The D-Squared Digest One Minute MBA Part 2: Managing the Risk of Getting Killed
Latest in an occasional series, applying the knowledge I gleaned from my business school education to various important problems of the day. I realise that these MBA posts wind some people up, but like most of the best stuff they teach you at business school, it�s four parts applied economics to three parts common sense. Special note for people who �hate MBAs� � I don�t actually have an MBA as my actual business school qualification was the London Business School�s Masters in Finance program � quite like an MBA but with more of a focus on financial markets and substantially cheaper. Anyway, onward with the basic MBA principles. This time round, I�m producing some general principles for dealing with a terrorist threat. In particular, the question of shooting people on the streets of London.
The optimal frequency of disasters is not zero. This graceful formulation is due to Prof. Richard Portes, who used to say it about emerging market financial crises. However, it�s a fundamental principle of risk management and one of entirely general application. Most dangers can be absolutely eliminated for all practical purposes, but only at unacceptable cost. This is true whether you�re thinking about �inconveniencing� people in the name of security (note here that the word �inconveniencing� is being used in the current newspaper sense as a portmanteau term for having to put up with having your rucksack searched, and allowing the police to detain you without charge for up to three months) or trying to design a rule of engagement for armed police which will avoid their shooting innocent people. If you�re trying to bring the risk down to zero, then you have almost certainly over-engineered. So you should design the system to leave some positive risk. Risk, by the way, is the risk that something very bad will happen; the fact that ex post something very bad did happen is not good evidence that ex ante the risk tradeoff made was the wrong one, nor is it evidence that the tradeoff needs to be changed going forward. Having said that, acceptance of the statistical inevitability of bad events needs to be tempered with another important principle:
By and large, you get the error rate that you are prepared to tolerate. The most usual case where you get taught this one is in photo developing shops, where the rate of defective prints can differ wildly between superficially identical units. The lesson learned was that, although mistakes are inevitable in any process involving people, the way that you deal with mistakes makes a big difference. If you wave it off as �no biggie�, then you will get more and more mistakes, and when it comes time to do something about it, it will be much harder because your employees have become conditioned to having their error rates accepted. On the other hand, if you make a big fuss about every single mistake and constantly work to understand how it was caused, then firstly, your processes will improve over time, and secondly the simple fact of having to file the report will encourage people to concentrate and make fewer mistakes. This is the basis of six-sigma reliability programs.
Application. It is an entirely salutary thing that the British police force has, as a general principle, an inquiry by an outside Force into every single case of death caused by police officers. On the other hand, it is entirely the opposite of salutary for the Commissioner of the Metropolitan Police to stand up at a press conference when his force have killed an innocent man and say �this sort of thing is bound to happen again and again�. While this may be true, factually, it is entirely wrong for the man whose job it is to make sure it doesn�t happen again to say so. When a surgeon chops the wrong leg off, you do not expect the hospital to say �well this is bound to happen and there�s nothing anyone can do about it� and the police should not hold themselves to a lower standard.
It is usually incorrect to believe that you are on the efficient frontier. This is a specifically business-school piece of wisdom, and one of the big points of departure between business school types and economists. Economists almost always think about things in terms of tradeoffs; more X means less Y. This is because economists think of things in terms of abstract idealised production functions with substitutable inputs and outputs. If you�re actually teaching people to manage factories, however, then you are thinking about concrete processes (particularly if it is a concrete factory) which are usually about as far from idealised as you can get. Unless your process is at the absolute bleeding edge of what is possible, tradeoffs are not necessarily the correct way to think about making improvements to it. Furthermore, if you have a complicated and non-optimised production process, it is not at all guaranteed that the tradeoff which the frontier implies is there at all; think about someone trying to get a complicated piece of software written who decides to trade off time against cost by adding more programmers to an already late project.
