Electronic copy available at: http://ssrn.com/abstract=1337708

Active vs. Passive Management: New Evidence from Exchange Traded Funds

by

Gerasimos G. Rompotis Senior Auditor- KPMG Greece

Researcher- National and Kapodistrian University of Athens, Greece 25 Ypsilantou Street

Peristeri, Athens, Greece GR 121 31

+0030 210 5776510 grompotis@kpmg.gr

Abstract

This paper expands the debate about “active vs. passive” management using data from active and passive ETFs listed in the U.S. market. The results reveal that the active ETFs underperform both the corresponding passive ETFs and the market indexes. With respect to risk-adjusted returns, both active and passive ETFs provide investors with no positive excess returns, an expectable finding for the passive ETFs but not for the active ETFs which are aimed at beating the market. Going further, the underperformance of active ETFs is depicted to the low performance rates such as the Sharpe or the Treynor ratios they receive relative to the passive ETFs and the indexes. Furthermore, regression analysis on the selectivity and market timing skills of ETF managers indicate that the managers of both the active and passive ETFs are lacking in such skills. However, the passive managers are not expected to have such skills. Finally, tracking error estimates indicate that the discrepancy between ETF and index returns is greater for active ETFs. However, this result is to be expected as the active ETFs do not target to replicate the performance of the indexes.

February 2009

JEL CLASSIFICATION: G12, G15 Keywords: Active ETFs, Performance, Managers, Selection Skills, Market Timing

Electronic copy available at: http://ssrn.com/abstract=1337708

1

1. Introduction The performance of actively managed mutual funds and the ability of managers to produce above-normal returns have been thoroughly examined by the literature. Blake et al. (1993), Malkiel (1995), Gruber (1996) find that the actively managed portfolios, on average, do not produce better returns than the market indexes or the passively managed index funds. The authors connect the inability of managers to “beat the market” with the increased expenses incurred by the active management. In contrast, Ippolito (1989) alleges that the fee- and expense-free risk-adjusted returns of mutual funds are comparable to returns delivered by index funds while the portfolio management fees do not affect funds’ performance. However, the findings of Ippolito are subject to the benchmark employed and experienced extended critique by the literature. In general, the inability of active mutual funds to generate persistent superior returns along with the low expenses of indexing contributed to the massive growth of indexing and to the boosting of the passively managed investing products, like index funds or ETFs. Two significant elements that affect the performance of active mutual funds are the security selection and the market timing ability of managers. The selection ability implies that the manager can predict the future return of individual securities and picks these that could perform better. The timing ability indicates that the manager increases successfully the portfolio’s exposure to the market index prior to market accessions and decreases exposure before the market recessions. On the other hand, passive managers are not supposed to possess any selectivity and timing skills since their core investing strategy is to invest to the holdings of a benchmark portfolio trying to replicate its performance. The findings of the literature on the selectivity skills of fund managers are ambiguous. Carhart (1997) finds no evidence on skilled or informed portfolio managers. However, Jensen (1969) and Elton et al. (1992) find limited evidence that managers who apply stock selection strategies can produce positive superior returns over long-run periods. Also, the literature accentuates the existence of selection ability over short-run periods, whose length does not usually exceed the one year. Goetzmann and Ibbotson (1994), Grinblatt et al. (1995), Wermers (1999) reveal some evidence on successful security selection, which partially explains the short-run persistence in mutual funds’ performance.

Considering the ability of managers to time the market, the records of the literature are as mixed as the findings about the stock selection ability. Treynor and Mazuy (1966), Henriksson and Merton (1981), Chang and Lewellen (1984), Henriksson (1984), Graham and Harvey (1996) report limited or non-existent market timing ability. The main characteristic of these studies is that returns are considered monthly or annually. The usage of more frequent return data could lead in different inferences about the managers’ market timing ability according to Bollen and Busse (2001) and Chance and Hemler (2001). Indeed, these authors adopt daily return data and demonstrate that the mutual fund managers exhibit significant timing abilities.

This paper expands the research on “active vs. passive” management and fund managers selection and timing skills using data of three active ETFs recently launched in the U.S. market and comparing them to the corresponding passively managed ETFs. These active ETFs were created by PowerShares. The study covers the period from the inception of active ETFs on April 30, 2008 up to the end of November, 2008.

Active and passive ETFs have significant structural differences. The very basic difference between them is that the passive ETFs are structured to track a specific public market index while the active ETFs aim at beating the market. Active ETFs are

Electronic copy available at: http://ssrn.com/abstract=1337708

2

assigned a benchmark index while they are usually designed to track a popular investment manager’s top picks, mirror any existing mutual fund or pursue a particular investment objective and target to offer investors above-average returns. Other differences between the passive and active ETFs concern the number of market makers required by each type of ETFs (at least two and one market maker for passive and active ETFs respectively), the minimum size of investment (not required by passive ETFs but required by active ETFs) and the relationship between the market maker and the ETF manager. More specifically, these investing participants are not related to each other for passive ETFs while the market maker and the manager of an active ETF belong to the same company.

