Balance Sheet Optimization: Can you do it?
A lot has been tried and written on the subject but rare who have been able to achieve meaningful results.
Thank you for Paola Brahimcha for the help
This paper approaches this topic from a different angle, the idea is to understand how components of a balance sheet react under a stressed environment and how the global Profit & Loss is impacted.
This is a holistic approach where Market Risk, Credit Risk, Liquidity Risk, Processing Risk, Margin Requirements and Regulatory Charges will be covered. These are all factors impacting the P&L i.e. the Balance Sheet.
The P&L is the heart of a Balance Sheet Optimization.
It is important to note that I will not try to solve the whole problem of the Balance Sheet Optimization, but more providing good building blocks using the existing technology found in the Capital Market business to achieve a very serious framework to optimize the Profit & Loss and understand the impact on the balance sheet.
Let us first understand what is Balance Sheet Optimization in simple terms (for the common mortals), a simple and complete description found on the web:
Balance Sheet Optimization helps financial institutions determine what to do with their assets to better meet regulatory requirements and achieve higher profitability.
Typical users include investment banks with large derivatives portfolios.
Basel III, EMIR, Dodd-Frank… regulations have imposed higher capital ratios and reduced the risks that banks can take. But reducing the number of assets and the risk that banks take reduces the return on equity, putting pressure on profits and the ability to deliver returns to shareholders. To meet these capital requirements, banks must determine which trades to clear, crush, offset, or exit.
Balance Sheet Optimization determines what to do with existing trades to meet regulatory requirements and maintain profitability.
I identified 5 pillars that can play into the Balance Sheet Optimization solution. There are shown below:
To understand how these pillars interact together, let us draw a high-level bank’s capital market functions heat map:
The chevrons highlighted in red are the ones that will largely affect the P&L (yellow chevron).
The process starts by:
Identifying the tradable (tangible) asset classes impacting the balance sheet, for example:
Some of the products are not typical traded financial instruments per say, hence the need to find a proxy financial product that is close enough to represent this type of investments (A synthetic 'pool' representation of activities). As an example, a bond can be used as a proxy to represent a real-estate investment …
Once the asset classes are identified and mapped to their proxies we need a good pricing and simulation engines to re-price these products.
We then classify the products (collateralized or not, liquid [high, medium, low], strategic / non-strategic investments…) and identify their corresponding pricing model (or use a more generic pricing model like an American Monte-Carlo) with the required ROI per activity (if possible). No need to go nuclear on pricing models, just keep it simple and fast, at the end it is just an approximation.
The input for these engines will be:
Product description (enough to price the product)
Shocked market data per a pre-defined stress scenario (ex. Brexit) using the “What-if” Market-Data simulation engines:
Rate up & down
Volatility Surface shape changing
Correlation surface shape changing
FX rate up & down
CDS up & down
Security prices up & down (per sector/geography)
Spreads against treasury bands up & down
Commodity prices up & down
Hedge baskets definition:
Appropriate hedging products
Behavioral assumptions scenarios using the “what-if” Behavioral simulation engine:
Early termination assumption on loans and deposits
Early call assumptions on bonds and loans
Rollover assumptions for deposits and issuances
Payment delay on loans
Rating downgrade on issuer and counterparty
Execute on new hedge recommendations by selling/buying at market price
Balance sheet evolution scenarios:
Extending my equity cash activity
Investing more is type bond A (buy) vs B (sell off)
Deposits grow by x% per year per currency
Reduce short term funding (within the leverage ratio limits)
Increase long term funding (within the leverage ratio limits)
Close non-strategic investments (sell off)
If option deep in the money, we are at an exercise date and American then exercise
If option out of the money, we are at an exercise date and American terminate
Increase in short term deposits will be done through pre-defined products per currency
Increase in long term deposits will be done though pre-defined products per currency
Any trade priced below a pre-set ROI (unprofitable) is flagged and the difference is set to be charged internally as an FTP
Scenarios must be prioritized i.e. which one will run 1st, 2nd, …
The output of these engines will be:
New NPV / Price
New risk sensitivities / Greeks
New hedge recommendations
New trades input recommendations
New trade closing recommendations
New trade life cycle recommendations
Next in the chain reaction:
A fictitious trade booking engine will execute the required transaction generated by the simulation engines. This will also trigger the back-office processing cycle [Confirmations, Settlements, Payments, Matching, Nettings,…]
A fictitious trade life cycle engine will execute on the new trade recommendations, this will also trigger the back-office processing cycle
A market risk engine will re-valuate the required indicators: SA-MR, VaR, IRRBB, Sensitivities
A credit risk engine will re-evaluate the required indicators: SA-CR, SA-CVA
A limit engine will re-calculate the new limits contributions for different rules (counterparts, compliances, market risk, credit risk, liquidity risk)
A margin calculation engine will recalculate the required margins based on the new pricing outputs:
Estimate the new Initial Margins (listed / bilateral)
Estimate any Variation Margins
For proxy products create fictitious VM by comparing prices before ‘chocks’ and after
A collateral engine will simulate the required collaterals to be called or received, this will also trigger the back-office processing cycle
A collateral optimization allocation engine will try to deliver the cheapest collateral and restructure the security and cash inventories accordingly, this will also trigger the back-office processing cycle
A liquidity risk engine will recalculate the impact of this new priced environment and positions on LCR/NSFR ratios and compare to agreed limits
A back-office engine will monitor the efficiency of execution under stressed environment by providing adequate KPIs and calculate the cost of transactions’ processing
A regulatory reporting engine will re-evaluate the new Capital requirements charges
A P&L engine will recalculate the new P&L based on this new environment (by including the cost of capital and processing cost).
A global viewer (with a slice & dice features) that will show before and after simulations the following:
Interest Rate risks (Sensitivities, VaR, SA-MR, IRRBB)
Credit Risk (SA-CR, SA-CVA)
Liquidity Risk (LCR, NSFR, including intra-day risk)
Limits (Market risk, Counterpart, Credit risk, Liquidity risks and compliance)
Positions & inventory (cash and securities)
Hedge baskets changes
Capital Charges requirements
Cost of Processing
List of operations / positions generating returns below required ROI
This solution will give a global view to managers and the opportunity to adjust their portfolio structures for a better outcome in case such scenario happens in real-life.
Sometime an outcome of the scenario analysis could be that certain desks are better shut down. But it might be better to assess the cost of rebuilding a desk from scratch rather than keeping in a controlled framework. Regulation can make certain activities (like equity repo) not so wanted, but what about the profitability and the institution balance in case this is shut down?
Common good sense in both scenario building and assessing the equilibrium between existing desks, along with the fact of not neglecting any risk, is key.
Machine Learning (the cherry on the pie):
We can push further and feed the outputs of this solution into Machine Learning [Deep Learning] engines that will allow the bank to:
Have a feedback loop to tweak the rules in order to determine what best setting suits such a simulated scenario
Identify the most susceptible clients ( then the banks should try reaching out to help these clients better define their risk strategy)
Rebalance portfolios to achieve better P&L i.e. better Balance Sheet
Nevertheless, having such chain reactions in a simulated environment has a great value in realigning businesses and strategies, helping clients and optimizing processes. Hence reducing cost , increasing efficiency and better services which will result in eventually a better healthier balance sheet.
All what was described above can be achieved using existing modern trading solutions.
FintekMinds has the knowledge and experience to help you build such a framework.
If you like the post, please share it on LinkedIn, Facebook or Twitter. It helps the spread of knowledge