Gujarat Malfunction: Debunking ‘Debunking The Facts’.

Recently, Asifa Khan, a member of the Gujarat BJP, and Zafar Sareshwala, a Gujarati businessman who opened a dialogue with Modi avowedly to improve the lot of Muslims in Gujarat, wrote an article in First Post. It was titled, “Debunking the ‘facts’ on Narendra Modi and Muslims”. Both the individuals have figured very prominently in a series of articles written by Madhu Kishwar under a magnum opus heading: ModiNama. The authors of First Post opinion piece were probably riled no end to have responded next day to the Indian Express article of Christopher Jaffrelot: “Silence speaks More”, which appeared on the 7th November. Why should Jaffrelot’s opinion matter so much in Modi’s scheme of things to deserve such swift counter attack? Is it because, though of little value in India’s electoral fortunes, it is perceived to carry weight with French and British policy makers? Be that as it may, Khan-Sareshwala duo chose the authority of Surjit Bhalla to *debunk* the facts often cited for “Modi’s Minority Malfunction”.
To the credit of Surjit Bhalla, whose several articles drew my scrutiny and reasoned criticism, he went by the evidence he found by the criteria he chose, he then interpreted it, and finally presented his findings uncoloured by prejudice. His most recent article of 26thOctober, “Lessons to be learnt from Narendra Modi’s Gujarat”, drew exactly opposite conclusions than the one he wrote on 13th December 2012, “The Modi Metric”. While both the articles worked from the identical baseline of year 1999-2000, the former benefited from the latest round of Monthly Per Capita Household Expenditure Survey (MPCE) held in 2011-2012. The December 2012 column was based on an earlier MPCE round conducted in 2009-2010, whose findings were held to be erroneous as these were expected to have been adversely affected by the financial crisis of 2008 and the draught of 2009. However, before we take a closer look at the figure work of Bhalla, I wish to draw attention to a an inaccurate comment he has made while comparing the approaches of eminent economists Amartya Sen and Jagdish Bhagwati..
Two of India’s leading economists—Jagdish Bhagwati and Amartya Sen—have been positing their distinct and separate models of economic development. While both desire and want development, their prescriptions have an important difference. Simply put, Mr. Bhagwati would let economic growth do most of the work while Amartya Sen would want active government intervention in the form of doles, preferably doles by right, to make right the misfortunes of the poor”.
One approach says economic growth would automatically deliver social goods (Hygiene and Sanitation, Food and Nutrition, Literacy and Education, and Health services). This may ultimately happen as it did in Europe and North America, but at great cost, displacement and upheaval. Moreover, in the industrial transformation of Europe, the plebeian’s sufferings were greatly cushioned by colonial conquests and plunder –a prospect firmly closed to former colonies. On this path, Growth and Development are sine qua non with immediate Displacement and Dislocation; and are accompanied by exhortations to Patience and promises of Deferred Benefits. In a parliamentary electoral democracy like India, which has ushered in political equality in principal –one person one vote, but has failed miserably to deliver after over six decades even basic social goods; peoples’ patience has worn dangerously thin. Add to this the fact that in a country of 1.2 billion if more than half of its members are *handicapped* to meaningfully contribute to growth and development; then at best they would simply be a drag on growth, and at worst would *blindly* oppose it tooth and nail for what they see as imminent worsening of their lot. Sen and et all therefore see providing social goods as a prerequisite to boost sustainable growth rather than expect these as latter’s *inevitable* fallout. It is not classroom Economics, but real life Political Economics that would set the course and determine the outcomes.
Bhalla’s article of a year agocarries comparative figures on education: **An illustration of performance relating to youth female education among Muslims is informative. Bihar is ranked fourth in performance, Gujarat 16th (out of 17 big states), and Himachal Pradesh is ranked first. The raw data for Muslims is an improvement from 2 to 4.1 years in Bihar, an improvement from 5.6 years to 6.4 years in Gujarat, and an improvement from 4.2 to 7.4 in HP. The improvement in non-minority youth female education for the three states is as follows: 1.9 years for Bihar, 1.8 years for Gujarat and 1.6 years for HP. Clearly, in terms of equity defined as relative performance of Muslim females, Gujarat is near the bottom**. The current articledoes not mention comparative education figures at all. Is it because of what Bhalla chose to present on two occasions or is it because this data was missing in the latest round of NSSO surveys? I do not have answer to that in the absence of access to NSSO data; but it can be safely surmised that the foregoing conclusion still stands validated. On another count, Bhalla has proved his own prediction, which a he made a year ago, wrong. He had then predicted, **It is unlikely that availability of data for the last two years, 2010 to 2012, will change any of the findings. Gujarat will remain one of the top states for GDP growth, and one of the worst states for equity and/ or inclusion. Interestingly, Bihar, over the last decade, is near tops in both growth and equity**. Not only he has chosen to repose his trust in the data for 2011-2012, but he has found his earlier conclusions turned on their heads. Since the latest NSSO round was undertaken in the shadow of the *questionable findings* of earlier 2009-2010 round, the sample size was considerably enlarged to make the sampling process more comprehensive and the findings more robust.This fact alongside two other facts mentioned earlier –financial crisis and drought year- persuaded Bhalla to give far more credence and weight, which he has justifiably done, to the latest round of MPCE survey conducted in 2011-2012. His findings are summarized in the table (see below) he gave with his article.
