If we examine what the Fed does in an overheated economy, we find events are changing rapidly. Most of the new information for the Fed comes from the last two quarterly results. Conversely, during long periods of low growth and low volatility, the Fed needs to look back two years, two sets of annual numbers to find anything innovative.
Accountants always use annual rates in comparing yields, for a very good reason; annually every economy completes one seasonal cycles, seasonal effects are cancelled. But in the two cases above, the first, the target rate is the six month rate, and in the second it is the two year rate. This follows directly from applying Shannon to information efficiency. The two previous significant events have the best predictive power, it is best to 'mark to market' all the significant events prior to the last two. I think this is going to be fundamental in spectrum constrained, self adapting statistics. In number theory this is about 'conserving divisibility'; divisibility ans spectrum play similar roles.
Consider the recent puzzle regarding seasonal adjustments, why did we have this Q1 slump start showing up? Well, it had to happen two years in a row before we were convinced.
Its like a combinatorial problem, we have a defined sequence of partitioned groups we can make, in which we maximally use prime numbers and make the best quotient ring. When we do this, pricing is most efficient,meaning we can multiply and divide prices for longer estimations with respect to some artificial variable like the year. Once we set the first partition, the prime twos separation, then we set the spectral conditions for the remaining aggregates, we know where the knee of the yield curve will lie, with respect to artificial time.
No comments:
Post a Comment