Exchange Traded Funds (ETF): History, Mechanism, Academic Literature Review and Research Perspectives

This paper is a response to the Russian ETF anomaly that demonstrates how costly for individual investors might be the absence of ETFs on domestic markets. Their absence and, therefore, the lack of investors’ familiarity with these products might be one of the several reasons why Russian investors have been buying ETF shares via Russian mutual funds rather than investing directly. Their extra spending thus reaching up to 36% of the capital invested over a 10-year horizon. This review aims to acquaint broader investor communities with exchange-traded funds, one of the most important financial innovations of the last 20 years, and provides a survey of the research done in this field. It reviews the literature on exchange-traded funds, provides a brief history of the market and these funds’ operational mechanism. This article complements the three previous ETF surveys, one of which is 10-year old and the other two are not comprehensive. The paper also suggests an alternative literature classification. In conclusion, some ideas for further research are proposed.


Introduction
The development of the exchange-traded funds market impresses by its dynamics. In December, 2014, the total value of ETFs' asset under management (AUM) reached 2.64 trln. USD ( Fig. 1), with the annual trade volume of 18 trln. USD. In the USA, for several years approximately a quarter of equity trade volume consisted of ETFs' shares. Currently, investors in most countries can trade ETFs on their domestic markets (Tab. 1).   Russia is still not on the list of countries where the annual volume of ETF trades is above one million USD. At the same time, exchange-traded funds became the most popular financial instrument among the Russian mutual fund industry. The way of investing is, however, unusual: unlike other countries, where people invest in ETFs directly, in Russia investors do it via domestic mutual funds (MF). However, when investing for 10 years in an ETF via a Russian fund of funds, an investor pays up to 36% of invested capital more in commissions than someone who invests in the same ETF directly (Tab. 2).

Table 2
Extra payment (as a percent of the invested capital) for investing in an ETF via a MF compared to direct investing.
Investment period, years 1 5 10 Extra payments, % of the invested capital 4-7% 14-18% 27-36% Source : Tarassov, 2016a For the last two years (2014)(2015), the absolute majority of the funds of funds have changed their investment strategy completely. Since then they have been investing only in one preselected western ETF out of the TOP 100 [Tarassov, 2016a]. During the same period, the group "funds of funds" has raised more money than any other fund category in Russia (Tab. 3). Madhavan [2014] focused mostly on research investigating recently emerged products: actively managed funds, leveraged and so-called smart-beta products.
This literature review is a logical supplement to the three previously published surveys.
The paper introduces an alternative literature classification, covers articles not included in earlier published reviews, and suggests some additions in ETF type's classification and systematization of the market development stages. The survey provides a brief history of the market and a description of the fund working mechanism. Additionally, in conclusion, some ideas for further research are proposed.
The existing ETF research is divided in three groups that unites six research areas: The first group is devoted to a traditional 4 ETF. It includes two areas: Are ETFs a substitute to index mutual funds? If yes, than to what extent? If the ETF is a substitute and is more convenient and less expensive, why do investors continue buying index mutual fund units?
These questions are reviewed from several angles: tax effectiveness, performance, transaction costs and tracking error.
Area 2. How do ETFs influence underlying assets of an index they track? This area also includes such topics as liquidity, arbitrage and hedging.
The second group of papers that is relatively new unites three areas: Area 3. How effective are the ETFs that track foreign 5 indexes?
Area 4. The ETF market development outside of the USA.
Area 5. ETFs other than those reflecting equity indexes. The introduction of a new generation funds: synthetic, actively managed, leveraged and so-called "smart-beta" products.

The third group of research covers:
Area 6. The application of ETFs for optimal portfolio construction.
As Deville noted, despite the extensive business literature devoted to ETFs, academic research did not pay practically any attention to these funds until 2000. Just after the NASDAQ ETF introduction, the situation changed. However, the quantity of research papers is not comparable with that of the ETF business literature. increased product range. The last part of the survey covers the papers about the application of ETFs for portfolio construction. Several suggestions for further research are included in the conclusion.

