I won't pretend I have great hopes for the upcoming G20 summit end of September. I struggled all summer to write anything on this blog, as it seemed we were slowly drifting to the "business as usual, the crisis was a healthy consolidation, there is nothing wrong with the system after all, let's just mend a few tiny details so we can say to voters 'we acted on ...' and everything will be fine." This was so at odd with what's required that it did not seem worth responding.

And yet the G20 is reconvening soon. Among the ideas to control (limit) risk-taking by the banks are mainly: - increasing capital, - decreasing leverage, - limiting variable compensations. Let's address the question of capital here.

The obvious first: increasing capital might sound equivalent to decreasing leverage but is not. The way the required capital is assessed heavily depends on the riskiness of the assets and liabilities of the bank. The riskier, the more capitalised. This is the whole idea of the Basel II agreements (and their predecessors). And while this makes perfect sense, the 2007 crisis showed the weakness of the approach: what happens when you under-estimate the risks? You don't take enough capital and, when things go bad, they actually quickly turn from bad to critical. Increasing capital requirements might in practise push bankers to focus on the less risky products. This sounds a good idea but the same weakness persists: less risky according to who? Increases in capital will in practise push bankers to favour products for which the risk might be the most under-estimated... preparing the nice next "black swan" (in which we'll again hear of "25-standard-devations events").
Limiting leverage is an important complement to increasing capital requirements for this reason. No matter how safe an investment appears, it never is perfectly safe. It thus makes sense to limit how much you can borrow only to make such investment. In short, making sure you invest money you have, or at least money you likely can redeem.

Increasing capital might favour future "black swan", under-estimated events and it might also push bankers to take more risks.

The pro-cyclical character of capital requirements has been discussed at length. Let's assume that rules for anti-cyclical reserves can be agreed on by the G20: when a crisis occurs, banks can use their cushions of capital and, when things go smoothly, banks have to replenish the cushion. Basically, the capital is equated with a cushion of cash.

The replenishment of the cushion will wait on banks' profits at times when other industries will have a nice ride. Banks will be the under-performing assets, by rule. I don't think I need to write much more. Clearly, such a status will challenge bankers and push them to take any imaginable risks they can to compensate. Increasing risks will lead to even more capital being legally required, you might say... but I just addressed this: the way to avoid this is to invest in under-estimated risks. It always is the same pattern: products for which bad outcomes might theoretically happen but on conditions that never historically did occur (never or extremely rarely). Bankers will be pushed to take "invisible" risks, and to create a bubble: when the risk appears (and capital is required), quickly take more new invisible risks! This will only push bankers to increase risks faster than the regulatory increase in capital... Like any good bubble, it may well work for years or even decades, and then... well... it will be handled by another government. 

Not only psychology but also poorly-mastered science are hindrances in such a situation. Humans think by analogy. When there is no analogy to be made (it never happened), an event is sure to be misevaluated (good or bad, but hopeful natures favour the positive-outcome interpretation). Moreover, statistics are regularly badly used in such context: with "extreme events", it is easy to badly estimate probabilities. This doesn't have to be a problem, as long as one remains aware of the uncertainty. Most people misusing statistics fall in the last trap: they do estimate (properly) "this event historically had a probability of 0.1% of occurrence per year" but they forget that, given the size of the sample they use, this is more like "between 0.001% and 0.2%". They easily miss a factor 2 (or more). Now, imagine that when luck turns, you lose what you have; that will be uncomfortable but you can still hope for a new beginning. Imagine that you forgot a factor 2, and lose not only the savings of your lifetime, but lose them twice. What's the hope left? Statistics are also often biased by conditionality. It may be possible to measure "historically, the probability is 1%" (let's assume the sample is large enough that this is a precise, fair, measure). It is very easy to forget "conditioned by the fact the economy was mostly driven by production of goods and not by services for 95% of the observed period." I already insisted on this earlier: the past does not predict the future. The economy evolves. Can you really draw conclusions by analogy to 1929? To a unique event? To a time when the economic activity and the social fabric of society were totally different from today's? Or can you draw conclusion by analogy to the petroleum crisis of the 70's, while our electricity is now mostly nuclear? If you can, this is great, because it also means the probability of a nuclear accident is null... I mean, there never was a nuclear explosion before the 40's... For millennia, not one.

Let's finish with some fun: let's assume the politicians can not agree on new capital rules but they can do so in a way that makes it impossible for bankers to run the risk bubble I just described... Like, really, banks are guaranteed to be under-performing assets during economic recoveries. As an investor, how likely are you to keep your shares in banks (when, like now, we might be at the start of a period of economic recovery)? Pretty small. You're likely to invest in banks when the economic cycle has matured and you believe it might consolidate (because then banks can release capital so they can maintain their profitability while others will struggle). Sounds logical? The consequences of everybody thinking this way though is likely to be as follows: at the beginning of recovery, bank share prices go down (everybody's selling) until the price is deflated enough that mergers and acquisitions in the financial sector happen. But who would buy a bank at the precise time it will need more capital? Only bankers, already suffering the problem, thinking they know how to handle it, thinking that's their business model... In short, the likely outcome is fewer but bigger banks. Does this still sound a good idea? I would have thought we all agreed that "too big to fail" is not a good idea! But OK, this is a period of economic recovery so banks will be big but will probably be more than surviving. Reach the end of the economic cycle. Now people buy bank shares, as a "defensive" sector. Did you say bubble in prices? Just banks "too big to fail" in a recession? We need to be seriously mad if we believe we can even draft capital rules that could prevent bankruptcy for sure. And if this was the case, the profitability of banks would be so low that banks would simply be government agencies... Why not? but I doubt lobbyists and investors will let this happen. The G20 will not fix this, they cannot. After all, Goldman Sachs is so profitable, it shows the private sector works fine. Does it not?