To the best of our knowledge, XTRA was the first exchange where actively managed ETFs were traded while the U.S. market launched active ETFs for the first time during 2008. The reason that active ETFs are not yet widely available relates to the arbitrage opportunities offered by passive ETFs but not offered by active ETFs. Arbitrage opportunities arise when a declination between the trading prices of ETFs and the value of underlying securities exists. The efficient arbitrage execution contributes to the narrowness of these divergences. Arbitrage bases on the in kind creation/redemption process of passive ETFs and it is attainable because the holdings of the tracking market indexes are publicly known throughout the trading day. In contrast, the stocks held by the actively managed ETFs are not publishable information until the end of the trading day because these stocks are picked by active ETFs so that they can beat the benchmark index. Therefore, if the holdings of active ETFs are disclosed frequently enough so that arbitrage could take place, their ability of beating the market weakens. In such a case, investors would simply let the fund managers do all of the research waiting for the disclosure of their choices and they would then buy the selected securities avoiding to pay the management fees. Thus, the arbitrage and the in kind creation/redemption are non-events for active ETFs.

Our results show that the active ETFs fail to beat the market. We assess market returns by alternatively use the benchmark assigned to each ETF and the Standard and Poor’s 500 Index. We use the two alternative benchmarks because PowerShares report the returns of their ETFs in reference both to the single benchmarks and to the S&P index. Moreover, the active ETFs (in two out of the three cases) underperform their passive ETF counterparts. Considering risk-adjusted returns, both active and passive ETFs do not provide investors with any positive excess return. While this finding is natural for passive ETFs, it is not for active ETFs which are aimed at beating the market and delivering above normal market returns. Going further, we rate the performance of active ETFs, passive ETFs and market indexes and find that the active ETFs basically receives lower ratings than the passive ETFs and market indexes. Considering the security picking and market timing skills of ETF managers, the results indicate lack of such skills for active ETFs. The same pattern applies to passive managers, who are not expected to have such skills anyway. In the last step, we compare the active and passive ETFs using the tracking error, which estimates the return difference between ETFs and indexes. Results indicate that in any case the active ETFs have greater tracking error than passive ETFs; a result that was to be expected since active ETFs do not target to replicate the performance of the indexes.

The remainder of this paper is organized as follows. In Section 2 we develop the methodology used in explaining and rating the various types of risk-adjusted return, in evaluating the selection and timing skills of ETF managers and measuring the tracking error of ETFs. Section 3 describes the data used in this study and provides the descriptive statistics of the sample. The empirical findings are presented in Section 4 and the summary and conclusions are discussed in Section 5.

3

2. Methodology 2.1 Risk-Adjusted Performance To analyze the managerial skill to achieve superior returns by picking stocks that could outperform the index we use the risk-adjusted return in Jensen’s model:

Rp,i-Rf = αp,i + βp,i(Rm-Rf) + εp,i (1)

where, Rp,i denotes the daily portfolio’s return for the sample’s ETF i. Rm represents the return of the market portfolio and Rf is the risk-free rate, expressed by the daily risk-free return provided by French’s website. The coefficient αp,i (Jensen’s alpha) is used to determine the excess return of the ETF i and measures the stock selection ability of the managers. If the market is efficient and the portfolio of ETF i is properly priced, the expected alpha should not be different to zero. Positive and significant alphas indicate that the manager adds value while negative and significant alphas indicate that the manager fails to well diversify the portfolio he manages or that he picks stocks that are overpriced. However, we should note that this analysis basically applies to the active ETFs because the flexibility for passive ETF managers to choose stocks or other securities that are different to these that compose the tracking index is limited or non-existent.

The coefficient βi measures the part of the ETF’s i statistical variance that cannot be mitigated by the diversification provided by the ETF portfolio, because it is correlated with the return of the other stocks that are included in portfolio. Beta stands for the systematic risk of ETF i and evaluates the degree of its sensitivity to the movements of the benchmark. εi represents the residuals of regression equation (1). 2.2 Rating Performance In this section, we rate the performance of ETFs. The first rate we use concerns the total percentage returns achieved by ETFs during the studying period. We calculate total return by subtracting the return on the first day of the studying period from the return on the last day of the period and dividing to the return on the first trading day. The total return of indexes is also calculated.

The second rating method we use is the Sharpe ratio, which is estimated via the following equation (2):

ip

fip

ip

RRS

,

,

, σ−

= (2)

where, ipR , denotes the average daily portfolio’s return for the ETF i and fR is the

average daily risk-free rate. σp.i is the standard deviation of ETF’s i return used as a proxy for the ETF’s risk. The Sharpe ratio is estimated by the division of ETF’s i excess return to its risk and is used to determine how well the return of the ETF i compensates the investor for the per unit risk they take. The higher the Sharpe ratio is, the better is the performance of ETF.

