Gold is Undervalued: A Look at Money Supply, the US Dollar, and Gold Prices

Jake Huneycutt | January 4, 2021

In the years after the Global Financial Crisis, no one would've ever accused me of being a "goldbug". I went long on real estate, homebuilders, and even a yacht dealer, but I thought gold was a bubble. I greatly annoyed my goldbug friends with predictions of a gold crash. At one point in 2011, I even shorted gold as a professional money manager.

In 2021, my view is quite different. In the Age of Bitcoin, I believe boring, old, traditional gold is undervalued, even after a pretty spectacular 5-year run. Moreover, the gold miners, in particular, may be one of the most attractive investment opportunities out there.

Gold is currently selling around $1900 per ounce. Based on our research, gold should be priced closer to the $2400 - 3000 price range, and well-run gold miners will reap the rewards of rising gold prices.

Gold's Historical Performance

I have deep criticisms of the way gold is frequently analyzed by both bulls and bears. Bulls often find weird metrics that allow them to justify outlandish positions such as "GOLD TO $20,000!" Bears, on the other hand, often tend to believe gold is inherently worthless; alternatively, some now believe Bitcoin will replace gold. My position is that gold is a proven store of value that is sometimes overpriced and sometimes underpriced; and we should analyze from a quantitative perspective to understand when it is mispriced.

We can see how gold has performed in relative terms by comparing the forward 1 year returns on gold with the yield on the 10-year US treasury.

In the chart above, we can see many periods of both outperformance (1985 - 87, most of 2000 - 2011, 2018 - present) as well as underperformance (1982 - 85, most of 1987 - 92, 1995 - 98, 2011 -15).

So why is this a period where gold is undervalued?

The first question we must answer is how do we value gold to begin with? I think there are a wide variety of methods; some much better than others. I've been analyzing dozens of different metrics relative to gold. In this article, I'm primarily going to focus on money supply, the strength of the US Dollar, and how that relates to the price of gold.

Money Supply and Dollar Strength

If gold is a store of value and a currency, its value should be closely aligned with money supply and the strength of the Dollar (or any other currency).

So let's examine money supply. There are a few paths we can take here, as there are a wide variety of money supply metrics. I opted to use M2 money supply over some alternatives due to its timeliness. However, there are a few issues with raw M2.

M2 only tells us the supply of money; it doesn't tell us how that supply relates to the underlying strength of the American economy. Money supply should increase as the US economy grows. Therefore, I also created a US Dollar Strength index. This examines NGDP units relative to M2. The idea is that if NGDP is rising due to real economic growth, M2 should also be rising, without the Dollar weakening. However, if M2 is increasing rapidly and NGDP is not keeping pace, then the US Dollar should be weakening.

What we can see from the US Dollar Strength Index is that the Dollar has severely weakened from its 1998 peak. Its rate of decline accelerated the past decade and it has plunged since Covid. There is a caveat here in that Covid may be suppressing NGDP for the short-term more so than the long-term and my index does not factor that in, but even accounting for that, it's without question that the US Dollar's strength has declined precipitously in the past year.

The main reason for this is that M2 money supply has skyrocketed at a rate much faster than economic growth. Raw M2 has grown 26% this year; meanwhile, economic productivity has fallen; a bad combination. While the Federal Reserve often gets the blame, the bigger culprit right now is the very large fiscal deficits being run by the US government. This started even before Covid, but the pandemic stimulus packages have created the largest budget deficits since World War II. I'm not saying all the stimulus is bad (indeed, I think the checks for individuals are good), but printing money out of thin air has consequences, and a massive chunk of this stimulus has gone to special interests.

And while budget deficits will likely shrink post-Covid, I don't see the overall trend of running large budget deficits changing. The age of "balanced budgets" is long over and there's not a modicum of fiscal balance in Washington right now. So don't be surprised by budget deficits upwards of 10% of GDP even post-Covid. This means further weakness for the US Dollar in the future.

That said, let's take a look at M2 money supply modified with a US Dollar strength adjustment.

While this data only goes back to 1980, we can see that this current period is, by far, experiencing the most rapid growth in money supply in the past 40 years. With a Dollar-strength adjustment, the rate of growth is more than double the early 80s inflationary peak and nearly 3 times as high as the Global Financial Crisis ("GFC") peak in mid 2009. Which is why comparisons to the GFC are off-base; the government response has been radically different this time.

Where's the Inflation?