Emergence and development
The idea to have an opportunity to buy or sell a portfolio in one trade used to be popular among a part of the market participants many years ago. Based on Gastineau [2002] and Deville [2008] a brief history of ETF market development is as follows.
The first modern portfolio trades took place among large investors at the end of the 1970s.
Mostly these portfolios consisted of companies' equity included in S&P 500. It took ca. 10 years until a North American small investor had the possibility to buy such a portfolio in one trade.
Although the products were successful, they could not overcome legal barriers in the USA, and only continued to be available in Canada.
After the break, in 1993, the first modern exchange traded fund was listed on the American USD) and the largest ETF (215 bln. USD under management) 6 .

In three years, Barclays Global Investors introduced a series of ETF products -World
Equity Benchmark Shares (WEBS), which were renamed later as iShares 7 MSCI. WEBS started tracking MSCI indexes 8 . Thanks to WEBS investors received easy access to trade in portfolios consisting of foreign assets.
In 1999, US investors could choose among ETFs tracking not only S&P 500, S&P 400, MSCI, but also DJI 9 , NASDAQ 100 and sectoral S&P. Researchers relate the trade volume explosion to the Cubes (Qubes, QQQ) introduction that tracks NASDAQ 100. In 2001, three ETFs that track S&P 500, DJI and NASDAQ 100 were traded not only on AMEX but also on NYSE. Boehmer and Boehmer [2003] proved that this was the key factor responsible for a radical bid-ask spread cut.

The ETF market today
In the 2000s the exchange trade funds gained popularity in the majority of the market economies (tab. 1 and 5). Institutional investors started buying ETFs more often.
Product range increase. ETFs are listed on NYSE.
A radical cut in funds' spread and commissions.
ETFs became popular among individual investors.
Legal barriers have been overcome for ETF creation.
IT solutions for index trading in seconds are established.
Individual investors more often prefer to trust their money to funds rather than to invest directly. (Fig.3) Passive investing gains more often in Active/Passive dispute. Last 10 years ETFs consist not only of companies' equity (tab.6). Practically all ETFs are passively managed funds tracking an index 11 .
The largest coverage in the business literature belongs to funds suggesting their own index construction criteria: weights, dividends, Sharpe ratio etc. This category received the name "smartbeta" funds. Their AUM reached 410 bln. USD. Another type of fund, "leveraged", has 23 bln.
USD under management.
However, the largest part of funds, with AUM 1 165 bln. USD, simply reflects the established indexes. These funds might be called classical funds.
There are 3 906 registered ETFs in the world. 407 started in 2014.

Brief working mechanism description
Based on Gastineau [2001], Deville [2008] and Investment Company Institute reports it is possible to describe the working mechanism of a traditional ETF as follows (Fig. 4). 10 Other funds tracking precious metals and natural resources prices (53 bln. USD) are ETF type funds that are not included in ETF statistics because of some legal nuances (please see the working mechanism description). 11 Actively managed funds' AUM is 2,7 bln. USD. There are two markets: the market between sponsors and APs and a secondary market.
Any investor can buy or sell ETF shares on the secondary market as she does with shares in other companies. 90% of ETF trade takes place on the secondary market 13 .
A sponsor and its AP exchange large modules of ETF shares for companies' shares. It is an "in-kind' buy or sell operation.
Thus, an AP has arbitrage possibility. When demand for an ETF's shares drives its prices up, the AP, buys underlying companies equity and exchanges it for ETF shares to sell them on the secondary market. This operation pushes the difference between the ETF price and that of underlying asset to a minimum. This mechanism practically guarantees negligible tracking error in most cases in the high liquidity fund sector.
The second advantage of such a mechanism is that an ETF, unlike a mutual fund, does not regularly calculate client's profit and so does not pay capital gains tax every quarter 14 . The third important difference between an ETF and a mutual fund is the relatively lower fund administration costs. An ETF, unlike a mutual fund, does not keep a client's files. A practical absence of the bid-ask spread in the high liquidity fund sector drives investors' expenses even further down.
An investor's profit from investing in a fund (ETF or MF) can be expressed as: (1) net profit; capital gain -price difference between fund's unit/share price at the moment of buying and at the moment of selling; ∑Т = Тах 1 + Тах 2: 1capital gains tax, calculated quarterly; 2capital gains tax, calculated after a fund's units/shares are sold; 1custody and broker commission; 2front-end and back-end; 3management fees; is a difference between price of buying and selling at one given moment; tracking error; In an ETF, in contrast to an MF: 1 = 0, 2 = 0.
The working mechanism, showed previously, is typical for the funds that physically buy assets included in the index the fund reflects. Usually such funds are called traditional funds.
For funds, that do not buy assets in a physical sense, the market developed the name "synthetic". To create such funds, a sponsor uses swaps, derivatives or statistical methods to mirror the prices of underlying assets. Goltz and Tang [2010] described the working mechanism and other detailsof synthetic exchange-traded funds. It is important to note that some of these relatively new products cannot be called ETFs in many countries. Due to legislation of these countries, these products receive the Before proceeding with the literature review, let us fix the classification of exchangetraded funds (Fig. 5).

