Docstoc

An Analysis of Bond Fund Performance

Document Sample
An Analysis of Bond Fund Performance Powered By Docstoc
					An Analysis of Bond Fund Performance

For Umbono Fund Managers
June 2009

Brett Dugmore



     Executive Summary

            This report analyses the performance of actively managed funds in the South
            African Bond market over the last 5 years.

            The aim is to measure the degree of out-performance of the active funds and to
            estimate how much of this performance could have been achieved using passive
            strategies.

            We find that the average bond fund out-performed the ALBI by about 80bps per
            annum (before fees) over the last five years (after accounting for market exposure,
            or “beta”).

            We analysed the performance using the Cadiz PCA model and found that the
            best-performing funds, in general, were those with the highest yields. The top-
            performing fund was an exception, but we present a plausible hypothesis for why
            this fund differs from the rest.

            On average, the returns due to the timing of yield curve shifts were positive, but
            negligible in magnitude.

            At a conservative estimate, we estimate that around half the out-performance
            (before fees) of the active funds could have been achieved from a passive mix of
            the GOVI index and the BESA credit indices.

            Strategies that invest in non-government bonds in South Africa are hampered by
            the poor liquidity of these bonds. To investigate this issue further, we extracted
            trade-by-trade data from the spreadsheets published daily by BESA and
            concluded that liquidity is so poor that there is almost no scope for dynamically
            managed strategies involving credit.

            We conclude that active management decisions relating to non-government bonds
            in South Africa are probably restricted to decisions as to whether to participate in
            placements and auctions of these bonds, with the expectation of holding the
            instruments to maturity.

            Our transition management experience indicates that even small corporate bond
            trades can have significant market impact.

            We suggest that a passive strategy consisting of an investment of 75% in a GOVI
            tracker and 25% in a corporate bond tracking fund (tracking a customised credit
            index) would be likely to produce similar performance to the average actively
            managed bond fund.
This strategy would have the advantages of relatively stable and predictable risk
characteristics compared with actively managed alternatives, lending itself well to
inclusion in a broader asset allocation framework.

Appendix A shows that there is no significant difference between unit trust returns
and institutional fund returns within asset management companies once fees have
been accounted for.

Appendix B gives detail of the Cadiz PCA model.

Appendix C shows the Cadiz PCA model results for a 3-year and a 1-year time
frame.
Contents

Introduction

1. Data

2. Basic return Analysis

3. Detailed Return Analysis

     a) Cadiz PCA Analysis
     b) Analysis of Fund Performance Relative to a Benchmark

4. Analysis of the RSA Corporate Bond Market

5. Conclusion

Appendix A: Comparative analysis of institutional and unit trust returns

Appendix B: The Cadiz PCA Model

Appendix C: Return Decomposition Tables
Introduction

The debate between active and passive fund management has been ongoing for a long
time with many arguments put forward in favour of both investment styles. An interesting
and objective analysis can be found in the article by David Loeper1 (2003), where he
concludes that in the long term the difference between active and passive management is
likely to be small, with passive management providing more certainty over future returns.

Loeper points out that there is a benefit to investing in actively managed funds provided
the investor is highly certain that skilful managers exist and that he is highly certain that he
can identify these managers in advance. He shows (given certain assumptions) that the
degree of certainty required is fairly high to compensate for the additional uncertainty
brought about by active management. This point is also touched on by Sharpe2 (1991) in
his well-known article “The Arithmetic of Active Management”.

The purpose of this report is not to add to the passive versus active debate, but rather to
measure the extent of historical out-performance by active fund managers in the local
bond market, and to estimate how much of this out-performance could have been
achieved through passive strategies. Our results do not imply anything about the actual
strategies followed by the active managers; we only wish to ascertain if some of the
performance could have been achieved through passive alternatives.

We begin with a simple comparative return analysis over three time periods: the last 1, 3
and 5 years and then use the proprietary Cadiz PCA model to attribute bond fund
performance to factors such as credit spreads (yield), curve exposure and active yield
curve management.

We compare the performance of active bond funds with passive combinations of
government and corporate bonds in order to estimate the extent to which credit could
account for observed out-performance of the active managers.

We conclude with an investigation of the liquidity and pricing of non-government bonds in
the domestic market. Our aim is to determine the feasibility of actively managed corporate
bond strategies.

In an appendix we compare the performance of unit trusts and institutional funds to
determine if retail and institutional bond investments are managed differently within asset
management companies.

1. Data

Our analysis is based on monthly total returns from two sectors of the industry:

a) Institutional Bond Funds: 30 funds, 60 months of data.
b) Bond Unit Trusts: 18 funds, 170 months of data.

The time series consist of monthly total returns (including reinvested coupons). The
institutional fund data is gross of fees, while the unit trust data is net of fees.

In Appendix A we compare unit trust and institutional returns, taking account of fees, and
conclude that there is no significant difference between performance within asset
1
 “Active vs. Passive Management”, April 23, 2003, Wealthcare Capital management.
2
 “The Arithmetic of Active management”, The Financial Analysts Journal, Vol 47, No1.
January/February 1991
management companies. The body of the report focuses on the institutional data because
there are more funds in this data set.

2. Basic Return Analysis

We analysed the 30 institutional fund return time series over three periods: the last 1, 3
and 5 years. Of the 30 funds, 18 had a five year history, 20 had a three year history and
23 had a 1 year history.

We focus on the 5 year results in the main document. We draw essentially the same
conclusions using 3 year and 1 year histories, but we can be statistically more certain
using the maximum number of data points. Some of the tables relating to the shorter time
periods are shown in Appendix C.

Figure 1 below compares the current government zero curve with the curve at the start of
each of the three periods.

Figure 1: Government zero curves




Table 1 below shows the total returns of various asset classes over the three periods.




