Even if you raise the poverty line as much as the Expert Group does, it turns out just 11 per cent of Indians are poor, not 37 per cent — that’s a huge poverty of estimation.
That the study of poverty is a major industry in India (and the World Bank) is a fact. Whether this is as it should be is another story. The Planning Commission of India, under the direction of its chief, Montek Ahluwalia, has set up an internal committee to recommend the future of the Commission, i.e., whether it has outlived its original purpose, and what its new role in life should be. This is indicative of clear, forward thinking. A pity, therefore, is that it comes on the heels of an old-fashioned publication by the same organisation, “The Expert Group to Review the Methodology for Estimation of Poverty”.
|POVERTY IN INDIA (% population, 2004-05)|
|Old Poverty Line||Rural||Urban||All India|
Measurement at 1987
Levels of accuracy
|New Poverty Line (36% higher for rural India, 16% for urban India)|
Measurement at 1987
Levels of accuracy
|Notes: 1) Inflation adjustment means that the poverty line in rural areas matches internal NSS inflation; this is the recommendation of the Expert Group.
2) The "1987 level of accuracy" refers to the ratio of survey to national accounts mean consumption of 71 per cent. This is taken as the "truth" implying that 29 per cent of Indian consumption does not accrue to anybody, rich or poor. See text and Bhalla (2002,2003) for details.
3) Calculations of poverty are for the NSS 2004-05 survey, unit level data. The three assumptions yield the three estimates. The most preferred estimate is "NSS-Consumption Measurement at 1987 levels of accuracy".
The analysis of poverty can get controversial, and complicated. There is technical stuff involved, of the sort not suitable for a family newspaper. But in reality, the calculation of poverty is very simple. A person is poor if her real consumption is below the stipulated level of consumption deemed the poverty line. If so simple, then why the controversy and why the need for the Expert Group and its volumes of publications?
The Group on poverty misses out on two crucial developments: poverty in India is becoming relative and the existing method of measuring poverty is prone to large and inexplicable errors. Let us deconstruct the route taken by the Group to come out with its assessment that the fraction of people poor in India was not around 21 per cent but close to double that amount at 37 per cent. The major departure for the Group is to change the poverty line, and increase it by 16 per cent for the urban areas and a whopping 36 per cent for rural India. The motivation — because the earlier poverty line was not comprehensive enough (it did not include estimates for expenditures on health and education) and was pointing to lower rates of inflation than actually existed. On both counts, the Group suffers from either a lack of appreciation of what a poverty line is, or myopia, or both.
A poverty line is just a level of consumption. How Ms Sita chooses to spend this amount — on food, liquor, movies, or children’s education — is ultimately her choice. In the olden days, the planner decided what Sita should spend on, and thus arrived at a justification for the poverty line. As I have written elsewhere, the developing countries should, and must, raise the poverty line. This to account for the growth that has taken place since the poverty line was first constructed in the 1960s. Coincidentally, the 26 per cent increase recommended by the Group in 2009 is almost identical to the recommendation contained in Bhalla (2003). Rather than state the obvious fact that the level of consumption that was considered non-poor in 1960 should be considered poor in 2010, the Group goes about catching the poverty nose in a typically Indian way (by going behind the neck, yoga-style). The basket of the urban poor around the old poverty line is defined to be the “new normal poor”. By using residence of an individual as the basis for defining the poor, the Group breaks new ground, is pioneering. It would have been simpler to just state that the absolute poverty line is being raised, period. Why would anyone object to that?
The convolutions that set in because of this desire to be technical (again, the details not suited to family consumption) lead the Group into avoidable errors. If they mention it once, they mention it twenty times: the CPI index for agricultural workers (CPIAL) has been understating inflation for rural workers; hence, one should use the internal price indices generated by the NSS surveys. The Group cites an important study by Prof Deaton, who showed that NSS inflation during 1999-2004 was 14 per cent compared to the CPIAL figure of 11 per cent. The Group fails to note that over the longer period, 1987-2004, the CPIAL was overstating inflation by 7 per cent. Hence, that is a basis for lowering the poverty line by 7 per cent rather than raising it by 3 per cent — an error gap of 10 per cent.
But these errors are minor compared to the most glaring omission of all — indeed, a deliberate commission. As is well known to the Group headed by Prof Tendulkar, until recently the head of the Statistical Commission of India (where I served for three years as a member), the National Sample Surveys have been suffering from extreme inefficiency. The amount of national consumption captured by the surveys reached a nadir of 48.7 per cent in the 2004-05 survey. This means that in the 2004-05 survey, 51 per cent of the consumption in the economy — of wheat, fruit, vegetables, airline travel, Rolls Royce etc. — did not accrue to the Ambanis, you, me or the poor. This is a large dose of missing consumption.
According to the NSS survey, this missing consumption does not exist, was never there, is a figment of your imagination. But it is there accruing to all households in the economy. And a detailed investigation (see Bhalla, Imagine There’s No Country) shows that the NSS surveys underestimate consumption for all households, rich and poor, on an approximately proportionate basis, i.e., consumption for all households needs to be increased by the same proportion in order that the lower survey mean and the higher national accounts mean match. An example can help illustrate. Assume that the 1987 NSS survey was approximately right. In that year, the survey mean was only 71 per cent of the national accounts mean. So, 29 per cent of total consumption was missing, and perhaps even accrued exclusively to the rich households. But we need to still allocate 22.3 per cent (71 - 48.7) of total consumption. Proportionate adjustment would mean that all consumption be increased by the factor 1.46 (71/48.7). On a conservative basis, the multiplier for the poor is reduced by 10 per cent, so instead of 1.46 it is 1.41. Note the extreme indulgence the Group shows for its perceived inflation error of 3 per cent between 1999 and 2004. And there is no concern, not even a footnote mention, of the glaring error of at least 41 per cent understatement in the consumption of the poor? Talk about missing the entire forest, and the trees as well.
What a difference the errors of omission and commission make to the poverty estimates is illustrated in the Table (above). Estimates of the head count ratio (HCR) of poverty are presented for the two different poverty lines. The big conclusions are as follows. First, a correct adjustment for inflation reduces the HCR by 3-5 percentage points (not raise it as believed by the Group). Second, a conservative adjustment to the 1987 level of efficiency reduces national poverty to 5 per cent. Since relative poverty is considerably more than 5 per cent, there is a clear case for substantially raising the poverty line. But if this poverty line is raised as per the recommendations of the Group, poverty in India is still only 11 per cent. This suggests that the method of measuring poverty needs to be changed — maybe an appointment of a new Expert Group, and one less steeped in preordained conclusions?
The author is Chairman of Oxus Investment