The scepticism of the economists in using multidimensional indices in development planning can be traced back to the Impossibility Theorem of Kenneth J. Arrow which stipulates that ranking of alternatives—economic or geographic states—reflecting the collective preference of any group, based on the rankings of the individual members of the group, would inevitably violate basic principles of rationality or axioms. By this logic, ranking of countries, states, districts or political regimes by employing a number of indicators would be theoretically untenable. This academic purity of social scientists—forcing them to be non-judgmental in many situations demanding immediate intervention—has been questioned by Amartya Sen and Mahbub ul Haq as they have ventured into ranking countries based on their levels of human development. One can see a number of national and global organisations now coming out with multi-dimensional indices, a very important one being the multidimensional poverty index (MPI), brought out annually by UNDP in collaboration with the Oxford Poverty and Human Development Initiative (OPHI) as a part of the Human Development Report. Several regional authorities and national governments have gone ahead and constructed an MPI at national, state and sub-state level, India being no exception.
A Glance through History
India’s concern to alleviate poverty and build a socialistic pattern of society led to attempts at measuring poverty largely through unidimensional indices. The Planning Commission had set up a committee in the early 1960s under the guidance of VM Dandekar which worked out normative expenditures of Rs 20 in rural and Rs 25 in urban areas, per capita per month, loosely based on Indian Council of Medical Research norms as the cut-off point (poverty line) below which people were termed as ‘poor’. Understandably, the issues relating to nutritional norms under different environmental and working conditions, appropriate consumption basket for households around poverty line, price indices for rural and urban areas in different states etc. have been revisited time and again by individual researchers as also by committees, task forces and working groups set up by government. The dominant view has been to anchor the poverty line at a certain nutritional norm, adding reasonable amount of non-food expenditure and determine the number of poor at the state level, taking into consideration variations in prices in their rural and urban areas. Construction of such a uni-dimensional poverty index, thus, involves complex calculations to identify a poverty basket of consumption, working out price indices for updating of the poverty line, and then applying these to the incomes or consumption of households for determining their poverty status.
There were dissenting voices like that of PV Sukhatme who argued that meeting the nutritional norms does not guarantee good health as the latter depends on several environmental parameters such as the quality of water, sanitation, and the living and working environment. BS Minhas demonstrated empirically that nutritional norms are often socially and culturally determined. Given that studies showed no correlation between nutritional adequacy and outcome indicators of health, morbidity or mortality, the Tendulkar Committee (2010) proposed the formal delinking of the poverty line from nutritional norms and set it well above that obtained from the complex calculations noted above, since the latter was clearly seen as inadequate given the increased out-of-pocket expenditure on health, education and basic amenities, particularly in rural areas.
Manifestations of Poverty
Importantly, India has witnessed several situations that can generally be considered as manifestations of poverty. Lakhs of migrants walked thousands of kilometres from cities to their homes during the pandemic. Millions got infected due to their congested living conditions and many died on the streets and outside hospitals due to lack of beds and oxygen. Experts, however, would attribute these phenomena to management and governance failures and not necessarily to poverty. Researchers associated with national and international organisations noted below have shown how poverty levels have not gone up and actually declined during the pandemic. The National Family Health Survey (NFHS) has recorded that anaemia among children aged 6-59 months and in pregnant women and adult males has gone up between its two rounds covering 2015-16 and 2019-21. Based on the uni-dimensional poverty measures noted above, one can still hold poverty to have gone down, explaining the above in terms of intra-household distribution of food, poor dietary habits, improper water/sanitation facilities etc. Achieving certain outcome indicators such as health, life expectancy, morbidity and mortality is not the basic concern in these computations.
Understandably, there has been an uproar about the working papers of the World Bank and IMF published last year that report low or no poverty in India in the years of the pandemic or before that. These raise a basic question of whether poverty estimates must be based on input indicators like consumption expenditure, income etc. or on direct indicators of well-being like life expectancy, health, education and access to basic amenities.
A Focus on Indian Poverty
After the rejection of the National Sample Survey (NSS) of consumption expenditure for 2017-18 by the government, no regular survey on this has been conducted except the Periodic Labour Force Survey with limited canvass, making it difficult to build a temporally comparable series on poverty. Given this limitation, Bhalla, Bhasin and Virmani have, in their IMF Working Paper, developed a method for extrapolation of the consumption expenditure of the NSS for 2011-12 to build a series up to 2019-20. They use the growth rate of the Private Final Consumption Expenditure (PFCE) to scale up the figures and bring in the distributional changes by allowing the household’s consumption to grow at the rate of the nominal per capita income of its state. Rural urban price differences are introduced by determining separate poverty lines for them. The underlying assumption is that the distribution will remain unchanged, both within the rural and urban segments in each state over the period 2012-20.
Extrapolating household expenditure by the growth in PFCE may bring in an upward bias since the latter is estimated as a residual in national income accounts and, therefore, includes many other components such as expenditure of unincorporated enterprises etc., as argued by Minhas. One wonders if there could be other multipliers for updating the figures, free from these problems. It would have been more appropriate to upscale the consumption of each household by the growth rate of income in the economic activity to which the household belongs. Furthermore, the growth rates of different commodities in the PFCE are significantly different and hence adjustments can be done commodity-wise, to bring focus on the items of consumption by the poor, across the states.
