Same War, Different Shock: How Markets Responded to the US-Iran Conflict the Second Time Around

In June I wrote about how the US-Iran war moved markets. The finding was that the war's fingerprint was real, ordered, and short. Markets responded to the conflict in roughly the way you would expect, for about two weeks. Then they moved on, well before the ceasefire was signed.

Then the ceasefire collapsed. In early July, fighting resumed, sanctions were reimposed, and the conflict restarted. I had a rare chance to test my own finding. Same dashboard, same 28 assets, same baseline model. Does the same shock hit the same way twice?

It did not. The pattern came back. The reaction did not.

The setup

Both rounds started with oil at nearly the same price. Round 1 opened at $67.02. Round 2 opened at $68.55. A $1.53 gap. This matters because if oil had entered Round 2 at $90 or $100, a smaller reaction could just mean "it was already elevated and had less room to run." That is not what happened. Oil entered July cheap. It had fallen 26% over the prior month as the June peace deal drained out the war premium.

If anything, Round 2 should have been a bigger surprise. A war restarting while oil is in free-fall, after weeks of markets pricing in peace, is a larger disruption to expectations than the first time. It was not.

The staircase came back

I re-ran the dashboard's analysis machinery anchored on July 7. For each asset, I measured how far it deviated from its baseline (the same single-factor market model fitted on pre-February data), then compared the actual deviations to the predicted exposure scores.

The pattern appeared. Predicted losers lost. Predicted winners gained. All four control-group assets (utilities, consumer staples, REITs, small-cap) showed no significant move, up from 3 of 4 clean in Round 1. The war was genuinely moving markets again.

But the shape was inverted. The two rounds start indistinguishable: 0.33 and 0.35 on day 2. From there they diverge completely. In February, the rank correlation between predicted exposure and actual outcome spiked to 0.64 by day 5 and decayed steadily to 0.12 by day 22. In July, it dipped, then built to its strongest at day 15 (0.58), and was still holding at 0.41 on day 22. The lines cross around day 12.

February spiked and faded. July built slowly and held.

The inversion follows from the composition of the two shocks. February was diffuse: a dozen markets moving at once for overlapping reasons. That produces a strong signal immediately and a noisy one soon after, as each market's own story reasserts itself and the common driver stops dominating. July had a single driver. One clean factor sorts a cross-section more slowly, because it takes time for a crude move to propagate into refiners, energy equities and airlines, but it sorts more durably, because there is only one story being told.

Half the size, a fifth of the instant reaction

Oil's numbers tell the story most clearly.

Day-one jump: In February, oil leapt 11.4% above its baseline on the day the war started (measured by regression discontinuity in time). In July, the same measurement gives 2.5%. A fifth of the instantaneous reaction to a restart of the same conflict.

Peak abnormal move: Oil reached +56.8% abnormal in Round 1 versus +30.5% in Round 2. In raw price terms, WTI peaked at $112.95 in February and $92.19 in July. Round 2 topped out $21 below a level Round 1 had already proven reachable four months earlier. There was headroom. Oil stopped climbing because buyers stopped buying, not because it hit a ceiling.

Speed to peak: Oil hit 90% of its eventual move by trading day 24 in Round 1. In Round 2, it took 13 trading days. The arc completed in about half the time.

The fear trade disappeared

This is where the two rounds diverge. In February, the shock was diffuse. Everything moved. In July, only one trade showed up.

The VIX (fear index): Round 1 went from 19.9 to 31.1 and stayed elevated above 20% of its baseline on 20 of 26 trading days. Round 2 went from 15.6 to 20.7 and was elevated on 3 of 20 days. As of now, it is back to 14.9. Part of that gap is a generally calmer market in July rather than anything specific to the war. The absolute move still differs sharply: +11.2 points against +5.1.

Gold: Scored +2 as a predicted war beneficiary. Gold fell in both rounds. In February, gold's abnormal return reached -19.5%. In July, -8.6%. Gold never played the safe-haven role in this conflict at all. It had peaked on January 29, a month before the war started, and declined through the entire conflict. A war that closes the Strait of Hormuz and never once produces a sustained gold bid is unusual, and it means the dashboard's +2 score for gold is simply wrong for this war. I did not rescore it. The whole force of the staircase comes from committing to predictions before seeing data. Quietly relabeling the assets that misbehaved would hollow that out.

Defense stocks: Scored +2 (strong predicted winner). Defense looks like a clean contrast until you change the measurement window. Over the first week, defense was roughly flat in February (+0.7%) and sharply negative in July (-5.4%). Over twenty days, the ranking reverses: -5.9% in February against +1.4% in July. A channel that returns whichever answer the window hands you is not evidence, so I am not treating defense as a leg of the February-versus-July argument.

What did work was purely physical-barrel: WTI +16.7%, Brent +16.1%, Chevron +11.2%, Exxon +7.5%, the energy sector ETF +6.5%. The demand-side hit showed up early, with airlines averaging -6.3% over the first week, but it had faded to nothing by week four (JETS -1.4%, Delta actually +1.3%, neither significant). Beyond the barrel trade, there was very little.

