Bond timing rules as one testable procedure

One input for the whole procedure

The archive describes a mechanical trading system that switched between long-term and short-term bond exposure. The sole market-state input was the 12-month rate of change in a gold-mining share index.

The case treated inflation perception, not the coupon itself, as the driver that made long-term bond prices more or less attractive over a subsequent 12-month window.

Weekly readings in the archive sample

Across 890 weekly readings from 1978 through late 1996, more extremely negative 12-month gold-mining-share change was associated with a higher frequency of higher bond prices 12 months later.

When that 12-month change was -30 percent or lower, subsequent 12-month bond prices were higher after 75 of 82 weekly readings.

When that 12-month change exceeded 100 percent, subsequent 12-month bond prices were lower after all 55 weekly readings in the sample.

Entry, abstention and listed dates

The stated rule-based entry and abstention rule was to hold long-term bonds when the 12-month gold-mining-share change was negative and to avoid long-term bonds when that change was positive.

A later version of the same mechanical procedure listed discrete buy and sell dates against a continuous three-contract Treasury-bond futures average used as a price proxy.

A bound set at entry

The case required a percentage stop-loss order for funds, or a predefined maximum-risk threshold for futures, to be set at entry rather than assuming any signal would eventually recover.