# About this document

This document collects the essential reports from the `financialmarkets` folder in one place, so that a reader can see what was done and what the results show. Except for this introduction, each chapter is a document from the folder, reproduced unchanged. The source file is named under each chapter title.

## The question

**Where are US long-term interest rates heading from 2026, and how will we know if the regime has changed?**

On 1 October 2026 the US 30-year Treasury yield reached 5.67%, its highest since 2002. In May 2002 a similar level looked like the floor, yet long rates then fell for 18 years. The work asks how to avoid making the same mistake in reverse.

## What was done

1. **A framework** (Chapter 2). Three AI-written notes and a yield chart were combined into a four-layer method:
    - split the yield into expected inflation, real rate and term premium;
    - score seven structural drivers;
    - classify the regime as A (disinflation), B (higher real rates, anchored inflation) or C (fiscal or inflation);
    - anticipate the policy reaction.

    The framework also defines a signposts dashboard and a five-metric "regime vote".
2. **A data room** (summarised in Chapter 1). Code that fetches, checks, stores and charts every statistic in the framework from nine sources (FRED, the New York Fed, the Treasury, Japan's Ministry of Finance and others), refreshed daily, with a written contract and automatic checks for each series.
3. **A forecast journal** (Chapter 3). Dated regime probabilities and 12-month yield ranges, the reasons behind them, and the readings that would change the view, to be scored on 1 October 2027.
4. **A back-test on 2002** (Chapters 4 to 6). The dashboard was rebuilt month by month for 1999–2006, using data as it was known at the time, to ask whether the framework would have flagged the disinflation regime that followed. The results were then applied to the 2026 view.

## What the results show

- **The forecast (1 Oct 2026).** Over three years, the data-based view gives B 55%, C 30% and A 15%; the gut view gives B 50%, C 35% and A 15%. The 10-year yield on 1 Oct 2027 is expected in 4.25–6.50% (central 5.4%) and the 30-year in 4.90–6.75% (central 5.85%), each about 3 times in 4. The sell-off so far is led by real rates and the Fed rather than by inflation fears, which is the signature of Regime B.
- **The headline vote is not making a call.** September 2026 reads A1 B2 C0, after B calls in May to August.
- **The 2002 back-test: the vote failed and the drivers worked.** In May 2002 the vote read A1 B2 C2, and it made no A call in any month from Nov 2001 to Dec 2006. The structural drivers scored −5 out of −7, which was the right direction. The range method put the actual May 2003 10-year yield (3.37%) below every historical case.
- **The live record agrees.** Running the same vote from 2003, it called A 13 years after the 2002 turn and called A again at the start of the 2021–22 inflation regime. It never called C in 285 months.
- **Confidence for 2026.** Trust the framework to describe where we are and to organise the thinking. Do not trust its headline vote to call a turn. In practice this means: lead with the drivers and the direction of change; keep regime probabilities near 50% (at most about 55%); treat yield ranges as too narrow; and size positions to survive the regime you did not expect.

## How to read it

| Chapter | Read it for |
|---|---|
| 1. How this folder was built | What was built, in what order, and why |
| 2. The framework | The method: layers, dashboard bands, scenarios and decision rules |
| 3. Forecast journal, 1 October 2026 | The calls and their reasons, plus an appendix explaining each watch metric |
| 4. How the 2002 back-test was done | Data problems and their fixes, validation, limitations |
| 5. The 2002 back-test report | Full results, month-by-month working and charts |
| 6. Confidence for 2026 | What the back-test means for the current view |
| Appendix A. Glossary | Terms and abbreviations |

For a 15-minute read, take this introduction, then Chapter 6, then section 1 of Chapters 3 and 5.

**Not included:** the step-by-step update guide, the three source notes, the one-page version of the journal, the change log, and the data room's code and data. All of them are in the folder.

*Not financial advice. This is an analytical framework for personal decision-making.*