RSS
Twitter
Every Economic Model Is Wrong... But That Doesn't Mean They're Not Useful
from the hating-on-economists dept
Fred Wilson has a thought-provoking post on his blog, asking what if the economic model you're basing business decisions on is wrong? He's building off of an Umair Haque blog post, asking whether or not economists matter any more, because the models they use are wrong. I think there's some truth to both of these pieces -- and I know both Fred and Umair, and enjoy exchanging ideas with both -- but it feels like they're throwing a pretty big baby out with the bathwater.
I don't think it's that the "old models" don't work any more. The problem is simply that they're models, not crystal balls. Every model has faults (that's why it's a model, rather than the real thing). What makes economics so difficult is that any real scenario involves too many variables to accurately model. Each modeling attempt tries to account for the most important of those variables, and assumes the impact of the rest washes out. However, changes in the world may impact one variable a lot more than expected, or change the relationship between variables -- and that's what throws a model out of whack. Every economic model is wrong because it simply cannot account for every human variable out there.
But, that doesn't mean you ignore what the models tell you. You can still learn quite a lot from those models -- including why they're wrong, and how to improve on them. So, for example, when Haque discusses Starbucks and Microsoft, as if basic economic models failed them, I disagree. I just think they didn't put enough weight into some important variables in their models. Starbucks, for example, failed to recognize the importance of "culture" in its equation, but that doesn't mean the whole equation was wrong.
It's certainly tempting to throw out the old models when they're proven incorrect, but that makes you lose a real learning opportunity. You should start from the position that all models are incorrect from the beginning -- and as each model is proven incorrect, figure out why so you can correct and improve, rather than simply tossing the whole thing out.
