Inter-Temporal ComparisonsOn theEPW editorial’s criticism relating to a problem of consistency with the World Bank’s methodology when it comes to making inter-temporal poverty compari-sons, Ravallion (2008b: 78) says: Our method – let us call it method A – entails converting the international poverty line to local currencies using the PPP at the bench-mark year… (now 2005) and then adjust-ing over time using the country’s consumer price index (CPI) to get to the prices prevail-ing in the relevant survey year. The editorial claims that it would only be a “fluke” if our method gave the same result if instead we had used the obvious alternative, method B, of converting all country-distributions to $ and then made the poverty calculations adjusting for inflation in theUS. It is not clear why methodB is the “obvi-ous alternative” to method A, nor that the editorial has claimed that it is only by fluke that methods A andB will “give the same result”! This makes it difficult to agree with the charge of a mistake having been made in the editorial (Ravallion 2008b: 78-79):The mistake made in the editorial (and by one or two other critics) is not to realise that if one uses method B then the PPP for the non-benchmark year must be calculated consist-ently with the differential rates of inflation between the US and the country in question.Specifically, theEPW editorial did not undertake any comparison between methodsA andB. For, why would method B be the “obvious alternative” to methodA, unless inflation in theUS were the same as inflation in every other country? And if such an identity of inflation rates did in fact obtain, thenB would not be an “alternative” toA, B would just be indistin-guishable from A! When Ravallion (2008b: 78-79) says that “the PPP for the non-benchmark year must be calculated con-sistently with the differential rates of in-flation between the US and the country in question”, one takes it he means the fol-lowing. Suppose in the benchmark year –
DISCUSSIONjanuary 31, 2009 EPW Economic & Political Weekly70Table: MDG Achievement at Different International Poverty LinesIPL Level in 1990 Baseline 2005 Target Annual Reduction Reduction Actual Reduction How Is the World Doing in 2005 Dollars (Millions of Poor) Reduction of Needed to Needed to be Achieved Regard to MDG-1? 27.5% Reach Target ‘On Track’ in 2005 1990-2005 (100% = Exactly on Track) (Millions)(%)(Millions)(Millions) $1.00/day 1,303.2 358.4 1.28 228.7 424.2 185%, much ahead of schedule$1.25/day 1,817.5 499.8 1.28 318.9 417.9 131%, ahead of schedule$2.00/day 2,753.6 757.2 1.28 483.2 155.8 32%, much behind schedule$2.50/day 3,076.6 846.1 1.28 539.9 -63.6 -12%, regressingSource: Pogge (2008) call itti – 1 US$ ≡rupeesx in terms of the PPP exchange rate. Suppose in some other, non-benchmark year – call itt2 – thepricelevel has risen by 20% vis-à-vis t1 in theUS and by 25% in India.Ravallion’s requirement, it seems, is that in yeart2, one must take it that 1.2 US$ ≡ rupees 1.25x, or (given that 1.25/1.2 = 1.042), that 1US$ ≡ rupees 1.042x. The EPW editorial’s claim is that it is only by fluke that theactualPPP int2 will turn out to be given by 1 US$ ≡rupees 1.042x. The problem of consistency alluded to in the editorial is now simply explained. What if a survey was conducted in yeart2, and it was dis-covered that thePPP exchange rate forthis year turns out to be: 1 US$ ≡ rupees1.05x? Ravallion (2008b: 79) effectively allows this possibility when he says: “Granted, there may well be an inconsistency if one compares PPPs obtained from different benchmark years, for then consistency of thePPPs with the differential ratesofin-flation from the nationalCPIs is not as-sured (given differences in the qualityof goods used and the weights)”. How then is the problem to be resolved? “We follow standard practice in other international comparisons of only doing the PPP conver-sion for the benchmark year, and then making inter-temporal comparisons using the national data” (ibid). What is being recommended, effectively, is that one must reject thet2 survey result in the cause of “consistent calculation”. But then, what if there is no reason to question the accu-racy of either thet1 or t2 surveys? Must we sacrifice fact for consistency? If that is what is being recommended, then it is no longer true of t2 that the poverty lines for theUS and for India in dollars and rupees respectively reflect identically the same purchasing power – which