Monthly plan to access Budget

Journalism of Courage

Data supports the claim that poverty decline in India has been faster since BJP came to office in 2014

The authoritative DHS household data for 2005-06, 2015-16, and 2019-21 establishes that the pace of decline in Indian poverty accelerated post 2014 elections. We must address our progress while continuing to shift goalposts.

A higher MPI suggests greater intensity of deprivation while a higher censored poverty rate is an important signal to policymakers to redirect policy focus. (Representational/File)

A recent article by Jean Dreze (‘Poverty, uncensored’, November 24) looks at our commentary on multidimensional poverty decline (‘Poverty is down, period’, November 4) and concludes that the debate on relative performance towards poverty alleviation during the UPA years and Narendra Modi’s tenure as prime minister should continue to be “open” even as his own conclusion is that multidimensional poverty (MPI) declined considerably faster in the Modi years.

The availability of multidimensional poverty data for 2005-06, 2015-16 and 2019-21, (note that poverty is likely biased upward due to the exogenous shock of Covid in the Modi period), means that we now have an estimate of poverty decline during the Modi years. Notwithstanding our inadvertent error in the reporting of 2011-12 values (IHDS not DHS data), the incontrovertible result remains that poverty decline was significantly faster during Modi’s tenure (period II, roughly 2015-21) than during the UPA years (hereafter period I, roughly 2005-15). We document this conclusion in detail below. But first, some transparent discussion of definition and measurement.

There are two sets of poverty estimates provided by the Oxford Poverty and Human Development Initiative (OPHI) that compile these data across countries, primarily from Demographic and Health Surveys (DHS). The first are uncensored estimates for individual indicators, which correspond to a simple question regarding whether a household is deprived (poor) in a given indicator — for example, nutrition. Alternatively, one can obtain an indicator-specific censored poverty estimate via a two-stage process. The first stage estimates the population that is multidimensionally poor; the second stage estimates the population that is poor in each indicator for the multi-dimensionally (MP) poor. For example, in 2005-06, the MP poor were 55.1 per cent; uncensored nutritionally poor were 57.3 per cent; and 44.3 per cent were censored nutritionally poor. In other words, close to 80 per cent of the nutritionally deprived are also multidimensionally poor. For the DHS India survey years, we had used censored estimates; Dreze uses uncensored estimates. Which definition should one use? The authors of MPI are clear in their preference, and recommendation: They believe that censored estimates are better for analysis and policy design.

Subscriber Only Stories

The logic for preferring censored estimates: For some individual indicators such as assets, some households may be considered as deprived (poor) even as they are relatively better off in other areas such as nutrition, sanitation, etc. Censored data helps shift the focus onto those who have been (multidimensionally) identified as poor. A higher MPI suggests greater intensity of deprivation while a higher censored poverty rate is an important signal to policymakers to redirect policy focus.

One other advantage of a censored approach is that it allows the capture of interlinkages between several poverty indicators. For example, environmental enteropathy is known to have a key role in nutrition absorption in children. Therefore, investments made towards providing sanitation facilities and piped water connections will have an impact on nutritional absorption.

We come back to our original goal, of analysing whether poverty decline was faster during the UPA than the NDA. Available data for MPI (and most indicators) point to a faster pace of poverty alleviation during the latter period (see table). Annual pace of improvement in the health, education and living standards indicators during 2005-15: 7.3, 10.0 and 9.6 per cent respectively. In the Modi years: 11, 8.4 and an outsized 17.2 per cent annual gain in living standards. What prompted this large improvement? Several policies, but more importantly an efficient redistribution combined with direct fiscal resources targeted specifically to reduce deprivation across individual indicators.

The table reports both uncensored and censored estimates for the average rate of change for all 11 indicators of multidimensional poverty — the master MPI index itself, and two indices (censored and uncensored) for each of the 10 indicators. What is of policy interest is whether the improvement (rate of decline of poverty) was larger in period II (2015-16 to 2020-21) than period I (2005-6 to 2015-16). As we had noted in our earlier article, the inclusive growth belief was that period I would show a greater improvement because the dominant component of poverty decline, growth in per capita consumption, was about 0.8 percentage point higher in period I (annual 3.8 per cent increase vs. 3 per cent in period II). However, and somewhat a priori surprising, the pace of MPI index decline was almost twice the pace in period II relative to period I! This result is strongly indicative of considerably more inclusive (and more efficient and less corrupt) growth in period II compared to period I.


The table also reports a performance index for poverty decline — the index is defined as the ratio of the rate of poverty decline in period II relative to period I. If the ratio is higher than 1, then there was an improvement in performance in period II; less than 1, a slowing of the rate of improvement.

Go down the uncensored list. For nine out of 11 indicators, the pace of poverty decline was faster in the Modi period II. Performance index average for all indicators — 1.6, or approximately a 60 per cent higher rate of decline in poverty. For the uncensored (Dreze’s preference), the average rate of improvement is only slightly lower at 1.55. Regardless of censored or uncensored poverty measurement, the average pace of poverty reduction was considerably faster during 2015-21.

For only four indicators is the rate of uncensored poverty decline lower in period II. Assets and school attendance are lower in period II for both uncensored and censored poverty. Incidentally, school attendance improvement is expected to be lower as one approaches 100 per cent enrolment — the pace of change from 20 to 25 per cent enrolment is 25 per cent versus a pace of only 1 per cent when enrolment increases from 95 to 96 per cent.


In contrast, Dreze dismisses the considerably faster pace of decline in MPI (from 6.9 per cent in period I to 11.9 per cent in period II) on the grounds that this occurred because of a “low base”! “Multidimensional poverty HCR, it turns out, decline faster in the second [Modi] period. This is not necessarily surprising, since a given percentage decline is easier to achieve from a lower base”. MPI poverty moved from 27.7 per cent in 2015-16 to 16.4 per cent in 2019-21 — this was not “acceptable” performance for Dreze. However, change from a considerably smaller base in school attendance (6.4 to 5.3 per cent) during these two years is, according to Dreze, indicative of a great slowdown in the rate of decline of poverty and hence completely acceptable to him!

In our article, we had concluded that the pace of poverty decline in India had accelerated post 2014. Despite Dreze’s polemics and obfuscation, there is no evidence to modify the earlier conclusion. One final point — Dreze claims that “most sample households in the 2019-21 NFHS survey were interviewed before the Covid crisis began”. Really? According to the sampling distribution reported in table 1.2 of the NFHS 2019-21 report, for slightly more than half of the sample interviews ended between January and April 2021 — or well after Covid had fully set in.

Can India do better on some of the individual MPI indicators? Of course, it can. But that does not mean we do not accept our progress and keep shifting the goalposts.

Bhalla is former executive director, IMF, representing India, Sri Lanka, Bangladesh and Bhutan. Bhasin is pursuing his PhD at SUNY, Albany. Views expressed are personal

First published on: 29-11-2022 at 07:22 IST
Next Story

Why India must stand with Palestine

Next Story