The implementation of traditional value-based selection strategies is now becoming more than a simple academic exercise. More advanced portfolio management strategies, based on client value, are now being adopted by an increased number of financial institutions, with simulated positive impacts on expected portfolio profitability (up to +85%) and on regulatory capital (up to -30%).

Although financial institutions have become more and more sophisticated, with some moving from decisioning based on probability of default (PD) to basing it on expected losses (PD*loss severity), there is still room for improvement of their current practices. It is not only a matter of default risk, but also of the potential value or profit that the application can generate. This depends on several elements, of which expected losses is only one. The terms of business (e.g., term, interest rate, etc.) can be adjusted to cover all costs and still make the transaction profitable. 

The figure below shows the simplified Profits & Loss statement of any lending decision. The risk component (or the unexpected loss linked to a credit decision) is represented by the cost of the capital at risk or economic capital, as derived through the Basel II advanced internal rating or any VaR model.

Table 1 - Example of profits & loss statement

Including the cost of risk is essential for a more holistic view of any lending decision. If it is easy to answer questions like “Who’s riskier?”, it may prove problematic to define by how much one applicant is riskier than another and quantify the economic impact of such difference. This is given by the cost of economic capital. Neglecting it causes an underestimation of the total costs. Setting the contract conditions (for instance, price and down payment) on the basis of such an estimate can impact negatively on the expected profitability, given that such a cost item can account for a considerable share of the total costs (15-20% in some cases).

It is common sense that riskier clients should be charged a higher price than less risky ones. However, selection or pricing strategies based on customer risk adjusted value are not quite common practice in the industry. In several markets, evidence shows that there is no link between price and risk, as shown in the figure below (the current price is the observed price for a retail credit portfolio). On the contrary, the risk adjusted price curve shows the minimum (theoretical) price that should be applied to each applicant in order to match all costs, cost of risk included: the higher the risk, the higher the price. These are the basic principles behind the risk based pricing strategy: the right price for each applicant (respecting legal price limit, where necessary). 

In many markets, though, several institutions act as price taker (adopting market prices) and adversely select some of their clients. This means that they accept some clients that they should have rejected as they would be unprofitable (area B in the figure below). As a consequence, a part of the portfolio is not generating value, maybe still leading to positive, but suboptimal, portfolio performance (as area A may be greater then area B).

Figure 1 - Current vs. risk based pricing strategy

Although some financial institutions have adopted some basic risk-based pricing strategies, portfolio performance can still be non-optimal. Deviations of the current price from the theoretical one are often attributed to cross-selling opportunities being discounted in the applied price. Somehow, such financial institutions could be aware of the client (life time) value and act accordingly. Again, evidence on some retail finance portfolios does not support such a belief.

As stated, financial institutions are reluctant to use a full risk-based pricing strategy but can still improve their current risk management practice. More advanced decision making processes based on risk-adjusted assessment of the application (or client) can still be applied, by:

• Deciding on a credit application on the basis of the expected risk adjusted performance, by using any risk adjusted performance measure, such as RARORAC (risk-adjusted return on risk adjusted capital);
• Varying the terms of business in order to reduce the costs and match revenues, leaving the price unchanged (risk-based processing).

In the first case, the expected risk-adjusted return of any application is calculated taking into account the P&L statement shown above, price being given (whereas in the risk-based pricing strategy, the price is unknown at this point). The application is accepted if this is positive or equal to zero (this means the application is expected to generate revenues greater or equal to the estimated costs). If RARORAC is used, the rule of thumb is:

If RARORAC ≥ 0 then Accept, otherwise Reject.

Under such an approach, the application is only selected if it can create value with the current terms of business. In fact, this performance measure allows financial institutions to create a more complete picture of an application (or client), as opposed to traditional performance measures (e.g., ROE) that neglect some cost elements, thus providing misleading indications. An example of the distortion caused by such measures is shown in Figure 2. In all cases, the real performance (risk-adjusted) is always lower than that derived by ROE. In two portfolios (2 and 3), the risk-adjusted performance is even negative, whilst the manager’s perception (as offered by ROE) was the opposite. Portfolios 2 and 3 have not created value, but instead lost some.

Figure 2 - Retail credit portfolios performance: ROE vs. RARORAC

Under the RARORAC selection approach, applications with a negative ratio have to be rejected as unprofitable. However, financial institutions can alternatively decide to change contract conditions (e.g., by asking for collateral) in order to reduce the costs (e.g., expected and unexpected losses) and then balance them with the revenues generated by applying the current price.

