The Elusive Objectives of Quantitative Finance
March 6, 2010 - David Buckham CEO of Monocle
It was somewhere over the North Sea, at 30 000 feet of altitude, flying through the thick oppressive blanket of North European winter, that the question first occurred to me. I had just spent six days in one of the cold Northern European cities, or, to be more specific, six very short days and five very long nights. I had been attempting to convince one of their largest banks of the benefits of dynamic capital management. This meant in more practical terms the building of quantitatively-based predictive models for risk.
It was December 2005. Many countries were climbing towards their own infamous peaks of national bank leverage, and in the process this particular capital was beginning to be considered more expensive than Tokyo. My taxi driver had driven us to the airport in a brand new black Mercedes S600. The country and region was in a state of suspended animation, spending was euphoric against a backdrop of unstable fundamentals. In 2005 however, in the face of an ongoing property and asset price boom, it had been a difficult task to argue for more stringent risk management, and virtually impossible to argue for the benefits of regulation.
A series of meetings had ultimately led to a single crucial gap in belief systems: that the bank, and in particular certain of its executives, believed that no quantitative model would ever meet the same risk discrimination standards of an experienced banker. The branch managers in the outlying cities and towns for example, knew their clients by name; no logistic regression, no stochastic cash flow model, no Merton contingent claims hypothesis, could come close to the fact of actually knowing one’s customers personally. Of course, and occupying the polar opposite end of this belief, there was my argument. It is precisely the fact of knowing someone for three decades, I pleaded, that clouds one’s judgement. Almost at a loss, I appealed to the studies of Kahneman and Tversky, to behavioural microeconomics, to the well documented fact of mankind’s overly optimistic view of itself. From the bankers’ point of view, however, given their cultural homogeneity and high levels of trust within their society, my argument appeared absurd. It was met, in fact, with wry smiles and audible sighs of impatience.
It was a last gasp effort: I tore an A4 page into 8 equal pieces, asked each executive to write down the answer to a statistical question, collected the answers, and then demonstrated the immense range of answers received from credit risk managers, all of whom knew each other very well, each of whom had a statistical background. The question relates to the statistical likelihood of 2 people amongst the 22 players and the referee on a football pitch sharing a birthday. The answer is a little less obvious than one would expect, and as I had hoped, the range of answers was enormous. “Don’t you see,” I pressed, attempting to mask any display of self satisfaction, “the protagonist, the subject, sees the world through his or her own prism? Credit managers, even in the face of factual questions of likelihood, even when sufficiently trained, come to vastly different conclusions?”
Now, several days later, high above the roiling Atlantic, a niggling doubt crept into my thoughts. Firstly, it was hardly a fair comparison – a quirky statistical quiz question pitted against the ancient art of credit extension. Besides, I had no idea whether the exercise had actually worked. Certainly, the response had been polite, but I had no real insight into the North European mindset. Secondly, and more importantly, was I actually right? Logistic regression, and credit scoring in general, had transformed the world of retail credit extension over the past 2 decades, but, in this particular region, in the SME market, never mind in the corporate market, were these models relevant at all? Could I prove, empirically, that an Altman-type model would categorically discriminate and predict risk more successfully than an individual credit manager could in a relatively homogeneous community? And then, leaping forward several steps in complexity, and recalling the heated argument I had had with the CFO, would capital optimisation really make much difference in an environment which was growing its asset value exponentially?
Four years later, and with the benefit of hindsight, it now seems apposite, and yet perhaps a little too easy, to write off my doubts that day on the plane over the Atlantic, as unfounded. It is too easy, I think, to view my initial arguments with the bank either as tragically ignored, unheard by executives, or at the very least, pandered to through a form of politeness; to claim that hubris stood in the way of the implementation of the science of risk measurement.
For one thing, it is highly debateable whether credit models would have made any difference, since the models may simply have been ignored within the governance structure. Second, this bank was hardly the first to raise doubts over the value of quantitative finance. Virtually every bank I had spoken with had expressed doubts over the Basel equations, never mind the treatments of certain products and the capital calibrations thereof. Thirdly, and most crucially, in spite of four decades of advances in financial mathematics and economics, in spite of half a dozen Nobel prize winners, in spite of the Chairman of the Federal Reserve actually believing that the economic cycle had been reined in, the capital markets, and the financial system in general, had virtually shuddered to a standstill, against what are apparently measurable risks.
Of course, one could simply argue that the models were not at fault, it was the fact that they were ignored by senior bankers. But this is too easy – and evidence supports the fact that the models themselves failed. Not all of them, but many. In fact, the science of financial mathematics and economics came under heavy attack shortly following the crisis. The Chicago school of economics, famous Nobel prize winners, the assumption of normality, Black and Scholes, the efficient markets hypothesis, the capital asset pricing model, in fact the very notion of measuring risk quantitatively, was severely attacked by a vast range of market participants.
It was the end of finance, as one paper put it – or at the very least, it was the end of the current MBA curricula in finance. So, my niggling doubts of December 2005, doubts which at that point in time could never be uttered, were now out in the open, in an apparently violent battle between polar opposite views.
A plethora of books followed, and they were all to follow partisan lines. On the one hand, financial theory was a lie, it had misled us for a generation, it had caused the crisis, and it needed to be expunged. In its most extreme form, as one theorist argues, we should view tail risk as black swan events. In spite of the fact that no solution is offered as to how to identify these black swans (other than to make use of a sort of Zen capability of imagining things that may never happen), this theory has been adopted with much gusto, not only by the general population, but also by a range of armchair economist politicians. The theory is not helped by the polemical language which simply denigrates anything that came before, or indeed after.
On the other side of the argument, certainly over the last 18 months, the voices have been quieter, the MBA programs continue, traders still use Black-Scholes, and the regulators press on with the use of capital models.
Noticing a widening of the gap between the believers and the non-believers in financial theory and economics, it occurred to me that my initial doubts, creeping into my consciousness as early as 2005, would not be solved by reading books published by the polemical non-believers, of whom Nassim Taleb is the main protagonist. Nor, for that matter, would my doubts be assuaged by a second reading of Hull-White, Vasicek, or Markowitz. In the same way that asking senior bankers to sit around and dream up black swans seems impractical, so too does the continued use of the phrase, ‘assuming a normal distribution’, without any acknowledgement of the principal failings of the assumptions of the bell curve.
This opinion section therefore wishes to explore the question – what is the value of quantitative finance? To discover examples and cases in which it has been successful, and to expose weaknesses in the continued use of particular theories. It is not intended, however, as a scientific discussion, or even as an empirical study. Rather it is intended merely to raise our level of doubt, to make us perhaps more cicumspect in the adoption, or in the denigration, of models. Particularly, the idea is to do this in the context of financial institution regulation as it is unfolding before us.
David Buckham, CEO of Monocle