Be Active With Your Bond Trackers

Transcription

1 Be Active With Your Bond Trackers October 2004 Noël Amenc Professor of Finance, EDHEC EDHEC Business School, and Director Risk and Asset Management Research Centre Jean-René Giraud C.E.O, EDHEC-Risk Advisory Philippe Malaise Professor of Finance, EDHEC Business School Lionel Martellini Professor of Finance, EDHEC Business School, and Scientific Director Risk and Asset Management Research Centre

2 Abstract Newly launched fixed-income Exchange-Traded Funds (ETFs) have specifically been designed to track bond market indices, and share many of the same benefits of equity ETFs, including in particular lower costs, transparency, buying and selling flexibility, all day tracking and trading. While it has often been argued that ETFs were natural investment vehicles for implementing passive indexing strategies, we show in this paper that the benefits of ETFs are actually much larger than traditionally reported, as these instruments can also be used to implement almost the full range of existing investment strategies. From a pure asset management perspective, bond portfolio managers typically attempt to put forward relative advantages to their competitors such as information advantage, technical or judgemental skills so as to beat their benchmark index. Broadly speaking, there are two kinds of active strategies. The first form of active bond portfolio strategies consists of trading on interest rate predictions, either on the basis of directional bets (bets on changes -or the absence of changes- in the level of the yield curve) or non-directional bets (bets on changes -or the absence of changes- in the slope or curvature of the yield curve). These strategies are also known as yield curve timing strategies. In this paper, we first explain in Section 1.1 how bond portfolio managers can use bond ETFs to implement butterfly strategies, which are the most common form of non-directional trading strategies. We further argue (Section 1.2) that ETFs can also be used to implement directional bets on an increase or decrease of interest rates under the form of maturity rotation strategies (decrease or increase in a portfolio duration in response to active views on changes in the level of interest rates). The second form of active bond portfolio strategies consists of trading on market inefficiencies. These strategies are also known as bond picking strategies. While these strategies can obviously not be implemented by trading in ETFs, it should be noted that they are usually focused on corporate bond markets, as opposed to Treasury bond markets. It can actually be shown that bond picking strategies (e.g. rich-cheap analysis) only generate modest returns when implemented in liquid efficient markets such as the Treasury bond markets in most developed countries. Therefore, the opportunity loss for an active manager to trade Treasury bond ETFs as opposed to individual Treasury bonds is very limited. This stands in sharp contrast to equity markets where the preferred form of active strategies consists of betting on a stock s specific risk. From a more general asset-liability management perspective, portfolio managers typically need to implement duration gap management strategies, aiming at offsetting the impact of changes on interest rates in the institution s balance sheet. In Section 2, we show that bond ETFs can also be used to implement such strategies. In order to assess the full extent to which the bond ETFs can be of interest to the final investor, it is important to document the economics of the vehicle in comparison to investing directly in the underlying securities. To address this question, we have surveyed financial institutions with the objective of estimating the various costs related to the management of a portfolio of individual bonds. The results we have found are reported in Section 3. They show that investing in ETFs involves modest management costs that are on average lower than those for direct management of bonds, notably for implementing strategic allocation decisions. This research has been supported by Euronext. EDHEC is one of the top five business schools in France. Its reputation is built on the high quality of its faculty and the privileged relationship with professionals that the school has cultivated since its establishment in EDHEC Business School has decided to draw on its extensive knowledge of the professional environment and has therefore focused its research on themes that satisfy the needs of professionals. 2 EDHEC pursues an active research policy in the field of finance. EDHEC-Risk Institute carries out numerous research programmes in the areas of asset allocation and risk management in both the traditional and alternative investment universes. Copyright 2012 EDHEC

