Surjit Bhalla, of course, has been pointing out for years that all poverty numbers are dramatically overstated since they're based on National Sample Survey data, which, with each passing year, capture less and less of the country's actual income as measured by GDP data -- in 1993-94, the NSS missed out 38 per cent of consumption and this rose to 52 per cent by 2004-05.

To put it crudely then, poverty estimates based on NSS data could be off by as much as half. The latest NCAER annual household survey, to use other data, points out that around 33 per cent of the bottom 40 per cent of households had a television set and another 12 per cent a two-wheeler in 2005 -- how they can all be poor is anyone's guess.

But this piece is not about debunking the Bank's latest poverty estimates using Bhalla's arguments or other data, no matter how correct they are. It's about the lack of internal consistency in even this data and this is something the Government of India needs to address since both use the same (under-estimating) NSS data but get dramatically different poverty numbers -- 27.5 per cent for GoI versus 42 per cent for the Bank.

It can't be just the different poverty lines used since India's official poverty line in 1993-94 was roughly identical to the 1.08 PPP dollars that the World Bank used then. So did the fact that the World Bank has now increased the poverty cut-off to 1.25 PPP dollars at the 2005 prices cause the sharp hike in poverty levels?

Sadly, it's not so simple. After every country's income level data are converted into US dollars at the market exchange rate, they are then translated into a PPP base, which takes into account the purchasing power in each country. Since this is done relative to the US dollar, the way to look at the change in the PPP exchange rate is to look at what's happened to US price levels. Once you do that, you find that 1.08 PPP dollars at the 1993 prices translates into 1.38 PPP dollars at the 2005 prices (the US GDP deflator, if you're into calculations, was 120 in 2005 versus 94.2 in 1993). In which case, the World Bank's 1.25 PPP dollars at the 2005 prices represents a lower income than 1.08 PPP dollars at the 1993 prices, and yet it shows higher poverty numbers!

So how does the World Bank get the higher poverty figures with lower poverty lines? Simple, it just lowers India's income levels, across the board, by� 36 per cent! It also does a similar number for a host of Asian countries like China, but that's of little consolation to us. For 2005, the World Bank website tells us, India's GDP in PPP dollars was 2,414 billion. Go to the World Bank's World Development Indicators CD, and the same figure is 3,780 billion PPP dollars! (You have to look at the CD, since all the figures on the website, even for the past, have been changed.)

And how were the figures revised so sharply downwards, since the US dollar numbers obviously haven't changed? PPP, to digress a bit, reflects the difference in costs between the US and, say, India. So, if a haircut costs a dollar in the US and just Rs 22 in India, the PPP exchange rate for haircuts will be 22 as compared to the market exchange rate of 44. If, however, you assume the Indian haircut is only half as good as a US one, the PPP exchange rate will also be 44. So, while calculating the relative exchange for the PPP, the World Bank (and the ADB, which actually did the Asian survey for it) assumed the productivity of services in India was a lot less than in the US!