Journal of Operational Risk
ISSN:
1744-6740 (print)
1755-2710 (online)
Editor-in-chief: Marcelo Cruz
A nonlinear analysis of operational risk events in Australian banks
Yifei Li, Neil Allan and John Evans
Need to know
- Cladistics analysis provides new insights into operational risk events in banks by identifying the combinations of underlying characteristics that were involved in each risk event.
- The analysis for Australian banks indicates that poor control has been by far the main characteristic of operational risk event losses.
- The analysis has identified the main characteristics involved with operational risk losses and these are shown to be consistent over time, thus enabling management to concentrate on these characteristics to reduce future losses.
Abstract
We propose a methodology applied to complex systems to analyze operational risk events in banks, with the objective of determining an understanding of the key characteristics and their relationships in initiating operational risk losses. We applied our methodology to operational risk losses in Australian banks over the period 2010-14. The analysis identified that there are a small number of characteristics that are common to many operational risk events, and these "level 1" characteristics are stable across time, which implies operational risk losses could be controlled by managing these characteristics. The methodology adds value to the existing analysis by identifying the main characteristics of operational risk events in a rigorous manner.
Introduction
The Basel Accord defines operational risk as “the risk of loss resulting from inadequate or failed internal processes, people and systems or from external events” (Basel Committee on Banking Supervision 2006). Figure 1 indicates the distribution of losses for Australian banks’ operational risk events in the ORIC International database over the period 2010–14. As can be seen, these are not trivial amounts.
The total operational risk loss reported to the ORIC database from 2010 to 2014 was A$3.75 billion, spread over ninety-four events, with the largest reported loss being almost A$700 million. Bear in mind that this understates the operational risk losses for this period, as not all Australian banks submit data. Further, all banks have a truncation point,11A truncation point is the level of loss below which the bank does not report the loss. These change over time, and observation of databases of operational risk events would suggest they are not strictly adhered to for reporting. so smaller losses are not always included. For comparison, the total shareholder equity of the Australian banks at June 2014 was A$225 billion, and their total net profit was A$32 billion (APRA 2014).
Another indicator of the relative importance of operational risk to Australian banks is the amount of capital they hold for the four major risk types, as presented in Figure 2. This shows the amount of risk capital (in billions of Australian dollars) for these risk types as a proportion of total capital being held by Australia’s biggest four banks in 2012; it can be seen that operational risk capital is the second largest amount being held, at almost 10% of total risk capital (Corrigan 2013).
Traditional analysis of operational risks has mostly attempted to fit statistical distributions to operational losses, with the intention that these distributions could be used to estimate future operational losses. But this has failed, especially for extreme operational risk events, as shown in Ganegoda and Evans (2013). Further, statistical analysis does not provide any information to help management to understand what might be causing the operational risk events. Given the significance of operational risk losses to the banks, analysis that could identify potential characteristics of the losses and thus enable management to reduce the losses by targeting their causes should be of interest to the banks and their regulators.
In this paper we shall explore the features of the Australian banking system that create challenges for traditional statistical analysis, and then introduce a methodology that can be used in the environment identified. We use our methodology to analyze the Australian banks’ operational losses over the period 2010–14 to establish their major characteristics.
2 The Australian banks’ operating environment
Australian banks operate in the global financial system, which can be shown to have the characteristics of so-called complex adaptive systems (CASs). It can also be shown that the operating environment within the banks is a CAS. It is this CAS feature of the banks’ operating environment that necessitates a different analysis than traditional statistical methodology (Daníelsson 2008) to appropriately understand the results of the interactions, adaptations and innovations that occur in the banks. In this section, we identify the general features of a CAS, and map these to the features of the Australian banks’ operating environment to demonstrate the applicability of the methodology used to analyze operational risk events in Australian banks.
CASs are observed in many disciplines (Newman 2011). There is no generally accepted formal definition of a CAS, but Mitchell (2006) postulated that it is generally seen as being a large network of simple components with no central control, in which emergent complex behavior is exhibited. Importantly, CASs do not have central organizational control, and, consequently, analyzing subsystems or individual agents’ reactions to events does not predict what the system as a whole will do (Guckenheimer and Ottino 2008). However, CASs do exhibit self-organization within and between the interconnected clusters arising from learning, adaptation and innovation (Mitleton-Kelly 2003).
The underlying characteristics of a CAS are usually then seen as follows.
