Humans are perfectly rational, self-interested agents.
We have perfect information when making decisions, and have stable preferences over time.
Given any two objects, we can always choose which one we prefer.
These are all examples of assumptions that economics has traditionally relied on, with some even treated as axiomatic.
I’m sure you could think of many ways that reality diverges from them: I could choose to lose $10 to punish someone who is acting unfairly, despite valuing my $10. Consumers don’t have full information about a business’ costs. Businesses rarely have complete information about consumers or the market. I could go on.
I’ve been studying economics for over six years and I’d be lying if I said there wasn’t a little voice in my head that said “this is bullshit” every time I came across an assumption like this.
I recently spoke with a behavioural economist to try and figure out why these assumptions are so widely used.
Now, behavioural economics is a field that recognizes people often behave irrationally due to psychological and social factors, like biases in our judgement. It tries to account for this, sometimes by relaxing assumptions to include more realistic ones (ie. bounded rationality). Naturally, I assumed this behavioural economist would also be critical of mainstream economics’ unrealistic assumptions.
I was wrong. He told me it’s more nuanced than that.
First, economists know these assumptions are unrealistic. They are not stupid. Making these assumptions gives us the ability to mathematically model choices and human behaviour, actually solve these models, and yield clear results.
There is a trade-off between realism and tractability.
A real economy has billions of moving parts, incomplete data and non-linear human behaviour. A model that tries to include all of it would be impossible to solve. An intractable model is too messy and complex to get clear results from.
While adding psychological realism might describe behaviour more accurately, it also gives your model less precision, more degrees of freedom, and can often lead to a model that can’t predict anything at all.
Well, why do we try and mathematically model something as complex as the economy in the first place?
It’s so that we can think about economic problems in a structured, logical, well-reasoned way. Without this basis, anyone can claim whatever they want about how the economy works and there’s no way to check if there is any validity to the claim.
Second, it turns out that what matters is not whether an assumption is realistic. It’s whether its lack of realism actually matters for the question you’re trying to answer. Whether it predicts well.
Take, for example, time consistency: the assumption that your preferences don’t change between today and tomorrow. If you’re studying a static risky choice where all payoffs happen today, assuming time consistency changes nothing about your answer. It just simplifies the math; makes the model solvable.
However, if you’re studying how people make decisions about education or investments, or what determines criminal activity; choices that trade off costs and benefits today against the future, the same time consistency assumption can substantially change your results.
I then asked: if behavioural economics has found a way to incorporate more realistic assumptions into models, why is mainstream economics still riddled with unrealistic assumptions?
He said that the “behavioural” way of thinking has pervaded the discipline significantly compared to 10 or 15 years ago. Economists won’t necessarily call themselves behavioural economists, but more and more research discusses behavioural elements or takes them into account, even if implicitly.
And when researchers do use simplifying assumptions that are known to be unrealistic, they’re increasingly explicit about whether those assumptions actually drive their main result.
His final point, and my key takeaway: the foundational concept of economics is that humans are goal-oriented, and that we change our behaviour in response to our environment. It’s hard to argue with this. And if you look at all the assumptions and axioms as a whole, despite their individual limitations, you can see how many of them stem from this foundational concept, with tractability filling the gaps.
His points made sense to me. I still wonder, though, why our models, despite all the effort we put into making them tractable, still yield inaccurate predictions.
Is an inaccurate prediction really better than no prediction at all?
Could there be real-life consequences to these inaccurate predictions? After all, these models are used by governments and central banks (for example) to make decisions; they do impact real people’s lives.
I wonder if the field’s tunnel vision on tractability comes at the expense of finding new ways to solve more “complex” models that incorporate more realistic assumptions.
We have some pretty powerful tools now. AI agents. We can simulate economies. Agent-based modeling is a thing; a pretty intriguing one at that (and incidentally, the topic of my next post).