This study compares the returns of exchange-traded funds (ETFs) and index funds against the performance of the most commonly tracked equity indexes. Of particular interest is the differential in cost (i.e., the expense ratio) and tax efficiency between index funds and comparable ETFs, in as much as low costs and tax efficiency are commonly cited advantages of ETFs (Chamberlain and Jordan, 2004).
At year-end 2004, there were 152 ETFs in the Morningstar Principia database, with total assets of more than $222 billion. The largest, and oldest, ETF is the Standard and Poor's Depositary Receipts Trust 1, or SPDR, which is designed to track the S&P 500 Index. Since its inception in January 1993, the SPDR has gathered in excess of $52 billion in assets.
Fund providers have steadily introduced new ETFs over the course of the past decade, as shown in Figure 1.
To avoid anomalous results based on limited data, this study focused on ETFs with at least a three-year performance history. Of the 152 ETFs on the market as of December 31, 2004, 97 qualify. This subset had a total of $199 billion in assets, representing about 90 percent of all ETF assets.
There were no bondindexed ETFs with a three-year performance history as of December 31, 2004, hence none were included in this analysis.
During the past decade, ETFs have entered into a market dominated by mutual funds, and particularly by index funds. At year-end 2004, there were 527 index funds (equity and bond) in the Morningstar Principia database, with over $504 billion in total net assets. The oldest currently existing index fund in the database is the Vanguard Small Cap Index Fund, which was brought to market in October 1960 (though it was not always an index fund). The next index fund (and at the time the first retail index mutual fund), the Vanguard 500 Index Fund, didn't arrive until August 1976, but is now the largest mutual fund (let alone index fund) in existence, with over $80 billion in net assets.
Many mutual funds now have multiple share classes, such as A shares, B shares, C shares, I shares, K shares, etc. In this study, only one share class (the class with the largest asset base) was included in the analysis-and is referred to as a "distinct" fund. Failing to remove redundant share classes over-represents funds with multiple share classes and skews the data. Of the 527 index funds, there were 213 distinct funds with at least a three-year performance history as of year-end 2004. These funds had a total of $380 billion in net assets, or nearly 76 percent of total index fund assets. As with ETFs, index funds began proliferating in the mid-1990s (see Figure 2).
This study also looked at the raw indexes that ETFs and equity-based index funds seek to mimic. As of December 31, 2004, there were 168 indexes in the Morningstar Principia database, all of which had a three-year performance history. Clearly, however, not all indexes are equally represented among ETFs and index funds. Therefore, in this study, the performance of index funds and ETFs was compared with the nine most commonly tracked indexes, which are listed in Figure 3. From among the 213 distinct index funds, a total of 132 "tracked" one of the indexes in Figure 3. (Note that the Principia database does not include information for the NASDAQ-100 Index; pricing data for this index was taken from Yahoo! Finance.)
Our analysis looked at the asset-weighted three-year annualized return, asset-weighted three-year tax cost ratio and asset- weighted expense ratio of each fund and index. The tax cost ratio is a statistic calculated by Morningstar and is a measure of tax efficiency. The lower the tax cost ratio the better.
All data were asset-weighted to reflect the different sizes (and therefore differential impact on investors) of index funds and ETFs.
Dow Jones Wilshire 5000
Our first comparison looked at ETFs and index funds that track the Dow Jones Wilshire 5000 Index (see Figure 4). This index offers a good example of the importance of asset weighting. The eight distinct index funds that track the Wilshire 5000 held a total of $36.9 billion in assets as of December 31, 2004. One of the funds (the Vanguard Total Stock Index Fund) held $31.7 billion, or 86 percent of the total. Therefore, the average annualized return, tax cost ratio and expense ratio for that particular fund were weighted accordingly (at 86 percent). The data for the other seven funds were weighted according to their respective share of the total net assets.
