Risk Analytics

The following sections are included:

• Foreword
• About the Author
• Contents

As evidenced by the plethora of academic and professional certification programs, Risk Management has definitely come into its own as a body of knowledge. The timing of this book could not have been better though unfortunate. The entire world is going through a Black Swan event (made popular by Nassim Nicholas Taleb) due to COVID-19. Losses to both firms and countries have reached trillion of dollars. On the well-known loss distribution, this occurrence is quite certain to be outside of the 99% confidence level or less than once in a 100 years. The event has significantly altered not only economics but also lives and society. Social distancing became a new term. What is most pertinent for the purpose of our discussion here is that it may not take another 100 years for a similar event to recur with the ever-growing global linkages. Besides, losses that are now in trillions may rise to tens or even hundreds of trillions. The Black Swan today may just be a regular White Swan in the near future. Losses of such magnitude become more frequent not unlike Category 5 hurricanes or tsunamis. All these are educated guesses but for those making investment decisions involving millions or billions of dollars, these prospects need to be analyzed and quantified rather than left to judgment…

With increased complexity and understanding, not all risks are the same. The Deputy President of a large state bank asked me to help with the creation of an Early Warning System (EWS) to enhance risk management. My question to her was: What does the bank want to be warned of ? Is it of dangers ahead or critical deficiencies within? A better understanding of the type of risk we wish to address is essential to designing and deploying the system needed to address it…

There are several building blocks to risk analytics. It is not uncommon that each is managed in silo. The repeated thrust of this book is that effective integration is needed to develop an automated process where analytics can support better decision-making. A brief examination of each component here is useful to understanding the gaps that may exist.

As mentioned earlier, an essential aspect of developing automated deployable risk solutions is to be able to integrate models and analysis into the database. This chapter shows can that can be accomplished…

Data are at the heart of analytics and models. The most advanced software cannot compensate for data deficient in critical aspects. The implications of data issues may not be apparent. For academic research purposes, they can be overlooked with little cost, but if million or even billion-dollar decisions are made based on the results generated, the gravity is of a different order. These deficiencies discussed here are based on actual work on real data. Before any risk analytics or modeling is carried out, data audit is a necessity to ensure that any results obtained are reliable.

It is almost impossible to support analytics directly from operational or legacy systems. Organizations that have tried quickly face issues too challenging even for professional IT service providers. There is increasing recognition of this matter, but data warehousing remains a hit and miss undertaking in most places.

The criticality of data warehousing has been recognized by the creation of cloud-based data warehouses. That is a boon to many types of industries, but for banks, it may not make a material difference. In most regulatory regimes, banks are required to keep data within their own premises. Even data centers are disallowed except for archived data. Creating a fully owned and controlled data warehouse is not an option.

In one bank, around $30m was spent on developing a data warehouse by the IT associate company over 5 years. When the project was completed, not a single piece of datum could be extracted and used for analysis. The bank CEO ordered that this blunder be kept a secret to avoid embarrassment. In reality, such outcomes are fairly common. Knowing the difference between operational and analytical data can help reduce the frustrations often encountered.

From years of consulting for and developing analytical solutions, I have come to realize that one of the biggest challenges in this area is the inability to distinguish analytical data from operational ones. From a meta-physical perspective, data are just data. There is no reason why analysis cannot be made of data that are available. It is unsurprising then that many decision-makers are frustrated in not having good insights to guide them and blame all responsible for data for their incompetence. At this juncture of the book, it is clear that not all data are alike. With every organization attempting to leverage on their data assets, I have always advised the need to create an analytical data sphere. What follows are some features that can help guide this attempt…

Stripped to its essence, financial service is about risk-taking. There is no tangible product or hedonic service provided to the client. In taking deposits, a bank is assuming the credit risk on behalf of the depositor. Insurance is about taking on an event risk. Perhaps the only type of financial service which requires nearly no risk-taking is passive fund management. Here, the financial institution offers diversification which is beyond the means of most individual customers. Risk analytics then is not an option in a financial institution…

There are many kinds of risk models and analytics, and the range is expanding. Excellent books have covered them in detail, so it is not the intent here to reinvent the wheel. Only the most common metrics are discussed first with a synopsis of the concern or theory followed more extensively by the ORE codes that can be used to obtain the results and store them in the Oracle database for decision support. Another chapter is devoted to using R in Microsoft Excel for the less time-critical analyses like those for strategic planning.

Banking is one domain where risk analytics has become prominent. This is largely due to a significantly more demanding risk management regime known as the New Basel Capital Accord or Basel II. More than 100 countries have committed to this standard and implemented local regulations to enforce compliance. It has been estimated that European banks have to spend tens of millions of Euros just to acquire the needed software and hardware to develop the metrics required…

Desktop analytics are useful for planning. R itself is now embedded in Microsoft Excel as Basic Excel R Toolkit or BERT, enhancing the use of the already popular business utility. What has been covered in the earlier chapters could be replicated in Excel and there are some types of analytics well suited to that platform. These are discussed in later chapters. The efforts incurred would, however, have gained limited returns if analytics like a rating system cannot be deployed in an automated process to support intensive and time-critical operational decision-making. In many developed countries, there is a demand for retail credit approval decisions to be made in no more than five seconds from the time of application. With today’s level of digital data flow, any financial institution that is too slow in making the necessary analysis will find itself outcompeted soon. The ORR development process described earlier will be used to discuss the merits of deployment for automation…

The use of periodic data like financial ratios to develop credit risk models is known as Point In Time (PIT) modeling. Theoretically, such models are good enough for predictions around the time of the data. They can be viewed as snapshots of financial health. Due to practical limitations, such models are used for a significant period, usually a year, before being recalibrated…

So far, the book has advocated development of deployable risk analytics integrated with a database. One or more models can process new data and produce results without the need for manual extraction and subsequent reinsertion of outputs. Assigning a risk rating to a new credit applicant is a business process that this design is well-suited for. Not only does complete automation save time and effort, it reduces the possibilities for errors and enhance competitiveness if processing is time critical like in the retail credit business. There was a time when this was known as Decision Support System (DSS) when IT was leveraged to expedite decision-making. There is, however, a wide range of risk analytics that can be done at a more considered pace on a desktop and presented at meetings. These are used to support strategic and even tactical decisions that do not have to be made soon. In fact, such analytics benefit from the typical “what-if” analysis where parameters are tweaked to evaluate the range of outcomes possible, usually at a meeting of corporate planners. In most business environments, Microsoft Excel or spreadsheet equivalents have been widely used for this purpose. With R now embedded as a free add-in, Excel is a useful tool for risk analytics that are not time critical. This chapter will cover some of the common topics.

With the increasingly widespread use of R, resources continue to be added to the ecosystem. Packages have already been mentioned in earlier chapters. Though this book is largely devoted to the use of Oracle R Enterprise (ORE), tools like an R editor can be very helpful in the coding process.

The following sections are included:

• Annex A: Meeting of Minds Questionnaire
• Annex B
• Index