Gold is not a haven in a hiking cycle

Special Report: Hikes, Weak Treasuries, Haven Demand — and Why the System Turned Bearish on Gold

2026-09-28 · Macro Analysis

> ⚠️ This is a macro analysis, not a forecast, and not investment advice. Sources and the full disclaimer are at the end.

---

A reading that looks self-contradictory

Since the middle of September, ETF Radar has read gold as bearish: GLD lean-negative at 0.60 confidence, IAU at 0.65, silver (SLV) at 0.55. Over the same weeks the Fed and the ECB have been raising rates, the Bank of Japan has been normalising, Treasuries have been selling off — and gold is supposed to be the haven.

Put side by side, it looks like the system read the sign backwards.

Take it apart and the contradiction lives in one common intuition: treating weak Treasuries as good news for gold. That is usually the wrong way round, and right now it is as cleanly wrong as it ever gets.

What follows is what the system actually had in front of it, why those numbers point down, and — just as important — which reservations it wrote into its own call.

Three things the system saw

One: real yields are rising hard.

Over the last twenty sessions:

Nominal yields rose fifty basis points, inflation expectations barely moved, so essentially all fifty landed on the real yield.

That is gold's most direct pricing variable. Gold pays nothing, so the cost of holding it is the risk-free real return given up. With the ten-year real yield going from 2.34% to 2.83%, holding gold got nearly half a percentage point more expensive inside a month.

Two: the dollar is broadly stronger — though it did weaken slightly against the yuan.

The broad dollar index went from 118.36 to 120.33, up 1.67% over twenty sessions. Gold is priced in dollars, and a stronger dollar presses on that price by itself.

But "the dollar is weakening" depends entirely on what you measure it against:

Against the yuan the dollar did weaken — by thirteen basis points in a month, which is barely a move at all. Gold is a globally dollar-priced asset, and the dollar it faces is the broad one, not any single pair.

For a holder thinking in yuan the currency leg made things slightly worse, not better: GLD fell 6.81% in dollars over twenty sessions while the yuan gained 0.13% against the dollar, which works out to roughly −6.93% in yuan terms.

Three: speculative money is leaving.

CFTC net speculative long positioning in gold:

Four consecutive weekly declines from the 25 August peak, 17,358 contracts or about 12%. This is a line of evidence independent of price — it says who is leaving, not which way the price went.

Correction one: weak Treasuries are not good news for gold

This is the crux.

Falling bond prices help gold in exactly one case: when money is fleeing bonds for gold, so that gold takes the haven bid. Gold rises, bonds fall.

There is another case, and it is the one happening now: bonds are falling because real yields are rising — the market demands a higher true return before it will lend. That kind of decline is not haven money in flight; it is the opportunity cost of money going up. For a zero-yield asset there is no upside in it at all.

Side by side, the two chains diverge after the third step:

Steps ① and ② look the same from a distance — bonds are moving either way — but only the left-hand chain passes haven demand through to gold. Step ④ is the variable gold actually faces, and it alone decides the direction of step ⑥.

Today's readings sit in the right-hand column: nominal +50bp, breakeven +1bp, real +49bp.

The way to tell the two apart is to look at inflation expectations.

• If an inflation scare drives nominal yields up and breakevens rise with them, gold usually benefits — it hedges debasement.
• If nominal yields rise while breakevens sit still (now: +50bp against +1bp), what rose was the real yield, and gold is on the wrong side of it.

Today's decomposition is about as unambiguous as these get: breakevens moved one basis point in twenty sessions.

Correction two: in a hiking cycle gold is not a haven, it is a zero-yield asset

The "haven" label hides something: gold's haven property is transmitted mostly through real yields, not directly.

The textbook haven move runs like this: risk event → money buys Treasuries → bond prices rise and real yields fall → gold's opportunity cost drops → gold rises. Gold rides the bond market's coat-tails.

So when bonds themselves are falling and real yields climbing, the first link in that chain is broken. Geopolitical risk can be high and still not offset a 2.8% annual real opportunity cost.

That is why the system holds the geopolitical arcs at low weight in this call: Gaza at 0.09 confidence, the Taiwan Strait at 0.08, US–Iran at 0.25, Korea at 0.12. They are all present. Against a 49-basis-point move in real yields they are not the same order of magnitude. The weight sits with "metals prices and demand" (short, 0.56) and "the Fed's 2026 rate cycle" (short, 0.36).

Three reservations the system wrote into its own call

Worth noting: the call is not made at full volume. Confidence is 0.60 rather than 0.8, and the reasons are in the thesis — all three point the other way.