Table 1: Comparative returns
                  Jibar    GOVI   ALBI   Top 40
Jan 04 - Jan 09     55.42% 60.59% 60.27% 123.53%
Jan 06 - Jan 09     34.41% 25.87% 25.52% 22.12%
Jan 08 - Jan 09     13.50% 14.34% 14.19% -27.15%

The three time periods displayed markedly different characteristics. Over 5 years, equities
out-performed the other asset classes. Over 3 years, the money market was the best
performing asset class, while over 1 year government bonds were the best performers.
This allows for an interesting comparison of bond performance over significantly different
market conditions. Although we only present the results for the 5 year period, it is worth
noting that our conclusions would be essentially the same using any of the three time
periods.

In Table 2 below we rank funds by total return over these time periods. The light green
cells represent top-quartile performance while the red cells represent bottom quartile
performance.

Table 2: Funds ranked by total return
                      Fund                       5-year       3-year    1-year
            Tri-Linear Fixed Income                     1           1         1
      STANLIB Core Bond Fund (Pooled)                   2           7        11
         STANLIB Core Bond Portfolio                    3          18        18
        Futuregrowth Yield Enhanced                     4           2         3
        PRESCIENT BOND QUANTPLUS                        5           3         4
          African Harvest Core Bond                     6           8        15
 Rand Merchant Bank Domestic Specialist Bond            7           4         2
                Prudential Bond                         8          12        14
         Investec Dynamic Bond Fund                     9           5         6
                SYmmETRY Bond                          10          10        23
      Coronation Active Bond Composite                 11           6        13
           Pure Fixed Interest Local                   12          11        12
       Advantage Moderate Bond FOF                     13           9        17
              Metropolitan Bond                        14          14        20
       OMAM Domestic Specialist Bond                   15          17        16
           SIM Duration Bond Fund                      16          16         8
           PRESCIENT BOND QUANT                        17          15        10
                Real Africa Bond                       18          13         9
                      ALBI                             19          20        22

We notice that only two funds were consistently in the top quartile and another two funds
were consistently in the bottom quartile. The remaining fund positions fluctuated over
time.

A different picture emerges if we rank the funds by their Sharpe ratios.

Table 3: Funds ranked by Sharpe ratio
                     Fund                        5-year       3-year    1-year
       PRESCIENT BOND QUANTPLUS                           1         1         1
               SYmmETRY Bond                              2         2         3
           Tri-Linear Fixed Income                        3         3         2
       Advantage Moderate Bond FOF                        4         4         5
        Futuregrowth Yield Enhanced                       5         5         8
          Pure Fixed Interest Local                       6         6         7
         African Harvest Core Bond                        7        10        17
 Rand Merchant Bank Domestic Specialist Bond              8         7         4
       STANLIB Core Bond Fund (Pooled)                  9        13        18
       Coronation Active Bond Composite                10        11        14
            PRESCIENT BOND QUANT                       11         8         6
          STANLIB Core Bond Portfolio                  12        20        22
                Real Africa Bond                       13        12        11
          Investec Dynamic Bond Fund                   14        14        15
               Metropolitan Bond                       15        15        21
                Prudential Bond                        16        16        23
        OMAM Domestic Specialist Bond                  17        17        16
            SIM Duration Bond Fund                     18        18        13
                      ALBI                             19        21        24

It is evident from Table 3 that when measured on a risk-adjusted basis, fund rankings
were far more consistent over the three time periods. This implies that while the relative
returns of the fund managers were variable, there was consistency in the relative ability of
fund managers to extract performance from the level of risk taken.

This highlights one advantage of a passively managed bond portfolio: the market
exposure (for example as measured by modified duration) is likely to be more stable and
predictable for a passively managed portfolio compared with an actively managed
portfolio. This has implications from a broader asset allocation perspective.

From the two tables above it is clear that gross of fees, all funds out-performed the ALBI.
Our aim in the next two sections is to investigate how this out-performance was achieved,
and specifically to determine if similar performance could have been achieved using a
passive strategy.




3. Detailed Return Analysis

Our aim in this section is to determine how much of the observed active fund out-
performance could be achieved through passive strategies. To this end we present two
analyses of the fund return data:

   In Section a) we apply the Cadiz PCA Model (outlined in Appendix B) in order to
   determine the contribution of yield curve shifts and yield curve timing to fund returns.

   In Section b) we apply the classical CAPM model to analyse excess fund returns
   relative to a benchmark.


a) Cadiz PCA model

Cadiz has developed a proprietary model that estimates the contribution of a number of
factors to a bond fund’s total return. The model using principal component analysis (PCA)
to measure the most common types of yield curve shifts over a historical period, and then
uses this information to “match up” fund returns with yield curve volatility.
This method allows us to estimate an attribution of fund returns to various factors. These
factors are credit spreads, yield curve shifts and yield curve timing. A detailed explanation
of the model is presented in Appendix B.

A useful feature of the model is that it incorporates a “timing” factor, which can potentially
detect a fund manager’s attempts to “time” the market. For example, a manager who
moves the modified duration of his fund shorter when expecting a rise in yields, and
longer when he expects a fall in yields is said to be “timing” the level shifts in the yield
curve.

It should be noted that fund returns, including returns due to timing, can generally be
achieved in many different ways. The model only allows us to draw conclusions about the
characteristics of the return time series. We can’t use the model outputs to draw
conclusions about a particular fund manager’s strategies. For example, protective option
strategies (e.g. rolling put options) can produce return series that have the same
characteristics as a fund manager with skill at timing the market as defined above.

In this report we will not attempt to draw conclusions about the strategies followed by
active managers, only to demonstrate that passive strategies can often account for some
of the characteristics of active management.

The possible sources of return identified by the Cadiz PCA model are:

       Yield: This represents the interest that accrues to the bond portfolio every month.
       The average credit spread of a portfolio would contribute to this component of
       return.

