Performance appraisal of exchange-traded funds using clustering and data envelopment analysis (XETRA, Germany)

Abstract

In the study hierarchical clustering approach and data envelopment analysis are applied in performance appraisal of ETFs traded on XETRA platform (Germany) for the period 2017-2018 year. The combination of methods allows to evaluate performance of ETFs within homogenous groups and by that increase reliability and validity of assessments. It also provides possibility for diversification because ETFs are grouped with regard to their movements of weekly returns. As a result, agglomerative hierarchical clustering algorithm with Ward’s method as a linkage and dissimilarity measure based on the Pearson correlation coefficient of weekly returns gives reliable and valid partitioning of ETFs into 30 clusters. CCR model with TER and downside risk as inputs, and average weekly return and upside potential as outputs shows performance evaluations of ETFs within their own clusters which do not correlate with ones based on traditional risk-adjusted measurements. All the measurements are used in constructing all-ETF portfolios in order to check for the effectiveness of proposed method. Comparison of annual returns of different all-ETF portfolios under assumption of buy-and-hold strategy and time horizon equal to 1 year demonstrates that portfolios which are built in consideration of CCR model scores outperform others which are based on traditional risk-adjusted measurements as well as randomly generated.