Phlogiston risk management series - How to get digital? Episode 1: information is not gold

In this new series of articles, I want to explore how current risk management practices are dealing, or rather struggling, with the digitalisation of society. My thesis is that risk management [with the financial sector as a case study[1] is based on old – pre-digitalisation - methods, techniques and knowledge principles. Even today, the direction is still focused on increasing objectivity in risk management, as if we are taking care of gold, instead of taking care of information. See the IFRS9 example below. Later, we will dig into the principal differences between gold and information.

In fact, the series is a challenge to make Philosophy meet Finance. That means, to raise the application of philosophy in, to increase the philosophising about financial economy. Philosophy and Finance have never been close, money stinks and is not worth our considerations versus contemplation is a waste of time and money. But if philosophers want to reflect on what is important in order to be relevant, and if bankers want to do the right things right in order to serve society, they'd better help each other, cause there is much to gain on both sides.

From my experience -a philosopher working in finance- I expect gains especially by applying philosophy in our thinking about risk, data, expertise, objectivity and centralisation. In this series, we will philosophise about these subjects.

Phlogiston in the title of this series sounds like something sophisticated you may want to have. The term is borrowed from Watson (1981), who provided the term “phlogiston theory of risk”. With this term, he[2]”caricatures objective risk as a unique substance, given off by a physical process, and at a rate which can be determined precisely by risk assessment.” With this term, Watson refers to risk managers who believe that risk is objective, and can be measured as if we measure the amount of water flowing from a tap, regardless of its context or social impact that may trigger responses to it.

This has large consequences. Since it can be quantitatively measured, it can be digitalised which allows the computerised processing of relevant data. And we can use statistics to model the risk objectively. Given that statistics is the Law of Large Numbers, we apply statistics preferably to the largest population, which creates incentives for standardisation and centralisation.

Although ideal for risk management and still widely used, the point made in this series is that phlogiston does not exist!

IFRS9 aims for objectivity, ends in expertise

IFRS9 is a recent example of a failed initiative to increase objectivity.

IFRS9 was implemented in the follow up of the credit crisis in 2008. The idea was that previously, management had too much discretion to determine credit impairments, hence credit provisions (estimated losses). This caused banks to add to the provisions in good times, and to subtract from provisions in bad times, smoothening credit losses (and profit) over time, thus showing stability.

Within the IFRS9 rules, banks have to build models to estimate credit losses objectively, properly indicating the state of their portfolio with adequate provisions instead of prudent or conservative provisions in good times (underestimating the value of banks) and too low provisions in bad times. Within IFRS9, impairments should be accurate, spot on. The models that banks were required to build would make the risk parameters point in time (adequate for the phase of the economic cycle), forward looking, and multi year meaning using economic forecasts for exposures with a horizon up to 30 years, using all data available at undue costs.

However, Noble prize winners hold that any economic prediction beyond the half year horizon is inaccurate. 30 years predictions are nonsense. From history, it is known that credit cycles can take 7 to 10 years, or even longer, requiring data from many years (>2 cycles) to become somewhat reliable. This means that these models will not be reliable the first decade after implementation in this set up. By lack of data, implementation of such empirical data driven models is purely expert based!

Finally, with the Covid 19 crisis, banks noticed that their historically oriented models were not build for these extreme economic conditions[3], rendering outcomes that were deemed inaccurate by management. If models are only allowed when build upon historical data, in fact they are principally unfit to make predictions in a changing world.

The result of the IFRS9 framework is that in the past year, for many banks, it was management (expert based) who decided with high discretion on the majority of the provisions with Top Level Adjustments. Despite what the models said, or better, correcting the models. The models were not able to cope with new situations (i.e. risk).

Changing ontology requires changing methodology

In an earlier thesis in the philosophy of financial economy[4] in 2006, it was shown that the virtual turn (now known as digitalisation) is a fact, is happening now, and requires one of the largest paradigm shifts to deal with the virtual world, that behaves different than the physical world we were used to. For example, when taking into account the recent highly increased dynamics and uncertainties, we need to focus on exchanges, dynamics and flows, instead of objects, stability and facts, respectively.

From the field of philosophy, we know that methods and techniques to capture reality must be aligned to the prevailing epistemology and ontology of what we want to capture, or in simple words, when the world around us changes, we need to rethink the methods and techniques we use to deal with it.