Figure 5. Exchange-traded funds classification
Source: created by the author.

An ETF versus an identical index fund
2.1.1. Taxes and performance. One of the major differences between a MF and an ETF is that a MF needs to distribute its profit quarterly 16 [Guedj and Huang, 2010] fund's performance before taxes was stronger than that of SPDR. The authors tried to find the possible reasons.
Firstly, an ETF holds dividends on accounts that do not bear interest and they do not reinvest them immediately in the way a MF does. The researchers calculated that this factor could play an important role for long-term investment. Secondly, the management fee of the Vanguard fund is less than that of SPDR. Thirdly, the tracking error of SPDR was found to be larger. 16 USA. Some countries do not have this rule.

Depends on the indexes tracked
Index, existed before the creation of ETF that tracks this index.
Index, created for an ETF. Sometimes it is a modified previously established index.
Fund uses swaps, derivatives and statistical methods to reflect price dynamics of an index.
Fund physically buys assets, included in index.

Classical | Smart-beta
However, the authors continued with the disadvantages of a MF. For example, a MF has to keep some reserves available for the situations when investors may want to sell their units. In general, MF's administration expenses are larger than these of an ETF. A MF has to keep books of every client.
Both of these papers [Elton, Gruber, Comer and Li, 2002;J. Poterba and J. Shoven, 2002] are cited very often when investigating ETF taxes and performances. However, these two papers use the data for the period of 1993-2000. Just after this period, ETF expenses decreased dramatically [Boehmer and Boehmer, 2003]. Gastineau [2004] responding to these papers argues that the reasons for the lower SPDR performance lie in the situation that an ETF could not ignore the trade demand at the end of a day (as opposed to a MF). Sometimes, the trades at the day end have the potential to perturbate the index tracking, particularly during the changes in the index itself. The author finds this to be a temporary problem. He also presents data showing, that after the period studied in those both papers the SPDR performance before taxes became better than that of Vanguard index 500 fund.

ETF influence on underlying assets. Liquidity, hedging and arbitrage
After the introduction of an ETF tracking DJI, the liquidity level of underlying assets increased [Hedge and McDermott, 2004]. The authors noted that the liquidity of the ETF became even higher than that of the underlying assets themselves. In general, derivatives' trade volume also increased. Additionally, their prices become more "fair", first, thanks to an increased liquidity level [Deville, Gresse, Severac, 2009] and effective arbitrage mechanism [Gastineau, 2001]. Camalia, Deville and Riva [2014] proved that possible liquidity problems in days of crisis could happen to ca. 20% of ETFs with the lowest day volumes. Their model shows that the ETF's liquidity level depends not only on the liquidity of underlying assets but also on several characteristics of the ETF itself, first of all, on its daily trade volume.
During the period 1998-2001 prices of ETFs and their underlying assets needed on average ten minutes in the USA and three hours in other countries to become equal [Engel, Sarkar, 2002]. During 2001-2010, it needed less than five minutes on average for SPDR tracking error to become practically equal to zero [Marshall, Nguyen, Visaltanachoti, 2013]. In this period, arbitrage operations resulted in earning of 6.7% p.a. on average. With an increase in the product range, the introduction of leveraged products and listing of ETFs on even more stock exchanges, the number of arbitrage opportunities will only increase. Alexander and Barbosa [2008], having investigated hedging possibilities against the possible volatility of the largest ETFs, developed a system for minimizing the risks. In particular, the authors focused on price changes around dividend payments days.