       Curve shifts: This factor measures returns that resulted from the funds’ average
       yield curve positioning over the full period.

       Timing: Returns arise from this factor when the fund manager actively changes
       his yield curve exposure in response to changing expectations of future yield curve
       shifts.

       Residual: This factor measures performance that is not explained by any of the
       previous three factors.

Analysis over last 5 years

Table 4 below summarises the results of the analysis over the last 5 years.
  Table 4: Return decomposition (5 years)
       Fund name               Return        Yield      Curve shifts     Timing       Residual     Adj R-square
 'Tri-Linear Fixed Income'    0.94% (1)    0.66% (16)   0.18% (16)     0.11% (1)     -0.00% (8)    92.67% (19)
'STANLIB Core Bond Fund
                              0.93% (2)    0.77% (1)     0.21% (2)     -0.03% (18)   0.00% (5)     97.12% (15)
              (Pooled)'
      'STANLIB Core Bond
                              0.89% (3)    0.75% (2)     0.22% (1)     -0.05% (19)   -0.00% (18)    98.37% (3)
             Portfolio'
      'Futuregrowth Yield
                              0.89% (4)    0.71% (4)    0.19% (12)     0.00% (7)     0.00% (6)      98.28% (7)
             Enhanced'
        'PRESCIENT BOND
                              0.89% (5)    0.74% (3)    0.16% (19)     0.01% (6)     0.00% (4)     97.61% (13)
           QUANTPLUS'
     'African Harvest Core
                              0.88% (6)    0.69% (8)    0.19% (13)     0.01% (5)     -0.00% (17)   97.15% (14)
                Bond'
    'Rand Merchant Bank
                              0.87% (7)    0.68% (12)    0.19% (9)     0.02% (4)     -0.00% (15)    98.30% (6)
Domestic Specialist Bond'
         'Prudential Bond'    0.87% (8)    0.67% (15)    0.21% (5)     0.02% (3)     0.00% (1)      98.25% (9)
  'Investec Dynamic Bond
                              0.87% (9)    0.63% (19)    0.20% (8)     0.05% (2)     -0.00% (10)   97.92% (12)
                Fund'
       'SYmmETRY Bond '       0.86% (10)   0.71% (5)    0.16% (18)     0.00% (9)     -0.00% (14)   96.98% (16)
 'Coronation Active Bond
                              0.86% (11)   0.69% (9)    0.19% (14)     0.00% (8)     0.00% (3)     96.03% (17)
            Composite'
'Pure Fixed Interest Local'   0.86% (12)   0.71% (6)    0.19% (15)     -0.02% (16)   -0.00% (19)    98.28% (8)
    'Advantage Moderate
                              0.85% (13)   0.70% (7)    0.17% (17)     -0.00% (12)   -0.00% (13)   94.76% (18)
             Bond FOF'
      'Metropolitan Bond'     0.85% (14)   0.69% (10)    0.20% (7)     -0.01% (15)   -0.00% (12)   98.08% (10)
        'OMAM Domestic
                              0.85% (15)   0.68% (13)    0.20% (6)     -0.01% (13)   -0.00% (7)     98.31% (5)
          Specialist Bond'
'SIM Duration Bond Fund'      0.85% (16)   0.67% (14)    0.21% (4)     -0.00% (11)   -0.00% (11)   98.05% (11)
        'PRESCIENT BOND
                              0.85% (17)   0.68% (11)   0.19% (11)     -0.01% (14)   0.00% (2)      98.62% (2)
               QUANT'
        'Real Africa Bond'    0.83% (18)   0.66% (17)   0.19% (10)     -0.00% (10)   -0.00% (8)     98.33% (4)
          'ALBI'              0.81% (19)   0.64% (18)    0.21% (3)     -0.02% (17)   -0.00% (16)    98.87% (1)
          Mean                  0.87%        0.69%        0.19%          0.00%         0.00%         97.47%
 Rank correlations with
                              100.00%       49.82%       -12.28%        35.09%
        Return


       The “Return” column shows the average monthly return of each fund.
       The next four columns give an estimated decomposition of the total return into factors.
       The final column gives an indication of the goodness-of-fit of the model.

  The table above highlights some interesting points:

       Yield was by far the largest contributor to return. Differences in yield between funds
       can be the result of differences in modified duration (due to the shape of the yield
       curve), or differences in the average credit spread of the portfolio.

       The second most significant contributor to return was the average yield curve
       positioning (e.g. modified duration). For example a fund with an average modified
       duration that was high during a period when yields fell, on average, would have a large
       positive contribution from this factor.

       Timing on average produced negligible returns, although these were positively
       correlated with return.
    The factor that had the highest (rank) correlation3 with total return was the yield factor.
    This implies that the yield of a fund was the best predictor of a fund’s ranking on the
    return table.

The top-performing fund stands out (and confuses the results somewhat). This fund
performed best over all time frames, yet it had close to the lowest yield. Most of the fund’s
out-performance was apparently generated through market timing, where it ranked first.

This fund’s return profile is similar to a passive strategy that generated out-performance
through protective option strategies, passively implemented. This strategy reduces the
fund’s yield significantly (due to the cost of purchasing the options), but boosts the
apparent market timing contribution, since the fund out-performs when there are negative
market movements. The fact that this particular fund had the lowest R-square is further
evidence that its returns were not generated through simple yield curve positioning or
traditional market timing to the same extent as other funds in the table.

Previous Cadiz research4 has shown that it would have been possible to produce
significant out-performance through passive option strategies, e.g. rolling put options or
rolling collar strategies.

The return profile of this fund is consistent with a rolling protective option strategy,
possibly with some yield enhancement through credit, although our analysis is not able to
determine whether this was, in fact, the strategy followed by the fund manager.

