I think Birdsall has it more or less right in setting the lower bound at PPP10 a day, and for the reasons she sets out. However, more variation might be allowed for her upper bound, depending on the mean PPP income of the country in question. For countries with higher mean incomes, the upper bound might usefully be set at the 90th percentile, whereas for those with lower mean incomes, it could be the 95th or even the 99th percentile. But it should be income, not consumption, based. If it is to be consumption based, then the lower bound should be moved down. Banerjee & Duflo and Ravallion have it quite wrong to set the lower bound at PPP2, mistaking what is arguably a useful yardstick for measuring (the progress in) poverty (reduction) at the world level for a realistic indicator of a standard of living. I would be willing to bet anything that neither of them could have anything that might be considered a middle class existence on PPP4 a day in any of the middle income countries of East Asia. That Birdsall's measure results in no middle class in South Asia: doesn't that capture the view that South Asia, or more specifically, India's, growth of the past couple of decades has not had as much spread effects as some might like to believe? Comparing her table 3 and 7 is also instructive. While urban India apparently has no middle class by her criteria, if we were to shift her upper bound up to the 99th percentile, then it would be 4% which, in absolute numbers, would be sufficient to support the malls sprouting up.