JP Morgan: Estimating Q2 Numbers and the Whale Trades Impact, Part 1
JP Morgan (NYSE:JPM) is due to report its Q2 numbers this Friday, July 13th. There is a ton of uncertainty regarding how big the so-called Whale Trade losses will be, and how good (or bad) the quarterly numbers will look. The range of estimated losses reported in the press varies from $2BB to $9BB on a worst-case basis. As of this writing, JP Morgan is trading around $33.90, down over 25% since early April, quite a drop.
Given that I spent several years trading in the credit default markets, I felt that (different from most equity analysts), I had a bit of an edge in understanding what the real trades looked like, and what the ultimate losses might be. Truth is, there isn't really a ton of information out there, but if you boil it down, I think it's possible to at least understand the basics of the trade, and hence the likely losses.
In summary, I think the trading losses will be far lower than $9BB, and likely closer to $5BB as of June 30th. With the stock losing $46BB of market value from the April peak, I view the stock as ridiculously oversold.
Here is what we know from various sources and news articles. I know this is somewhat old news, but worth putting it all together to build a model:
- JP Morgan's CIO division bought High Yield Credit Default Swap (CDS) protection, which is called a Long CDS position, and essentially a short credit trade. That is, if the HY index fell in value, then JP Morgan would make money. This is a sensible hedge for a bank that has a $724BB loan portfolio. If some of these loans default, then likely the HY index will trade down, and the bank can make up the loan losses from the HY Index short.
- There also seems to be reports of a tranche trade in the investment grade index, done last year or early this year. Again, it makes sense that JP Morgan had a short trade on in the senior most tranche of an IG index. It's arguably a good hedge for a company long tons of loans. A tranche trade works like this. You have exposure to say the top 70% of the index. So, first losses apply to the bottom tranche holders, and you don't take losses until the bottom 30% of the index has defaulted. The problem was, after the ECB implemented its LTRO lending program, systematic risk fell (i.e. the odds of correlated defaults fell). When systematic risk falls, then senior tranches tighten quickly. It's not terribly likely that 30% of the index defaults anymore. So, if JPM was short the top tranche, the ECB kind of screwed up this trade for it, taking correlation down, and causing some pain here.
- The markets improved in early 2012, so the bank began to lose money in its short HY Index trade, and also in its tranche trade (as correlation broke down). So to offset these losses, the bank then sold protection on an investment grade index. That is, it got long investment grade credits to offset the bad shorts. It did this via selling protection on the 10-year CDX.NA.IG9, which matures in December 2017. Typically this is called the IG9 Index. However, the HY market is far more volatile than the investment grade index. It's full of far dicier credits, so naturally it had to sell much more protection (and hence got much longer) in the IG9 than the notional of the HY trade. For example, if JPM was short 20BB of HY credit, then it would need to get long say $50-60BB of notional investment grade bond credit to neutralize the trade.
- When the markets then went against it during the sell-off in March and April, it seems the HY index trade wasn't working. It also appears that the HY CDS trade the bank put on was of shorter duration too. At the same time the IG9 (investment grade CDS) trade was hurting it. So, to offset this, JPM then tried to get short credit in another IG CDS index. This one was likely the IG17 CDS index, which matures in 2016.
- Notional size of each of these 4 trades is a complete mystery. I have seen exposure numbers of $66BB, $84BB, $100BB, and then a range of $150-250BB. What makes sense is that likely the gross exposure numbers are in the $200BB range, and the net exposures are the lower ones, say in the $100BB range. Why? Because based on what has been reported, JPM had a short HY bond index trade on, a long investment grade bond index trade on, and also a short-dated investment grade bond trade on. It also was short a senior investment grade last loss piece. (Now, I am simplifying a bit. It did this all synthetically. Put properly, JPM sold IG9 protection, bought protection on HY CDS, was long an IG senior tranche CDS, and probably short protection on the IG17 CDS index too.) Confusing I know but keep reading.
How I Estimate the Trading Position and Losses
The first $2.0BB of mark to market losses occurred from the end of the March quarter, to May 10th. That is 40 days. Some have reported that the DV01 (dollar value of a basis point) is $200mm. That is, the trade widened on JP Morgan by 21 basis points, and hence the company lost $2.0BB. That is $200mm X 21 bps.
