Suppose you invented a policy, some kind of economic miracle, which doubled the incomes of the poorest ten per cent of families without the Government spending a pound. That would reduce benefit spending. It would also increase tax revenues from the poorest. The same method that the IFS are using in their reports would show the effects of that policy as horribly regressive, cutting spending on the poor and shifting the fiscal scales against them.
Of course that is an extreme and artificial example. But it shows the big problem with the IFS analysis, which essentially assumes that the fortunes of the poor add up to the amount of Government money spent on them.