Felix Salmon’s speculative analysis of what exactly happened at JP Morgan is very worth reading. Even if he doesn’t have the details exactly right, he gives a good illustration of how JP Morgan could have gotten into a position where a unit that supposedly was responsible for reducing risk incurred a $2 billion trading loss.

In a nutshell, the narrative runs thusly:

  • “JP Morgan accumulated enormous new deposits in the wake of the financial crisis, both by acquiring banks and by attracting big new clients wanting the safety of a too-big-to-fail bank.” This deposits needed to be invested profitably.
  • Since there was little demand for loans, the bank chose to acquire assets in the financial markets. These assets were acquired by the same unit responsible for hedging risks elsewhere in the bank – the Chief Investment Office, headed by Ina Drew.
  • “Jamie Dimon, like everybody else, was worried about a Europe-induced financial crisis at the end of 2011, and so he told Drew to put on positions which would protect against such a crisis. She did so — only this time around, the crisis never happened, and Drew’s positions had to be unwound.”
  • “When that meltdown didn’t happen, simply selling those positions would involve realizing a substantial loss. And so rather than selling the positions, Drew decided to put on new, profitable positions which would offset the old hedge.” These new positions, needless to say, weren’t perfect matches for the old positions. While they usually moved in tandem – and hence acted as hedges – they could, potentially, move in opposite directions. And eventually they did.

Salmon concludes:

How did Iksil’s trade go so horribly, massively, wrong? Partly it’s because his position was so big and so public. When hedge funds worked out what he was doing, they managed to get the word out, using stories in Bloomberg and the WSJ. And then it was just a matter of watching the market do what it always does, when it smells blood: I’m told that Boaz Weinstein’s Saba, for one, made a lot of money taking the other side of Iksil’s trade.

Taking a much bigger-picture view, however, what was really going on here was that JP Morgan had hundreds of billions of dollars in excess deposits, thanks to its too-big-to-fail status. And rather than lending out that money and boosting the economy, Jamie Dimon decided to simply play with it in financial markets, just as a hedge fund would.

That’s a comfortable conclusion on one level, because it suggests that the problem is that JP Morgan was doing what it wasn’t supposed to: trading, rather than hedging. If only they had been following the Volcker Rule – or if only that rule were enforced more strictly – things like this wouldn’t happen.

But there are two problems with this conclusion: it’s very difficult to distinguish hedging from trading, and, on the other hand, it’s very difficult to distinguish investing from trading. To explain why, I’m going to have to explain what the CIO’s role is.

*     *     *

I don’t know Drew, nor do I know how things worked at JP Morgan specifically, but from my past experience the CIO’s job generally looks something like the following. The various business units of a bank accumulate portfolios which, in turn, expose the bank to risk. The CIO’s job is to look at the overall portfolio of the bank, and assess it for two kinds of risk: specific risk and systemic risk. Specific risk is over- (or under-) exposure to a particular credit. Systemic risk is risk that isn’t tied to a particular position, but arises from overall portfolio exposure to a particular market factor. And the CIO’s job is to hedge these risks if they are either simply too large or not risks the bank wants to take.

So, to pick a couple of simple examples: if the bank were under-exposed to the auto industry (because the bank had no strong auto company relationships), the CIO might purchase auto-related assets. If the bank were over-exposed to Japanese commercial real estate, the CIO might sell assets in this sector. If the bank had a view that the risk of the Euro unraveling was significant, which would have widespread negative repercussions across the bank’s portfolio, the bank might purchase an asset that produced a positive return if the Euro were to unravel.

These decisions by the CIO are going to be dynamic; the hedges they execute are going to change as the market moves, as the bank’s portfolio evolves, and as the bank’s “views” – about the economy as a whole, specific assets or asset classes, risk factors like the unraveling of the Euro, etc. – change. So it makes sense for the CIO to transact with a view to liquidity – the hedge may need to be taken off. As well, it makes sense for the CIO to try to execute hedges at the best available prices – which means being able to choose between different instruments that have similar (though not identical) sensitivities but that may trade at different prices. For both reasons, it makes sense for the CIO to have the full panoply of derivatives available to transact.