- •
The agents in the system interact with each other, but not necessarily with all other agents.
- •
The outcome of the system arising from some event cannot be predicted by observing the likely outcomes for the individual agents acting independently.
- •
The agents have their own reactions to events but are also influenced by the expected reactions and actual reactions of others that they interact with.
- •
Reactions to seemingly small events can have a significant effect on the reaction of the overall system.
It should be noted that CASs are open and are therefore influenced by both the arrival of new agents to the system and the exit of existing agents, which further adds to the complexity of the system. The interconnectedness between the agents may also change.
We next demonstrate that the financial markets within which the Australian banks operate have the necessary characteristics to be classified as CASs. We then consider the operating environments within the Australian banks themselves, which by necessity are influenced by their external environment.
The Australian financial system consists of
- •
markets in which financial instruments are traded,
- •
agents in the financial system who carry out particular functions.
In terms of understanding the outcomes of the financial system, it is the agents’ behavior that is important, as the markets themselves are just processes for trading securities.
The major agents in the Australian financial system are
- •
bankers, who predominantly facilitate cashflows in the economy and between Australia and other economies, borrow and lend cash and operate currency swaps on behalf of clients;
- •
asset managers, who accumulate funds from investors and buy and sell a range of securities;
- •
security brokers, who act on behalf of investors to buy and sell securities;
- •
insurance companies, who accept the risk of some event damaging to the insured in return for premiums;
- •
retirement funds, which predominantly accumulate cash from individuals to fund their retirement and invest in assets;
- •
individuals, who predominantly borrow and lend money and use the financial payments system to facilitate payments and receipts;
- •
businesses, who predominantly borrow and lend money and use the financial payments system to facilitate payments and receipts;
- •
regulators, who control entry and sustainability of the agents and behavior of the agents in the financial market, primarily to maintain the stability of the system; and
- •
exchanges, which provide facilities for the sale and purchase of securities.
All of these agents interact with each other in the system and with agents in other economies. New agents can enter the system and existing agents can exit the system.
In terms of mapping the characteristics of financial markets to the generally accepted characteristics of a CAS, we can state the following.
- •
The financial system agents interact with each other as follows:
- (1)
bankers borrow money from and lend money to the other agents, carrying out currency swaps for the other agents and facilitating cashflows between the agents;
- (2)
asset managers use the banking system to transfer cash as required, arrange currency swaps as required and borrow money for the leveraging of investments;
- (3)
security brokers buy and sell securities on behalf of bankers, asset managers, retirement funds, insurance companies, individuals and businesses and use the banking system for cashflow facilitation;
- (4)
insurance companies insure potential losses for bankers, asset managers, security brokers, retirement funds, individuals and businesses, buy and sell securities through security brokers and use the banking system for management of cashflows;
- (5)
retirement funds lend funds to bankers, manage cashflows through the banking system, buy and sell securities through security brokers, use asset managers to invest assets, buy insurance from insurance companies, receive cash from or on behalf of individuals and pay retirement benefits;
- (6)
individuals use the banking system for cashflow management, lend money to and borrow money from banks, invest through asset managers, buy and sell securities through security brokers, buy insurance from insurance companies and invest through retirement funds to receive retirement benefits;
- (7)
businesses use the banking system for cashflow management, borrow from and lend to banks, and buy insurances from insurance companies;
- (8)
regulators control the amount of capital and behavior of agents;
- (9)
exchanges provide markets and clearing facilities for the sale and purchase of financial securities.
- (1)
- •
The agents have their own reactions to events but are also influenced by the reactions of others to the same event. The October 1987 US stock market crash shows how this can occur, and, to quote Gerald Corrigan, former president of the New York Fed, just before the October 1987 crash (Corrigan 1987): “In recent years the pace of change and innovation in financial markets and institutions here and around the world has increased enormously, as have the speed, volume and value of financial transactions. The period has also seen a greatly heightened degree of aggressive competition in the financial sector. All of this is taking place in the context of a legal and a regulatory framework which is increasingly outdated and ill-equipped to meet the challenges of the day. This has led to concern that the fragility of the system has increased, in part because the degree of operational, liquidity and credit interdependency has risen sharply.”
