State a view — a European recession, an oil shock — and let the model carry its consequences to every country and every asset class. Then read what that view costs in plausibility.
Ageing lowers the equilibrium rate of interest, and so pushes central banks closer to their lower bound. Here is by how much, across the OECD, under demographic pressure alone — every economy anchored to the same starting point.
A purely demographic counterfactual: everything that is not demographic stays at its 2024 value, and every economy starts from the same position — that of the United States in 2024.
| Economy | 2024 | 2050 | 2075 | Spells / century | Length (qtr) | Natural rate |
|---|
Move the slider and Korea travels from 8 % to 77 %. The ranking never moves. That is what a model gives you — not a number, a robust ordering.
Notice who is at the top. It is not the oldest economies but the fastest-ageing ones: Korea, Chile, Costa Rica, Colombia, Türkiye. The indicator is driven by the change from today's position, and economies that are already old have paid most of their bill.
Ageing also raises the frequency of the bound far more than its depth: between 2024 and 2075 Korea's spell rate rises twentyfold, its expected duration only threefold. A central bank in an ageing economy should size its toolkit for frequency, not for depth.
From public data to an exported allocation, with no detour through a spreadsheet.
Economies linked by their trade flows, plus a global block — oil, commodities, world equity. A view on Germany propagates to its partners through the structure of the model itself, not through expert judgement.
State absolute, relative, averaged or cumulated views, each with its own confidence. The model returns the conditioned distribution — and the price of your assumptions, view by view.
A cohort engine on United Nations data, coupled to an overlapping-generations model. Demography stops being scenery: it puts a floor under the interest rates of your scenarios.
Each asset class is tied to the variables of its own region. And the link runs both ways: a view on European equities moves the distribution of European GDP, exactly.
The classical constructions — mean-variance, minimum variance, equal risk contribution — and a proprietary family that equalises contributions to drawdown, the risk that actually triggers redemptions.
Every result carries its data vintage, model version, random seed and fingerprint. It replays identically two years later, and exports to your portfolio system.
The question nobody puts a number on. Geodesis answers it with one — and decomposes it view by view.
Published results, read straight from the public tier. Nothing here is computed on demand — which is why this page stays up when the engine does not.
No vintage has been published yet.
Showing the latest of 0 published results.
The dividing line is not the data, it is the computation. Reading is free — and complete.
The free tiers read results that are already computed and published at each vintage. They touch neither our database nor our compute queues — and stay up even when the engine does not.
Create an account →Each one names an assumption that, ignored, makes a result read as something it is not — a floor for a forecast, a plausibility for a probability of occurrence. They are served by the platform rather than written into this page, so that the site and every export say exactly the same thing.
The published tier is unreachable; the limits are also printed in every export.
What the research programme withdrew from its own results, why, and what replaced it. Any prediction withdrawn from the corpus is withdrawn from the product, and the reason is published here, without an account. It is not an admission of weakness: it is the most convincing proof a model-validation team can receive that a result still on display has been through a discipline of refutation.
The published tier is unreachable; the register
is served at /public/v1/withdrawals.