Why economic analysis doesn't work well.

[william the wie]

This is going to get technical real technical real quick I'm mostly trying to figure out what erroneous math tools are being used to screw up the economy. If you don't understand something ask but none of the people on the thread may know the answer so be aware of that. I'm trying to discover how much economists don't know they don't know but think they do.

Most of the laws of physics are algorithms. All laws of economics are highly dependable rules of thumb but this difference is usually forgotten and that leads to disasters. Then there is the matter of control systems like the fed. So far as I can tell the board of the Fed and its open market committee have never made any statement indicating that they are aware of, much less use, the following tools:

Fuzzy logic. Developed in 1964 it is a branch of set theory in math. The classic example is the transition from cups to mugs some cups are not mugs and some mugs are not cups but most are both.

Catastrophe theory. That name is not much used anymore but tipping points, strange attractors in Chaos theory and quantum leaps are all catastrophes: entering a singularity and the smooth path you were on is suddenly taking you someplace else entirely.

But the big problem is that economic models are based on continuous functions that do not exist in reality however continuous functions are easy to work with. So the Fed, White House and congress are using conclusively disproven economic laws to get into deeper trouble. For example the Efficient Market Hypothesis predicts that fat tails or black swans should happen 5% of the time. They happen 8% of the time is the consensus analysis or index options at the margin are underpriced by 60% which is a not so minor problem. So what other errors are built into our economy?

 

[flacaltenn]

I do have some experience in fuzzy logic for control systems (not economics) and it seems to me that it is suitable for economic modeling. Instead of math equations, tools or regressions, it simply reduces to the control system to a logical "word" problem. When C < B then A goes UP. It gives a "relational" if not mathematical way of modeling close-loop systems.

I also know that neural network techniques are used to find previously unknown patterns in economic data. You select an input of parameters that you think are important and throw data training sets of these parameters to a multi-dimensional polynomial combo of all the inputs. By training it, the net optimizes the output truth by exhaustively exploring all the possible coefficients for the polynomial combos. Problem with this technique is -- it might perform BRILLIANTLY -- but extracting the actual mathematical relationship from the trained net is not trivial. So you may never know HOW the net became such a brilliant predictor.

Don't really know why these advanced tools are so important. Like the weather, you can look out your window and guess the weather pretty accurately with a few primitive tools.

I think the larger issue is that economic modeling and theory hasn't kept pace with American economic reality. And old assumptions based on when we had a manufacturing sector and a reasonable national debt are gonna get us in more trouble than the scary modern tools..

 

[Trajan]

the only yardstick I'll employ, Time= Money.

 

[del]

"if you laid all the economists in the world end to end, they still wouldn't reach a conclusion."-anon

 

[william the wie]

Yeah, but I am still waiting to hear that the Fed is no more than a quarter century behind the curve.