Bank Regulatory Capital has been in the news a lot recently. The Spanish banks have their own set of capital issues. There has been a lot of discussion about Too Big To Fail (“TBTF”) in the U.S. with regulators demanding more and banks fighting it. After JPM’s surprise loss this month, the debate over the proper regulatory framework and capital requirements will reach a fever pitch. That is great, but maybe it is also time to step back and think about what capital is supposed to do, and with that as a guideline, think of rules that make sense.
What Is Regulatory Capital?
Simply it is the amount of capital that the banks are required to hold against their assets. Generically, though the concepts have been evolving with various Basel Accords, regulatory capital for various debt instruments is 8% of the bank’s risk-weighted assets. Risk weights vary dramatically – from 0% to high-rated sovereigns to 50% for an investment grade corporate bond rated “A” to over 150% for high-yield corporate bonds below “BB-“. Under Basel II, there are also regulatory capital charges for CDS with respect to counterparty exposures (under a complicated formula), and rules under which CDS provides partial offsets for cash assets regulatory capital requirements. But those risk weights are multiplied by the 8%. So “AA”-rated sovereign uses NO regulatory capital (infinite regulatory capital leverage), a “A”-corporate is 4% charge (25x leverage), and a high yield “B” bond is 12% charge (8.3x leverage). From regulatory capital perspective, therefore, the banks are incentivized to put on large positions of highly-rated debt.
What Is Regulatory Capital Supposed To Do?
Surprisingly enough, there are actually lots of different answers. Many believe that it should cover a bank’s potential losses in a portfolio with some cushion. That is wrong. No bank ever gets the luxury of seeing if the over the long run the capital is enough to cover actual losses. The real world doesn’t work like that.
Bank runs start when people become concerned that if a bank had to liquidate its portfolio, there wouldn’t be enough money to do that. So no matter what style of accounting a bank uses, at some basic level, investors react to the perceived value of the assets, not eventual value. That is a key distinction.
Depositors reacting that way are the biggest threat, because if they pull deposits, then banks have to sell assets, even those they hoped to hold until maturity. If the bank isn’t relying on short term funding, the run is less likely, and they have more ability to weather the storm.
So, regulatory capital, or capital adequacy, or just plain capital needs to address the worst of eventual loss and potential mark to market loss. Since mark to market loss risk is almost always worse than eventual loss risk that has to be the key focus. The way a bank funds itself is also important. The less it has in “demand” deposits or short term debt, the more control they have over the asset side of the balance sheet (no forced selling), so the liability side has to play a role in capital determination.
Let’s start at a very basic level. Owning $1 million, $10 million, or $100 million of an individual bond is very different in the real world. In some probability of default and eventual recovery analysis the regulators would demand a fixed percentage of capital for each position. If default expectation was 10% with a 50% recovery, then the bank would hold 5% capital against the position.
But that isn’t right in the real world. No matter what the risk of eventual loss is, the risk in a trading environment is much different as the position grows. To sell a $1 million position, you can call any number of dealers and get out at a price. If the first place you call is out grabbing coffee, you don’t care and just dial another dealer.
If you have $10 million, you have to be a bit more selective. You need to work with someone a little bigger, and possibly more specializing in that bond. Or you can alternatively “spray” the street and sell small size to a bunch of dealers. In any case, the level of difficulty of cutting the asset position has increased.
In the case that you loaded the boat and bought $100 million, selling it will require some effort. That is now potentially a market moving side. Instant liquidity is impossible. You are going to have to work with someone. Exiting will take a period of time and even then likely will move the market against you.
So in a “progressive” capital system, if 5% is the “right” expected loss and capital that can be applied to the small position. But at $10 million maybe it increases slightly to 6% but by the time you get to $100 million it is 10%. It isn’t saying that you cannot be large and outsized in a position, but it is saying that you better think that extra risk is worth the extra capital charged.
Notice, that nowhere have I mentioned ratings yet, and I won’t, at least not directly. Figuring out the “expected” loss may require some ratings based rule, or internal models, I’m willing to concede that. But the liquidity has nothing to do with ratings. The liquidity will be applied diligently across all assets.
The financial crisis in the U.S. and that in Europe have one thing in common – it was huge positions to what was formerly thought to be a low risk asset that turned ugly that caused the problems. In the U.S. it was primarily AAA tranches of various forms of MBS and ABS. In Europe it was primarily sovereign debt. For all the fear of HY bonds, other than Drexel, not once have they been at the heart of a financial institutions meltdown – and even the Drexel case was less about the risk of the bonds themselves.
So by ratcheting up capital as position sizes increase, you stop banks from building positions that are too large. That liquidity premium increases with size is there, because you can almost guarantee that by a time a bank is selling one of these formerly “safe” assets, they are selling into a falling market.
Now we have to apply the same concept across portfolios. If you have 50 positions of $10 million each, you have more risk than having 20 positions of $10 million each. So again, the concept of additional, or incremental regulatory capital charges for larger total portfolios needs to be applied. Some of this will be picked up in the asset by asset capital charge, but it is important to look at the overall portfolio. Bigger portfolios add more risk in an unwind scenario. That needs to be accounted for.
Now we circle back to “expected loss”. To use a concept like “expected” loss, you need a large enough sample of assets to assume that it will perform as “expected”. For fixed income this concept is particularly important because in the end, there are only two states – defaulted and trading at recovery value or paid in full. Over a well diversified portfolio, I’m willing to accept this concept – so long as the additional capital charges for size are in place, but we need to look at problems with applying it.
