Rethinking social safety nets in a changing society

This paper was coauthored by Debasis Barik, Pallavi Choudhuri, Bijay Chouhan, Om Prakash Sharma, Dinesh Kumar Tiwari (NCAER) and Sharan Sharma (University of Maryland College Park and NCAER)

Historically, India’s approach to social safety nets has involved identifying the poor and providing them with priority access to social protection. Analysing data from the India Human Development Survey, collected in three waves across 2004-05, 2011-12 and 2022-24, this article finds that households face considerable transition in and out of poverty as the economy grows, making it difficult to identify and target the poor in a precise manner.

This is the third article in the Ideas@IPF2024 series.

Throughout the 21^st century, poverty in India has consistently declined. Preliminary estimates¹ from three rounds of the India Human Development Surveys (IHDS), organised by the University of Maryland and the National Council of Applied Economic Research (NCAER), document a drop in poverty headcount ratio from 38.6% in IHDS-1 in 2004-05 to 21.2% in IHDS-2 in 2011-12, further declining to 8.5% by IHDS-3 in 2022-23. These poverty estimates use the inflation-adjusted poverty line, first proposed by the Tendulkar Committee. This trend is consistent with data from other sources, including the newly available Household Consumption Expenditure Survey (HCES) conducted by the National Statistical Office, although the exact estimates differ.

The importance of transient poverty increases as poverty declines

Most of the discourse around poverty in India is ‘static’. It focuses on poor households at any given time but pays little attention to movement in and out of poverty, possibly due to a lack of longitudinal data. Even during periods of relative economic stability, when poverty rates move slowly, individual households may experience income and consumption changes as household composition changes, workers retire, and children grow up and find employment. Illness, marriage expenses, and natural disasters may also influence income. During periods of economic transition, unforeseen circumstances may become particularly important. India has also seen widening regional divergence in economic growth, with a college graduate in Tamil Nadu earning much more than their peer in Madhya Pradesh. Opportunities in some occupations have declined while those in others have expanded.

This suggests a need to examine household economic status from a longitudinal perspective. Figures 1 and 2 show the transition in household consumption, showing both entry into and exit from extreme poverty between 2004-05 and 2011-12 and between 2011-12 and 2022-24.

Figure 1 documents the movement in and out of poverty for 38% of households that were poor and 62% of households that were not poor in 2004-05. By 2011-12, poverty had declined, and 25% of families had moved out of poverty, with only 13% still mired in poverty. At the same time, of the 62% of non-poor households in the prior wave, 8% had now become poor. These newly poor comprised about 39% of all impoverished households in 2011-12. A similar exercise with 2022-24 data in Figure 2 shows a similar trend, with 18.1% of households moving out of poverty and 5.3% falling back. The change between the two periods is that the overall poverty level in 2022-24 is substantially lower, and the newly poor form a more significant part of all poor households - about 62%.

Figure 1. Poverty transition between IHDS wave 1 and 2

Figure 2. Rising share of transient poverty between IHDS wave 2 and 3

For the poverty transitions between 2004-05 and 2011-12, we undertake a detailed analysis of the characteristics of entry into and exit from poverty (Thorat et al. 2017), which documents that this movement out of poverty is greater for historically disadvantaged groups like the Scheduled Castes (SC). However, these groups remain highly vulnerable to falling back into poverty, suggesting a precarious perch in lower-middle-class status. Other studies using data from IHDS waves 1 and 2 have tried to examine underlying conditions at IHDS-1, such as household size, presence of the elderly, land ownership, and caste/religion that may predict the potential of slipping into poverty in IHDS-2 (Bandyopadhyay and Bhattacharya 2022), and found that their ex-ante vulnerability measure is a positive and significant predictor of future poverty. However, the coefficients are relatively small, reflecting our inability to predict falling into poverty perfectly ex-ante. This is in sharp contrast to the observations from studies on 20^th century poverty (Mehta et al. 2011), which found considerable persistence of chronic poverty among SCs, Scheduled Tribes (ST), landless households, and large households with many family members. It is important to note that chronic poverty for STs is higher across centuries.

Dynamic nature of poverty makes it difficult to identify the poor

Social protection programmes aim to introduce cost efficiency by limiting their benefits to the poorest households. India began implementing this approach by moving from a universal public distribution system (PDS) for foodgrain to a targeted PDS in 1997. This required the designation of households as being poor, or in the words of Indian officialdom, being ‘Below Poverty Line’ (or BPL) households.² The Ministry of Rural Development made this identification through nationwide censuses in 1992, 1997, 2002 and 2011 (Saxena 2015).

The latest exercise in 2011 was conducted through the Socio-Economic Caste Census (SECC) by the Ministry of Rural Development in rural areas and the Ministry of Housing and Urban Poverty Alleviation in urban areas. Identification of households as being deprived was based on criteria for automatic exclusion (for example, having an automobile or government job), automatic inclusion (for example, primitive tribal groups or people living on alms), and a graded score based on occupation, living conditions, caste/tribe, and family composition. These criteria were selected for use in rural areas based on the recommendation of a working group headed by Dr NC Saxena for rural areas (Ministry of Rural Development, 2009) and Prof SR Hashim for urban areas (Planning Commission, 2012). Local government authorities then make a provision for the validation of the target households. While this method produces a ranked list, the cut-off of households deemed eligible is determined by the proportion of households in a state considered to be poor based on the 2011-12 National Sample Survey.

