Consumption-based measures don’t accurately estimate poverty
Since the publication of poverty estimates purportedly based on the Tendulkar methodology and the 2009-10 consumption survey of the National Sample Survey Organisation (NSSO), many in Parliament and outside, from different political parties, have questioned its conclusions. Concomitantly, media reactions have speculated on poverty’s relationship with fertility, growth, specific schemes, et al. But, India’s poverty, like itself, refuses to classify itself in simple boxes.
Beyond the happenstance of poverty decline in an odd state being less than another, there is no strong and obvious relationship to growth in incomes, whether agricultural or non-agricultural; population, urban or rural, or to the performance of schemes like NREGS. Might it be found in district-level relationships to economic and demographic structure?
It might, but there are good reasons why it might not. The headcount ratio, that is, the share of people below a certain level of consumption, called the poverty line, is a blunt measure. States with a high proportion near the poverty line will show a large fall in headcount ratio for relatively small increases in overall consumption, while states with a large proportion well below the poverty line will show smaller reductions, even if they have higher increases in consumption. This indifference to inequality below the poverty line weakens the relationship between growth and the headcount ratio and is the essence of Amartya Sen’s critique of the measure.
Besides, using a consumption measure introduces oddities. Between two households with the same income, use of a private school may classify one above the poverty line while the one using a government school may stay below, since the higher expenditure on education would be included while calculating expenditure, but the possibly higher saving of the second household ignored.
The Planning Commission mentions that the “rate of [poverty] reduction was much lower than the average in Bihar, Uttar Pradesh and Chhattisgarh”. These three states are dissimilar. The population share of the bottom decile (lowest 10 per cent) of national consumption in UP is similar to its population share, but in Bihar (as in Jharkhand, Orissa and Madhya Pradesh), this share is more than twice (thrice in Chhattisgarh) their share of population. Poverty runs deeper in Bihar and Chhattisgarh, and thus equivalent income growth may lead to less poverty reduction than in UP. Besides, people in deep poverty are unlikely to possess the requisite characteristics to participate effectively in the “modern” economy and may need a very different approach.
More broadly, the distribution of consumption differs across states. The inter-decile interval, that is, the difference between one group of 10 per cent of households and the next, is much less in Bihar, that is, they are more tightly bunched, than Maharashtra. Many states have large proportions (around a fifth of the official poor) within a 50-rupee band above and below their specific poverty lines. This can be seen as fragile poverty reduction, where small changes add a fifth more to poverty, or as a large poverty reduction that will happen presently. Because of this sensitivity to small changes in the poverty line, similar households will be treated very differently, if such headcount cut-offs are used to allocate benefits.
Such pitfalls of targeting are one reason why the UID is not a silver bullet. It is a tool for traceability, that is, it can track where money goes, but it cannot ensure that it goes to deserving persons — indeed there may be no clear “deserving person”. In cash transfer schemes elsewhere, beneficiary selection is quasi-universal within geographies and often the prerogative of the local government. Votaries of cash transfers from an omniscient Delhi cannot digest that.
Nor do they accept that these poverty estimates cannot be a report card on pro-poor schemes, without much more work. Is the “higher than average” performance of Maharashtra because its NREGS is targeted at the bottom of the distribution (which it is), or because it is more urbanised? Is better BPL targeting the reason in Orissa and MP (we can’t say, since the BPL question was removed from the 2009-10 survey!)?
Notwithstanding protestations, there is little evidence of manipulated changes in the poverty lines, though the apparently non-comparable inclusion of mid-day meals in calculating household consumption (Mint, March 26) will raise suspicions. Increases in the poverty line in all states (an average of 8.5 per cent per year) have been more than state-specific inflation rates. Yet, government offers little justification; instead, it concedes and establishes a technical group to “revisit the methodology for estimating poverty in a manner which is consistent with the current realities”. Have “current realities” really changed that much since the Tendulkar Committee? Is household consumption of Rs 5,200 a month in Delhi that indefensible as a poverty line?
This heat is generated, in part, because the headcount ratio is used by the Planning Commission as a tool of inter-se resource allocation between states. Regardless of the number identified in the BPL survey (another misguided attempt at targeting), resource transfers were capped by the state’s share of the poor as calculated from the NSSO survey. Shorn of this influential but unsupportable function, which is slated to go, it can be what it is – one of many measures of progress (or the lack thereof). Since the data is publicly available, researchers can discover many other patterns, e.g., women workers moving away from rural handloom to readymade garments. If media and politicians paid as much attention then as now, we might get a discourse on the nature of our progress that is as rich and varied as our reality.
The writer is senior fellow at the Centre for Policy Research, New Delhiexpress@expressindia.com