Document Sample
finale Powered By Docstoc
					    Does Active Management Pay in Italy? A Study of Mutual Fund

                       Performance in the Period 1989-1999

                                    Giovanni Radicella

                                First Version: November 2000
                                  This Version: March 2001


Mutual funds (in the form of open-end unit trusts) were introduced in the Italian financial system in
1984. In this paper we assessed mutual fund performance in Italy in the decade April 1989-March
1999. We used absolute performance data, portfolio compositions and net inflows data to measure
the security selection and market timing skills with different performance measure models such as a
linear market model, a multi-factor model, a quadratic model and a conditional model.
Overall, the after-cost risk-adjusted extra-performance was not significantly different from zero
implying that the information and analysis superiority went to cover mostly costs (mainly
management fees and taxes). However, this resulted from two different contributions: security
selection and style rotation were able to add value whereas market timing activity had a negative
impact on financial performances. Besides, we showed that investing into mutual funds in Italy was
a good way to diversify into financial markets according to the modern financial theories.
Does Active Management Pay in Italy? A Study of Mutual Fund
Performance in the Period 1989-1999.

1. Introduction
           Italian mutual funds (in the form of open-end unit trusts) made their first appearance in the

arena of Italian financial markets in the second half of 1984. Their absence in Italy before that date

still marked a difference with more mature and developed financial markets, where mutual funds

had been existing for a long time.1

           By now the Italian market is filled with a wide range of funds belonging to different

categories: domestic equity funds, balanced funds, Euro equity funds, emerging markets funds,

Euro bond funds, global bond funds, money market funds, etc. At the end of 1999 the Italian mutual

fund industry consisted of 823 funds: 344 bond funds, 61 balanced funds and 356 equity funds.

(Table 1).

Table 1
Number of funds (end period data). Source: Assogestioni
Year            Bond Funds                 Balanced Funds              Equity Funds                       Total
1984                     4                         3                           3                          10
1985                     19                        17                          5                          41
1986                     26                        29                          5                          60
1987                     31                        33                          8                          72
1988                     48                        47                          20                         115
1989                     60                        54                          35                         149
1990                     74                        55                          54                         183
1991                     78                        60                          83                         221
1992                     96                        61                          98                         255
1993                     116                       62                          114                        292
1994                     147                       59                          148                        354
1995                     198                       58                          201                        457
1996                     239                       57                          235                        531
1997                     296                       53                          277                        626
1998                     325                       57                          321                        703
1999                     344                       61                          356                        823*

*The total number also includes the new categories introduced in 1999: liquidity funds (33) and flexible funds (29).

           At that time the amount of assets under management was ITL 920,306 bn, approximately

Euro 475.3 bn. Equity funds accounted for 29.5%, balanced funds accounted for 10.75% and bond

    In US for example the first mutual funds dated back to 1924.
funds for 54.1% (Table 2). The rate of growth was really impressive especially in the last three

years when assets under management increased more than five times.2 By 1998, the Italian mutual

fund industry became one of the largest in the world (Table 3): in particular in Europe it ranked

after Spain and France, but ahead of Germany and UK.

Table 2
Total Net Assets of Mutual Funds (end period data- Billion ITL and % of total). Source: Assogestioni
Year    Bond Funds                Balanced Funds             Equity Funds               Total % growth
1984    698      (60.28%)         224      (19.34%)          236     (20.38%)           1,158
1985    6,864 (34.65%)            7,324 (37.02%)             5,585 (28.23%)             19,773 1607.5%
1986    18,000 (27.66%)           28,216 (43.36%)            18,863 (28.98%)            65,079 229.1%
1987    22,701 (38.19%)           21,264 (35.77%)            15,484 (26.05%)            59,449     -8.7%
1988    18,585 (36.04%)           17,790 (34.50%)            15,188 (29.46%)            51,563 -13.3%
1989    16,161 (32.87%)           16,391 (33.34%)            16,616 (33.79%)            49,168     -4.6%
1990    19,757 (41.70%)           13,157 (27.77%)            14,468 (30.53%)            47,382     -3.6%
1991    31,042 (55.23%)           11,659 (20.75%)            13,500 (24.02%)            56,201    18.6%
1992    37,742 (62.22%)           10,082 (16.63%)            12,830 (21.15%)            60,654      7.9%
1993    72,000 (65.40%)           15,950 (14.49%)            22,143 (20.11%)            110,093 81.5%
1994    73,294 (56.31%)           19,318 (14.84%)            37,555 (28.85%)            130,167 18.2%
1995    76,574 (60.39%)           14,339 (11.31%)            35,878 (28.30%)            126,791 -2.6%
1996    149,340 (75.82%)          12,869 (6.53%)             34,748 (17.64%)            196,957 55.3%
1997    267,201 (72.73%)          22,179 (6.04%)             78,007 (21.23%)            367,387 86.5%
1998    521,688 (72.37%)          55,917 (7.76%)             143,217 (19.87%)           720,822 96.2%
1999    498,035 (54.11%)          98,981 (10.75%)            271,772 (29.53%)           920,306* 27.7%

