We evaluate the costs of three different programs for alleviating long term poverty: (i) workfare, or grants made contingent on a work requirement, (ii) universal welfare, or unconditional grants, and (iii) meanstested welfare, or grants conditioned on private labor market earnings. We identify the conditions under which each of these programs are cost efficient and compare them with the corresponding conditions for short term poverty alleviation. We find that the arguments for using work requirements are both strenghtened and weakened with long term poverty. We also find that as the fraction of really needy persons increases, it may be optimal to give up any attempt to screen the genuine poor from the spurious cases, even when screening is made easier by the possibility to observe earnings. (It is never optimal to refrain from screening when poverty is of short term).
