Each day seems to reveal more losing investors in Bernie Madoff's house of cards. See Scheer, Probe of Madoff Intensifies, Bloomberg.com. Dec. 21, 2008 (suggesting 4000 investors, with problems going back to the 1970s). As the number of losing investors increases, the interest in figuing out the tax angles in the Madoff ponzi scheme increase. How realistic is the possibility of a theft writeoff? See, e.g., Browning, Tax Deduction May Help Duped Investors. Depends, at least in part, on the investor's real status (for tax purposes)--equity holder (theft write off might be ok, and not as repugnant as 'CG preferential rates, theft write-off ordinary rates' as it seemed when I first thought about it, if most of the money would have been short-term capital gain because of the predominant use of option strategies, as one reader pointed out) or debt holder (with a bad debt loss rather than a theft loss). Of course, it also has to be a theft (generally, a state law definition) for the theft loss rules to apply, but that may not be a big hurdle for something that looks at this point an awful lot like embezzlement. Is confession to fraud enough, or will conviction be necessary? When will everyone know that there are no assets hidden away that could be liquidated to pay off claimants? Are there assets hidden in Liechtenstein banks that could pay back investors? They may not know for quite a while. These timing and recovery possibilities are big issues with theft losses. Roth & Co. seems to think the IRS will be fairly lenient in its application of the rules in this case: see this post. Maybe.
For those wanting more, there is a discussion among Proskauer Rose attorneys that has been posted on the firm's website, titled simply "The Bernie Madoff Case." It includes both an audio link and a transcript. It first summarizes the litigational aspects of the case, from arrest and seizure of assets; handling by SIPC --which insures up to half a million per client for those invested through BMIS, Madoff's firm that was a broker-dealer conducting market making and investment advisory services; the stay on any litigation by claimants and transfer of the case to a bankruptcy judge; to the fact that the company's assets are likely to provide little or nothing for claimants. Then it talks about the treatment (legal and tax purposes) of payments, There are two kinds--redemptions (return of investment) and dividends (earnings). Since BMIS was really a ponzi scheme, there could be a fraudulent conveyance issue with respect to both (at least, for redemptions and for purported earnings paid out), but investors could keep their redemption payouts if they are able to use the "good faith" exception, while profits payments are more likely to be subject to reclaim as fraudulently conveyed to investors rather than to creditors.
Here's what the Proskauer tax attorney says about those losses:
Looking first at the losses of investment, there the law is clear that on the basis of the facts, as we know them, a theft loss will be available. This is a very good loss from a tax standpoint. It is an ordinary loss that’s deductible against ordinary income. It is not subject to limitations, like the 2% floor, and it does form the basis for what's known as a 'net operating loss', which carries back, with regard to a theft loss--it carries back for three years--and it is available to carry forward for 20 years. So, this loss is very effective.
I might add that to be a net operating loss, losses have to exceed income--this loss by itself may not result in a net operating loss.
What about investors who paid tax on earnings payouts that were only phantom profits, as it turns out? There, the Proskauer attorney acknowledges the lack of certainty about the options: refund claims only apply for open years; while authority from 2004 that cannot be cited as precedent suggests a theft loss claim is possible, the Service could well review the matter in this context and decide it differently; and the applicability of the claim of right doctrine under section 1341 is uncertain in this context.