When the Detail is in the Devil!

Anyone, who is even remotely conscious of the perils of interpreting numbers, would know of *false positive* or *true negative* diagnosis that can befuddle one’s labored analysis. However, for one, who has made up his shoe and now seeks to fit somehow the foot of *facts* into it, such niceties do not deter. Such was the endeavor in an Op-Ed piece, Gandhi’s Experiments with Populism, in Indian Express of July 14th by its much acclaimed regular columnist. The Columnist, Mr. Surjit Bhalla – Chairman of Oxus Investments, seems keen to prove that NDA government (1999-2004) delivered better governance than UPA government (2004-2009). He is free to have his views, but not his facts. He is persistent in his pursuit as can be seen from his earlier article, The Real Scandal, also in Indian Express of February 4th. An account of that article was taken in this post: Oxus Sux! O!! No Proof Required.
In the present article the columnist posits that << the facts suggest that the NDA achieved a lot more in terms of inclusive growth than UPA 1>>. Given the title of his column, No Proof Required, he stresses untiringly that << it is critical that all the facts be presented, in all their complexity, and that the facts and interpretation be properly vetted>>. This too we make our starting point in exploring his presentation. But, before that here is a small disclaimer. The analysis is based solely on data as provided by him here and is only limited to deciding if the conclusions he draws pass the test of logic and reasonable inference. Negation of his assertion should not be taken as proof that the opposite is true (Caution: keep in mind perils of (mis)interpretation). The lynchpin of his article is a table, which appeared in the print edition, but seems to be missing in the e-version. Since without the *table* it would indeed prove trying even to follow what he has to say let alone discuss it, I have prepared a table, which appears below.

