Task #4 v0.1 Asset Pool Design Note

Status: draft after Kaite 2026-05-17 comments. Purpose: re-ground the all-weather asset pool from macro logic before more backtests.

1. Design Principle

The domestic all-weather portfolio should earn domestic broad-liquidity expansion and nominal economic growth/inflation, not short-term statistical artifacts. Asset inclusion must pass two tests:

1. Macro explainability: why should this asset respond to growth, inflation, credit, rates, or policy liquidity?

2. Empirical validation: after the macro relation is understood, test whether realized behavior matches the expected conditional pattern.

If macro logic cannot explain a relationship, statistical fit alone is not enough.

2. Risk Definition

Current v0 uses asset return volatility as a proxy for risk. This is insufficient for v0.1. The better abstraction is risk exposure to macro drivers:

- Growth beta: equity/corporate profit sensitivity.

- Inflation beta: commodity/gold/pricing-power sensitivity.

- Duration beta: real/nominal rate sensitivity.

- Credit/liquidity beta: credit spread, financing condition, policy put, broad liquidity.

- Policy/structural beta: assets supported by policy direction or scale expansion.

Risk parity should balance these drivers, not mechanically overweight the lowest-vol ETF.

3. Regime Framework

Keep the four-regime growth x inflation framework as a first-order map:

Regime labels are monitoring context. Portfolio allocation should not become one-regime tactical betting.

4. Asset Pool: Stable Core Categories

The pool should emphasize large, important, scalable domestic asset categories, plus policy-supported categories with large scale-growth potential.

Equity Growth Bucket

Role: growth/profit beta and broad liquidity upside.

Candidate proxies:

- CSI 300 ETF: large-cap/core equity.

- CSI 500 ETF: mid-cap equity.

- CSI 1000 ETF: small-cap/high-beta equity; include point-in-time after ETF availability.

- ChiNext ETF: growth/innovation equity; include point-in-time.

- STAR 50 ETF: hard-tech/strategic industry equity; include point-in-time.

Regime: mostly R1/R2, with sub-bucket diversification by market-cap/growth style.

Duration Bucket

Role: recession/deflation hedge, rates-down beneficiary.

Candidate proxies:

- 10Y Treasury ETF.

- 5-10Y / active Treasury ETF.

- 30Y Treasury ETF if history/liquidity acceptable, point-in-time.

Regime: R4.

Credit / Carry Bucket

Role: domestic liquidity expansion, credit spread/carry, policy support. This is a real domestic asset category and should not be removed merely because volatility is low.

Candidate proxies:

- 城投债 ETF.

- 政金债 ETF where available.

- High-grade credit ETF where available.

Treatment question for v0.1:

- Do not let credit/carry dominate a growth sleeve through inverse-vol.

- Preferred treatment: its own credit/liquidity sleeve or sub-bucket with explicit risk budget, not hidden inside R1.

- Test variants: credit as R1 sub-bucket vs separate credit/carry sleeve.

Commodity / Inflation Bucket

Role: inflation and supply-shock protection. Current v0 is incomplete.

Candidate proxies:

- Gold ETF: monetary/inflation/geopolitical hedge.

- Non-ferrous / metals: industrial inflation and global cycle.

- Energy-chemical futures ETF (159981.XSHE): must include candidate; important macro asset class.

- Agriculture / soybean meal ETF: food/feed/agricultural supply shock.

- Coal/steel ETFs may be sector equity proxies rather than pure commodity; include carefully as candidate, not default.

Regime: R2/R3 depending on sub-asset.

Defensive Equity / Dividend Bucket

Role: lower-beta equity, cash-flow/valuation anchor, recession resilience.

Candidate proxies:

- Dividend ETF.

Regime: R4 and possibly R3 as defensive equity.

Cash / Collateral Bucket

Role: liquidity buffer, collateral/funding, operational reserve. Not a macro return sleeve.

Candidate proxies:

- Money-market ETF.

- Short-duration bond ETF.

Treatment: outside regime sleeves. Do not use cash-like assets to define R4 volatility.

5. Allocation Logic: What Must Change from v0

v0 flaw: Stage 1 inverse-vol inside a sleeve allows low-vol credit/cash-like assets to dominate. This is not macro risk parity; it is low-vol carry concentration.

v0.1 design requirements:

1. Separate macro-driver buckets before applying statistical weighting.

2. Within each bucket/sleeve, prevent one asset from dominating just because recent volatility is low.

3. Test longer volatility windows and EWMA, because 60d can understate slow-moving credit/duration risk.

4. Use point-in-time asset availability. New ETFs enter only after listed date.

5. Document universe shifts and their backtest effect.

6. Rebalancing

Current v0: monthly rebalance.

v0.1 should test:

- Monthly scheduled rebalance.

- Deviation rebalance: trigger if actual asset weight deviates from target by >5% NAV or by a relative threshold.

- Trading cost and liquidity constraints.

Bridgewater exact cadence is not fully public. We should not claim more precision than public sources support.

7. Risk Control / Crisis Mode

Normal environment:

- Let the model rebalance risk exposures.

- Avoid frequent subjective intervention.

Crisis mode:

- Define observable triggers for broad liquidity contraction / funding stress / policy regime break.

- When triggered, human discretionary override is allowed because this is rare and high-impact.

Candidate mechanical warning indicators for v0.1 research:

- Broad equity drawdown and realized vol spike.

- Credit spread / credit ETF drawdown abnormality.

- Funding-rate spike / DR007 stress.

- Rapid M2/TSF deterioration where data is available.

- Cross-asset correlation breakdown.

8. Immediate Next Step

Before coding Layer 2 full backtest, finalize candidate asset pool table:

- Asset category.

- Macro role.

- Regime/bucket assignment.

- ETF proxy.

- Listed date.

- Liquidity/history adequacy.

- Include in v0.1 or defer.

Then rerun variants:

- Vol windows: 60 / 120 / 252 / EWMA.

- Credit treatment: R1 sub-bucket vs separate credit/carry sleeve.

- Sleeve-internal cap as safety net, not primary driver.

- Deviation rebalance and risk-reduction rules.

9. Candidate ETF Inventory Snapshot

Generated files:

- research/asset-pool-candidate-etfs-v0.1.csv

- research/asset-pool-candidate-etfs-v0.1.md

Initial observations from RQData:

- Core existing proxies have full or near-full history from 2015, except newer bond/commodity ETFs.

- CSI 1000 512100.XSHG starts 2016-11-04; point-in-time inclusion is straightforward.

- ChiNext 159915.XSHE has full 2015+ history and good liquidity.

- STAR 50 588000.XSHG starts 2020-11-16; must be point-in-time.

- 30Y Treasury 511090.XSHG starts 2023-06-13; useful conceptually but short history.

- Policy bank bond 511520.XSHG starts 2022-10-25; conceptually important but short history.

- Energy-chemical futures 159981.XSHE starts 2020-01-17 and has adequate recent turnover; this is the cleanest energy/chemicals inflation proxy found.

- Coal 515220.XSHG and steel 515210.XSHG start 2020-03-02; these are sector equity proxies, not pure commodity futures, so they should be treated carefully.

- Broad commodity 510170.XSHG has full history but low recent turnover relative to core ETFs; review liquidity before inclusion.

- Old high-grade credit ETF 511280.XSHG delisted in 2021 and recent turnover field is not meaningful; defer.

Implication: v0.1 should support point-in-time universe entry by listed date. A single static 2015 universe would either omit important modern assets or create false history.