ETFs, tracking foreign 17 assets
Continuing with liquidity and arbitrage issues, Ackert and Tian [2008] found that the liquidity level of funds tracking foreign assets is lower than that of those tracking major American indexes. The authors also demonstrate that a lower liquidity level results in lower arbitrage possibilities. Ackert and Tian concluded that the mechanism that minimizes tracking error in such cases does not work correctly. They calculated that this mechanism has a U-shaped function depending on liquidity and works to full potential only after the trade volume reaches a certain level. Svetina [2010] and Poterba and Shoven [2002], prove that, in general, the effectiveness of ETFs tracking foreign assets is lower than that of funds reflecting domestic stocks. The authors mention the following possible reasons: tax retention from dividend payments in the issuer-country, lower liquidity, substantially larger tracking error, time difference and higher transaction costs.
ETFs tracking global emerging markets have an even larger tracking error. In addition, if, in order to lower transaction costs and overcome liquidity constraints, a fund is organized in a synthetic way, its tracking error is even higher [Blitz, Huij, 2012]. However, classical ETFs reflecting global indexes have a better Sharp ratio than identical MFs [Harper, Madura, Schnusenberg, 2006]. Huang and Lin [2011] concluded that there is no important difference in performance between direct investing in foreign assets directly, investing via index MFs or via ETFs. Miffre [2007] notes that the ETFs, in addition to providing an investor with very high international diversification level, have another very important advantage, compared to the MFs,the possibility of short selling.

ETF market development outside the USA
As shown in table 1, ETFs are traded on the majority of stock exchanges. There is research into how the domestic ETFs perform in many countries. Most of these papers come to a similar conclusion regarding the performance of domestic ETFs and the positive influence on domestic markets. European research, in additional, focuses on new generation products and the application of ETFs for portfolio optimization [Deville et al., 2003;Deville et al., 2009;Deville et al., 2014;Musavian, Hirsch, 2002;Simon, Sternberg, 2005;Rompotis, 2012;Zanotti, Russo, 2005]. The authors demonstrate that many European ETFs have a larger tracking error than the US ETFs, supposing that this is the result of the non-traditional structure of the funds. Chu [2011] also finds that tracking error on the Chinese market is, in general, larger than that in the USA. In addition to "standard' issues, Yao [2012] investigates conflicts of interest, manipulation and insight problems on the Chinese ETF market. Finally, another peculiarity of the Chinese ETF market is that several research papers, discussed below, are devoted to gold ETFs, which is not typical for developed markets.

ETFs that reflect price dynamics of assets other than equity. New generation funds:
actively managed, leveraged, synthetic and "smart-beta"  [Naumenko, Chystyakova, 2015] and a higher liquidity level than underlying assets [Camalia, Deville, Riva, 2014]. However, in general, the level of liquidity and transparency continues to be an object of criticism. European regulators try to find a compromise between defenders of synthetic and traditional ways of ETF structure [Millet, 2013]. However, since 2010 the industry has been returning to a traditional form, in particular in the area of equity funds. Competition for retail clients, who prefer better transparency, lowered the status of synthetic ETFs from the level they reached in 2006-2008. 18 Goltz and Tang [2010] described the working mechanism and other details of this type of exchange-traded funds.
3.3.4. Actively managed exchange-traded funds not only confirmed the observations that active MFs do not outperform passive MFs [Gruber, 1996;Malkiel, 2005], but also demonstrated even worse overall return [Rompotis, 2009].
3.3.5. Currency ETFs. Ivanov [2015] investigates reasons of tracking error of these funds.
The author calculated that spread, fund commissions and transaction costs are responsible for it.
Ivanov suggests that his methodology is applicable to gold and natural resources funds. The author does not compare currency funds with other instruments.
3.3.6. «Smart-bets» ETFs. Madhavan [2014] describes in detail the working mechanism, methods of calculation of various indicators and advantages of such funds. He adds that creating index criteria for these ETFs is an intensive activity that means that such funds could be perceived as active funds.

