Some say the root cause of the global financial crisis was a few regional financiers selling risky mortgages to poor people. How can that be? The subprime mortgage market is a fraction of the U.S. mortgage market, which is a fraction of the U.S. credit market, which is a fraction of the global credit market. How can defaults in a sub-sub-submarket destroy banks on two continents and send several countries to the brink of bankruptcy?

In his landmark analysis of the Three Mile Island nuclear accident, sociologist Charles Perrow argued that that meltdown was not caused by any particular component or operator failure. Rather, it was caused by a number of small component failures that interacted in unpredicted ways. Those failures ranged from a pressure-relief valve that didn’t reseat properly to a meter that gave deceptive information to some perceptual errors by operators.

If a system has high interactive complexity, then by definition it is hard to predict the impact of a particular action on other elements within the system. If the system is tightly coupled, to use Perrow’s terminology, an action at one point propagates rapidly throughout. Perrow argued that systems with high interactive complexity and tight coupling are more prone to systems accidents.

The financial sector meltdown in the fall of 2008 was a systems accident. However, rather than failures interacting, the byproducts of innovations interacted. With the possible exception of some regulatory decisions — and the definite exception of illegal activities — many of the factors leading up to the financial crisis involved innovations to the system, rather than existing system components that failed. Within the financial industry’s system of rewards and values, people innovated to improve their job performance. One group innovated by creating the subprime mortgage category; another by finding a way to lend people down payments; a third by working out how to turn mortgages into securities; a fourth by figuring out how to price those securities. The list goes on.

Innovations alter the systems in which they are embedded in three ways. First, wherever innovation occurs, the relationship with the rest of the system changes. In this case, the innovations created mismatches between the actual financial system and the financial system the regulators thought they were regulating. It is the interactions of these mismatches that caused the systems accident. Second, some innovations increase the complexity of the system. Third, some innovations tighten the system’s coupling, increasing the probability