Additionally to Sharpe ratio, we estimate the Treynor ratio which is expressed by the next formula (3):

4

ip

fip

ip

RRT

,

,

, β−

= (3)

where, ipR , and fR are defined as above and ip,β is the systematic risk of ETF i.

Again, the higher the Treynor ratio is, the better is the performance of ETF. The last criterion used to rate the performance of ETFs is the Jensen’s alpha derived from the regression equation (1). Positive (negative) and significant values of alpha are connected with value-adding (non value-adding) portfolio management and therefore ETFs with such alphas should be placed at the top (bottom) of the ranking. 2.3 Testing Market Timing Ability In general, the market timing ability of fund managers implies the efficient increase or decrease in the portfolio’s exposure on equities prior to market accessions or decreases, respectively. However, the manager’s market timing ability is affected by the investing objective of the fund, which probably obliges the manager to keep the same maximum or minimum portion of equities in his portfolio. In this case, the manager tries to time the market just by adjusting the equity synthesis of the portfolio, substituting the overpriced stocks by undervalued stocks. The successful market timing is also affected by the restricted or unrestricted usage of leverage and derivative products.

To test the market timing ability of ETF managers we use the model developed by Treynor and Mazuy (1966) and described in Bollen and Busse (2001) and (2004). The model is expressed by the following equation (4):

Rp,i-Rf = αp,i + βp,i(Rm-Rf) + γp,i(Rm-Rf)

2 + εp,i (4) where, Rp,i, Rm and Rf are defined as above and γp,i measures timing ability. If the manager increases (decreases) efficiently the portfolio’s exposure to market index prior to market accessions (recessions), γp,i will be positive resulted from a convex function of portfolio’s return with respect to market return. We note that we apply model (4) to passive ETFs too but we do not expect the passive ETF managers to present any meaningful timing skills as they are not required to. They just need to follow the tracking index and to timely adjust their portfolios whenever the synthesis of the benchmark changes.

Bollen and Busse (2001) suggest that the daily tests on market timing are more powerful than monthly tests and that mutual funds exhibit more significant timing efficiency in daily tests than in monthly tests. In our paper, we evaluate the market timing skills of ETF managers using daily data and expecting statistically significant estimations for γ coefficients of models (4) for active ETFs and insignificant γ estimates for passive ETFs. 2.4 Estimating Tracking Error The deviation between the performance of index funds and the performance of the corresponding indexes is defined as “tracking error” and this issue has attracted great interest in the literature. The most common issue in passive portfolio management is that fund managers usually fail to replicate accurately the return of their corresponding indexes. In our study, we choose the three most commonly used methods of tracking

5

error measurement described in Frino and Gallagher (2001). The first one, ΤΕ1,Ρ, is straightforward and defined as the standard error of performance regression (1).

The second one, ΤΕ2,Ρ, computes the tracking error by calculating the average of absolute differences between the returns of ETFs and the corresponding indexes. We take into account the absolute value of return differences because either a positive or a negative difference reflects a performance deviation. This estimation is expressed in model (5):

ΤΕ2, Ρ = n

n

t

pte∑=1 (5)

where Pte is the absolute return difference in day t.

Finally, the third method, ΤΕ3,Ρ, computes the standard deviation of return differences between ETFs and their indexes. The estimation of this tracking error is presented in equation (6):

ΤΕ3, Ρ = 2

11

1 )(∑=

− −n

t

pptnee (6)

where pte is the return difference in day t and pe is the average return difference over n

days. We should note that the investing strategy of active ETFs concerns the

outperformance of the market while the passive ETFs aim at replicating the market. As a result, the tracking error of active ETFs should be higher than this of passive ETFs. In addition, we note that tracking error is basically an efficient performance comparison among passively managed investing products but it is applied to actively managed funds too. 3. Data and Descriptive Statistics Active ETFs were first introduced in the U.S. market on April 30, 2008 by PowerShares. More specifically, Powershares launched 4 active ETFs. On the other hand, the history of passively managed ETFs is much longer. The first passive ETF listed in AMEX in early 90’s was the SPDRS which track the performance of the Standard and Poor’s 500 Index. Currently, there are five active ETFs available in the U.S. market.

