Doing the subsidy maths March 14, 2013 at 12:35 am

There has been a blog-fight between Bloomberg, whose editorial suggested that large US banks enjoy an 80 bps funding subsidy from the tax payer, and Matt Levine, who came to, well, a lower number. Now, I don’t really have a dog in this fight, but I was amused to note that SIFMA, a trade association, quoting the IMF, came to a 20bps subsidy.

Let’s assume that the subsidy is indeed 20 bps, and moreover that that 20 applies just to non-deposit funding. We will take JPMorgan, as that seems to be the paradigmatic example. JP has roughly speaking $2.4T of assets, funded by $1.2T of deposits, $200B of shareholder’s funds, and $1T of debt (quite a bit of it short term). So suppose JP enjoys a 20bps subsidy on that $1T*. That comes to $2B. Two billion dollars. To put this number in context, JPM’s last dividend payment was roughly $1.1B (30 cents a share last quarter to 3.8B shares). So the annual state subsidy JP gets, using trade association numbers, covers 40% of what JP gives shareholders. Um. I don’t know about you, but if this is even vaguely plausible, then the US taxpayer could legitimately be quite peeved about it‡.

*Obviously the 20 is a blended number; it won’t apply equally to all maturities of debt, nor equally to secured vs. unsecured funding.

‡For an earlier discussion of the UK taxpayer, see here.

5 Responses to “Doing the subsidy maths”

  1. The subsidy maths – for long dated debt the subsidy is way more than 20pts. For short dated debt it is way less than 20pts.

    Ask yourself this way – how comfortable are you having 1 day exposures to JPM? Does the subsidy massively change that?

    How comfortable are you having 10 year exposures to JPM? Does the implicit guarantee change that.

    Answer is obvious – Credit Curves get more expensive for long exposures. There is a reason.

    The subsidy is larger for long exposures.


  2. It seems to me there’s two ways to measure the “too big to fail subsidy.” Essentially it’s a put option (in the expected, if not guaranteed sense) for (at least some) creditors of the TBTF banks.

    There’s two ways to measure the value of these put options. One is to use their expected payouts, the other is to use their market implied value. As a very large volume of history has shown the former is less than the latter. I.e. options tend to pay out less than they are purchased for, particularly deep out of the money put options.

    What’s the correct way to calculate the cost of options that are valued at say $15, but only on average pay out $5? I’d say in terms of the cost imposed on the taxpayer/currency holder (depends on if the bailout comes from tax revenue or an expanded money supply) is $5.

    There’s good economic justification for why put options, especially deep out of the money systematic put options, earn a net premium. In the event that the major asset markets shed that much value, most investors will be highly capital constrained and in danger of going bust. Avoiding even a small loss in this state of the world is paying higher than its expected value. Gambler’s ruin is the instructive parable here.

    One of the primary benefits of fiat currency though is it allows for a player, the central bank, that is unaffected by the Kelly criterion. The central bank can never go bust because it can always print more money. This gives the central bank a unique position as lender of last resort to step in at times when assets are priced below expected value and stabilize markets by being the only player able to arbitrarily finance any position of any size at any time.

    To a central bank with a printing press this role is basically costless, whereas for any other capital constrained participant guaranteeing this would demand a very high premium.

  3. John – Um, yes, of course. That’s why I called it a blended number.

    As to your question, I would use a large exposure (23A for Americans) regime, and limit my 1 day exposure to any bank to, say, 10% of my capital. JP is no different. And yes, I’d want a (much) smaller limit on my ten year exposure to them. But I don’t think I implied anything different, did I?

  4. Doug – I agree with all of that. The problem, though, is that those OTM puts are very sensitive to the assumed (real world) return distribution. Make the tails a bit fatter and they get a lot more expensive. So measuring the subsidy that way has a lot of model risk. Also, in some sense, central banks are writing liquidity not equity puts – and those are even harder to price. But conceptually absolutely I think your model is right.