The announcement in June by the Government of India’s Press Information Bureau (PIB) that in new estimates by the World Bank (WB) poverty in India had dramatically fallen and India was now one of the most equal countries in the world caused much puzzlement and wonder. The Government of India and its supporters lost no opportunity to tom-tom this progress in the war against poverty.

How accurate and reliable though are these numbers? One has to dig deep into the methodology and data underlying these estimates to understand how the WB has come up with its latest numbers on poverty and inequality. What one finds at the end are inexplicable changes to past methodology and the use of indefensible assumptions, which raise questions about the credibility of these estimates. This is the result in part of the Government of India abdicating its responsibility and not updating national poverty estimates, contributing to the unreliability of the WB’s numbers.

A careful study of the data and methodology is therefore required. The innumerable changes and assumptions made at all levels make this a difficult task and not for the faint hearted to understand. But this task has to be done; for unless they are questioned the World Bank and the Government of India will use the denseness of the underlying methodologies to keep publicising these numbers.

 World Bank’s Estimates of Poverty

For some decades now, the WB has been monitoring poverty and inequality for several countries as part of its mandate and puts out periodic updates. It uses the Purchasing Power Parity (PPP) exchange rate of the International Comparison Programme (ICP) to make comparisons between countries. The latest update put out in June 2025 is based on the revised WB international poverty line of $3 a day for extreme poverty, which uses the PPP exchange rate based on the 2021 cycle of the ICP. According to these new estimates, poverty in India is estimated to be 5.25% in 2022-23, declining from 27.12% in 2011-12. In terms of population this translates into 75 million poor in 2022-23, compared to 344 million poor in 2011-12.

However, this was not the only estimate that the WB released this year for India. Less than six weeks earlier, the WB had released estimates of poverty for India based on the $2.15 per day poverty line, drawn up with the PPP exchange rates of the 2017 ICP cycle. These April 2025 estimates were put out during the annual Spring Meetings of the WB and the International Monetary Fund. They suggested a different set of estimates of poverty in India though still showing a sharp decline: 2.35% in 2022-23, down from 16.2% in 2011-12. In terms of the number of poor, the April 2025 estimates showed a decline from 206 million poor in 2011-12 to 33.7 million in 2022-23.

Both these estimates were also released by the PIB as vindication of the poverty reduction efforts of the present government. The PIB subsequently released another press release also based on the WB estimates claiming India to be the fourth most equal country among all countries for which the WB maintains and releases data. The inequality estimates are also part of the regular revisions by the WB. The endorsement by the Indian government of the WB estimates ensured that these were covered widely by the Indian media.

Laying out the Issues

There are three major methodological sets of issues and problems which undermine the quality of the WB’s recent estimates.

The first is the comparability of the consumption expenditure data that the National Sample Survey (NSS) collects, which form the basis for the WB’s numbers. The Household Consumption Expenditure Survey (HCES) of 2022-23 conducted by the NSS 1The HCES data and report is now available for 2022-23 and 2023-24. However, the WB estimates are only based on 2022-23 data.  is the first consumption expenditure data available for India after 2011-12. However, there have been a series of independent assessments by scholars reaching a unanimous conclusion that the 2022-23 data on consumption expenditure is not comparable with the earlier NSS rounds and therefore not suitable for tracking poverty over time. (See Anand 2024, Mehrotra and Kumar 2024, Sinha 2025, Subramanian 2024, Waghmare 2024, Manna 2024 and Himanshu et al 2025.)

The WB acknowledges the issues of non-comparability but it has yet gone ahead with preparing its latest estimates of poverty for India by using the HCES 2022-23, after noting a caveat 2Section 3 of the methodological note on India poverty measurement discusses some of the issues along with the caveat. See (World Bank 2025a). .

The second set of issues relates to the various adjustments that the WB has done to the data, adjustments which, we will see, are not warranted and contribute to the estimation of low poverty numbers and low levels of inequality. The WB’s estimates are also problematic because they are inconsistent with the Bank’s own past practices and principles used to estimate poverty for various countries, including India. These relate to issues in the use of PPP exchange rates but also to the choice of consumption aggregates used to measure poverty and inequality. These are also at variance with the WB’s own guidelines laid out in the Commission on Global Poverty chaired by Tony Atkinson (World Bank 2016, also known as the Atkinson commission report), the last comprehensive set of guiding principles for global poverty measurement by the World Bank 3The Commission on Global Poverty (World Bank 2016) chaired by Tony Atkinson laid out the basic principles and guidelines that the World Bank should follow for maintaining and updating the global poverty estimates. .

