A study of the performance of exchange traded funds

Abstract

Exchange traded funds (ETFs) are collective investment vehicles that have known a growing interest over the past years. Yet, only a few studies were dedicated to the measurement of their performance. In this thesis, I examine one of the most widely used performance measure for passive management, the information ratio. I analyze its weaknesses and assumptions, and justify the need for a new performance measure that will be applicable to ETFs. The information ratio does not work well when the tracking difference is negative and it does not take into account the magnitude of the tracking error (Roncalli, 2014). Moreover, it assumes that the excess returns are normally distributed. I select a sample of 30 ETFs and show that their excess returns are not normally distributed. Therefore, I develop a new performance measure that takes into account the skewness and the kurtosis of the distribution. Moreover, I consider the work of Hübner (2005) to take into account the relative performance of the benchmark. I apply this new performance measure on the ETFs from the sample and analyze the results. As expected, the ranking obtained seems to show a positive correlation between performances of ETFs tracking the same benchmark. Moreover, a rolling window analysis highlights the stability of the measure when using windows of different widths. In order to assess the quality of the measure, I first test its robustness in the measurement of performance persistence. I show that, using well known statistical tests, the measure is relatively robust in measuring performance persistence since the results indicate persistence for the sample. The same tests are then performed on the information ratio. Since the results also show persistence of the sample, it means that the new performance measure is good at identifying persistent winners when the sample is composed of funds that are persistent in their performance. However, the results of the tests of persistence indicate slightly better results for the information ratio. I explore two hypotheses to explain this result. Firstly, it can be due to the characteristics of the ETFs from the sample. Secondly, it can be explained by performance manipulations on the information ratio.