3 1. Implementing Active Portfolio Strategies with Bond ETFs In the framework of active fixed-income management, it is believed that there are investment and arbitrage opportunities that yield on average a higher return than the cost incurred to implement them. So as to identify these opportunities, portfolio managers put forward advantages relative to their competitors such as better access to information, or better ability to process available information. The first form of active bond portfolio strategies consists of trading on interest rate predictions, either on the basis of directional bets (bets on changes -or the absence of changesin the level of the yield curve) or non-directional bets (bets on changes -or the absence of changes- in the slope or curvature of the yield curve). These strategies are also known as yield curve timing strategies. In this section, we explain how bond portfolio managers can use bond ETFs to implement butterfly strategies, which are the most common form of non-directional trading strategies. We further argue that ETFs can also be used to implement directional bets on an increase or decrease in interest rates under the form of maturity rotation strategies (decrease or increase in a portfolio duration in response to active views on changes in the level of interest rates). In what follows, we show how investors can use bond ETFs to implement portable alpha strategies based on non-directional as well as directional bets on changes in the level of interest rates. For illustration purposes, we focus on the ETF tracking the EuroMTS Index, which models the totalreturn of the Eurozone bond market and currently covers the sovereign sector. It is calculated each day both in real-time and with two daily fixings using prices from the MTS system. The EuroMTS Index was formerly the CNO Etrix. The EuroMTS Index reproduces the performance of the Euro zone government bond market by modelling the performance of a limited portfolio of bonds chosen to represent the performance of the wider market. The contribution of each representative bond s price and coupon accrual to the final index is determined by its issuer s size relative to the entire Eurozone government bond market. The EuroMTS Index is comprised of six Sub-Indices based on specified maturity ranges, and dedicated ETFs will be launched to replicate the 3-5 year and year segments. 1.1 Using Bond ETFs to Implement Non-directional Bets on Changes in Interest Rates: Butterfly Strategies A butterfly is one of the most common fixed-income active strategies used by practitioners to bet on parallel shifts in term structure, structured so as to generate a profit whatever the direction of this shift. 1 It is the combination of a barbell 2 (called the wings of the butterfly) and a bullet 3 (called the body of the butterfly). The purpose of the trade is to adjust the weights of these components so that the transaction is cash-neutral and has a $ duration equal to zero. As recalled above, the latter property guarantees quasi-perfect interest-rate neutrality when only small parallel shifts affect the yield curve. Because the strategy is usually structured so as to have positive convexity, the investor is then certain to enjoy a positive pay-off if the yield curve is affected by a positive or negative parallel shift. In what follows we perform the following experiment. On the sample period ranging from 1 January 1999 to 31 December 2003, we implement at the beginning of each month, a butterfly trade that satisfies the dollar and dollar duration constraints using ETFs based on the EuroMTS index and the 3-5 and sub-indices See Martellini, Priaulet and Priaulet (2003) for an in-depth analysis of this strategy. 2 - A barbell portfolio is constructed by concentrating investments on the short-term and the long-term ends of the yield curve. 3 - A bullet portfolio is constructed by concentrating investments on a particular maturity of the yield curve. 4 - One should perhaps talk instead about euro and euro-duration constraints. 3

4 To obtain an estimate of the $ duration, we use duration estimates available for EuroMTS indices, and turn them into $ duration estimates through the following formula: Note that we assume away the presence of transaction costs in the experiment, and we also assume that we can trade any infinitesimal quantities of each ETF. Figure 1 below shows the distribution of monthly returns on that strategy based on a short position on the broad-based EuroMTS index, and long positions in both the 3-5 year and maturity sub-indices, designed so as to achieve cash and (modified) duration neutrality and positive convexity. Figure 1: Distribution of monthly returns on a systematic butterfly strategy. This figure shows the distribution of monthly returns on that strategy based on a short position on the broadbased Euro-MTS index, and long positions in both the 3-5 year and maturity sub-indices, designed so as to achieve cash and (modified) duration neutrality and positive convexity. We note that in this sample a positive return is generated in 100% of the months when the proceeds from the short positions are assumed to be invested at the risk-free rate (Eonia rate). Note also that a butterfly strategy is designed to generate a profit only in the case of a parallel shift in the yield curve, and it is therefore natural that returns below the risk-free rate are generated as well. The purpose of this experiment is simply to demonstrate that bond ETFs can be used similarly to individual bonds as investment vehicles to implement popular alpha bond strategies such as butterflies. 4 We know that the yield curve is potentially affected by many movements other than parallel shifts. These include in particular pure slope and curvature movements, as well as combinations of level, slope and curvature movements (see for example Litterman and Scheinkman (1991)). It is in general fairly complex to know under what exact market conditions a given butterfly might generate a positive or a negative pay-off when all these possible movements are accounted for. There are actually many different kinds of butterflies (some of which are not cash-neutral), which are structured so as to generate a positive pay-off in case a particular move in the yield curve occurs.