Our sample consists of three active ETFs and three corresponding passive ETFs that in pairs have the same index of reference. More specifically, our active ETFs are the Active AlphaQ Fund, Active Alpha Multi-Cap Fund and the Active Low Duration Fund. The first active ETF rate companies with more than $400 million market capitalization traded in the United States. Weekly, the Fund's Sub-Adviser generates a master stock list that ranks these stocks based on its proprietary methodology. Stocks are selected based on factors such as strong earnings growth, low valuations and positive money flow. Fund managers then define their universe as the 100 largest Nasdaq-listed Global Market Securities from the master stock list. The manager of the second active ETF of the sample adopts the same investing approach but he defines his investing choices from the 2,000 largest stocks of companies with varying capitalizations from his list. Finally, the third active ETF invests at least 80% of its assets in a portfolio of U.S. government, corporate and agency debt securities. The fund is limited to 25% of its total assets in non-investment grade securities. Under normal market conditions, the fund's effective duration will be in the range of zero to three years. The corresponding benchmarks are the Nasdaq 100 Index, Russell 3000 Index

6

and the Barclays Capital 1-3 Yr US Treasury Index. The respective passive ETFs are the Powershares Trust Series 1, the iShares Russell 3000 Index Fund and the iShares Barclays 1-3 Year Tr. BF. The main strategy of these passive ETFs concerns the investing in indexes’ components to the same weightings and rebalances whenever it is necessary due to changes to indexes’ holdings.

We should point out that PowerShares report the returns of their ETFs with respect both to the returns of their own benchmarks and the returns of the Standard and Poor’s 500 Index. Therefore, we alternatively use this index as a proxy for the U.S. stock market so as to additionally compare the returns of active ETFs to the returns of the market.

We gathered our data by several resources. In particular, the trading prices of active ETFs were supplied from yahoo.finance, the prices of passive ETFs were collected from Nasdaq.com. Nasdaq.com also provided us with the closing prices of Nasdaq 100 Index and Standard and Poor’s 500 Index. Finally, we gathered the historical data of the rest two indexes from iShares.com.

Table 1 presents the descriptive statistics of ETFs and indexes. More specifically, the table presents the average and median daily returns, the standard deviation of returns along with the minimum and maximum returns occurred during the studying period. The results indicate that the equity active ETFs underperform both their benchmarks and the corresponding passive ETFs and they also underperform the Standard and Poor’s 500 Index. In contrast, the bond active ETF outperforms its passive counterpart but not the benchmark. Moreover, the results show that the actively managed ETFs are riskier either than the passive ETFs or the benchmarks. Overall, the results demonstrate that the active ETFs fail to deliver higher returns than the passive investing products or the market indexes while they load investors with greater risk. 4. Empirical Results 4.1 Regression Results in Explaining Risk-Adjusted Performance The regression results for the risk-adjusted performance are presented in Table 2. Included in the table are the value of Jensen’s alpha which stands for the risk-adjusted return, the beta coefficient of each ETF, t-statistics which evaluate the statistical significance of the estimates, R-square values and the number of observations. According to the results, both active and passive ETFs do not produce any material above-normal return. The alpha estimates of each active or passive ETF is negative while the statistical significance of the estimates is poor. Particularly, there are two significant alpha estimates. The first one concerns the regression of the first active ETF’s return on the return of the S&P 500 Index. This coefficient is significantly negative at the 10% level indicating that the specific ETF offer investors negative excess return with respect to the market return. The other significant alpha estimate relates to the third passive ETF investing in the Barclays Capital 1-3 Yr US Treasury Index. This alpha coefficient is negative and significant at the 10% level indicating that this ETF achieves significant underperformance in relation to the benchmark return.

Overall, the results suggest that the ETFs follow the market return failing to produce any significant abnormal returns. In fact, since the alpha estimate of the benchmarks is a priori zero, we ascertain that both the active and passive ETFs perform worse than the indexes. These findings were expectable for the passive ETFs, whose effort to replicate the index return is negatively affected by factors such as the expenses. However, the findings contradict the main assertion of active ETFs which are

7

expected to beat the market offering investors above the average returns. Our results are in line with the relevant findings of the literature about the U.S. mutual funds [see, Blake, Elton and Gruber (1993), Malkiel (1995), Gruber (1996)]. Interpreting results as an indicator of market’s efficiency, we perceive that the U.S. ETF market is efficient enough offering managers low or no chances of gaining irregular returns. Considering the results as an indicator of the selection skills of the active ETF managers, the negative alpha estimates show that these managers either fail to diversify properly the portfolios they manage or fail to add value for investors by detecting and picking stocks that are undervalued. However, we should point out that the study covers a period that is characterized by the dramatic financial distress, market crisis and prices volatility. These features should make us be very sceptic when judging about the skills of the active ETF managers.

Considering beta estimates, the beta of the first active ETF is significantly lower than the unity. This pattern applies when we regress the performance of the ETF either on the return of its own benchmark or on the performance of the S&P 500 Index. On the other hand, the beta estimates of the passive ETFs are higher and closer to the unity. These results indicate an investing conservatism for the first active ETF with respect to the market and the passive ETF competitor.

Considering the beta coefficients of the second pair of active and passive ETFs, the results in Table 2 show that the beta of the active ETF is higher than the beta of the passive counterpart being simultaneously statistically indifferent from the unity. In this case, the active ETF is more aggressive than the passive one. Finally, the beta of the bond active ETF is significantly lower than the unity and inferior to the beta estimate of the corresponding passive ETF. Therefore, the third active ETF is less aggressive than the market and its passive counterpart.

