EXCHANGE TRADED FUNDS VERSUS TRADITIONAL MUTUAL FUNDS: A comparative analysis on the Italian Market
Giovanna Zanotti[1] & Cristiano Russo [2]
April 2005
ABSTRACT
During the last few years, Exchange Traded Funds (ETF) were characterised by a very significant growth. Born in 1993 in the United States, they didn’t appear in Europe until April 2002. Even though Italian market was born just in September 2002, nowadays it is the most important national reality in Europe, together with Germany. This paper aims to summarize results gotten from a complete compared analysis between traditional mutual funds, and ETFs that literature indicate as potential substitute products. Some econometric models helped us to evaluate management style (passive vs active) and funds capability to reach their own goals. we compare historical and risk-adjusted performances of ETFs versus mutual funds, to better appreciate investors returns. We also consider differences and common sides in commission structures and in management fees in particular.
I. Introduction
Nowadays Italian Exchange Traded Funds market is the most important reality in Europe, together with the German one: European asset under management (AUM) of ETFs market at February 2005 is 36.9 billions dollar, while Italian reality gains 12.4 billions dollar with more than 33% market share.[3]
Nevertheless Italian Exchange Traded Funds market, has got a quite recent origin: even if MTF[4] segment was born already in July 2002 as a negotiation site for closed-end funds, the first three ETFs were listed only the 30th September 2002. The goodness of these new generation instruments was suddenly appreciated by investors, since before the end of 2002, 5 more ETFs were listed. Starting from 2003 the growth was exponencial: just in that year 5 more products were listed by a number of 73.364 contracts traded, equal to 1.6 billions euros (+482% compared to 2002). As June 2004, 17 ETFs were listed on MTF, issued by 5 different brands: that year was also created the first bond index-linked ETFs. The raising diversification level of supply, high liquidity, pricing efficiency and good disclosure requirements are all factors that strongly affected the growth. As far as liquidity is concerned, MTF rules require at least one specialist to be a liquidity provider by each ETF listed. Furthermore market rules require a 100 basis points maximum bid-ask spread, for almost all the Exchange Traded Funds listed. Actually in the last quarter of 2003 the effective bid-ask spread was on average definitely much lower than the ruling value implied, as shown in table 1.
Table 1Average bid-ask spread
BENCHMARK / ETF / 2°
Quart
2004 / 1°
Quart
2004 / 4°
Quart
2003 / MAX
SPREAD
ITALY
S&P/MIB / S&P/MIB MU / 0,08% / 0,07% / 0,06% / 0,25%
CORPORATE BONDS
iBoxx EUR Liquid Corp / iBoxx EUR Liquid Corp / 0,10% / * / * / 0,75%
GOVERNMENT BONDS
EuroMTS Global / EuroMTS Global MU / 0,10% / * / * / 0,27%
USA (“Technology”)
NASDAQ-100 Index / NASDAQ-100 E.T. / 0,15% / 0,16% / 0,21% / 1,50%
MSCI USA IT / MSCI US Tech MU / 0,30% / 0,28% / 0,45% / 1,00%
USA (Blue Chips)
S&P 500 / iShares S&P 500 / 0,24% / 0,24% / 0,38% / 1,00%
DJ Industrial Average / DJ Industrial Average MU / 0,16% / 0,18% / 0,26% / 1,00%
EUROPE (Euro Area)
DJ Euro Stoxx 50 / DJ Euro Stoxx 50 MU / 0,17% / 0,17% / 0,18% / 1,00%
DJ Euro Stoxx 50 / iShares DJ Euro Stoxx 50 / 0,19% / 0,18% / 0,18% / 1,00%
S&P Euro / SPDR Euro Fund / 0,74% / 0,73% / 0,95% / 1,00%
MSCI Euro / B1 – MSCI Euro / 0,20% / 0,21% / 0,20% / 1,00%
FTSE Euro 100 / iShares FTSE Euro 100 / 0,80% / * / * / 1,00%
EUROPE
S&P Europe 350 / SPDR Europe 350 Fund / 0,41% / 0,63% / 0,75% / 1,00%
DJ Stoxx 50 / iShares DJ Stoxx 50 / 0,15% / 0,19% / 0,36% / 1,00%
FTSE Eurotop 100 / iShares FTSE Eurotop 100 / 0,56% / * / * / 1,00%
WORLD
DJ Global Titas 50 / DJ Global Titas 50 MU / 0,32% / 0,29% / 0,45% / 2,00%
ETHIC
Ethical Index Euro / B1 – Ethical Index Euro / 0,47% / 0,53% / 0,52% / 1,00%
Source: Borsa Italiana S.p.a. Situation as November 2004
As far as pricing efficiency is concerned, premium-discount quotations are very limited both in frequency and magnitude side. The average absolute value stated during the second quarter of 2004 was just 12 bps and this value is very low if compared for example with US International Exchange Traded Funds, which value is 35 bps.
