When a war starts in the Middle East, most people expect two things: oil goes up, and gold goes up. In 2026, only the first one happened the way the textbooks say.
Brent crude jumped from around $70 a barrel before the war to above $120 at its peak. Gold, the classic "safe haven", fell sharply at times and spent much of the summer under pressure. Then the Fed raised interest rates in September, and gold bounced.
Why? To answer that, I went back to a well-known research paper by economist Lutz Kilian, "Not All Oil Price Shocks Are Alike" (first released as a CEPR discussion paper in 2006, and later revised and published in the American Economic Review in 2009). This note follows the 2006 version, so some details may differ from the published paper. It was written long before 2026, but its main message fits today's market very well: the same oil price rise can mean very different things, depending on what caused it.
This note has three parts, so you can always see whose idea you are reading:
Part 1: What the paper says. A plain-words summary of Kilian's findings. These ideas are his, not mine.
Part 2: My own analysis. How I connect the paper to gold, what 2026 looks like through it, three later oil moves I tested, and where I think it falls short. This part is my own reading.
Part 3: What to watch and the takeaway. The data points I will follow, and the main lessons, split into what comes from the paper and what comes from me.
Part 1: What the paper says
This part summarises Kilian's paper. The findings are his, explained in plain words.
The big idea: four kinds of oil shock
Most people talk about "the oil price" as if it moves for only one reason. Kilian says this is too simple. He splits the real (inflation-adjusted) oil price into four types of shock:
Political supply shocks: wars or revolutions in OPEC countries that cut production.
Other supply shocks: production changes from cartel decisions or non-OPEC countries.
Global demand shocks: the world economy is booming and needs more industrial commodities, including oil.
Oil-specific demand shocks: buyers rush to hold more oil because they fear future shortages. Kilian links this to "precautionary demand", which is basically fear of running out.
To measure the strength of the global economy, he built a clever index from dry-cargo shipping rates (the cost of moving grain, coal, iron ore and similar goods by sea). He then studied the period from 1975 to 2005.
What the paper found
1. Supply cuts alone did not drive the big price jumps. Looking at history, the major oil price rises since the 1970s came mostly from strong global demand and from fear-driven demand, not from physical production losses. The paper found that political supply disruptions did not raise the real oil price in a statistically significant way. Part of the reason: other producers tended to pump more to fill the gap.
2. Fear moves prices faster than barrels do. The clearest example is 1990, when Iraq invaded Kuwait. The paper finds that the price spike was almost entirely due to precautionary demand, meaning fear about future supply, not lost oil. It was the market's expectations that moved, and they can change almost overnight.
3. The same price rise can hurt the economy in different ways. This is the most useful finding for investors. In the paper, oil-specific (fear-driven) demand shocks had the clearest negative effect on US growth and, at the same time, pushed up consumer prices persistently. That is stagflation: weaker growth and higher inflation together. A price rise driven by a booming world economy looked very different: growth got a short-term boost first.
4. Central banks matter. The paper notes that a central bank might raise interest rates early in response to an oil supply shock, and that how much growth suffers can depend on how the central bank reacts.
5. A cushion decides everything. Using the 1984–88 "Tanker War" as a test, Kilian showed that attacks on ships did not always move oil prices. Tankers in the Gulf were hit again and again, at times more than 30 in a month, yet prices mostly did not follow, because other producers could replace the lost barrels. Fear even fell sharply in 1986, when OPEC's control broke down and oil looked plentiful. The market stayed calm when other countries had spare capacity to step in. A war only shocks prices when that cushion is missing.
One honest note: the paper itself says its results are tentative and that many estimates are imprecise. It is a framework to think with, not a calculator for today's price.
Grey bars show moves driven mainly by demand or calm markets. Blue bars show moves tied to lost or feared supply. The sections below explain each one.
Part 2: My own analysis
From here on, the ideas are my own reading of how the paper applies to gold and to 2026. The paper itself does not cover them.