The rankings in panel B have been worked from data not made available in his article. Therefore, and since these are only ordinal numbers (Rankings), no further scrutiny is possible. However, panel A data not only permits further analysis, but also cross checking for the integrity of the data set and for validating the conclusions drawn. After stating the NSSO figures in main columns 2 and 3, the column 4 gives derived figures of “change in poverty level %”, which are simply arithmetical difference in poverty % numbers between 2009-2010 less 1999-2000 and 2011-2012 less 1999-2000. Based on this Bhalla has drawn the *obvious* conclusion that: “The latest 2011/12 NSSO data radically changes the conclusions and interpretation of the nature of inclusive growth in Gujarat. The sharpest decline in poverty between 2009/10 and 2011/12 is observed for the Muslims, the very community against which a Gujarati-Modi bias is assumed and presumed. The poverty ratio for Muslims, which had not shown much change between 1999/00 and 2009/10, now collapses to only a 11.4% level from the high 37.6% level observed just two years earlier.At this level, the Muslim poverty rate is marginally below the 12.4% poverty rate of the non-disadvantaged group consisting of OBCs and upper-caste Hindus”. In other words, Muslims are doing economically (poverty numbers) better than non-disadvantaged group, and far better than OBC (18.9%). This dramatic result (comparable poverty level figures for 2009-2010 to 1999-2000 period are Muslims 37.6% and non-disadvantaged group 14.8%) barely conceals an *eye-sore* crying out desperately for attention: the poverty level in fact increased to 4% from 2.6% for the “Other group” in just two years between 2011-2012 and 2009-2010, whereas all other classes (SC/ST-5.80%, Muslims-26.2%, OBC-6.8%) showed significant declines. How did this anomaly escape attention? Is it consistent with reality? A closer look at the figures in panel A shows further inconsistencies. See the table below.
The figures in first sub-column of the 3rdand 4th main columns are taken as it is from the two sub-columns of last main column in Bhalla’s panel A. The figures in the first sub-column of 5thmain column (2011-2012 and 2009-2010) have been worked out in much the same way they have been worked out in the last main column of panel A. The second sub-column of 3rd, 4th, and 5th main columns give the contribution % made by each of the 4 sub-groups –SC/ST, Muslim, OBC, and Others- to the change in overall poverty level %. Therefore, contribution % of SC/ST and Muslims should equal contribution % of Disadvantaged group; and likewise OBC and Others should equal Non-Disadvantaged group. There are two additional Rows added called *Disadvntgd(Cal.)* and *Non-Disadvntgd(Cal.)* that appear immediately after the rows *Disadvantaged* and *Non-Disadvantaged*.respectively. While the latter pair is a direct product of respective changes in poverty level % and share in population %; the former pair is interpolated (the abbreviation cal. stands for interpolated)from the sum of contribution of each category % (Disadvntgd(Cal.)=SC/ST + Muslim, and Non-Disadvntgd(Cal.)=OBC + Others) adjusted for its combined share in population. Ideally, both these pairs would have matched perfectly but for the rounding off errors had the data presented been consistent. Even a cursory look at the table shows that not in a single case there is an agreement in any pair. In fact, some pairs differ widely and significantly in their values. This certainly raises a question mark about the quality of the data.
There is another game changing consequence. Regarding the rankings of the panel B Bhalla says “In the accompanying table, the lower panel B reports on the performance ranks for the different population categories. The rank for each category is based on the poverty decline in that category with reference to the decline observed for the non-disadvantaged. For Gujarat, the non-disadvantaged have a poverty decline of 15 ppts between 1999/00 and 2011/12. For SC-STs, the poverty decline is 28 ppts for the same period. So, the excess poverty decline for SC-ST is 13 ppts, and this is the third-largest excess decline in the country for SC-STs. In the case of Muslims, Gujarat was the second-best performing state,..”. As against this, the adjusted or non-disadvntgd(Cal.) poverty decline stands at 17.88% (2.88%-point improvement over Bhalla’s figure), and thereby reduces the excess poverty decline for SC/ST to 10% -a scaling down by full 3%-point. Similarly, the excess poverty decline for Muslims too is scaled down to 10%from 13%. This 3%-point negative change from Gujarat’s perspective would conceivably change Gujarat’s rankings significantly. If one were to compare poverty decline % of disadvntgd(Cal.) as a whole with poverty decline % of non-disadvntgd(Cal.), then the excess poverty decline of former now reduces by whole 5%-points to 10% (17.88%-27.85%) from the earlier 15% (14.90%-29.80%). In fact, another glaring error here that has surprisingly escaped notice: if change in poverty level % for SC/ST is -27.80% and for Muslims is -28.00%, then how conceivably the joint poverty decline for the two (-29.80%) can be greater than individual poverty declines.
However, the meanest hole in the fresh narrative by Bhalla has been made by the biggest surprise of all. Bhalla has gone to some lengths to justify why he reposes far more confidence in the NSSO data of 2011-12 round than of 2009-2010 round. Bhalla would therefore undoubtedly be conversant with the concept of Outliers in statistics: “Outliers are often indicative of measurement error or at times indicate that the population has a heavy-tailed distribution. In the former case one wishes to discard them or use statistics that are robust to outliers, while in the latter case they indicate that the distribution has high kurtosis or has two distributions mixed”. The last column of last table gives the change in change in poverty level % between two periods of 1999-2000 to 2009-2010 and 1999-2010 to 2011-2012. Muslims have the most abnormal change in change in poverty level % of 1455.56% when compared to SC/ST (26.36%), OBC (51.13%), and Others (-8.54%). If this figure of 1455.56% is not an outlier, then tell me I am shamming. This single figure casts a heavy doubt over the whole edifice of otherwise careful surmise. Bhalla has yet gone ahead presenting his findings completely oblivious to the insurmountable obstacle of this outlier.
0 0 0 0 0 0 0 0 0 0 0 0 0
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: