Key Takeaways
- • Man Numeric argues EM country-level risk is significantly higher than in developed markets, making the country timing opportunity structurally larger in EM, though stock selection models are typically not built to exploit it.
- • A standalone fundamental country model blending 12-month price momentum and relative valuation produced a positive cumulative PnL with a gross Sharpe ratio under 0.5 and significant turnover, with over 40% of value added since 2008 generated in the past 12 months.
- • The proposed macro-based country timing model combines fundamental economic logic with machine learning and a dual time-series and cross-sectional lens, achieving a realised Sharpe ratio of 0.85 and 28% lower turnover than the fundamental approach over the 2008-2026 backtest window.
- • The authors argue country tilts from bottom-up stock selection are positively correlated with existing alpha, limiting diversification, and that a systematic country overlay informed by macro data can address this without compromising the bottom-up process.
Summary
Systematic EM managers are increasingly looking to pair stock selection with a complementary country timing component, according to a Man Numeric research piece. The firm notes that stock selection remains the primary driver of alpha in EM equities, while country-level returns are dominated by macro forces such as monetary policy cycles, terms of trade, external demand and financial conditions. Country timing has traditionally been the domain of discretionary managers, while systematic managers have typically treated country exposures as risks to be contained.
A risk decomposition of EM versus developed markets, with data as at January 2008 to June 2026, shows that while stock-level risk dominates both, country-level risk is significantly higher in EM. Cross-country return dispersion remains wider in EM than in DM, reflecting differences in economic structure, policy regimes and external vulnerabilities.
The article evaluates a simple cross-sectional model blending relative valuation and 12-month price momentum into long/short country positions, sized by conviction and capped at plus or minus 5%. Applied across all EM countries, the model generated positive cumulative PnL, but its gross Sharpe ratio was under 0.5 with significant turnover. Over 40% of the value added since 2008 came in the past 12 months, driven by the South Korean rally, and the model would have been vulnerable to the 2022 Russia drawdown. The authors also note that country tilts from such models tend to correlate with stock-picking alpha, which limits diversification.
The proposed alternative is a macro-based model that combines fundamental economic logic with machine learning over point-in-time macroeconomic datasets. It views markets through a time-series lens, assessing each country against its own history, and a cross-sectional lens, comparing countries against each other. The model is illustrated with a single-country case study of South Korea from 2021 to 2026, which the authors caution should not be generalised across markets. Its South Korea position shifted repeatedly as drivers rotated from sector dynamics to trade and terms-of-trade signals, in contrast to the static multi-year long tilt of the fundamental model.
In backtesting, using only data available at each point in time, the macro model delivered a realised Sharpe ratio of 0.85 and 28% lower turnover than the standalone fundamental approach over the 2008-2026 window, according to the article. The piece is presented as research from Man Numeric and does not include independent verification of these results.
A risk decomposition of EM versus developed markets, with data as at January 2008 to June 2026, shows that while stock-level risk dominates both, country-level risk is significantly higher in EM. Cross-country return dispersion remains wider in EM than in DM, reflecting differences in economic structure, policy regimes and external vulnerabilities.
The article evaluates a simple cross-sectional model blending relative valuation and 12-month price momentum into long/short country positions, sized by conviction and capped at plus or minus 5%. Applied across all EM countries, the model generated positive cumulative PnL, but its gross Sharpe ratio was under 0.5 with significant turnover. Over 40% of the value added since 2008 came in the past 12 months, driven by the South Korean rally, and the model would have been vulnerable to the 2022 Russia drawdown. The authors also note that country tilts from such models tend to correlate with stock-picking alpha, which limits diversification.
The proposed alternative is a macro-based model that combines fundamental economic logic with machine learning over point-in-time macroeconomic datasets. It views markets through a time-series lens, assessing each country against its own history, and a cross-sectional lens, comparing countries against each other. The model is illustrated with a single-country case study of South Korea from 2021 to 2026, which the authors caution should not be generalised across markets. Its South Korea position shifted repeatedly as drivers rotated from sector dynamics to trade and terms-of-trade signals, in contrast to the static multi-year long tilt of the fundamental model.
In backtesting, using only data available at each point in time, the macro model delivered a realised Sharpe ratio of 0.85 and 28% lower turnover than the standalone fundamental approach over the 2008-2026 window, according to the article. The piece is presented as research from Man Numeric and does not include independent verification of these results.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.