Everyday Economics

“Complexity breeds precision; it does not necessarily breed accuracy.”

There is something about human nature that is inextricably drawn to precise numbers. Perhaps we feel that it provides a level of certainty and control in a world that is anything but certain and controllable. Or maybe I’m just making that up. To be fair, I am an economist and not a behavioral scientist.

However, there is undeniably something about precision that we find attractive.

In our everlasting quest for precision, we continue to build more complicated mathematical, statistical, and analytical tools to improve our predictions. In economics, GDP will be calculated with seasonality, controls for industry, supply shocks, and a million and one other things. When we spend so much time and energy creating something that is supposed to be accurate, perhaps it’s natural for us to feel that it better be.

But the reality is that oftentimes, our precision is misguided. Our models make a great deal of assumptions in order to generate estimates. And perhaps more importantly, there are a great deal of assumptions that they fail to make, and things which they fail to take into account.

For example, let’s say that I’m trying to project revenues for my business.

I can follow a very simple approach and calculate my expected revenue per customer, multiplied by my expected number of customers. Let’s say that this model gives me a projection of $5000.

Similarly, I could segment my model by customer demographic, product type, etc. This model is much more precise and gives me a projected revenue of $5240.

So now I have two numbers, and the question is which one do I use? Well, the $5240 estimate appears to be more accurate—it takes into account a lot of things that the simple model doesn’t. More importantly, it looks to be a more precise estimate, and so we naturally gravitate towards thinking about this as the best model.

But is it the best model? Well, the best answer is maybe. It’s possible that the complex model accounts for the relevant variables accurately, and therefore generates a more accurate prediction. On the other hand, every additional variable creates an additional opportunity for a mistake or faulty assumption to be made. For example, what if you assume that you’re going to sell an equal number of products A and B, when in reality, you’ll sell almost exclusively Product A? Suddenly, the complex model will actually get you further away from your accurate answer.

Complexity breeds precision; it does not necessarily breed accuracy.

And this is the crux of the matter. Not only does complexity increase the likelihood for certain errors, but it becomes a double-edged sword by convincing us that it’s all the more certain for its complexity.

The fact of the matter is that $5000 is unlikely to be exact. But also, $5240 is unlikely to be exact. We are, almost by definition, conducting inexact science.

The challenge is that when you’re working with simple models, you assume that they have a degree of inaccuracy, and you take this into account. As the model becomes increasingly complex, you assume that the inaccuracy becomes increasingly negligible. But the truth is that regardless of how you choose to conduct your predictions, they’re always going to be some degree of wrong.

This line of thinking can help prevent you from wasting too much time digging through the weeds and from the human curse of thinking that we know more than we do—something that arguably extends far beyond revenue projections.