5 Generally speaking, implementing a butterfly trade amounts to placing a bet on the spread between the body and the arithmetic average of both wings. If we assume that the yield curve is only affected by parallel shifts, the change in the spread is always zero, and the strategy is generating a profit due to the positive convexity property. More generally, investors are short or long the body of the butterfly depending on their anticipations on the change in the spread. An efficient method that can be used to measure the sensitivity of a butterfly to interest rate risk in a more general context consists of calculating the level, slope and curvature $ durations in the Nelson and Siegel (1987) model, which provides a popular parameterisation of the zero-coupon yield curve. The model for the yield curve contains four parameters and allows one to obtain the standard increasing, decreasing, flat and inverted shapes (see for example Martellini, Priaulet and Priaulet (2003)). 1.2 Directional Bets on Changes in Interest Rates: Maturity Rotation Strategies A significant part of the expertise of active bond portfolio managers is the ability to implement successful bets on interest rate changes. If they think that interest rates will decrease in level, they will lengthen the $ duration or modified duration of their portfolio so as to optimise absolute or relative capital gains. On the other hand, if they think that interest rates will increase in level, they will shorten the $ duration or modified duration of their portfolio so as to minimise the absolute or relative capital losses. There is actually now a consensus in empirical finance that expected asset returns, and also variances and covariances, are, to some extent, predictable. Pioneering work on the predictability of asset class returns in the U.S. market was carried out by Keim and Stambaugh (1986), Campbell (1987), Campbell and Shiller (1988), Fama and French (1989), and Ferson and Harvey (1991). More recently, some authors started to investigate this phenomenon on an international basis by studying the predictability of asset class returns in various national markets (see, for example, Bekaert and Hodrick (1992), Ferson and Harvey (1993, 1995), or Harvey (1995)). The use of predetermined variables to predict asset returns has produced new insights into asset pricing models, and the literature on optimal portfolio selection has recognised that these insights can be exploited to improve on existing policies based upon unconditional estimates. For example, Kandel and Stambaugh (1996) argue that even a low level of statistical predictability can generate economic significance and abnormal returns may be attained even if the market is successfully timed only 1 out of 100 times Timing Rotation Strategies TAA strategies were traditionally concerned with allocating wealth between two asset classes, typically shifting between stocks and bonds. More recently, more complex style timing strategies have been successfully tested and implemented within each of these two asset classes (stocks and bonds). For example, Keim and Stambaugh (1986) find that several ex-ante observable variables based on asset price levels predict, among other things, ex-post risk premiums on U.S. Government bonds of various maturities. Different maturity indices strategies perform somewhat differently in different times and economic conditions, and there is evidence of predictability in these patterns. Using multifactor models for the return on bond indices, where the factors are chosen to measure the many dimensions of financial risks (market, volatility, credit and liquidity risks), one may be able to implement a strategy that generates abnormal return from timing between different maturity sub-indices. In an attempt to assess the performance of a maturity style timer with various levels of forecasting ability, we calculate the profit generated by investing 100% at the beginning of each month in the index with the highest return in the following month. More specifically, we show in Table 5

6 4 the performance of a tactical style timer over the sample period ranging from January 1999 to December In this experiment, 100% of the portfolio is invested in the best performing index (3-5 or year) with various degrees of predictive ability depicted by hit ratios ranging from 50% (no predictive ability) to 100% (perfect timer). The benchmark is 50% invested in each index, with rebalancing taking place at the beginning of each month to bring the allocation back to neutrality. As can be seen from the results in Table 1, we find that the performance of a style timer with perfect forecasting ability who invests 100% of a portfolio at the beginning of the year in the best performing style for the year generates an impressive information ratio equal to 4.8. Table 1: Performance of a tactical style timing strategy. This table shows the performance of a tactical style timing portfolio on the period ranging from January 1999 to December In this experiment, 100% of the portfolio is invested in the best performing index (3-5 or year) with various degrees of predictive ability depicted by hit ratios ranging from 50% (no predictive ability) to 100% (perfect timer). The benchmark is 50% invested in each index, with rebalancing taking place at the beginning of each month to bring the allocation back to neutrality. Obviously, the assumption of perfect timing ability is not realistic. It can be argued that a realistic performance for a successful style timer is consistent with a hit ratio of around 65% (see for example Amenc et al. (2003)). Such a level of hit ratio allows for a 1.27% excess return and a 1.52% annual tracking error with respect to the equally-weighted benchmark, which results in a comfortable information ratio equal to 0.83 (see Grinold and Kahn (2000) for an empirical distribution of information ratios among active managers). To test the robustness of the results, we repeat the experiment 100 times by drawing the successful months randomly, while maintaining a 65% hit ratio level. Figure 2 below shows the distribution of performance obtained by a style timer, as we let the successful 65% of the months vary across the 60 (5x12) months in the sample. As can be seen from Figure 2, the performance of the style timer is not a mere artefact of a particular choice of the winning months in the sample. This shows the robustness of abnormal performance that can be generated by a realistic timing strategy. Figure 2: Performance distribution of a style timer with a 65% hit. 6 This figure shows the distribution of performance obtained by a style timer with a 65% hit ratio as we let the months found to be successful vary across the 60 (5x12) months in the entire sample.

