Opportunity: The Federal Reserve’s Comprehensive Capital Analysis and Review (CCAR) program, has opened up a $100+ billion annual market in optimizing the Net Interest Margin (NIM) at significant US banks. An extra basis point of NIM is worth $200mm/yr to a bank the size of JPM. The computer floating point execution needed to run NIM optimization is inexpensive and the cost is expected to continue to drop for another 5-7 years, even considering Colwell’s 2013 comments regarding the End of Moore’s Law. These financial and technology developments premier on a stage where US Bank Net Interest Margin is at a 30 year low and in an interest rate environment where rates are expected to remain low (see Rieder).
Problems: Problem 1– Net Interest Margin has only been a quantitative focus at large US Bank Treasury departments for the last several years for regulatory reporting, so the level of quantitive security modeling is elementary and non-standard. The prepayment and cashflow models for the standard accrual portfolio securities including mortgages, loans, and credit cards are well established in two-way secondary trading securitized markets. This implies that the quantitative modeling error is controlled enough to support tight spreads and liquidity in the secondary market. But these security models have not completely moved from the small Front Office Trading portfolios (the Trading book) to the large Treasury accrual portfolios (the Banking book). Problem 2 – The stochastic modeling for the econometric and market variables needed for complete Banking book Monte Carlo simulation is new field without a mature publication history. There is however a mature publication history for Monte Carlo simulation of Trading book CMBS and RMBS valuation (where the prepayment models depend on stochastically perturbed econometric variables) so there is some reason to expect this problem is not hard to solve. Finally Problem 3 – Bank technology tends to lag and remains very conservative with respect to upgrades. Simulation of 1+bn-security aggregate USD accrual portfolios with static CCAR scenarios strains the Banks’ technology organization for a variety of long-standing reasons.
Solution: The Princeton Bank Consortium presents aggregate Net Interest Margin projections and optimizations based on data from the US Federal Reserve. This website presents the data from a novel large-portfolio (e.g., aggregate US Bank Assets and Liabilities) numerical optimization implementation of the Net Interest Margin Optimization (NIMo) algorithm. The PBC will develop a standard set of security models to be used across individual bank’s accrual portfolios as reported to the Fed. The PBC will use the US Federal Reserve’s forecasted market and econometric variables to drive the Monte Carlo simulation. Finally, the PBC will side step the proprietary, legal, and slow technology limitations of working with a single bank and instead simulate and optimize all the CCAR reported assets and liabilities using optimized versions of the open source code.
Broadly, the idea is to automate the US Banking System’s implicit aggregate discretionary capital allocation plan to optimize the new business revenue (e.g., funding loans with deposits) relative to dynamic operational and market constraints. We can then run explanatories to fit our standard model parameters on individual US Banks to complete NIM P&L attribution. We expect the results will be useful for both Banking System Monitoring and Capital Planning.