ETF for optimal portfolio construction
De Freitas and Baker [2005] posed the question of whether exchange-traded funds themselves are a complex solution for optimal portfolio construction. In this paper, the authors gave a positive answer to this question. Furthermore, they devoted a special role in optimal portfolio construction to bond ETFs.
The application of exchange-traded funds for portfolio construction was found also in new paradigm -Core-Satellite Theory. Core-Satellite Theory, the paradigm that many leading investment companies (Goldman Sachs, UBS, Vanguard etc.) started following, divides a portfolio in two parts: core and satellites. The core consists of passive instruments like bond ETFs. The satellites invest in selected potentially more profitable and more risky strategies. According to a survey conducted in 2010, most investors, following this strategy, use ETFs in core [Goltz, Tang, 2010].
Amenc, Goltz and Grigorou [2010], describe the next stage of this method of portfolio construction -Dynamic Core-Satellite Theory. The dynamic application of bond and equity ETFs increases potential profit without adding new risks. Puelz, Carvalho and Hahn [2015] analyze the latest trends in the wealth management industrysharp AUM increase in two companies: Wealthfront and Betterment. These companies offer automatic asset management using ETFs. The authors, having developed a model of picking the right exchange-traded funds, prove that one may build an optimal portfolio with a high Sharpe ratio by using a very limited number of ETFs.

Conclusion and research suggestions
Despite the increasing amount of academic research focused on ETFs, there are still several questions that are answered only partly or not covered at all.
There is still no answer to the most frequently posed questionwhy classical traditional ETFs have not yet replaced identical index mutual funds. One research direction might be an investigation of the differences between MFs' active marketing, including distribution channels, and the virtual absence of ETF marketing activities. For example, some MFs have a budget for incentivizing intermediaries while ETFs are not engaged in such activities. On the other hand, various countries introduce barriers for advisors to receive commission from product issuers.
Moreover, there is a clear tendency in the remuneration system of independent advisors to evolve towards fixed commission paid by clients. The potential influence of these latest trends on the ETF market has not been studied yet.
There are no studies of the physiological aspect of the choice between MFs and ETFs. In general, the investor's attitude towards ETFs from a behavioral finance point of view has not yet been investigated. Tarassov [2016b] studies the relationship between people's predisposition to categorical thinking and non-optimal index investing, including choosing an index MF instead of an identical ETF, might be a small contribution to this area. Additionally, a lack of index-investing culture in countries with immature financial markets could be one of the reasons for even more widespread irrational behavior of investors, e.g. the Russian ETF anomaly [Tarassov, 2016a].
Another important question might be whether classical traditional exchange-traded funds, thanks to their transparency and absence of specific risk, could be attractive for investors who do not trust the stock market and invest only in saving deposits (other fixed income instruments) or real estate. While in the USA, where more than a half of households participate in the stock market (directly or via funds) 19 , such research might be less important, in countries where the stock market investors' share is at a minimal level such a study may probably have a stronger impact. What could be a trigger for capital owners, investing in real estate, to turn their attention to more liquid real estate ETFs? Probably, constructing a model that would allow a comparison of direct investment in real estate with investment via an exchange-traded fund would be a challenging topic for research.
Another interesting topic for study is the application of ETFs for optimal portfolio construction in the concept of "passive management revolution" or in "focus shifting from a single security picking capabilities importance to importance of asset class choosing" 20 . The research into application of ETFs in Dynamic Core-Satellite Theory is a very good start in this direction.
The question about regulation regarding non-traditional funds remains slightly separate. So far, there is not yet a comparative study between different approaches in the USA and in Europe. Is it important to limit the industry only by classical traditional ETFs? Does trust in the ETF market decrease with an increase in the number of funds with complicated structure? Or, on the contrary, does a broader product range results in more attention to the ETF market?
There are no studies on how ETFs have been introduced in transition economies and why there are still stock exchanges without this type of funds. There is no research into ETFs' potential impact on these countries markets.