1. Introduction
Exchange traded funds (ETFs) were introduced in 1993 when the American Stock Exchange (AMEX) listed the Standard and Poor’s Depositary Receipts (SPDRs), which track the S&P 500. In Europe the first ETFs, tracking the Euro Stoxx 50 and the Stoxx Europe 50, were listed on the German market in 2000. Since the introduction of these products, the industry has experienced rapid growth, and according to a BlackRock Investment Institute report, at the end of June 2011, 1,185 and 1,039 ETFs were listed in Europe and in the USA, respectively. The estimated value of the assets under management for the European ETFs is $321.2 billion, while in the US it is $973.5 billion.
1。介紹
在1993年被引入美國證券交易所（AMEX）上市的標準普爾存托憑證（SPDRs的），跟蹤標準普爾500指數的交易所買賣基金（ETF）。在歐洲第一的ETF，跟蹤Euro Stoxx 50指數和歐洲斯托克50，于2000年在德國市場上上市。推出這些產品以來，該行業已經歷了快速的增長，并根據向貝萊德投資研究所的報告，在2011年6月底，1,185和1,039只ETF上市，在歐洲和美國。歐洲交易所買賣基金所管理的資產估計價值為321.2十億美元，而在美國，它是973.5十億美元。

This study investigates the tracking error of traditional and synthetic ETFs2 traded in Europe. The objective of ETFs is to track the returns of the benchmark index as closely as possible. Traditional ETFs attempt to fulfill this objective by holding the benchmark underlying securities, while synthetic ETFs use derivatives contracts, primarily total return swaps. However, neither type of ETF can guarantee that their performance perfectly matches the returns of the benchmark index.In Europe, ETFs are UCITS funds.3 One interesting feature of the European market that does not exist in the US because of regulatory constraints,4 is the synthetic replication method that was introduced in the French market in 2001. Synthetic ETFs hold a basket of securities that usually do not match the index’s underlying securities and than swap their return with that of the benchmark index. The swap counterparty is usually the parent company of the ETF provider. http://www.ttc-switch.com/dissertation_writing/Ecommerce/
本研究探討在歐洲交易的傳統和合成ETFs2的跟蹤誤差。交易所買賣基金的目的是盡可能地跟蹤基準指數的回報。傳統ETF試圖達到這個目的，通過持有標的證券基準，
而合成交易所買賣基金使用衍生工具合約，主要是總回報掉期。然而，無論是類型的ETF可以保證他們的表現完美的的基準index.In歐洲的回報相匹配，交易所買賣基金的UCITS funds.3歐洲市場的一個有趣的功能，在美國不存在由于監管方面的限制，4是合成復制的方法，于2001年在法國市場推出。合成ETF持有一籃子證券，通常不匹配指數的標的證券和不是交換與基準指數收益率。掉期對手通常是ETF供應商的母公司。


Synthetic ETFs appear to have some advantages over traditional funds, but they can also incorporate additional risks. The synthetic replication model makes it possible to create ETFs that track indices that otherwise would be very difficult to reach because of restrictions on foreign investments. Moreover, synthetic ETF providers claim that the synthetic replication method is more efficient and produces a lower tracking error when compared with traditional ETFs. However, the main concern with this replication strategy is counterparty risk. According to UCITS regulations, counterparty risk cannot exceed 10% of the ETF’s net asset value (NAV). However, if the swap counterparty defaults on its obligations, the ETFs might face a loss failing to track the return of the benchmark index. Ramaswamy (2011) notes that the synthetic replication strategy transforms the tracking error risk into counterparty risk and highlights the potential systemic risks that this replication strategy can create. Large liquidation of synthetic ETFs in periods of higher counterparty risk could force to sell the collateral, often illiquid assets. This in turn can hinder the correct functioning of the markets. He concludes that the market risk of these products can be underestimated. Furthermore, the synthetic replication method represents a serious threat to the traditional flagship qualities of ETFs, i.e., simplicity and transparency.
  合成ETF似乎比傳統的資金有一定的優勢，但他們也可以加入額外的風險。合成復制模型使得它可以創建的ETF，跟蹤指數，否則會很難達到，因為對外國投資的限制。此外，合成ETF供應商聲稱，合成復制的方法更有效，并產生一個較低的跟蹤誤差時，與傳統ETF相比。但是這種復制策略，主要關注的是交易對手風險。根據UCITS規例，
交易對手風險不能超過ETF的資產凈值（NAV）的10％。然而，如果掉期對手方未能履行其責任，交易所買賣基金可能面臨虧損，未能跟蹤基準指數的回報。拉馬斯瓦米（2011）指出，合成復制策略，將追蹤誤差風險，交易對手風險，并強調潛在的系統性風險，這種復制策略可以創建。大清算合成ETF的交易對手風險較高的時期，可以強制出售抵押品，往往是流動性不足的資產。反過來，這可能會妨礙市場的正常運作。他得出結論認為，這些產品的市場風險可能被低估了。此外，合成復制的方法，代表傳統的旗艦本色交易所買賣基金，即，簡單性和透明度的嚴重威脅。
  
Previous empirical studies identify the main factors that give rise to tracking errors, including transaction costs, index-composition changes, corporate activity, fund cash flows, index volatility, the reinvestment of dividends, and index replication strategies (Chiang 1998; Elton,
Gruber, Comer, and Li 2002; Frino and Gallagher 2002). The volatility of the exchange rate is a further source of tracking error (Shin and Soydemir 2010).

