Improving capabilities internally to deliver stress tests in a more comprehensive way and increase the efficiency of stress testing

By Pradyumna Javalekar, Executive Director, Head of CVA Stress Testing, JP Morgan Chase & Co.

What, for you, are the benefits of attending a conference like the ‘Stress Testing Europe Summit’ and what have attendees learnt from your session?

I have been involved in stress testing at two of the world’s leading banks for many years of my career. Over the years I have realized that the application and breadth of stress testing is vast and the challenges involved in its implementation are many. There are various forms and motivations behind stress testing – there is capital stress testing (to test the firm’s capital position in an extreme scenario), risk appetite (to see if losses will be within a risk appetite defined by senior management), ad-hoc stress tests to determine the impact of topical issues etc. Some of these are mandated by regulators and some are conceptualized internally and in many cases there is an overlap between the two.

I believe that conferences like this one are a great way to compare notes, discuss best practices and learn about aspects of stress testing that one has potentially never heard before. There are enormous challenges in setting up a sustainable stress testing framework and process and the bigger and more complex the institution, the more ominous the task. The key is to take it one step at a time and a certain level of accuracy at a time. Simplify things in the beginning so that you start getting some meaningful results and then peel off the many layers of complexities one at a time. At every point in the process of building such a framework, it is important to keep an eye on consistency and scalability. Models, databases and processes should be kept similar and fungible as much as possible. I discussed this at the conference and we all had much to learn from each other.

What is the best way for risk professionals to develop their databases?

In terms of databases, there are the inputs and the outputs. You need to think about both very carefully. The inputs include the firm’s current positions/risk sensitivities/balance sheet/capital etc. as well as scenario information (macro-economic variables or “shocks”). This information needs to be stored in a consistent format across asset class, product and business segment as much as possible. The scenario information though, might need some tailoring by asset class depending on the granularity of the scenario definition.

In terms of outputs, how you store the scenario results plays a significant role in how quickly you can produce meaningful reports and summaries for risk managers and senior management. The more granular your results are, the easier it is to aggregate across segments and produce variations of different scenarios. However, on many occasions, you would use a “full revaluation” approach in which case granularity may be lost. It is important to strike a balance between the accuracy and the granularity/flexibility in the way you store the outputs.

Can you provide our readers with an overview of Automation of creation and execution of stress test models?

First of all, stress testing models should be fit for purpose. Complexity is not always a good thing. A stress test is meant to capture “tail risks” and is by design a significant approximation of an extreme tail event. However, a lot of the time stress testing models are driven by regulatory requirements which do require a certain degree of accuracy and/or conservatism. Nonetheless, for organizations developing new models for internal stress testing purposes, they should evaluate the cost/benefit of developing more complex models.

In terms of automation and efficiency, the ideal framework is one in which you can simply re-use the models you use on a business as usual basis to evaluate the firm’s position on a regular basis, e.g. valuation models or budget processes, for stress testing, i.e. you feed the scenario variables/shocks into your existing BAU models and re-evaluate your firm’s position in a stress scenario. Particularly for market-to-market assets, this is referred to as “full revaluation”. If such a framework does not seem feasible, the aim should be to build fungible models, i.e. models that can be applied to as many different business segments/asset classes and purposes as possible. For example, “sensitivity based” stress models are quite powerful because they are typically agnostic to the exact nature of the inputs (i.e. the asset class or business segment) and are quite generic in nature.

What are the main challenges and opportunities of handling complex products in Stress Testing?

Complex products typically do not fit into existing models easily and require workarounds until the models are enhanced to accommodate them. Again, here a cost/benefit analysis is necessary – ideally you want to enhance your existing models and capture more products as it is strategic and will help in the long term but if the volume or materiality of such products is low and the cost very high, it may be pragmatic to settle for a conservative workaround. In terms of regulators, although they rightly push for greater risk capture, in my experience they do appreciate the difficulties involved and are not always against a reasonable and conservative alternative.

It might work well if firms can come up with a generic solution to complex products instead of bespoke solutions for each product. A workaround that is more generic in nature, for example that can applied equally well to a complex interest rates product and an FX product, is a more scalable and sustainable alternative in the long run.

What do you foresee for collaboration across the bank in the next 12 months?

It is next to impossible to achieve meaningful results from stress testing without significant collaboration across various functions in the bank, particularly Risk, Finance and the Front office. In particular for our firm, although our stress testing framework has relatively stabilized over the years, the impending Brexit and a move towards European centres brings new challenges. Depending on the outcome of Brexit and the direction of the bank in terms of its European business, we will need to collaborate deeply with colleagues in our continental European centres and adapt more to European regulators, particularly on stress testing requirements.