A portfolio with 10 $1 million positions is different than a portfolio with 1 $1 million position and 1 $9 million position. The “expected” loss might be the same, but the deviation of outcomes is much different. To illustrate a point, let’s assume a 50% recovery (which in the real world is also a variable that moves around over time and is different for each company). In the first portfolio you could have anywhere from 0 to 10 defaults, so the range of outcomes is 0 loss to $5 million loss, but the different scenarios will center around the estimated default probability and the correlation. A little bit too statistical for my tastes, but I think it’s workable and can be done in a way that creates a sound banking system.
The second portfolio also has the same range of outcomes, but in reality there are only 4 scenarios. Both default, both survive or only 1 defaults. In the both default or both survive scenario it looks a lot like the 10 name portfolio. But if only 1 name defaults, you become very sensitive to which name defaults. If it is the large name, then the bank will have a loss that would require 9 defaults in the well diversified portfolio.
So expected loss has to be adjusted for large positions as well. The more “smooth” a portfolio is in terms of size, the more sense it makes to apply an expected loss rate uniformly. But large individual positions in that will cause disproportionate problems in the downside if there was a bad decision made.
So expected losses can be applied uniformly across a well diversified portfolio of where each position is of similar. For less diverse portfolios, or individual positions that are relatively large, the expected loss has to be adjusted upwards to provide more protection for bad decisions. This is in addition to the “liquidity” surcharge.
Will any of this ensure TBTF doesn’t happen? No, but it will more properly adjust capital to take into account unwind risk, and bad credit selection.
How You Use It Matters
Now we enter into the “complexity” charge arena.
Hedges are once again front and center. The only “perfect” hedge is selling an asset. Not owing an asset is the best form of risk control. There is a cost to not owning the asset – lack of income, possible relationship damage, but that is the only way to ensure there are no losses.
What about if I buy CDS against the position? How much risk relief should the bank get? Not as much as they currently get.
Now the bank has 2 positions a loan/bond and a CDS. The bank has money tied up in the loan. That is not a problem if they sold it but having a hedge means they continue to rely on borrowing.
They need to monitor the loan. Book P&L on the loan. Etc. So even if they were “fully” hedged, they still need to do almost all of the work they would do if they owned the loan outright.
In the case of loans they maintain a direct relationship with the lender. That may influence their decisions, a key consideration in times of stress.
On the CDS, they now need someone to monitor that trade. They actually created new people cost by having this trade. They have also created counterparty exposure where they had none. The big problem here is that if the underlying name is tight and the counterparty is good quality, minimal (possibly no capital) is attributed to the short position. That is wrong in the first place, because by the time that protection has value (the underlying company credit is deteriorating) the counterparty is also likely in worse shape.
Now, let’s say that this bank is known to have a huge position of long assets and short CDS. Will the hedge perform as meant to in an unwind situation? No it won’t, not at all. So again, the regulations have to focus more on unwind, then final outcome because that is what drives banks.
If the asset attracts X capital and the hedge attracts Y capital, I’m not sure that the two positions shouldn’t attract X+Y capital. Maybe there should be some discount for the “offsetting nature”, but certainly not full capital relief. This “hedge” is also a trade. The risk profile looks very different than having sold the loan and the capital should reflect that.
Complex Books and Trades
Complex books aren’t the same as complex trades, but both should require escalating capital.
A bank with 10 derivative trades is simpler to monitor than 1 with 100, or 1 with 1,000. The concept is similar to notional escalation. The potential for error and the time and effort required to unwind goes up. So the base case capital should increase as the number of trades goes up. This will be great for cash bond and loan trading. Banks can own a bond, not own it, or be short it. Derivatives are a complex mess of buys and sells. Banks can continue to behave as they do now, or they can work with counterparties to kill redundant trades, or better yet, move it to an exchange. In any case, it will be their decision to run a complex book versus a simple one with less capital (if they choose complex books then increase the capital requirement acceleration).
This doesn’t prevent banks from trading, but it does reduce their unnecessary complexity or charges them appropriately.
Complex trades are ones that require models or some other means of valuing. Massive capital charges and hard limits are required here. No hedge accounting. Nothing. If something needs to be complex and model driven, they can use it, but the capital has to assume the worst case – that the position is far worse than anyone believes it could be. Tranches that spread through the entire system, some run “no delta” but others plugged into correlation models are a prime example of this. The bank can put on those positions, but they are complex, not simple, and should be treated as such.
More often is better than once
Capital charges should be based on an average of 6 days in a period rather than just some quarter end assessment that is relatively easy to game. The average should be based on each month end, and 1 day selected randomly by regulators, and only announced after close of business the day that was selected. Banks can do this calculation in their sleep. If they are a small bank it is easy. If they are a big complex bank they have the systems. If they can’t do it, the question is why not? How are they running their business without being able to analyze risk at the fairly basic levels these capital rules would require?
I think making the calculation something the banks face every day, rather than once a quarter would go a long way to encouraging banks to have excess collateral. They don’t have the option of just cleaning it up for a nice and predictable quarterly report.
This gets far more complex, but a bank that has nothing but short term financing is far riskier than one that only issued 10 year bonds. I’m not sure how to provide those benefits back to the banks that do the best job funding, but it should be a component of any regulatory overhaul.
Will this fix TBTF? No, but it seems fair. It allocates equity the way a hedge fund would. It is more complex than existing systems but protects against mistakes and unwinds. Small, simple banks will attract less capital as a percentage of assets than a small complex bank, and large complex banks will attract the capital that means if they “fail” they won’t bring down the financial system. This could be applied to FDIC premiums or anything else, but it is time to start analyzing banks correctly from a regulatory standpoint.