I started Geekpad at school, back when that name seemed like a great idea, as a blog about almost everything: tech, sport, COVID. Two things happened since. I realised the part I enjoyed most was looking at what markets were doing, building a view from the data, and putting a number on it. And I realised “Geekpad” was not a name anyone would take a bond call from. So version two has a new name and does one thing properly: markets.
Every call on Marketpad is backed by its own modelling, not just a hunch and a headline. For the first piece, that meant:
- A debt-dynamics simulator calibrated to the IMF’s latest projections, to stress-test how yields, growth and fiscal policy move debt paths
- Splitting Treasury yields into expected Fed rates and term premium using the New York Fed’s ACM model
- A fair-value model for France’s spread over Bunds, plus a backtest of how similar premiums have behaved since 2005
The modelling also changed my mind once: a UK call I was confident about didn’t survive a better-specified model, so it never made the call log. That’s the point of testing.
The call log is the part I’m most excited about, and the part most likely to keep me honest. It’s easy to be right in hindsight. It’s harder when the entry level is timestamped and the stop is written down in advance. Closed calls stay up, wins and losses alike.
The first piece is out: “Discipline isn’t enough: why bond markets reward credibility, not austerity”. It asks why markets punish some indebted governments more than others, regardless of their debt-to-GDP ratios, and argues the answer lies in the path and credibility of a country’s fiscal plan, not its level of debt.
It also opens my first two calls:
- US 2s30s steepener, opened at +75bp
- France to keep trading wider than Italy into the 2027 election, opened at +45bp
Pushback is very welcome, especially from anyone on the other side of either trade. Get in touch.