- •
Small reactions to events by a small group of agents can have a surprisingly significant effect on the reaction of the system. The May 2010 “flash crash” is a good example, where it is argued that a few traders using trading algorithms caused a major shift in the US security market over a very short period of time, initially negatively and then positively. While there was not much resultant overall change, the sudden and significant changes caused other financial system agents around the world to have to react to what was transpiring to manage their own portfolio of affected assets. To quote Haldane (2013, p. 243), “an external event strikes. Fear grips the system which, in consequence, seizes. The resulting collateral damage is wide and deep. Yet the triggering event is, with hindsight, found to have been rather modest. The flap of a butterfly’s wing in New York or Guangdong generates a hurricane for the world economy. The dynamics appear chaotic, mathematically and metaphorically.” Apparently, flash crashes are not isolated occurrences, but have occurred quite often over the past century (Aldridge 2013).
- •
The financial system is broadly an open system, although there are barriers to entry for some agents, such as banks and insurance companies, which are required to have significant capital and management expertise. There are exits from and entries to the system, which makes its analysis complex, as the dynamics of the interactions of the new and departing agents with the existing agents need to be understood. The financial markets also see partial exits, as some agents exit parts of their business from time to time and enter new businesses.
It is clear that the financial system in which the Australian banks operate exhibits all of the classic characteristics of a CAS, with the consequence that analysis of the likely reaction of individual agents to an event is not going to yield valid results for the system as a whole, and, by implication, historical linear-based statistical analysis will not provide a reliable estimate of the future. A risk analysis of the financial system that uses statistical distribution analysis and constant correlations across agents will therefore fail to provide a reasonable indication of outcomes for the system as a whole, particularly at the extremes, where, as we shall demonstrate, the reaction of a small number of agents can have a significant ripple effect across other agents.
Considering the operating environment within a bank, we demonstrate that Australian banks interact with other agents both domestically and internationally in a CAS. All of this interaction between the banks and other agents necessitates operations within the banks to administer the necessary functions. Administrative functions require systems to record and activate processes and people to carry out the functions. It is this “people involvement” that creates a CAS within the bank, primarily through bank staff needing to interact with each other to achieve desired outcomes for other agents, and with systems to record and execute services and to achieve the desired outcomes in the financial system for their particular bank. The reactions of the overall bank cannot then be analyzed by considering the likely reaction of individual bank staff within the bank, as they are influenced by historical observed reactions and their views of likely current reactions to events.
These characteristics of the bank operations create a CAS within the bank, which results in significant difficulties in successfully analyzing operational risk events with statistical techniques that assume linear relationships.
3 Analysis of complex adaptive systems
In analyzing CASs, the aim is to understand the interrelationship between the agents and their overall likely reaction to events.
The analysis of CASs evolves from systematics, which has existed in various forms over the years; Hutchinson (2013) describes the systematics process as “the field of biology that deals with the classification of all organisms and hypothesizes how they may be related. A significant discipline within systematics is that of taxonomy. This is the process of describing organisms, or taxa as they are referred to, grouping them and assigning scientific names. This is complemented by phylogenetics which deals with the creation of phylogenetic trees, generally accepted as hypothetical depictions of the evolutionary relationships between taxa. In the mid-twentieth century the popular methodology for developing phylogenetic trees was phenetics, however, this was largely replaced by the end of the twentieth century and now cladistics is the prevailing methodology, in which the trees are referred to as cladograms. It uses a parsimony criterion that searches for the most efficient tree. That is the tree with the fewest gains and losses of characters.”
In analyzing historical relationships in biology, it needs to be recognized that, while the analysis relies on facts, the algorithms used to determine the ancestry are subject to model error and assumptions. A comprehensive analysis of the alternative methods of cladistics analysis can be found in Makarenkov et al (2006).
As an example of cladistics analysis, Figure 3 shows a simplistic tree of the possible evolution of lizards, crocodiles and birds (Benton 2005). This tree is simply indicating that lizards, crocodiles and birds have a common ancestor, Diapsida. The green triangle simply indicates the most distant ancestor (not shown here) with characteristics common to all other animals in the tree, and the red diamond is the next ancestor with characteristics common to the animals to the right of Diapsida.
There is a growing literature researching nonbiological systems using cladistics methodology to give a better understanding of the consequences of the systems being CASs. As observed by Mitleton-Kelly (2003), “one way of looking at complex human systems is to examine the generic characteristics of natural complex systems and to consider whether they are relevant or appropriate to social systems”.
