Limitations

52

53 choosing 10 years are twofold. Firstly, 10 years of data generates a lot of extra data points. As mentioned in the calculation above, the choice of five years already means that there is over 200,000 data points, and this should of course be doubled with 10 years of data, meaning over 400,000 data points. This amount of data is a lot to process. Secondly, and most important, using 10 years of data means that there are a lot of funds which have incomplete data, meaning they have only existed for parts of the time horizon. This means that there are times where there is a lot more data points, which could easily change the conclusions drawn in the analysis. This should of course be avoided. The reasoning for not having only one year of data is easy to grasp. This would create way too much uncertainty around if it is the right year which has been chosen.

Because of the reasoning above, a time horizon of five years has been chosen to balance out the positives and negatives of a long and short time horizon. That said, it is indeed very interesting to see whether there is any change in the conclusions drawn when using a longer time horizon.

Because of that, the discussion section, section 6.3.1 will both discuss the impact in detail and report results using a longer time horizon, which can then be compared with the original results to assess the impact.

4.5.3 Benchmarks

The analysis conducted relies heavily on market returns, which might as well be called benchmarks.

These benchmarks are chosen to fit the type of funds they are used for, but for several of the asset types the benchmarks are not at all perfect proxies for market returns. There are two limitations related to the benchmarks used, which will be handled separately in the following.

Firstly, the benchmark used for bond-focused funds, money market-focused funds and for mixed assets-focused funds is not a perfect proxy for the respective market returns. For the bond-focused funds the benchmark used is the "Vanguard Total Bond Market ETF". The name seems to indicate that it is the perfect benchmark but actually, the ETF only covers the total bond market in the United States. This is of course not ideal. The same problem appears for the money market-focused funds, where the ETF used also covers only United States. For the mixed assets-focused funds the problem is more that mixed assets are a fairly broad definition, as there is no clear indication of which assets they focus on. For these funds, a benchmark is constructed as 60% of the benchmark for equity-focused funds and 40% of the benchmark for bond-focused funds. This is by no means ideal, as the distribution can be far from correct. Nevertheless, this has been used in lack of a better

54 benchmark. The impact of this can be substantial but it is a problem which is very hard to

overcome, as there seems to be no better benchmark to use for the three types of asset classes.

Because of that the conclusions drawn regarding these types of funds have a significantly decrease in their validity. Another impact of this is that the analysis will focus mainly on the equity-focused funds, as the validity of the conclusions drawn here are significantly higher.

Secondly, the benchmark used for equity-focused funds throughout the analysis are a global benchmark, meaning that it is an ETF tracking the global equity market. This is on line with the theory behind CAPM since CAPM would argue that all rational investors would invest in this portfolio, but there are also arguments for using another approach. Another approach could be to use a benchmark for whatever geographical focus area the funds have. This could especially make sense in the regressions where only funds focusing on a specific geographical focus area are

included. That said, the analysis will still use the global indices as this is the approach most consistent with the theory behind the CAPM and the Fama-French Three-Factor Model frameworks.

The impact of this limitation can quite easily be investigated with the way the model is set up. The only change compared to the original setup is that the market return benchmark is changed in the regressions focusing on a specific region, whether it is emerging markets, Nordic markets or global markets. The impact of the limitation will be further discussed in section 6.3.2 covering the how the results change when using local benchmarks.

4.5.4 Level of Activeness

The operationalisation used throughout this paper of using the fee as a proxy for the distinction between actively and passively managed funds has some flaws built into it. This operationalisation is by no means linear, implying that a high fee not necessarily means a high level of activity. This connection between fee and the level of activeness in the fund is not in scope for this paper but could instead be an area of further research to be done. As Financial Times reports 'Some tracker funds cost 10 times more than rivals' (Financial Times, 2017), where the title says it all. A tracker fund is another word for a passive fund and therefore it is a clear indication of that the link between fee and level of activeness of a fund.

55 4.5.5 Tax

As it previously has been mentioned, taxation of fund returns for private investors are different across countries, even in the Nordics. Furthermore, for some countries it is even different how domestic and foreign investment funds are taxed. Taxation is of course of major importance when private investors choose to invest in an investment fund, but it is not a problem which will be handled in this paper. Reason for that is that it should have minimal impact on the choice of funds and especially minimal impact on whether an investor should choose an actively or a passively managed fund.