Harford’s Adapt: Why Success Always Starts with Failure


Jason Collins


February 17, 2012

Natural selection operates through heritable variation in traits and differential reproductive success due to those traits. Many combinations of genes and mutations are failures, but the variation in traits creates a natural experiment in which highly evolved solutions to the environment can develop.

In his excellent book Adapt: Why Success Always Starts with Failure, Tim Harford applies this evolutionary concept to business, war, accidents and other human pursuits. How did on-the-ground experimentation lead to a better outcome in Iraq? How does Google or W.L. Gore & Associates develop new ideas? Harford’s argument is that by allowing low-level experimentation, solutions to highly complex problems are more likely to be found than through top-down decree.

Unlike much of the work in areas such as evolutionary economics, which use an evolutionary analogy to describe business activities or other social phenomena, Harford moves beyond the descriptive and asks how these processes can improve policy, reduce accidents and improve business outcomes.

For example, Harford’s encourages more government experimentation. Politicians tend towards large, sweeping plans, which can have unintended consequences and allow little opportunity for alternative approaches to be examined. If governments were more tolerant of failure (the lack of tolerance a function of the electorate as much as politicians), they could allow many options to be tried, with the best and most successful then applied on a broader scale. Harford also questions whether government should offer prizes or alternative incentive mechanisms to encourage private sector solutions where existing incentives such as patents have credibility issues.

One of my favourite sections of the book was Harford’s discussion of accidents. Most of the problems Harford examines in the book are complex and “loosely coupled”, which allows experimentation with failure. But what if the system is tightly coupled, meaning that failures threaten the survival of the entire system? This concept reminded me of work by Robert May, which undermined the belief that increased network complexity led to stability.

The concept of “normal accidents”, taken from a book of that title by Charles Perrow, is compelling. If a system is complex, things will go wrong. Safety measures that increase complexity can increase the potential for problems. As such, the question changes from “how do we stop accidents” to how do we mitigate their damage when they inevitably occur? This takes us to the concept of decoupling. When applied to the financial system, can financial institutions be decoupled from the broader system so that we can let them fail?

Climate change is also addressed, as Harford takes on the soft target of the well-meaning environmentalist. Decisions as to which options are most “environmentally friendly” are inevitably problematic as it is impossible for someone to understand the full network of cause and effect underlying their decision. In deciding which type of coffee is most environmentally friendly, how do you consider the inputs to the coffee, the cup, the building in which it you purchased it and the manner in which the barista got to work? As Harford points out, it is only through the decentralised price system that this information can be reliably provided to the consumer, while also providing incentives for the less well-meaning to change their behaviour.

Normally I am indifferent to criticisms of the well-meaning environmentalist, as the people who mock Harford’s environmentalist are often those who oppose measures to introduce a carbon price. Thankfully, Harford takes the relatively rare option of pointing out the flaws of a piecemeal approach in a complex world but providing an option to address the problem.

Harford also demonstrated his strong understanding of evolution by including the concept of survivability in his recommendations for how to implement his evolutionary approach to problems. While some people talk about evolution being for the good of the species, it is actually only good for those individuals that survive. In future generations, the survivors are the species. Drawing on experiments involving the adaptation of guppies in response to predation, Harford writes:

Adapting is not necessarily something we do. It may well be something that is done to us. We may think of ourselves as Professor Endler, but we’re actually the guppies. No individual guppy adapted, but some guppies avoided being eaten and some did not. …

As the pike cichlid closes in for a meal, it’s little consolation to the polka-dotted guppy that its failure is helping clear space for a thriving population of pebble-coloured nieces and nephews. A struggling entrepreneur is just as unlikely to be comforted by the thought that the failure of her start-up is part of a wealth-generating process of creative destruction. …

[U]nlike Amazon, or geniuses like Mitchell or Capecchi, or a pebble-coloured guppy, we don’t all get it right first time. Fortunately we have something that guppies do not: the ability to adapt as we go along.

Most of the individuals are toast. As a result, to apply the evolutionary ideas to your own life, you should experiment more, but you want to undertake experiments that you can survive. You are only one guppy.