This is a reprint without changes or updates from a posting on this site on 18 January 2008.
Someone had asked about it, and so I looked and found it. I stopped updating the spreadsheet described in this posting at least three years ago and have not run any correlation analysis since.
I was not using the native regression in Excel, but a bolt-on tool that was Excel friend and fairly nice, if a bit pricey.
As I have stated in the past, I suspect that the greatest correlation now is to manipulation of price, especially since about 2013. Trading in precious metals has become very political since they almost let it slip lose in 2012.
I use the term 'we' quite a bit. I had some help behind the scenes at that time. It seems odd now. The time period covered in this analysis is roughly 2003 to 2008. I had done and published quite a bit of this sort of thing on my old blog Jesse's Crossroads Cafe. I used to trade futures actively and would put up ten minute charts of the SP futures three or four times a day.
As I have mentioned several times I think the Dollar DX index is passe, All you have to do is look at its weighting to see it is a child of Bretton Woods.
One thing I did not mention here is that the weightings of the variables did differ over periods of time. They were averaged and not always usefully so. I did quite a bit of work trying to 'crack' the variables looking at their own subcomponent factors. This I never did publish anywhere.
What Is Driving the Price of Gold?
We last took a look at the price of gold a few years ago, and not surprisingly found a high correlation to M3. Since the Fed no longer supplies this data, we thought it might be interesting to see what a fresh analysis turned up for this leg of the gold bull market.
The spreadsheet we used contains weekly data on 35 categories of economic and market measures since January, 2003 which is 265 observations and more than enough for statistical validity. Most of the data comes from the St. Louis Federal Reserve official database.
We aren't going to go into the methodology we used to find correlations, as it gets a bit technical and very tedious. Let's just say its all about finding the prime candidates, and then trying a significant number of 'better or worse' fits. The measure of 'fit' used is the R-Square Adjusted which is expressed as a percentage. The higher the percentage, the more the model explains the price of gold. And before the quant geeks come out of the woodwork, we stipulate that we have simplified and rounded both the equations and the concepts for more generalized readers.
The US dollar is the most obvious factor to check as a driver for the price of gold (in dollars), but our analysis showed that the dollar only explains about 58% of the price of gold since 2003. Money supply is the next most obvious factor. Since we no longer have M3 available as data, we went a different broad measure of money supply, Money of Zero Maturity, MZM. It is what the Fed refers to as the best measure of liquidity in the system, and is M2 less small-denomination time deposits plus institutional money funds. If it included eurodollars and net repos, it would be roughly equivalent to M3.
As you can see from this equation, the Price of Gold (POG) equals .261 times the Money of Zero Maturity supply NSA (Not Seasonally Adjusted). Since .261 is a positive number, we say the correlation is positive, meaning as MZM goes up, the price of gold will go up. The actual number itself means little since we are comparing the price of gold in dollars versus the MZM in billions. The R-Square Adjusted is about 89% which is a very high correlation for a single variable.
POG = 0.26 * MZM NSA billions - 1281
R-Square Adjusted 89%
The way we would state the above result is that the Price of Gold is positively correlated with MZM (NSA) to about 89% from 2003 to today. While the money supply as measured by the broad liquidity measure MZM is increasing, the price of gold will be increasing over time, with an accuracy of about 89%. You could say that each billion in MZM results in about 26 cents to the price of gold, but that is a little misleading since its happening over such a long period of time.
Now, 89 percent sounds good and it is for one variable. As the usual suspects go, liquidity supply of the US dollar is the prime candidate. But we wanted to add some of the other suspects in combinations, to see if we can improve on that without getting ridiculous. When we worked many years ago at Bell Labs, we sometimes saw techs taking projects like this to an impractical degree of fineness, certainly well beyond anything that might be applied to the practical problem at hand. We used to call it "trying to measure the depth of the ocean with a micrometer."
Without getting into too many details, about 50 software runs later we arrived at the following best fit for the price of gold since 2003.
POG = 0.1607 MZM NSA billions + 34.3 EFF + 12.3 Moody's Baa - 740
R-Square Adjusted 94.6%
EFF is the Effective Fed Funds Rate. This is the market expectation of what the Fed Funds rate as expressed as a volume weighted average of all the actual transactions. Moody's Baa is the interest rate for Baa corporate bonds. Its a measure of perceived riskiness in the corporate environment.
So we would say that the price of gold is positively correlated to the growth in the liquid money supply (MZM) and negatively correlated to the higher short term official interest rates and positively correlated to corporate risk. with about 95% accuracy. Makes sense? Passes the red face test? Pretty much we think.
So, if you think on the whole that MZM will keep increasing and the Fed will be lowering short term rates, with a dash of corporate risk in the mix, the price of gold should continue to do well over the long run. Since these variables also feed into the valuation of the US dollar as expressed as DX, without the noise of currency manipulation, we should see a similar negative correlation to the dollar over time.
Well, you might say, that's all very well and good if you are a long term holder of gold for five or more years AND things remain as they are, but what about the shorter term price of gold?
We've been doing a lot of work in this area, and most of it would become incredibly complicated very quickly if we tried to explain it here. Let's just say that the relationship to money supply and EFF is definitely still there, but with a great deal more noise in the model, even if the statistical sample is no smaller than one year. This is where DX comes back into play as a modifier and adds something to the mix. By introducing a risk variable like VIX we have been able to take the R-Square up to 94%.
The market place of buyers and sellers obviously sets the price of gold. As the saying goes, in the short run it's a voting machine (with appropriate antics) and in the longer term its a fundamental discounting machine; what drives it in the short term is somewhat different from what drives it in the longer term.
One might ask, "why don't you factor in Central Bank gold sales?" Prior to 2003 we think they were a significant factor in the price of gold, and several people did quite a bit of work in this area. Since 2002 the data leads us to believe that central bank gold sales have had an increasingly weak and temporary effect on the direction of the price of gold. Why engage in complexity when the data analysis is so straightforward without it?
In summary, the data indicates that since 2003 the price of gold in US dollars is strongly related to the growth in a broad money supply measure like MZM or M3. What the market thinks the Fed intends to do with short term interest rates and therefore money supply growth, Effective Fed Funds, is also a powerful factor. Finally, the perception of riskiness in the business world has a smaller but significant effect, as we see in using Moody's Baa rates and also the VIX.
We expected DX to play a stronger role in driving this leg of the gold bull market, but apparently it is playing a role only in the short term wiggles. If it is money supply expansion and lower short term interest rates that has been driving the price of gold for the past five years, with a bit of riskiness tossed in for spice, then the outlook for the price of gold over the forseeable future looks bright. In some future pieces we will touch upon deflation and credit crunches, but for now those remain possibilities and not certainties.
So what drives the price of gold? In this case, as in so many other financial questions, it always seems that we must follow the money.