Counting the poor: some methodological issues4 min read . Updated: 24 Sep 2008, 01:02 AM IST
Counting the poor: some methodological issues
Counting the poor: some methodological issues
This panel was set up on Supreme Court orders, following a ruling on a lawsuit filed by human rights group People’s Union of Civil Liberties regarding the BPL survey of 2002, the use of which was stayed until February 2006.
The 2002 survey is known partly due to methodological flaws in the way the poor were identified and partly due to faulty implementation of the census exercise. The expert group will have to deal with the first part of the problem, which is to design suitable measures to identify the poor based on quantifiable and readily observable criteria.
In other words, the proxies selected as part of the new methodology should be able to identify the poor without the time-consuming and difficult measure of consumption expenditure that is the source for official poverty estimates.
But can there be such robust proxies? Previous evidence for proxies used in the 2002 BPL survey suggests this may be a near impossible task in the rural context. Even if the proxies selected are a close approximation of the characteristics of the poor, the grouping will also have to design suitable aggregation rules for such proxies.
The present BPL census method is based on 13 indicators, each being given a score of zero to four. These indicators are educational status of children, migration, possession of clothing, food security, sanitation, housing and so on. The final score is the sum of scores of all 13 indicators. The problem with this method is that this procedure establishes, in effect, cardinal equivalences across what are essentially ordinal rankings of alternative status of households in respect of individual indicators.
In other words, not having one square meal a day throughout the year is treated equivalent to open defecation or not possessing electrical appliances. The cut-off scores are then decided to ensure the total number of BPL families does not exceed a certain limit based on the official estimate of the number of poor in 1993-94.
Moreover, by explicitly linking the choice of the cut-off score to a pre-set estimate of poverty by reference to the official poverty line, the census also nullifies a potential advantage: of providing an alternative measure of poverty that seeks to go beyond consumption poverty. This is particularly so given the large anomalies in the present official figures.
What has been the track record so far of the three BPL surveys (1992, 1997 and 2002) conducted? Though the number of proxies has gone from one in the 1992 survey to 13 in 2002, the errors of exclusion and inclusion remain above acceptable limits. Errors of exclusion are those that misclassify the poor in the non-poor category, while errors of inclusion include the non-poor in the poor category.
These errors can be calculated using the 61st round (2004-05) of consumer expenditure data of the National Sample Survey Organisation. The 61st round has, for the first time, included a question on possession of BPL cards by households in rural areas. It also included questions on possessing durable assets. With this, and other household characteristics, it is possible to construct BPL scores using the same indicators used in the official BPL census.
A simple cross-tabulation of the status of a household using the two methods suggests that only 39% of the households identified as poor using the official poverty estimation methodology of the Planning Commission possess a BPL or Antyodaya card.
This means that 61% of households—who are poor on the basis of their consumption expenditure being less than the official poverty line—are excluded from the net of BPL census. On the other hand, 25% of the households belonging to the non-poor category by consumption expenditure poverty method possess BPL or Antyodaya cards.
Incidentally, the percentage of population that report possessing either of these cards is the same as that obtained from the use of official poverty estimation methodology of the plan panel.
But what happens when the exercise is done using same proxy indicators as those used by the official BPL census methodology? This exercise for 2004-05 was done by Jyotsna Jalan and Rinku Murgai (2007) using the 61st round of consumer expenditure data. It concluded that more than 45% of the poor are excluded by the present BPL census method.
The authors tried alternative methods of aggregation. Under all such simulations, the errors of exclusion and inclusion remained substantial. It was practically impossible to bring these errors below 33% even with expanding the list of proxies and using weighted aggregation methods.
Given these complexities, the committee has a real challenge to devise suitable proxies for identifying the poor using a BPL census. Needless to say, a universal entitlement approach with self-selection of the beneficiaries would be the best option. However, this option has not found favour with either the bureaucracy or the political parties.
Himanshu is assistant professor at Jawaharlal Nehru University and visiting fellow at Centre de Sciences Humaines, New Delhi. Farm Truths looks at issues in agriculture and runs on alternate Wednesdays. Respond to this column at email@example.com
Also Read Himanshu’s earlier columns