The issues raised by the WB’s methodological changes and adjustments to the NSS data are relevant not only for the quality of its estimates for India but also raise fundamental questions about the WB’s measurement of poverty in many developing countries.

The third set of issues has to deal with the Government of India’s own inability to update the national poverty line. Poverty estimates are based on a poverty line which has to be periodically revised based on the latest available consumption expenditure data. But no such revision has been made for many years. There is no official poverty line that can be applied to the HCES 2022-23 to estimate poverty in India. The last official committee on poverty lines was the Suresh Tendulkar committee (2009) which, suggested the Mixed Recall Period (MRP)-based poverty line. Subsequently, another committee was with C. Rangarajan as chairman was set up. This committee h suggested the use of the Modified Mixed Recall Period (MMRP)-based consumption expenditure for poverty estimation. (See Appendix for a description of the different recall periods used by the NSS in collecting information on consumption from respondents.) Unfortunately, the recommendations of the Rangarajan committee have not been accepted by the government. This means there is no officially accepted poverty line available that is based on the MMRP recall period and can be applied on the 2022-23 HCES consumption expenditure data.

The endorsement of the WB estimates has allowed the government to escape its own obligations of setting the poverty line and releasing official poverty estimates. A blind endorsement of the methodology used by the WB undermines the well-established principles in India of poverty measurement. The issues raised by the WB’s methodological changes and adjustments to the NSS data are relevant not only for the quality of its estimates for India but also raise fundamental questions about the WB’s measurement of poverty in many developing countries.

These three sets of issues are discussed in detail below.

What is New in the WB Poverty Estimates

The June 2025 updates to the global poverty estimates are part of the regular updates of estimates of poverty and inequality carried out by the WB on its Poverty and Inequality Platform (PIP). The international poverty lines suggested by the WB are presented in United States Dollars (USD). The USD poverty lines are converted into national currency poverty lines using the PPP estimates from the ICP. Major updates to the WB poverty estimates are necessitated as and when new rounds of ICP PPP data are available.

The WB uses three different poverty lines, each for countries grouped according to three levels of per capita income as measured by the PPP exchange rates. With the ICP 2017 PPP, the poverty lines used by the WB were $2.15 for low-income countries, $3.65 for lower middle-income countries and $6.85 for the upper middle-income countries. The latest round of PPP estimates by the ICP are available for 2021, which are an update from the 2017 PPPs that the WB was using so far. With every change in the PPP, the global poverty lines are also revised. The revisions are required given the way the WB arrives at these poverty lines: the international poverty lines are typically the median of the PPP converted poverty lines for countries in each group of countries. With the new ICP PPP measures, country groupings change and so therefore do the poverty lines applicable to these groups of countries.

The latest revision of international poverty lines based on the ICP 2021 has led to the WB revising the three poverty lines to $3 for the low-income countries, $4.20 for the lower middle-income countries and $8.30 for the upper middle-income countries (see Foster et al 2025 for details). The poverty line for the lower-income countries has been increased the most (40%) whereas the increase for lower middle-income countries has been the least. An important source of the changes arises from the fact there have been changes in the national poverty lines for the countries in each of the groupings.

As the WB’s methodological note suggests, the entire decline in South Asia is driven by updated poverty estimates for India.

The changes imply an increase in the global population living under extreme poverty for 2022 from 713 million based on earlier estimates to 838 million using new estimates. The revisions in international poverty lines have led to the population living under extreme poverty increasing for all regions of the world for which WB maintains the estimates -- except in South Asia. Even the poverty headcount ratio has seen an increase for all regions, except in South Asia. The overall poverty ratio for South Asia witnesses a decline from 9.7% to 7.3%, with the number of poor under extreme poverty declining from 186.2 million to 141.5 million.

Apart from new and updated national poverty lines for several countries, another significant change in the June 2025 revision is the inclusion of new data for India. As the methodological note suggests, the entire decline in south Asia is driven by updated poverty estimates for India.