Barring the last two columns of the table and some colourful embellishment, the table reproduces faithfully the original. The table has actually 2 distinct parts: first, gives the household consumption data from different rounds of National Sample Surveys by NSSO; and second gives the head count ratio of poverty as estimated by Tendulkar committee. The fifth & sixth columns give the percentage per annum (PPA) growth rates as calculated by Bhalla for household consumption & poverty ratio (sic). Original table simply had the heading *Growth Rates* for column 5 & 6. I had to change it to *(Surjit Bhalla’s) Log Growth Rates* for reasons we shall soon come to see. The legend below the table informed that all growth rates are *Log Growth rates. Therefore what Bhalla first calculated and presented in column 5 & 6 are PPA Log Growth rates and then below each of these two sets (consumption & poverty) he calculated what he terms Relative Growth Rates (RGRs). RGRs are calculated by simply deducting relevant growth rate for OTHER Group from corresponding growth rates for SC, ST, Muslim & OBC.
Bhalla’s main point is short & simple: <<What can constitute inclusion is if the poorer groups (for example, SC, ST, Muslim, OBC) had a faster pace of consumption growth (and/ or poverty reduction) than the other non-(SC, ST, Muslim, OBC) group>>. From the perusal of the table it is clear that OTHER Group outperformed all 4 categories in household consumption growth under both NDA & UPA dispensation except in case of ST(scheduled Tribes) under UPA. However, as the intensity of outperformance by OTHER group over SC, Muslim & OBC (cells highlighted in green) is less in case of NDA than UPA; he concludes that NDA was more inclusive. When it comes to poverty reduction numbers, he talks in generalities: <<Parallel computations for the percentage change in the poverty ratio shows that each group performed relatively worse than the other group. So out of eight, for only one indicator (consumption growth of STs) was growth more inclusive in the UPA regime>>; without making any direct comparison between NDA & UPA numbers like in case of consumption data:<< For each indicator, the levels in the three years — 1999-2000, 2004-05, and 2009-10 — are presented along with growth (change) and relative growth. The latter is the difference in the pace of (percentage) growth for the group relative to the pace of growth of the other group. For example, per capita consumption of SCs grew at a 0.7 per cent rate in 1999-2004 and a 1.1 per cent rate in 2004-09. The other group’s consumption increased from a pace of 0.9 to 1.7 ppa. So relative to this group, the SCs had a 0.2 ppa lower pace during the NDA and a 0.6 ppa lower pace during the UPA>>.
Before we examine his aforesaid main point, there are certain glaring mistakes in his calculations that need to be mentioned.
1.  Growth Rates: The calculations of growth rates made by Bhalla are utterly wrong. Take for example the PPA growth rate of 0.7% for SC during NDA rule. Applying this rate to consumption level in 1999 of 477 will yield a figure of only 494 in 2004 as against 516 given by NSSO. Similarly, in case of UPA for the same group, by applying PPA growth rate of 1.1% would raise consumption of 516 in 2004 to only 545 in 2009 against the surveyed consumption of 586. Same trail of errors continues in all the remaining 10 computations of growth rates. See the (Natural) Log Growth Rates under column seven & eight for correct growth rates.
2.  Relative Growth Rates: Brazen subtraction of one growth rate from the other to derive a relative growth rate may be a simple exercise, but may turn out to be a stupid one as it disregards the underlying weightages or complexities of original rates. For example, take the case of poverty Ratios. In column 5, the relative growth rates (i.e. relative poverty reduction rates) for Muslim and OBC are given as 0.3 & 0.4. These positive rates may create a false impression that incidence of poverty has increased for these two groups, something not borne out by the survey data. Moreover, all the cells highlighted in RED are not even arithmetically accurate. In column 5, in place of -0.3, 0.3 and 0.4 for ST, Muslim & OBC, the correct figures should be 0.3, -0.3 and -0.6. Similarly, in column 6, in place of -2.4, -1.8, -2.3 & -2.4 for SC, ST, Muslim & OBC, the correct figures should be 2.4, 1.8, 2.3 & 2.3.
A better measure of Relative Growth Rates (RGRs) or Relative (Poverty) Reduction Rates (RRRs) is presented in columns 7 & 8. The RGR and RRR are worked out for each separate group by taking their Growth Rates or Poverty Reduction Rates as a percentage of those for population as a whole. The relative consumption growth rates now not only look superior under UPA dispensation than NDA, but in case of UPA, except the Muslims, they are above or very close to national consumption growth rate. In case of NDA, these are far below the national rate except in case of OBC & Others. Coming to poverty reduction too similar reversal of Bhalla’s case presents itself. The relative poverty reduction rate is above the national rate for all categories except the SC, under UPA. Whereas under NDA, SC, ST & Others look to have fared far worse than national poverty reduction rate.
The column thus makes not only unjustified claims, but has been grossly negligent in basing them on calculations that can be independently verified to be inaccurate. But, there are other issues too with such *back of the envelope analysis*. Planning commission has been very fond of toying with the poverty line and over the years has set up numerous committees to grapple with defining and counting the poor. Whatever may be the divergence of opinions about what constitutes poverty line and how to measure it, but one thing is for sure that counting poor in most official exercises is done only in the economic sense. Same is true about Tendulkar’s poverty line. Therefore, based on this unitary measure, *inclusion* is to be counted whenever any person moves out of poverty (on some criteria, however controversial) irrespective of his caste, creed or religion. There are deprivation measures, which take into account social, locational, health, hygiene, access, gender, age, religion, and such other crucial factors into account besides economic deprivation, and those would truly yield the kind of information columnist perhaps had in mind. But, these were nowhere considered in Bhalla’s column. Another interesting observation is that RGRs and RRRs show far more divergence in case of NDA than UPA. See graphs below.

*Why is it so?*, is worth investigating, but beyond the scope of this post. In conclusion, it is not the case here that UPA has done better in terms of inclusion contrary to what the Columnist had set out to prove; but to show the pitfalls in trying to fit the foot into the shoe of preconceived notions.

O O O O O O O O O O O O O O

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: