Rather then deal with it in the comment section of Nick's blog, I do it here where I can edit my errors.
Note: This is my third time through, and I may still get it mostly wrong, but here goes.
The banking sector is modeled as a channel with bandwidth H. This channel encodes two transaction series, one of frequent events with small quants, and another of much larger quants happening seldom. These are sequence short and long. The two channel are optimum when the sum -iLog(i) for each of two channels is maximum.
The, i, is the inventory growth rate for either short or long, as a percentage, required for acceptable inventory risk. The short line is downward sloping, lowering reserve requirements allows for more real Y. If long term reserves were lowered we get fewer but much larger trades. The lower the long term rates, the better to hold gold long term. Raising long term rates is making the curve steep, affording more gains to the wholesale side of distribution, investment picks up.
We made a little approximation, we treated i as applying to both long and short, but the construction still holds. In channel theory we would have two members of the set i: {i1,i2} and we require i1Log(i1) and i2Log(i2) to be within an integer.
Even in the smooth world of stochastic calculus, the economist needs to get onto the side of the yield curve as power spectrum, incompletely known. Then the LM curve is just more bandwidth (gain), and we can see a random walk when both processes have equal spectrum.
It gets confusing to undergrads, like me, because the actual Y is composed of two independent parts, both represented on one axis. In both cases we assume either a Gaussian channel, or a Gaussian system function.
No comments:
Post a Comment