We should note that theoretically, conservatism helps funds to be less sensitive to the market movements protecting them in a bear market but it restrains them in a bull market. This means that funds’ return declines less than benchmarks’ return when the market goes down but funds’ return ascends slowly in comparison to indexes’ return when the market moves upwards. However, this theoretic aphorism is not empirically verified by the results of our study since in the current bear market the conservatism of active ETFs indicated by the results (in two of three cases) does not prevent them from declining more with respect to the market returns. 4.2 Results in Rating Performance In this section we present the rating of ETFs and indexes performance according to the total return for the period, the Sharpe ratio, the Treynor ratio and the Jensen’s alpha. The results are presented in two sub-panels in Table 3. The first panel presents the ratings of ETFs with respect to their own benchmarks while the second one presents the ratings of ETFs in regards of the ratings received by the S&P 500 Index.

The ratings confirm the underperformance of the active ETFs relative to the passive ETFs and the indexes described in the previous two sections. More specifically, Panel A reports that the ratings of any type assigned to the first active ETF are inferior to the ratings given to the passive ETF competitor and the Russell 3000 Index. This is the case for the second active ETF too. Considering the third active ETF, its rating is inferior to the rating of the benchmark and basically superior to the rating of the passive ETF (but not in terms of Treynor ratio).

Panel B reports that the rating of equity ETFs, both the active and the passive ones, is basically inferior to the rating of the S&P 500 Index. Regardless the rating method, the performance of ETFs is assigned lower rating in comparison to the index’s

8

rating. Considering the bond ETFs, we see that these ETFs achieve positive total returns during the studying period while the rates they receive in accordance to the Sharpe and Treynor ratios are positive. Comparing the equity and bond investing instruments, we see that investors could replace their equity holdings with bond ones so as to mitigate the losses from the rapid and deep decline in stock markets during the period we assess. 4.3 Regression Results in Testing Market Timing This section discusses the estimations of the Treynor and Mazuy 1966 model (4) which evaluates the market timing skills of ETF managers. The regression results are presented in Table 4. Presented in the table are the values of alpha, beta and gamma coefficients, the values of the relevant t-statistics, the R-squares and the number of daily trading observations available for each ETF. The gamma coefficient assesses the market timing skills of the managers. Positive and significant gamma estimates indicate that the managers efficiently time the market. Negative or insignificant gamma estimations imply that the ETF managers are lacking in sufficient market timing skills. Moreover, we note that the model mainly applies to active ETFs since passively managed ETFs are not supposed to time the market. According to the results in Table 4, the active ETF managers do not possess any significant market timing skills. The gamma estimates of the applied model are either positive or negative but insignificant at any acceptable level. For the equity active ETFs, the finding of non-existent timing skills for the managers applies either when we use the specific benchmark of each ETF or the S&P 500 Index.

Considering the results for the passive ETFs, the gamma estimates do not differ statistically from zero when the benchmarks’ returns are used as a proxy for the market performance. Interestingly, when the S&P 500 Index is used as the market performance the model derives statistically significant gamma estimations for passive equity ETFs. The estimate for the first ETF is positive and the estimate for the second one is negative. However, their values (0.010 and -0.003 respectively) is not considerably different from zero. Focusing on the active ETFs, the results of the applied model demonstrate that their managers failed to predict the depth of the unexpected financial crisis and adjust accordingly their investing decisions. Overall, our results are consistent with the previous findings of Treynor and Mazuy (1966), Henriksson and Merton (1981), Chang and Lewellen (1984), Henriksson (1984), Graham and Harvey (1996) on the time abilities of active mutual fund managers. All of these studies report restricted on non-existent market timing ability for mutual fund managers. The common feature of these studies is that the returns are considered on a monthly or an annually basis while, in our study, returns are considered daily. Therefore, our findings are more comparable to the results of Bollen and Busse (2001) and Chance and Hemler (2001). These authors apply daily tests on the market timing efficiency of mutual funds revealing that the managers possess material market timing skills. However, this inference does not apply to active ETF managers, at least according to our results. 4.4 Results in Estimating Tracking Error The last segment of the paper presents the estimations of ETFs’ tracking error. Tracking error denotes the difference in returns of ETFs and indexes. Tracking error calculations are presented in Table 5. The first three columns show the results of each one of the three types and the fourth column is the average tracking error of the three alternative estimates.

