Following a methodology proposed by Jantzen and Volpert (2012), we use IRS Adjusted Gross Income (AGI) data for the United States (1921-2012) to estimate two Gini-like indices representing inequality at the bottom and the top of the income distribution. We also calculate the overall Gini index as a function of the parameters underlying the two indices. Our findings can be summarized as follows. First, we find that the increase in the Gini index from the mid 1940s to the late 1970s seems to be mostly explained by an increase in inequality at the bottom of the income distribution, which more than offsets the decrease in inequality at the top. The implication is that middle incomes gained relative to high incomes, but especially relative to low incomes. Conversely, it is rising inequality at the top that appears to drive the rise in the Gini index since 1981. Second, inequality at the top of the income distribution follows a U-shaped trajectory over time, similar to the pattern of the share of top incomes documented by Piketty and Saez (2003, 2006) and Atkinson, Piketty, and Saez (2011). Third, the welfare effects of the different forces behind an increasing Gini index can be evaluated in light of the Lorenz-dominance criterion proposed by Atkinson (1970): both top-driven and bottom-driven increases in the index appear not to imply strict Lorenz dominance by previous income distributions, and therefore are not associated with lower welfare in an absolute sense. In a relative sense, however, once average growth rates over the two periods are taken into account, the top-driven increase in inequality since 1981 appears to have been welfare reducing.