This post aims to use the bare minimum of statistical modelling to build a view on the gold price that is consistent. While I don't discount the importance of issues around physical demand for gold, for the sake of simplicity we focus on the gold price as a variable determined by financial forces than the minutiae of physical demand/supply balances.
As this post is a little long, the key points are as follows
- Gold is surprisingly well behaved against the US dollar and real US interest rates over long periods.
- Simple regression models work well for these periods, but gold has had sudden shifts in behaviour (both bullish and bearish).
- This gives us two key questions in testing a gold price view
- 1. how much will gold move given the prevailing model and assumptions?
- 2. what might cause the model to change?
In part two, well look in more detail at applying this framework to current conditions. But first a little analysis.
It would appear from data that over long periods of time that gold behaves fairly consistently as an alternative to other highly liquid financial variables, mainly the US dollar and real interest rates on US Government bonds. While the relationship is in reality heavily intertwined between these variables (for the wonkish creating an issue with multicolinearity), simple linear regression models of gold using the the above variables do have significant explanatory power over significant periods of time.
To demonstrate this, the chart below shows the gold price vs. simple linear regression models of gold vs. the US Trade Weighted Index and real interest rates. Estimates were calculated over defined time periods as noted in the chart
What's interesting about this chart is these models work surprisingly well for long periods of time, but then tend to break down. When it does breakdown, gold tends to have a sudden shift in relative behaviour. To make it a bit clearer, the below charts focus in on particular time periods.
This chart focusses in on the late 1980s and the early 1990s. Surprisingly, the model that works reasonably well in the late 80s is consistently too bullish on prices in the 1990s.
To be sure, given the reasonable accuracy of the 90s model, the gold price appeared to have a structural bearish shift through the economic cycle from the late 1980s to the 1990s.
This model in the 1990s, however, was too bearish at the turn of the millenium, particularly from 2005 onwards.
While the 2000-20008 model is not as precise as some of the others, directionally it does work enough to illustrate the point.
In today's context this is a surprise, as conditions then were almost the total opposite of where we are today.
Even this bullish model broke down during the turmoil of 2008, which appeared to set a whole new paradigm for gold vs. our key variables. There have been signs that this model is starting to break down somewhat, which we will explore in part two.
One might think I've only highlighted a bunch of models that work until they don't, which is not very useful. But it's the reasons that they break down which are critical to the framework.
For example, it demonstrates that the Lehman meltdown in 2008 caused a fundamental bullish shift in the gold market. Similarly, the dawn of lower inflation and interest rates in 1990s had a fundamental bearish impact. The 2000s up to the financial crisis was interesting, as it seems that loose monetary conditions globally had a bullish impact on gold well beyond falling interest rates and a rapidly dropping US dollar.
So what does this all mean? I think a good way of thinking about future gold prices is to first think about what you think about the US$ and real interest rates, then use that against the current model. This helps quantify what gold should do depending on other variables for which there seems to be more sensible views (but also some crazy ones).
Second is to think about potential looming events that might shift the prevailing paradigm. As the last 30 years has shown, these are not always bullish and not always intuitive. For example, the high-risk taking behaviour of the mid-2000s hardly lends itself to the current commentary on why gold prices should go higher, but generated some of the strongest price increases.
In part two we will look more closely at recent performance within this framework and some potential future scenarios.