WB Estimates of Poverty and Inequality for India

India has been part of the WB global poverty estimates since when these estimates were first prepared. India’s official estimates of poverty were based on the thick round (quinquennial) consumption expenditure surveys of the then National Sample Survey Organisation (NSSO, now renamed as the NSS). Until now, the consumption expenditure survey which was officially accepted by the Government of India and used by the WB for its poverty estimates was carried out in 2011-12. A consumption expenditure survey was conducted in 2017-18 but it was junked by the government after its report was leaked in the media. In the absence of data after 2011-12, until now the WB was using different alternatives to estimate poverty for India. 4WB estimates of poverty for India after 2011-12 were based on survey-to-survey imputation-based estimates. See Sinha Roy and Van der Weide (2022) for details.  Those WB poverty estimates for India were arrived at using data from the private sector generated Consumer Pyramid Household Survey (CPHS) (Sinha Roy and van der Weide 2022). According to these estimates, poverty in India using the $2.15 poverty line was estimated at 22.9% for 2011-12, which declined to 13% in 2021-22 with the number of poor declining by 107 million.

The availability of the consumption expenditure data from the HCES 2022-23 has led to the WB using this data for its June 2025 revisions (and earlier for the April 2025 revisions). The WB acknowledges the serious issues of non-comparability of the HCES 2022-23 with the earlier data on consumption expenditure. These are primarily to do with the changes in sampling design, aggregation of items, and methodology of data collection. In the HCES 2022-23, data were for the first time collected by splitting the consumption survey into three rounds with each of them canvassed in separate visits, separated by a monthly interval. This method of data collection is unique not only for India but also internationally. All of these issues raise questions on the usability of the HCES 2022-23 for any inter-temporal comparison of poverty in India.

Most of these issues are highlighted by the WB in its methodological report for India released in May 2025 (World Bank, 2025a). However, it yet decided to go ahead with using the same HCES data. This is surprising, especially given that the WB has hitherto been careful about not including a national consumption/income survey if there are issues of comparability (See World Bank 2025b). There is a precedent. The WB had decided to exclude the 55th round of the NSS consumption expenditure survey (1999-0) in India from its poverty estimates, following the controversy on comparability of that survey with earlier ones 5Even today, there are no estimates of poverty and inequality for 1999-00 in the WB PIP platform for 1999-00 for India. . But in the case of HCES 2022-23, for its own reasons it has made an exception.

While the decision to use the HCES 2022-23 itself is problematic, the WB has made changes as well in the way the consumption aggregates are used for poverty and inequality estimations. This again is a departure from past practice of the WB.

Adjustments of Indian data from HCES 2022-23

Nation-wide survey data on consumption expenditure that have been collected in India from the 1950s onwards have hitherto been regarded very highly for the quality of collection and robustness of the estimates of consumption expenditure. India being a pioneer in collection of data on consumption expenditure, several developing countries have adopted similar methodologies to conduct their surveys on consumption.

As we have seen, these have also been part of the WB’s estimates of global poverty, except for years when there were issues of comparability such as in 1999-00. The NSS data have so far been used without any adjustment and modification of the aggregate consumption data as available in the unit records or from the NSS reports. Until 2022-23 HCES, the WB had used the URP estimates of the consumption expenditure survey for estimating poverty and inequality. These are available until 2011-12 with the HCES 2022-23 shifting to MMRP-based consumption expenditure estimates. (The difference between URP and MMRP estimates are important to understand if we are to understand the changes made by the WB to its poverty estimates for India. What the URP, MRP and MMRP are have been detailed in footnote 4 above)

For the June 2025 update, the WB has decided to use an adjusted estimate of consumption expenditure from the HCES 2022-23. There are four adjustments that the WB has made. These relate to how the WB treats the expenditure on (1) durable goods,( 2) hospitalization, taxes, and a select set of non-food items, (3) housing rentals and (4) treatment of subsidized or free items.

The apparent reasoning behind making changes in the way these items/item-groups are treated in estimating consumption expenditure of households is the shift from consumption expenditure to real welfare aggregates as the measure of income/consumption (World Bank 2025a).

This concept of “welfare aggregate” that the WB has used in its June 2025 revisions to measure poverty in place of consumer expenditure is a departure from its own practice of measuring poverty and inequality. It was neither recommended by the Atkinson commission report (World Bank 2016) which examined the issues in global poverty measurement and lay down guidelines for best practices, nor is it consistent with its own practices. It appears to be applied only to India and not to other countries, posing questions about comparability of the WB’s India estimates with those for other countries. The WB note makes no mention of applying the same concept to measurement of poverty to other countries (for many of which such data may not be available).

The idea of using welfare aggregate as against consumption expenditure as the metric on which poverty is estimated is not new and has been advocated by Hentschel and Lanjouw (1996), Deaton and Zaidi (2002) and more recently by Mancini and Vechhi (2022). While all of them agree on consumption expenditure as the parameter for poverty measurement, differences arise on the components of consumption to include while constructing the consumption expenditure variable.