When adapting the cladistics methodology to other systems, we need to ask whether the characteristics of the biological system can be mapped to the system being considered (Mitleton-Kelly 2003). The related concepts of CAS and cladistics analysis have been used for the following.
- •
Identifying the linkages of characteristics causing significant derivative-based financial failures. Allan and Corrigan (2013) looked at the characteristics for some large international operational risk events involving derivatives and produced the tree in Figure 4, which shows the most likely linkages that resulted in these events. Their analysis indicated that, by using cladistics analysis, it can be shown that the most common characteristics involving these derivative events were fraud, normal trading activity that went wrong and the fact that derivatives were involved. The clear message to institutions trading in derivatives was that just carrying out that activity should heighten the institution’s awareness of things that can go wrong, and should they do so, that they should be aware of likely fraud, as well as checking that the trading is being carried out correctly.
- •
Identifying emerging risks in financial institutions. Allan et al (2013) observed that “the identification and assessment of emerging risks can become more robust by using a systems approach that enables a clearer understanding of the underlying dynamics that exist between the key factors of the risks themselves. It is possible to identify interactions in a system that may propagate hitherto unseen risks. Emerging risks can be viewed as evolving risks from a complex system. It is also known that such systems exhibit signals in advance of an observable change in overall performance. Knowing how to spot and interpret those signs is the key to building a scientific and robust emerging risk process.”
- •
Explaining the continuation of booms and busts in financial markets. Hommes and Wagener (2009) considered two types of investors, one following a fundamental value methodology and the other following a trend methodology, with both rationally bounded within a capital market exhibiting CAS characteristics. The authors demonstrated that their interaction results in booms and busts as the fundamentalist enters and leaves the market, causing the trend-following investor to also enter and leave the market. They concluded that “the most important difference with a representative rational agent framework is probably that in a heterogeneous bounded rational world, asset price fluctuations exhibit excess volatility”. Hommes and Wagener (2010) demonstrated a similar result.
- •
Explaining the need for corporations to maintain innovation in products in order to maintain market share. McCarthy (2005) cited as examples the margarine and cash register industries; the former showed little volatility in market share as participants developed innovative products to maintain their market share, but the latter showed significant market share volatility due to low levels of innovation. Markose (2005) concluded that “either outcome is in keeping with the Red Queen principle in that players cannot maintain status quo along a critical dimension unless they can keep up with their peers”,22The Red Queen principle is derived from Alice Through the Looking Glass (Carroll 1871), where the exchange went: “Well, in our country,” said Alice, still panting a little, “you’d generally get to somewhere else – if you ran very fast for a long time, as we’ve been doing.”“A slow kind of country!” said the Queen. “Now here, you see, it takes all the running you can do to keep in the same place.” which would not be a conclusion drawn from market equilibrium economics. Markose also suggested the Red Queen principle applies to investment managers who need to innovate through their investment processes, but these are also interrelated, with the consequence of the process being innovation that changes the dynamics of the market.
- •
Explaining the emergence of industrial clusters. Andriani (2003) explained that industrial clusters will emerge where there is a strong coevolutionary coupling between businesses, and their connectivity will create an environment in which they are sensitive to others in their cluster. This does not mean businesses will not fail; as Andriani points out, in Silicon Valley the loose coupling is the engineering teams that move from one business to another.
There is, then, a growing body of literature that analyzes quite diverse systems by assuming they are CASs, and that explains phenomenons that were unexplainable by traditional theories.
4 Justification of cladistics analysis for bank operational risk events
To date, there has been an emphasis on applying statistical modeling to operational risk losses. However, one of the issues for modeling operational losses in financial institutions is that, as demonstrated by Daníelsson (2008), financial markets and financial institutions exhibit the features of a CAS, which do not lend themselves to statistical modeling. The main reasons for this are that statistical analysis assumes constant, and therefore replicable, linear relationships within the system being analyzed; this is not true for the financial systems and the institutions that operate within them, as they adapt and innovate over time.
In determining whether cladistics analysis methodology is appropriate for analyzing operational risk events, it is important to note that these risk events occur within agents in the financial systems, which have been demonstrated to be CASs. For operational risk events that have occurred, we can determine using historical data the major characteristics that were involved – these parallel the characteristics used in biology. The algorithm that can be used to determine the optimum interconnectedness of species in biological evolution can thus also be used to determine how various observed characteristics have combined in a financial system to produce the operational risk event. This is because the algorithm for cladistics analysis of operational risk events would simply link the characteristics of the events that have occurred, such that the combinations of operational risk events result in the smallest number of combinations of the observed characteristics, which is consistent with the view that this produces the outcome closest to the real situation, as postulated by Lin et al (2007). Cladistics analysis is therefore a valid methodology for analyzing the Australian banks’ operational risk events.