*:Liquidity and flexible funds account for 51,518 bn ITL, i.e. 5.6% of the total.

Table 3
Ranking of the main OECD countries by managed funds as a percentage of GDP in 1998 (Stock market capitalization in
parenthesis). Source: Assogestioni

1) USA              60.1% (122%)              9) Korea*           24.2% (17%)         17) N. Zeland       12.7%    (32%)
2) France           40.4% (51%)              10) Sveden           22.5% (95%)         18) Denmark         10.3%    (42%)
3) Spain            40.0% (47%)              11) Netherl.         21.5% (137%)        19) Japan            9.2%    (52%)
4) Italy            34.7% (38%)              12) Belgium          20.9% (71%)         20) Germany         8.4%     (39%)
5) Canada           32.9% (61%)              13) Portugal         20.5% (30%)         21) South Afr.      8.4%     (51%)
6) Australia*       27.8% (10%)              14) UK               19.3% (143%)        22) Taiwan           7.2%    (18%)
7) Ireland*         29.2% (59%)              15) Austria          13.7% (62%)         23) Norvay           7.0%    (18%)
8) Greece           24.6% (34%)              16) Brazil           13.3%   (7%)        24) Finland          4.2%    (88%)
                                                                                      25) India            1.9%     (3%)
*: previous period data

           As a consequence, a very lively debate is taking place in Italy about the ability of mutual

funds to add value through their financial performance. For example in June 1998 a study by

Mediobanca which concluded that Italian funds were unable to offer a higher return than Italian T-

Bills in the period 1990-1998, caused a very harsh reaction by Assogestioni (the association which

represents all Italian funds) and a proliferation of several articles on press by contributors who

    The stunning growth rate is evident even if we adjust for the asset rate of return and look at net inflows. Between 1996
defended one party or the other. Of course this kind of interventions could not be very detailed and

could hardly go beyond a simple qualitative argument and generic conclusions. This study wants to

contribute in a rigorous way to this debate analyzing in a scientific way the performance of mutual

funds in Italy in the period 1989-1999.

        Besides, with the recent launch of pension funds also in Italy, performance evaluation will

be even more important for plan sponsors when choosing the portfolio managers who have to

manage pension contributions.

        Finally, managed portfolios evaluation can be a variable worth being monitored also by

monetary authorities to prevent a financial markets turmoil: for example the difficulties experienced

by Long Term Capital Management were among the reasons contributing to the decision of the

Federal Reserve to reduce official interest rates in autumn 1998 to avoid a crisis in the payment


        From a theoretical point of view, a performance evaluation study can be important also to

verify the assumption of market efficiency, or, better, which definition of market efficiency fits

actual data in the best way. In fact, before Grossman-Stiglitz (1980), market efficiency was defined

in its semi-strong definition as a situation where all agents had exactly the same information so that

all risk adjusted returns were zero. If this were actually the case, performance evaluation studies

would find that performances of actively managed portfolios, net of management costs, would be

less than those of passively managed portfolios, because portfolio managers trading activity would

not be based on superior information but would be a simple waste of time and resources. With

Grossman-Stiglitz (1980) instead, a new concept of ‘dynamically efficient market’ was introduced.

The crucial assumption was that acquiring information could be costly and that agents would

acquire information only if expected profits deriving from this activity were positive. In this case

performance studies would find that net extra-performance of actively managed portfolios would be

and 1998 net inflows were respectively 45.8%, 72.6% and 85.1% of the previous period net asset value.
zero or positive and conclude that management fees rewarded managers for their research and

trading activity.