Application. Talking about the �tradeoff� between the number of innocent people we kill versus the number of lives we save by killing guilty people, is probably misplaced. Most of what we want to do at present to save innocent people involves gathering better intelligence. And this would make us less likely, not more, to kill innocent people. Conversely, it is not at all obvious, given our current state of information, that encouraging Met officers to be even quicker on the trigger than they already are, would make us any safer at all. Since we are looking for less than a dozen people in a city of seven million, a simple Bayesian calculation suggests that until the state of intelligence reaches some threshold level, a random selection from the pool of suspicious people is more likely to be innocent than guilty (and even if guilty, is more likely than not to not actually have a bomb on him at the time, reducing the potential benefit even more.
Ambiguous instructions infallibly generate errors. This is a lot of the reason why so much �management-speak� is so pedantic and goes so humourously out of its way to state the obvious. (Scott Adams has made a career out of this). When you�re describing a process, you state the obvious so that everyone knows the same obvious (this principle was copied from the military, who are also keen on stating the obvious), and in order to minimise the number of �exceptions� � cases which do not fit into any of the categories covered by rules. When faced with exceptions, particularly under stress conditions, people often either freeze (failing to put the case into any category, leading to inaction when action is needed) or panic (place the case into an inappropriate category, usually leading to excessive action).
Application. The inquiry will make this clear, but it certainly looks as if the Menezes shooting was an �exception� in this sense for both the police officers and Mr de Menezes. A police officer who is unclear about the rules of engagement could quite understandably decide to err on the side of preventing a suicide bomb. A Brazilian electrician living in a quite dangerous part of London could quite understandably decide that running away from armed men was a good idea if he didn�t know that we had a shoot-to-kill policy. And it is very worrying indeed to me that nobody, including anybody briefing the media at the Met, seems to know when it was that this policy was introduced.
Depending on who you listen to, the new-style �shoot to hope to kill to protect� policy was brought in as part of a top-secret operation in early 2002 as part of �Operation Kratos� by Lord Stevens. Or possibly introduced in 2003. Or possibly in ACPO guidance following the second Stanley inquest. There are also suggestions (can�t find on internet but clearly remember from broadcast news) that the armed officers deployed on regular patrol duty after 7/7 were given a whole new set of instructions. It appears that the UK has had at least three shoot-to-kill policies, above and beyond the standard understanding that guns are for shooting and shooting carries a danger of death. In fact we’ve had so many shoot-to-kill policies that it’s rather surprising the news didn’t leak out to Joe Public. I think any inquiry should very certainly be looking closely at whether, somewhere between Lord Stevens� standing orders, the 2003 terrorism changes and the 2004 ACPO note, there is a single coherent and current set of rules of engagement and whether these rules were communicated effectively to the officers on the ground. If they weren�t, then someone at a desk somewhere is guilty of something that looks to me to be not entirely unlike negligent homicide.
In any case, when one is told at a press conference that �we have had a shoot-to-kill on suspicion policy for the last few years�, the correct response is not �oh really? thanks for keeping us in the loop�. The kind of policy that did for Mr de Menezes is not a small change to the general arrangement of things in the UK. It is the sort of thing that ought to happen in the House of Commons; not necessarily through primary legislation, but at the very least the Home Secretary should have made the announcement and allowed for debate. It is just not good enough to have something like this done through secret administrative order; I would say the same about changes to the planning guidelines for supermarkets and this is more important.
Anticipated events do not change the information set. This is one of the cornerstones of efficient markets theory but again, is of much more general application. If you expect something to happen then the fact that it has, in fact, happened, is not new news to you. Since we had all expected that London was going to be bombed, sooner or later, it is clearly wrong to say that �everything has changed� as a result of the bombing. If it is a good idea now to pass laws against �glorifying terrorism� and to allow the police to hold terrorist subjects for three months without charge (it isn�t) then it was a good idea three weeks ago (it wasn�t). This makes it slightly more heartening that we introduced our shoot-to-kill policy a long time ago, on the basis of proper planning, although I suppose that at least if we�d introduced it in a panic this week we would have known that we were doing it.
And that�s about it. That will be $18,000 please.