27 September 2014

New Poverty Estimate: Some Concerns

This blog post should be read as a complement to my recent note on Reading Between the Poverty Lines in the Economic and Political Weekly (27 September 2014). The note has been written in response to the Report of the Expert Group to Review the Methodology for Measurement of Poverty (Chair: C Rangarajan, hereafter Rangarajan report/method) that was submitted to the government on 30 June 2014 and is now available in the public domain. In this blog, I will raise some concerns based on my immediate reactions on this new method, but before that a quick background.

Source: Cartoonscape, The Hindu, 8 July 2014

A Working Group suggestion of 1962 based on a balanced diet requirement by the Indian Council of Medical Research (ICMR) in 1958 arrived at monthly per cpaita poverty lines of Rs.20 for rural areas and Rs.25 for urban areas in 1960-61 prices.

In January 1979, the Report of the Task Force on Projection of Minimum Needs and Effective Consumption Demands (Chair: YK Alagh, hereafter Alagh report/method) used age-sex-activity specific calorie allowances recommended by the Nutrition Expert Group of 1968 to arrive at a per capita per day average calorie requirement of 2435 kcal for rural areas and 2095 kcal for urban areas (for convenience rounded off to 2400 kcal for rural areas and 2100 kcal for urban areas and referred to as the calorie norms for rural and urban areas respectively). Based on consumption expenditure around these average requirements, they arrived at monthly per capita poverty lines of Rs.49.09 for rural areas and Rs.56.64 for urban areas in 1973-74 prices. While anchored in 'calorie norms', these poverty lines are behaviourally determined and indicate the purchasing power needed to meet these norms with some margin for consumption of non-food items. This gives a consumption basket around the poverty line.

The July 1993 Report of The Expert Group on Estimation of Proportion and Number of Poor (Chair: Late DT Lakdawala, hereafter Lakdawala report/method) used the Alagh report's average calorie requirements and the commodity basket associated with the poverty lines for the base year, 1973-74. However, to address variations in prices and consumption patterns across states (and also over time) they used appropriate state-specific consumer price indices (separate for rural and urban areas) and applied the Fisher's index to arrive at state-specific poverty lines.  These poverty lines, as indicated earlier, indicated the purchasing power needed to meet the calorie norm, but the actual consumption could be different that gives us  a commodity basket pegged around the poverty lines for the base year.   

Over time, the updated poverty lines (say, for 2004-05) pegged to the 1973-74 base years differed substantially from the average calorie requirement/calorie norm (see Utsa Patnaik's A Critical Look at some Propositions on Consumption and Poverty). This and other considerations led to the setting up of an  Expert Group with SD Tendulkar as its chairperson. Their report (hereafter Tendulkar report/method) submitted in November 2009 had to do away with the dated average calorie requirement, but they did not have any alternative reference to begin with. Hence, for pragmatic considerations, they begin by using the proportion of poor for urban India, 25.7 per cent for 2004-05, computed by the Lakdawala method, as a base. The difference being that they shift away from the use of monthly uniform recall period (URP) data to the mixed recall period data that combines annual recall for five low frequency items (clothing, footwear, consumer durable, education, and institutional health expenditure) with monthly recall for the remaining items. 

Given the proportion of poor for urban India, the Tendulkar report computed a commodity basket for the third decile group (20-30 per cent). Now, the consumption amount for this decile group became the new base associated with the poverty line basket, which was used to arrive at all-India rural and also rural and urban state-specific poverty line baskets from an appropriate state-specific decile group. 

Another departure for the Tedulkar report was to discard the consumer price indices and instead use prices determined endogenously from the household level consumption expenditure data by taking the average per capita expenditure of a commodity per unit consumed. Like the Lakdawal report, the Tendulkar report also used the Fisher price index to obtain state-specific poverty lines and also for future updating of these poverty lines. Instead of separate poverty line baskets for rural and urban areas, as was the case for both Alagh and Lakdawala methods, the Tendulkar method had only the urban poverty line basket for 2004-05 as base. 

A number of questions have been raised on the Tendulkar method, see Poverty Estimates in India: Old and New Methods, 2004-05 for a critical discussion. This necessitated the search for a new method.

The Rangarajan Method
Meanwhile, the ICMR came up with the Nutrient Requirements and Recommended Dietary Allowances for Indians in 2010. Based on this and taking a cue from the Alagh method, the Rangarajan method superimposes the age-sex-activity specific composition and updates the average per capita per day dietary requirements for calorie (2155 kcal in rural and 2090 kcal in urban), protein (48 grams in rural and 50 grams in urban) and fat (28 grams in rural and 26 grams in urban). From a purchasing power perspective, all these are possible at a monthly per capita food expenditure of Rs.554 (sixth fractile group, 25-30 per cent) for rural India and Rs.656 (fourth fractile group, 15-20 per cent) for urban India.   

In addition, expenditure in the median fractile class (45-50 per cent) is considered as the norm for four essential non-food  items, viz., education, clothing, shelter (rent) and mobility (conveyance). This turns out to Rs.141 for rural and Rs.407 for urban areas.

