New Zealand is a small, open economy: its long-term interest rates, fuel and currency are set as much in Washington, the bond market and the Black Sea as in Wellington. This model maps the feedback loops involved, reads them against the latest data, and tests published commentary against the whole system rather than the slice a single article covers.
What the latest data says the system is doing. Each statement is generated from the readings, not written by hand, so it changes when the data does. Every loop is scored gaining, fading, mixed or steady from the indicators wired to it.
| Indicator | Latest | As at | Change | Trend | Signals loops | Source |
|---|
Published commentary run through the model. Each test has a review date; once its prediction can be judged it is scored held, partly, failed or void and moves to the archive, which keeps a public scorecard. A test can be reopened for a later round, such as checking election plans against the first Budget. Each test records which loops a piece uses, which relevant loops it leaves out, whether it stops at New Zealand's border, what it implicitly predicts, and the indicator that will show whether it holds. The test judges the analysis, never the author's motives.
The point of the model. A common shortcut is to treat New Zealand as a fishbowl, explained by the OCR and local politics alone. The loops show a split instead: one channel the Reserve Bank steers, and several it can only lean against.
Most NZ mortgages fix for one to two years and price off short swap rates, which follow the OCR (B13). Over household mortgage costs, the RBNZ has a firmer grip than the Fed has over US 30-year mortgages. Repricing is also faster, because fixes roll within about 18 months and there's no 30-year lock-in to cushion it.
NZ 10-year yields follow the US benchmark through offshore holders of NZ bonds (B15), softened by bond programme management (B12). Fuel comes through world prices and freight, since there's been no domestic refinery since 2022. The Fed–RBNZ rate gap moves the NZD (B14). The RBNZ can lean against all of these but can't set them.
R1, R2: heavy sovereign borrowing, long yields at multi-decade highs
Offshore holders sell or buy NZ bonds alongside US Treasuries (B15). The bond programme can soften the move (B12).
Feeds the Crown's interest bill (R14) and longer-term and wholesale funding
B2: hiking against inflation
A wider US–NZ rate gap weakens the NZD. A higher OCR supports it (B14).
Fuel and imported goods cost more in NZD
R8, R12, B3
Bought from Asian refiners at world margins plus freight
Diesel runs through freight into almost every price. Short-run demand barely moves.
Fuel, the NZD and domestic demand
B13: the RBNZ's own loop
Repricing within about 18 months, then household spending
R8: grain and oil-seed routes at risk
Importers pay more for replacement supply
NZ sells food, so higher world food prices lift export receipts and support the NZD
| Indicator | Latest | As at | Change | Trend | Source |
|---|
Peter Senge's system archetypes are recurring structures that turn up across very different systems. Seven of them are at work here. Each card shows the generic structure, where it sits in this model, and the leverage point it points to.
Commentary often reads as a one-way spiral. With its balancing loops in place, the same structure cycles, because loop dominance shifts as delays play out. Each turn starts from a higher debt base, so the shape is closer to a ratchet than a circle. Click a phase to see which loops dominate it.
B11 flight to quality beats R10 crowding out and R13 flight from sovereigns. Weak data sends money into long government bonds, yields fall, cuts land on a falling long end, and mortgages and confidence repair. Most past recessions resolved this way.
Signal: the 30-year yield falls on weak data and stays down after the first Fed cut.
R1, R10 and R13 beat B11. Recession widens deficits, fear money goes to private and hard assets instead of government paper, and long yields hold up even as the Fed cuts (R4). The exits are B10 (inflation shrinks the debt, savers pay) or B6 forced by the market.
Signal: long yields and inflation expectations rise after the first cut, while equities, credit or gold take the bid.
The statement "Where fear money is going" on the Reading now tab is the live version of this fork. Until the Fed cuts into a weakening economy, the data can lean one way but can't decide it.
Most economic commentary works inside a boundary: one market, one country, one cause, often one villain. That boundary is where its blind spots live. This model draws the boundary wider. It holds the feedback loops that connect government borrowing, central banks, energy, war, households and New Zealand, and asks of each new commentary: which loops does it use, which does it leave out, and what does the latest data say?
No model is complete. The aim is to cover most of the system most of the time, and to say plainly what sits outside it.
An arrow means a change in one variable pushes another. A + means they move the same way and a − the opposite way. A ‖ marks a delay. Follow the arrows round a closed loop. If the loop has an even number of − signs it is reinforcing (R) and amplifies whatever is happening, in either direction. If the number is odd it is balancing (B) and pushes back toward some limit or goal. Cycles happen because reinforcing loops run until delayed balancing loops catch up.
Loops are marked by origin: from the commentary that prompted them, added by the model because the commentary left them out, or a hypothesis under test.
Each indicator is wired to the loops it measures, with the direction that means the loop is gaining. Over the indicator's window, usually 91 days, a move past its threshold counts as up or down. A move in the loop's direction scores as gaining, the other way as fading, and both together as mixed. An indicator whose latest observation is older than its staleness limit is flagged, never hidden. A failed data fetch keeps the last good value and marks it stale, and nothing is estimated.
This scoring is deliberately simple. It shows which loops the data is currently feeding, not their size, and it can't see a loop no indicator is wired to.
Most series come from FRED (Federal Reserve Bank of St. Louis), refreshed by script. The dairy index comes from Global Dairy Trade's auction feed, and NZ pump prices from Gaspy's national averages. Brent futures come from World Oil Monitor's daily history, which relays Yahoo Finance. The RBA cash rate and Australia's 10-year yield come from the Reserve Bank of Australia's statistical tables. Readings none of these carry are entered by hand with their source and as-at date: the OCR, NZ CPI, the NZ 2-year swap, ISM survey indices, the Conference Board index, gold, and the net migration loss to Australia (Stats NZ, twice a year). Monthly series such as the NZ 10-year yield lag by a month or more. The as-at date beside every reading says how fresh it is.
The model is US-weighted, because that's where the global benchmark is set. Australia is the second channel: its bond market, its central bank, the parent banks behind most NZ lending, and the jobs that draw New Zealanders across the Tasman. China, Japan and the UK are left out by design. Loop status measures direction, not strength.
Built for Strategic Kata with AI assistance, which is what made building and testing a model this size practical for one person. The structure, the tests and the verdicts are reviewed by the author before they are published. Nothing here is financial advice.