The poverty of questions

Poverty has certainly reduced in India but we need to ask more questions

The Planning Commission has released the results of the National Sample Survey Organisation’s (NSSO) Household Consumption Expenditure Survey. This is the survey that gives us the figures for the extent of poverty  in India. According to this survey, 29.8% of Indians were poor during the period 2009-10, a decline of 7.3 percentage points from the survey of 2004-5. The Planning Commission has also announced that it used Rs 28.6 per person per day in urban areas and Rs 22.4 per person per day in rural areas as the poverty line.  This is the national average, but each state has its own rural and urban poverty lines.

The numbers have caused much headline-grabbing outrage and have been subject to a lot of misplaced mockery, as they have always been. In the November 2011 issue of Pragati, we examined many of the common criticisms of the poverty numbers and explained how they were mistaken. While it is unfortunate that the same ill-informed arguments are being repeated, the intent of this article is not to cover that ground again. Instead, we will attempt to go in depth and examine some new issues.

It is worth doing a quick review of how the poverty line and the poverty headcount is calculated. The NSSO measures poverty in the most direct way that is feasible. Households across the country are sampled and surveyed for their consumption over a previous recall period. They are asked what they consumed and how much they paid for it. The value of their total consumption is compared against what it would take to buy a “poverty line basket” of goods and services (food, education, health and other essentials) using the same prices that people in their state have reported paying, in the same survey. For example, in rural Odisha in 2009-10, those who were surveyed reported spending Rs 567 per person to buy the basket. So Rs 567 per person per month is the poverty line for rural Odisha, and any household that reported spending less than that amount is classified as poor. Using this definition, the NSSO classified 37% of the rural Odiyas, or 153,000 people, as poor during the period 2009-10. The important point is to note that the NSSO survey by itself does not classify individual Indians as poor or non-poor. Remember that it has carried out a sample survey, not a census. It just estimates the overall number, and publishes the poverty line. The poverty line could be used to determine eligibility for various government schemes and budgetary allocations, but currently it is not. The Planning Commission’s 2011 attempt to do so was shot down by the states.

Photo: World Bank

Is this methodology perfect? Certainly not. There are flaws, some of which we will discuss. But we must point out that arguing over the definition of who is poor is rather pointless.  Let us use an analogy from the other end of the spectrum here — just as we disagree over what is poverty, we also disagree over what affluence is. If you are an Indian reading this article, chances are that an overwhelming majority of you are from the top 10% of the population, and if you were asked to classify yourself, you would probably put yourself in the middle-class, reserving the epithet “rich” for people like Mukesh Ambani.

We can argue till kingdom come over whether you are rich or middle-class, or we can recognise that some things are objectively determined, while others are a matter of classification. The objective fact is that you travel to work by car while Mr Ambani does by helicopter. Whether your commuting by car makes you rich is a matter of terminology and classification, which may differ from person to person. The criteria for classification shouldn’t be entirely arbitrary — what criteria you use should depend on what you want to measure.

But one thing is sure — if we use one set of criteria to classify people as poor or rich, you should take care to use the same criteria when you want to compare across time. If you use “commutes by helicopter” as your condition for being rich in 2004, but use “commutes by car or helicopter” in 2009, you will get a large increase in the number of rich people. But your data will not be comparable.

The Planning Commission has been updating and expanding its definition of poverty in response to criticism. The latest update to the definition came about in 2009, when the working group under economist Suresh Tendulkar updated not only the poverty lines, but also made changes to the methodology by which poverty is measured.  Sure enough, this report has come under attack from people like P Sainath and Dilip D’Souza for the sin of confusing them. Mr Sainath, in his article in the Hindu conflates an increase in the poverty number because of an updated definition with a real increase in the number of the poor. Mr D’Souza in his article in the Hindustan Times does not appear  to realise that his question about how the updated definition affects previous numbers is already answered in the Suresh Tendulkar report that he refers to. More importantly, the NSSO reports and data are available publicly, albeit for a price.  If anyone wishes to analyse whether the survey numbers have  misstated the trend, it is possible.  But that would require going beyond the headlines and press releases.