The question arises if money supply is growing so rapidly, why isn't inflation also rising? This a complex question, but I can divide it into 2 different answers. The first one is that the velocity of money has also plunged. This mean that while there's a lot more money in the US economy, it's mostly just parked right now, therefore not creating too much inflation. I suspect that it will not stay this way in 2021 and 2022 after more Americans get vaccinatted for Covid and people start behaving normally again.

The second answer is that, perhaps there is already inflation. Our common inflation estimators, CPI and PCE, are both flawed in numerous ways. While I won't explore that issue in depth in this article, the short version is that CPI uses a metric called "Owners' Equivalent Rent" rather than housing prices as one of the most critical elements of the CPI calculation.

It turns out there's a huge divergence in rental prices and housing prices right now. Owners' Equivalent Rent has fallen to 2.3% growth due to Covid. Housing prices, in contrast, have seen accelerated growth. The Case-Shiller 20-City Index shows 6.6% annual price growth in its most recent reading (Sep 2020). Meanwhile, the more conservative All Transactions Price Index still shows 4.7% growth. Keep in mind that while these numbers are far from records, they are huge when you consider that NGDP declined 1.8% from the prior year. This strongly suggests stagflation is occurring.

If we replace OER with the All Transaction Housing Price Index in CPI, annual CPI growth increases from 1.18% to 1.76%. If we instead use Case-Shiller 20, inflation jumps from 1.18% to 2.54%. Which is to say, that our traditional measures of inflation are significantly underestimating real inflation right now.

A true measure of inflation is probably closer to that 1.7% - 2.5% range, which seems pretty high considering that we are theoretically living through one of the most disinflationary events of our lifetimes, the Covid-19 pandemic. If inflation is already around 2% now with an extremely subdued economy, what happens once the majority of the US population is vaccinated and money starts flowing again?

Predicting the Price of Gold Using Adjusted Money Supply

With all that out of the way, we can make a simple guess at what the price of gold should be based on our US Dollar strength adjusted M2 money supply figures. We examined the historical ratios for gold price relative to our custom metric; we found the median, the 25 percentile (conservative) and the 75 percentile (aggressive). From that, we can see how this metric predicts gold prices in the chart below.

While it's certainly not a perfect predictor, when it's wrong, it tends to be early, rather than late, which is why I view as such a good metric for investment purposes. In January 1981, it said the price of gold was too high; gold fell nearly 50% over the next 4 years. Around 1998 - 99, it told us that gold was signficantly undervalued. It took several years for gold to fully correct, but between 2000 - 2004, gold rose 50%, and it mostly continued to rise straight through to 2011. However, the metric suggested that gold was overvalued by late 2009. Once again, and true enough, in 2012 - 13, gold experienced a huge correction, falling about 25% in 2 years, before declining even further.

Since around 2014, however, our metric has suggested that gold is underpriced and the disparity grew much larger after Covid. In fact, the numbers it is suggesting do seem somewhat insane. I would raise some caution that the extraordinary nature of this pandemic is probably distorting these figures a bit, and I'd go more on the conservative side, but even using the conservative price estimate (using our 25 percentile valuation ratio) puts gold at around $2,400 per ounce, meaning it would need to rise about 25% to reach that number.

The median predictor suggests gold should be priced around $3,400 per ounce; which I personally feel is a bit high, but I also wouldn't discount completely. It's very easy to underestimate an event before it happens and I've found myself being an uber-bull on certain stocks before, and still being "too conservative". So my personal feel is that gold to $2,400 is reasonable in the next 3 years and it might have additional upside beyond that.

All this said, there is a data-oriented counter-argument as well.

Cash Costs

While I think money supply paints a compelling picture for why gold is undervalued, I do also want to point to another metric that I created that tells a slightly different story: the cash costs of mining gold.

On a basic level, gold should virtually always be valued higher than the cash costs of production. Remember, that cash costs including things like labor, energy, and transportation, but it does not include capital expenditures, so the cash costs are not the true all-in costs of gold production; and indeed, the incentive of gold miners is to make their cash costs seem lower than reality, so we should always take the "cash costs" with a grain of salt. The reason I decided to use cash costs, however, is because there's a longer history of gold miners reporting their cash costs than their all-in production costs; so this is a "more data is better" decision moreso than a judgement on which metric is superior (both have their merits).

The chart below shows the price of gold compared to the cash costs of mining for one gold miner, Barrick Gold. I used Barrick because they had a long track record that made it easier for analysis, though, I think a blended-average of several miners would be even better. Also note I created a "predicted price" based on the media gold / cash costs ratio.