This concept of “welfare aggregate” that the WB has used in its June 2025 revisions to measure poverty in place of consumer expenditure is a departure from its own practice of measuring poverty and inequality.

Conceptually, welfare represents the value of goods and services actually consumed by the households as against consumption expenditure which is the money the household spend on purchasing all the items of consumption. The difference between the two arises in case of goods that are purchased but not entirely consumed by the household in the reference period. This distinction applies to all the goods and services purchased by the households, but its application is empirically difficult, as most household surveys only collect information on expenditure rather than on actual consumption of the goods purchased. Moreover, if the idea of poverty is based on the money-metric value of goods and services commanded by the household as a proxy of disposable income, then the relevant category is consumption expenditure and not welfare.

Despite the academic literature, this distinction was found to be irrelevant for measurement of poverty and therefore does not find any mention in the 2016 Atkinson commission report. Further, despite its intuitive appeal, the empirical difficulty of obtaining data on actual consumption as against total expenditure meant that most countries have continued to use consumption expenditure as the metric for poverty measurement.

Inclusions and Exclusions

The logical justification for inclusion of free and subsidised goods and exclusion of expenditure on the first three categories listed above in this measure of the “welfare” aggregate is very much debatable. 6In the India methodological note (Section 4, page 79) “Theoretical considerations on welfare measurement” describes the shift in approach. It argues, “Using consumption as a metric for individual welfare requires adjusting household consumption expenditure to account for welfare-improving non-monetary benefits and household composition. For example, expenditures on debt repayments or medical emergencies do not enhance welfare but reduce the disposable incomes for spending on welfare-enhancing consumption. On the other hand, access to public services like free healthcare, subsidized food, or education can reduce private spending, meaning lower expenditures may not always reflect lower welfare. By excluding certain welfare-reducing expenditures and incorporating non-monetary benefits, the welfare aggregate offers a more accurate and policy-relevant measure of poverty, ensuring that interventions effectively target those most in need.” The explanation equates spending on debt payments to medical expenses as non-welfare enhancing expenditures. This in itself is problematic. While debt payments do not enhance welfare, medical expenses are welfare enhancing. It certainly improves the health status of a person and thereby improves welfare. It is absurd to argue that expenditure on medical emergencies reduces other welfare enhancing expenditures and therefore should be excluded.  Why the adjustments have been made now is inexplicable. Almost all of these categories of expenditures have existed in the NSS consumption surveys since the very first survey in the 1950s with the NSS having well-defined rules for inclusion and exclusion in aggregate household consumption expenditure. These were also accepted by the WB until the latest revisions.

Inclusion of Government Transfers in Private Expenditure

Perhaps the most controversial of these changes is the way WB now treats consumption of free and subsidized goods and services in the consumption surveys. The usual practice of the NSS and in majority of the countries where surveys of consumption expenditure have been carried out is to exclude these from the expenditures of households as these households do not make outlays on them. However, the NSS does make an exception for free goods and services received by the households as part of wages as these are income in kind 7Since these are incomes earned in-kind, these are not transfers but part of household income. See Deaton and Zaidi (2002). . All other free transfers including those from the government were not included as these were surveys of expenditure and no outlay was incurred by households on these items.

Free goods and services have been made available by governments to Indian citizens for a long time. These include food as well as non-food items given free or at subsidised rates by state governments as well as the central government. Among these, free cereals (mainly rice and wheat) have been provided by many states for some time now. With the central government also making rice and wheat sold through the Public Distribution System (PDS) free during the COVID pandemic, a significant proportion of the Indian population received free foodgrains. This scheme, as part of the National Food Security Act (NFSA), has now been extended until 2028. Consumption surveys by the NSS had a separate item group for items received from the PDS. However, these were valued at the prices paid by the consumers including zero prices if received free of cost. Since HCES 2022-23, the NSS has introduced a separate item code for food grains and non-food items received free of cost. The food items now include coarse cereals, pulses, salt, sugar and edible oil. The NSS also collects information on free provision of non-food items such as laptops, mobile phones, tablets, bicycles/motorcycles, clothing and footwear that are part of school uniforms.

As Ravallion (1998) and Deaton and Zaidi (2002) pointed out, consumption expenditure of households represents the money-metric utility underlying welfare measurement of households 8In fact, Deaton and Zaidi (2002) explicitly argue against using the value of free transfers in the value of consumption expenditure. Hentschel and Lanjouw (1996) and Mancini and Vechhi (2022) also recommend exclusion of free transfers from consumption expenditure, particularly for goods which are rationed. . For any money-metric measure of poverty what matters is the household’s actual expenditure on goods and services.

