Interestingly, the argument I had with Dean Baker over makework jobs is a mirror image of the debate over the poverty rate, in which conservatives "prove" that there's no such thing as poverty because we really spend quite a lot on poor people, and liberals point out that it still sucks to be poor.
Again, this is fundamentally a debate over what we're trying to measure. If the poverty rate is supposed to be a measure of how many people are trying to live on the puny incomes that fall beneath the line, then of course conservatives are right: it doesn't measure that very well. We provide the poor with various forms of aid, from food stamps to transfer payments to Medicaid, that top up their purchasing power by 50% or more. (I've heard it convincingly argued that things like Medicaid and housing subsidies mean that the legally poor in New York are in much less danger of missing a meal than the lower middle class.)
But the poverty measure also measures something else: how many people in the United States are unable (for whatever reason) to secure the basics of life for themselves without substantial government aid. If we include government aid, we are in danger of obscuring that important figure, because we will be measuring variations in government aid as well as variations in the underlying economy.
It is true that (just as with government work relief) the government funds are not entirely exogenous--at any level of benefits, some people on them prefer welfare to work. Rationally so--the poor can face marginal tax rates that approach 100% on extra income, when you factor in the loss of benefits like housing, medical care, and subsidized childcare. But the change in the "pure" figure from year to year is still telling us something about the distribution of wealth in the economy that adding in benefits will obscure.
Most of the people who have been angrily emailing me have treated the question of make work jobs as if it were a referendum on government spending. One can believe that the WPA was a good idea, and also believe that one should leave it out of the unemployment statistics when trying to estimate the extent of economic recovery. Similarly, one can be in favor of poverty spending, and still think it's important to know what poverty looks like before taking the spending into account.