Judgment Playbook

Second-order thinking

Keep asking: and then what?

When you need itThe first effect is obvious; the reaction to it is not5 minute readtaught in course chapter 7
AlternativesManage uncertainty

What it is

A shop cuts its prices, and sales go up. Everyone sees that first effect. Later, customers learn to wait for the next discount, and the profit on each sale shrinks. These reactions come after the decision is made, and they can undo the gain.

Second-order thinking means following a decision past its first effect. You ask what it does at once, how the people it touches will respond, and what their response does to your goal.

Where it comes from

The idea has no single author. The French economist Frédéric Bastiat put it clearly in 1850. His pamphlet was called Ce qu'on voit et ce qu'on ne voit pas ("What is seen and what is not seen"). He wrote: "an act, a habit, an institution, a law produces not only one effect, but a series of effects. Of these effects, the first alone is immediate; it appears simultaneously with its cause; it is seen. The other effects emerge only subsequently; they are not seen."

In 1936 the sociologist Robert Merton named the problem in the American Sociological Review: "The Unanticipated Consequences of Purposive Social Action". He meant what people did not foresee when they acted on purpose. He blamed ignorance, mistakes and caring only about the result you want now.

In 1971 Jay Forrester, who studied complex systems, explained why it happens, in "Counterintuitive Behavior of Social Systems". Systems made of people run on many feedback loops, where an effect comes back round and changes its own cause. We are used to short chains of cause and effect, so we misread such systems.

Business readers mostly know it as "second-level thinking", from chapter 1 of The Most Important Thing (2011) by the investor Howard Marks. In Principles (2017) another investor, Ray Dalio, calls it weighing second- and third-order consequences.

What it corrects

Say you plan to cut a price or change how you sell. You can measure the direct effect now, but not the reactions. A good analyst builds the argument on what can be measured, because that is where the data is. And the analyst usually gets that first step right. Working harder does not help: the numbers get more exact, but they still describe only that step. The reactions drop out of the forecast for two reasons. The people who react are outsiders, and they act in their own interest, not yours. And they react late, after the decision has been reviewed and called a success.

How it works

  1. Write the direct effect in one sentence. Name the number you expect it to move.
  2. List who will feel that effect: customers, competitors, distributors, staff, regulators.
  3. For each of them, say what is now in their interest to do.
  4. Take the strongest response and follow it one step further: what does it do to your number?
  5. Say how long the reaction takes to reach your number: weeks, quarters or years.
  6. Check that your recommendation holds once the reaction hits your number. Name one early signal of the reaction from outside your own systems.

Worked example

On 20 November 1989 Mexico City introduced Hoy No Circula ("No driving today"). Each car had to stay off the road one weekday a week. The last digit of its number plate set the day. The rules covered 2.3 million vehicles, so about 460,000 stayed off on any weekday. Enforcement was strict, and almost everyone obeyed. The second-order reaction was just as simple as the rule. A family with two cars whose plates end in different digits could drive every day.

In 2008 the economist Lucas Davis studied the result in the Journal of Political Economy. He used readings from air-monitoring stations. He checked every pollutant in several ways, and found no sign that the air got cleaner. The number of registered vehicles rose by about 325,000 when the programme started (probably between 51,000 and 597,000). New car sales rose about 15%, but that rise came to under 2% of all the cars on the road. So nearly all the extra cars were used ones, from elsewhere in Mexico and the United States. Older cars pollute more. Metro and bus use did not rise, and petrol sales did not fall. The ban kept each car off the road on its day, but the extra cars cancelled the gain.

In a Business Case Weekly case

In the Nike case, the decision in June 2020 is about a plan called Consumer Direct Acceleration. Nike would drop many wholesale partners, the shops that resell its goods, and sell straight to customers. The first-order argument on management's slide is convincing: more profit per pair and fast online growth in the pandemic.

Think one step further, and you will ask two questions. What will a retailer do with the shelf space Nike gives back? What will a shopper do when the shoe is missing from the store they walk into? Both reactions happen in someone else's shop, and management's own dashboard cannot show them. So the case asks you for tripwires: warning signs chosen in advance and measured outside Nike's own stores and apps. Even then, you will not know yet whether the plan was right.

In your answer

  • "The direct effect is X, and I expect it to last about N months."
  • "The one with most to gain from reacting is Y. The cheapest thing Y can do is Z."
  • "If Y does Z, the gain falls from A to B. That changes my recommendation like this: …"
  • "The signal I would watch is outside our own channel. It is …"

Common misuse

The usual mistake is a chain of five guesses. Each sounds likely, but none has a number, and the chain ends in a story nobody will check. Nobody can prove such a chain wrong. So it never overturns a recommendation, and it never supports one. Test each step with one question. Can you name who acts, what they gain, and what you would see when they start? A step without all three proves nothing. Stop at the last step that has all three, and put a number on it.

References

Practise it on a case

More on thisWikipedia explainer

Steel-man

Next method

Write the strongest honest version of the option you plan to reject, so that its supporters would say “yes, that is my argument”.

Open the next method