|
Vose Consulting and Denali Alaskan Federal Credit Union formed a strategic partnership in 2008 for the purpose of exploring areas within the credit union industry where Quantitative Risk Analysis (QRA) can improve decision making processes. QRA has been gaining increasing acceptance in financial institutions of all sizes as a fundamental decision support tool because it offers decision makers more and better information under conditions of uncertainty.
Traditional 'deterministic' approaches typically involve single point estimates of future events despite the fact we are often dealing with all manner of uncertainty and variability. Consider Discounted Cash Flow methods. Traditional Net Present Value (NPV) and Internal Rate of Return (IRR) calculations require us to make assumptions about the timing and amount of future cash flows. The assumptions are single point estimates and so are the results of the calculations are also single point estimates, e.g. $100,000 and 7%. What a decision maker also needs to know is how likely is it that a given project’s NPV and IRR will be above particular thresholds. By capturing the uncertainty and variability in the timing and amount of cash flows, QRA provides results in probabilistic terms. Example: the likelihood that the project will produce a profit at or above $80,000 is 75% and there is a 40% probability that the IRR will be above 6.8%.
While knowing the profitability profile of a project provides an advanced view when compared to deterministic methods, a higher level of understanding can be obtained when the profit profile for two projects can be compared. For example, imagine an organization was trying to decide whether to invest a particular amount of capital in one of two possible projects – Project A and Project B. Each project has a lot of uncertainty regarding the actual amount of profit to expect however they both have the same average expected return. A traditional deterministic analysis would most likely look at the two projects as being essentially or effectively equal in terms of their risk vs. return.
However, a QRA approach to making the decision should delve more deeply into the underlying processes that a fundamentally driving the profit for each project. In that case, the two projects may appear to be quite different from a risk vs. return point of view. Consider the probabilistic profit profiles for each project shown below.
 
These charts can be understood as representing the probability (vertical axis) of any particular profit level (horizontal axis). The chart below shows the profit profiles superimposed on the same set of axis.
Looking at these two alternatives from a QRA perspective provides a much different, and richer, set of information. We can see from the chart that the two projects have essentially the same average value. We can also see that the range of possibilities for Project A is much wider than for Project B. Project A's profit profile has two "peaks" which implies that the possible results if we choose Project A will most often be closer to one of these peaks than to the average value as would be expected for project B.
While these charts clearly demonstrate that the projects have much different profit profiles, we don’t have enough information to know which Plan is the best choice. In order to make the best choice we would need to understand the priorities and risk tolerance of the credit union's management and board. A credit union with a larger appetite for risk may opt for Plan A as it has a greater likelihood of a greater profit. In the case of a credit union that would prefer a more certain (less risky) range of outcomes, Project B may be preferable.
In the current economic environment, understanding and managing risk and opportunity are, more than ever, essential to a credit union's future survival and prosperity. QRA provides decision makers with a better understanding of the risks and uncertainties affecting your organization through their explicit identification and quantification.
QRA offers many analytical tools, including assessment of uncertainty of the assumptions on which a model or forecast is based, time-series forecasting, eliciting and modeling expert opinion, and determination of key metrics that measure the uncertainties.
Vose Consulting and Denali Alaskan Federal Credit Union have experience in, and can help you with, among others, Value at Risk (VaR), Credit Risk, Interest Rate Risk, Market Portfolio Risk, Capital Management and Operational Risk. Vose Consulting’s sister company, Vose Software, has also commercially released a software tool, ModelRisk™, that puts a new, state of the art of risk analysis modeling onto the desktop of credit union executives.
Examples of analytics applied to credit union problems:
Interest Rate Modeling:
Arguably, two of the most important exogenous factors affecting credit unions, future market interest rates as well as the paths of future interest rates are particularly suited to modeling by risk analysis techniques. While it is truly impossible to accurately produce single point forecasts of future interest rates, there are a number of techniques available for generating probabilistic ranges of future interest rates. In other words it is possible to model future interest rates where the outputs are ranges of possible future rates as well as the probability of the actual rate falling within each range.
We have produced, and been using, an interest rate model based on an accepted technique called one-factor modeling. Our model is based on the historical rates, means, trends and volatility and is used to generate possible future interest rate paths and scenarios. Below is an example yield curve chart of the modeled range of possible interest rates for 12 months in the future.
The yield curve chart can be read as follows:
- The dashed line shows the current yield curve for Federal Reserve Board rates ranging from Fed Funds through 10Yr Constant Maturity.
- For each rate there is a range of possible outcomes 12 months into the future.