Inclusion of transfers from the government is a broader measure of welfare. Any change in the extent of poverty by the inclusion of such transfers is then not just the change in households’ own income/consumption and thereby command over commodities, but also reflects changes in government expenditure. Any change in poverty then becomes not just a function of changes in income of households but also of government expenditure.

These adjustments have not been done for all the countries… raising questions about why the WB has done so for India.

This approach is also problematic since the choice of which goods and services to impute and include in the consumption expenditure is an arbitrary one. Thus, the WB, for its part, has imputed and added the value of free food grains along with clothing and footwear provided free to school children, but excluded free goods such as laptops, bicycle, mobile and so on. On a fundamental level, all services that the governments provide are also welfare enhancing and therefore there is a case to include the value of free education, free health services, clean air, better law and order and so on in the “expenditure” of households. Further, since consumption expenditure will then include not just expenditure on goods purchased but also on services, the selective imputation of free and subsidised goods for certain items defies logic.

Even for its latest revision of poverty numbers, while the WB has included free uniforms and footwear in 2022-23; these are not included in the estimates for 2011-12 since this information was not available for that year. Although the share of free uniforms and footwear in total consumption expenditure is small, the issue is of non-comparability with previous estimates. These adjustments have not been done for all countries since most countries do not include data on free transfers and the value of such transfers in the consumption expenditure, raising questions about why the WB has done so for India.

The issue of inclusion of free transfers is not entirely new. Since the absolute poverty line refers to the cost of basic needs defined in terms of quantities valued at prevailing prices, any assessment of welfare should exclude free transfers as no price is being paid. This has been the approach taken so far by the WB and also in official poverty estimates of India. The same practice is followed in most of the developing world. However, as the WB report argues, a move from consumption expenditure to welfare would require the inclusion of these transfers. But then it is imperative that the poverty lines used to measure poverty on welfare are constructed to account for the free transfers at the implicit prices. An updated poverty line that included the value of imputed transfers should ideally leave the poverty rate unchanged compared to a poverty line without free transfers valued in private consumption expenditure. Himanshu and Sen (2013) use such a methodology to examine the impact of free or subsidised transfers to households. However, this is not the case in most countries whose poverty lines are used to arrive at the global poverty line. In the absence of a welfare metric poverty line, the choice is limited to using consumption expenditure as the base for poverty measurement.

All in all, these adjustments of inclusion and exclusion made by the WB to the HCES data, the development of this new measure of “welfare aggregate” in place of the money-metric utility based on consumption expenditure that has so far under-grided the measurement of poverty, and the application of this measure only to India and only for 2022-23 together cast serious doubts about the veracity of the June 2025 revisions to the India numbers.

Exclusion of Durables and Other Items

While there is a case for including free transfers in the measurement of consumer expenditure as long as the corresponding poverty lines also include the value of such transfer for determining the cost of basic needs, the WB approach to exclude some categories of consumption in the measurement of poverty defies logic. It also runs counter to its own assertion of moving away from a consumption expenditure-based measure to welfare-centric measure for poverty estimation. Four set of items, durables, certain non-food non-durable expenditures, hospitalisation expenses and rental expenses have been excluded while arriving at the aggregate consumption expenditure of households. Of these, rental values are not available for all households, particularly in rural areas. The NSS has been using the category of imputed rent of self-occupied dwellings, but these are in any case excluded from aggregate consumption expenditure.

However, the remaining three items have always been part of the consumption aggregate of households and have not been excluded by the NSS or so far by the WB.

For the latest revisions, the WB has excluded these items from consumption expenditure on the ground that they are not an accurate measure of the value of welfare derived from the consumption of these items. The reason for exclusions are not that these expenditures do not yield any welfare to the consuming households but the difficulty in assessing the value of welfare derived from the purchase of these items.

The implication of these arbitrary and irrational inclusions and exclusions is that the WB has created its own version of consumption expenditure from the HCES 2022-23.

The underlying argument that purchase price of a durable item may not accurately reflect the welfare derived from the purchase in the reference year is not just erroneous but also goes against the basic principle of using consumption as a proxy for welfare derived from consuming these goods and services. The fact that durables are multi-year assets and therefore their use value will differ from year to year does not justify excluding the welfare aspect of such purchases in the year of their purchase. This is conceptually no different from the expenditure on clothing and footwear which many, particularly poor households, use not only in the month that it has been purchased but also for several years 9In fact, the Classification of Individual Consumption According to Purpose (COICOP), the international reference classification of household expenditures (United Nations, 2018) treats them as semi-durable. . It is also counter-intuitive given that the WB has decided to include the value of free transfers in their estimates because they provide utility to the households. But it yet excludes the whole set of durables because of its inability to correctly estimate the welfare gains from such purchases.