We will use cladistics analysis to provide insights into the characteristics of Australian banks’ operational risk events from 2010 through 2014.
5 Data
The data for this study was obtained from RiskBusiness OpRisk IntelliSet (RBI), which collects information related to operational risk events in the financial sector, particularly for banks and insurance companies. The data set is based on public news reports and documents, and it is updated on a daily basis.
In order to eliminate differences that might arise across different industry sectors, for our analysis we selected data for the Australian banking industry for the period 2010–14. After removing events with the same characteristics, the sample consisted of eighty-six risk events.
Table 3 in Appendix A online sets out the risk events we included in our analysis.
6 Methodology
Our research methodology is based on that adopted by McCarthy (2005) and Allan and Corrigan (2013). Our first step was to select the characteristics of the operational risk events to be analyzed in the study. These were chosen so as to be observable from the descriptions of the risk events, and we chose an extensive set of events so as to capture as many characteristics as possible. The characteristics determined for the Australian banks’ risk events are shown in Tables 4 and 5 in Appendix A online.
It should be noted that the characteristics selected are just features of the risk events, and are not necessarily drivers of any particular risk event. This applies in particular to the “single person” and “multiple people” characteristics, which simply identify whether there was one person or more than one person involved in a particular group of risk events. A similar rationale applies to the business lines as characteristics.
To analyze the interrelationships between the characteristics of the operational risk events, we used the maximum parsimony methodology, which chooses the solution that uses the least number of steps needed to explain a relationship or phenomenon (Farris 1983); in our case, this is the cladistics tree that best represents the grouping of characteristics involved in a group of operational risk events. We adopt Camin–Sokal parsimony (Camin and Sokal 1965) for constructing the cladogram.
In this study, we used Evolutionary Risk Analysis (ERA) software from Systemic Consult Ltd to produce trees of related characteristics for the operational risk events. (Other software is also available to undertake this analysis.)
7 Results
In order to gain maximum insight into how the various characteristics we identified may relate to the operational risk events, and, in particular, into the stability and evolution of the relationships, we analyzed the results both with and without “business line” included as a characteristic, and for differing time periods. As well as analyzing one-year results, we considered cumulative results to better understand how the relationships between the characteristics change over time.
Figures 5 and 6 in Appendix B online show the results of the analysis, with and without business lines, in the typical “tree” format. The analysis without business lines on a year-by-year basis is shown in Figures 7–9 in Appendix C online. Figures 10–14 in Appendix D online show the analysis without business lines over a longer period.
The “trees” are read from left to right, with the characteristics on the leftmost side being the most common to the group of risk events shown on the right-hand side and joined by the branches.
The leftmost characteristics can be thought of as the “level 1” characteristics, ie, they show up more than other characteristics and are common to more of the operational risk events than the other characteristics. But, for any particular risk event to occur, all the characteristics in the branches leading to the risk event on the right-hand side of the tree must have occurred, ie, based on the analysis, in most instances, you need more than one characteristic to cause a risk event. It is also possible for a level 1 characteristic to show up as a level 2 or subsequent characteristic linked to another level 1 characteristic.
While it is tempting to interpret the trees resulting from the analysis as indicating path dependency, there is a significant difference between the analysis we have done and that required to demonstrate path dependency. Our methodology simply identifies the characteristics involved in operational risk events and orders the risk events into groups that exhibit common characteristics, such that there are the smallest number of branches at level 1 and at subsequent levels. It is also important to note that level 1 characteristics do not necessarily have to have occurred first, as would be required for evidence of path dependency. In addition, it should be appreciated that we are analyzing operational risk events across the Australian banking industry, and not for any one particular bank, where it may be possible to demonstrate dependency of the characteristics based on better information as to what has caused particular characteristics to arise.