But once you say that, you’ve basically said that the CIO is running an internal hedge fund of sorts. You can’t readily tell the difference between a trading position and an optimal hedge (optimal being defined as maximizing some combination of liquidity and return). The only difference, really, between the CIO’s job and the job of a hedge-fund manager is that the CIO gets risk dumped in her lap by actual business units – that is to say, she has axes to trade against.

This has long been Dimon’s argument against the Volcker Rule: that, because you can’t distinguish between trading and hedging, the rule cannot be applied, and so all the rule does is waste traders’ and risk-managers’ time with semantic distinctions with no real meaning in terms of risk. And he has – and still has – a point.

But that’s not where the discussion ends. The problem isn’t so much that banks are allowed to trade, as that they find trading so profitable. And that, in turn, is related to, on the one hand, the fact that residual risks of hedged trading can never be fully captured by risk-management systems and, on the other hand, capital rules that encourage risk-reduction rather than risk-disclosure.

*     *     *

Let’s take a look at Salmon’s speculative narrative. Early on in his narrative, Salmon says: “With lots of deposits coming in, and little corporate demand for loans, it was easy for all that money to find its way to the Chief Investment Office, which could take any amount of liabilities (deposits are liabilities, for a bank) and turn them into assets generating billions of dollars in profits.”

There’s a problem with this sentence. “Demand” for loans and “supply” of cash balances do not exist independent of prices. If corporations don’t want to borrow money, instead preferring to lend their money to banks, that suggests that banks are paying too much interest on deposits, and demanding too much interest on loans. At lower interest rates, corporations would be reluctant to deposit money in the bank, and more interested in borrowing.

Why don’t rates continue to drop? This might be another instance of the zero-bound problem – there has to be a risk differential between government debt and corporate debt, so if the risk-free rate can’t go below zero then corporate debt yields can’t go below some positive number. Or you could describe it a question of return on capital – investors in bank equity expect a certain rate of return on their money, and banks can’t achieve that rate of return from rates at which businesses would be willing to borrow.

But in either case, if banks can’t make money lending, how can they make money trading? To answer that question, we need to get into capital rules.

Banks use different capital rules to handle “hold-to-maturity” positions like corporate loans versus trading positions like credit derivatives, and the capital rules are designed to give the bank credit for the reduced risk associated with hedging. All of which makes sense: a hedged position is less-risky than an unhedged position. But how do you know how well a position is hedged? How do you measure the residual risk?

The most important tool for most large banks for measuring risk for capital purposes specifically is VaR, or Value at Risk, which was developed (funnily enough) by JP Morgan. Without, again, going into too much detail, VaR uses the past behavior of an entire portfolio as the basis for predicting the possible future behavior of that portfolio, and regulatory capital schemes based on VaR require banks to hold sufficient capital to protect against what by historic standards would be extreme moves. VaR, when it was introduced, was considered an improvement on factor-based risk-metrics, because such metrics were only as good as their list of factors – leave a factor out, including one you’ve never heard of, and your risk is wrong. Since VaR used actual prices for the entire portfolio, by definition all factors are included.

But, of course, they are only included to the degree that past performance (in the sense of volatilities and correlations between assets) are predictive of future performance. Which, in the financial world, they generally are . . . until they aren’t. Volatilities can suddenly go to previously unseen levels. Correlations between previously uncorrelated assets can suddenly go to one. Correlations that have historically hovered around one can suddenly go to negative one. All these things can happen at once – indeed, they usually do. When these things happen, losses vastly exceed the predictions of VaR.

Our capital rules, overall, are designed to charge banks as accurately as possible for measured risks. They are not designed to prevent banks from taking unmeasured or unmeasurable risks.

