Let me concede the obvious upfront — Societe Generale's cross-asset desk has been right about gold more often than wrong over the last three cycles, and the "conviction broadens" framing they've been circulating this year does land on something real. Money that ignored the metal from 2013 through 2019 has been moving back. Central banks are net buyers on a scale the World Gold Council reports quarterly. That much is genuine. What I have noticed, though, after reading every sell-side gold note that crossed my desk this quarter, is that they all miss the same three things — and the omission is not accidental.
The pattern is consistent enough that it stops being coincidence and starts being culture. Sell-side gold notes are written for a client base that pays for reassurance dressed as analysis, and "conviction broadens" is a phrase engineered for exactly that trade. It sounds like a call. It functions like a hedge. The reader who has been long can nod. The reader who has been underweight can start to move without admitting error. Everyone wins except the person who wanted to know what the flows actually mean.
What They All Get Wrong
The shared error — and I mean shared, I could not tell you which desk wrote which paragraph if you stripped the letterheads — is that they treat "conviction broadens" as a directional statement. It is not. It is a compositional statement, and the difference matters.
Here is what I mean. When a note says conviction has broadened, what has almost always happened underneath is that the *mix* of buyers has changed. The buyer profile in 2015 was ETF-heavy, dollar-sensitive, price-elastic. The buyer profile now is central-bank heavy, geopolitically motivated, and largely price-insensitive at the marginal ton. Those are two different flow regimes producing the same directional print on the tape. Calling both of them "conviction" flattens the analysis into something an intern could write.
The second failure follows from the first. Once you treat all incremental demand as one substance, you end up predicting price action from aggregate flows without asking which portion of those flows is reversible. Central bank purchases, once made, do not exit at a fifty-dollar drawdown. ETF flows do — that is literally what the historical tape shows every time the metal breaks a technical level in a hurry. So a note that lumps them together and calls the total "conviction" is describing a stack of dry wood and wet wood and telling you the whole pile will burn evenly. It will not.
The third failure is the most quiet. Every note I read this quarter uses the word "broadening" without defining it. Broadening across what? Across investor type? Across geography? Across time-horizon buckets? The word does real work in the reader's head without ever being cashed out on the page. That is the tell. When a piece of financial writing depends on a load-bearing word it refuses to define, you are reading marketing, not research.
You can test this yourself. Take any current gold note. Circle the word "conviction" or "broadening" wherever it appears. Now try to replace each instance with a specific, measurable claim — "central bank purchases have risen from X to Y over Z quarters, per WGC" or "European private-bank allocation to gold has moved from A basis points to B." Most of the time you will not be able to do it, because the underlying specificity is not there. The word is doing the lift the number should be doing.
What Is Almost Always Missing
What never appears in these notes — and I have looked — is the composition table. Not a table of prices, a table of *buyers*. If you told me who bought each incremental ton over the last eight quarters and at what implied price sensitivity, I could tell you something useful about where the metal goes next. If you tell me the total demand number and call it broadening, you have told me almost nothing I can act on.
The second absence is the motive layer. Central banks buy gold for reasons that have almost nothing to do with the twelve-month price forecast. They buy it to reduce concentration in one particular reserve currency after watching another country's reserves get frozen. That is a diplomatic act priced in tons, and it will continue at prices that would make an investment committee walk away. A note that models this demand with the same elasticity assumptions it uses for wealth-management flows is going to be wrong about the floor and wrong about the ceiling.
The third absence is the paper-physical gap. Most retail exposure to gold — and increasingly a lot of institutional exposure — is paper. Futures, ETF units, structured notes, leveraged CFD positions at retail brokers. The math of what happens when a meaningful percentage of that paper wants to convert to physical is not discussed, ever, in these notes. It is the plumbing risk sitting underneath every "bull market broadens" thesis and nobody wants to write about it because nobody wants to be the desk that flagged it first.
*The London vault capacity numbers are published quarterly. The gap between outstanding paper claims and deliverable metal is not.*
The fourth absence is regulatory friction. Sell-side notes on gold rarely mention that the retail vehicles feeding "broadening" demand — leveraged CFDs on spot gold, in particular — have been under active restriction across multiple jurisdictions for years. ESMA capped retail leverage on gold at 20:1 back in 2018. ASIC followed with equivalent restrictions in 2020. The retail broker landscape that services this demand — the FCA-regulated books at operators like Exness, FXTM, and HF Markets, the ASIC-regulated books at AvaTrade and FBS — is a fundamentally different beast from what it was during the 2011 rally. Retail flow into gold is regulated flow now. That changes the reflexivity of any "broadening" thesis in ways nobody prices in.
*The regulator's phone line is answered on the third ring. Nobody at the desk that wrote the note has ever called it.*
What I Would Say Instead
Here is the reframing. Instead of "conviction broadens," which is a directional word masquerading as a structural one, I would say the *motive stack* has diversified. That is a longer phrase and it does not fit a chart title, which is one reason nobody uses it. It happens to be closer to what is actually going on.
The motive stack for gold right now has at least four layers, and the sell-side notes conflate all of them. Layer one is reserve diversification by central banks — this is the largest and least price-sensitive component. Layer two is sovereign wealth funds hedging currency exposure. Layer three is family-office and private-bank allocation. Layer four is retail speculation via paper products at leveraged brokers.
Let me show you why this matters, because when you separate the layers, the math gets interesting.
