« Rambling thoughts on Canadian house prices and global savings gluts | Main | Spam attack! »


TrackBack URL for this entry:

Listed below are links to weblogs that reference Too big a percentage to fail?:


Feed You can follow this conversation by subscribing to the comment feed for this post.

These are interesting questions, but I dont think your assumptions are quite correct. For example, although the US banking system is composed of thousands of banks, the big players were the major culprits. The market shares of CITI and BofA alone account for a very very substantial share of the retail banking world (dont know the exact #, but I am pretty sure it would compare to the share of a Canadian Bank in Canada). In the investment banking world, there were only 5 major US banks, plus a handful of of foreign banks. The same is largely true in Europe. Deutsche, BNP, or Barclays, for example, all account for a very large market share in their respective countries.

Canada did for the most part escape the worst of the credit crisis as all the major banks remained well-capitalized (just a few billions in losses here and there...). But I dont think it is correct to attribute this to a particular feature of the banking system. The fact that we have 5 big banks (6 with NBC) seems to be more circumstantial. I believe that as far as we can tell so far, luck and circumstances played a more important role in safeguarding the banking system than the competitive environment.

[q]Now, how would we model those two types of risk, mathematically?[/q]

What you've proposed is known as a one-factor risk model. It can be used in the context of a Gaussian copula default model (see Hull, "Options, Futures and Other Derivatives" chap. 22), but there are several "issues" with such a model. Specifically, Gaussian copulas have tended to underestimate the risk of cascade/systemic defaults. Such events are rare enough that historical data on them is not a good way to determine correlation coefficients. You can use CDS prices to attempt to measure correlations, but as recent events have taught us, market prices might not fully reflect the probabilities of such cascades (plus, market prices only reflect probabilities in the risk-neutral measure, while financial regulators probably need to be concerned with real-world probabilities).

I think that your question is actually one of the most important in modern finance, and one of the least well understood. There appears to be a whole regime where the performance of firms is much more strongly coupled than in normal times, and any physicist will tell you that strongly coupled dynamics are extraordinarily difficult to model. Plus, we're data poor.


Sounds like monoculture risk.

This indirectly raises a related, but different, question. If all the banks comprising 50% of the market are making the same decisions the same way, why do they think there will be high returns for their actions?

From the point of view of risk assessment, you have a problem the moment you can actually justify putting all your eggs in one basket. You don't need to monitor the basket - the very existence of the basket should be evidence of a problem.

Not sure about modeling the risk... The rocket scientist quants had every incentive to model the risks and weren't able to do it. I suppose if one had information about every debt contract in the economy and a supercomputer it would be possible to find the most dangerous failure modes and try to short circuit them... Too much like 'big brother' for my taste.

Practically, why not make a rule that says banks & financial institutions can't make bets they can't afford to loose? That's how the investment banks did it when they where using their own money, and they where stable through all sorts of economic and financial upheavals from the late 19th century until they went public in the '90s (Bear in 1985). It didn't take long after they went public for the whole thing to blow-up. What changed? The people making the decisions weren't investing their own money!

Maybe I'm being too simplistic ... I dunno.

In the 1907 crisis, wasn't it private investment bankers who bailed-out the system? Not that I want to go back to the Gilded Age but my goodness, how times have changed! I didn't see Dick Fuld or Ken Lewis or John Thain pledging any of their personal wealth to fight the good fight. They just sidled up to the gov't trough and plunged their snouts into the slops without any hesitation at all.

Collectively, I bet the boards and senior executives of Bear, Merrill, GS, etc ... could easily have raised a few tens of billions, at a minimum. Maybe not enough to save the system, but at least we'd be sure of their incentives...

good point.

yes I agree, but didn't Keynes already say this? (Something about the prudent banker being imprudent in the same way as everybody else?)

reason: Keynes (and others) have noted that bankers tend to make the same mistakes at the same time. But I keep hearing how the problem is that we allow banks to get "too big to fail". My point is just that keeping banks small won't solve this problem.

Patrick: "Practically, why not make a rule that says banks & financial institutions can't make bets they can't afford to loose?" First, because they would tend to break that rule (how could we enforce it?). But second, and more importantly, because there wouldn't be any banking at all if we did succeed in enforcing that rule. There is always the possibility that all of a bank's investments will fail. We wouldn't have banks as we now understand them; we would have mutual funds. Though, that might be the way to go.

jg: yes, I wasn't altogether serious when I made the point about 6 banks being optimal. And maybe Canada just got lucky (so far). But there definitely were correlated mistakes, in the US and elsewhere. Breaking the big banks up into lots of little banks wouldn't have fixed this problem. It was a system-wide mistake and failure.

MattM: over my head, I'm afraid ;). But I was thinking more about the underlying model of the risk structure, rather than the probability distribution that results. (I'm not expressing myself clearly).

Chris S: "From the point of view of risk assessment, you have a problem the moment you can actually justify putting all your eggs in one basket. You don't need to monitor the basket - the very existence of the basket should be evidence of a problem." You lost me there, I'm afraid.

Nick: I'm not sure what you mean by "underlying model of the risk structure".

MattM: Nor am I ;). Roughly what I have in mind is that there is some distribution of characteristics F(Ci) over the banks, and some probability distribution P(Si) over the exogenous shocks, and F and P interact in some way to yield both individual-specific mistakes and population-specific mistakes. Some sort of underlying story about the decisionmakers and their information that could yield both individual and population mistakes.