At year-end 2004, there was only one ETF that tracked the Wilshire 5000-the Vanguard Total Stock Market (TSM) VIPER (Vanguard Index Participation Equity Receipt). There were eight distinct index funds tracking the Wilshire 5000. As would be expected, the raw Dow Jones Wilshire 5000 Index had a higher three-year annualized return compared to the performance of both the index funds and the ETF. (Note: The benchmark for both the Vanguard Total Stock Market Fund and its VIPERs share class has since switched to the MSCI U.S. Broad Market Index. Also, it should be mentioned, that VIPERs are unique among ETFs in that they are one share class of a traditional mutual fund, and may therefore be expected to have different performance characteristics, especially regarding tax efficiency, when compared to stand-alone ETFs.)
The three-year annualized return of the ETF was 5.36 percent, or seven basis points higher than the asset-weighted average return of the eight index funds. The differential in expense ratio between the index funds and the ETF was seven basis points in favor of the ETF, thus the differential in performance is entirely explained by the expense ratio, rather than tracking error.
The asset-weighted return of the index funds was 18 basis points lower than that of the raw index. The return of the ETF was 11 basis points lower than the raw index. In both cases, the differential in return is four basis points less than the expense ratio. This suggests that the funds and the ETF are delivering performance that partially offsets their expense ratios.
The asset-weighted average tax cost ratio of the index funds was the same as the tax cost ratio of the ETF.
Standard & Poor's 500 Index
For many people, the Standard & Poor's 500 Index is the king of U.S. equity indexes. At year-end 2004, 67 index funds tracked it, with total assets of over $203 billion (and recall that this is only the total number of distinct funds). In addition, there were two ETFs tracking the index, the SPDR and the iShares S&P 500 fund. Remarkably, these two ETFs had $64 billion in assets at year-end 2004.
The average three-year asset-weighted return of the two ETFs was within 20 basis points of the return of the S&P 500 Index (see Figure 5). Given their average expense ratio of 11 basis points, the tracking error of the ETFs was only nine basis points. Tracking error, in this application, is defined as the differential in performance between the raw index and the fund or ETF that is not explained by the expense ratio.
The asset-weighted three-year return of the 67 mutual funds that track the S&P 500 was 14 basis points lower than the raw index, or six basis points less than would be expected given the average expense ratio of 20 basis points. Moreover, the asset-weighted return of the index funds was six basis points higher than the asset-weighted return of the two ETFs. In spite of higher average expense ratios, the S&P 500 index funds outperformed the ETFs.
The tax-cost ratio of the index funds was 10 basis points better than the tax-cost ratio of the ETFs.
S&P 500/Barra Value Index
For the S&P 500/Barra Value Index, we compared a single iShares ETF against the average of three index funds (Figure 6). The average index fund performance demonstrated a negligible advantage in tax cost ratio, but significantly worse raw performance in comparison to the ETF.
The ETFs had a 44 basis point advantage in three-year annualized return, posting returns of 6.27 percent compared to 5.83 percent for the funds. The ETF return was 19 basis points lower than the raw index return, which with an average expense ratio of 18 basis points, represents virtually no tracking error. By contrast, the index fund average return was 63 basis points lower than the raw index, of which 44 basis points is accounted for by the expense ratio. Thus, the average tracking error of these three index funds was 19 basis points.
As shown in Figure 7, the two index funds that tracked the S&P/Barra Large Growth Index demonstrated superior tax efficiency compared with the iShares Barra Growth ETF. However, the dramatically larger average expense ratio among the index funds negates their tax efficiency advantage. The ETF had a positive return of 0.44 percent, 50 basis points higher than the -0.06 percent asset-weighted return for the index funds. The ETF return was within 19 basis points of the raw index, which is essentially accounted for by its expense ratio of 18 basis points. The average return of the index funds was lower than the raw index return by 69 basis points, or five basis points better than expected given an average expense ratio of 74 basis points.
