Pricing Derivatives at an Investment Bank

[ValueArb]

From Matt Levine's column today.

 

Quote

Derivatives structuring

Disclosure! I used to sell customized derivatives at Goldman Sachs Group Inc. There are roughly four steps in pricing a complex derivative to show to a client:

1. You need a pricing model for that type of derivative. This will be built by quants and will live in the bank’s systems (or on Bloomberg), ready to be applied to particular cases; you pick the model and then fill in the terms of the trade.
2. You need to get market data (prices, volatilities, interest rates, etc.) to input into the model. This data will be ingested and will also live on the bank’s systems (or on Bloomberg), ready to be used by the model. 
3. You might need to adjust the market data, somewhat subjectively, to account for the size and risk and liquidity and terms of your particular trade. If your market data feed says that the implied volatility of a six-month 100-share call option is 35%, but you are selling a five-year option on 10 million shares, you might not want to use 35%.
4. Then the model will give you a price, and you will look at it and ask yourself “how much more can I charge the client for this trade?” If the trade is fairly standard and the client is an aggressive hedge fund with its own model who is bidding out the trade to six banks and will call your boss to scream at her if your price is wide, you will quote pretty much what the model says is fair value. If the trade is unique and the client is an assistant treasurer at a sleepy corporate client who is grateful to you for taking him out to an occasional steak dinner, you will add like 2% edge to the model price. This is the most important part of the job.

 

I'm assuming the model automatically adds a markup for the bank's profit, otherwise selling derivatives at expected value makes no sense. Either way, it seems like a very problematic process. What happens if you have a banker who prices some very risky derivatives without incorporating their extra risk and you only find out later after they blow up into a huge loss for the bank? I presume there is another layer of approval he skipped in order to make the description simpler. 

 

But the other thought that occurred to me is what if the client has some inside information that radically changes the expected value of the derivative? How does the bank ensure their bankers can ferret it out when thats occurring, or when the derivative is so specialized that the risk of informational imbalance is there?

[dwy000]

1 hour ago, ValueArb said:

From Matt Levine's column today.

 

 

I'm assuming the model automatically adds a markup for the bank's profit, otherwise selling derivatives at expected value makes no sense. Either way, it seems like a very problematic process. What happens if you have a banker who prices some very risky derivatives without incorporating their extra risk and you only find out later after they blow up into a huge loss for the bank? I presume there is another layer of approval he skipped in order to make the description simpler. 

 

But the other thought that occurred to me is what if the client has some inside information that radically changes the expected value of the derivative? How does the bank ensure their bankers can ferret it out when thats occurring, or when the derivative is so specialized that the risk of informational imbalance is there?

There's kind of 2 levels to the pricing. The bookrunner will give the salesperson the price for the derivative.  That price includes his/her cost to hedge and often a little juice for the book.  Keep in mind the bookrunner generally has thousands of trades in their book and their pricing will reflect where the book is positioned at time of quote. Often, if their book is heavily weighted in the other direction they will price the trade low because its a natural offsetting position for them anyways.  The salesperson will then tell them to add on the desired P&L and the bookrunner will give them a final price for the client.  Rarely is that P&L under $10k, especially for a corporate.  The more complex or custom a product and the more unsophisticated the corporate, the higher that goes.  

 

The risk to the bank is with the bookrunner.  It's their job to hedge out the excess risk.  Note, excess risk, because the bookrunner is running a large, active book with risk as their job on a day to day basis.  The quants, the bookrunner's boss(es), the risk manager all maintain a high degree of oversight over the book as a whole and it is managed for a whole host of risks (most commonly referenced as daily VaR). Risks are managed real time based upon 100 underlying variables and reported out constantly.  That way you know of 2 bookunners are each within their risk limits but combined might exceed the banks' desired limit.  I used to hear salespeople batching at the bookrunners's price and his response is like "I'm way long vega in the 4 year and need to add gamma in the 5 so that's the price.  Take it out of your P&L if the client is complaining."

 

If the client has inside knowledge on something and the bookrunner doesn't there really isn't much impact because he/she will hedge it out with other counterparts who have the same info he/she does.  

 

There's a lot more to it but it's definitely not as susceptible to accidental risk as people think.  Absolutely nobody at the bank wants to risk a blow up because that ends their bonus, their job and likely their career.  

Edited October 26, 2023 by dwy000

[ValueArb]

Thanks, great perspective.