I thought American and European banks were in a 'race to the bottom,' meaning competition forced them to spend their reserves of assets. If so wouldn't the number of banks in a market be meaningless on success or failure of banks? Unless number of banks creates controls competition level?

Dpends on your game-theoric lobbying assumptions. The big bank knows it is too big to fail and can gamble. The smaller banks obviously less too-big-to-fail in the aggregate but how little anti-trust behaviour do they exhibit? There are other assumptions to like whether you bail out the first little banks to fail to try to force non-panic or whether you competantly bail out the strongest remaining actors, but the big issue is do the little banks lobby competitively or co-operatively? Which is it? Depends on merger rules and anti-trust enforced regulations among other things. Maybe easiest to sketch out each scenario?

Phillip, I don't see how little banks are immune to "we can take these risks because the government will bail us out" mentality. A small bank can watch what its competitors and say "If I fail, they'll be failing too. So I have to take these risks or my bank will lose profits, and if we fail the government will be forced to bail us out."

Again, that is an example of co-operative behaviour. The small bank could also fail faster or attempt to merge to become more essential...it really depends on regulatory response and expected response. If RBC decided to load up on Iceland Market Cap and died, I doubt they'd be bailed out much, but if they were the first victim of a bank run or Cdn housing collapse, probably get some bailout, probably bailed out. Much more intricate then just assuming the USA 21st century banking experiment will replay itself as global norm (and the effects of wasting trillions or more are yet to be fully played out).

Phillip I don't see your point. What is your argument?

My point is: I need pot to live in a world of 7 billion retards.

Nick: Sorry, got lost in the snark... I didn't mean to imply that modeling was useless or futile. On the mutual fund thing: I think Nassim Taleb is advocating for extremely 'boring' banking, much less use of debt, and much more use of equity.

It seems to me that a functional financial system should not have got so thoroughly fooled. Maybe there's an information problem. The real estate market is not liquid and it hasn't been easy to short (I think there are now Case Shiller Index Futures on the CME). As far as I can tell, the only way to short it up until recently were CDS: bet that someone with big RE exposure would go belly up. But by that time, it's too late; the system has already blown-up.

"My point is: I need pot to live in a world of 7 billion retards."

lol.. comment of the year

Sorry if I'm repeating myself: in the early 1960's, mutual funds were the newest fad for the common man. Spreading the risk, how could you lose? Except that the business expansion model was based on one thing; "black gold" allowed the car culture and everything related or not to expand in an orderly manner using the existing regulatory structures.

(A recent comment on economist's view is an attempt to describe one phase of this expansion, please see - http://tinyurl.com/lb2fez -
Qoute "..The entire country went down the wrong road beginning in 1980 with Reagan and no energy policy and the pressing down of wages and unions. A lot of imbalances started to build despite a 'Great Moderation' that looks in hindsight to be just dumb luck and easily and foolishly exploited oil fields in the North Sea, Mexico and Alaska..")

Countries as "healthy" as Canada still have some time to diversify in a true manner. Countries which have overinvested in unsustainable infrastructure will continue to be a drag on the world economy. Suggestions for the proposed recovery were presented recently by the Pope, as discussed in economist's view today -
http://economistsview.typepad.com/economistsview/2009/07/the-caritas-in-veritate-social-justiice.html#more .

Nick, remember that "Money as Debt" cartoon? There is a new version that addresses some of your prior concerns. I would be very interested to read your opinions.


opal_rumours makes a couple of interesting points here, that go beyond my original post's point. Let me rephrase and expand on them:

1. If we have some sort of (say) Cournot oligoploy model, then the degree of competition increases with the number of banks, and so banks' monopoly profits will decline as the number of banks increases. This can have two effects on risks: first, a lower flow of profits may mean a lower capital ratio, and so a given loss will be more likely to cause the bank to fail; second, with a higher risk of failing anyway, due to increased competition, banks may have a bigger incentive to take on more risk, since if you do the safe thing you will probably go out of business anyway.

2. The moral hazard problem of bailouts means that banks that expect to get bailed out will take riskier bets. That's the standard moral hazard problem. Opal_rumours adds a new (to me) twist). If a bank takes an idiosyncratic risk, and loses, it knows it will not be bailed out (if it's small). But if it takes the same risk as all the others, and loses, it knows it will be bailed out ("they can't let us all fail"). So moral hazard can itself explain why banks' mistakes tend to be correlated.

pointbite: I may have a look at that video sometime, but it takes a long time to come to the important (to me) bits (so much easier to skim words than videos).

You could use tax incentives to encourage diversity. For new markets and new business lines banks could be given a first mover new tax credit in exchange for losing a % of bailout equivalent to % of business constituted by new line (new nation takeover or finance sector). The regular tax credit should reduce the amount of bailout needed in recessions; should pay for itself and give some finance business cycle smoothing.
I guess the banks could still act conspiratorily in cartelling while aucting specific tax credit for first-mover new market entry, but throw enough players in the auction and eventually some banks will want to lose bailout insurance in reutrnfor being first in land to bank in India. A stronger version of this is to give credits for banks that are already diverse (LB and NA are Que-based, BWB is Prairie, many local credit unions, part of ScotiaBank was Latin America, or already operate on diverse business cycles (gold banks?). Might want to penalize big players and credit new entrants just to encourage creative destruction.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad