Risk and Robustness: A Conversation with Nassim Nicholas Taleb

In April 2013, structural problems caused the collapse of a garment factory building in Savar, Bangladesh, resulting in 1,129 dead and  2,515 injured. (Rijans/Flickr)

Why is orange juice more dangerous than terrorism? Why is suppressing volatility dangerous? And why does international relations tend toward anecdotal sensationalism? Nassim Nicholas Taleb, best-selling author of The Black Swan and Antifragile explains in this interview with Walter Kemp, IPI’s senior director for Europe and Central Asia. Mr. Taleb—a former businessman—is a philosopher and professor of Risk Engineering at New York University’s Polytechnic Institute.   

How can you predict events if, as you say, dominant risks are silent?

The type of work that IPI does has a much higher degree of unpredictability than anything else because the interactions that you look at belong to complex (rather than ordinary) systems. You might be able to predict the weather tomorrow, but you cannot predict the outcome of stock markets, wars, diplomatic negotiations, or things like that. So based on that, you have to completely change your approach. Rather than naively fooling yourself into thinking that you can predict, you should focus on building systems that can handle unpredictability. By not accepting unpredictability, you are doing something wrong.

What are some characteristics of systems that can handle shocks?  

First, systems that are optimized or lack redundancies tend to blow up. For example, a system that has a lot of debt and then faces a crisis is usually out of business. It’s like nature gave us two kidneys. The second kidney is a buffer—you don’t have to constantly predict the chances of losing the first one, because you have a back-up. So systems that have such redundancies are less vulnerable. Second, systems that are highly concentrated or centralized cannot handle unpredictability. Generally, you want the army to be centralized, but nothing else.  

But centralization is only one factor. You also have to take into account size: size is what matters.

When you have ethnic diversity in a country, and you want to integrate people, the tendency is to assimilate or eradicate the differences. If you want stability, it would be better to give people control over their own municipality. Good fences make excellent neighbors. And it also fosters peace. Just think: you have a house with a yard. You will get along with your next-door neighbor much better if you build a fence then if you share a yard—or even worse, share the house.

Then when you say that, historically, the most robust unit of governance is the city-state, what you actually mean is a small unit of governance like a municipality.

Well, it should have the scale of a city-state.

You make the caveat that it should operate under some kind of empire.

Exactly, that’s what you need. Let me explain. Take the city-states of the Mediterranean, like the city-state of Beirut where I’m from. Instead of being nation-builders, the Phoenicians thrived from trade. These were city-states, and they were stable. And they managed to get along with empires by paying protection money, like the mafia. For example, Nebuchadnezzar came down, and they settled with him. As a model, the city-state has worked in history vastly better than a nation.

Would you say that the European Union is a pretty good construct because it has a supra-state framework that enables a high degree of decentralization or subsidiarity?


Don’t you think it’s a problem when Scotland, the Basque country or Catalonia go their own way?

No, today it’s not a big deal. It would probably make them feel cosmetically better to be by themselves. If they are all part of the EU they have the same rights. They are not going to fight each other.   

So independence for some of these regions could actually have a stabilizing influence?

The most stabilizing thing is for people to feel that they control their municipality.

In that vein, why is Lebanon more stable than Syria?

First, it’s the scale. Lebanon is smaller. It has a fifth of the population of Syria, but a similar GDP. Lebanon has decentralization working in its favor. And the state is weak. It is always very good to have a weak state. Centralization by the Ba’ath party destroyed Syria. It is a very bad idea to overstabilize things. Same thing for political systems. It’s a very bad idea to overstabilize Saudi Arabia. This weakens its ability to adapt. Same thing happened in Egypt. Instead of letting it find its natural equilibrium, and just worry about the big events, now there is micro-management for the sake of stability—and this produces the opposite result. The Arab world is rife with mistakes. Sectarianism is nothing but centralization. Centralization makes one ethnic group dominate another. So decentralization actually increases rather than decreases stability.

You have identified markers of country fragility like centralization, absence of diversity, debt, and lack of political variability. If you see that there is a highly constrained system, what are the triggers that would make that country explode?

Never predict catalysts. The big mistake that historians make is to confuse catalysts for causes. Let me give you an analogy. You have a bridge that is very fragile. Eventually it is going to collapse. Why should we focus on the car or truck that could make it collapse? It is extremely unwise to try to predict an event. Instead, you should look at the underlying fragility and try to modify it and make it more stable so we don’t have to predict why and when something will collapse. Don’t try to predict when the bridge will collapse, just build a safer bridge.

Why do you say that measuring risk is “predictive and sissy”?

Because measuring risk doesn’t work.  

What about stock market analysts and investors?

There is a lot of luck in that business. Don’t be “fooled by randomness” (the title of my first nontechnical book). You can’t measure risk. If you have a million monkeys, at least one of them will be lucky. Collectively, money managers underperform the stock market. Some get lucky—but they only look good in retrospect.

