However, such regressional analysis cannot be taken as conclusive. For example,
1) USD appreciated against many other currencies during this period;
2) The daily movement is restricted to 0.15%
The mayin group has established a system to monitor the correlation coefficients (i.e. the basket "weight"). In addition, they have tracked an indicator to see if there is structural change in the formula. They found that in the middle of March 2006 there is finally (statistically significant) change in the formula. i.e. it is likely that RMB and USD finally decoupled.
- From 1 November 2005 onwards, the empirical fluctuation process (black line) was compared against its boundary for rejection of stability (the red line). While the process stayed below the boundary, we could not reject the null that the currency regime was unchanged compared to that prevailing in the history period. But now the process has crossed the boundary, so we know something has changed.
The 'fluctuation process' involves some complicated statistical theories. My understanding is that it means the old formula (with USD=99%+) no long holds. It is still too early to say if RMB has been decoupled from USD, but it would be a new formula, and in this formula USD does not represent 99%+ weight in the basket. It could be 80% or 98%. We need to wait for more data before we know what it is. Or, it will be changed by the time we figured it out. However, this does not neccessarily mean that RMB is "manipulated" arbitrarily, it could mean that there is an objective mechanism and built into which is something to ensure it will move by the time we gather enough data to plot out the weight composition. i.e. a 'moving target' for speculators.
Related: see Stephen Roach on the rate of crawl.