2. A brief review of the empirical literature

        The extra-performance of a managed portfolio versus a benchmark portfolio can derive from

two sources. One source of performance is the ability of the portfolio manager to ‘time the market’

and is called market timing. This occurs when the portfolio manager correctly anticipates the

‘bearish’ and ‘bullish’ periods and effectively implements strategies to benefit from this by

increasing the beta3 of the portfolio in the first case and by reducing it in the second case.

        The second source of performance is ‘security selection’ and is due to the ability of the

portfolio manager to select (avoid) securities which are expected to yield positive (negative)

abnormal returns4.

        So, if a manager has a better information than the market and/or superior analysis skills,

his/her portfolio will show a positive extra-performance and his/her job will add value compared to

a passively managed portfolio.

        Most of the classical empirical studies on mutual fund performance used the Jensen alpha

(JA) as performance measure. Introduced by Jensen (1968) and derived from the Capital Asset

Pricing Model, this measure (JA) is the difference between the average actual excess return 5 of a

portfolio and its theoretical excess return which is a function of its beta. That is

=M(Rp)-p*M(Rm)                                                                                       (1)

where M(Rm) is the average excess return of the market/benchmark portfolio, p=Cov(Rp,Rm)/m2

and m2 is the variance of the market portfolio return. In other words, JA is the vertical distance

between the portfolio and the ‘Security Market Line (Fig.1).

  By beta in this example we mean the same as in the CAPM. See Sharpe (1964), Lintner (1965), Huang-Litzemberger
  Where the expression ‘abnormal return’ indicates a risk-adjusted return measured according to some models such as
the CAPM.
  By excess returns of a security we mean the difference between the return of that security and the risk-free rate.
To estimate the JA for a portfolio, one can estimate a simple market-model equation such as:

Rp=+p*Rm +                                                                                 (1a)

where the JA and the systematic risk are obtained simultaneously.

                                         Fig. 1: Jensen alpha and Security Market Line
                                                               Jensen alpha
                         excess return

                                                                       Security Market Line


       Since the early Jensen study was published in the sixties, a lot of studies were produced to

assess the investment performance of mutual fund managers.

       Jensen himself (1968) found that for the 1945-64 period, returns to investors in U.S. funds

after management fees (and before load fees) were on average, about 1% per year below the

security market line, and the average return on more than half of his funds was below this line. Only

when all the published expenses of the funds were added back, did the average return on the funds

scatter randomly around the market line.

       Ippolito (1989) provided an extensive analysis of 143 U.S. funds for the period 1965-1984.

He found that fund returns, before load fees but after other expenses, were on average 0.83% per

year above the market line.

       But Elton, Gruber, Das and Hklavka (1993) found that Ippolito’s conclusions were flawed

because he did not consider that in his sample there were funds that invested also in non-S&P

stocks whereas he used the S&P 500 as a benchmark. Thus, they proposed a three-factor model

consisting of S&P 500 index, non-S&P 500 stocks, government and corporate bonds to evaluate the

performance of mutual funds in the period 1965-84. With this correction, they found that mutual
funds produced an abnormal return of –1.1% per year when measured against their benchmark. This

work could be seen as the one that determined the end of a first generation of performance

evaluation studies. Indeed, it highlighted that results strongly depended on the chosen benchmark

and that it was reasonable to use more indices.

           And in fact, with the evolution of the asset pricing models and the introduction of the multi-

factor models such as the APT, the most recent performance evaluation studies followed the

provocative article of Fama-French (1992). This study stated that the CAPM did not hold because 

did not seem to help to explain the cross-section of average stock returns. Moreover, the

combination of size and book-to-market equity were better predictors of returns6.

           Hence, if these variables were significant predictor of stock returns, managers were likely to

condition their investment decisions to them, and should be judged against the return that these two

variables predicted.

           Carhart (1997) is an example of this recent stream of literature: he used a four factor model

that could be interpreted either as a model of market equilibrium with four risk factors, or as a

performance attribution model, where the coefficients on the factor mimicking portfolios indicated

the proportion of mean return attributable to four elementary strategies (high versus low beta stocks,

large versus small market capitalization stocks, ‘value’ versus ‘growth’ stocks and one year return

‘momentum’ versus ‘contrarian’ stock). The author found that during the 1962-1993 period, the

market value and momentum of stocks explained the bulk of the difference in returns between best

and worst performing equity funds.

