What makes a strong assumption?

Why false assumptions can be good assumptions.

Economists often assume that agents are “rational” - but that’s obviously false. After all, it’s well known that humans are prone to a range of cognitive biases: scope neglect, status quo bias, availability bias… you name it! Clearly, this suggests that the assumption of “rationality” is terrible - right?

Well, no. I personally hear this kind of reasoning a lot (i.e. “if an assumption is false it must be bad”), but I think it’s pretty mistaken. In particular, I think these criticisms boil down to a misunderstanding about what assumptions are for.

To see why, let’s go back to our previous example. Although this assumption is indeed “wrong”, there are many things that it helps explain. For example, it tells us why people are more likely to buy a good if its price decreases, and why stock market crashes are fundamentally hard to predict (if everyone somehow knew that the stock market would crash in a week from now, then they’d all sell immediately, making the stock market crash today). It plays an important part in explaining why inflation occurs, why countries go to war, etc. Much of your bog-standard economics textbook depends on assuming that agents behave “rationally”, i.e. they tend to find the correct way to achieve their objectives. As you can see, just from this simple assumption, you get a huge amount of explanatory power about a truly massive range of real-world phenomena.

This of course does not imply that the assumption is always valid. There are indeed countless cases where humans wander very far from perfectly rational behaviour, littered throughout the expanding literature on behavioural economics. In trying to explain these cases, it would be rather silly to start off with the assumption of rationality.

What this example tells us is that you can’t just judge whether or not “rationality” is a good assumption without accounting for the context. In some cases it’s a good assumption, and in others not so much. The difference is whether or not making the assumption provides a lot of explanatory power for some question of interest, at low cost.¹ So it’s perfectly possible for an assumption to be false, but still be a good assumption in some context! False assumptions can be good assumptions.

I think the same is true across pretty much all of science. Newton’s law of gravity in some sense is “false” (e.g. it doesn’t do a great job at explaining phenomena around black holes), but assuming it’s right is still super useful in a wide array of less “extreme” scenarios. Gases in the real world aren’t ideal gases (e.g. they’re not comprised of perfect point particles), but you can learn a great deal about gases by assuming they are. And so on!

Perhaps this might all sound totally obvious to you. Great! At least personally, I hear equivocations between “false assumption” and “bad assumption” pretty frequently, and I regard this as somewhat of a fallacy. So if you agree with me then please call it out!

There are a bunch of other corollaries of this. When you make assumptions, it’s important to be specific about what situation you’re referring to - this helps clarify what the role of the assumption is in your claims or arguments. When you’re reviewing the work of others, you should also pay attention to the most important assumptions - the ones that could really impact their conclusions. If you think an assumption is bad, explain why in the context of what the assumption is being used to explain.

At least in my experience, paying attention to these details can be a very rewarding part of the research process. On the other hand, it’s also slightly embarrassing when I catch myself making fallacious arguments of this form!