Catherine Rampell has an excellent blog post on this question, here is one bit:
Since the mid-1990s, the biggest increases in spending have gone to those who were middle class or hovering around the poverty line. Meanwhile, Americans in deep poverty — that is, with household earnings of less than 50 percent of the official poverty line — saw no change in their benefits in the decade leading up to the housing bubble. In fact, if you strip out Medicare and Medicaid, federal social spending on those in extreme poverty fell between 1993 and 2004.
Then, during the Great Recession and not-so-great recovery, automatic stabilizers kicked in and Congress passed new, mostly temporary, stimulus measures (such as unemployment-insurance benefit extensions). As a result, spending on the social safety net increased sharply and this time for a broader swath of Americans, including the very poor, “near-poor” and middle class. But it still rose more for people above the poverty line than it did for the very poor, Moffitt found.
Other public policies not captured by Moffitt’s calculations have also effectively diverted funds away from the very poorest Americans. Consider the rise of “merit-based,” non-means-tested financial aid at public colleges or the increasing number of tax breaks and loopholes known as “tax expenditures,” more than half of which accrue to the top income quintile.
Since the early 1990s, politicians have deliberately shifted funds away from those perceived to be the most needy and toward those perceived to be the most deserving. The bipartisan 1996 welfare reform — like the multiple expansions of the earned-income tax credit — was explicit about rewarding the working poor rather than the non-working poor. As a result, total spending per capita on “welfare” slid by about two-thirds over the past two decades, even as the poverty rate for families has stayed about the same. Many welfare reformers would consider this a triumph. If you believe many of the poorest families are not out of work by choice, though, you might have a more nuanced view.
Meanwhile, there is probably greater political cover for expanding the safety net for the middle class (that is, the non-destitute). As mid-skill, mid-wage jobs have disappeared — what’s known as the hollowing-out of the labor market — middle-class families have lost ground and are demanding more government help. These middle-class families, alongside the elderly, are also substantially more likely to vote than are the poor. The feds have whittled away at welfare, and (almost) nobody has said boo; touch programs that the middle class relies on, and electoral retribution may be fierce.
The piece is interesting throughout.