The WB also excludes certain non-food non-durable expenditures on account of a lack of precision on the welfare provided by these purchases. These include jewellery and wrist watches but also furniture. This again defies logic. While jewellery may be a store of value, it is incorrect to argue that it does not provide any welfare to the households. Possession of jewellery is an essential cultural capital for many rural societies and is part of social customs and rituals. But even if one agrees on exclusion in the case of jewellery, the exclusion of wrist watches and a large group of expenditures including furniture is inexplicable. The WB has also excluded hospitalisation expenses on account of these not being welfare enhancing 10Mancini and Vechhi (2022) who examine the Deaton and Zaidi (2002) recommendations also suggest inclusion of health expenditure as part of the consumption expenditure. . At the same time, it has included non-institutional medical expenses in aggregate consumption expenditure. It again defies logic as to how non-institutional medical expenses are welfare enhancing but institutional medical care is not.

The implication of these arbitrary and irrational inclusions and exclusions is that the WB has created its own version of consumption expenditure from the HCES 2022-23. This aggregate is heavily biased towards food expenditure and has excluded a large majority of non-food expenditures made by households. The choice of these adjustments is clearly in favour of increasing the consumption expenditure of poorer households while reducing the expenditure of non-poor households.

The ICP 2021 PPP Exchange Rates

As mentioned earlier, major revisions in WB poverty lines and estimates are necessary as and when the Bank incorporates the new PPP exchange rates from the ICP. This time, the WB has incorporated the new PPP exchange rates for 2021. Estimates of PPP exchange rates by the ICP have improved over time, but serious issues remain particularly for large countries such as India.

Table 1 provides the reference PPP exchange rates based on the different ICP cycles and the corresponding national poverty lines in Indian currency. There are two reference PPPs available from the ICP: according to final consumption expenditure by households and non-profit institutions serving households (NPISHs) including spending on housing (Code:9100000) and another one excluding housing expenditure (Code: 9260000). The difference lies mainly in the use of rental expenditure in the PPP which has been a difficult issue in the ICP programme. As is obvious from Table 1, the PPP exchange rates are lower with housing included in the consumption basket. Since the PPP exchange rates are available for 2021, the 2022 poverty lines have been drawn up by updating the 2021 poverty lines using national CPI inflation factors.

Table 1: ICP PPP Exchange Rates and $3 Poverty Lines in Rs per month

Current global poverty comparisons by the WB employ three poverty lines, with $3.0 (PPP) as the lowest poverty line. India’s national poverty line has hitherto broadly corresponded to the lowest international poverty line converted using the PPP exchange rates. While these are similar to the poverty line obtained (in Rs) for 2011-12 by using the Tendulkar committee methodology, the WB revision of the poverty lines to $3.0 per day (based on ICP 2021) from $2.15 per day (based on ICP 2017) raises the nominal poverty lines in 2011. However, these are similar to the poverty lines for 2022-23 when updated using either the Tendulkar or Rangarajan committee methods. One consequence of this is that poverty is then estimated to be significantly higher in 2011-12 for India when compared to the Tendulkar committee estimates, but significantly lower in 2022-23.

This results in a much sharper reduction in poverty in 2022-23 compared to 2011-12 than is otherwise obtained using the updated Tendulkar or Rangarajan committee poverty lines.

The ICP revisions also sit uncomfortably with the implicit domestic inflation estimates. The corresponding index for the rate of inflation between 2011 and 2022 for the two PPP exchange rates are 1.35 for household expenditure including housing and 1.47 for household expenditure excluding housing. As against this, the actual inflation from the CPI indices during the same period is 1.87, while using the Fisher indices it is 2.09. Note, it also shows a decline in PPP exchange rates on expenditure with housing between 2017 and 2021, which implies a lower nominal poverty line in 2021 compared to 2017 which is difficult to reconcile with an increase in prices by a factor of 1.29 based on national CPI.

