- Design
- Two-way fixed-effects panel models use country and year effects with asset-weighted banking aggregates, lagged terms and dependence-robust uncertainty. They are panel estimators, not design-based difference-in-differences studies.
- Outcome and measure
- Separate specifications examine national financialisation proxies and manufacturing value added. Original units and proxies remain distinct.
- Coverage
- Fifteen dual-banking settings are examined over the comparative study's 2000–2022 window, subject to measure-specific availability and complete-case requirements.
- Public interpretation
- Some preferred models are consistent with a negative finance–manufacturing association, but results vary by proxy, lag and episode. Short-run predictive-precedence tests do not establish a universal direction.
Limitations
- Endogeneity, omitted institutions and unobserved heterogeneity remain possible.
- Granger tests assess predictive content at selected horizons, not structural causality.
- Official aggregates inherit the classifications of the system being critically examined.
- The 15-country dual-banking sample does not represent every financial system.