Our conclusion is that, even though there are some anomalies, in general high yield was a
good indicator of overall performance – a strong indication that fund managers relied on
high yielding corporate bonds to enhance returns. This prompts the question of whether it
was necessary (or even possible) to actively manage the credit exposure, or if this return
enhancement could have been achieved through a passive holding of non-government
bonds. We address this question in the next section, using a classical CAPM model.

b) Analysis of fund performance relative to a Benchmark

The Cadiz PCA model implemented in the previous section provided strong evidence of
the importance of portfolio yield in generating active returns.

The question we will address in this section is whether this performance could be
achieved from a passive position in credit. To use the terminology of two articles by Jane
Li5, we are interested in determining if credit bonds form an “exotic beta” component of
returns. For example, if a government bond index and a corporate bond index form two
“markets” within the bond universe, then we can measure beta relative to both indices,
with government bonds forming the “traditional” beta and corporate bonds forming “exotic”
beta.

 In this section we measure beta relative to various indices by fitting a classical CAPM
regression model of the form6:

3
  This is the correlation coefficient of the funds’ ranks based on total return and the funds’ ranks
based on returns arising from the yield component.
4
  “3-month R153 Option Strategies”, Brett Dugmore, Cadiz Research, 2002
5
  “Practical Applications of Active and Passive Investment Management: Examining Real Alpha
and Exotic Beta”, Jane Li, Fundquest, 2007 and “Practical Applications of Exotic Beta”, Jane Li,
Fundquest, 2008.
6
  This model originated in work of H. Markowitz. See, for example, “Portfolio Selection”, H
Markowitz, The Journal of Finance, 7 (1): 77–91.
where the dependent variable is the excess return of a bond fund (over the risk-free rate)
and the independent variable is the excess return of an index such as the ALBI.

Table 5 below presents the results of this analysis for 18 funds over the last five years,
relative to the All-Bond Index.

Table 5: Performance relative to the ALBI, last 5 years
                                                            ALBI
                   Fund Name                        Alpha          Beta
       Advantage Moderate Bond FOF                  0.06%          0.83
          African Harvest Core Bond                 0.08%          0.93
      Coronation Active Bond Composite              0.06%          0.91
        Futuregrowth Yield Enhanced                 0.09%          0.92
         Investec Dynamic Bond Fund                 0.06%          0.98
              Metropolitan Bond                     0.05%          0.96
       OMAM Domestic Specialist Bond                0.05%          0.97
           PRESCIENT BOND QUANT                     0.04%          0.91
        PRESCIENT BOND QUANTPLUS                    0.10%          0.74
                Prudential Bond                     0.07%          0.99
           Pure Fixed Interest Local                0.06%          0.88
 Rand Merchant Bank Domestic Specialist Bond        0.07%          0.93
                Real Africa Bond                    0.03%          0.93
           SIM Duration Bond Fund                   0.04%          0.99
      STANLIB Core Bond Fund (Pooled)               0.13%          1.01
         STANLIB Core Bond Portfolio                0.09%          1.01
                SYmmETRY Bond                       0.07%          0.78
            Tri-Linear Fixed Income                 0.14%          0.92
                     Mean                           0.07%          0.92

The column headed “Alpha” shows the average excess monthly return over the ALBI once
the fund’s market exposure was accounted for. The column headed “Beta” shows the
average market exposure of each fund relative to the ALBI. A value of 1 represents the
same level of exposure as the ALBI.

The average monthly alpha of the 18 funds was 0.07%. This means that on average an
active fund out-performed the ALBI by 7 bps per month after taking market exposure into
account. The average fund beta was 0.92 which means that the average active fund had
less exposure to the bond market than the ALBI.

In the context of this model credit spreads (arising from positions in corporate bonds) will
translate into alpha. The addition of a credit index as a source of “exotic beta” will allow us
to estimate the addition return due to changes in credit spreads but bonds with constant
credit spreads will not be detected by the model, and the additional return due to the credit
spread will still form part of the alpha.
In Table 6 below we include two credit indices7 (published by BESA in conjunction with
Standard Bank) as two additional sources of “exotic beta”. Note that the credit indices
have only 39 months of data and so the analysis in this section is based on this shorter
time period.


Table 6: Performance relative to government bonds and credit


                   Fund Name                          Alpha    GOVI     Credit Fix     Credit Floating
        Advantage Moderate Bond FOF                  -0.06%    0.90       -0.14             0.37
          African Harvest Core Bond                   0.04%    1.04       -0.09             0.08
      Coronation Active Bond Composite                0.04%    0.92        0.05             0.14
         Futuregrowth Yield Enhanced                  0.11%    1.05       -0.09            -0.02
         Investec Dynamic Bond Fund                   0.09%    1.11       -0.08            -0.02
              Metropolitan Bond                       0.02%    0.84        0.16             0.11
       OMAM Domestic Specialist Bond                  0.00%    1.01       -0.05             0.13
           PRESCIENT BOND QUANT                       0.01%    0.82        0.09             0.07
        PRESCIENT BOND QUANTPLUS                      0.06%    0.78       -0.03            -0.06
                Prudential Bond                       0.05%    1.02        0.00             0.01
           Pure Fixed Interest Local                  0.02%    0.87        0.03             0.08
    Rand Merchant Bank Domestic Specialist
                      Bond                           0.06%      0.88        0.08            0.07
                Real Africa Bond                     0.02%      0.97       -0.02            0.08
           SIM Duration Bond Fund                    0.01%      0.98        0.02            0.19
       STANLIB Core Bond Fund (Pooled)               0.04%      0.68        0.35            0.22
         STANLIB Core Bond Portfolio                 0.03%      0.87        0.18            0.06
                SYmmETRY Bond                        0.02%      0.76        0.00            0.00
            Tri-Linear Fixed Income                  0.10%      1.06       -0.19            0.27
                      Mean                           0.04%      0.92       0.02             0.10

Table 6 shows that the addition of credit as an additional source of beta reduced the
average fund alpha to 4bps per month. This implies that changing credit spreads
enhanced the active fund returns by an average of 3bps.