7 1.2.2 Absolute Return Approach Maturity rotation strategies can not only be implemented in a long-only context, but can also be used to generate absolute return benefits. Table 2 below shows the performance of a strategy for a style timer with a 65% hit ratio, and long and short positions implemented with a neutral exposure allowing 100% of initial capital to be invested in the risk-free rate (Eonia), so as to satisfy a level of leverage equal to 2. An example of allocation would be 50% in the index that is perceived as likely to underperform, and +50% in the index that is perceived as likely to outperform. It should be noted that, while the satellite s gross leverage (i.e. sum of absolute values of long and short positions) is equal to 2, the net leverage is equal to 0. Mixing long and short positions on bond ETFs allows investors to neutralise their exposure to interest rate risk. More accurately, when an investor is implementing such a strategy, the investor stays exposed to changes in the slope, as opposed to changes in the level, of the term structure of interest rates. Table 2: Absolute return approach. In this experiment, we focus on a 65% hit ratio, with long and short positions implemented with a neutral exposure allowing 100% of initial capital to be invested in the risk-free rate (Eonia) Table 2 suggests that the benefits of maturity rotation strategies can be implemented in an absolute return approach, which allows for the portability of the abnormal performance to a core portfolio invested in a broad-based index. This is what we turn to next. 2. Using Bond ETFs in the Context of Asset Liability Management In the previous section, we argued that ETFs can be used efficiently by investors who seek to implement active bets consistent with their view on future changes in the yield curve. These active bets can be beneficial if they are intentional and explicitly reflect the active views of the manager. On the other hand, if these bets are unintended they can be costly and introduce an obviously undesirable element of luck into the management process. The situation is relatively straightforward to manage in a pure asset management context, but is more complex in the presence of liability constraints. It is important to note that the proper objective of a pension fund is not the management of a portfolio s tracking error with respect to market indices, but instead the management of deviations from risk factors affecting the current value of its liabilities. The index that best represents the client objectives and rules is indeed the most appropriate benchmark. While no existing commercial index is perfectly suited to representing a specific investor s liability structure, Treasury bond indices are typically the asset class most correlated with institutional investors liabilities, and as a result should be predominant in institutional investors allocation. For example, a typical pension plan is exposed to significant interest rate risk emanating from a duration mismatch between assets and liabilities. This exposure is permanent and represents a large, unacknowledged, strategic bet on interest rates and the mismatch in duration exposes the plan to uncompensated risk. Assuming that most pension plans have an average duration of between 10 and 15 years, the opportunity cost and dramatic risk/reward difference of using a broad aggregate bond index with a duration that will be significantly lower than that of the liabilities is obvious. A given decrease 7

8 in the level of interest rates will have a more dramatic impact on the value of the liabilities than on the value of the assets, with a dramatic decrease in surplus size as a consequence. In what follows, we show how to use ETFs based on short and long maturity indices to manage the remaining duration gap that can exist between the investor s asset and liabilities. After reviewing the basics of duration gap analysis, we show how maturity sub-indices can be used as completeness portfolios to neutralise unintended biases in the allocation induced by the presence of liability constraints. 2.1 Duration Gap Analysis A mismatch between assets and liabilities leads institutions to enjoy an exposure to interest rate risk. For example, banks or insurance companies experience interest rate risk if changes in market interest rates cause bank profits to fluctuate. If a bank has more rate-sensitive liabilities than assets, a rise in interest rates will reduce bank profits and a decline in interest rates will raise bank profits. Let us define the gap as the difference between the present value of interest rate-sensitive assets and the present value of interest rate-sensitive liabilities. The interest rate risk with respect to the gap is then equal to the change in interest rate multiplied by the gap. It is well known that duration measures the elasticity of the assets' (or liabilities') market value (MV) with respect to a change in the interest rate. Duration is the weighted sum of the maturities of the payments in the financial instruments, where the weights are equal to the present value of the payment divided by the present value of the asset or liability. This allows us to measure the effects of interest rate changes on an institution s (say a bank s) net worth NW where A denotes bank assets and L bank liabilities. We now recall that $ duration can be used to compute an approximation of the absolute profit and-loss of portfolio with cash-flows F i for a small change Δy in the yield to maturity y, as can be seen from the following relationship: Absolute P & L N V $Dur Δy where, the derivative of the bond value function with respect to the yield to maturity is known as the $ duration of the portfolio. On the other hand, modified duration, or sensitivity, can be used to compute the relative profit-and-loss of the portfolio with cash-flows, as can be seen from the following relationship: Relative P & L N V x MDur x Δy where 8

9 By the definition of duration as an elasticity measure, we therefore have and where Dur A and Dur L represent duration measures for assets and liabilities, respectively. Hence, we finally get The terms in the first set of parentheses (scaled by assets A) represent the duration gap faced by the institution; that is, So, we have (2) A positive duration gap indicates that assets are more interest rate sensitive than liabilities, on average. Thus, when interest rates rise (fall), assets will fall proportionately more (less) in value than liabilities and the market value of equity will fall (rise) accordingly. On the other hand, a negative duration gap indicates that weighted liabilities are more interest rate sensitive than assets. Thus, when interest rates rise (fall), assets will fall proportionately less (more) in value than liabilities and the market value of equity will rise (fall). The fiduciary is responsible for managing the net asset/liability interest rate position. Most fiduciaries maintain a mismatch between what are typically short-duration assets and longduration benefit obligations. In an upward-sloping yield curve environment, a plan implicitly pays a cost of carry to bet on rapidly rising rates - a strategy that has proven detrimental in the past several years. On the level of the overall company, this duration mismatch may be exacerbated because the sponsor typically has medium- to long-duration debt opposite interest-insensitive (zeroduration) assets. In Table 3 below, we show the balance sheet of a hypothetical institution (insurance company). Please note that the numbers and assumptions used in the example are not necessarily realistic, as they merely serve an illustrative purpose. A similar example can be built for any type of institutional investor (bank, pension fund, etc.). Table 3: Balance sheet of a hypothetical insurance company. The numbers in the table are expressed in millions of euros. 9