Many of the things that you say are counterintuitive, for example, that volatility is a good thing.

A necessary thing, not a good thing. Any organic system needs a certain amount of volatility to operate, otherwise it becomes constrained. And this can cause a major explosion later.

Isn’t there a threshold where too much volatility becomes dangerous?

No, it’s the exact opposite. Take forest fires. If you leave a forest alone, it will have its own fires. They will come and die down, come and die down, and they will be pretty much harmless. If you repress every forest fire, the first major fire that you cannot control will destroy the whole forest. It’s the same things with states and how they solve their problems. If you have constant volatility and tension, the system can cope. Look at Switzerland—it has constant volatility and tension, but at the lowest possible level. I would rather have a lot of small problems than one large problem.  

But surely you should try to avoid having fires.

You don’t avoid having big fires by avoiding small fires. There is a big difference between volatility and risk. Volatility can actually reduce risk. Look at the stock market—there are many small variations, but few crashes. The more you try to compress small variations, the more you can cause a crash. Or if you work in a highly sterile environment, the more likely you are to get sick because your system is overstabilized and not used to dealing with deviation.

You wrote in an article in Foreign Affairs, called the “Black Swan of Cairo,” that there is no freedom without noise and no stability without volatility. At the moment, the world seems like a pretty noisy and unstable place. Is that good?

The world is not unstable. People always think there is more uncertainty and volatility today than there was yesterday. I would be careful about that. The world is certainly much more stable than it was in 1945. The number of casualties since the Second World War is much more limited than preceding periods. There is a feeling of volatility because people read the newspaper and have this naïve mapping. Statistical truth is much different than sensationalist, anecdotal journalism. The problem that I notice in your field of international relations is that it is vastly more based on anecdotal sensationalism that leads to bad policies rather than pure truth. Look at terrorism. It is a dangerous thing, but there are vastly more critical risks that we face for the health of people than terrorism. Like orange juice, stuff like that, bacteria that may be antibiotic resistant. That’s a big risk, not terrorism. But terrorism is concentrated news. If a car bomb blows up and kills four people, that is bigger news than if four different people are murdered in four different cities. Or suicides—there are more suicides than homicides in much of the world. These don’t really make the newspapers. So you have a perception of stability in one case (wrongly), and a perception of instability in another case (equally wrongly).   

Isn’t the problem low predictable high impact events? I take your point that more people probably die from shark attacks or slipping in the bath than terrorist attacks, but terrorist attacks are dangerous.

Again, remove the anecdotal and focus on the statistics. Shark attacks are less likely than car accidents. It is statistically proven. Now terrorism. It is vastly less dangerous than other types of warfare. What we really need to worry about is what can kill a large number of people. The big danger is a nuclear problem or the spread of diseases rather than terrorism. As long as terrorists don’t get ahold of a dirty bomb, they can’t really harm us. They can’t do a Stalin on us, or a Mao, who killed  tens of millions of people.

How can we strengthen social antibodies against risks? How can we reduce fragility and increase resilience?

The first thing is to let communities operate on their own—bottom up. Government shouldn’t waste time on improving small things like education. Let the communities decide themselves. When people govern their own communities, they are not tempted by war. Indeed, historically, most city-states have been peaceful. Nation-states like war, city-states like commerce, families like stability, and individuals like entertainment.

And diplomats?

The job of a diplomat is to get us out of trouble at the UN on Monday. That’s fine. But their larger mission should be to show the rest how to build systems that are mature enough that you don’t have to do that.   

How do you explain that the current international system is so bad at dealing with the crises of today?

Because we are not designed to understand our environment very well. Big states of today—like the US—have built up huge capabilities to fight armies. The best way for people to stop fighting is to do commerce together. Instead of focusing on having a big army that can fight China, they should focus on developing militias instead. States of today are not equipped to deal with a militia—someone willing to kill himself and die (groups like ISIS). Governance works much better on a small scale than a large scale. If you live in a small community, you have eye contact with people. If you work in Washington, you deal with spreadsheets. A spreadsheet doesn’t make you feel shame; going to church does. Cost over-runs are much lower in community level projects than mega-projects. No lobbyist can hijack a local system.

Today we have unprecedented access to information, and yet leaders and analysts are so bad at anticipating events. How do you explain it?

In a system, information comes accompanied with noise. But a lot of data will have a lot more noise than information. Noise rises faster than information. So thanks to our access to data and its wide availability, you get a lot of noise, and people start to talk about noise. In the old days, people had less noise, and more information. A big problem that we face today is that there is insufficient understanding of the nature of complex systems related to politics and international relations by people in government. They don’t know enough about the properties of these systems and their interaction. Government people should look at complexity theory to get insights about the nature of systems that self-manage under unpredictability.