From oil to gold
The paper is about oil, not gold. So here is how I connect the two. Gold does not react to oil directly. It reacts to what oil does to inflation, interest rates and fear:
Oil up → inflation expectations up. In theory this helps gold, because gold is seen as protection against rising prices.
Inflation up → central bank turns more hawkish. Higher interest rates (and higher real yields, meaning yields after inflation) make gold less attractive, because gold pays no interest. This is the "opportunity cost" of holding it.
War and fear → safe-haven buying. This supports gold, but it usually works in short bursts.
Gold is therefore pulled in two directions at once. Which side wins depends on the type of shock. If it is the stagflation type, the Fed faces a hard choice between fighting inflation and protecting growth, and the market's guess about that choice decides which channel wins.
2026 against the framework
Here is the year so far, in simple steps:
Late February: the shock. The US and Israel began a war with Iran, and Iran effectively closed the Strait of Hormuz, through which about 20% of the world's oil and LNG normally flows. Brent was around $70 before the war.
March: gold falls, not rises. Gold dropped to around $4,570 in what was on track to be its biggest weekly fall since 1983, as higher energy prices pushed rate-cut hopes away.
April: peak oil stress. A ceasefire was agreed on 8 April, but the Strait stayed largely shut. By 30 April, Brent was above $120 (it touched about $126 during the day). Industry tracker Kpler estimated that at least 11 million barrels a day of crude output had gone offline.
July: gold's low. With renewed US strikes on Iran and rate-hike bets growing, gold slipped to around $4,019.
August: the war premium fades but does not vanish. Brent was trading near $90, and only a handful of ships a day were crossing Hormuz, a tiny fraction of normal traffic.
September: the hike. The Fed raised rates for the first time in about three years, with Brent near $105. Gold rebounded about 2% to roughly $4,360.
What does Kilian's framework say about this?
Read through Kilian's framework, this looks like a mix of a political supply shock and fear about how long the Strait stays shut. The paper links fear-driven shocks most strongly to stagflation, which is what makes a central bank's job hardest. It also helps explain the gold puzzle: in March and July the rate channel outweighed gold's safe-haven appeal, and the September hike removed some of the uncertainty. This is an interpretation, not a proven cause. The dollar, central bank buying and positioning also matter.
Stress test: three oil moves after 2005
Kilian's data stop in 2005, so here are three later oil moves, read through his framework (my own reading, not a re-run of his model). The test for each: was fear of future supply the driver? In all three, it was not.
1. 2003 to 2007: the emerging-market run-up. Oil rose from about $30 to nearly $100 by early 2008 and peaked near $147 in July 2008. The driver was strong demand from China and India with flat supply. Kilian's paper already finds the 2003 to 2005 rise was almost entirely demand, and later studies (Hamilton, 2009; Kilian and Murphy, 2014) agree. Kilian and Murphy use inventory data to test for hoarding and rule it out as the main cause.
2. 2008: the crash. Oil fell from about $147 in July to roughly $35 to $40 by December as global demand collapsed in the financial crisis. Dry-cargo shipping rates fell by more than 90% between May and December. Fear of shortage vanished, as it did in the 1997 to 1998 Asian crisis.
3. 2020: the drop and the negative price. Brent fell from about $65 to the low $20s by April, and US WTI futures briefly traded below zero on 20 April. Lockdowns cut demand, a Saudi-Russia price war added supply, and nearly full storage at Cushing, Oklahoma pushed holders of expiring contracts to pay to avoid delivery. Fear worked in reverse: the worry was too much oil, not too little.
Episode | Main driver | Fear of future supply |
|---|---|---|
2003 to 2007 | Demand boom | Small |
2008 | Demand collapse | Gone |
2020 | Demand collapse, supply surge | Reversed |
2026 | Lost Gulf supply | Large |
To tell causes apart, check whether shipping rates and metals move with oil (demand), whether inventories are rising (hoarding), and whether near-term oil prices sit far above later ones (scarcity now).
Where the paper may not fit 2026
I see four reasons to be careful.
1. The "cushion" is missing. In the paper's history, other producers usually offset political supply cuts, so prices stayed calm. This time, much of the world's spare capacity sits inside the Gulf, behind the closed Strait, so the offset is weak. This is a major difference from most of the 1975–2005 sample.
