The liberal imperative

Noble efforts deserve the closest scrutiny. Which is why The Great Indian Poverty Debate engages the mind as much as the conscience. It's a debate of critical value; economic reforms remain vulnerable to their "impact on poverty", as Montek Singh Ahluwalia puts it.

But first, take this test. How many times did you eat""hmmm, insert lip-smacker of choice""last week? Next: how many times last month?

There's a good chance you'd overstate the frequency in reply to the second question. You don't mean to, of course""it gets 'biased' inherently by the first question. And good statisticians, jibes be damned, see themselves as bias busters.

The big bust-up of recent years has been academia's debunking of India's official poverty figure having declined to 26 per cent in 1999-2000 from 36 per cent in 1993-94.

This drastic 10-point drop, scoff sceptics, can be traced to a distortion of the National Sample Survey (NSS) of Household Consumption in 1999-2000. Respondents were asked for expenses on items consumed over the past week, as well as expenses on the same over the past month (as was the norm).

So people over-reported their monthly expenses, and this was sneakily used in comparisons to push poverty down. Actual poverty ought to be higher.

Nonsense, say others, NSS undercounts consumption. Surjit Bhalla uses separate data on wages to revise the poverty figure downwards""to under half of 26 per cent.

The book's focal point, however, is Deaton's own revision. The Princeton professor offers a daring statistical device that aims to negate NSS' recall-distortion bias, and thereby derive an adjusted poverty figure.

Forget the consumption items involved in the double-recall jeopardy, he urges. Instead, use six items that were left unmolested as a thin-slice sample to compare poverty.

You need the probability of a household being 'six-item poor' (which is available from the otherwise flawed 1999-2000 NSS data), modified by the conditional probability of a household being 'overall poor' given that it is 'six-item poor' (taken from valid 1993-94 data), under the plausible assumption that the relationship between overall and six-item poverty stays stable.

India's probable poverty figure, by Deaton's NSS-blending formula, works out to around 29 per cent.

It is indeed daring. The theoretical logic of the formula may hold good, but does it give the probability of being poor ""or the probability of being poor and six-item poor?

Perhaps the distinction should be blurred for convenience. Perhaps. But to anyone acutely and-sensitive, it gnaws: could poverty be even higher than 29 per cent? Uh-oh.

That's the book's effect. Once you get over the shock of India's poverty turning on a petty recall span, you'd be well advised to take the time to space out ... and marvel at those living in thraldom to figures wrenched out from their own computations. Familiar?

Misery is misery. And as Amartya Sen groans in The Argumentative Indian (chapter 2), India's record in asymmetry is too terrible for further neglect.

Patriarchal policies may have biases too, but to assign the job of evoking that smile to the 'Invisible Hand' may be to shirk responsibility.

THE GREAT INDIAN POVERTY DEBATE

Angus Deaton & Valerie Kozel (ed)
Macmillan India
Price: Rs 760;
Pages: 600