In banks and financial institutions, there is a typical barrier that exists between Finance and Risk. The Finance world is very “exact”. A balance sheet or a Profit and Loss (P&L) statement have to be very precise.
In the Risk Management world, things are slightly different. It is common that numbers are created as a result from statistical analysis or stochastic simulations; recipes vary in order to create those numbers. It is indeed fascinating.
Balance sheet forecasting is a technique being used by Finance professionals since a long time ago; although most of the time, the forecasting models being performed are fairly simple. Perhaps by applying a more analytic approach to the balance sheet and other Finance artefacts, such P&L or income statements could provide us with some leverage.
Assets, liabilities and equity can be broken down into small categories and end up with dozens of items (or factors for the purpose of our discussion). Now, let’s imagine we have access to the general ledger system of a bank (or any business), and that we could extract a time series of those factors in the balance sheet.
We could simulate future balances for each one of those factors and add them up as a portfolio, which will be seen as projected upcoming balance sheets.
If we do this process for thousands of times and multiple time steps, we could create a distribution and apply a confidence level (e.g. 99%) so a worst case scenario could be forecasted on the balance sheet.
Since we have the balance sheet time series we got from the GL system, we could also calculate volatilities of each one of the factors and compute correlations amongst them.
A similar approach could be applied to income statement or P&L data, as well.
By treating the company’s finances with risk analytics techniques, potential benefits could include:
In a bank, the data from the GL most likely would reflect balances on different types of retail products (e.g. home loans, credit cards, deposits, personal loans, etc.). The Monte Carlo Simulation discussed above could be applied to each one of those product types and forecast a potential state by product type.
This becomes very powerful for a decision-making process and to learn possible growth on the different product types. Also, exercises on mix optimisation could be performed if this feature was available.
On the not-so-uplifting side, a potential challenge could be seen in preserving inter-relationships across the balance sheet factors. In this regard, we could create functions as part of the model, in such way that a segregation of dependent and independent factors becomes available. Also, we need to be mindful that reconciliation of forecasts might be required at different levels of the GL hierarchy.
As it can be seen, there are plenty of opportunities that could be targeted by using this approach. In the same fashion that we have explored benefits from a Finance perspective, a Marketing lens could also be applied, opening up other opportunities.