As mentioned earlier, this model will interpret constant credit spreads as an alpha
contribution, even if they actually arise from the credit index component. South African
non-government bonds are highly illiquid and this has the effect of keeping credit spreads
stable in the absence of trade. This lack of liquidity makes a rigorous analysis difficult, but
it is probable that the contribution of passive credit is actually higher than the 3bps
estimated above.

An additional consideration is the fact that the credit indices used above are heavily
weighted towards large bond issues by state-owned enterprises such as Eskom. These
bonds have had tighter credit spreads than some of the riskier corporate debt. It would be
possible for a fund manager to enhance his alpha by holding predominantly higher-


7
  BESA, in conjunction with Standard Bank, publishes a “composite credit” index with two sub-
indices, the “fixed credit” and “floating credit” indices. These indices are constructed on the same
basis as the ALBI, using only corporate bonds. We use the two sub-indices as two additional
“exotic beta” variables.
yielding corporate bonds instead of the market-cap weighted proportions in the indices
above.

We conclude that approximately half the out-performance exhibited by active bond funds
can be achieved through a passive holding of non-government bonds. Further attribution
analysis is hampered by the lack of liquidity in the local corporate bond market, but it
seems likely that, in reality, the contribution due to passive credit is higher than the model
is able to detect.


4. Analysis of the RSA corporate bond market

The previous section has highlighted the important role that non-government bonds play
in the performance of domestic bond funds. Around half the observed out-performance
could have been achieved with a passive position in listed corporate bond indices.
Unfortunately these bonds, in general, are highly illiquid and any analysis is hampered by
the fact that the major data vendors do not provide important statistics, such as daily
volumes traded, for most of the listed non-government bonds.

We are interested in the feasibility of dynamic investment strategies involving RSA
corporate bonds. To do this we require accurate trade-by-trade data for non-government
bonds, so we have reconstructed time series of volumes traded for all listed non-
government bonds from January 2006 to the present using raw data provided by the Bond
Exchange.

We extracted daily trade data for the 561 non-government bonds that were in existence
any time within the last 783 trading days. To obtain an initial indication of liquidity we
constructed a histogram of the maximum number of days between trades for non-
government bonds. This is shown below in Figure 2.

Figure 2: Maximum number of trading days between trades
The histogram highlights the illiquidity of non-government bonds. For example, almost
10% of the bonds experienced periods of inactivity lasting between 8 and 38 days. Over
33% of the bonds experienced periods of inactivity lasting longer than 128 working days
(approximately 6 months).

We found that average monthly turnover for a non-government bond was about 7% of the
issued nominal. This is compared with about 500% for a liquid government bond. The
histogram in Figure 3 below summarizes the turnover statistics.

Figure 3: Percentage Turnover




From this chart we can see that over 45% of stocks had a monthly turnover of less than
2%.

These numbers highlight the problem of stale prices in the corporate bond market. If a
bond hasn’t traded for six months, then the credit spread reported by BESA is six months
old. Nonetheless, this credit spread is used by fund managers to value their portfolios.
This results in an understatement of the true riskiness of bond portfolios that include these
bonds.

Based on the turnover statistics, buy-and-hold appears to be the only feasible strategy
and it is apparent that no large dynamic credit strategies are being implemented. The
active decisions relating to credit are probably restricted to decisions about whether to
buy-and-hold (or liquidate existing positions) based on perceived credit risk.

An additional concern highlighted by our analysis is the mark-to-market process followed
by BESA. We have identified a number of instances where illiquid instruments have
traded, but the traded levels were not reflected in the mark-to-market statistics. In
particular, the FRN traded spreads appear to be frequently wrong.
In addition, there are instances where the credit spread on a corporate bond changes
dramatically due to new trade, but the spreads on other bonds issued by the same entity
have not been updated because they haven’t traded.

It is likely that credit spread volatility is generally understated by the BESA mark-to-market
process and thus the risk-adjusted performance of the credit indices is probably
overstated.

In conclusion, the South African corporate bond market is not liquid enough for the
implementation of actively managed credit positions. The data suggests that most
participants in this sector of the market implement trades on a buy-and-hold basis. This
situation lends itself to a passively managed credit component within a bond tracking fund.
The main issue here is defining a suitable credit index that is investible, given the lack of
liquidity. The solution would probably involve participation by the tracking fund in bond
auctions and placements as they arise, in a way that is consistent with the subsequent
inclusion of these issues in the benchmark that is being tracked.

5. Proposal for a Passively Managed Strategy

Our results have shown that there is evidence that actively managed funds have out-
performed the ALBI, although there is some lack of consistency in the ranking of the
funds’ absolute returns.

We propose a passive strategy that is likely to produce returns similar to the average
active fund, but with more control over the risk characteristics of the fund. In particular, the
modified duration of the fund is likely to be more stable and predictable than an actively
managed bond fund.

In particular, we suggest the following process:

       Establish a strategic level of exposure to credit. For example the average active
       fund was 26% invested in instruments with a credit rating lower than AAA, so a
       mandate that allows 75% to be invested in government bonds and 25% in non-
       government bonds may be reasonable.

       Use the 75% component to track the GOVI index. This is relatively straightforward
       since there are currently 10 government bonds in the index, all of which have good
       liquidity.

       Use the remaining 25% to track a corporate bond index. The lack of liquidity in the
       corporate bond market would prevent immediate investment in one of the existing
       credit indices, but investment could be phased into a custom credit index.

