Bhopal Revisited

S Ravi Rajan who has written extensively on Bhopal calls it a “natural laboratory” which allows one to study and understand matters such as environmental and societal violence in relation to complex, modern technology (see this essay). In particular, discussing the Bhopal gas leak, which he calls not quite unpredictable or unusual, he states and I quote:

“ …….. Yet, they are far from being freak incidents, results of a stochastic roll of the dice of history.........In light of this history, what happened on December 3, 1984, was clearly not accidental in the sense of a chance, random, unpredictable event......"

The history Ravi Rajan is referring to includes the various accidents at the Bhopal factory of Union Carbide, prior to 1984. This fact juxtaposed with the faulty plant design, departure of skilled staff and manning by under-qualified staff makes his hypothesis seem quite convincing and the accident almost inevitable.

However, closer examination suggests that only is the analysis based on the study of just one major disaster but there is a far more serious problem with it.

It is manifestly wrong (methodologically) to argue backwards from the event after it has occurred. From the knowledge of the disaster having taken place it is simple to trace it to a set of particular circumstances and underlying reasons. It is simple but it is also deeply flawed.

For what of the hundreds of instances where similar initial conditions prevailed but no accident/disaster took place?

To understand this better it is instructive to invoke the ideas of Nassim N Taleb discussed in his book The Black Swan (this is definitely the must read book of 2007 - irrespective of your interests or background. The present analysis owes a lot to this book. If unfamiliar with his ideas see 1 and 2).

A Black Swan event has three characteristics:

1. small probability of occurrence
2. large impact
3. and, retrospective predictability.

As we have seen in an earlier post the Bhopal gas leakage of 1984 was an extremely rare event with huge impact. And yet with perfect hindsight it is rationalised and explained as something not unusual or unpredictable.

However, a proper analysis ought to start with a universe of factories (or a subset of them where the pre-conditions/indicators that have been identified -– small accidents, lack of preparedness, staff issues – are present) and see how many lead to extremely, serious and huge accidents. This would help test if the hypothesis has any predictive power.

In other words rather than start from Bhopal (a known disaster site) one ought to start from factories and move forward. The analysis as done by Ravi Rajan starts with the disaster and then moves backwards to underlying causes and is thereby flawed.

A similar analysis applies in case of the warnings about the hazards from the Bhopal factory and which were largely ignored. Such warnings were and continue to be made in the case of numerous other factories; warnings which were and have been ignored by and large. But the consequences too have never materialised. Even environmentalists didn’t take the warnings seriously till the gas leakage took place in Bhopal.

So should the solitary Bhopal catastrophe lead one to conclude that the other disaster-free plants had appropriate designs, trained staff and high level of preparedness? If so it would make Bhopal even more of an outlier but more importantly it would defy all what is generally known and accepted about the conditions in Indian factories – certainly till the 1980s.

Or alternatively if the plants were not designed safely, had staffing issues and a history of small accidents then the hypothesis as propounded by S Ravi Rajan is faulty since in such factories (with certain pre-conditions identified) no major accident took place. These indicators, it would seem are poor predictors of industrial accidents and Bhopal would be unusual and unpredictable, contrary to Rajan’s thesis.

This still begs the question: what turns some to be Black Swans?

Well, the short answer is that we don’t know; nor can we know prospectively else they wouldn’t be Black Swans.

Think 9/11. In retrospect it is all too clear. But what seems predictable now was inconceivable back then.

It is also very likely that industries have learnt from Bhopal and other disasters and taken steps to avoid them. We will never hear or read about disasters avoided since they never happened!

More Black Swans?

Black Swans are rare but they do occur – though not exactly in the same way. History rhymes, it doesn’t repeat.

It is a frightening thought that though they can’t be predicted disasters will occur and perhaps with extremely devastating consequences.

Should we be worried? Yes. Can we act to stop them? No, and for reasons outlined above.

However, while we all seem to be fixated on low probability events with huge consequences we tend to ignore the more probable ones with smaller distributed impacts.

So while we focus on Bhopal, we ignore the everyday small accidents in factories. We worry about airplane crashes (though adjusted for passenger kilometres air travel may be safer than road travel) while everyday 250 persons die in road accidents in India. To put that number in perspective it is nearly 100,000 people every year. While we are concerned about terrorism (less than 5,000 deaths every year on average in India), millions of infants will not see their first birthday and tens of thousands of women die in childbirth every year.

Not only do these seemingly small matters have more bearing on our lives but they are also far more but amenable to corrective action.

(this post may be read along with the one below)