This study aims to investigate whether exchange-traded fund (ETF) replication strategy influences tracking error. While synthetic ETFs are often criticised for their complexity and perceived risk, empirical evidence on their tracking performance remains mixed. This paper provides a systematic comparison between replication methods using a large data set and robust statistical techniques to assess whether differences in tracking error are economically significant.
A panel of ETFs using different replication strategies is analysed over 5- and 10-year horizons. Tracking error is computed relative to benchmark indices. Both ordinary least squares and robust regression methods are used, controlling for fund characteristics and enabling identification of effects that are sensitive to outliers and heteroskedasticity.
Parametric results indicate negligible differences between replication strategies. However, robust regressions show that synthetic ETFs exhibit significantly lower tracking error, by up to 0.20 and 0.30 percentage points over 5- and 10-year horizons, respectively. These findings suggest that synthetic replication may improve tracking efficiency over longer horizons. Investors, asset managers and policymakers can use these findings to better assess the trade-offs between replication strategies.
The analysis is based on historical data and may not fully capture changing market conditions or regulatory developments. Future research could examine intraday tracking, alternative asset classes and stress-period dynamics.
This study offers a large-scale comparison of ETF replication strategies using both parametric and robust methods. It demonstrates that reliance on conventional estimators may obscure economically meaningful differences and that synthetic replication can reduce tracking error.
