Tuesday, 21 April 2009

CVAs or "The Magic of your Own Credit on Profits"

There has been much discussion in recent days regarding the surprisingly positive results coming out of US banks. There has been an equal amount of disection of those same results, and in particular those of Citi.

In Citi's Q1 earnings report, it mentions the following commentary for its fixed income division:

"A net $2.5 billion positive CVA on derivative positions, excluding monolines, mainly due to the widening of Citi's CDS spreads"

A number of commentators have rightly focused on this, with most describing it as a one-off gain as a result of Citi essentially making bets, using CDS, on its own creditworthiness. However, I believe that the reality is somewhat different, and in fact more troubling. The analysis below is something I have wanted to blog about for a while, and now seems the perfect opportunity. A warning - to really demonstrate the point, I've used a numerical example - please stick with it, as I feel it really hits the point home.

Good old-fashioned cash products

Once upon a time we lived in a financial world that was largely cash based e.g. cash equities, cash bonds. These would be traded on active markets, with high price transparency. A liquid cash bond, say a 10yr Citi (C) issue, would be widely traded, and participants would be able to get executable prices throughout the day. What would the bond price depend on? Among other things, market interest rates, liquidity in the market, and the perceived creditworthiness of Citi.

Let's say Goldman Sachs (GS) buys $1m worth of the Citi bond at par. A few days later, some bad news spreads round the market about Citi, and the price falls to 70. How do both companies account for it?

GS - The bond was bought by the trading desk seeking to make a short term profit. As such, GS marks the bond to market, in this case taking a $300k loss. For those accountants among you, the entry would be:

DR Principal Transactions $300k

CR Financial Assets $300k

C - Most firms when they issue bonds use the amortised cost method to account for the bonds i.e. they accrue interest over the life of the bond less any cash payments made to holder. When it first issued the bond, Citi recognised a liability for $1m, and the liability stays that way despite the fall in fair value of the bond. This makes perfect sense, because Citi still expects to be a going concern and still expects to repay the £1m at maturity.

Fancy new derivatives

Fast forward to 2009. What is a derivative? There are all sorts of 'official' definitions floating about, but for now we will say it is a legal contract between two parties to exchange certain cashflows at certain future dates. Keep this in mind whilst reading the next couple of sections.

Counterparty CVA

Now imagine instead that GS enters into an OTC derivative trade with C, sometime in the past. At year end, GS's theoretical model says that the swap has a FV of $1m in favour of GS i.e. it is an asset for GS, a liability for C.

What does GS have on its balance sheet? It has a $1m financial asset, with Citi as the counterparty. GS marks the position to market i.e. at fair value. What is the definition of an asset's fair value? Accounting standards differ slightly in their wording, but in principal it is the price you could sell that asset for in an arm's length transaction. So GS should mark to $1m then? No. Remember, GS doesn't have a derivative financial asset, it has a derivative financial asset with C as the counterparty. I.e. fair value in relation to OTCs effectively says 'At what price would a 3rd party be willing to take your place in a trade?' A buyer would pay more to 'swap places' with GS if the counterparty was JP Morgan, than if it was Citi, because Citi is a greater credit risk.

So how does GS establish this fair value, taking into account the credit worthiness of its counterparty? This is an OTC trade, which by definition is not exchange traded. In the cash bond example above, GS could find the allowance for the increased credit risk of C because it was incorporated into the market price of the bond. For the OTC, GS turns to indicative measures - in this case CDS spreads on C. Let's say CDS on Citi debt is trading at a spread of 500bps at year end. Goldman incorporates this spread into the risk-free rate used in its pricing model, and it spits out a FV of $700k i.e. $300k less.

GS therefore takes a counterparty Credit Valuation Adjustment (CVA) of $300k. This seems fairly prudent- in effect, GS is taking a $300k credit reserve against potential default by a counterparty. For accounting geeks like me, the entries would be:

DR Principal Transactions $300k

CR Financial Assets $300k

Own Credit CVA

Now what about Citi? It has the same pricing model as GS, which calculates a FV of $1m. So surely C recognises a financial liability of $1m? No. Again, accounting standards differ slightly in wording, but in principal the fair value of a liability is the price at which an entity could extinguish any future obligations, in an arm's length transaction. I.e. "What price would a counterparty be willing to cancel/settle the trade at now?" Well, we have seen above that GS, and probably the rest of the market, would accept $700k to cancel the deal. So, Citi gets to write the liability down to $700k. In accounting terms:

DR Financial Liabilities $300k

CR Principal Transactions $300k

Yes, you have read it correctly - Citi has made a profit of $300k as a result of becoming less creditworthy!!! In accounting/industry spreak, Citi has made an Own Credit CVA of $300k.

