Criticism of the complexity argument

Rube Goldberg Machine

Photo credit: Wikipedia.

This is the next in a series of posts about complexity and implementation. In When Things Break I presented the idea of asking  what happens when one part of a system breaks. In The Broken ATM I discussed experiences during fieldwork in rural Brazil that got me thinking along these lines. Then in Wiring Government Programs in Series or Parallel I summarized Pressman and Wildavsky’s argument that the more “clearances” a program needs the more likely it is to fail. 

The most obvious response to Pressman and Wildavsky is that complicated things work without breaking all the time. Ernest R. Alexander writes:

…when they are routine they go unnoticed: they are so much the stuff of our daily lives that we take them for granted. The product of an incredibly complex series of technological and organizational inter- dependencies is the ‘on-time’ arrivalof your flight, the completion of your telephone call, or the fact of your accident–and incident–free drive on the highway. These only surface to our consciousness in their absence: when the flight is late or our baggage is lost, when our call to Bangkok ends up in Bucaramanga, or when we find ourselves stuck in solid steel on the Santa Monica freeway.

He continues:

These examples might suggest a contrast between (usually) smoothly working technology and the ‘messy’ program implementation the Pressman-Wildavsky model is referring to; such an inference is incorrect. Relatively ‘low-tech’ activities are also the product of complex processes and systems, and their outcomes are so expected that they, too, go unnoticed.

For all its inefficiencies, the welfare system doesn’t evoke surprise when a single mother gets her AFDC check, but indignation when she does not. Other experiences which also result from implementation through multilinked and complex interdependencies, such as receiving your mail, having your garbage picked up, or filling a cup of coffee at the corner machine in the office, are noticed only when interrupted.

Recall that Pressman and Wildavsky use the example of each part working 90% or 99% of the time and how few components something needs to have before the joint probability of everything working at once drops to 50%. Alexander points out that these percentages are assumed arbitrarily. Clearly, the many components in my iPhone have to do a lot better than .99 or the thing would never work. So it’s a question of how things actually work in the world, as opposed to in a simplified, arbitrary mode. (See my post Tidy Theory vs. Messy Reality for further discussion of this idea).

Next time I’ll look at some responses that modify Pressman and Wildavsky’s approach. I also want to talk about the ideological implications of this debate.

(For a more detailed discussion of this material see p. 23-29 of the literature review to my dissertation on Brazil’s Zero Hunger Program. If the parts about the metaphor of building blocks are confusing, I present the idea on p. 21.)


Alexander, Ernest R. 1989. “Improbable Implementation: The Pressman–Wildavsky Paradox Revisited.” Journal of Public Policy 9(04): 451–65.