This is a methodological limitation of VaR. And it’s a limitation that is most significant with well-hedged positions, because these are the positions that, from a VaR perspective, appear to have very little risk – and therefore require very little capital. Because they appear to have little risk and require very little capital, when they appear to be profitable it makes sense for a bank to acquire them in very large size. And then, when those historic relationships break down, even briefly, the bank can be exposed to very large losses.

*     *     *

So let’s go back to our CIO. She has a big portfolio dumped on her by the bank. It’s exposed to a systemic risk of some sort, which she dutifully hedges. The hedge loses money – that happens some time – and now the time comes to either take the hedge off (realizing the loss) or “hedge the hedge” with another instrument. Which does she choose?

Realizing the loss should not be a significant factor, because a trading portfolio is always mark-to-market, and both derivative and repo positions involve daily posting of collateral. There should not be significant consequences to either earnings or cashflow from unwinding the trade, even if it has gone against them. If there are, there’s an accounting problem somewhere.

So the choice comes down to that question of optimizing the combination of return and liquidity. It could well be that putting on a new “hedge-for-the-hedge” wins on both counts – the new trade could be in an apparently more-liquid instrument, and it could be trading at more favorable prices. If that’s the case, then why wouldn’t the CIO opt for it? Of course she would.

But if she does, then what does her book look like? Well, in addition to whatever else is in it (the various risks dumped on her by the different business units), there’s this huge trade matching one hedge against another. What’s the risk on this trade? The risk is the residual difference – the “basis” – between two highly correlated assets. So the trade is called a “basis trade.”

Basis trades are the bread-and-butter of so-called “market-neutral” hedge funds, and the CIO’s portfolio, in this hypothetical, looks an awful lot like such a hedge fund. Which is precisely what the Volcker Rule was supposed to prevent. But it got there via a set of completely reasonable decisions. So what can be done to achieve the Rule’s objectives more effectively?

I can’t see how the Volcker Rule could be applied to eliminate the choice of putting on a hedge-for-a-hedge – there’s nothing to distinguish this move from any other aspect of hedging. But I can see how capital rules could be tweaked to make basis trades less-attractive. If putting on a “hedge-for-a-hedge” is actually more profitable than unwinding the original hedge, that means the residual capital that needs to be held against the basis trade is low enough that the modest spread between the two highly-correlated assets generates an adequate return on capital – indeed, a return on capital that compares favorably with taking that capital and putting it against loans to actual businesses. That is the problem.

If you don’t want banks to trade, then you also have to disincentivize hedging, because the two cannot be readily distinguished. The CIO’s job should be, as it is, hedging systemic and specific risks in the bank’s portfolio. But any time the bank hedges, it acquires basis risk – the risk that the hedge won’t match up with the risk being hedged – and this basis risk, by definition, cannot be hedged and cannot be measured. By anything. So if banks are acquiring a lot of basis trades, that is prima facie evidence that the capital rules are out of whack, and that banks are running large risks that are not being captured by their systems.

The brute-force solution to this problem that I tend to go back to is some kind of charge on gross book size, sufficiently punitive that banks would be strongly incentivized to eliminate hedges – not re-hedge them with new instruments, but actually eliminate the trades – when they were no longer necessary for their primary purpose, and to only consider hedging when there is sufficient warrant. Such a charge would make derivatives books look much more like cash books, with regular processes for eliminating inventory. And it would basically make it uneconomical for banks to hold basis trades.

And what would banks do then? They might look harder for opportunities to lend profitably in the real economy. They might look harder for opportunities to make their own operations more efficient – paying their employees less and their shareholders more. They might lobby Congress for a more expansionist monetary policy, because they would need an upward-sloping real yield curve to make money. No doubt they will also find new ways to get themselves (and us) in trouble. The game never really ends. But if we want to play it better, we shouldn’t focus on scolding the banks, but rather on making them pay an appropriate price for engaging in activities outside their core function of turning savings into capital.