Holding the S&P 500 rather than a foreign equity basket earned about 3.45 percentage points of relative return over the two trading days after the Iran escalation. Both sides lost money. The S&P 500 fell 0.91% in dollars over 2 and 3 March; the synthetic comparison basket fell 4.36%. The paper establishes that gap after the event and tests no trading rule. Its finding is relative resilience in a falling market.
Yoshimori's model meets the March 2026 Iran shock
Yoshimori's four-part model makes index prices revert toward fundamental value. Geopolitical shocks raise perceived risk, which then subsides as news arrives. Investors move between optimist and pessimist states. An imitation term lets sentiment settle into either of two states when herding becomes strong enough and fear roughly balances information. A sufficiently large shock can then trap investors in pessimism.
Oil affects fundamental value through a net coefficient that declines as a country's net energy exposure increases. Domestic producers gain as energy users pay more. The third proposition carries the allocation argument: when two markets have equal sensitivity to common risk and sentiment, those effects cancel from their price gap. The remaining difference reflects energy exposure. The parameters are illustrative calibrations (κ=0.3) and remain unestimated.
The test concerns early March 2026. After averaging about 118 in February, the daily Geopolitical Risk (GPR) index reached 500.8 on Monday 2 March. By 6 March, Brent stood 34% above its 27 February level. A constant-mean event study puts the S&P 500's event-day abnormal return at -0.01%. Over five days, the cumulative abnormal return is -2.26%. Its 95% interval spans -7.43% to +2.91%, which the paper acknowledges "does not rule out a sizeable decline".
For the cross-market comparison, Yoshimori uses a synthetic control, a weighted basket fitted to follow the S&P 500 before the shock. The basket combines dollar returns on the S&P/TSX, All Ordinaries, Hang Seng, Nikkei 225 and Euro Stoxx 50. Fitting daily returns over the preceding 60 days assigns weights of 0.334 to Canada, 0.283 to Australia, 0.248 to the euro area, 0.135 to Hong Kong and zero to Japan. The resulting two-day gap of 3.45 points exceeds 99.6% of the gaps at 455 pseudo-event dates (in-time p 0.004).
A Canada-heavy basket cuts the 3.45-point gap to 0.98
Across fitting windows of 20, 40, 60 and 120 days, the gap stays between 3.45 and 3.90 points. Extending the fit to 250 days puts 0.97 of the basket in Canada. The gap then shrinks to 0.98 points (p 0.26).
Yoshimori acknowledges the limitation directly. Against Canada alone, he writes, the gap is small and insignificant. The finding therefore cannot describe US outperformance against foreign markets generally. Canada's status as a net oil exporter agrees with his hypothesis, although he concedes that this "does not by itself resolve this sensitivity".
His sole defence of the 60-day window concerns the April 2025 tariff turmoil included in the 250-day fit. US and Canadian equities traded on related news during that episode. He accordingly interprets the 60-day result as "a comparison with a basket weighted towards oil-importing markets". Excluding Canada raises the gap to 4.61 (p 0.009).
Japan's zero weight means the Nikkei's decline contributes nothing to the 3.45. The euro area's 0.248 weight matters: removing it reduces the gap to 2.29 (p 0.037). Over the two days, the Euro Stoxx 50 fell 8.02% in dollars and the Nikkei fell 5.51%.
The inference carries less weight than the 0.004 might suggest. In the in-space placebo comparison, the US ranks 2 of 6, or roughly 1/3. Statistical significance comes from the in-time placebos, whose two-day windows overlap. And fitting daily co-movement left a trend mismatch: over the fitting window, the S&P 500 trailed the basket by about 10.6 points. The paper's drift adjustments increase the estimated gap to 3.80 and 3.88.
How much can two exporting markets and four importers tell us?
The net oil exporters lost an average of -1.38% over the two days, against -5.33% for importers. This ordering supplies the paper's evidence for an oil channel. Australia complicates the story. An oil importer and an overall energy exporter, it fell 4.73%, more than Hong Kong's 3.07%. The Australian figure also includes 27 February because the series is missing that day.
Changing the horizon changes the ranking. Over the full month, the difference between exporters and importers classified by oil contracts from 3.95 to 1.62 points. Classifying them by total energy instead expands the difference from 3.05 to 4.17. With only two markets on one side, the preferred classification depends on the window.
The sector check is more revealing. Removing energy from the S&P 500 using weights of 2 to 5% leaves a gap of 3.47 to 3.49 points. Removing it from the TSX too, using 13 to 21%, produces 3.54 to 3.66. The model gives producer earnings the job of cushioning exporters against an oil shock. Taking those producers out barely changes the result.
The paper reaches the same conclusion: differences in index weights for energy producers "cannot explain the US outperformance". The protection lay somewhere in the other 95% or more of the S&P. Yoshimori acknowledges that this exercise does not establish a channel through energy-using firms. His pre-event regression assigns oil changes a coefficient of +0.017 in US returns (p 0.194).
The energy explanation remains a hypothesis: the paper says the independent effect of net energy exposure was "not measured".
Lagged VIX and oil offer no return signal
The VAR gives no predictive result from lagged volatility or oil. Lagged VIX changes have a coefficient of 0.046 for next-day S&P returns, with a standard error of 0.039. The estimate is insignificant. Lagged oil changes enter at 0.024 (0.022), and the system's largest companion root is 0.153.
Almost the entire volatility-return relationship occurs on the same day. Return and VIX residuals have a correlation of -0.843. A contemporaneous regression using the VIX change has a coefficient of -0.405 and an R² of 0.713. Much of that fit follows from the VIX being priced from S&P options. The paper recognises this by describing the regression as a descriptive decomposition.
The relative trade faded, too. On 23 March, the raw gap remained at 4.06 points. It had dropped to 0.98 by 30 March (p 0.628). The S&P fell 3.8% over the final three sessions, finishing 8.10% below its 27 February close. Mean-adjusted gaps remain at 4.68 and 4.96, though neither exceeds placebo noise at 21 days (p 0.165 and 0.145).
Of that 8.10% decline, the pre-event regression leaves -6.16 points unexplained. The paper offers a sustained supply-driven rise in Brent, which exceeded $120, as a possible explanation. It says the data cannot identify that channel.
We could not test these results ourselves. Our data contains neither the daily Caldara-Iacoviello index nor the five foreign indices. We also lack Brent. Substituting WTI or US-only equities would alter the mechanism being tested.
The paper supplies a prior for cross-country risk allocation. On an oil spike following a Middle East shock, its evidence favours underweighting Europe and Japan against US and Canadian equities. The paper reports that the US lead over the euro area and Japan persisted to 30 March, with the S&P 500 down 8.10%. Australia and Hong Kong had meanwhile moved ahead of the US. Turning this prior into a sizing rule would require a pooled cross-section across many GPR spikes, which Yoshimori identifies as his own next step.