I don't think it's nearly that simple. Below is a hypothetical but realistically possible view of the trades the company did. I conservatively estimated net exposure of around $110BB, and gross exposure of $250BB. We do know that the biggest component of this trade was in the IG9 index too. You can see trading volume spikes in this index.
Bought protection Sr Tranche
Sold Protection (Long Credit)
CDX.NA.IG9 Index (Dec 2017)
Sold Protection (Long Credit)
CDX.NA.IG17 Index (Dec 2016)
Bought protection (Short Credit)
High Yield Component likely (Short Duration?)
The Gross Exposure is the sum of the absolute value of the longs and the shorts. Note again that the amounts here will turn out definitely wrong. Nobody outside the bank knows the notional amounts of any of these. But these amounts fit with press reports of exposure (net and gross), and also fit the trading levels below.
That is, if you know what positions the firm had, the trading levels of those positions, and the losses incurred, you can infer the notional amounts, or at least ballpark them for modeling purposes.
This also fits with reports that JP Morgan had a flattener trade on. Basically, the bank was short protection (long credit) on the longer maturity indices (2016 and 2017), and long protection (short credit) on the near-term maturities. That is essentially short the index out to 2017, and long the index for a year or two. If the curve flattens, then the higher duration part of the portfolio trades up much more than the short dated bond indices. This was the bet JP Morgan had on.
However, the curve steepened, which is pretty typical in bad markets. That is, unless liquidity dries up and its looks like a pile of defaults is imminent. But with the ECB providing long-term loans, this wasn't the case. (Put differently, correlation fell).
The takeaway: While JPM was hypothetically net long by around $110BB, the longs and shorts add up to $250BB of gross exposure. So, while the DV01 may seem like its $200mm, that is only if the longs and the shorts go against the bank simultaneously. This is an important point. This trade seemingly went wrong for JPM in every possible way leading up the reported losses.
In a normal market, my math suggests that the DV01 was probably more like $70mm. It's no different if you as an investor buy JP Morgan stock, and short Goldman Sachs stock. Normally they move together, and you would think your exposure is the long less the short. But if the long goes down, and the short goes up, it's a painful squeeze and the gross exposure becomes problematic.
Testing this seems to validate the trades. We know losses were $2BB on May 10th from the end of March:
The losses come out to around $2BB in the far right column, consistent with the first reports on the trade.
On May 17th, press reports pegged the losses at around $3BB plus. Based on where IG9 and IG17 were trading on these dates, again we can estimate the losses. We also know that IG9 was around 128 on May 10th, and around 145ish on May 17th. The squeeze was on.
Note: typical five-year CDS contracts in investment grade land work so that a 20 bps move in the credit default swap implies a 1% move in the value of the swap on a mark to market basis. That is, if 5-year IG9 widens by 20 bps (from say 100 to 120), then on a $10mm position, the mark to market change in value would be 1% of 10mm or $100,000. You could also say that that is roughly a DV01 (dollar value of a basis point) of $5,000. It's probably closer to $4500 from my recollection on a 5-year IG swap, but this is more conservative. In fact I used 17bps because the IG9 trade is out 5.5 years, which is a conservative $5800 DV01.
Exiting the Trade
As of last week, JP Morgan had supposedly taken off 65%-70% of this trade. This is a good chart of price and volume of the IG9 follows.
The heavy volume in late June (the green bar) here must be JPM taking off this trade, and clearly at a terrible level. My guess is they top ticked it between 155 to 170 bps. Volume of $30BB here along with other likely trades prior to this is consistent with taking of 65%-70% of $140BB too. In fact I might be overestimating the size of this one component of the trade.
This new level impliy roughly $5BB of losses below:
On the whole, I accept any criticism of my estimated notional amounts for the Whale trades. Pick whatever numbers you like and use the CDS DV01 numbers to do the math yourself. Regardless though, I think this is a reasonable framework to estimate the losses from these Whale Trades.
This $5BB is importantly far lower than the $9BB worst case number that is being tossed around in the press. That is the good news. And if roughly 2/3s of the trades have been unwound, the exposure going forward is likely minimal given the marks and the size.
Next I will take these numbers to estimate what Q2 results may look like. Click here for Part 2.