feeds from sacw.net
print versionThe Economic Times, February 7, 2018
Word has it that World Bank economists use "obviously fabricated" data from time to time. These are not Sitaram Yechury or Medha Patkar’s words, but those of Paul Romer, former chief economist of the World Bank, in a recent email exchange reported by Financial Times. Romer retracted them later, but this "may not end the controversy", as The Economist mildly put it.
This not the first time that World Bank economists skate on thin ice. Another recent example concerns the widely-quoted estimate of $11 billion annual savings (or potential savings) due to Aadhaar. The original source of this estimate is the Bank’s World Development Report 2016. On page 195, the report mentions that many subsidies in India "are being converted to direct transfers using digital ID, potentially saving over $11 billion per year in government expenditures through reduced leakage and efficiency gains".
By way of source, a footnote cites a brief of the Consultative Group to Assist the Poor, prepared by Shweta Banerjee. But the $11 billion figure cited in that brief is not a savings figure at all. It is just an estimate of GoI’s total annual expenditure on ’major cash transfers’.
After we drew attention to this error, others pursued the matter and the World Bank arranged for soft copies of WDR 2016 to carry a correction. The correction, however, is most intriguing, as the sentence quoted earlier remains unchanged. It is merely the note referring to Banerjee’s work that’s now replaced with a much longer footnote.
The latter invokes two studies, one suggesting that biometric smart cards helped to reduce the National Rural Employment Guarantee Act (NREGA) leakages by 10.8% in Andhra Pradesh, and another showing that the 2013-14 experiment with direct benefit transfer of liquefied petroleum gas (LPG) subsidies (DBT-L) reduced the purchase of subsidised LPG cylinders by 11-14%, "suggesting a reduction in subsidy diversion". In both cases, technical hurdles "abounded", to quote the first study.
Answer Before the Question
But the footnote soldiers on, "Extrapolating these leakage reduction rates to all government of India welfare programmes" amounting to roughly $70 billion to $100 billion government expenditures "yields savings in the range of $8 billion to $14 billion or an average of $11 billion potential less spending." Voila!
The authors, in short, managed to return exactly where they were after along somersault.
But there is a catch: how come GoI is now spending "$70 billion to $100 billion" on "welfare programmes" when Banerjee had placed the relevant total at $11 billion? The World Bank graciously responded to a request for clarification.
It turns out that the grand total of $100 billion (more than 4% of India’s GDP in 2015-16, the reference year) is a capacious amalgam of multiple expenditure heads: half of all "central sector schemes and project", all centrally sponsored schemes, Finance Commission grants, etc. This would include expenditure on roads, bridges, horticulture, cattle, dairy and even wildlife.
The lower total of $70 billion is similarly bloated by including fertiliser and petroleum subsidies. Could it be that this imaginative definition of welfare programmes was retrofitted "so to speak" to the pre-determined ’estimate’ of $11 billion savings? All in all, the revised footnote looks more like an attempt to paper over the cracks than a piece of serious research. Two further observations are due.
First, the savings estimates do not pertain to Aadhaar as such but to the use of digital technology in subsidy transfers. Indeed, the smart cards study invoked in the said footnote does not involve Aadhaar at all. Even in the DBT-L experiment, Aadhaar is only one component of the project, and “a direct bank transfer can itself increase enforcement, irrespective of UID. The $11 billion figure is routinely construed as an estimate of"Aadhaar-enabled savings". But that is not what it is.
Second, the literature on evidence-based policy emphasises the need for caution with the ’external validity’ of quantitative analysis what has been found to apply in one context may or may not apply in another.
To illustrate, a biometric system may help to plug NREGA leakages in Andhra Pradesh, but not public distribution system (PDS) leakages in Jharkhand.
Indeed, there is evidence that Aadhaar-based biometric authentication led not only to severe exclusion problems, but also to a revival of corruption when it was imposed on the PDS in Jharkhand last year.
Aadhaar Card With Holes
If one were to extrapolate from the Jharkhand experience without attention to external validity, as the World Bank did from the Andhra Pradesh study, one would ’estimate’ that Aadhaar is all set to cause billions of dollars of additional leakages in welfare programmes. The World Bank’s extrapolations break all the rules of external validity.
The footnote mentioned earlier redeems itself with a salutary warning at the end, “It should be noted that realising these potential savings for all government transfers is conditional on accountable institutions to complement the investment in digital technology.†If only this sentence were widely quoted, instead of the shaky estimates that precede it.
(Dreze and Khera are development economists with Ranchi University and IIT-Delhi respectively)
