Tail risk & incentives

The trouble with Northern Ireland’s water supply reminds me of the financial crisis. Both are examples of the dangers of tail risk.
For years, there was an under-investment in the province’s water supply. This freed up cash in the good years, but left the region vulnerable to disaster in the event of low-probability extreme weather of the sort we’ve seen.
This behaviour was like hoovering up pennies in front of a steam-roller. It yields a small return in good times, but risks catastrophe if you trip up. This is, of course, just what financial institutions did before 2008. In selling credit insurance or holding mortgage derivatives they too made steady low returns, only to blow up when a low-probability event hit.
These are not the only examples. BAA’s low spending on de-icing equipment, or BP’s under-spending on maintenance under Lord Browne, are also cases of firms taking on tail risk and suffering.
But here’s the thing. Such behaviour is not merely the result of particular ownership or market structures: banks were shareholder-owned but NI Water is state-owned; banks were in a competitive(ish) environment but BAA and NI Water are, in effect, monopolies. This raises the possibility that there are incentives to take on too much tail risk which exist across various types of ownership and market structure. There are three here:
1. Losses are socialized. In the case of banks, losses were borne not just by bankers but by tax-payers and workers made redundant by recession. In the case of BAA and NI Water, it is customers who lose consumer surplus. In BP’s case, it was workers who were killed in the Texas City explosion.
2. Positive feedback loops. Imagine there’s a strategy - be it selling insurance or cutting spending on de-icing gear - which yields £3 a year with a 98% probability but loses £150 with 2% probability. This has a negative expected value. However, there’s a two-thirds chance of it consistently paying off over 20 years. During this time, the manager who irrationally implements it will become increasingly over-confident, and his over-confidence will induce his employers to invest more faith and bonuses in him. And rival firms will wish to emulate him. The upshot will be a selection in favour of an irrational strategy.
3. Uncertainty. The above calculation assumes what often cannot be proven - that a strategy is in fact irrational. What’s more likely is that the odds are unknowable. This allows bosses to claim, as Paul does, that spending money to avoid extreme events would be irrational. When the extreme event occurs, therefore, they can argue that they were unlucky rather than incompetent. But the very possibility of being able to make this defence emboldens them to take the risk, in the knowledge that their incompetence can‘t be proven.
Because of this, we can’t say for sure that NI Water, BAA or bankers were definitely incompetent.
What we can say, though, is that there are systemic incentives to over-invest in tail risk even outside of banking.

Comments

[Richard Gadsden]

In the context of your point 1, there is also an agent/owner issue. The director of the company has much higher upside incentives than downside ones - the worst than can happen on the downside is that he loses his job and his reputation is damaged enough that he can't find a new one.

That's not exactly wiping him out, where a company that suffers as a result of this big downside can be badly damaged (e.g. BP) or even completely wiped out (Lehman Brothers).

Aligning the incentives better between the owner and the agent is a long-standing problem in modern capitalism, but I don't think you have a solution either.

[Steve]

The Minister also came on the radio saying that even pipes laid in 2008 had cracked, so perhaps it's all just down to the weather?

[dcomerf]

Weitzman's Dismal Theorem says that, as risk averse consumers, we should be almost infinitely averse to this tail risk. Backs up the principle-agent analysis above.

[InfoholicUK]

As you point out, the trouble with evaluating tail-risk is that because the events are by definition rare, it is difficult to evaluate with any accuracy either their likelihood or impact, because you have so few observation points. Your £150@2% analogy looks like a no-brainer, but what if it's actually 0.2% or 0.02%, or that loss could be £1500 or £15,000. Decision making is the easy part, its the generation of the statistics that is hard.

(Although I would have to say that, given I build Monte-Carlo risk engines for a living)

[Metatone]

I think that while there are elements of unknowability, they would be much improved if losses were not socialized. If BAA (and others) had to pay meaningful compensation to those whose travel plans were disrupted, I'd be much more sympathetic to the idea that they made a mistake with the probabilities.

As it stands, there is a massive power and risk imbalance. If I miss my flight by 30 mins because of unforeseen weather difficulties, I lose up to the cost of my ticket (depending on the exact ticket.) If they fail to run my flight for 5 hours... I get... possibly nothing and up to £5 to spend in the airport cafe.

Now I can take out insurance... but so can they... but they have no incentive to...