10 From that example, we can calculate the duration gap. The duration gap is then 2.88 (920/1000) 6.76 = years. The duration of assets is significantly lower than the duration of liabilities. As a result, liability value changes more than asset value. This is an implicit bet on an increase in interest rates. Indeed, if interest rates increase, the value of the liabilities will decrease faster than the value of the assets, which will increase the surplus level. This bet can be beneficial if it is consistent with the manager s view about interest rates. On the other hand, in the absence of any active view on interest rates, the implicit bet induced by the mismatch between assets and liabilities introduces an undesirable element of luck into the management process. Assuming that the yield curve is flat at an 8% level, the modified duration gap is 3.34/1.08 = Hence, by equation (2), if interest rates decrease by 1%, the surplus, or net worth of the company, will lose value by 3.09% of the assets. 2.2 Duration Gap Management: Using Maturity ETFs as Completeness Portfolios to Neutralize Unintended Biases in the Core Portfolio induced by the Presence of Liability Constraints. Institutional investors can use maturity indices as completeness portfolios to neutralise these unintended biases in the allocation induced by the presence of liability constraints. Given that it is never easy for an institution to modify the structure of its liabilities, it is recommended that the investor trade in a bond ETF so as to get closer to the objective of a zero duration gap. Let us consider a situation where the institutional investor from the example in Table 3 liquidates the 900 (million) euros held in 3 and 6 year T-Bonds to invest in a broad-based European bond index such as the EuroMTS index, in an attempt to enjoy the benefits of a well-diversified portfolio. As recalled above, this is optimal from a pure asset management standpoint in the absence of active views on bond markets. On the sample period ranging from 1 January 1998 to 31 December 2003, the average sensitivity of the index is equal to 5.102, with a standard deviation equal to Let us calculate what the (modified) duration gap would be under the assumption of the assets fully invested in a bond ETF with modified duration equal to The (modified) duration gap is then equal to (920/1000) 6.76/1.08 = years. While holding the broad-based index as a proxy for the European bond market portfolio may be optimal from a passive asset management standpoint, it is therefore not entirely optimal in this example from an ALM standpoint as the remaining duration gap induces an implicit, and unintended, exposure to interest rate risk. 10 We now investigate how maturity indices can be used as completeness portfolios to customise the strategic allocation in an attempt to neutralise the interest sensitivity. As the duration of the assets is shorter than that of the liabilities, we seek to include in the portfolio an ETF on a longer duration index, namely the ETF written on the year EuroMTS sub-index. The average duration of this index over the sample period is (with a standard deviation). The (modified) duration of the assets should be equal to (920/1000) 6.76/1.08 =5.758 so as to yield a zero duration gap.

11 This can be achieved by investing the percentage p in the long maturity sub-index, where p is a solution to the following equation: p (1-p) = 5.758, that is p = 20%, or 180 million euros transferred from the core broad-based portfolio to the completeness sub-index portfolio. It would therefore appear that investing in ETFs written on broad-based indices and maturity sub-indices immunisation theory has the power to completely eliminate interest rate risk. Of course, this is not possible in practice, for a number of reasons. First, it should be noted that duration hedging only works for small yield changes, because the price of a bond as a function of yield is non-linear. In other words, the $ duration of a bond changes as the yield changes. Accounting for larger shifts in the term structure of interest rates can easily be done within the framework presented above. Another set of simplifying assumptions that duration gap management techniques are based on is the assumption that yield curve shifts are parallel, i.e. yield curve changes are small and perfectly correlated along the yield curve. Empirical evidence suggests however that short rates move more than long rates, and the amplitude of these changes can be very significant. An empirical analysis of the dynamics of the interest rate term structure suggests that two or three factors account for most of the yield curve changes. They can be interpreted, respectively, as a level, slope and curvature factors (see Section 3). There are different ways to generalise the above method to account for non-parallel deformations of the term structure, as well as option-like dependencies of assets and liabilities with respect to the interest rate level. We refer the reader to Martellini, Priaulet and Priaulet (2003) for an overview of the state-of-the art techniques that can be used for interest rate risk hedging in a generalised context. 3. The Cost-Benefits of Using Bond ETFs If there is little that portfolio managers typically do that cannot be done with bond ETFs, one legitimate question is what is the most convenient and cost-efficient way of implementing bond portfolio strategies. We have taken the opportunity to survey financial institutions with the objective of estimating the various costs related to the management of a portfolio of bonds comprising the underlying securities of the EuroMTS index. The results show that investing in ETFs involves modest management costs that are on average lower than those for direct management of bonds, notably for implementing strategic allocation decisions. 3.1 Objectives In order to assess the full extent to which the EuroMTS tracker Exchange Traded Fund can be of interest to the final investor, it is important to document the economics of the vehicle in comparison to investing directly in the underlying securities. The cost of buying and holding the EuroMTS tracker can be described by three important parameters: - Transaction costs (brokerage fees and bid/ask spread offered by the market makers to acquire and sell the ETF); - Yearly management fees; - Custody costs On the other hand, managing a portfolio of holdings matching the underlying securities of the index results in a fairly different set of direct and indirect costs: - Transaction costs (brokerage fees and market bid/ask spread offered by the market makers on the underlying government bonds); - Infrastructure and human resources costs related to the management of the portfolio; - Infrastructure and human resources costs related to the execution of orders; - Infrastructure and human resources costs related to the middle and back offices (settlement, coupons, etc.); 11

