Gold Lost the Line. The Dollar Kept the Case.

On Monday newswires carried reports of possible official-sector selling: Kazakhstan’s central bank, according to an interview with its deputy governor carried by several outlets, aims to cut gold’s share of its reserves by about three percentage points, to roughly 75 per cent, by the end of the year, half through options and half through physical sales, with high Treasury yields cited as the reason (newswires, 12 October 2026). The desk has not seen the original interview, and no formal policy announcement or completed sale has been independently verified. The same reports quote the deputy governor as saying a gold price below $4,000 could bring the share to target on its own, so a lower percentage need not mean outright selling. The World Gold Council reported that Kazakhstan added seven tonnes in August, taking gold to about 79 per cent of its reserves (World Gold Council, central bank gold statistics, 6 October 2026). Separately, newswires reported a sell-side forecast for gold to average $4,000 an ounce in the fourth quarter of 2026 (newswires, 12 October 2026). Gold closed Monday near 4,171 on the desk’s feed and traded near 4,188 in Asia on Tuesday. The Treasury cash market was shut on Monday for Columbus Day, so Friday’s is still the last Treasury settle, and Tuesday’s session is live; nothing in this note is scored on it. Gold closed the week near 4,194 on the desk’s feed, inside a one-month range of 4,067 to 4,400, and has not closed above 4,449 since the stab on 4 September (desk feed, 4 September to 9 October 2026).

The rates market had already done the work the headline describes. From 20 August to Friday’s settle the two-year Treasury rose from 4.19 to 4.80 per cent, sixty-one basis points, while the ten-year rose fifty-five to 5.24 and the thirty-year thirty-seven to 5.60 (Treasury par yield curve, 20 August and 9 October 2026). On matched dates, 20 August to 8 October, the latest day with both series published, the ten-year real yield went from 2.35 to 2.87 while the ten-year breakeven went nowhere, 2.34 then 2.35, against a nominal ten-year up fifty-three basis points to 5.22 (Federal Reserve H.15 and Federal Reserve Bank of St. Louis breakeven series, 20 August and 8 October 2026). The breakeven printed 2.33 on Friday 9 October; the real yield for that day had not been published when this note was written, and there is no 12 October observation for either series because the bond market was shut. Almost the entire increase in the ten-year is accounted for by higher measured real yields, with inflation compensation broadly unchanged, so the evidence favours a real-rate repricing over an inflation-expectations shock. Gold’s decline is consistent with the rising opportunity cost of holding a non-yielding asset, although the relationship alone cannot establish how much of the price move higher real yields caused.

The desk wrote the gold line in August. Here is how it scored.

Every line in this table was in print before the tape answered it. The 20 August note set out both directional cases for gold and named the combination that would decide between them. That combination has been printing since early September, and the desk logged it as live on 11 September. The same week’s auction and buyback tests were scored in Duration Cleared. The Crowd Did Not. (12 October) and are not repeated here.