This paper develops and applies new measures of poverty that overcome a number of specific methodological flaws in the official US poverty statistics. Sen's distribution sensitive index of poverty and each of its components are estimated at several distinct poverty thresholds for the period 1961-1996. Distribution sensitive measures of urban poverty are corrected for interarea differences in the cost of living and for comprehensive incomes. Recently developed statistical inference procedures are applied. Official poverty statistics are shown to be seriously misleading in some time periods and the choice of a poverty line affects conclusions concerning changes in poverty.