Recessions, in fact, appear to affect disproportionately the extreme poor, rather than those closer to the federal poverty threshold or the "near poor," those whose income is less than twice the federal poverty threshold.
Consider this: in 2010, 6.7 percent of Americans were among the extreme poor, as compared to 5.2 percent in 2007 and 4.5 percent in 2000. That's a 50 percent increase in the fraction of extremely poor individuals - the greatest increase, by far, of any income group relative to the poverty threshold.
The unambiguous statistical trend since 2000 has been large increases in the fraction of Americans at the extreme end of poverty, with little to no change in the fraction of Americans considered "near poor." The poor, in other words, are getting poorer - or more precisely, poverty in America is becoming an increasingly extreme and unequal phenomenon.
Observe, for example, that since 2008, the percentages of individuals making between one-half and three-fourths of the poverty threshold, and between three-fourths and up to the threshold, have seen the second and third largest growth since 2000 - 27 percent and 21 percent respectively. The former has increased from 3 percent of Americans to 3.8 percent, the latter from 3.8 percent to 4.6 percent. Meanwhile, the fraction of near-poor individuals has remained roughly stable in the 18 percent range since 2000.
If you're an economist, you might notice that the statistical behavior of the two most extreme poor groups in the first graph - less than half of the threshold, and between half and three-fourths of it - is dramatically different from the other income brackets. They appear highly sensitive to recession, rising as sharply as a fraction of the population during the recessions of 1981, 1990, 2000, and 2008. All other poverty-threshold groups show little to no cyclical behavior.
The graphs and original article are on RSN