Choosing Relative Momentum for International ETF Allocation
Summary
The document compares relative momentum against country-specific factor models for allocating among four international equity ETFs based on performance versus the US market. It favors ranking the country funds against the US benchmark, arguing that separate models for a small set of assets would require estimating many parameters and could produce noisy, overfit signals. The comparison is qualitative; it presents no backtest or measured performance evidence.
Relative momentum is described as a simple, interpretable complement to an existing US time-series momentum strategy, but it may concentrate holdings in a small number of countries. Its diversification benefit depends on relative trends. Country-specific factors might diversify more steadily if they are sufficiently independent, though their usefulness depends on model complexity and data quality. The document suggests expanding toward markets with distinct macro drivers for momentum, or markets with cleaner factor data for factor models. These are recommendations rather than tested findings, and results would depend on signal design and implementation.
Key ideas
- Relative momentum ranks international equity ETFs by performance against a US benchmark.
- Estimating independent country models across a small universe can create noisy signals and overfitting risk.
- Relative momentum may concentrate exposure in the countries with the strongest relative trends.
- Factor models could offer steadier diversification if their signals are sufficiently distinct and supported by reliable data.
- The document's recommendations are qualitative and are not backed by reported performance tests.
Tags
Full text
# Choosing between two approaches for an international ETF strategy # Choosing between two approaches for an international ETF strategy I'm building a strategy that allocates to international equity ETFs (EZU, INDA, EWJ, MCHI) when they outperform SPY. I'm deciding between two approaches: - Relative momentum - Signal based on each country's momentum relative to SPY - Country-specific factor models - Independent signals per country based on local drivers, filtered for low correlation with our existing SPY TSMOM strategy Which approach is more appropriate given the small universe size? If I were to expand the universe, how would the choice of approach influence which countries to add? Additionally, what is the return vs diversification tradeoff for each approach? Thank you. ## Answer by Divyanshu Verma (score 0) https://quant.stackexchange.com/a/85581 Relative momentum is almost certainly the better fit, country-specific factor models are going to be noisy and overfit-prone. Building independent signal models for four countries means estimating a lot of parameters relative to the data you have, and the low-correlation-with-TSMOM filter sounds good in theory, but with only four assets you'll likely end up holding one or two positions most of the time, which defeats the purpose. Relative momentum is also a natural complement to your existing SPY TSMOM strategy. Ranking these ETFs against SPY is a clean, interpretable rule that tends to work well because macro and risk-on/risk-off flows drive broad country-level price moves before they show up in factor fundamentals. On the return vs diversification tradeoff: relative momentum is primarily return-seeking. You're holding whichever countries are beating the US, so you're often concentrated in one or two regions. The diversification benefit is real but conditional on a persistent relative trend existing. Factor models, if built properly, can give more stable diversification since the signals are somewhat orthogonal to pure market direction, but that stability comes at the cost of complexity and data quality risk, especially for INDA and MCHI. For universe expansion, relative momentum pushes you toward countries with distinct macro regimes and low return correlation to your existing four, things like EWZ or EWY that respond to commodity cycles or export demand. Factor models push you toward developed markets where clean factor data actually exists, since adding frontier or thin ETFs would undermine the whole premise. My recommendation is to start with relative momentum, get it running cleanly, then layer in country-specific signals as a sizing modifier once you have a solid baseline.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.