One: part of the downside is already priced. One cited arc carries price_in=True. The market is already pricing this direction, so what is left to earn is the remainder, not the whole.

Two: money is still coming in. GLD's 30-day net inflow is 1.41% of assets. Price falling while money arrives is a disagreement, not a consensus. Shorting into that means facing the people who have not left.

Three: options are extraordinarily cheap, which is the dangerous part. GLD's implied-volatility rank is 2% over a 43-day window, and the gold volatility index GVZ has a z-score of −1.51, three days into a low-vol regime. Almost nobody is paying for a violent move in either direction. A very low volatility price plus money still flowing in is the standard recipe for a squeeze — any geopolitical or policy surprise would make this short uncomfortable.

The system wrote all three into the thesis and cut its confidence accordingly. That is the right behaviour: direction and conviction are two different questions.

Why the miners were not marked down with the metal

An easily missed detail: on the same day the system reads gold miners completely differently — GDX lean-positive at 0.09, GDXJ neutral at 0.0, against GLD and IAU at lean-negative 0.60 and 0.65.

The prices explain it:

The miners fell harder than the metal — they carry leverage to it — yet remain far ahead over sixty days. Their price also contains output, costs and takeover prospects, which is a separate story that does not flip in step with real yields.

Two directions on one theme is not the system contradicting itself. They are not the same instrument.

What would overturn this

Three tells, all readable daily, in order of importance:

One: breakevens start moving. The ten-year breakeven is at 2.34% and has moved a single basis point in twenty sessions. If it starts rising with nominal yields, what is climbing is no longer the real yield, and gold's case reverses on the spot. This is the number to watch, not the gold price.

Two: real yields top out. The five-year real yield rose 57 basis points in twenty sessions, faster than the ten-year. That pace is hard to sustain; once it stalls or reverses, the short loses its main support.

Three: the haven chain reconnects. That means stocks and bonds going back to moving inversely — bonds rallying on a risk event instead of falling with everything else. The system currently reads the regime as "rates/inflation-driven risk-off, bonds sell too, not a haven", which is exactly the broken state. When it resets, gold's haven property starts working again.

As for the geopolitical events themselves — Gaza, the Taiwan Strait, US–Iran, Korea — they are all in the system's reading, at low weight. Not because they do not matter, but because against 49 basis points of real-yield repricing they are not yet enough to change the direction.

In closing

Back to the reading that looked self-contradictory.

The contradiction comes from an intuition that is widely used and often misapplied: that weak Treasuries and haven demand are both good for gold. In fact weak Treasuries right now mean rising real yields, which is gold's most direct headwind; and haven demand reaches gold only through rising bond prices, so while bonds themselves are falling that channel is closed.

The system's reasoning is simple: nominal yields rose fifty basis points in twenty sessions while breakevens rose one, so what rose was entirely real. Add a broad dollar up 1.67% — the 0.13% softening against the yuan is small enough that, converted, it actually widens the loss in yuan terms — and four straight weeks of speculative longs leaving, down 12%. So far price agrees: GLD is down 6.81% over twenty sessions against the S&P's 0.17%.

And it still caps conviction at 0.60, because three counter-arguments are just as clear: some of the move is priced, money is still arriving, and implied volatility sits in its 2nd percentile. That last one matters most — a volatility price that low means nobody has paid for a surprise, which is precisely the environment in which a short gets hurt.

This is not a conclusion that gold must fall. It is a judgement that, against a real-yield move of this size, the bull case for gold is missing its most important support. When that support returns, the place to look is breakevens, not the gold price.

---

Data through 2026-09-28. Rates and currencies from FRED: DFII10, DFII5, T10YIE, DGS10, DTWEXBGS, DEXCHUS, DEXUSEU, DEXJPUS (all through 2026-09-25). Positioning is CFTC net speculative long in gold, report date 2026-09-22. ETF prices, fund flows, implied-volatility ranks and the system's own readings come from data already stored in ETF Radar.

---

Disclaimer

This piece was produced and assembled by an automated research system. It is general information and educational discussion only. It is not investment advice and not a recommendation to buy or sell any security or commodity, and it has not been tailored to anyone's financial situation, investment objectives or risk tolerance.

All figures come from public sources and may contain collection, parsing or definitional errors; past data does not indicate future performance. The system readings quoted here are the output of an automated model. Models are wrong sometimes, and their confidence levels are not probability guarantees.

The operator of this site is not registered with the Ontario Securities Commission or any securities regulator. Investing involves risk and decisions are your own; consult a licensed professional adviser before acting.