I have mixed feelings about this part of Obama's proposed stimulus package. On the one hand, if not for pervasive income mismeasurement by the tax system it would be madness not to treat losses as fully refundable. Otherwise, one discourages risk-bearing (since the government in effect says "heads we win, tails you lose") and offers inefficient incentives for corporate conglomerates (so one activity's losses can be deducted against another's gains). But with pervasive income mismeasurement and tax sheltering opportunities, it's more problematic. Imagine what Enron would have done had losses been refundable.
So extending NOL carrybacks is a move towards refundability, with mixed merits, but it happens after the fact for losses that have already occurred. This takes care of the tax planning problem for losses to date, but also eliminates the significance of making planning decisions more neutral ex ante (at least for those past decisions).
Perhaps another factor to consider here is that, as discussed in Alan Auerbach's recent paper (with Rosanne Altshuler) that we will be discussing at the NYU Tax Policy Colloquium on Thursday, Jan. 22, losses at the corporate level have become more common, predating the recession and apparently reflecting an economic shift towards greater divergence in business outcomes. That presumably strengthens the case for refundability by showing that it's more of a problem.
From the standpoint of Keynesian stimulus, the issues are somewhat different. Unclear to me that handing $$ to companies that happen to have experienced big losses recently is necessarily the best way to encourage further economic activity. If they have been losing money lately, are they the ones who would invest and hire more if someone handed them money? Is being cash constrained their big problem? Are they the ones to whom handing dollars would be most stimulative? (As opposed to consumers, or else the businesses - if any exist these grim days - that actually would run out and hire & invest if only they had the cash in hand.)
I'm reminded of TARP's fiasco in handing money to banks that did not respond by lending it out again because they didn't want to lend, didn't consider it a good move in the current economic environment, as opposed to being cash-constrained or simply paralyzed by the state of their balance sheets.
So this idea is not necessarily bad policy, but it might not be especially good stimulus.