9

The results show that the tracking error of active ETFs is higher than that of the passively managed ETFs. This pattern applies both to the return comparison between ETFs and their assigned indexes of reference and the performance differences between ETFs and the S&P 500 Index. Considering the individual equity active ETFs, the maximum average tracking error concerns the Active Alpha Multi-Cap Fund, which tracks the Russell 3000 Index. The magnitude of the average tracking error of this ETF is equal to 1.753. On the contrary, the corresponding passive ETF presents lower tracking error relative to the QQQQ which tracks the Nasdaq100 Index. The maximum average tracking of QQQQ is equal to 0.344. Comparing the equity ETFs to the bond ETF, the results indicate that the tracking error estimates of the bond ETF is significantly lower than the tracking errors of the equity ETFs. This finding applies either to active or passive ETFs and it is reasonable since in general bond ETFs are less risky and expensive than equity ETFs and they consequently derive lower tracking errors in relation to equity ETFs. Overall, the fact that the active ETFs produce higher tracking error relative to the passive ETFs is a reasonable and expectable finding since the active ETFs aim at beating the market returns rather than replicating them. 5. Summary and Conclusions The ability of mutual fund managers to beat the market by applying efficient selection and market timing strategies has been thoroughly examined by the literature. The findings on this issue are ambiguous, which means that there are studies that report material outperformance of actively managed mutual funds with respect to the market returns but there are also studies that demonstrate that the active mutual funds basically underperform the market indexes and the passively managed index funds. In this paper, we expand the research on the “active vs. passive” management debate by providing new evidence from U.S. listed ETFs.

The results of our study are summarized as follows: At first, considering the raw return we find that the active ETFs basically derive lower percentage returns than the passive ETFs and the market indexes. Simultaneously, the active ETFs are found to be more risky than the passive counterparts and the benchmarks. Considering risk-adjusted performance expressed by Jensen’s alpha, we find no evidence that the active ETF managers achieve significant excess returns with respect to the average market returns and risk-adjusted performance of the passively managed ETFs, which are not supposed to deliver any excess return anyway. The underperformance of active ETFs is also depicted to the lower ratings (total return and Sharpe and Treynor ratios) they receive relative to the indexes and passive ETFs.

The findings about risk-adjusted performance may suggest that the U.S. ETF market is efficiently enough providing managers with no significant opportunities for gaining abnormal returns. On the other hand, the results can be interpreted as an indicator for the managers’ lacking in selection skills. This indicator implies that the active ETF managers fail to detect and select the stocks that are undervalued. However, we should bear in mind that the unprecedented crisis in stock markets worldwide impoverishes the investing choices on the managers.

Considering the market timing abilities of ETF managers, the results indicate that managers are not able to time the market efficiently. In regression analysis, we find that market timing coefficients (γ estimates) are basically negative and insignificant. While passive ETF managers are simply follow the market without trying to time it, the active ETFs are supposed to time the market efficient so as to help the portfolios they

10

manage to outperform the market return. Therefore, our findings indicate that the active ETFs fail to do what they are required to. Moreover, the lack of market timing skills may contributes to the failure of active ETFs to achieve significant abnormal returns.

Finally, with respect to the tracking ability of ETFs, we find that the active ETFs have greater tracking error than the passive ones. However, the active ETFs do not aim at replicating the market returns and therefore the findings are reasonable.

Overall, our empirical results about the performance of the active and passive ETFs are in line with the previous findings of the literature on the performance of active mutual funds. The lack of significant risk-adjusted performance for active ETFs due to inadequate selection and market timing skills support the existing literature on mutual fund performance and managerial behavior with a new set of data having different operating characteristics.

11

References Blake, C.R., Elton, E.J. and Gruber, M.J., 1993, “The Performance of Bond Mutual Funds”, Journal of Business 66 (3), pp. 371-403. Bollen, N.P., and Busse, J.A., 2001, “On the timing ability of mutual fund Managers”, Journal of Finance 56, pp. 1075-1094. Bollen, N.P., and Busse, J.A., 2004, “Short-Term Persistence in Mutual Fund Performance”, Review of Financial Studies 18 (2), pp. 569-597. Chance, D.M., and Hemler, M.L., 2001, “The Performance of Professional Market Timers:. Daily Evidence from Executed Strategies”, Journal of Financial Economics 62, pp. 377-411. Chang, E.C., and Lewellen, W.G., 1984, “Market Timing and Mutual Fund Investment Performance”, Journal of Business 57(1), pp. 57-72. Carhart, M.M., 1997, “On Persistence in Mutual Fund Performance”, Journal of Finance 52 (1), pp. 56-82. Edelen, M. R., 1999, “Investor Flows and the Assessed Performance of Open-End Mutual Funds”, Journal of Financial Economics 53, pp. 439-466. Elton, E.J., Gruber, M.J., Das, S., and Hlavka, M., 1993, “Efficiency with costly information: A reinterpretation of evidence from managed portfolios”, Review of Financial Studies 6, pp. 1-22. Frino, Alex and David R. Gallagher, “Tracking S&P 500 Index Funds”, Journal of Portfolio Management, 2001, Vol. 28 (1), pp. 44-55. Goetzmann, W.N. and Ibbotson, R.G., 1994, “Do Winners Repeat? Patterns in Mutual Fund Performance”, Journal of Portfolio Management 20, pp. 9-18. Graham, J. and Harvey, C.R., 1996, “Market Timing Ability and Volatility Implied in Investment Newsletters' Asset Allocation Recommendations,” Journal of Financial Economics 42, pp. 397-421. Grinblatt, M., Titman, S., and Wermers R., 1995, “Momentum Investment Strategies, Portfolio Performance, and Herding: A Study of Mutual Fund Behavior”, American Economic Review 85 (5), pp. 1088-1105. Gruber, M.J., 1996, “Another Puzzle: The Growth in Actively Managed Mutual Funds”, Journal of Finance 51, pp. 783-810. Henriksson, R.D., 1984, “Market Timing and Mutual Fund Performance: An Empirical investigation”, Journal of Business 57, pp. 73-96. Jensen, M.C., 1969, “Risk, The Pricing of Capital Assets, and The Evaluation of Investment Portfolios,” Journal of Business 42, pp. 167-247.