The rapid rise of applications of Artificial Intelligence and Machine Learning, in a growing number of fields, can be regarded as an accelerator for the further digitalisation of society. That increases the need for this paradigm shift further.

But in my recent experiences, the observation is usually that the direction is opposite to what is required. Instead of embracing the principles of the virtual world, we try harder every time to conquer it with the (old) principles of the physical world, treating information as gold.

Information is not gold

The below table summarises some principal differences between information and gold.

Information really is different! For example, on the Internet, all times and places are connected, hence annihilated according to the classic meaning of time and place as bounded periods or areas. Castells[5] speaks of Timeless time and Space of flows, referring to another logic of space and time. In timeless time, all times are mixed, that is, history is as much available as is the present or the future. Humans can make comparisons between times, by way of reviewing the representations of these times, without interference of physical reality, without real resistance. In space of flows, locations become characterised and determined by the flows they are connected to. It is not relevant where your hotel is, as soon as global business men visit it, it will adhere to global norms and values. Adherence to the protocols or codes of specific flows is a prerequisite to be connected. In a binary access, your pin code is not halfway right; it is either 1 or 0.

Risk is subjective

30 years ago, Pidgeon[6] studied the perception of risk from a social science perspective. p.89: “From the perspective of the social sciences, risk perception involves people’s beliefs, attitudes, judgements and feelings, as well as the wider social or cultural values and dispositions that people adopt, towards hazards and their benefits.”

Pidgeon et al. conclude that risk perception cannot be reduced to a single subjective correlate of a particular mathematical model of risk, such as the product of probabilities and consequences, because this imposes unduly restrictive assumptions about what is an essentially human and social phenomenon.

In my own words, this means that a local credit analyst, aware of local circumstances and developments, possibly having met clients physically, reading the local newspapers and member of several local social networks, knows so much more about the credit risks in his portfolio of local clients than can ever be quantified and centrally processed and managed. When information has a relationship with risk, it is clear from this example that risk being centrally managed is higher than risk overseen and managed locally (with the feet in the mud). Risk is subjective to the (be)holder of it.

In the current Capital Regulations for banks, even after the Credit Crisis of 2008, focus still is on determining objective probabilities and severities or consequences (of default). This approach assumes that reality is sufficiently registered in data of good quality and availability regarding risks that can be objectively determined by perfect models, ignoring aspects that cannot be quantified, human expertise and interventions. Models based on algorithmic rationality, and many assumptions, can take full control. This does provide a comfortable basis for regulation of banks, not requiring any other expertise than statistics, but it does not reflect reality. And assumes phlogiston risk concepts.

Next steps

In line with the new laws of information society, sharing information and knowledge is multiplying it, increases its value. Only in interactions, awareness is created. I can do many things, but offer the perspective of the other (as the philosopher Levinas described so well). That is what I seek and ask of you when offering my knowledge: interact and provide feedback to make progress together!

In this series, I want to take some further steps to explore whether not only the perception of risk is subjective, but also the knowledge of risk (Epistemology), and even the nature of risk itself (Ontology) should be shifted to account for digitalisation. For example, by early warning systems, even the ontology of risk shifts due to digitalisation, since risk as uncertainty is the inverse of information. Early warning systems provide information about future states to enable changing the future. Information is then reflexive, turning back on itself. Different from gold.

New ways of information generation and processing then create new risks and change current risks.

How can we manage the value of digital information, recognising its differences compared to what we have known for thousands of years?

Towards a new episode ...

[1] Choose your own grounds for your wars, Sun Tzu [2] Pidgeon, N.; Hood, C.; Jones, D.; Turner, B.; Gibson, R., in of Risk Analysis, perception and management, Chapter 5, Risk perception, The Royal Society, London, 1992, p.94. [3] which in the Netherlands did not turn out so extreme, economically. [4] A. Kukler, The EC movement, or the quantitative turn in banking, A case study of the virtual turn in society, Febo Druk BV, 2006 [5] Castells, M., The rise of the network society, 2nd edition, volume 1 from The information Age: economy, society and culture, Blackwell Publishing, 2003 [6] Pidgeon, N.; Hood, C.; Jones, D.; Turner, B.; Gibson, R., in of Risk Analysis, perception and management, Chapter 5, Risk perception, The Royal Society, London, 1992