There are three kinds of lies: lies, white lies and statistics, or so goes the saying. If there is one scholar who has made it his life’s mission to uncover attempts at using statistics to project lies and unmask truths about the world using statistics, it is this year’s Nobel laureate in economics, Angus Deaton of Princeton University.
Deaton has been a key figure in resolving a long-drawn, contentious debate on poverty statistics in India. The Nobel Prize Committee cited Deaton’s work “on measuring living standards and poverty in developing countries with the help of household surveys” while awarding him the prize.
Poverty estimates in India have been based on consumption expenditure surveys conducted by the National Sample Survey Office (NSSO). These estimates became the subject of a major controversy in the late 1990s after a revised methodology used by NSSO suggested an increase in consumption expenditure, which translated into a big decline in poverty levels. Coming as it did in the midst of a spirited debate on the role of globalization and liberalization in reducing poverty and inequality in the world, and also in India, the survey attracted wide attention.
The key point of contention was the recall period used for the survey. Critics argued that due to a change in the recall period, the consumption figures and poverty estimates of 1999-2000 were no longer comparable with earlier estimates. This put a big question mark on the data and on the extent of poverty decline in the 1990s.
Re-examining the data to excavate parts of it that were seemingly uncontaminated by the methodological changes, Deaton and his co-author, economist Jean Drèze, showed there was indeed a decline in poverty, although the decline was less than what the official numbers stated.
In a review of the great Indian debate on poverty estimation, Deaton and Valerie Kozel of the World Bank argued that what was a statistical exercise became a political one because of the zeal of influential policymakers to show that the Indian economy was lifting all boats as it grew. Similar ideological persuasions also inform writing on poverty numbers, with different authors using methodological approaches with implicit assumptions that are almost designed to arrive at findings that match their ideological predilections.
They cite the example of the debate around the discrepancy between the estimates of consumption derived from surveys versus those from national accounts and show how some conservative economists made use of national account numbers to justify their position that poverty is much less than what the official numbers show.
Drawing on the work of reputed Indian scholars such as the late B.S. Minhas, Deaton and Kozel argued that while national account numbers may seem intuitively more credible compared with survey numbers, in reality the manner in which they are calculated involves a number of assumptions and use of old survey information. As such, it is not wise to scale up poverty estimates based on national account estimates, a practice India abandoned many years ago.
Deaton harked back to this theme in his recent statement to the Indian media after winning the Nobel prize.
“I have used data from India’s famous National Sample Surveys to measure poverty,” wrote Deaton. “Perhaps the biggest threat to these measures is that there is an enormous discrepancy between the national accounts statistics (NAS) and the surveys. The surveys ‘find’ less consumption than do the national accounts, whose measures also grow more rapidly. While I am sure that part of the problem lies with the surveys—as more people spend more on a wider variety of things, the total is harder to capture—but there are weaknesses on the NAS side too, and I have been distressed over the years that critics of the surveys have got a lot more attention than critics of the growth measures. Perhaps no one wants to risk a change that will diminish India’s spectacular (at least as measured) rate of growth?”
Deaton’s critique is significant given the widespread discrepancies associated with India’s new gross domestic product series, and will give strength to its critics.
While Deaton has critiqued the work of conservative economists, he has also sparred with those on the left end of the spectrum employing, once again, careful empirical calculations and theoretical rigour.
When left-leaning economist Utsa Patnaik claimed that poverty levels were much higher once one took into account real calorie consumption levels, Deaton and Drèze rebutted her, arguing instead that falling calorie consumption was not necessarily a sign of deprivation.
Lower physical activity and better access to health had reduced calorie requirements for a large section of Indians, they suggested. They cited the example of Kerala, a relatively well-off state, having a low calorie intake to highlight their point that well-being is not linearly related to calorie intakes.
Although Deaton significantly influenced the debate around India’s poverty line, he has not shied away from critiquing the non-transparent and complicated poverty estimation procedure in place currently. He has also advocated the delinking of poverty estimates—which, in his own words, are mere statistical tools for comparison—from all welfare entitlements.
A 2014 Indira Gandhi Institute of Development Research working paper by C. Rangarajan and S. Mahendra Dev provides a useful summary of poverty estimation methods that have been employed in India so far.
