The Scrap Over Poverty Statistics

What should we make of the latest scrap over the Tendulkar committee report? Here are some thoughts.

 Poverty isn’t a binary variable. There is no switch that, when turned on, defines a household as poor vs.  non-poor.  There are various degrees of deprivation, and we have differing intuitions about at what level of deprivation we should classify a family as poor. Part of the root of the outrage over the seemingly low household income (Rs. 26 per day per person in villages to Rs. 32 per person per day in cities)  comes from the fact that our intuition about what constitutes poverty has changed.

 My uncle started his career after completing his graduation in the mid-70s in Bombay’s weather office. He was single and lived alone then, but he’d send part of his salary home to his family. Towards the end of the month, his money would run out, and the last few days of the month, he’d be able to cook and eat only one meal a day.

 Then, as now, if you were a graduate and you were earning an entry level salary in a government firm, you would be categorized into the middle-class – lower middle-class to be sure, but middle-class nonetheless. When did you last hear of a middle-class person lacking for food in India? But that’s how things were till the 70s, and my uncle’s situation wouldn’t excite comment then.  One can only imagine the situation of the others who were poorer than my uncle.

This is the background to the question of where to draw the appropriate poverty line. The NSS is a very good statistical organization. It has the difficult job of collecting data about an economy that is largely informal  and undocumented. It had to draw a poverty line ages back when we were a much poorer economy than now.  How do you estimate the income of the poor, when they have uncertain, intermittent and variable incomes? If you survey a poor household and ask them how much they earned in the past year,  there is very little chance that you will get a reasonably accurate answer, even if the respondents made a good faith attempt to provide one – and who gives good faith answers to questions about income?

So they decided to  draw a poverty line based on how many calories the household got to consume per day – this, they did by asking the surveyed families how much they bought or procured in the past “recall” period.  For example, they’d ask something like: “Did you buy rice in the past 7 days? How much?”

Now, it turns out that choosing the recall period is a dicey undertaking. If you ask my household if we bought any rice in the past 7 days, the answer is likely to be “no” because we don’t actually buy rice that often – more like once a month. So, a seven day recall period will overestimate poverty among the middle-class.

 But poor households who lead a hand-to-mouth existence do buy rice every week, if not more frequently. If you set the recall period at 30 days, you are likely to get inaccurate answers from the poor, because people’s memories aren’t very good. Because it is easier to recall a single purchase than to recall multiple  small purchases, the inaccuracy is likely to be more when the poor are asked.

 The NSS has been making adjustments to the recall period between its quinquennial  surveys. Sometime in the 90s, it changed the recall period from  7 days to 30 days for staples. There were accusations that part of the supposed decline in poverty was because of this. They were eventually debunked, but this is an example of the kind of complexity our poverty numbers have to deal with.

 Then there is the question of the basket of goods you ask about.  It is notoriously difficult to keep it constant across time. My ancient finance professor’s famous example illustrates this. Back in the early 70s (yes, he was ancient) he was in IIM Ahmedabad doing an analysis of what items of consumption went into the basket that determined the consumer price index. The index was constructed in the 60s, and then Gujarati women wore ghaghras. But, in the 70s, as my professor said “ghaghras were out!” – i.e. women were no longer wearing  them, but they were still on the index.  If you replace ghaghra-choli with sarees, blouses and petticoats, is the index the same?

 Whenever you ask about any set of consumption items, it is difficult to ensure comparability across time. You wouldn’t expect this to be a huge problem when it comes to poverty measurement – after all, you wouldn’t expect the composition of basic foodstuffs to change, but it turns out that it has, in subtle ways. The poor have been consuming less of cereals, but more of everything else.  They have also been consuming fewer calories even though their overall purchase patterns indicate that they can afford more.

 So, to summarize the NSS has the challenge of measuring poverty in a world where the consumption patterns of the poor have been changing, while ensuring that the poverty line is comparable across time. The Tendulkar committee that submitted its report in 2009 was constituted to try to resolve these complexities. You can read the report here.  They tried to resolve the issue by moving away from the calorie intake norm to an overall poverty line basket. In the process, they apparently raised the poverty line – i.e. a person who would not have been poor earlier would now be classified as poor under the new norm.