Finally turning at disclosure, the market seems to be very efficient as well. Portfolio Composition Files are public and always available to all investors, who are also allowed to get real time Net Asset Values.
II. The analysis
Literature use to identify ETFs as the new potential substitute product of mutual funds. We compared the two product making reference to the Italian market. The ETFs sample utilized, is composed by 7 within the 10 most important products listed on MTF by Asset Under Management: this way we could consider more than 80% of market share supply, as showed in the table below.
Table 2
Exchange Traded Funds in sample
ETF Name / Benchmark / AUM (Mln €) / Mkt ShareDJ EuroStoxx 50 Master Unit / DJ Euro Stoxx 50 / 2.356,16 / 29,92%
iShares DJ Euro Stoxx 50 / DJ Euro Stoxx 50 / 2.181,61 / 27,71%
iShares S&P 500 / S&P 500 / 755,45 / 9,59%
iShares DJ Stoxx 50 / DJ Stoxx 50 / 508,59 / 6,46%
DJIA Master Unit / DJ Industrial Average / 204,95 / 2,60%
iShares FTSE Euro 100 / FTSE Euro 100 / 158,75 / 2,02%
B1 MSCI Euro / MSCI Euro / 140,85 / 1,79%
Situation as 30/09/2004
On the other hand, traditional mutual funds sample is selected by a benchmark-criteria: this means that we directly compared to each ETF, those funds whose benchmark stated in the prospectus is the same. This way we were able to select 142 instruments between more than 3000 funds registered with Assogestioni[5]: the following table summarizes the whole dataset.
Table 3
Number of traditional mutual funds analyzed by each ETF
ETF Name / 1 Year52 data / 2 Years
104 data / 3 Years
156 data / 4 Years
208 data / ETF Life
DJ EURO STOXX 50 MU* / 11 / 10 / 6 / 4 (187 data)
ISHARES DJ EURO STOXX 50* / 11 / 10 / 6 / 3 / 4 (236 data)
ISHARES DJ STOXX 50 / 3 / 2 / 2 / 1 / 1 (236 data)
B1 MSCI EURO / 68 / 65 (86 data)
ISHARES FTS EURO 100 / 8 / 6 / 6 / 6 (165 data)
ISHARES S&P 500 / 50 / 45 / 40 (135 data)
DJIA MASTER UNIT / 2 / 2 / 1 / 1 (182 data)
Total / 142 / 65 / 15 / 4
* Those two ETFs have got the same benchmark, so that traditional mutual funds compared are the same ones.
Database set-up
Both Exchange Traded Funds and traditional mutual funds are included in the huge OICR family. About those instruments, it’s very common to select the value of the individual share as the basic element to study. By using Datastream we got 5 years-length historical series of prices. We chose weekly frequency prices, because it would have been the best solution for the existing trade-off between volatility and minimum number of observation in sample.[6] All return rates were continuously compounded on annual basis.