12

Ippolito, R, 1989, “Efficiency with Costly Information: A Study of Mutual Fund Performance, 1965-1984”. Quarterly Journal of Economics 104 (1), pp 1-23. Malkiel, B.G., “Returns from Investing in Equity Mutual Funds 1971 to 1991”, Journal of Finance, 1995 50 (2), pp. 549-572. Treynor, J., and Mazuy, K., 1966, “Can Mutual Funds Outguess the Market?”, Harvard Business Review 44 (4), pp.131-136. Wermers, Russ, 1999, “Mutual Fund Herding and the Impact on Stock Prices”, Journal of Finance 54 (2), pp. 581-622.

13

Table 1 Descriptive Statistics

This table reports the mean and median daily return, the standard deviation and the minimum and maximum return, ratio of active ETFs, passive ETFs, benchmarks, Standard and Poor’s 500 Index, the SPDRS which track the S&P 500 Index (passively managed ETF) and the risk-free rate during the period 05/01/2008- 11/28/2008.

Names Symb. Mean Median Stdev Min Max Obs. Active ETF Active AlphaQ Fund PQY -0.295 0.000 3.041 -10.412 13.571 148 Passive ETF Powershares Trust Series 1 QQQQ -0.286 -0.269 2.834 -8.956 12.165 148 Index NASDAQ 100 Index -0.280 -0.194 3.009 -10.519 12.580 148

Active ETF Active Alpha Multi-Cap Fund PQZ -0.388 0.000 3.613 -12.375 13.919 148 Passive ETF iShares Russell 3000 Index Fund IWY -0.258 -0.052 2.857 -9.072 10.429 148 Index RUSSELL 3000 Index -0.259 -0.031 2.970 -9.143 11.475 148

Active ETF Active Low Duration Fund PLK 0.019 0.000 0.703 -2.564 2.863 148 Passive ETF iShares Barclays 1-3 Year Tr. BF SHY 0.012 0.036 0.193 -0.561 0.712 148 Index Barclays Capital 1-3 Yr US Tr. In 0.026 0.035 0.197 -0.913 0.761 148

Index S&P 500 Index -0.253 -0.016 2.869 -9.035 11.580 148

14

Table 2 Performance Regression Results

Rp,i-Rf = αp,i + βp,i(Rm-Rf) + εp,i This table presents the results of performance regression. We regress the risk-adjusted daily return of ETFs on the risk-adjusted daily return either of their benchmarks or the Standard and Poor’s 500 Index. Ri is the return of ETFs, Rf is the risk-free return and Rm denotes the market’s return. T-tests for alpha and beta coefficient evaluate the difference of the estimates from zero and unity respectively. *Statistically significant at the 1% level. **Statistically significant at the 5% level. ***Statistically significant at the 10% level.

Symb. Underlying α t-test β t-test R2 Obs. Active ETF PQY NASDAQ 100 Index -0.085 -1.349 0.781* -5.111 0.720 148 Passive ETF QQQQ NASDAQ 100 Index -0.005 0.003 1.003 0.271 0.989 148 Active ETF PQY S&P 500 Index -0.123*** -1.936 0.729* -3.552 0.619 148 Passive ETF QQQQ S&P 500 Index -0.071 -0.847 0.882* -3.209 0.806 148

Active ETF PQZ RUSSELL 3000 Index -0.103 -0.882 1.061 0.821 0.741 148 Passive ETF IWY RUSSELL 3000 Index -0.008 -0.994 0.973** -2.483 0.994 148 Active ETF PQZ S&P 500 Index -0.128 -0.873 0.992 -0.115 0.605 148 Passive ETF IWY S&P 500 Index -0.014 -0.672 0.977 -1.299 0.931 148

Active ETF PLK Barclays Capital 1-3 Yr US Tr. 0.000 -0.024 0.646 -1.562 0.342 148 Passive ETF SHY Barclays Capital 1-3 Yr US Tr. -0.009*** -1.665 0.818* -4.029 0.701 148

15

Table 3 Performance Rating

This table presents the rating of funds’ and benchmarks’ performance per group according to total return, Sharpe ratio, Treynor ratio and Jensen’s alpha. ***Statistically significant at the 10% level.