2010 | 2011–12 | 2013–14 | 2010–14 | |
---|---|---|---|---|
ATM | ||||
Bank cross-selling | ||||
Complex products | ||||
Complex transaction | ||||
Computer hacking | ||||
Credit card | ||||
Crime | ||||
Derivatives | ||||
Employment issues | ||||
External fraud | ||||
Human error | ||||
Insurance | ||||
Internal fraud | ||||
International transaction | ||||
Legal issue | ||||
Manual process | ||||
Misleading information | ||||
Money laundering | ||||
Multiple people involved | ||||
Offshore fund | ||||
Overcharging | ||||
Poor controls | ||||
Regulatory failure | ||||
Single person | ||||
Software system |
2010 | 2011–12 | 2013–14 | 2010–14 | |
---|---|---|---|---|
ATM | ||||
Bank cross-selling | ||||
Complex products | ||||
Complex transaction | ||||
Computer hacking | ||||
Credit card | ||||
Crime | ||||
Derivatives | ||||
Employment issues | ||||
External fraud | ||||
Human error | ||||
Insurance | ||||
Internal fraud | ||||
International transaction | ||||
Legal issue | ||||
Manual process | ||||
Misleading information | ||||
Money laundering | ||||
Multiple people involved | ||||
Offshore fund | ||||
Overcharging | ||||
Poor controls | ||||
Regulatory failure | ||||
Single person | ||||
Software system |
2010 | 2010–11 | 2010–12 | 2010–13 | 2010–14 | |
---|---|---|---|---|---|
ATM | |||||
Bank cross-selling | |||||
Complex products | |||||
Complex transaction | |||||
Computer hacking | |||||
Credit card | |||||
Crime | |||||
Derivatives | |||||
Employment issues | |||||
External fraud | |||||
Human error | |||||
Insurance | |||||
Internal fraud | |||||
International transaction | |||||
Legal issue | |||||
Manual process | |||||
Misleading information | |||||
Money laundering | |||||
Multiple people involved | |||||
Offshore fund | |||||
Overcharging | |||||
Poor controls | |||||
Regulatory failure | |||||
Single person | |||||
Software system |
2010 | 2010–11 | 2010–12 | 2010–13 | 2010–14 | |
---|---|---|---|---|---|
ATM | |||||
Bank cross-selling | |||||
Complex products | |||||
Complex transaction | |||||
Computer hacking | |||||
Credit card | |||||
Crime | |||||
Derivatives | |||||
Employment issues | |||||
External fraud | |||||
Human error | |||||
Insurance | |||||
Internal fraud | |||||
International transaction | |||||
Legal issue | |||||
Manual process | |||||
Misleading information | |||||
Money laundering | |||||
Multiple people involved | |||||
Offshore fund | |||||
Overcharging | |||||
Poor controls | |||||
Regulatory failure | |||||
Single person | |||||
Software system |
In Tables 1–4 we summarize the results from the trees in Appendices A–D. These show the periods in which each characteristic appears as a level 1 or level 2 characteristic and illustrate the relative importance of each characteristic that we have included in the analysis.
The results indicate the following.
- •
For individual years, poor controls are a major level 1 characteristic, followed by external fraud, legal issues and single persons. Level 2 results suggest that complex products, internal fraud and regulatory failures need to be watched, as they may well emerge as major characteristics.
- •
For longer periods, where the analysis is simply adding more events without considering when they were reported, poor controls, external fraud, legal issues and single persons emerge again, as would be expected, and at level 2 complex products, internal fraud and regulatory failures appear, as well as a new emerging characteristic, multiple people. The emergence of multiple people as a level 2 characteristic is the result both of combining more events and of the algorithm used, which allowed this characteristic to emerge when it would not have done so in individual years.
- •
There are a lot of traditional characteristics that are assumed to drive operational risk events that do not appear to be very important, and this suggests they could be ignored, or at least given less attention by management.
- •
Both multiple people and single person characteristics can occur, ie, risk events are not uniquely characterized by requiring a single person to be involved or by requiring groups of people to be involved, and risk prevention needs to recognize this possibility and not concentrate on only one of these characteristics.
- •
The emergence around 2012 of complex products as a characteristic linked to poor controls indicates that enhanced controls need to be adopted when complex products are introduced. Note that while this new relationship is shown as occurring in 2012, it may have occurred previously but just not been reported until 2012.
- •
The combinations of characteristics changed across time. By way of illustration, in each year the “poor controls” level 1 characteristic combined with new characteristics to result in an increasing number of risk events. In 2010, it combined with regulatory failure and human error, resulting in three risk events. However, by 2014, if we consider both the poor controls and the combined poor controls/external fraud branches, these had combined with regulatory failure, complex transaction, money laundering, crime, internal and external fraud, multiple and single people involved, computer hacking, misleading information and an offshore fund being involved, resulting in eight significant events. This clearly illustrates the need to have improved controls in place to avoid even greater losses.
