Being stochastic isn’t what makes it unpredictable. The movements of molecules in a gas are stochastic, but at large scales gas pressure is perfectly predictable.
What you need for unpredictability is a combination of stochastic input, a high degree of sensitivity to initial conditions, and non-linear responses to those conditions. That’s what defines a complex system, along with the fact that connections between objects are more important to understanding the system than are the objects themselves. Macro generally fails at this, attempting to reduce complex relationships down to simple aggregates that can be manipulated with equations essentially derived from physics.
Even our explanations for observed events are simplified. We look for and are offered simple explanations for why unemployment is where it is, why more women don’t go into a certain field, why markets rise and fall, etc.
Ask an ecologist, “Why did that river change its shape?”, and you aren’t getting a simple explanation. Instead, it’s always some long and complicated chain of events: Wolves were introduced into the area. That scared the grazing animals away from drinking ponds, which enabled beavers to come in, which flooded the area… But then they’ll admit that this was wholly unpredictable, because maybe instead introducing the wolves caused a reduction in rabbits and other small animals, which in turn caused a completely different effect. It wasn’t kjowable until after the fact, and even then all you could learn is how the system evolved that time. Rerunning the experiment might have given a completely different result.
Even understanding the system perfectly doesn’t make it predictable. We understand ant hills extremely well. We xan modelmthem i computers and produce the same behaviur, etc. But even all that kjowledge won’t allow you to predict the evolution of an anthill. Sensitivity to intial conditions: two ants are foraging, and both about tomfind major food supplies. But one ant has to climb over a twig and the other doesn’t and wins the race. So the whole dolony winds up evolving in that directly. The butterfly effect, applied to ants.
People aren’t ants. We’re far less predictable. And human social and economic systems arent just complex, they are complex adaptive systems. That means our ability to learn from past behaviour is very limited, yet that’s exactly what we try to do by building models and testing them with sequestered data from past performance. What if that past performance was a random walk, and if you could repeat the same ‘stimulus’ again and again, you’d get a different result every time? Complexity theory says that’s exactly what would happen.