12 - Settlement and custody costs for the underlying bonds; - Specific local tax conditions applied to bond transactions. The objective of this study is to estimate the various costs related to the management of a portfolio of bonds comprising the underlying securities of the EuroMTS index in the context of a final investor. 3.2 Methodology In order to achieve the objective stated, we have analysed the monthly composition of the EuroMTS Global index and highlighted the impact of the rebalancing, as well as the impact of securities entering/leaving the index or moving from one maturity bucket to another. This analysis has been carried out on 10 months of historical data (May 2003-February 2004) to allow for a good level of representativity. In order to provide an estimate of the costs associated with managing a portfolio with equivalent holdings, we have conducted two surveys with French investors and asset managers with the objective of determining: - The average direct and indirect costs related to European government bond transactions - The average direct and indirect costs associated with the holding of government bonds and the settlement of relevant transactions. These surveys took place over a four-week period (12 January February 2004) during which 29 financial institutions received a questionnaire or were interviewed on questions related to costs. The sample of respondents covers final institutional investors (dealing rooms, pension funds and insurance companies) as well as fund managers operating in France. The sample was carefully constructed to focus on institutions with considerable investment activity in government bonds from the Eurozone in order to allow for a detailed analysis of their cost base. It is important to stress that the sample comprises large and small size institutions with varying volumes of trading in government bonds (both in number of trades and nominal amounts). We focused on analysing the costs related to managing the portfolio rather than the actual price charged to clients for these services, which is a commercial consideration that we have not included in our investigations. 3.3 Cost Benchmarking Methodology Financial institutions have been surveyed using a questionnaire followed by an interview aimed at clarifying the feedback and ensuring consistency across the various responses. The structure of the sample allows for a good representation of direct investors and asset managers of all size. Figure 3 shows the distribution of respondents by nature, while Figure 4 shows the distributions of assets under management of the institutions in the sample. Figure 3: Distribution of respondents by nature. 12

13 Figure 4: Distribution of assets under management of the institutions in the sample Assumptions When government bond investments did not constitute the only activity of the financial institution, our approach to assessing costs related to government bond activity has been one of the following: - When possible, the financial institution was asked to provide the overall costs for each activity and allocate the part related to government bonds only; - Alternatively, when allocation is not available, total costs have been allocated based on the proportion of government bonds in relation to overall assets under management. It should be noted that the treatment of taxes is dependent upon the legal nature of the investor (insurance company, pension fund, UCITS, other type of funds) and has therefore not been included in this study. Transaction costs have been split as follows in order to allow for a detailed analysis by our respondents: - Market spread (bid/ask) including brokerage fees (based on the assumption that a medium term maturity bond is a good proxy for the average transaction required for rebalancing the portfolio, average spread related to this medium term maturity bond is therefore reported); - Intermediation costs if orders are transmitted and executed through a separate desk invoicing for the service. Direct settlement costs have been included in the overall cost of the custodian (only one respondent was in a position to segregate transaction-related costs from pure custody costs). These transaction costs have been expressed as a percentage of the net proceeds of the transactions and reflect the cost associated with any change in the weight of a given security either for rebalancing or for modifying the overall amount invested (net assets of the portfolio). It should be noted that the average spread measured has been documented globally, without reference to the maturity. Even if differences in spreads can be noted when maturity increases, the balanced nature of the EuroMTS Index confirms that an equally weighted average spread is a fair representation of the spread investors will face on the market. On the other hand, fixed costs comprise a series of costs related to the management and the custody of the assets as well as the infrastructure required to trade: - Front office costs (management of the portfolio) in terms of human resources and systems (portfolio management systems, analytics, etc.); - Execution costs in terms of human resources and systems (price feeds, Bloomberg, etc.). Where orders are directly handled by the front office, no costs have been included in this envelope; - Middle and back office costs in terms of human resources and systems. It should be noted that these middle and back office costs only relate to the trading activity (settlement, management of 13

