Multicausality and the economics of causation and control

We often think of causation as a simply linear flow from causer to causee. A hand pushes a box across a table, and we say that the hand, or the force imparted by it, causes the box to move.

In economics and development, causality is not so simple. The most well known example of this is the case of reciprocal externality. Suppose that Alice is Bob’s neighbor. Bob likes to go to bed early, while Alice prefers to throw loud parties late into the night. Alice’s parties make it hard for Bob to go to sleep, but Alice really values the fun parties and the friendships and connections that come out of them. This situation is an externality. The question is, who caused it?

As Nobel prize winner Ronald Coase showed, you can’t simply say that either Alice or Bob caused the externality. Instead, both parties must be present for the externality to occur. If Alice wasn’t neighbors with Bob, there wouldn’t be any parties to disturb him. But if Bob wasn’t neighbors with Alice, there wouldn’t be any people for Alice’s parties to disturb. The externality is a situation, and all parties have to be present and doing what they’re doing for the situation to be happening the way that it is.

Externalities, as Ronald Coase showed, don’t have a single cause. Instead, externalities are inherently what Esther Thelen called multicausal. Multicausality is an entirely different concept from the idea of multiple causal factors. It’s not, “Alice and Bob each contribute 50% of the externality”, it’s “Alice and Bob both have to be in place, doing what they’re doing, for the externality to happen.” Neither party can cause any amount of externality on their own.

What could be attributed to a single party—to Alice or Bob alone—is efficient control. Maybe it would be really easy for Bob to buy a pair of earplugs so the parties don’t bother him anymore. If it would cost Bob $5 to get a pair of earplugs that are totally comfortable for him and block all noise, we might say he’s “causing” the externality if he refuses to buy the earplugs because it would be so easy for him to change the situation so that the externality goes away. Or maybe Alice could easily move her parties elsewhere so Bob wouldn’t be bothered, in which case it would be more tempting to say that Alice is causing the conflict. Here, causation isn’t some kind of exact, logical distinction: it’s a practical claim about which part of the system offers the most control for the least cost. If you can push Bob a little and get a big effect on the overall situation, we’ll say that Bob causes things to happen. If we can adjust Alice inexpensively and get a big effect, we’ll attribute causation to Alice.

We can see this in the simple example of a hand pushing an box across a table. Causation can’t simply be attributed to the hand because without the box being present, the hand won’t be able to push the box. And without the table being present, the hand can’t push the box across the table. Again, this is multicausality. However, because the changing what the hand is doing in the context of the overall situation can yield a lot of change for very little cost, we say as a practical matter that the hand is the thing causing the effects.

Development is a highly multicausal process. The genes can’t be said to cause development because the genes need the cells to make use of them and for the environment to shape what happens. The same limitation also applies to bioelectricity: the same bioelectric signals will yield different effects depending on the cells reading them and the environmental situation. A tremendous amount of confusion in the history of development was due to scientists seeking single factors, or even multiple interacting causal factors, to explain development when development is actually multicausal. The entire nature-nurture debate is an example of the kind of conceptual issue created by a failure to recognize the multicausality of a system. There are many related issues in medical research.

Macroeconomist Scott Sumner has described the importance of understanding multicausality in macroeconomics, with two recent really good posts explaining his views. When you drive a car, there are many factors influencing the speed of the car, like how hard the driver is pushing down on the gas, whether the car is going uphill or downhill, what the wind is doing, etc. However, as a practical matter, if one factor can achieve very large effects on the outcome at little cost, which stepping on the gas usually does, then it is correct as an economic decision to attribute causation to the decision to step on the gas, even if the situation overall is multicausal. Causation is a decision about how to most effectively interact with a situation. It’s economics, not physics or logic.