About dividends paid by those securities composing fund’s portfolio, we decided not to keep in consideration those kinds of income. Actually, that doesn’t mean we didn’t consider dividends impact: both the Exchange Traded Funds and traditional mutual funds analyzed, use to reinvest dividends within the fund itself.
Methodology
The topics we aimed to evaluate in particular are:
. Management style and capability to reach the objectives as stated in the prospectus of the fund
. Historical and risk-adjusted performances
. Commission structure and costs linked
The best way to evaluate ETFs and mutual funds efficiency concerning with the management style, is to check how does funds value behaves if compared to its benchmark. So we implemented some econometric analysis, by using a double-variable linear regression model, estimated by OLS techniques. The basic estimation model we used, can finally be expressed as follows:
[Basic Model]
This model is used for both Exchange Traded Funds and traditional mutual funds, by standardized holding periods.
To check the quality of results coming out from regressions, we kept into consideration some well known econometrical and statistical tests[7]: this way we were able to investigate about any violations oh model hypothesis such as normality of residuals distribution, non autocorrelation and omoschedasticity. As soon as that kind of problems came out, we made some modifications to the basic model, to get results in parameters of regressions that could be affordable.
To solve autocorrelation problem we opted for a solution that implies an addiction of further time gaps. The following expression we used adds one more time delay:
[R1 Model]
This model is the consequence of a first order residuals autocorrelation hypothesis (εt = ρεt-1+ μt); anyway when autocorrelation problems kept persisting, we were assuming second or third order hypothesis.[8]
This kind of modification could generate any eteroschedasticity problems. In this case a solution is to split the whole sample in two (T1 e T2) or even more sub-samples, so that in each of them, regressions would be omoschedastical. Then it’s possible to rebuild historical series just by dividing each observation by partial standard error of regression (σn). Formally this new model can be defined as follows:
, with t e t-1 є T1
, with t e t-1 є T2
[M1 Model]
If eteroschedasticity problems keep persisting further, that’s probably because of a structural break in regression coefficients.[9] After having individuated the break-point within the sample, we could consider any dummy as follows:
[D1 Model]
If the model does not imply any time delay, it would be the same as this one, but just without those t-1 indexed-parameters.[10]
Only after these modifications, when necessary, it’s possible to express any affordable valuation about Exchange Traded Funds management style.
After having analyzed the style management of those products, we wanted to directly verify how they performed in the last years. First step was to simply calculate historical performances and correlation for each fund, with its own benchmark. Then we also estimated risk-adjusted performances by using some well known coefficients such as information ratio and sharpe ratio: in particular, to get this last coefficient we also had to verify the average tracking error which, in this job, is defined as the difference between the average return of the investment fund and the average return of its own benchmark at the same time. This approach allowed us to evaluate how efficiently fund managers were able to solve risk-return trade-off.
In the end we took into consideration commission structures of Exchange Traded Funds and traditional mutual funds. Actually cost sources are quite different to each other: so we tried to catch common elements between ETFs and mutual funds costs, to be able to make a direct compared analysis. As the matter of facts we finally considered the following elements:
. entrance fees
. exit fees
. management fees
. performance fees
Those ones are the most important costs sources of traditional mutual funds, even if actually they are not the only ones quite frequently.
Expectations about results
Considering both Exchange Traded Funds and traditional mutual funds characteristics, it’s possible to have some expectations about results of the analysis. Because of their passive management style, ETFs should have a quite similar profile to their own benchmark: this means that from linear regressions we expect a value of β close to 1, while the capability of the fund to gain active returns should be very low, so that α value should be statistically equal to zero. We consequently expect very high correlations with benchmarks and very low tracking error as well. Otherwise is a little bit more difficult to have such defined expectations about traditional mutual funds. This is because of the fact that concerning with the prospectus of most of them, the objective of management is always to try and outperform the benchmark: this means even strong mismatches versus benchmark tracks which would suggest very assorted values of coefficients α and β. The only expectation that seems to be reasonable, is that coefficient α should be statistically different from zero.