So what does this tell us? First off, it's a decent predictor. The price of gold rarely veers too far off that 2 - 3x cash costs range. So if, for instance, the cash costs of gold were $600 and the price of gold were $3000 (a 5x ratio), we could surmise that gold was overvalued. That said, there is one glaring problem; cash costs appear to be a lagging indicator, which is problematic for investment predictions. We'd much rather be "early" than "late" in investment.

You can see this most clearly in the 2009 - 14 timeframe. Our predictions always seem to be late in the run-up. Then, in 2013, our prediction shoots all the way up to $2100 per ounce right at the time that gold plunged from $1700 down to $1200. Thus, the cash costs may work better as a "sanity check" than an actual predictor. So while cash costs make sense to look at, I would not use it as a main predictor.

That said, even using the more conservative cash costs as a basis for gold's true valuation, it's not unreasonable to go with something in the range of $2,100 - 2,200, around 3x the cash costs; particularly when we consider how much higher even the conservative M2-based predictions go.

Other Considerations

While I wanted to focus mainly on historical gold prices relative to money supply, there are several other eye-opening metrics on gold. I won't do a deep-dive analysis in this article, but one I will point out one that's quite important, which is the level of gold and silver mining production.

There are a few interesting takeaways from that data. Peak US gold production occurred in 1998 during the longest gold bear market since the US exited the gold standard. The state of the market is very different in 2021. We're actualy coming off the lowest levels of gold and silver production since 1988 in the US, in spite of surging prices. It should be noted that this measure looks at the US and most gold production is happening abroad, but the US is still the world's 4th largest gold producer and global supply has been relatively flat for years.

But even examining overseas shows us a potential supply crunch. Australia, the world's 2nd largest gold producer, is expected to reach peak production in 2021. In that same linked article, Gold Fields CEO Nick Holland is quoted as saying that global gold production has probably "peaked".

So we have a situation where government budget deficits are surging, M2 money supply is likely rising at the highest rate since World War II, and global gold production will likely flat-line or decline over the next few years. To me, this is a pretty bullish scenario for gold; albeit, it was obviously better to have gotten on the train at $1500 or $1700 (where I jumped in), but even at $1900 or $2000, I don't think this is the end of gold's bull run. Don't be shocked to see $2500 or even $3000 gold within 2-5 years, particularly if the US government continues running large budget deficits.


This is just a touch of some of the work we've been doing at Aardvark Data. We're building algorithms to predict the price of gold and other commodities and we've examined dozens of factors in our analysis. However, the strength-adjusted money supply is one of the most interesting and valuable predictors.

For my view, the strength adjusted money supply is the best long-run predictor for the price of gold and the cash costs are a good "sanity check". Moreover, production / supply is also a key factor. Based on this, I think the conservative estimate (at the 25 percentile) for gold from the strength-adjusted M2 is reasonable. That would put gold's true intrinsic valuation somewhere around $2,400 per ounce; about 25% higher than the current price around $1,900.

While it's a work in progress, our more robust algorithms are also producing predictions more in line with a $2400 - 3200 price point for gold per ounce. These algorithms are based on several long-term factors. When combined with the rest of our analysis, I find the overall case for investment in gold to be quite strong right now.

These predicted price points would be significantly higher at 3.35x than the typical cash cost ratio prediction (typically in the 2-3x range), but ultimately, we should expect some of this inflation to filter down and I expect the cash costs for mining gold will increase over the next few years, particularly given the likely scarcity of new production.

Based on our analysis, I am buying gold via the Perth Mint Physical Gold ETF ($AAAU). I recommend Perth Mint or GraniteShares Gold Trust ($BAR) if you want to buy a gold ETF.

Even more so than gold, I think the gold miners could be an excellent investment right now. I am long on Barrick Gold ($GOLD). As I'm still in the early stages of evaluating the miners, I don't necessarily offer Barrick as the "best gold miner to buy"; it's merely the one that I felt most comfortable after some very rudimentary research. I'm open to suggestions on others.

This is still a "work in progress" for us at Aardvark Data, and our algorithms may change in the future as we find further improvements and better data sources, but I've felt confident enough from our current research to start buying in and publish some of our research.

About Aardvark Data

Aardvark Data develops quant finance, advanced analytics, machine learning, and data storytelling solutions for clients. We are currently looking for co-founders with skills in data engineering, full-stack development, and marketing / social media. Our tech stack includes Python, Vue.js, D3 / Vega-Lite, and SQL. Inqueries should be directed to info at aardvark-data dot com.

Author: Jake Huneycutt, Aardvark Data

Last Updated: 4 January 2021, 06:00 am EST

Sources: US Federal Reserve