This article finds that both traditional and synthetic European ETFs are affected by a significant tracking error. This analysis provides evidence that ETFs that follow a synthetic replication strategy rather than holding the underlying benchmark securities, enjoy a lower tracking
error and a higher tax efficiency. Furthermore, they are particularly efficient when tracking emerging-market benchmarks. However, synthetic ETFs underperform both the benchmarks and the traditional counterparts

This paper extends the previous literature by examining the tracking ability of traditional and synthetic European ETFs offered by the leading providers in Europe. The synthetic replication method, although widely used by European ETF providers, has never been deeply analyzed by previous studies. However, synthetic ETFs became a serious concern for regulators6 also because of the potential implications for the stability of the financial system. This research is motivated by the need to provide investors, who seek a transparent and cost-efficient passive investment strategy, with more insight regarding the tracking ability of traditional and synthetic ETFs.

The remainder of the paper is organized as follows. The next section reviews the related literature. Section 3 presents the data and the empirical design. Section 4 reports the results and the final section concludes.

2. Literature Review
Previous academic research, based primarily on US and Australian markets, documents significant tracking errors generated by index funds and ETFs (Elton, Gruber, Comer, and Li 2002; Frino and Gallagher 2001 and 2002). Elton, Gruber, Comer, and Li (2002) and Harper, Madura, and Schnusenberg (2006) compare the return of the ETFs and the corresponding index return. Roll(1992), Pope and Yadav (1994) and Larsen and Resnick (1998) identify three metrics to measure the tracking error as the dispersion of the fund’s NAV return relative to the benchmark return. Tracking error can also be evaluated using market prices instead of the NAV (Harper, Madura, and Schnusenberg 2006). Any market-price deviations from NAVs should disappear quickly because of the in-kind and in-cash creation/redemption processes (Elton, Gruber, Comer, and Li 2002; Engle and Sarkar 2006), but the market price tracking error can substantially deviate from the NAV tracking error. DeFusco, Ivanov, and Karels (2011) show that the pricing deviations of the Spiders, Diamonds and Cubes7 are different from zero. The authors claim that the pricing deviation can be considered an additional cost of administering an ETF. Furthermore, seasonal patterns in tracking errors have been detected (Frino and Gallagher 2001 and 2002; Frino, Gallagher, Neubert, and Oetomo 2004; Rompotis 2010).

Chu (2011) investigates the tracking errors of ETFs traded in Hong Kong and finds that they are higher compared with those in the US and Australia. He assumes that one possible explanation could be the use of synthetic investment tools instead of holding the underlying stocks. Other recent research by Blitz, Huij and Swinkels (2010) examines the tracking error of
European index funds and ETFs as measured by their underperformance against the gross total return indices. They find that European funds underperform their benchmarks and that dividend withholding taxes and fund expenses have similar explanatory power.

3. Data and Method
In this section, the data and the method used in this research are described.
3.1. Data
The analyzed sample comprises traditional and synthetic ETFs listed in Europe. According to the BlackRock Investment Institute report, at the end of June 2011, 1,185 ETFs were listed in Europe. For the purposes of this study, the ETFs that track the major European and global stock market indices are selected. To ensure a significant data history, only the ETFs that were created before September 2007 are considered to allow for the analysis of a common period of four years
that starts with September 2007 and ends in August 2011. This four-year period has been selected as a trade-off because it offers a meaningful data history while also including a significant number of both traditional and synthetic ETFs from the leading providers of ETFs in Europe.

The final sample consists of 48 ETFs that track 20 different benchmark indices: 21 are traditional ETFs and 27 use a synthetic replication method. Providers comprise Blackrock (iShares),
Sociétè Générale (Lyxor), Deutsche Bank (db x-trackers), State Street Global Advisors, BBVA, UBS, Amundi, EasyEtf, Powershares, and Credit Suisse. The replication method, as well as total
expense ratio (TER) and other characteristics are acquired from the fund prospectuses. The funds are domiciled in Ireland, France, Spain, and Luxembourg. This sample is one of the most diverse for ETFs in the academic literature in terms of the benchmark indices, the providers and the
replication strategies included. Table 1 reports the profiles of the 48 ETFs, including the name, replication method, net or gross benchmark index, primary listing, inception date, and total expense ratio (TER).


For the other three regressions that test the significance of the determinants of the tracking error measured by TEAAD,TESDRD and TESER, the coefficients on TER are statistically indistinguishable from zero. Thus, while fund expenses generate the tracking error measured by
TERD, i.e., reduce the performance of the ETF, they do not generate variability of the return differences of the ETF. The coefficient on SYNT is negative and significant. This also confirms the descriptive statistics. Synthetic ETFs are more efficient in tracking benchmark indices because they generate a lower tracking errors compared to traditional ETFs. The coefficients on OPT and EMERG are positive and significant, that shows ETFs that track the underlying benchmark by holding a subset of the constituent securities and ETFs that track emerging-market indices, generate a higher tracking error. The coefficient on the interaction term SYNT × EMERG is negative and statistically significant as expected. It shows that synthetic ETFs are more efficient when tracking emerging markets benchmark indices compared to the ETFs that follow a traditional replication method. Finally, the coefficients on the dummy variables LU, FR, and ES are significant. Therefore the domicile of the fund is significant in explaining the tracking error.