This paper revisits Korpi and Palme’s (1998) classic article: “The Paradox of Redistribution,” which influentially argued that the more countries target welfare transfers at the poor, the less they reduce poverty. Korpi and Palme showed that income redistribution is greater when transfers are distributed broadly rather than concentrated on the most vulnerable, and this is partly because universal programs are more popular. Our study advances beyond Korpi and Palme by: a) updating the analyses to the 2000s; b) innovatively measuring the distribution of transfers; d) including a broader set of dimensions than just targeting and universalism; c) utilizing multilevel models of individuals nested in welfare states; and d) examining both poverty and preferences for redistribution. Using the Luxembourg Income Study (LIS) data on the distribution of welfare transfers actually received by households, we develop novel measures of the extensity, universalism, generosity, targeting and dualization of transfers. With multi-level logit models of individuals nested in countries, we then examine the association between these dimensions of transfers with poverty (using the LIS for 19 rich democracies; N=1,032,016) and with redistribution preferences (using the International Social Survey Program “Role of Government Module” for 14 rich democracies; N=14,258). Consistent with Korpi and Palme, we find that the dimensions of welfare transfers do matter to poverty and redistribution preferences (net of a variety of individual-level controls). First, poverty is very strongly and significantly negatively associated with extensity of welfare transfers – the percent of average household income coming from transfers. Second, targeting, universalism, generosity, and extensity are not significantly associated with poverty. Third, redistribution preferences are significantly positively associated with the universalism – the uniformity of transfers across the population. Fourth, redistribution preferences are also significantly reduced by dualization – the heterogeneity of transfers within the poor. Fifth, extensity, generosity and targeting are not significantly associated with redistribution preferences. Therefore, we revise the “paradox of redistribution”, which claimed that universalism reduced poverty because it was more popular and targeting worsened poverty because it was unpopular. Instead, we demonstrate that the paradox is that extensity is most effective at reducing poverty, but universalism is most popular. Therefore, what reduces poverty is not popular, and what is popular does not reduce poverty. Moreover, unlike Korpi and Palme, we find no clear association between targeting and either poverty or preferences. We provide supplementary evidence by showing that income inequality is negatively associated with extensity, and Left party affiliation is positively associated with universalism. Finally, by examining the interrelationships between different dimensions of welfare transfers, we demonstrate that there is no clear association between universalism and extensity.