III. Results
The following table shows in details econometric model utilized to get affordable results on which is possible to express any opinion about Exchange Traded Funds.
Table 4Implemented Econometric Models
ETF Name / 1 Year / 2 Years / 3 Years / 4 Years / ETF Life
DJ EURO STOXX 50 MASTER UNIT / Basic / D1 / M (4-2) A / *** / R1
ISHARES DJ EURO STOXX 50 / Basic / Basic / Basic / Basic / Basic
ISHARES DJ STOXX 50 / Basic / Basic / Basic / Basic / Basic
B1 MSCI EURO / R1 / *** / *** / *** / D1
ISHARES FTS EURO 100 / R1 / R1 / R1 / *** / R1
ISHARES S&P 500 / R1 / M (3-1) B / *** / *** / R1
DJIA MASTER UNIT / R1 / M (2-2) C / M (2-2) C / *** / M (3-2) D
A In this case the regression model is M1-type: anyway it is modified by considering 4 sub-samples and 2 temporal delays.
B In this case the regression model is M1-type: anyway it is modified by considering 3 sub-samples and 1 temporal delay.
C In those cases the regression model is M1-type: anyway it is modified by considering 2 sub-samples and 2 temporal delays.
D In this case the regression model is M1-type: anyway it is modified by considering 3 sub-samples and 2 temporal delays.
Results of econometric analysis show that Exchange Traded Funds are always able to get the goal stated in the prospectus in a very efficient way. Correlation of historical series of returns between ETFs and their own benchmark are always higher than 99% and that implies a very low tracking error along the whole holding periods set.
In particular, coefficient values are consistent with expectations. We always found α values statistically equal to zero: that stress the attitude of Exchange Traded Funds managers not to look for extra-earnings by leaving any opportunity of active trading operations. On the other hand, β values are always close to 1, even if the fact that values are always a little bit under 1, in particular in the farther holding periods, could be read as a sign of prudence concerning with management style. Those results gain once more consistency, because of the fact that R-squared coefficient is always very high and in most of the cases close to 100%.[11]
While about Exchange Traded Funds we can definitely say that they are able to get their own goals, we cannot affirm the same as to traditional mutual funds.
Even if concerning with what stated in their own prospectus, those products should try and outperform the index, that’s not what concretely happens. Econometric analysis on traditional mutual funds gave a kind of unexpected results: in most of the cases α values are statistically equal to zero. This would have been reasonable in a 12 months horizon, since the efficiency of this kind of products is much higher through longer holding periods: but as the matter of fact, α values seem to be unlinked with holding periods. On the other hand, β values are very assorted as we expected.
Econometric analysis can already give some suggestions about Exchange Traded Funds and traditional mutual funds, but that’s not enough to have a fulfil idea of the comparison: since some quality problems with regressions on mutual funds occurred, that’s suitable to look for further confirmations by analyzing performances.
Just with reference to historical performances, we can find a confirmation with the passive management style: as the matter of fact returns of ETFs and their own benchmark are always very similar.
Table 5
Historical Performances Standings
1 Year / 2 Years / 3 Years / 4 Years / ETF LifeDJ Eurostoxx 50 MU / 4 / 13
8 / 13
4 / 13
8 / 13
36 / 52
19 / 52
2 / 5
5 / 5
3 / 4
2 / 4
2 / 10
4 / 10
17 / 70
24 / 70 / 7 / 12
9 / 12
7 / 12
9 / 12
20 / 47
17 / 47
2 / 4
3 / 4
3 / 4
2 / 4
2 / 8
3 / 8 / 7 / 8
8 / 8
6 / 8
8 / 8 / 5 / 6
6 / 6
4 / 5
5 / 5
34 / 44
11/ 44
2 / 3
3 / 3
2 / 3
3 / 3
7 / 8
6 / 8
21 / 67
24 / 67
DJ Euro Stoxx 50 (bench)
iShares DJ Eurostoxx 50 / 3 / 5
5 / 5
DJ Euro Stoxx 50 (bench)
iShares S&P 500
S&P 500 (bench)
iShares DJ Stoxx 50 / 2 / 4
3 / 4
3 / 3
2 / 3
6 / 8
7 / 8 / 2 / 3
3 / 3
DJ Stoxx 50 (bench)
DJIA Master Unit
DJ Industrial Average (bench)
iShares FTSE Euro 100
FTSE Euro 100 (bench)
B1 MSCI Euro
MSCI Euro (bench)
The table above shows the position of each ETF compared to its own benchmark, in a list of those instrument from the most to the less performing.[12] The first evidence can be expressed comparing Exchange Traded Funds versus benchmarks: they use to perform a little bit better than their own benchmark in most of the cases. The following chart allows to better understand tracks behaviour.