14 coupons, etc.) and do not cover the administrative and accounting costs related to the portfolio and dependent upon the nature of the investment vehicle; - Custody costs (related to assets under custody as well as transaction settlement costs). All fixed costs have been calculated in on a yearly basis and brought back to the assets under management as a percentage of these assets. Responses have been filtered to exclude elements that could not be considered as realistic (i.e. absence of response or non-relevance of a response) and are presented through their lowest, highest and average value. The weighted average has been excluded to avoid favouring large institutions that might have cost structures that are not representative of the universe Assets under Management The following table summarises the assets under management of our sample as well as the average portfolio structure. Table 4: Information on assets under management of institutions in the sample as well as the average portfolio structure. Our study comprises institutions managing large and small portfolios of investments covering the whole range of Euro government bonds included in the EuroMTS Global index Analysis of Transaction Costs Transaction costs have been analysed as a percentage of the net proceeds on an average basis. Table 5: Information on transaction costs in the sample. Transaction volumes described by our respondents are significant and can therefore form an appropriate basis for assessing the costs related to the rebalancing of our Euro-MTS Index portfolio Analysis of Fixed Costs Fixed costs have been assessed on a yearly basis and initially expressed in Euros. Table 6 summarises the details of our findings. Table 6: Information on fixed costs in the sample Absence of lowest figure corresponds to respondents for which the cost associated with the infrastructure or the resource is marginal.

15 Costs have also been assessed relative to the assets under management (Euro government bonds only) and are illustrated in Table 7. Table 8: Information on fixed costs in the sample expressed as a percentage of the assets under management. 3.4 EuroMTS Global Passive Replication Methodology In order to assess the level of transactions implied by passively tracking the index (i.e. by adjusting the portfolio on a monthly basis for changes in weights and incoming/outgoing securities), we have summed up the total of absolute relative changes on a monthly basis. It should be noted that when a bond expires during a given period (i.e. in between two rebalancings of the index), it is sold at the end of the previous period in order to avoid unbalances from the time the bond matures until the index is rebalanced. The EuroMTS Global Index included a list of 146 securities during the period we investigated Assumptions It is assumed that every transaction can be executed at a maximum cost of the spread available on the market for that bond from the bid price. Initial costs to buy the bonds and constitute the initial portfolio have not been included in these calculations, which also assume a constant level of assets under management (i.e. no subscriptions and redemptions ) Changes in Index Constituents The list of securities constituting the index might vary from one month to the other for one of the following reasons: - Bond maturing; - Maturity bond diminishing resulting in a move from one maturity bucket to another; - Bond liquidity conditions (the weight of individual bonds is dependent upon the relative market capitalisation as well as upon the size of each issuer s entire bond portfolio for a given maturity range). These changes result in monthly investments or disinvestments in specific securities and also result in changes in the weightings of other index constituents. Table 9 summarises the net amount of transactions induced by the monthly rebalancing of the index over a period of ten months. 15

16 Table 9: Information on the net amount of transactions induced by the monthly rebalancing of the index over a period of ten months. The passive replication of the entire index portfolio results in total net proceeds of % of the portfolio, excluding the cost of constituting the initial portfolio at a 100 basis. On a yearly basis, this level of rebalancing corresponds to % of the portfolio. 3.5 Global Cost Analysis The global cost of passively replicating a portfolio that has a nominal value of 100 with the EuroMTS Index over a period of ten months can be summarised as follows: Table 10: Information on the global cost of passive replication of the EuroMTS index over a period of ten months. 3.6 Findings The objective of this study was to analyse the various cost components related to the passive replication of the EuroMTS Global Index with a portfolio of bonds. We have assessed the costs related to trading and holding a portfolio of bonds that replicated the index over a 10-month period. The following table (Table 11) provides the final comparison between the costs of trading and holding the EuroMTS Global index tracker versus the portfolio of underlying bonds. 6 Table 11: Final comparison between the costs of trading and holding the EuroMTS Global index tracker versus the portfolio of underlying bonds This cost analysis does not take into acocunt the fact that annual management fees on ETFs can be partially offset by lending the ETF shares to investors who sell short the ETF. The security lending process is arguably developping more significantly on the ETFs market compared to the individual bond market.