This process is graphically detailed in Figure 3. The case of a RARORAC < 0 is in the area where the current (or market price) is lower than the risk-adjusted one (the minimum price). The contract terms do not grant adequate coverage for all costs in their current state.

Figure 3 - Risk-based pricing strategy

The new conditions are set to reduce the costs, pushing the minimum acceptable (risk-adjusted or theoretical) price back to the one currently applied. At this stage, one critical issue is the choice of which conditions to adjust in order to achieve the 'right' risk-profit relation, given their different impacts. The table below provides the typical risk drivers and their effectiveness on risk correction and cost reduction.

Table 2 - Risk drivers and their impact on costs

Some market players already follow such a strategy, yet end with partial evaluation of the overall expected performance, the new strategy being driven more by a partial view rather than more rigorous approaches (e.g., risk is not adequately considered, customer life time value is over estimated, etc.).

Risk-based processing an application proves to be more effective when collateral can be included in the contract conditions. Reducing the loan amount (eventually together with the term) would leave some credit needs unsatisfied, thus forcing the client to meet them through other financial institutions in the market place. As a consequence, the client global risk profile would be somewhat underestimated. Some may argue that when the applicant receives a counter offer for a lower amount, for instance, in the case of personal loans, he/she will actually reduce or limit his/her needs. Psychologically, the applicant may be willing to change their plans. However, that may not work when there is 'real' need for finance, like financing primary needs.

Going real
Experian Decision Analytics has already developed, implemented and tested a risk-based decisioning solution. Fully parameterised, the solution is integrated in the client’s process flow (which can be origination or account management) as an ad hoc module of Strategy Manager (and Decision Agent). A typical implementation is as shown in Figure 4.

Figure 4 - Typical architecture of risk-based decisioning solution

The Decision Agent is able to derive all or either of the following outputs: the minimum risk-adjusted price, the RARORAC and the set of new conditions (under the risk-based processing). Such elements are then integrated into the client’s decision making process flow, leading to the final decision (e.g., Accept/Reject, change conditions, etc.). Final decisions are stored in the analytical data mart, allowing future analyses and reporting activities, in order to monitor these strategies and refine them if necessary.

Conclusions
Managing risk more actively is now becoming key to success in more and more competitive markets, and the three strategies (risk-based pricing, risk-based or RARORAC selection and risk-based processing) are being used for a variety of needs: some financial institutions are facing the problem of managing new market niches, such as sub prime clients; some others are reviewing the strategies (combination of price, term, down payment, etc.) for applications generated through the dealer network; some are simply facing the need to increase sales. The above strategies aim at maximising the expected risk-adjusted portfolio profitability, with just a few constraints that can be easily considered, such as the reject rate.

Although profit maximisation can be a very appealing target, it can have unexpected effects on financial institutions’ other key performance metrics, like expected portfolio bad rate, expected losses, provisions, etc. Therefore, these strategies should be applied in a flexible way, balancing appropriately the different strategic targets. For instance, blindly applying a risk-based strategy to all applicants would have the perverse effect of increasing adverse selection rather than eliminating or reducing it. If no entry barriers are set, high risk clients that face difficulties in obtaining credit may be willing to pay more in order to get it. The financial institution applying the risk-based pricing strategy will end up attracting such customers, with obvious impacts in terms of the expetced quality of the portfolio.

More realistically, the right approach can be based on a combination of different strategies, for example, to keep selecting applicants depending on their probability of default, and then adjusting the price in all or part of the accepts.
This is just one example. More complex strategies can be adopted that consider the various objectives and constraints that have to be met.

Most importantly, so far only maximisation has been mentioned. An institution willing to achieve the maximum benefits (e.g., portfolio profit) given a set of constrains (e.g., max level of expected losses, max amount of economic capital, etc.) has to face much more complex problems, of which the above is just one element. Strategy optimisation is then the right solution.

Finally, the problem can be even more complex if the view shifts from application to customer level. Estimating customer value throughout the expected duration of its relationship with the financial institution radically changes the operational framework. The P&L statement in Table 1 will have to reflect this new dimension, embedding all the new costs and revenue items, deriving from cross-sell opportunities, induced clients, etc.

Daniele Vergari & Emiliano Forti - R&D Consultants, Experian Decision Analytics - taken from the presentation given at the Experian Day, Italy, March 2007

Contact us for further discussions about this article

Top