These anomalies are partly a result of how the ICP exchange rates are calculated. The fact that 2021 ICP leads to a lower nominal poverty line in 2021 compared to 2017 is also not unique. The adoption of the 2011 ICP led to the anomaly of India’s nominal poverty line being reduced in nominal terms for 2005 despite national prices having increased rising much faster. The adoption of the 2011 ICP led to reduction in the nominal national poverty lines in 65 countries prompting the Atkinson commission to suggest a freeze in using new ICP estimates until 2030. 

There is considerable literature on this issue but some of the issues specific to developing countries, and particularly in Asia, are associated with the large urban bias in the price data (See Deaton and Dupriez 2011 and Himanshu 2008). This is particularly true for housing prices which are collected separately in urban areas, unlike all other data that are collected on a basis similar to the household expenditure schedule of NSS (See Deaton and Heston 2010). For India, the weighting scheme for ICP 2021 remains the 2011-12 consumption surveys as the 2022-23 survey was not available at the time of finalisation of the ICP 2021 PPP exchange rates. Finally, since rental data are not available in consumption surveys in most countries (including India where it is not included in MPCE or CPI measures in rural areas), the ICP uses a hybrid approach of rental data from urban areas and volume of housing from other sources. This partially explains the lower rate of inflation reported by the PPP exchange rates with housing, compared to the ones without housing.

Revision of Historical Estimates of Poverty and Inequality

The revision in poverty estimates for India by the WB not only affects the estimates of poverty for the last decade but also for earlier years.

The changes in poverty estimates for earlier years are primarily due to the shift of the WB from the URP-based consumption estimates to MMRP-based estimates. First, the adjustments use a sector-consumption level adjustment factor to convert them into MMRP estimates. Second, unlike for 2011-12 and 2022-23, the revised estimates do not use state-sector specific poverty lines but are arrived at using national poverty lines (state-sector specific since 1993-94). Third, the consumption aggregates for the years prior to 2011-12 do not make any changes to the NSS reported consumption expenditure as done with the 2011-12 and 2022-23 data to reflect “welfare”. As a result, the estimates of poverty have changed and so has the pace of poverty reduction over time. Even the Gini based inequality estimates have changed due to the change in the recall period.

Table 2 presents the Bank’s September 2024 estimates of poverty for India and the revised estimates (June 2025) for the years since 1977-78.

Table 2: Poverty Head Count Ratio and Gini for India as estimated by the World Bank (adjusted MMRP)

While the earlier WB estimates were consistent with trends in poverty reduction using official Indian poverty lines, trends from the new WB series are at variance. For example, it is well known that the 1980s, represented by 1983 and 1987-88 witnessed a faster poverty reduction than was seen earlier. A large literature exists with varying poverty lines, all of which concluded that the period between 1983 and 1987-88 was a period of rapid poverty reduction in India. The revised WB estimates, however, suggest poverty actually increasing between 1983 and 1987-88. Similar is the case with the period between 1993-94 and 2004-05, which had witnessed poverty reduction of over 5 percentage points, but the revised WB estimates suggest almost unchanged poverty between 1993-94 and 2004-05.

WB Estimates of Inequality

The shift to MMRP in estimating consumption expenditure also changes the trends in inequality over the years. More importantly, it even changes the level of inequality every year. That is expected since MMRP consumption estimates are consistently higher than corresponding URP estimates at the lower end of the distribution, whereas these are lower at the higher end of the distribution. In general, Gini coefficients are the highest for URP estimates followed by MRP estimates and the lowest for MMRP estimates 11For 2011-12 where all three estimates are available, the Gini coefficient for rural areas for URP, MRP and MMRP estimates are 30.7, 28.4 and 28.3. The corresponding estimates for urban India are 38.5, 37.2 and 36.3. .

Just changing the recall period changes the consumption Gini every year. The Gini for 1993-94 and later are also lower due to the use of spatial price deflators. For trends, between 1993-94 and 2004-05, consumption inequality based on URP estimates increased sharply from 31.56 to 34.01, but the revised MMRP estimates suggest a much more moderate increase from 26.11 in 1993-94 to 27.83 in 2004-05.

Without the adjustments, the consumption Gini of India would put India at the 18th rank amongst countries with comparable consumption inequality.

The estimates for 2011-12 and 2022-23, even though on MMRP recall, are non-comparable to the earlier years due to the adjustments made to the consumption expenditure for the later years. The net impact of the inclusion of the free transfers and exclusion of durables and other items is to lower the inequality estimates by 4 percentage points, compared to the unadjusted estimates. For 2011-12, the adjusted estimates suggest a Gini of 28.8 as against the unadjusted estimates of 32.7. Similarly for 2022-23, WB adjustments reduce the original Gini from 29.5 to 25.5.