Percentage | |
of total | |
Characteristic | losses |
Poor controls | 60.1 |
Internal fraud | 27.7 |
Legal issues | 22.8 |
Regulatory issues | 16.6 |
Bank cross-selling | 11.4 |
External fraud | 9.7 |
Overcharging | 8.1 |
Crime | 7.0 |
Money laundering | 1.0 |
Misleading information | 0.4 |
Employment issues | 0.1 |
Computer hacking | 0.0 |
Human error | 0.0 |
To illustrate the overall importance of the characteristics, Table 5 shows the percentage of total operational risk event losses where the characteristic is present.
Clearly, poor control is a major characteristic of Australian banks’ operational risk events, and the analysis indicates that attention to controls would have a significant effect on reducing operational risk event losses. Alternatively, failing to improve controls in the banks is likely to result in increasing losses as new combinations of characteristics emerge.
Importantly, there is consistency across the individual years for the level 1 and level 2 characteristics, indicating their stability. This is important information for management, as it indicates the maximum effect on operational risk events could be achieved by concentrating on these characteristics.
For the “with business lines” results, there is a particularly interesting result, in that only the retail and trading and sales business lines emerge as level 1 characteristics. This suggests that
- •
just being in the retail banking or trading lines of business in Australia creates heightened exposure to operational risk events, although other characteristics are required for risk events to occur; and
- •
other lines of business are not level 1 characteristics, which is interesting, as Basel II stipulates for the “prescribed method” of determining operational risk capital for banks that lines of business are to be used.33We have carried out a similar analysis of US banks’ operational risk events, which indicates that in the United States business lines are important characteristics.
8 Applications
We have demonstrated that cladistics analysis can yield valuable information as to the characteristics of operational risk events in Australian banks. The application of this methodology within a bank may yield even more information given the greater depth of analysis possible. In particular, information regarding why particular characteristics have emerged should be available to a bank. The analysis could then be used to
- •
identify characteristics that could, with changed processes, yield cost-effective reduced operational risk losses;
- •
identify business activities that require greater compliance management, as, by their nature, these are exposed to more operational risk losses than other business activities;
- •
identify emerging characteristics that otherwise may go undetected; and
- •
provide inputs to scenario modeling for the estimation of future operational risk losses.
9 Conclusion
In this paper, we show that cladistics analysis can assist Australian banks in understanding the characteristics of their operational risk events. Our analysis based on Australian bank data shows significant stability in level 1 characteristics. It also shows that the number of level 1 characteristics is not large, making their management feasible. The methodology would also support informed scenario testing for estimation of future risk events.
Declaration of interest
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. No organization has provided any funding for this study.
Acknowledgements
The authors thank Joshua Corrigan, previously of Milliman, and Amandha Ganegoda, Operational Risk Manager at ANZ, for their comments and support during the analysis undertaken for this paper. The authors also thank Caroline Coombe from ORIC International for her support and for access to the data, and an anonymous referee who made valuable suggestions.
References
Aldridge, I. (2013). High-Frequency Trading: A Practical Guide to Algorithmic Strategies and Trading Systems, 2nd edn. Wiley.
Allan, N., and Corrigan, J. (2013). Emerging risk assessment: latest practice and innovation. Paper presented at the Actuaries Summit, Sydney (http://bit.ly/2jbCmP1).
Allan, N., Cantle, N., Godfrey, P., and Yin, Y. (2012). A review of the use of complex systems applied to risk appetite and emerging risks in ERM practice: recommendations for practical tools to help risk professionals tackle the problems of risk appetite and emerging risk. British Actuarial Journal 18(1), 163–234 (http://doi.org/bwnf).
Andriani, P. (2003). Evolutionary dynamics of industrial clusters. In Complex Systems and Evolutionary Perspectives on Organisations: The Application of Complexity Theory to Organisations, Mitleton-Kelly, E. (ed). Pergamon, Oxford.
APRA (2014). Quarterly authorised deposit-taking institution performance. Report, June, Australian Prudential Regulation Authority (http://bit.ly/2hOisvN).