violates pre-cisely the cardinal property attributed to itsIPL by the World Bank. (I may add that this problem of consistency would not ever “surface” if one had information on PPP exchangeratesonly for the benchmark year; but then, empirical non-verifiability is not proof of the absence of analytical inconsistency!)Of course, I do not dispute that an in-consistency will arise from comparison of PPPs from different benchmark years – let us call this mode of inter-temporal com-parison methodC. Under methodC, we will have to forsakehorizontal consistency – by which I mean that the actual relative rates of inflation between the US and the country in question will not be reflected in the PPPs of the years under comparison. Under methodA, as we have seen, we will have to forsakevertical consistency – by which I mean that if the actual relative rates of inflation between the US and the country in question are preserved, then in the non-benchmark year the calculated PPP between theUS and the country in question may not coincide with the actual PPP. It is in no way incidental to the prob-lem of “poverty identification” that failure to fix the poverty line in relation to some level of income explicitly derived from a common standard of human requirements, is linked to the problem that the analyst is constrained to resort to one or theother of Methods A andC of inter-temporal poverty comparison. Both methods, un-fortunately, lead to inconsistent outcomes. One does not have to choose between the two inconsistencies, unless one has opted, for whatever reason, to confine one’s uni-verse of available identification-related alternatives to MethodsA andC. Indeed, those not committed to “dollar-a-day” type international poverty lines are not, as a matter of logic, required to engage with this issue: in general, in a just world, one should be spared the necessity of deal-ing with a problem which is not of one’s creation! On Artefacts and Reality In the Editorial, it was observed that…it turns out that the lower the poverty line employed, the more flattering is the result-ing trend decline in poverty. This suggests that the trends we obtain are not so much a reflection of what is actually happening to poverty on the ground, as a verification of the consistency of arithmetic. Ravallion interprets this as pointing to the possibility of exploiting access to the a priori knowledge that as poverty tends to either zero or infinity the rate of poverty reduction will also tend to zero. This strikes me as being a simplistic interpreta-tion. A more substantively relevant con-sideration would be a tendency reported in Pogge (2008), and echoed in Reddy (2008): this is reflected in the table repro-duced from Pogge (2008) (itself based on data provided in Chen and Ravallion 2008), and my suggestion is that the reader should find it intriguing.Ravallion states: “The fact that we find that the rate of progress against poverty is slightly higher for $1 per day than $2 per day is hardly a sign that our results are arithmetic artefacts (2008b: 79).” I submit that the choice ofIPL is not anything like so innocuous as this statement might suggest. The table reveals that with anIPL of $1 (2005 PPP), the world can claim to be 85% ahead of schedule in realising the Millennium Development Goal-1; but with anIPL of $2, the world would have to be found 68% behind the warranted sched-ule for meeting MDG-1. The level at which theIPL is pitched is nothing if not crucial, and given a justifiable absence of extreme aversion to a Type-I error, the $2.50 IPL should be seen to be more reasonable than the $1.25 IPL.Or so one could argue if one’s interest in the matter led one in that direction, and if one also suspected that the cumulative density function (cdf) of income had a tendency to behave over time as shown in the figure (p 71). If the rate of growth of low income levels were faster than that of high income levels, then it is conceivable that the cdfs in an initial periodt1 and in a later periodt2 could be as pictured in the figure. If one knew or suspected this to be the case, and if one also wished to demon-strate a deterioration in poverty, then one would have a strategic reason for pitching the poverty line at a “high” level like zH, to the right of z*, in the figure – and one
t1 t2 t1 t2 Income

Dear Reader,

To continue reading, become a subscriber.

Explore our attractive subscription offers.

Click here

Back to Top