Panel A: Rating of ETFs with respect to their benchmarks Symbols Total Return Sharpe Treynor Jensen

Active ETF PQY -39.672 -0.097 -0.378 -0.085 Passive ETF QQQQ -38.318 -0.101 -0.285 -0.005 Index NASDAQ 100 Index -38.168 -0.093 -0.280 0.000

Active ETF PQZ -48.970 -0.107 -0.366 -0.103 Passive ETF IWY -35.788 -0.090 -0.265 -0.008 Index RUSSELL 3000 Index -36.133 -0.087 -0.259 0.000

Active ETF PLK 2.503 0.027 0.030 0.000 Passive ETF SHY 1.727 0.061 0.014 -0.009*** Index Barclays Capital 1-3 Yr US Tr. 3.841 0.130 0.026 0.000

Panel B: Rating of ETFs with respect to the S&P 500 Index Symbols Total Return Sharpe Treynor Jensen

Active ETF PQY -39.672 -0.097 -0.405 -0.123*** Passive ETF QQQQ -38.318 -0.101 -0.325 -0.071 Active ETF PQZ -48.970 -0.107 -0.391 -0.128 Passive ETF IWY -35.788 -0.090 -0.264 -0.014 Active ETF PLK 2.503 0.027 N/A N/A Passive ETF SHY 1.727 0.061 N/A N/A Index S&P 500 Index -35.317 -0.088 -0.253 0.000

16

Table 4 Market Timing Regression Results

Rp,i-Rf = αp + βp(Rm-Rf) + γp,i(Rm-Rf)2 + εp,i

This table reports the results of regression which evaluates the timing ability of ETF agers. Ri is the return of funds, Rf is the risk-free return and Rm denotes the market’s return. T-test counts for the statistical significance of coefficients. The timing ability implies that the managers respond efficiently to the movements of the stock market and revise the portfolios they manage. The timing ability is assessed via the regression’s gamma estimate. *Statistical significant at the 1% level. **Statistically significant at the 5% level. ***Statistically significant at the 10% level.

Symb. Underlying α t-test β t-test γ t-test R2 Obs. PQY NASDAQ 100 Index -0.107 -1.398 0.780* 18.122 0.002 0.511 0.721 148

QQQQ NASDAQ 100 Index 0.002 0.799 1.001* 122.47 -0.001 -1.760*** 0.989 148 PQY S&P 500 Index -0.136*** -1.712 0.728* 9.363 0.002 0.188 0.619 148

QQQQ S&P 500 Index -0.157 -1.602 0.872* 23.339 0.010*** 1.827 0.810 148

PQZ RUSSELL 3000 Index -0.089 -0.862 1.068* 14.381 -0.003 -0.274 0.752 148 IWY RUSSELL 3000 Index 0.004 0.624 0.973* 91.877 -0.001 -1.591 0.994 148 PQZ S&P 500 Index -0.117 -0.694 0.994* 15.067 -0.001 -0.133 0.605 148 IWY S&P 500 Index 0.008 0.367 0.976* 57.874 -0.003*** -1.748 0.934 148

PLK Barclays Capital 1-3 Yr -0.013 -0.497 0.658** 2.035 0.319 0.396 0.346 148 SHY Barclays Capital 1-3 Yr -0.011*** -1.784 0.820* 17.988 0.054 0.700 0.702 148

17

Table 5 Tracking Error Results

This table presents the estimations of tracking rrror, which reflects the deviation between the return of ETFs and their underlying indexes. We apply three distinct methods in tracking error estimating, labeling them as ΤΕ1, ΤΕ2, and ΤΕ3. TE1 is the standard deviation of return differences between ETFs and indexes. TE2 is the absolute average return difference between ETFs and indexes. TE3 is the standard errors of ETFs’ performance regression.

Symbol Underlying TE1 TE2 TE3 Average TE(1+2+3)

PQY NASDAQ 100 Index 2.022 1.249 1.643 1.638 QQQQ NASDAQ 100 Index 0.482 0.232 0.318 0.344 PQY S&P 500 Index 2.401 1.430 1.919 1.917

QQQQ S&P 500 Index 1.312 0.812 1.254 1.126

PQZ RUSSELL 3000 Index 1.941 1.461 1.857 1.753 IWY RUSSELL 3000 Index 0.309 0.186 0.232 0.242 PQZ S&P 500 Index 2.372 1.616 2.294 2.094 IWY S&P 500 Index 0.983 0.329 0.777 0.696

PLK Barclays Capital 1-3 Yr 0.743 0.413 0.589 0.582 SHY Barclays Capital 1-3 Yr 0.127 0.084 0.106 0.106