Without the adjustments, the consumption Gini of India would put India at the 18th rank amongst countries with comparable consumption inequality. India’s consumption Gini at 29.5 is similar to the reported consumption Gini of Bhutan (28.5), Pakistan (29.6), Nepal (30.0) and China (30.4). While India is no different from other South Asian countries in the level of inequality, the PIB’s attempt to claim that India is among the most equal countries also makes the basic mistake of comparing apples and oranges. The rankings are based on data which are income based for some countries while for some others they are consumption based. It is well known that consumption inequality is 7-14% lower than the income inequality for the same population. But even within groups of countries with consumption surveys, changes in the recall period for collecting data, and the methodology of data collection can affect the estimates of inequality.

In fact, the WB also highlights the difficulty of comparing income and consumption Ginis. It also points out that India’s income inequality estimates from the World Inequality Database (WID) are relatively high and have also shown a sharp increase from 52 in 2004 to 62 in 2023, making it the second most unequal country after South Africa. There may be issues with the income Gini from the WID as these are synthetic estimates based on income tax data. But even survey-based estimates of income inequality from the India Human Development Survey (IHDS) suggest income inequality at 0.52 in 2005, not very different from the WID estimates.

Conclusions

The debate on measurement of poverty in India predates by decades any attempt by the WB to estimate global poverty. Much of this debate has largely been based on India’s official poverty lines. While the government actively participated in this process by releasing official estimates of poverty with independent expert groups revising the poverty line, it was also helped by the availability of consumption expenditure data from the NSS surveys. The NSS surveys were seen as pioneers in the measurement of consumption expenditure.

However, there has been a steady erosion in the reputation that India has enjoyed for its approach to poverty measurement. The decision to junk the 2017-18 consumption survey was not based on any analytical exercise and the changes in 2022-23 have not been based on any robust analytical exercise. On most issues of survey methodology, the NSS would generally take the lead through pilot and bridge surveys which were absent before the changes made in the 2022-23 survey. Another serious problem is the absence of any official poverty line after 2011-12, even though the Rangarajan committee report was submitted more than a decade ago.

The WB’s exercise is a global exercise that tracks progress on sustainable development goals: the SDG-1 and SDG-10. Given its mandate, it has continued to release these estimates at regular intervals for global comparisons. While the WB uses consumption for a large number of countries for measuring poverty and inequality, it also uses a significantly large number of income surveys for countries where consumption surveys are not available. But even in consumption surveys, there are issues of comparability across countries which makes difficult any exercise to arrive at a definitive ranking of countries on either poverty or inequality. Nonetheless, these estimates serve the useful purpose of prioritising poverty and inequality on the global agenda. While there is certainly scope for improvement in how different countries measure poverty and inequality, these are largely within the domain of each country’s national priorities, statistical capacity and, above all, the local and sub-national context.

In the 2025 WB numbers, there appears to be a complete variance between the poverty estimates based on the official Indian poverty line methodology and the global estimates…

However, as a global multilateral institution and as a repository of global data on poverty and inequality, it is the responsibility of the WB to generate knowledge on the best practices for measurement of poverty and inequality. Such a process would require the WB to insist that countries follow these best practices for measurement which allows comparability over time and across countries, while countries retain their national priorities. The Commission on Global Poverty chaired by Tony Atkinson was one such exercise. Among the many recommendations that the commission made, two are worth reiterating. One was the need for greater coherence between national poverty estimates and the global poverty estimates (Recommendations 1 and 2). In the 2025 WB numbers, there appears to be a complete variance between the poverty estimates based on the official Indian poverty line methodology and the global estimates, both in terms of the modifications made to the consumption expenditure data and the poverty line drawn up. The second major recommendation was that given the problems with the ICP and the variance between inflation measures from the national CPI and the ICP, the ICP 2011 would be kept as the basis for updating the poverty line until 2030 (Recommendation 10). Even this has not been acted on.

While the WB acknowledges the difficulty of comparing consumption surveys of different countries which vary in recall period, coverage of items and the methodology of data collection, it nevertheless puts them together for country-wise rankings. But the WB estimates are neither suitable for domestic policymaking nor are these rankings based on strictly comparable data. The priority for national governments including India is to set-up their own committees for examining some of these issues. India also needs to revise its own poverty line, which is more than a decade old. This must be based on an independent committee’s recommendations. The celebration about the WB estimates cannot hide the government’s incompetence in being unable to update a more than one-decade-old poverty line or revising its own methodology for estimating poverty or inequality.