Apparently, this was not enough, if you go by the chatter in the media.  The issue has been framed rather cleverly – 32 rupees per day is a number that will tug at our heart strings. But it is 32 rupees, per person per day. If you translate that number to per household per month it amounts to Rs.5,760 per month for a household of 6  (six being the average size of a poor household )

 It is no one’s case that one can live comfortably with that income. A few days back, Mint  ran a story where they profiled families that were just above the poverty line, aiming to show how bad their lives were. Well, call me heartless, but what leapt out to me from the profiles was that the concerns of the families were things like sending children to a good school and owning a modest home, things middle-class readers can relate to.  Is that poverty? Perhaps it is, but I note that this is different from the kind of poverty that I read about in 1987, about the woman in Kalahandi, Orissa, who had to sell her daughter so that the rest of the family could get something to eat. Her case made the headlines when Rajiv Gandhi visited Orissa and tried to tell him what she did – and the officials mistranslated her account to spare Gandhi’s delicate feelings.

In fairness to the opponents of the current poverty line, I must point out that they have always complained that the current poverty line is more of a destitution line than a true poverty line – and they are right.

 But this has always been the case. Perhaps they want to make the following case: We are a richer society now. We have more resources to help the poor, and fewer of the most destitute. We should be able to use our resources to help greater numbers of disadvantaged, which is why we should raise the poverty line.

 If that is the case they want to make, I can agree – as long as they don’t use the statistical illusion of higher numbers of poor caused by a raised poverty line to argue that the number of the poor has gone up because of the economic reforms.

 But I must also sound a warning note – raising the poverty line will give rise to a  version of the searching-under-the-streetlight problem. The problem is this:

 A cop finds a drunk man in a parking lot late at night, searching the ground under the only street light in that parking lot. He asks what the guy is doing, and the drunk replies that he dropped his car keys and is looking for them. Asked where he was when he dropped the keys, the drunk waves towards a car in the darkness. Asked why he’s searching under the street light, he says that if the keys are actually over in the darkness, he’d never find them anyway.


The poverty version of this problem is the “Studying Rural Poverty in Thane District” problem. I used to note that there was a disproportionately large number of studies, news reports, etc. on poverty among the Warlis in Thane. It occurred to me that the reason for this is that Thane is the closest one can encounter rural poverty if one sets off from Mumbai. And because many NGOs, schools of social work and reporters are based in Mumbai, the Warlis are a convenient subject for field trips, project work and news reports.

 If you raise the poverty line, it will be easier to encounter the poor, and it will be that much easier to ignore the destitute.

4 thoughts on “The Scrap Over Poverty Statistics

  1. Good post, but aren’t you trying to counter a different issue? What you write seems more relevant to the determination of ‘how many poor households’ rather than ‘what makes for a poor household’ ?

    “So they decided to draw a poverty line based on how many calories the household got to consume per day”

    Not sure what this means. Why would a *poverty line* be set based on *actual consumption* of poor?

    Utsa Patnaik writes in the Hindu that the poverty line was basically set in 1974 based upon converting a ‘representative basket’ of 2400 calories of consumption for the rural poor and 2100 calories for the urban poor. This was later revised to 2200 and 2100. Since then, the rupee value of this basket has simply been revised according to inflation indices. Her argument (implied) was that the 17x increase in prices factored in by the Planning Commission in the latest report was factually incorrect, that the basket should be measured again rather than any arbitrary indices. She didn’t provide any data in her support, appealing rather to its obviousness, but the basic line of argumentation is correct. Determining a consumption, mainly food consumption, driven poverty line is basically about setting the *right* level of nutrition, converting that to an appropriate basket of goods and then getting the prices right.

    Even if the actual food consumption patterns have changed to substitute cereals by other foodstuffs (that are presumably more expensive per calorie), this does not change the determination of the poverty line. It could be a good argument against a polemic of ‘poverty has worsened, in fact the the poor are getting lesser food than earlier’ , but what does it have to do with the rupee value of a nutrition figure at current prices?

  2. Ritwik, I am writing a longer piece in Pragati which addresses your questions (in fact, you can blame your comment for my Pragati piece being 2000 words long) But here is the paper by Angus Deaton that measures what you were referring to:

    In short, the impact of the incorrectness in CPI calculations in 1%.

Comments are closed.