There may be valid criticisms over the comparability of data over time. Prof Himanshu of the JNU makes the case in the Mint newspaper that poverty is understated by 1.6 percentage points in 2009 because the mid-day meals for school children are included in the consumption estimates, and because this is a change from the 2004 statistics, the comparison is misleading.  Other than the valid debate over whether government welfare payments should be included in consumption figures for poverty calculation purposes,  this data can give us information on a more important area. Are welfare spends like mid-day meals and NREGA wages reaching the right people? Is the consumption increase among the disadvantaged commensurate with the spending on welfare schemes?

There may also be questions over survey design and sample sizes. The Northeastern states have shown wild swings in poverty numbers between 2004 and 2009. For example, Tripura has shown a decline from 40% to 17.4% while Manipur has shown a drastic increase from 37% to 47%. What accounts for such drastic swings? Is it a real change or is it an artifact of small sample sizes? If so, what can be done to improve data collection in smaller states?

The quality of the data provided by the NSSO has been questioned in the past. While the NSSO data comes from the most direct source that one can think of — by a direct survey of people, the quality of the data depends on the surveyors’ sincerity and knowledge, the design of the survey questions, and the willingness and ability of the respondents to recall and provide accurate information.  The data provided by the NSSO has been compared against the National Accounts Statistics (NAS) provided by the Central Statistical Organisation.  The NAS are  the data that are used to calculate India’s GDP. It turns out that there is a discrepancy in the numbers — Montek Singh Ahluwalia referred to this discrepancy when he was under fire for the supposedly low poverty numbers, and the media immediately took this reference as an admission that the poverty numbers were low because of this discrepancy.

If they had bothered to check, they would have realised that the truth is the other way round. As compared to the NAS, the NSSO numbers overstate poverty. They always have, and the discrepancy has been increasing over time. Back in the 90s, the NSSO numbers did not show any decline in poverty (or for that matter any increase in consumption, whether among the rich or the poor) while the NAS numbers did. Since then, the NSSO numbers have been showing a robust decline in poverty, but the NAS numbers have been showing even more of a decline.

Prof Tendulkar and K Sundaram  carried out a study in 2009 to understand the reasons for the difference, and while they identified many reasons, an important one is that the NSSO survey design is relatively static. An item of consumption will show up in the results only if it is asked about in the survey. As Indians broaden their consumption patterns, the discrepancy between what they are asked about and what they consume will increase. The NAS are relatively more dynamic. They collect information from a wider variety of data sources and the methodology is updated continually.

In fact, the NAS-NSS story is not the only one. The NSS numbers have been cross-checked against the Market Information Survey of Households (MISH) and once again, the result is clear — the NSSO numbers overstate poverty.

So yes, poverty has reduced while economic growth has occurred. But how? We are often reminded that the plural of anecdote is not data. What is meant by this is that we shouldn’t generalise wildly from isolated examples. Yes,  but in another sense, what else is data, but a collection of anecdotes? Each of the families that the NSSO’s HCE survey interviewed probably have a story to tell. How did they improve their lot?

The poorest Indians are the landless labourers. They do not have steady jobs, and they do not have an assured income. In 2004-5, many of them did not have enough to eat, and were below the poverty line. In 2009, a significant number fared much better. Did this improvement come about because they found better paying casual jobs? If so, where did they find them? Was it construction labour in the cities? Was it service jobs in malls? Was it farm jobs? Or was it due to the NREGA? These are the types of questions that will help resolve the debate on whether growth is sufficient or if you also need welfare schemes. Given the volume of the data the NSSO collects, there are bound to be pointers in the data to these questions, if not outright answers.

Likewise, what has been the impact of the inflation that India has experienced over the past two years on poverty? The poorest Indians spend over 60% of their wages on food, and food has suffered the worst inflation. The poorest Indians are also farm labourers, and to an extent the increase in prices may have translated to a rise in wages. Has that happened or have the price increases benefited middlemen, leaving the poorest worse off?

How have different states managed to achieve different levels of poverty reduction? Is there a correlation with the policies they followed? What about inequality? Has it increased? Have different segments of the population benefited differentially from the economic reforms? Has there been a segment that has been left worse off? These are the kind of questions that we should focus on once we get over the headline grabbing outrage.