Excerpt fromt the Interview with Russ Roberts:
complexity cannot tolerate leverage, since no room for error. Since 1980, tripling in leverage, in U.S. and in Europe, ratio of leverage to GDP. High exposure to errors. Commodity prices, similar argument: can have rise in wheat prices in response to very small imbalance in demand, followed by rapid collapse. Not long ago talking about inflation, now talking about deflation. Speculative issue in commodity prices: how much comes from similar mistakes to Greenspan. Globalization has side effects; produces fat tails. In competition, consider two banks, one robust and one not. With globalization, everybody pushed to do outsourcing; concentration of U.S. and German firms outsourcing to, say, Bangalore. Probability of failure low because only one source of randomness. But suppose there is a problem in Bangalore--political, storms--that disrupts communication there? What happens to all these firms? French bank, lost money on a rogue trader. If you had 10 small banks each with 5 billion, wouldn't have a problem because easy to liquidate. With one bank, should a rogue trader happen, his position will be 10 times the size. Assume no linearities in execution cost. Liquidating $50 billion is more costly than 10 times liquidating $5 billion. Lumping makes you more vulnerable to large events. One large error is a lot worse than 10 smaller errors.