       We suggest constructing a custom index that only includes bonds with credit
       spreads that are sufficiently wide, and perhaps implements a less concentrated
       weighting scheme (e.g. equally weighted) than that followed by the BESA credit
       indices. This would ensure that the index was not dominated by the bonds issued
       by state-owned enterprises such as ESKOM. The index construction would need
       to be specified in advance, and could possibly be expanded to incorporate a
       degree of non-listed credit.
Other Cadiz research has shown that this strategy could be further enhanced by including
passively managed protective option overlays. In particular, an earlier Cadiz report8 has
shown that the implementation of rolling protection (either bought put options or zero-
premium collars) in the local bond market would have significantly improved risk-adjusted
returns.


6. Conclusion

We have analysed the returns of local institutional bond funds in an attempt to measure
the extent to which these funds have out-performed the All-Bond index and to ascertain
how much of this out-performance could have been achieved through passively managed
strategies.

We found that on average over the last 5 years, institutional bond funds out performed the
ALBI by about 80bps per annum before fees.

An analysis using the proprietary Cadiz PCA model indicated that portfolio yield (via credit
spreads on non-government bonds) was probably a large contributor to this out-
performance.

We applied classical CAPM techniques to analyse the possible contribution of a passive
credit strategy and found that, at a conservative estimate, about 36bps of the total out-
performance could have been achieved by passive holdings of listed credit (subject to
liquidity considerations).

The Cadiz model showed little evidence that the average fund generated out-performance
through active yield curve strategies. In particular we investigated the liquidity of corporate
bonds in South Africa and conclude that the lack of liquidity of these bonds makes it
unlikely that many funds implement active strategies involving corporate bonds.

Possible additional sources of out-performance that our models cannot detect are unlisted
credit, non-vanilla fixed income instruments, or option strategies (most likely passive).

We outline a possible passive strategy that we believe is likely to produce after-fee returns
that are close to the average active bond fund.




8
    “3-month R153 Option Strategies”, Brett Dugmore, Cadiz Research, 2002
Appendix A: Comparative analysis of institutional and unit trust returns

In this appendix we compare the performance of the unit trust bond funds with the
institutional bond funds. For each fund we find the average out-performance over the
period of the fund’s existence, up to a maximum of 60 months previously.

   Unit Trusts

   Historically, bond unit trusts have under-performed the ALBI on average.
   Of the 18 funds analysed, only three consistently beat the ALBI after fees.
   The average underperformance was 25bps p.a.
   Before fees, all funds out-performed. Average out-performance was 75bps p.a.

   Institutional

   The institutional data consists of returns before fees.
   The average out-performance was 62bps p.a.
   This is based on a different number of funds and different time period to UT analysis.

Unit Trust versus institutional Comparison

We matched 15 Unit Trust funds with institutional counterparts. Table 7 below shows the
paired funds.

Table 7: Matched Institutional and Unit Trust bond funds
 Institutional                                     Unit Trust
 ABSA Bond Fund                                    ABSA Bond A
 Coronation Active Bond Composite                  Coronation Bond R
 Coronation Core Bond                              Coronation Bond R
 Coronation Domestic Specialist Bond               Coronation Bond R
 Futuregrowth Yield Enhanced                       Futuregrowth Bond R
 Investec Bond Fund                                Investec Gilt R
 Investec Dynamic Bond Fund                        Investec Gilt R
 Investec Triple Alpha                             Investec Gilt R
 Metropolitan Bond                                 Metropolitan Gilt
 OMAM Domestic Specialist Bond                     Old Mutual Gilt
 Prudential Bond                                   Prudential High Yield Bond A
 Rand Merchant Bank Domestic Specialist Bond       RMB Bond
 STANLIB Core Bond Fund (Pooled)                   STANLIB Bond A
 STANLIB Core Bond Portfolio                       STANLIB Bond A
 SIM Duration Bond Fund                            SIM Bond Plus R

Unit trust returns are generally published net of fees, while our institutional data is gross of
fees. Using a fee schedule, we added back the annual management fee to the unit trust
returns, and then compared the returns with the corresponding institutional fund.

We found that on average the Unit Trust funds out-performed the corresponding
Institutional funds by 5bps p.a. However, there was no statistically significant difference
between Unit Trust and Institutional performance before fees.
  Appendix B: The Cadiz Model

  Please contact Cadiz Securities if you are interested in the detail of the Bond PCA Model.
  Detail of this model can also be found in: “Bond Fund Analysis Using Principal
  Components”, Brett Dugmore, Cadiz Securities, 2005.