17 The difference in management costs can be explained by the very high level of portfolio turnover related to changes in the EuroMTS index implying high transaction and settlement costs. The results show that, on average, it is more advantageous to use ETFs than to invest directly in all the underlying securities in order to implement bond investment management. More specifically, while the transaction costs are on average higher for an ETF than for direct management, this drawback is more than counterbalanced by the differential in management costs. Therefore, when it involves implementing the strategic allocation choices, which by nature generate few transactions, the ETFs are a very good solution in terms of both ease of use and costs. However, in the case of very dynamic active management strategies, the use of ETFs could turn out to be more costly, due to the number of transactions carried out. 4. Conclusion Asset allocation is the initial and key decision any investor should make. In the context of asset/ liability management, the proper asset allocation objective is not the management of a portfolio s tracking error with respect to market indices, but instead the management of deviations from risk factors affecting the current value of its liabilities. The index that best represents the client objectives and rules is indeed the most appropriate benchmark. While no existing commercial index is perfectly suited to representing a specific investor s liability structure, Treasury bond indices are typically the asset class most correlated with institutional investors liabilities, and as a result should be predominant in institutional investors allocation. Newly launched fixed income ETFs have specifically been designed to track bond market indices, which is the reason why it has often been argued that ETFs were natural investment vehicles to implement passive indexing strategies. In this paper, we have argued that the benefits of ETFs are actually much larger than traditionally reported, as these instruments can also be used to implement almost the full range of existing investment strategies. If there is little that portfolio managers typically do that cannot be done with bond ETFs, one legitimate question is what is the most convenient and cost-efficient way of implementing bond portfolio strategies. To answer that question, we surveyed financial institutions with the objective of estimating the various costs related to the management of a portfolio of individual bonds, and we found that ETFs ease of use allows investors to practice sound and sophisticated risk management, without necessarily having to deploy extensive human resources and large material infrastructures. We have also found that ETFs presented management costs that were lower on average than those relating to the direct management of individual bonds. 5. References Bekaert, G., and R. Hodrick, 1992, Characterizing predictable components in excess returns on equity and foreign exchange markets, Journal of Finance, 47, Campbell, J., 1987, Stock returns and the term structure, Journal of Financial Economics, 18, Campbell, J., and R. Shiller, 1988, Stock prices, earnings, and expected dividends, Journal of Finance, 43, Elton, E., M. Gruber, S. Das, and M. Hlavka, 1993, Efficiency with costly information: a reinterpretation of evidence from managed portfolios, Review of Financial Studies, 6, 1, Fama, E., French, K., Business conditions and expected returns on stocks and bonds. Journal of Financial Economics, 25,

18 Grinold, R., and R. Kahn, 2000, Active portfolio management: A quantitative approach to providing superior returns and controlling risk, Second Edition, McGraw Hill. Harvey, C., 1995, Predictable risk and returns in emerging markets, Review of Financial Studies, Kandel, S., and R. Stambaugh, 1996, On the predictability of stock returns: An asset allocation perspective, Journal of Finance, 51, Keim, D., and R. Stambaugh, 1986, Predicting returns in the stock and bond markets, Journal of Financial Economics, 17, Ferson, W., and C. Harvey, 1991, Sources of predictability in portfolio returns, Financial Analysts Journal, May/June, Ferson, W., and C. Harvey, 1993, The risk and predictability of international equity returns, Review of Financial Studies, 6, Ferson, W., and C. Harvey, 1995, Predictability and time-varying risk in world equity markets, Research in Finance, 13, Litterman, R., and J. Scheinkman, 1991, Common factors affecting bond returns, Journal of Fixed-Income, 1, 1, Martellini, L., Priaulet, P., and S. Priaulet, Fixed-income securities: Valuation, risk management and portfolio strategies, John Wiley and Sons, Standard Life Investments, 2003, Bridging the pensions gap, Global Bytes, uk.standardlifeinvestments.com/content/strategy/strategy_index.html. Watson Wyatt, 2003, Global asset study (ongoing); as cited by "Finanz und Wirtschaft" (28/01/2004), 18

19 EDHEC-Risk Institute is part of EDHEC Business School, one of Europe s leading business schools and a member of the select group of academic institutions worldwide to have earned the triple crown of international accreditations (AACSB, EQUIS, Association of MBAs). Established in 2001, EDHEC-Risk Institute has become the premier European centre for applied financial research. In partnership with large financial institutions, its team of 85 permanent professors, engineers and support staff implements six research programmes and ten research chairs focusing on asset allocation and risk management in the traditional and alternative investment universes. The results of the research programmes and chairs are disseminated through the three EDHEC-Risk Institute locations in London, Nice, and Singapore. EDHEC-Risk Institute validates the academic quality of its output through publications in leading scholarly journals, implements a multifaceted communications policy to inform investors and asset managers on state-ofthe-art concepts and techniques, and forms business partnerships to launch innovative products. Its executive education arm helps professionals to upgrade their skills with advanced risk and investment management seminars and degree courses, including the EDHEC-Risk Institute PhD in Finance. Copyright 2012 EDHEC-Risk Institute For more information, please contact: Carolyn Essid on or by to: carolyn.essid@edhec-risk.com EDHEC-Risk Institute promenade des Anglais BP Nice Cedex 3 - France EDHEC Risk Institute Europe 10 Fleet Place - Ludgate London EC4M 7RB - United Kingdom EDHEC Risk Institute Asia 1 George Street - #07-02 Singapore