Here we finally have it! Time for some data pre-analysis to figure out individual component relationships and a few interesting facts on the distributions overall before we go into heavy-duty multi-variable time series regressions and GARCH data manipulations. This is pretty much the initial step and helps us make some more sense out of this chart:
Now a problem with this chart and the data in general is that it is of “poor” continuity, meaning that the monetary base values are seldom updated so a lot of time series relationships will return total gibberish, and the exchange notes outstanding for example is a really poor predictor as it is either static or continuously rising. This will be a bit of a nut to crack. For example, the HSI changes a lot when the monetary base remains relatively stable, and that introduces a lot of errors that are non-trivial to deal with, but this series of articles will at least start on this journey. For those of you wondering how the monetary base and the HSI are related on a level-to-level comparison, that’s what the next chart is for.
Note here that the clusters right before HK$1.1-, HK$1.25- and HK$1.35 trillion monetary base represent the flat areas of the monetary base in the top chart, and indicates the times when the index oscillates around a flat MB-value. The areas where the clusters aren’t are much more descriptive of monetary base expansion, and those seem very centered around the trend given! If the values in the clusters are removed, the R^2 value ought to jump massively as the plentiful cluster errors are removed. This still looks like spurious correlation for serious statisticians, but fundamentally there is more money chasing the same amount of assets, so although the day-to-day errors drown out the ability to value changes in the HSI from monetary statistics, it would provide a way to better understand fundamental analysis, specifically stating that a rather major change in market valuations of the Hong Kong underlying economy is needed if the HSI is to fall below 22 000 points in this monetary base regime.
Looking into the individual components of the monetary base also allows some interesting data for the analysis. Let’s start with the total monetary base versus all coins, HKMA-issued currency and certificates of indebtedness that represent currency printed by the three Hong Kong banks that do so:
Now, since this data uses daily changes (one of the few combinations that stand up to that strict test) rather than levels, it shows how much cash impacts the total value of the monetary base. According to the line slope (0.879) cash is about 88% of the change of the monetary base change even though it represents ~25% of the total monetary base at any given time. this data sort of works mainly because cash is regularly updated due to normal economic activity inside the economy, like withdrawals, payments and deposits by everything from consumers to fundamental businesses to banks. It’s not a massive surprise that this data works! the R^2 value is a little bit disappointing, but still clears the 0.15 value that time series regressions normally require, and we can on this pre-analysis say that the cash balances seem important to the overall monetary base and is a clear candidate for more modeling!
What I am perhaps most interested in is the closing aggregate balances, since these represent the total level of interbank and bank-to-HKMA accounts. When this changes it strongly indicates that the economy is seeing inflows of foreign capital that means the HK$/US$ peg needs to be defended by swelling the monetary base quickly. As we’ve seen before, this means that CAB’s don’t change a lot with normal economic activity like cash, thus leading to even worse clustering effects. Surprise! They’re still here when regressing over the monetary base, which again changes on a daily basis thanks to the effects of cash circulation in the economy:
Looking at this data, it is pretty obvious that the center cluster is too strong to really allow proper mapping of the rest of the distribution. Still, if that is ignored, the rest is leaving a vague hockey-stick shape, where the left area is relatively flat and the right area monetary base is increasing at nearly 1-to-1 with the few times that the closing aggregate balance is increasing, with the slope of the regression line “dirtied” by the 0-cluster and the left area of the graph. We still have an R^2 above 0.15 though even with these data problems, indicating that the individual points of the right-hand data are very good indeed!
So… if the closing aggregate balance isn’t reflecting lower values of the monetary base when it decreases… then where on earth does the money go? While this data is obviously clouded in the cluster, thankfully there is more data to play with!
OEFBN stands for Other Exchange Fund Bills and Notes” and generally constitutes the main electronically traded liquidity in Hong Kong that is backed by the HKMA. Of course, this situation happens because we have not been close to the lower HK$ peg band for a very long time, and the charts would look completely different if we were.