Crude oil changes sides when GCC trading calendars align

A risk desk can use the broad tail result here, but should be wary of the oil signal. Spillovers among Gulf equities, energy futures and gold are higher in both tails than at the median on either trading calendar. Yet 8 of 30 node classifications change sign when the calendar changes, about one in four, and the authors acknowledge that several markets are calendar-sensitive.

Ten series on different clocks

Mgadmi, Belhouichet, Gherghina, Moussa, Regaïeg and Hachicha study ten daily series from January 2, 2022 to February 27, 2024. The six GCC equity benchmarks are TASI (Saudi), QASI (Qatar), ADXI (Abu Dhabi), MSM-30 (Oman), BASI (Bahrain) and the FTSE NASDAQ Kuwait index. They add continuous front-month futures for WTI, natural gas and gold. The remaining series is the Caldara-Iacoviello daily geopolitical risk (GPR) index, a count of newspaper coverage of geopolitical tension.

Their method comes from Ando, Greenwood-Nimmo and Shin. They fit a quantile VAR(1) at τ = 0.05, 0.50 and 0.95, then use a generalized forecast-error variance decomposition at a 10-day horizon to assign each series' forecast variance to shocks elsewhere in the system. The total connectedness index (TCI) averages the shares attributed to other series. NET subtracts what an asset receives from what it sends; a positive NET makes it a net transmitter. The paper's claimed addition is a single quantile system for GPR, Gulf equities and commodities, rather than the mostly mean-based equity-oil links studied earlier.

The clocks matter. Gulf exchanges trade Sunday to Thursday, futures trade Monday to Friday, and GPR appears every day. For the main sample, the authors carry the last observation forward on a continuous calendar. They call this LOCF; it yields 785 returns. Their second sample, "All-Markets-Open", contains the 376 days on which all ten series trade.

What holds across calendars?

The aggregate tail result holds. LOCF gives TCI of 72.35% at the median, 82.99% at τ = 0.05 and 79.97% at τ = 0.95. On common days, the corresponding figures are 73.41%, 86.76% and 86.14%. Both tails exceed the median in each sample. In a jitter audit of the zero-return ties created by LOCF, tail TCI stays unchanged across 15 runs. Median TCI ranges from 71.45% to 73.97%.

Some individual roles persist. TASI sends more than it receives in all six quantile-calendar cells; ADXI, gas and gold receive more in all six. GPR sends more in five of six. Its exception is the LOCF median, at -4.41. The authors are careful with the gold finding, saying its receiver role "should not be construed as evidence of safe-haven characteristics".

Across calendars, signs agree in 22 of 30 cells, or 73.33%.

One cell in four changes sides.

Oil's reversal

WTI leads the lower-tail transmitters under LOCF, with a NET of +102.31. Restrict the sample to common days and it becomes a receiver at -15.34. GPR moves in the opposite direction, from +40.80 to +113.21. The abstract acknowledges the oil reversal as "sensitive to the calendar-alignment procedure". The authors defend the broader tail pattern despite changes in individual roles, and the TCI figures support that defence.

The LOCF median table raises a harder question about direction. It assigns 61.76% of WTI's forecast-error variance to TASI and just 0.76% to WTI's own shocks. A US crude futures contract drawing 61.76% of its variance from a Saudi index while accounting for 0.76% itself is hard to interpret economically.

Gulf exchanges close hours before WTI settles. A lag-1 VAR fitted to daily closes can turn that timing gap into apparent directional spillover; keeping only common days leaves the gap in place. The paper describes commodity entries as having "dominant diagonal values", although the median LOCF diagonals for oil, gas and gold are 0.76%, 0.14% and 15.00%.

QASI changes sign at every quantile when the calendar changes. The authors call its balance "marginal net directional balance". Under LOCF, however, QASI's NET is +75.62 at the median and +60.57 in the lower tail. The upper-tail +7.82 is the marginal figure, while the sign flips at all three quantiles.

Thin tails and uncertain gaps

At τ = 0.05, the 376 common-day returns supply about 19 tail observations for 11 coefficients per equation (our arithmetic). A 200-day rolling window supplies about 10. The paper's text gives no confidence interval for any TCI, leaving the difference between 72.35% and 82.99% formally untested there. Rolling median TCI averages 82.0% under LOCF, compared with a static 72.35%. We did not find an explanation for that gap.

The authors themselves list the short sample, its limited extreme-tail observations and the missing macro controls among the limitations.

A stress gauge before a hedge

The paper promises "actionable insights for cross-asset hedging". Its estimates describe shares of 10-day forecast variance within the sample. On common days in the lower tail, GPR's NET of +113.21 makes it the largest net sender, ahead of TASI at +57.32. Whether trading on GPR would earn or save money remains untested: GPR is a log-differenced article count with daily variance of 2354.98, versus 4.56 for WTI. We did not find a portfolio or out-of-sample test anywhere in the paper.

We could not run one either. We lack GCC index data and the daily Caldara-Iacoviello series needed to rerun it. Replacing GPR with a news-derived proxy would change the variable.

Static tail TCI of 80% to 87%, against 72% to 73% at the median, offers a reasonable stress gauge for tightening cross-asset limits in Gulf books. The node rankings are too calendar-sensitive to size hedges from. Oil retaining a positive lower-tail NET after Gulf and futures prices are sampled at the same clock time would change that assessment.