Second Order Effects
knock-on effects · unintended consequences
The consequences of the consequences. A first-order effect is what an action does; second-order effects are what the system does in response, usually later, often in the opposite direction, and rarely attributed to the action that caused them.
In practice
Cut the price to win the quarter. The quarter is won. The buyers who signed at the discount set the reference price, the renewal is negotiated from it, and the sales team has learned that discounting is how a quarter gets won.
The common mistake
Stopping at the first order because the first order is the part that is measured. The delay does the damage: by the time the second-order effect arrives, the decision that produced it is no longer a candidate explanation for it.
Every action lands in something that reacts. The first-order effect is the intended one and arrives on schedule; the second-order effects are the system's response to having been acted on, and they arrive late enough that nobody connects them to the cause.
The canonical statements
Bastiat (1850) gave the structure its first clear statement: the good economist accounts for what is not seen. Robert Merton (1936) made it a research programme, setting out the sources of unanticipated consequences of purposive social actionMerton, R. K. (1936). 'The Unanticipated Consequences of Purposive Social Action.' American Sociological Review 1(6). Merton's sources include ignorance, error, the 'imperious immediacy of interest' in which the actor's focus on the near-term outcome excludes consideration of further ones, and self-defeating prediction.. His second source is the one that recurs in organisations: attention to the immediate result actively excludes consideration of later ones, so the failure is structural rather than a lapse.
Jay Forrester (1971) reached the same conclusion from modelling rather than sociology — complex systems present obvious intervention points that are reliably the wrong ones, and absorb well-aimed policy. The concrete version is the cobra effect, named by Horst Siebert (2001) after a colonial bounty on cobra skins that produced cobra farms, and an identical structure appears wherever a payment is attached to a proxy. Goodhart's law is that result stated as a law of measurement; Campbell (1976) published it independently for social indicators in the same decade.
Why delay is the mechanism
Second-order effects would be manageable if they were merely indirect. What makes them reliably invisible is that they are separated from their cause in time. Cutting marketing raises margin this quarter and empties the pipeline two quarters later; by then the cut is not on the list of suspects, and the decision looks like it worked for long enough to be repeated. In feedback terms the first-order effect travels along a short path and the second along a long one, and the organisation's reporting cycle is usually shorter than the long path — so it is, structurally, unable to see them.
The objection, which is a strong one
Albert Hirschman (1991) identified the appeal to second-order effects as one of three standard rhetorical forms used to oppose any reform, and named it the perversity thesisHirschman, A. O. (1991). The Rhetoric of Reaction: Perversity, Futility, Jeopardy. Harvard University Press. The perversity thesis holds that an intervention will produce the exact opposite of its stated aim. Hirschman's point is not that it is always false but that it is asserted far more often than it is demonstrated, and that its rhetorical attraction is independent of its truth.: the claim that an action will produce the reverse of its intent. Hirschman's charge is not that perverse effects never occur but that the claim is unfalsifiable in advance, available against every proposal, and satisfying to make. Anyone can generate a plausible second-order story for any change, and generating one is not analysis.
The discipline this imposes is specific. A second-order argument is worth something when it names the mechanism, the party whose incentive changes, and roughly when the effect should appear — which makes it checkable. Without those three it is an objection dressed as foresight, and it is the cheapest objection available.
What it rules out
It rules out evaluating a decision on the reporting cycle that produced it, since the relevant effects arrive after it closes. It rules out attaching a payment to a proxy and expecting the proxy to keep measuring what it measured. And it rules out 'it worked' as a conclusion when the interval observed is shorter than the loop being acted on.
It does not rule out acting. Hirschman's warning is the necessary counterweight: an unlimited requirement to anticipate consequences is a general argument for doing nothing, and that is a decision with second-order effects of its own.
Sources
Bastiat, F. (1850). Ce qu'on voit et ce qu'on ne voit pas. · Campbell, D. T. (1976). 'Assessing the Impact of Planned Social Change.' Occasional Paper Series, Dartmouth. · Forrester, J. W. (1971). 'Counterintuitive Behavior of Social Systems.' Technology Review 73(3). · Hirschman, A. O. (1991). The Rhetoric of Reaction. Harvard University Press. · Merton, R. K. (1936). 'The Unanticipated Consequences of Purposive Social Action.' American Sociological Review 1(6). · Siebert, H. (2001). Der Kobra-Effekt. Deutsche Verlags-Anstalt.
Concept web
Open the full webQuestions
What are second-order effects?
The consequences of the consequences — what a system does in response to being acted on, as opposed to what the action directly did. They typically arrive later than the first-order effect and often run in the opposite direction.
Why are second-order effects so easy to miss?
Because they are separated from their cause in time. By the time the effect arrives, the decision is no longer a candidate explanation, so it looks successful long enough to be repeated. Merton (1936) called the underlying cause the imperious immediacy of interest.
What is the cobra effect?
A bounty on cobra skins intended to reduce cobras instead produced cobra farming. Named by Horst Siebert (2001), it is the standard illustration of an incentive attached to a proxy producing the reverse of its aim — the same structure as Goodhart's law.
What is wrong with second-order reasoning?
It is unfalsifiable in advance and available against every proposal. Hirschman (1991) called this the perversity thesis and showed it functions as rhetoric more often than analysis. A second-order claim earns attention only when it names the mechanism, whose incentive changes, and when the effect should appear.