Not a One-Time Item, but a Two-Timer

Most blogger and analyst commentary that I have read on this Citi gain and others (see below) describes an Own Credit CVA as a 'one-off' or 'one-time' item.

Unfortunately, they could not be more wrong! In the above trade, what would happen if at the end of the next quarter, Citi's CDS spread fell to 20bps i.e. it becomes more creditworthy? Lets assume the model FV is still $1m. GS plugs the new Citi CDS into the risk-free rate used by the model, at it spits out a FV of fairly close to $1m again. GS can therefore release the CVA of $300k that it had reserved for earlier. The double entry for GS would be:

DR Financial Asset $300k

CR Principal Transactions $300k

I.e. GS makes a profit of $300k in Q1.

Now Citi. The fair value that counterparties are willing to cancel the trade at appears to be back to $1m. Therefore, C writes the liability up to $1m again. The double entry is:

DR Principal Transactions $300k

CR Financial Liabilities $300k

I.e. Citi makes a loss of $300k in Q1.

This is what makes the Citi CVA a two-time item - if Citi continues as a going concern and its credit spread falls, it will at some point have to reverse the $2.5bn profit it gained in Q1 2009 i.e. it will have to recognise a $2.5bn loss.

But wait there's more...

I've shown above that there are two types of CVA - a counterparty CVA that a bank applies to its assets, and an Own Credit CVA that a bank can apply on its OTC liabilities. The $2.5bn reported by Citi is a NET amount. I.e. at the very least, $2.5bn will have to be reversed at some stage. Let's say Citi made a Counterparty CVA of $1bn in Q1 (not unreasonable on a balance sheet with $2.2 trillion in assets). That would mean that the Own Credit CVA was actually $3.5bn i.e. an extra billion to reverse in future.

So did Citi purchase CDS on itself?

I'm pretty sure they haven't - Citi's press release mentioned CDS spreads in the context of CVAs, not proprietory CDS positions on its own debt. Plus how would Citi ever gain? OK, it buys protection on itself initially. Its credit spread widens, and the CDS is worth more - but how would they ever unwind this in the market? Usually you would do a back to back CDS trade selling protection at the higher spread, and pocket the difference. But what would motivate a counterparty to buy protection on Citi from Citi?

I've worked on CVA calculations for financial institutions, and currently the best proxy for credit worthiness is a CDS spread, given the illiquidity in cash bond markets, which makes me firmly believe Citi's mention of CDS relates to CVAs.

Stop picking on Citi

I think I've bashed Citi enough in the above sections. I'll give them a rest, and instead present to you a few Own Credit CVA snippets from around the banking world:

RBS - Pg 207 of the 2008 annual report shows own credit gains of £2.823bn ($4.15bn)

HSBC - Pg 31 of the 2008 annual report shows $6.57bn of own credit gains, on $6.5bn of profit for the year.

Barclays - Pg 4 of the 2008 annual report shows own credit gains of £1.663bn ($2.44bn)

ML/BoA - Hard to be 100% on this one, but looks like a $2.2bn CVA gain on structured notes in Q1 (from the Q1 press release)

Together with Citi, that means approximately $17.9bn will have to be charged back to bank P&Ls. Or put another way - in the last few months, the above banks have reported $17.9bn of profit they probably will never actually realise (this is on top of the general MTM issues on illiquid assets).

In Conclusion

Hopefully the above analysis has helped. Own Credit CVAs are an unintended consequence of fair value accounting. Any firm that is a going concern will never realise the profits shown, and will in fact have to reverse them in future periods, thus showing apparent losses. I'm a big supporter of fair value, as I believe it helped expose the problems in the banking system much earlier than if we were using historic cost. However, nothing is perfect - and it is these imperfections in fair value, and their possible cures, that I will be looking at in future posts. If you have any questions/queries/comments, please leave them in the comments section, or email me at holdingtoaccount@googlemail.com.