S&P MidCap 400 Index
Among trackers of the S&P MidCap 400 Index, two ETFs outperformed 11 index funds in the area of tax efficiency (Figure 8). In terms of performance, the index funds had the advantage by 14 basis points. This is surprising in light of the fact that the asset-weighted average expense ratio of the index funds was 23 basis points higher than the expense ratio of the two ETFs.
The 11 index funds demonstrated an average tracking error of eight basis points: 47 of the 55 basis points in differential can be accounted for by the expense ratio. The tracking error of the ETFs was 45 basis points.
S&P SmallCap 600 Index
Eight distinct index funds track the S&P SmallCap 600 Index, whereas only one ETF tracks the same index (see Figure 9). (There are, however, iShares ETFs tracking both the S&P SmallCap 600/Barra Value Index and the S&P SmallCap 600/Barra Growth Index).
The SmallCap ETF had a tax cost ratio of 0.26, only three basis points better than the 0.29 asset-weighted average for the eight index funds. As for performance, the ETF was superior to the eight index funds by 30 basis points-which is explained by the 32 basis point differential in expense ratios.
Russell 2000 Index
In Figure 10, the iShares ETF that tracks the Russell 2000 Index was compared against the asset-weighted average of 10 index funds. The ETF had a superior tax cost ratio by 12 basis points. The three-year return of the iShares ETF trailed the Russell 2000 Index by 15 basis points, but that was five basis points better than anticipated given its expense ratio of 20 basis points. The average three year return of the 10 index funds was 10.69 percent, or 79 basis points behind the raw index and 64 basis points behind the ETF. The underperformance of the index funds is larger than would be expected given an average expense ratio of 60 basis points.
For the NASDAQ-100 Index, the asset-weighted average annualized return of the seven index funds trailed the index by 96 basis points over the three-year period ending December 31, 2004. That's remarkable considering the average expense ratio of 114 basis points-the funds managed to pick up 18 basis points through performance.
The NASDAQ-100 ETF (QQQQ) turned in a stronger performance, trailing the benchmark by just 20 basis points. This performance was entirely explained by the fund's expense ratio (20 basis points). The tax cost ratio between the ETF and the index funds was identical.
MSCI EAFE Index
The Morgan Stanley Capital International EAFE Index is tracked by 16 distinct funds and one ETF. The iShares ETF posted a higher return than the index funds by a margin of 48 basis points. Surprisingly, the iShares ETF was within four basis points of the return of the EAFE index, or 31 basis points better than expected given its expense ratio of 35 basis points. The asset- weighted return of the 16 index funds was 52 basis points lower than the raw index, which nearly coincides with the average expense ratio of 53 basis points. Finally, the ETF had a 16 basis point better tax cost ratio.
To summarize, among the nine equity indexes analyzed, ETFs that tracked them had a lower expense ratio than corresponding index funds in every case. ETFs also had superior threeyear returns in seven out of nine cases. Index funds that track the S&P 500 and S&P MidCap 400 demonstrated superior assetweighted performance compared to the corresponding ETFs.
On the tax question, ETFs also came out generally on top. ETFs had better tax efficiency in four out of nine cases, and in four cases, the tax efficiency of ETFs and index funds were essentially the same. In only one case (the S&P 500/Barra Growth Index) was the tax efficiency of index funds clearly superior to that of the corresponding ETF.
In Figure 13, we list the best performing index fund and ETF relative to each of the nine equity indexes, based on the three- year performance window ending December 31, 2004. For this particular table we also included the S&P SmallCap 600/Barra Value and Growth indexes, and the corresponding best performing index fund and ETF. It is interesting to note that the three- year return of the best performing index fund was generally higher than the return of the best performing ETF. This suggests that some index funds can, and will, compete with the lower cost structure of ETFs.
Chamberlain, Mark and Jay Jordan, "An Introduction to Exchange-traded Funds", Barclays Global Investors Services, http://www.ishares.com/material_download.jhtml?relativePath=/repository/material/downloads/intro_to_etfs.pdf