I’ve been studying dynamic econometric models. Yes, that sounds scary. But that lead me to a deeper degree of understanding about economic reality (understanding the first and second derivatives, that is).
Let me explain. A dynamic econometric model is a model that, to predict a dependent value, takes into account the values of past periods of time.
Why is that interesting? Well, when you are meddling with economics you find several cases:
- In the macroeconomics IS/LM model where yields are related to consumption, investment and government spending, there is a multiplier between those values, that means that, for example, with increased government spending this money will spread into the system, to consumers, to investment and ultimately to a yield increase. There will be a multiplier because of the recursive interchange of money. Even though that can mean less investment by substitution effect, the long run will be better. That’s why government deficit can help overcome recessions.
- Or when you pour more money into the system (if you are a central bank, that’s for sure) then a part of this money will go into bank accounts, a part again into lenders, and another part into reserves. But the global quantity will be higher than what you have poured into the system. Another multiplier.
But these processes have two things in common. First of all, they share the fact that the initial change is increased in effect by a multiplier, so there is some kind of amplification. And they also share the fact that these increased effects do not happen instantly but by realimentation: there’s an initial increase that is a new input to a second increase, and the process goes on an on.
The process converges, that means that there’s a point of equilibrium which it goes closer to. And it is not a short run process, but takes time to see the change.
So when we are writing the equations of those processes we are assuming that everything happens instantaneously. And that’s untrue. As we care about economic magnitudes not only on a yearly basis but for shorter periods of time, we should devise more accurate ways to describe reality.
As an engineer I can also see it from another point of view. Recursive digital filters are used in digital signal processing. Given a response there are ways to find which way the filter must be. There you have tools like the z transform. In fact you can emulate all kinds of analogic filters just with the right digital signal processing device.
That will sound strange and far-fetched for most economists but, believe me, both techniques have a lot of things in common.