2. The "fear" shock is a leftover. Kilian defines the oil-specific demand shock as whatever is not explained by the other three. He admits it could include things like changes in inventory policy. So "fear" is a reasonable label, but it is not directly measured.
3. The data end in 2005. The sample is before US shale oil, before years of near-zero rates and quantitative easing, and before today's very large central bank gold purchases. These could all change how oil and gold respond to shocks. The stress test above checks the framework on three later episodes, using my own reading, not Kilian's model.
4. It is a US-focused study. It looks at US growth and inflation. A shock that hits Europe or Asia harder, or a strong or weak dollar, can change the outcome for gold priced in dollars.
Part 3: What to watch and the takeaway
The data points below are the ones I will be following. The takeaway separates the paper's findings from my own conclusions.
What I will watch next
Hormuz traffic: the number of ships crossing per day. This is the clearest sign of whether the missing cushion is coming back.
The oil futures curve: a steep drop between near and far months would suggest the market sees the shock as temporary, while a flat curve would suggest fear is sticking.
Real yields and the Fed's next move: because this is the channel that seems to have mattered most for gold this year.
Central bank gold buying: a possible source of support for gold that did not exist in Kilian's sample.
The takeaway
2026 in one line: Brent rose from about $70 to above $120 after the Iran war closed the Strait of Hormuz, gold fell to around $4,000 by July, and the Fed's September rate hike brought a rebound to about $4,360.
The main lesson is simple: don't ask "what is the oil price doing?" Ask "why is it moving?" A rise driven by a booming world economy, a physical supply cut or a wave of fear can lead to very different results for inflation, growth, interest rates and, in the end, gold. Here is what I take away, split by whose idea it is.
From the paper (Kilian's findings)
The cause of an oil price rise matters more than the size of the rise.
Global demand and fear of future supply explain most big oil moves, more than lost barrels do.
Fear-driven oil shocks are the most likely to bring stagflation, which is the hardest problem for a central bank.
A war only moves prices when spare capacity elsewhere cannot cover the loss.
From my own analysis
In 2026, oil reached gold mainly through the Fed. Higher energy prices raised rate-hike expectations and real yields, and for much of the year that outweighed gold's safe-haven appeal.
2026 looks like a lost-supply and fear shock, closer to 1979 and 1990 than to the demand-led moves of 2003 to 2008 or the collapse of 2020.
The paper's cushion is missing this time, because spare capacity sits behind the closed Strait of Hormuz. This is the biggest reason I would not apply its history one-for-one.
This is a reading, not a forecast. Real yields, Hormuz traffic and central bank gold buying will show whether it holds.
Bottom line: in 2026, gold has moved with the Fed's response to the oil shock more than with the war itself.
In the next note, I will look at the gold side in more detail and test whether the "safe haven" label still holds when interest rates are rising.
This note is for information and education only. It is not investment advice, a recommendation, or an offer to buy or sell any security or commodity. The views are my own and do not represent those of any employer. Market data are approximate and taken from public news reports; please check primary sources (EIA, LBMA, FRED) before relying on them.
Main paper discussed: Lutz Kilian, "Not All Oil Price Shocks Are Alike: Disentangling Demand and Supply Shocks in the Crude Oil Market", CEPR Discussion Paper No. 5994 (December 2006); later revised and published in the American Economic Review, 2009. This note follows the 2006 version, so details may differ from the published paper.
2026 market context: Reuters, AP, S&P Global, Kitco and Trading Economics news reports on Brent, gold and the Fed, March–September 2026.
Later episodes (2003 to 2008, 2020): public price data, plus James D. Hamilton, "Causes and Consequences of the Oil Shock of 2007–08", Brookings Papers on Economic Activity, Spring 2009, pp. 215–261; and Lutz Kilian and Daniel P. Murphy, "The Role of Inventories and Speculative Trading in the Global Market for Crude Oil", Journal of Applied Econometrics 29(3), 2014, pp. 454–478.