The most significant contribution of the study is that it includes the provisioning of extra food grains by the government into the real expenditure of the poor while noting that the off-take of the poor from the public distribution system (PDS) has doubled after the implementation of the National Food Security Act, 2014. This, however, opens up the possibility of bringing in the changes in the level of state engagement in the provisioning of essentials, including free gas cylinders, electricity etc. within the framework of calculations. However, this must be supplemented by an assessment of the disengagement of the state in crucial social sectors such as education and health. Incremental private expenditure of the poor on health and education in recent years, as estimated through NSS data, would have to be considered by revising up the poverty line.
The paper by Sutirtha Sinha Roy and Roy van der Weide for World Bank explores the possibility of using Centre for Monitoring Indian Economy (CMIE) data in poverty calculations, after correcting for the unrepresentative character of the panel data, by modifying the weightages of households for aggregation. Indeed, the asset position and level of basic amenities are generally higher in the CMIE sample than in more robust national sources, basically due to the former’s sampling procedure. Many people find this study more
acceptable not because of the methodology but because the magnitude of poverty ‘appears more reasonable’ than that of the IMF.
It is important to note that recently, Bhalla has sought to conclude the poverty debate by shifting from the earlier computation discussed above, to multi-dimensional poverty, measured through physical deprivation indicators from the NFHS. Computing poverty based on deprivations in health, education and access to basic amenities, along with nutrition, is likely to enhance the importance that poverty numbers enjoy in policy debates, if not in policy making. Responding to this, John Dreze has made a few methodological points regarding the figures quoted by Bhalla but tends to accept the usage of MPI in policy making. It is important that both the scholars utilise the processed results from the Global Multidimensional Poverty Index published earlier this year. Dreze favours assessing achievements in pre-Modi and Modi periods using uncensored poverty data (without excluding the households whose poverty score is less than 33 per cent of the total possible score), thereby making a departure from the MPI framework. He finds no difference between the rate of decline in poverty in the two periods while the Global Report, quoted by Bhalla, demonstrate a higher rate of decline in the second period.
Importantly, the OPHI methodology supports the usage of censored figures as it stipulates that a household should not be considered multi-dimensionally poor because of its deficit in any one or two of the ten proposed indicators since such deficits can be compensated by ‘no deficit’ in a good number of other indicators. One can, however, argue, like Dreze, that the deficit in certain critical indicators cannot be compensated through others, dismissing thereby, the need to reduce the number of deprived people through censoring. In accepting either position, one would be exposed to criticisms, unless one takes the decision of censoring by the nature of each indicator.
Objectivity of the MPI Methodology
While constructing the MPI, a researcher can select a set of indicators, choose their sources and design the weightages to obtain, to an extent, a trend and pattern that he/she desires. The list of indicators and their weights developed by the UNDP-OPHI understandably have not been adopted universally in different subcontinents, countries and sub-regions. Most of these, have selected indicators under three
dimensions—health, education and living standards—assigning equal weights. Individual indicators have, however, been constructed differently, so much so that the figures in the regional and national reports cannot be compared with those in the Global Report. Although some guidance is available regarding the broad framework of analysis from the OPHI, the exact selection of indicators depends on the regional/national study team. The list of indicators and their weights, therefore, varies significantly across regions and countries due to the local consultative processes and the socio-political priorities of the government, despite attempts being made to withstand political pressures representing short-term interests. It is important to point out that NITI, in its Baseline Report, has recasted the indicators of the Global Report in the Indian context, using NFHS data for 2015-16.
The NITI Report employs 12 indicators against 10 of the Global Report. It adds maternal health under ‘Health’ and banking facility under ‘Living Standards’ as additional indicators but adjusts the weighting system to ensure that each dimension has equal weight. The Arab Multi-Dimensional Poverty Report, brought out by the UNESCWA, with which the present author was associated as research advisor, is based on the same three dimensions. This, too, has two additional indicators: early pregnancy or female genital mutilation (FGM) under Health, and congestion under Living Standards. This was based on extensive consultations with the Arab states and other stakeholders. The exact indicators and their deprivation thresholds in this report, as also in the NITI report are mostly different from those in the Global MPI. The robustness and acceptability of the results in any report will inevitably depend on the scientific basis of the consultative process for developing the indicators. While the inclusion of FGM as an indicator of poverty was contested by several members of the Arab League, the inclusion of bank accounts at par with drinking water or sanitation in promoting Living Standards is unlikely to go unchallenged in India.
An important point is that the MPI is not a headcount ratio of the multi-dimensionally poor. It multiplies the latter by the depth of deprivation, computed by averaging the deprivation of all individuals over all indicators. This adds to its robustness. This adjusted headcount is the MPI, and any temporal or cross-sectional comparison must be made based on this index. The shift from income poverty to multi-dimensional poverty may, thus, be seen as desirable. There is, however, a host of issues regarding the choice of indicators, their data sources, scaling and weightages. These need to be addressed via a consultative process, keeping it above the short-term politics in a regime.
Amitabh Kundu is senior fellow at the World Resources Institute, New Delhi office