Chart 6
Dividends impact is very clear: the continuous reinvestment policy of dividends allows the ETF share values to slightly outperform the benchmark, which usually is a price-index.[13]
Another evidence from table 5, is that Exchange Traded Funds use to perform well above the mid-standing position in most of the cases: this means that more or less 50% of traditional mutual funds gained lower returns than Exchange Traded Funds along the sample. This statement gets much more importance if it takes into consideration that our 5 years sample goes back from October 2004: that means an almost complete bearish market environment. In this case, it would have been reasonable to exploit the opportunity that active management gives to shift from the benchmark: but as the matter of fact, that didn’t happen.
Considering also the level of risk connected with investment instruments, it’s possible to express more evaluations. Information ratio values, confirm the high capability of Exchange Traded Funds to reply benchmark returns efficiently: both tracking error and standard deviation values are always very low, so that ETFs can finally be seen as well performing passive instruments. On the other hand a direct comparison can be made by observing sharpe ratio values both of ETFs and traditional mutual funds. The following table was built by the same criteria as table 5.
Table 7
Sharpe Ratios Standings
1 Year / 2 Years / 3 Years / 4 Years / ETF LifeDJ Eurostoxx 50 MU / 5 / 13 / 7 / 12 / 5 / 8 / 6 / 6
DJ Euro Stoxx 50 (bench) / 8 / 13 / 9 / 12 / 7 / 8 / 3 / 6
iShares DJ Eurostoxx 50 / 5 / 13 / 7 / 12 / 4 / 8 / 2 / 5 / 2 / 5
DJ Euro Stoxx 50 (bench) / 8 / 13 / 9 / 12 / 7 / 8 / 3 / 5 / 3 / 5
iShares S&P 500 / 28 / 52 / 16 / 47 / 15 / 42
S&P 500 (bench) / 20 / 52 / 14 / 47 / 12 / 42
iShares DJ Stoxx 50 / 3 / 5 / 2 / 4 / 2 / 4 / 1 / 3 / 2 / 3
DJ Stoxx 50 (bench) / 5 / 5 / 3 / 4 / 3 / 4 / 2 / 3 / 3 / 3
DJIA Master Unit / 3 / 4 / 3 / 4 / 3 / 3 / 3 / 3
DJ Industrial Average (bench) / 2 / 4 / 2 / 4 / 2 / 3 / 2 / 3
iShares FTSE Euro 100 / 4 / 10 / 1 / 8 / 3 / 8 / 2 / 8
FTSE Euro 100 (bench) / 5 / 10 / 3 / 8 / 2 / 8 / 1 / 8
B1 MSCI Euro / 25 / 70 / 47 / 67
MSCI Euro (bench) / 31 / 70 / 43 / 67
Results summarized in Table 7 confirm those opinions already expressed by observing historical performances only: evaluations about Exchange Traded Funds that come out from Risk-Adjusted coefficients, are even better. For each ETF, sharpe ratios standings use to be better than historical returns ones, so that the number of traditional mutual funds that can be defined underperforming, definitely increase. We can finally summarize those results we got, without making any distinction between each Exchange Traded Funds. Cumulative data expressed by the following chart can show a quite interesting trend.