Doubts have been raised about the comparability of the size distributions and of the poverty measures based on them from the 50th (1993-94) and the 55th (1999-2000) rounds of Consumer Expenditure Surveys (CES) carried out by the National Sample Survey Organisation. We resolve the comparability problems (Section I) by using the unit level records of the 50th round of CES and those relating to consumer expenditure from the employment-unemployment survey (EUS) of the 55th round. In particular, we show that the estimated monthly per capita consumer expenditure (MPCE) from the 55th round of CES based on the 30-day recall have not been biased upwards (as maintained by the critics) by an alleged extrapolation by the respondents of their prior responses to questions on the 7-day recall (if the latter were canvassed first) and that therefore they are comparable to the (recalculated) ones from the 50th round estimates. Using comparable estimates of four measures of poverty and the Sen index at the all-India level (Section II), it is shown that poverty in India has declined in the 1990s in terms of all the five measures of poverty in rural India and in the country as a whole and in urban India on all the measures of poverty except the number of urban poor. Also, normalised for the length of the time interval and the base year levels of poverty measures, the average annual rate of reduction in poverty was higher in the last six years of the 1990s than that recorded during the ten and a half years period preceding 1993-94. This is so on all the five poverty measures and this difference is particularly significant in respect of the number of poor