  Appendix C: Return decomposition tables

  Table 8: Return decomposition (5 years)
       Fund name               Return        Yield      Curve shifts     Timing       Residual     Adj R-square
 'Tri-Linear Fixed Income'    0.94% (1)    0.66% (16)   0.18% (16)     0.11% (1)     -0.00% (8)    92.67% (19)
'STANLIB Core Bond Fund
                              0.93% (2)    0.77% (1)     0.21% (2)     -0.03% (18)   0.00% (5)     97.12% (15)
              (Pooled)'
      'STANLIB Core Bond
                              0.89% (3)    0.75% (2)     0.22% (1)     -0.05% (19)   -0.00% (18)    98.37% (3)
             Portfolio'
      'Futuregrowth Yield
                              0.89% (4)    0.71% (4)    0.19% (12)     0.00% (7)     0.00% (6)      98.28% (7)
             Enhanced'
        'PRESCIENT BOND
                              0.89% (5)    0.74% (3)    0.16% (19)     0.01% (6)     0.00% (4)     97.61% (13)
           QUANTPLUS'
     'African Harvest Core
                              0.88% (6)    0.69% (8)    0.19% (13)     0.01% (5)     -0.00% (17)   97.15% (14)
                Bond'
    'Rand Merchant Bank
                              0.87% (7)    0.68% (12)    0.19% (9)     0.02% (4)     -0.00% (15)    98.30% (6)
Domestic Specialist Bond'
         'Prudential Bond'    0.87% (8)    0.67% (15)    0.21% (5)     0.02% (3)     0.00% (1)      98.25% (9)
  'Investec Dynamic Bond
                              0.87% (9)    0.63% (19)    0.20% (8)     0.05% (2)     -0.00% (10)   97.92% (12)
                Fund'
       'SYmmETRY Bond '       0.86% (10)   0.71% (5)    0.16% (18)     0.00% (9)     -0.00% (14)   96.98% (16)
 'Coronation Active Bond
                              0.86% (11)   0.69% (9)    0.19% (14)     0.00% (8)     0.00% (3)     96.03% (17)
            Composite'
'Pure Fixed Interest Local'   0.86% (12)   0.71% (6)    0.19% (15)     -0.02% (16)   -0.00% (19)    98.28% (8)
    'Advantage Moderate
                              0.85% (13)   0.70% (7)    0.17% (17)     -0.00% (12)   -0.00% (13)   94.76% (18)
             Bond FOF'
      'Metropolitan Bond'     0.85% (14)   0.69% (10)    0.20% (7)     -0.01% (15)   -0.00% (12)   98.08% (10)
        'OMAM Domestic
                              0.85% (15)   0.68% (13)    0.20% (6)     -0.01% (13)   -0.00% (7)     98.31% (5)
          Specialist Bond'
'SIM Duration Bond Fund'      0.85% (16)   0.67% (14)    0.21% (4)     -0.00% (11)   -0.00% (11)   98.05% (11)
        'PRESCIENT BOND
                              0.85% (17)   0.68% (11)   0.19% (11)     -0.01% (14)   0.00% (2)      98.62% (2)
               QUANT'
        'Real Africa Bond'    0.83% (18)   0.66% (17)   0.19% (10)     -0.00% (10)   -0.00% (8)     98.33% (4)
          'ALBI'              0.81% (19)   0.64% (18)    0.21% (3)     -0.02% (17)   -0.00% (16)    98.87% (1)
          Mean                  0.87%        0.69%        0.19%          0.00%         0.00%         97.47%
 Rank correlations with
                              100.00%       49.82%       -12.28%        35.09%
        Return
  Table 2: Return decomposition (3 years)
       Fund name               Return        Yield      Curve shifts     Timing       Residual     Adj R-square

 'Tri-Linear Fixed Income'    0.81% (1)    0.80% (18)   -0.14% (4)     0.18% (1)     0.00% (14)    92.39% (21)
      'Futuregrowth Yield
           Enhanced'          0.73% (2)    0.88% (3)    -0.14% (9)     0.02% (11)    0.00% (2)      98.71% (5)
       'PRESCIENT BOND
          QUANTPLUS'          0.72% (3)    0.80% (13)   -0.11% (1)     0.04% (6)     0.00% (13)    98.30% (15)
    'Rand Merchant Bank
Domestic Specialist Bond'     0.70% (4)    0.82% (10)   -0.14% (8)     0.05% (4)     0.00% (1)      98.54% (8)
  'Investec Dynamic Bond
              Fund'           0.70% (5)    0.80% (16)   -0.16% (15)    0.09% (2)     0.00% (16)    98.45% (12)
 'Coronation Active Bond
           Composite'         0.69% (6)    0.89% (2)    -0.16% (16)    -0.01% (18)   0.00% (4)     98.19% (16)
'STANLIB Core Bond Fund
            (Pooled)'         0.69% (7)    0.89% (1)    -0.16% (17)    -0.01% (19)   0.00% (6)     97.61% (19)
     'African Harvest Core
              Bond'           0.69% (8)    0.84% (7)    -0.15% (11)    0.02% (8)     0.00% (3)     98.51% (10)
    'Advantage Moderate
           Bond FOF'          0.68% (9)    0.82% (11)   -0.14% (7)     0.02% (7)     0.00% (9)     94.67% (20)
   'SYmmETRY Bond '           0.68% (10)   0.77% (20)   -0.11% (2)     0.04% (5)     0.00% (5)     98.34% (14)
'Pure Fixed Interest Local'   0.68% (11)   0.84% (6)    -0.14% (6)     -0.01% (15)   0.00% (10)     98.55% (7)
    'Prudential Bond'         0.67% (12)   0.80% (17)   -0.15% (13)    0.06% (3)     0.00% (15)    98.51% (11)
    'Real Africa Bond'        0.67% (13)   0.82% (9)    -0.15% (10)    0.01% (13)    0.00% (7)      98.82% (3)
  'Metropolitan Bond'         0.67% (14)   0.86% (5)    -0.16% (20)    -0.01% (17)   0.00% (11)     98.66% (6)
   'PRESCIENT BOND
        QUANT'                0.66% (15)   0.80% (15)   -0.14% (5)     0.02% (10)    0.00% (16)     98.88% (2)
'SIM Duration Bond Fund'      0.66% (16)   0.84% (8)    -0.17% (21)    0.02% (12)    -0.00% (21)   98.12% (17)
    'OMAM Domestic
     Specialist Bond'         0.64% (17)   0.81% (12)   -0.15% (14)    0.01% (14)    0.00% (7)     98.41% (13)
   'STANLIB Core Bond
        Portfolio'            0.64% (18)   0.86% (4)    -0.16% (19)    -0.04% (21)   0.00% (18)     98.75% (4)
   'Argon Core Bond'          0.63% (19)   0.79% (19)   -0.15% (12)    0.02% (9)     0.00% (11)     98.54% (9)
          'ALBI'              0.61% (20)   0.80% (14)   -0.16% (18)    -0.01% (16)   0.00% (18)     99.03% (1)
     'JMB BondPlus'           0.57% (21)   0.73% (21)   -0.12% (3)     -0.02% (20)   -0.00% (20)   98.09% (18)
          Mean                  0.68%
Correlations with Return      100.00%       46.23%        43.96%        44.56%
  Table 3: Return decomposition (1 year)
       Fund name               Return        Yield      Curve shifts     Timing       Residual     Adj R-square