Until the recommendations of the expert group under Rangarajan were submitted to the government in 2012, the poverty line was based on (or rather, supposed to be based on) the fulfilment of a minimum calorie intake for each person. The estimated calorie requirement decided by the 1979 task force was taken to be 2,400 kilocalories per person per day in rural areas and 2,100 kilocalories in urban areas.
The poverty line was to be set at an expenditure that would guarantee the aforesaid calorie requirements, given the observed consumer behaviour from NSSO data. The calculations yielded a poverty line of Rs49.09 per capita per month in rural areas for 1973-74 and Rs56.46 in urban areas.
Another expert group set up under the leadership of D.T. Lakdawala, which submitted its report in 1993, laid down the procedure for calculating state-wise poverty estimates and suggested that the original poverty line expenditure be adjusted by consumer price indices for agricultural labour in rural areas and industrial workers in urban areas.
The result of all these endeavours was the availability of state-specific rural and urban poverty lines and the percentage of people below poverty line expenditure, often referred to as the head count ratio.
The Suresh D. Tendulkar committee report, published in late 2009, for the first time expanded the scope of poverty norms by including expenses on health and education. Tendulkar’s results, however, relied on average calorie intake based on actual consumption figures reported by NSSO, which turned out to be lower than the normative estimates that were used earlier.
So, even after including expenses on education and health, Tendulkar’s revised numbers were roughly similar to the old urban poverty line (and higher than the old rural poverty line), after adjusting for inflation. Tendulkar justified this by arguing that the consensus within the profession suggested that the old urban poverty line gave a far more accurate picture than the old rural poverty line did. As a result of the Tendulkar committee report, the overall poverty ratio rose by 10 percentage points to 37% for 2004-05, compared with the earlier methodology.
But the estimates were widely criticized for being too low, and another expert group led by Rangarajan was set up to look into the issue. The Rangarajan committee report, published in 2014, made use of a different recall period, and also different nutrition levels. The committee included protein and fat intake in addition to a normative calorie requirement level to arrive at poverty estimates.
Rangarajan’s method has led to an increase in head count ratio by 19 percentage points in rural areas and 41 percentage points in urban areas from the previous methodology, with a poverty line expenditure of Rs972 for rural areas and Rs1,407 for urban areas for 2011-12.
As is clear from the discussion so far, poverty line expenditure (and the head count ratio) is not sacrosanct in itself. Rather, it denotes a person’s ability to achieve a minimum consumption level, which in India’s case has been calorie (now protein and fat as well) intake.
In a scathing attack on the Rangarajan committee report, Deaton and Drèze argued that the committee’s method was both theoretically and empirically implausible. Although the use of normative calorie norms was sanctified by long use (at least till the Tendulkar committee abandoned it), it had no underlying logic to it, they wrote. They highlighted that the evidence did not suggest a straightforward link between calorie intake and nutritional well-being.
Despite far more complicated calculations, the Rangarajan poverty lines provided counterintuitive results, with urban poverty turning out to be higher than rural poverty in half of India’s major states, they wrote.
“In short, the Rangarajan expert group method is both theoretically and empirically implausible,” wrote Deaton and Drèze. “What then is the way forward? Appointing another expert group is unlikely to serve the purpose, given the record of previous expert groups. Perhaps the time has come to abandon the elusive search for a technical method of deriving a poverty line that can be interpreted, in some normative sense, as the minimum cost of dignified living. Would it not be simpler, and more useful, to regard the poverty line as a mere statistical benchmark, and set it in a simple and transparent manner that the public can understand?”
Deaton and Kozel had stressed the role of transparency in resolving contentious debates on key government statistics such as those relating to poverty. They also argued that long gaps in data collection complicate problems of comparability and make contentious interpretations more likely.
In his work, Deaton has taken care to honestly note the assumptions used to arrive at his own findings, and cautioned that his own estimation of poverty in India, based on partial and corrected data, is sub-optimal compared with having clean and comprehensive survey data.
Deaton makes a similar point in his press statement too: “My work shows how important it is that independent researchers should have access to data, so that government statistics can be checked, and so that the democratic debate within India can be informed by the different interpretations of different scholars. High quality, open, transparent and uncensored data is needed to support democracy.”
Deaton’s comments should be a wake-up call for a radical overhaul of India’s statistical systems.
Economics Express runs weekly and features interesting reads from the world of economics and finance.
Comments are welcome at firstname.lastname@example.org