- Based on the model’s outputs:
- 100% of the possible rates 12 months in the future fall between the blue lines.
- 50% of the possible outcomes fall between the red lines.
- 50% of the outcomes fall above (or below) the green line.
- This chart could be interpreted as showing a relatively high possibility of the yield curve becoming flatter over the next 12 months.
It is also informative to look at the possible paths a particular rate might take over the next 12 months. The chart below shows the range of possible rates for the 10Yr rate.
This chart can be read as follows:
- The dashed line shows the prior 12 months of historical rates for the 10Yr Constant Maturity.
- For each of the next 12 months:
- 100% of the outcomes fall between the blue lines.
- 50% of the outcomes fall between the red lines.
- 50% of the outcomes fall above (and below) the green line.
Interest Rate Risk:
After we had built the probabilistic model of future interest rates we have applied it in a number of analyses. One of the principle areas is Interest Rate Risk (IRR) to a credit union. The traditional method for assessing IRR is to shock the balance sheet and income statement by applying instantaneous positive and negative interest rates shocks of 100, 200 & 300 basis points. This traditional approach has a number of built-in assumptions:
-
Instantaneous rate changes.
-
All rates move an identical number of basis points.
-
There is no information available regarding to probabilities of where rates may go in the future.
All of these assumptions are worse than unrealistic, they are, in fact, extremely unlikely to ever occur. Thus, the results of this analytical approach, likewise, are known to be wildly inaccurate before the analysis is performed. Interest rates do not make instantaneous moves of these magnitudes. While rates of various maturities do tend to move in a correlated fashion they do not move in lockstep. From any current level of interest rates there is a range of possible future outcomes, but there is no consideration given to the likelihood of any particular outcome.
Once we developed a probabilistic interest rate model, we then had the capability to build a model that can measure probabilistic IRR.
Our model uses the following parameters:
- Current balance sheet and income statements.
- Cash flows from current loan and deposit portfolios.
- Management opinion about:
- Future balance sheet account levels.
- Interest income and expense.
- Fee Income
- Operating expenses.
- One-off expense and income adjustments.
- Allowance for loan loss.
- Charge offs.
- Non-maturity deposit decay rates.
- Probabilistic interest rate model.
These elements combine to form the basis of a very flexible model for producing a number of probabilistic metrics that are of interest to credit union management, the Asset-Liability Committee and Board of Directors. A few examples are:
-
Net income.
-
Return on assets.
-
Net interest income.
-
Net capital.
-
Loan to Share Ratios.
-
Operational efficiency ratios.
-
Key financial ratios.
A couple of example output charts are shown below.
This is a probabilistic trend chart showing the cumulative net income over the next 12 months where probability of the outcomes can be read in a similar fashion to the charts above.
Another metric of particular interest is the probability of net capital 12 months in the future falling within designated levels. The chart below shows another example of the type of outputs available through QRA:
In this case the model is showing that the example credit union would have an 86% probability of being adequately capitalized in 12 months, but a 14% probability of being undercapitalized. If these were the results of a real analysis, credit union management could easily test any number (or combination) of alternative courses of action to improve the balance sheet and income statement forecasts to reduce the probability of being undercapitalized.
Other areas where quantitative analysis has been applied to credit union decision making:
-
New Branch:
-
There are a number of uncertain factors affecting:
-
Member growth.
- Market growth.
- Expenses.
- Product mix.
- Cannibalization of existing branches.
- Given a particular set of up front expenses and costs, a probabilistic model can help understand when a new branch might expect to break even and what a worst case scenario might be regarding how long it might be cash flow negative.
- Plastic Card Bond Coverage:
- Credit unions are often presented with various deductible options for plastic card coverage. Using historical data on claims frequency and severity, QRA can be used to provide decision support when selecting the deductible/premium option that is most likely to result in the lowest total cost of risk to the credit union.
- Litigation Support:
- There are often a number of uncertain variables when making decisions about whether to settle a legal claim vs. going to trial:
-
Strength of the case and probability of winning.
- Expectations regarding the level of awards if there is a trial.
- Attorney contingency fees and other costs may be variable based on whether there is a trial.
- A probabilistic model can be very valuable when negotiating a settlement and when deciding whether (and when) to take a particular offer.
If you are working at a credit union or community bank, have found the discussion above interesting and would like more information, please feel free to contact us anytime or download a free whitepaper here.
For additional information, please contact Tony Gurule via phone at (303) 768-8669 or email at tony@voseconsulting.com.
| 