'Tri-Linear Fixed Income'     1.62% (1)    0.53% (8)    0.69% (21)     0.46% (1)     -0.00% (15)   93.40% (24)
  'Rand Merchant Bank
Domestic Specialist Bond'     1.31% (2)    0.47% (10)    0.80% (9)     0.10% (7)     -0.00% (16)   98.97% (15)
   'Futuregrowth Yield
        Enhanced'             1.29% (3)    0.53% (7)    0.78% (15)     0.03% (12)    -0.00% (23)    99.65% (2)
    'PRESCIENT BOND
       QUANTPLUS'             1.28% (4)    0.55% (4)    0.61% (24)     0.15% (4)     -0.00% (22)   98.95% (16)
 'Investec Triple Alpha'      1.26% (5)    0.37% (22)    0.83% (5)     0.13% (6)     -0.00% (24)    99.53% (6)
'Investec Dynamic Bond
          Fund'               1.26% (6)    0.28% (24)    0.81% (7)     0.23% (2)     -0.00% (10)   99.32% (11)
'CAHAM Dynamic Bonds'         1.25% (7)    0.41% (19)   0.73% (20)     0.16% (3)     -0.00% (2)    99.32% (10)
'SIM Duration Bond Fund'      1.24% (8)    0.54% (6)     0.82% (6)     -0.06% (23)   -0.00% (3)    98.08% (21)
    'Real Africa Bond'        1.23% (9)    0.50% (9)    0.79% (12)     -0.01% (18)   -0.00% (7)     99.38% (9)
    'PRESCIENT BOND
         QUANT'               1.23% (10)   0.46% (11)   0.76% (18)     0.05% (9)     -0.00% (12)   99.28% (12)
'STANLIB Core Bond Fund
        (Pooled)'             1.23% (11)   0.54% (5)     0.85% (1)     -0.11% (24)   -0.00% (9)    98.03% (22)
'Pure Fixed Interest Local'   1.22% (12)   0.57% (3)    0.74% (19)     -0.04% (21)   -0.00% (17)   99.21% (13)
 'Coronation Active Bond
        Composite'            1.22% (13)   0.60% (1)    0.80% (10)     -0.13% (25)   0.00% (1)     98.86% (19)
     'Prudential Bond'        1.21% (14)   0.30% (23)    0.83% (4)     0.14% (5)     -0.00% (7)     99.60% (3)
  'African Harvest Core
           Bond'              1.20% (15)   0.46% (14)   0.78% (14)     0.02% (14)    -0.00% (19)    99.43% (8)
    'OMAM Domestic
      Specialist Bond'        1.20% (16)   0.45% (15)   0.80% (11)     0.01% (16)    -0.00% (6)    98.88% (17)
 'Advantage Moderate
         Bond FOF'            1.19% (17)   0.59% (2)    0.60% (25)     0.03% (11)    -0.00% (14)   93.52% (23)
   'STANLIB Core Bond
         Portfolio'           1.18% (18)   0.42% (16)    0.85% (2)     -0.03% (19)   -0.00% (18)   99.19% (14)
   'Argon Core Bond'          1.17% (19)   0.42% (17)   0.79% (13)     0.02% (15)    -0.00% (20)    99.45% (7)
   'Metropolitan Bond'        1.17% (20)   0.46% (13)    0.80% (8)     -0.04% (20)   -0.00% (4)    98.88% (18)
'Trident Capital-PQ Active
          Bond'               1.16% (21)   0.41% (21)   0.78% (16)     0.03% (13)    -0.00% (4)    98.83% (20)
          'ALBI'              1.16% (22)   0.42% (18)    0.85% (3)     -0.05% (22)   -0.00% (21)    99.55% (5)
   'SYmmETRY Bond '           1.10% (23)   0.46% (12)   0.63% (22)     0.04% (10)    -0.00% (13)    99.85% (1)
     'JMB BondPlus'           1.00% (24)   0.41% (20)   0.62% (23)     0.00% (17)    -0.00% (11)    99.55% (4)
          Mean                1.16% (22)
Correlations with Return      100.00%       18.87%       -15.93%        51.11%
DISCLAIMER
REGULATORY INFORMATION
CADIZ SECURITIES (PTY) LTD
Physical Address:
Ground Floor, Fernwood House
The Oval, 1 Oakdale Road
Claremont
7700
Telephone number: 021 657 8300 / 011 853 8000
Internet website: www.cadiz.co.za
FSP number: 13062


Cadiz Securities (Pty) Ltd (“Cadiz”) is an authorised Financial Services Provider.
Cadiz does not warrant the accuracy, relevance, completeness or fitness of the information contained
herein for use by the recipient or any other party for whatsoever purpose. The opinions and
recommendations contained herein are and must be construed solely as statements of opinion and
not statements of fact. This document is further provided purely for information purposes and must
not be seen to constitute financial advice or an offer to buy or sell any securities stated herein.
Neither Cadiz nor its staff derives any direct or indirect financial benefit for the views expressed
herein.
All returns are Rand returns, unless specifically stated otherwise.
Market fluctuations and changes in rates of exchange or taxation may have an effect on the value,
price or income of investments and an investor may not get back the full amount invested as a result
of such market fluctuations. Past performance is not necessarily a guide to future investment
performance. Performance is further affected by uncertainties such as changes in government policy
and other legal and regulatory developments.
All investment transactions with clients and their intermediaries are done on an arms length basis.
No part of this document may be reproduced or disseminated without the prior permission of Cadiz.

				
DOCUMENT INFO