February was "the world is dangerous." July was "barrels are tight for a few weeks." That conclusion rests on two things that hold cleanly: the VIX contrast and the sheer difference in oil's reaction size, which runs between a third and four-fifths of Round 1 at every horizon I tested. Gold and defense, which I first reached for as supporting evidence, do not show a clean February-versus-July split. The thesis is narrower than the four-channel version I started with, but the two legs that hold are strong ones.

One finding that emerged while checking the rest: energy equities (Exxon, Chevron, XLE) tracked crude from day one in July, whereas in February they sat flat through the first week while oil futures spiked as much as 32%. The equity market transmitted the oil move immediately the second time instead of lagging it by weeks. In that specific sense, the second reaction was faster.

The giveback

I measured how much of each market's abnormal move reversed before any de-escalation news arrived. This separates "markets calmed down because they saw a headline" from "markets calmed down on their own."

For Round 1, the cutoff is April 7 (the ceasefire was announced April 8, on trading day 27). For Round 2, the cutoff is August 3 (the Oman talks were announced August 4, on trading day 21).

In February, oil had given back nothing when the ceasefire arrived. It was still climbing. In July, oil gave back nearly half its move while the fighting was still active and no deal was in sight. WTI peaked at $92 on July 23 and was back to $80 by August 3. Roughly three-quarters of the total giveback happened before the Oman announcement.

The honest caveats

This comparison has real limitations.

The two shocks may not be the same size. This is the most important limitation and the one price data cannot settle. Round 1 began with the Strait of Hormuz closed. Round 2 began with strikes on 80+ targets and a naval blockade reimposed. Those are related but not identical supply events. If February physically stopped more barrels than July did, then a smaller July reaction is markets correctly pricing a smaller disruption, not markets becoming desensitized to the same one. Every number in this post is consistent with both readings. Separating them requires physical data I have not brought in: tanker transits through the strait, loadings out of Kharg Island, floating storage.

The February baseline is stale. The market model was fitted on pre-February data. By July that is roughly 10 months old. Betas may have drifted, which means Round 2's abnormal returns are measured against a baseline that may no longer be accurate. One concrete symptom: China (FXI) posted +13.4% abnormal at twenty days on an exposure score of zero, a larger move than Chevron's. A market with no plausible war exposure moving that much is a reminder that the single-factor baseline leaves a great deal unexplained. The direction and rough magnitude of these results hold. The second decimal does not.

February's "0% giveback" is partly because it never got the chance. The ceasefire arrived on trading day 27 while oil was still climbing. "R1 gave back nothing" partly means the peace news came before markets had time to reverse on their own. The defensible comparison is: within each round's available pre-de-escalation window, Round 1 showed no self-reversal and Round 2 showed roughly 50%.

Round 2 has only 24 trading days of data. The day-15 rank correlation (0.58) rests on a handful of observations. It is directionally interesting but not as reliable as the Round 1 findings, which had more data behind them.

Natural gas is doing its own thing. Natural gas showed -22% abnormal in Round 2 despite being scored +1. This is almost certainly storage and weather dynamics, not the war. It drags the Round 2 staircase down and is worth flagging rather than explaining away.

The exposure scores are less reliable than they look. At twenty days, 10 of 20 directional assets moved against their own predicted score. Gold and natural gas are the most visible misses, but they are not the only ones. The staircase works because oil and the correctly-negative names (Turkey, emerging markets, Treasuries) carry the ordering. The pattern is narrower than "markets sorted themselves by war exposure" implies. It is more accurately described as: the oil trade and the oil-victim trade sorted correctly, and enough else went along for the ordering to hold.

This post is about data, not about the human costs of the conflict, which are real and separate from anything markets can measure.

What this means

The same war, restarting under nearly identical oil-price conditions, produced a fundamentally different market response. The physical-supply trade showed up both times. The fear premium that accompanied it in February did not show up in July. And the physical trade itself was unwound by markets on their own timeline, without waiting for diplomacy.

There are two readings, and this analysis cannot fully separate them.

Markets learn. The first time a geopolitical shock hits, it carries genuine uncertainty about how bad it could get. The second time, the range of outcomes is known, because markets have already lived through one round. They price the physical disruption (oil, shipping, fuel costs) and skip the fear premium.

Or the second shock was simply smaller. A blockade is not a closed strait. If July threatened fewer barrels than February did, a smaller reaction is just accurate pricing, and there is nothing to explain.

The evidence tilts toward the first. A market calmly repricing a smaller disruption has no particular reason to unwind half that repricing while the disruption is still ongoing, and no reason for the fear gauge to stay flat throughout. But the tilt is not decisive, and I would want tanker-transit data before claiming more than that.

What I can say without qualification is narrower, and still worth saying. The first time, markets believed the disruption would last. The second time, they priced it and stopped believing while it was still happening.

---

This piece accompanies the Ripple Effect dashboard, which tracks 28 financial assets through the US-Iran conflict. The Round 1 analysis is here.