Following Akerlof (1978), ‘proxy-means testing’ (correlating proxies, such as occupation and household characteristics, with poverty) without verifiable income for social benefits has a long history (see also Banerjee et al. (2024)). However, the validity of specific criteria used in identifying deprived households as a part of issuing BPL cards has come under considerable criticism (Alkire and Seth 2013, Drèze and Khera 2010, Sharan, 2011), even from one of the originators of the identification schema (Saxena 2015). These reviews found that many poor households were excluded from the BPL list, while many non-poor households were included.

When inclusions and exclusions were observed, they were often attributed to the local political economy, which allows some individuals to negotiate access to BPL cards, regardless of their economic status. Research shows that elite capture and social networks play an important role in who can get a BPL card (Besley et al. 2005, Panda 2015). With some innovation, improving the identification of the poor may be possible using carefully crafted inclusion and exclusion criteria (for instance, see Asri et al. (2022)). However, little attention has been paid to how these targeting strategies may fare in a rapidly changing economy.

As we have shown earlier, tracing the movement of households in and out of poverty becomes more critical as chronic poverty declines. However, the cost and logistic difficulties in undertaking major initiatives like SECC imply that these exercises will be infrequent and may not be effective in an era of rapid change.

Using data from different waves of IHDS, we examine the correlation between possession of BPL cards³ and per capita consumption expenditure. The results highlight that while the poor are more likely to hold a BPL card, we find both types of households at each expenditure level. Although the distributions have progressively converged (results not shown here), exclusion and inclusion errors remain.

At the time of IHDS waves 1 and 2, the BPL cards would have been issued using the 2002 survey, but by IHDS-3, the SECC survey 2011 was used for BPL designation. The total number of households eligible for BPL cards grew slightly between IHDS-1 and IHDS-2, but expanded substantially by the time IHDS-3 was conducted. This expansion was due to implementation of the National Food Security Act (NFSA), which mandated that 75% of rural and 50% of urban households be covered for highly subsidised food distribution. This massive expansion should have addressed the exclusion errors, and all poor households should have received BPL (or Priority Household (PHH)) cards. In contrast, inclusion errors would have increased due to the programme's expansion, an acceptable form of error under NFSA.

This hope has been only partially fulfilled. While BPL cards became more common in 2022-24, about 30% of the poor still do not have access to them, nor do 35% of the households above the poverty line but below a zone where they can still slide into poverty (taken to be twice the poverty line). Ironically, while there was an increase in the number of non-poor households (from 30% to 56%) between IHDS-2 and IHDS-3, households with BPL card access increased from 59% to 66%.

However, focusing on elite capture may overstate the issue. A part of the exclusion of the poor may be due to the original design for BPL cards being linked to residential locations, which led to the exclusion of migrants. The 'One Nation, One Ration' card initiative may help enhance the portability of BPL cards, reducing vulnerability among poor urban migrant workers. A larger problem may be that this inconsistency is due to poverty decline, with many poor households (who correctly held BPL cards when the cards were issued) having now moved up to the non-poor category but retaining these cards due to infrequent identification of the poor. Conversely, households that were not poor during the SECC survey may have fallen into the poverty or non-poor but vulnerable category. However, their initial designation could only be changed with new assessments; hence, they were excluded from gaining BPL cards.

Imprecise identification of the poor has concrete consequences

Eligibility for many benefits, particularly access to government-provided health insurance under the Ayushman Bharat Yojana, is tied to the household’s designation as poor. However, working in selected occupations may also confer some of these benefits. We find that households in different consumption categories, with and without BPL cards, differ substantially in their access to public insurance. Whether they are poor, vulnerable, or rich, among households with a BPL card, about 40-43% have access to government insurance, including both central and state schemes. Among households without a BPL card, this number drops to about 23-25%. While private insurance access is also rising, it does not fill the hole for the poor households without a BPL card.

The dynamic nature of poverty suggests that identifying poor households and determining whether to provide them with social protection only once in a decade may miss out on those who fall into poverty in the intervening time, while continuing to support those who no longer need social protection.

I4I is now on Substack. Please click here (@Ideas for India) to subscribe to our channel for quick updates on our content.

Notes:

Results are based on initial data from the IHDS-3 survey that have not been scrutinised for data quality and use preliminary sampling weights without considering attrition between surveys. Hence, the result should be treated with caution.
In recent years the terminology has changed to ‘Priority Households’ as the segment covered has grown to expand a larger share of the population, but BPL remains popular in common parlance, and we continue to use it in our writing.
Other categories of households eligible for programmes like Annapurna Yojana (a scheme under which 10 kilograms of food grains per month are distributed free of cost to individuals above the age of 65 years with no or meager subsistence) and Antyodaya Anna Yojana (a scheme which aims to provide food to poor families at highly subsidised rates) are combined with BPL households here.