Suppose — and these are illustrative proportions, not published figures — that central bank buying accounts for roughly a quarter of net incremental demand in a given quarter. Assume that portion has a price elasticity near zero over any twelve-month window: those buyers do not care whether they paid $2,200 or $2,600 for the last ton. Now assume the remaining three-quarters is elastic, with an implied stop-loss discipline that kicks in on a ten to fifteen percent drawdown from recent highs. Do the arithmetic. In a scenario where price drops fifteen percent from a local high, roughly seventy-five percent of the marginal buyer base is now either exiting or hedged. The price-inelastic quarter keeps buying. The price does not go to zero. It re-bases at a lower level, then holds there because the inelastic bid is still in the market. That is a completely different price path from what a "conviction is intact" narrative predicts, and it is completely different from what a "the bubble is popping" narrative predicts. It is the actual physics of a mixed-motive market and it is what the notes should be modeling.
Now take the same math and run it forward. If central bank buying accelerates by twenty percent — plausible on any given geopolitical rupture — while retail elastic demand stays flat, the composition of the total shifts. The inelastic share moves from twenty-five percent to something closer to thirty. The floor rises. Retail can still panic on any given day, but the trough of that panic is higher than it would have been the quarter before. That is what "broadening" would mean if the term were used honestly: the inelastic share of the buyer base is growing, which structurally raises the floor without necessarily raising the ceiling.
Listen — I know the notes are pleasant to read. I know your PM likes them. But if you are trying to size a gold position on the back of "conviction broadens," you are sizing on a phrase that has been engineered to feel like analysis without doing the work of it. When I blew up my first commodity book, it was not because I read a bad note. It was because I read a good-sounding note and forgot to ask what the words were doing. Ask the question. Circle the word. Try to replace it with a number. If you cannot, you have your answer about what the note is worth.
The thing I would tell the sell-side desks, if any of them were asking, is that "broadening" is not wrong — it is just unfinished. The interesting question, the one that has not been answered in any note I have read this year, is whether the inelastic share of gold demand has crossed a threshold at which the metal starts behaving less like an investment asset and more like a reserve asset, priced in the currency it is displacing rather than the currency it is quoted in. If it has, the analytical framework that produced every one of these notes is out of date. If it has not, we are still in the world where a fifteen percent drawdown liquidates most of the marginal buyer and the "conviction" phrase is worth exactly what it costs to print.
FAQ
Why does the "conviction broadens" framing sound convincing even when it is thin?
Because it lets both a bull and a hesitant reader nod at the same sentence. The phrase promises structural change without committing to a testable claim. Sell-side desks write for a mixed audience — long-holders needing reassurance, underweight readers looking for a face-saving reason to move. A phrase that flatters both is more commercially useful than a specific forecast, which is why it recurs across desks that otherwise share no analytical DNA.
What data would actually validate the "broadening" thesis if it were real?
A quarter-by-quarter decomposition of net gold demand into central bank, sovereign wealth, private-bank, and retail buckets, with each bucket's implied price elasticity estimated from historical drawdown behavior. The World Gold Council publishes the top-line quarterly demand figures. What is not published — and what would settle the question — is the elasticity coefficient by buyer type at current price levels. Without that layer, "broadening" is a vibe.
How much of retail gold exposure is paper versus physical in 2026?
The exact split is not disclosed in any single public dataset, but the direction is clear: retail leveraged exposure via CFDs at brokers operating under FCA, CySEC, ASIC and FSCA licenses is a materially larger share of the retail base than it was in 2011. That matters because paper positions unwind faster and at higher magnitudes on drawdowns than physical holdings do, which changes the reflexivity of any bull thesis.
Do retail leverage restrictions on gold change the analysis?
Yes, and this is under-discussed. Since ESMA capped retail leverage on gold at 20:1 in 2018 and ASIC followed in 2020, the retail flow into the metal has been structurally different from prior cycles. Higher-leverage retail books have migrated to offshore-regulated venues, while onshore retail exposure is smaller per capita. The aggregate retail bid is less volatile than it used to be, which slightly raises the floor and slightly compresses the peaks.
What is the risk that central bank buying reverses?
Central bank gold purchases, once settled into reserves, historically show very low reversal rates over any twelve-month window. The motive is diplomatic and structural rather than tactical. That does not mean a reversal is impossible — a sovereign in acute liquidity distress will sell — but the base rate is low enough that modeling central bank demand as sticky is closer to right than modeling it as elastic.
Why don't sell-side notes disclose their buyer-composition assumptions?
Two reasons. First, the underlying elasticity data is genuinely hard to source and defend, so most desks work with rough assumptions that would not survive publication. Second, the commercial function of the note is to move the client toward action, not to expose the analytical uncertainty behind the recommendation. Disclosure of composition assumptions would invite questions the desk does not want to answer under a subscription pricing model.
If the "motive stack" has really diversified, what is the practical implication for a reader trying to size a position?
Size the position against the layer of the motive stack you actually understand. If your thesis rests on central bank buying continuing, size for a slower, less volatile path with a higher floor. If your thesis is a retail-flow momentum trade, size much smaller and set stops tighter, because that layer of demand is the first to exit on a technical break. Do not size the position against the aggregate demand number, because the aggregate is a blended object that behaves like none of its components.
Is there a single question the sell-side gold desks still owe their readers?
Yes. Whether the inelastic share of gold demand has crossed a threshold at which the metal is being repriced against the currency it is displacing in reserves rather than the currency it is quoted in. If the answer is yes, most existing frameworks are stale. If the answer is no, "conviction broadens" is a phrase that describes nothing you can trade. Nobody has published a defensible answer. If you have one, write.