by Alasdair Macleod, GoldMoney:
Over the past few months, gold prices have completely detached from our model-predicted prices. While we have seen deviations between actual and predicted prices in the past, those deviations were always temporary. What we are witnessing now is the paradigm shift we alluded to in our report from March 2023. Some explanations for this deviation we presented in that report are still valid. However, it appears increasingly likely that the main reason for this development is the central banks’ having lost control over the gold price.
TRUTH LIVES on at https://sgtreport.tv/
Exhibit 1: Gold prices have completely detached from model-predicted values
$/ozt
Source: Goldmoney Research
Earlier this year, we published a two-part report Gold prices reflect a shift in paradigm – Part I and Part II, (March 15 & 16, 2023) in which we explored the thesis that the gold market exhibits a permanent paradigm shift. Historically, gold prices followed three drivers: Real-interest rate expectations, longer-dated energy prices, and central bank policy (net gold sales and QE). In 2016 we presented a gold price model to our readers, which showed that most of the price changes in gold can be explained with these three drivers. Deviations of the observed price from the model were usually short lived and prices eventually converged with the underlying drivers. Readers unfamiliar with our model can catch up here (Gold Price Framework Vol. 2: The energy side of the equation, May 28, 2018, here (Part II, 10 July 2018) and here (Part III, 24 August 2018), as well as some follow up reports that built on the model (Gold Price Framework Update – the New Cycle Accelerates, 28 January 2021) and (Gold prices continue to weather the rate storm, 13 April, 2022.
When the Fed began to hike rates in late 2021 as a reaction to burgeoning inflation, gold prices did first what the model would predict: They began to decline. Rising interest rates usually lead to higher real interest rate expectations if long-term rates rise faster than long-term inflation expectations. Real-interest rate expectations (as measured by 10-year TIPS yields) are strongly inversely correlated with gold prices as shown in our model. In the past, gold prices often followed real interest rates almost tick-by-tick intraday without any other news.
The decline in gold prices in early 2022 on the back of rising real-interest rate expectations was somewhat cushioned by rising energy prices. Long-dated oil prices rallied from $63.90/bbl in December 2021 to $75.91/bbl by July 2022. In our model, the rise in 10-year tips lowered predicted gold prices by $400/ozt while the rise in deferred oil prices increased it by $100/ozt (see Exhibit 2). On net, by August 2022, gold stood at $1737/ozt, just $25/ozt over our model predicted price of $1712.
Exhibit 2: The effect on gold by the rise in 10-year TIPS yields was initially offset by rising longer-dated oil prices
$/bbl (LHS), % (RHS)
Source: FRED, Goldmoney Research
However, while model-predicted prices continued to decline on the back of relentlessly rising real-interest rate expectations – and later the retracement of long-dated oil prices – gold prices turned and started to go up again. By fall 2022, we began noticing that, once again, gold prices had meaningfully deviated from predicted values. By the time we wrote our March 2023 note, gold traded already $450/ozt over the model predicted price, an absolute record at the time. Now the delta is a staggering $668/ozt. At the time of writing, gold is trading at $1870/ozt. But based on our model, it should be closer to $1202/ozt (see Exhibit 1). The chart illustrates clearly how detached gold prices have become from our model, and thus, the underlying drivers.
In our March 2023 note, we explored several theories that attempted to explain this large discrepancy between actual and predicted prices and we discussed whether we thought this was just a temporary phenomenon or whether this was something more permanent.
The first observation was that central banks of non-OECD countries have been on a massive buying spree from late summer 2022. In our March 2023 report, we highlighted that the IMFs estimate of net central bank purchases was way too low in our view. We highlighted that even the much more optimistic estimates by the World Gold Council might be too low. For our model, we are using the high end of estimates from the WGC, but if actual gold purchases were even higher, then our model-predicted price would be too low. In addition, we also explained why we think our model may underestimate the extent to which central bank purchases drive the price. Historically, changes in CB holding were relatively small, and often the reporting time did not match the actual purchase. Particularly non- OECD central banks have been very opaque when it comes to gold purchases. That means our econometric models cannot properly attribute changes in the gold price to changes in central bank gold holdings. We concluded that the true impact on the price of gold is likely larger than what our model predicts. Thus, in our March 2023 note we came to the conclusion that strong central bank gold purchases might partially explain why our model was underpredicting prices.
However, since then central banks became net sellers again. Data from the World Gold Council shows that central banks were large net gold sellers in March, April and May of this year, and only became net gold buyers again in June, July and August.
Exhibit 3: After a few months of large increases, central banks turned to net sellers in summer 2023 again
Tonnes month-over-month
Source: WGC, Goldmoney Research
On net, central banks didn’t add more gold than normal so far in 2023. In fact, despite the strong rebound over the past few months, central banks added less gold in 2023 than on average since 2009. Hence, we can conclude that the large deviation of actual and model predicted gold prices was and is not due to abnormally high central bank gold purchases.