Basel Committee on Banking Supervision (2006). Basel II: international convergence of capital measurement and capital standards: a revised framework – comprehensive version. Report, June, Bank for International Settlements (http://bit.ly/2h2ocBt).
Benton, M. J. (2005). Vertebrate Palaeontology. Blackwell, Oxford.
Camin, J. H., and Sokal, R. R. (1965). A method for deducing branching sequences in phylogeny. Evolution 19(2), 311–326 (http://doi.org/dbh4gm).
Carroll, L. (1871). Through the Looking Glass, and What Alice Found There. Macmillan, London.
Corrigan, E. G. (1987). Financial Market Structure: A Longer View. Report, Federal Reserve Bank of New York.
Corrigan, J. (2013). Operational risk modelling. Presentation for the Actuaries Institute (http://bit.ly/2iLvIj7).
Daníelsson, J. (2008). Blame the models. Journal of Financial Stability 4(4), 321–328 (http://doi.org/cvsvpq).
Farris, J. S. (1983). The logical basis of phylogenetic analysis. In Advances in Cladistics, Volume 2, Proceedings of the Second Meeting of the Willi Hennig Society, Platnick, N. I., and Funk, V. A. (eds), pp. 7–36. Columbia University Press, New York.
Ganegoda, A., and Evans, J. (2013). A scaling model for severity of operational losses using generalized additive models for location scale and shape (GAMLSS). Annals of Actuarial Science 7(1), 61–100 (http://doi.org/bppn).
Guckenheimer, J., and Ottino, J. M. (2008). Foundations for complex systems research in the physical sciences and engineering. Report, September, National Science Foundation Workshop (http://bit.ly/2j4XPxk).
Haldane, A. G. (2013). Rethinking the financial network. In Fragile Stabilität – stabile Fragilität, Jansen, S. A., Schröter, E., and Stehr, N. (eds). Springer Fachmedien Wiesbaden (http://doi.org/bwnh).
Hommes, C., and Wagener, F. (2009). Complex evolutionary systems in behavioral finance. In Handbook of Financial Markets: Dynamics and Evolution, Chapter 4, pp. 217–276. Elsevier (http://doi.org/dn22sz).
Hommes, C., and Wagener, F. (2010). Does eductive stability imply evolutionary stability? Journal of Economic Behavior and Organization 75(1), 25–39 (http://doi.org/dcbr2h).
Hutchinson, T. (2013). Development of electric and hybrid motorcycles. Paper presented at 4th Annual Research Conference, Industrial Doctorate Centre in Systems in Systems, University of Bristol, UK.
Lin, Y. M., Fang, S. C., and Thorne, J. L. (2007). A tabu search algorithm for maximum parsimony phylogeny inference. European Journal of Operational Research 176, 1908–1917.
Makarenkov, V., Kevorkov, D., and Legendre, P. (2006). Phylogenetic network construction approaches. Applied Mycology and Biotechnology 6, 61–97 (http://doi.org/bj7qsn).
Markose, S. M. (2005). Computability and evolutionary complexity: markets as complex adaptive systems (CAS). Economic Journal 115(504), F159–F192 (http://doi.org/bjcf4s).
McCarthy, I. P. (2005). Toward a phylogenetic reconstruction of organizational life. Journal of Bioeconomics 7(3), 271–307 (http://doi.org/fvq9q3).
Mitchell, M. (2006). Complex systems: network thinking. Artificial Intelligence 170, 1194–1212 (http://doi.org/cn9ccq).
Mitleton-Kelly, E. (ed.) (2003). Complex Systems and Evolutionary Perspectives on Organisations: The Application of Complexity Theory to Organisations. Pergamon, Oxford.
Newman, M. E. J. (2011). Resource letter CS-1: complex systems. American Journal of Physics 79(8), 800–810 (http://doi.org/bfb5ng).
Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.
To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe
You are currently unable to print this content. Please contact info@risk.net to find out more.
You are currently unable to copy this content. Please contact info@risk.net to find out more.
Copyright Infopro Digital Limited. All rights reserved.
As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (point 2.4), printing is limited to a single copy.
If you would like to purchase additional rights please email info@risk.net
Copyright Infopro Digital Limited. All rights reserved.
You may share this content using our article tools. As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (clause 2.4), an Authorised User may only make one copy of the materials for their own personal use. You must also comply with the restrictions in clause 2.5.
If you would like to purchase additional rights please email info@risk.net