For all other non-food items, the expenditure for the fractile group identified with the poverty line basket is behaviourally determined. This turns out to Rs.277 for rural and Rs.344 for urban areas. 

Adding up the above three gives monthly per capita poverty lines of Rs.972 for rural and Rs.1407 for urban areas. These give the associated poverty line baskets.

Like the Lakdawala method, the Rangarajan method uses Fisher price index on the all-India poverty line basket to state-specific poverty line baskets to obtain the state-specific poverty lines for rural and urban areas. The Rangarajan method continues with the Tendulkar method's use of prices from the household level consumption expenditure data. The state-specif poverty lines help compute the rural and urban state-specific poverty ratios and their weighted average gives the all-India poverty ratios.

A departure in the Rangarajan method is that its computations are based on the modified mixed recall period (MMRP) consumption expenditure data, which, from 2009-10, also collected weekly consumption expenditure data for some high frequency food items (edible oil, egg, fish and meat, vegetables, fruits, spices, beverages, refreshments, processed food, pan, tobacco and intoxicants). Under MMRP, like MRP, the five low frequency items continue to be based on annual recall. For all other items that are not identified as low or high frequency items, the monthly recall continues.

In short, the new method is an updating of the 'calorie norm' that is age-sex-activity standardised (Alagh method) to obtain poverty line baskets at the all-India level for rural and urban areas and then it applies the Fisher's index to obtain state-specific poverty lines (Lakdawala method) by using prices from the household level consumption expenditure data (Tendulkar method). In addition, the Rangarajan method uses MMRP and also adds a normative criterion for four non-essential food items.Thus, it is an improvement over the earlier methods. Nevertheless, it raises some concerns, which I elaborate below.

Some Concerns
The combining of expenditure from two fractile groups is statistically plausible, but it poses a behavioral dilemma. This is so because the aggregate all-India poverty line is derived from two different fractile groups - one for food and most non-food items and the other for four essential non-food items. An alternative could have been dual poverty lines, as suggested in my note, Reading Between the Poverty Lines (Economic and Political Weekly, 27 September 2014).

The aggregate all-India rural and urban poverty lines are not meant for computing poverty ratios. If computed then then this poverty ratio will not be sub-group consistent.  In other words, the poverty ratio computed through this method will not be equal to the poverty ratio that is a weighted average of the state-specific poverty ratios.

In particular, the quantities used at the all-India poverty line basket will not be weighted averages of the quantities associated with the state-specific poverty line baskets. Once we question the all-India poverty lines then its usage in the Fisher price index to derive the state-specific poverty lines also come into question.

If one uses the 2010 calorie allowances suggested by ICMR for each state separately then by superimposing the state-specific age-sex-activity composition one will arrive at state-specific average requirements. There is an argument that different average requirements call for different poverty benchmarks. However, continuing with the Lakdawala method, the Rangarajan method uses only the the all-India average requirement to ensure comparability.

In addition to calorie deprivation, the allowances of ICMR 2010 can also be used to compute protein and fat deficiencies. Comparing and contrasting these three deprivations might be insightful.

As already mentioned, the poverty lines indicate the purchasing power needed to meet the average requirements. In short, all the state-specific poverty lines (separately for rural and urban areas) would be identified with the same purchasing power. This, along with the fact that the all-India poverty lines are not sub-group consistent, calls for caution in comparing the all-India poverty lines with purchasing power parity (PPP) in dollar terms.

The Rangarajan method points out that the proportion of poor at the combined all-India level declined from 38.2 per cent in 2009-10 to 29.5 per cent in 2011-12; that is, 8.7 percentage point decline in poor. While this is commendable, one should not read too much into this decline, as 2009-10 was a drought year.

It is possible that some non-poor population just above the poverty line would have become poor if they would not have been the recipient of some welfare schemes. If possible, quantifying these would show the impact of such interventions.

While, the current poverty lines are an improvement over the earlier methods, the inclusion of four non-essential items is not based on an absolute norm and they do still miss out on some other essential items like health and sanitation among others.

Finally, even if one keeps aside the above points, the Rangarajan method points out that in 2011-12, the proportion of poor are 30.9 per cent in rural and 26.4 per cent in urban areas. This is still a substantial amount and a matter of serious concern from a public policy perspective.

Concluding Remarks
Notwithstanding the improvements in the Rangarajan method, there are important methodological and public policy concerns. One possible alternative that the Rangarajan report could have followed or at least given as an alternative was to calculate region- (rural and urban) and state-specific average calorie requirement. They being not comparable can be a matter of methodological debate, but its need was already pointed out by S Guhan in the Lakdawala report. And, in any case, this approach of having separate calorie norms is already being followed for rural and urban India and it has not come in the way of comparison at the combined level. In future, a move towards measuring multiple dimensions of poverty would also be helpful.

An elaborate version of this blog post has been published as Reading Between the Poverty Lines in the Economic and Political Weekly (27 September 2014).

No comments:

Post a Comment