           In this paper we assessed mutual fund performance in Italy in the decade April 1989-March

1999. We used absolute performance data, portfolio compositions and net inflows data to measure

the security selection and market timing skills with different performance measure models such as a

linear market model, a multi-factor approach a la Fama-French, a quadratic model and a conditional


    For a critique of Fama-French results see Kothari-Shanken-Sloan (1995).
3. The Data
        As we already mentioned, this study deals with the performance evaluation of a sample of

Italian funds. This sample contains the funds that, according to Assogestioni classification,

belonged to the category of domestic equity and domestic balanced funds. According to this

classification, in the period covered by this study, domestic equity funds invested more than 70% of

their asset in Italian stocks, domestic balanced funds could invest between 15% and 50% into equity

and between 50% and 85% into bonds subject to the constraint that Italian assets exceeded 60% of

total assets.

        Among these funds we selected the ones that at the date of 31 March 1999 had been existing

for at least ten years. These funds added to a total of 41. For each of them, the end-of-month price

of a unit in the fund (net of all fees and expenses except load fees) and the dividends distributed in

the period 31/04/89-31/03/99 (8904-9903) were taken from Money Mate, the main data provider on

Italian mutual funds. Appendix I lists these funds, their starting date and net asset values at the start

of 1999. Table 4 below reports some useful summary statistics about this sample of funds. In the

first quarter of 1999, our sample represented 52.72% of the total assets under management of the

related categories. In particular the asset value of the domestic equity funds in our sample accounted

for 21.6% of all domestic equity funds, whereas the assets of our domestic balanced funds

accounted for 94% of the total assets of the category.

Table 4
Summary statistics for 41 domestic equity and balanced mutual. Management fees and NAV are as of the first quarter
1999, average annual return are on annual compounding basis.
                          Total     NAV       Average management fee7       Average annual return
                                   (bn ITL) Eq. weight.     Val. weight.   Eq. weight. Val weight.
All funds                 41       47,898     1.37%          1.39%          11.53%      11.37%
Equity (domestic)         13       11,194     1.59%          1.63%          13.17%      13.16%
Balanced (domestic)       28       36,704     1.27%          1.31%          10.77%      10.83%

  Many funds charge some more fees in the form of load fees and some fancy performance-related fees that do not have
any sensible meaning. It is not rare for example to see a performance-related fee linked to the over-performance of a
fund with respect to the rate of inflation, which is not the correct benchmark for that fund. It is worth noting, however,
that only a small fraction of all these fees goes to asset management companies. In fact, they mainly remunerate the
banks which are in most cases the main distributors and owners of asset management companies. Even if the peculiar
structure of the mutual fund industry in Italy is an extremely interesting topic, it is not the object of this work.

4. The ‘Jensen alpha’ in a linear market model

         Before applying equation 1a to compute JA for our sample of Italian mutual funds, we took

explicitly into account the fact that balanced funds could have a non-negligible stake of the

portfolio invested also in government bonds. Thus, we estimated:

Ri=+m*Rm +I (i)*g*Rg                                                                                    (2)

where Ri is the monthly net excess return of the fund8, Rm is the excess return of the Comit total

return index, Rg is the excess return of the JP Morgan Italian government total return index and I is

equal to 1 if the fund is a balanced one, or 0 if the fund is an equity fund.

         The average JA was 38 basis points on annual basis. (Table 5 and Fig.2). Only one fund had

a significantly negative alpha, while, among the 25 which showed a positive alpha, only 2 passed

the test at a 95% confidence level. The average monthly alpha was not significantly positive and not

significantly different across the two categories of analysed funds.

Analogously, an equally weighted portfolio made of all the 41 funds would yielded an annual JA

equal to 41 basis points, not significantly different from zero.

Table 5
Summary statistics for the monthly Jensen alpha for 41 domestic equity and balanced mutual funds in the period 8904-
9903 (equation 2)

                          Full sample (41 funds)     Equity Funds (13 funds)      Balanced Funds (28 funds)

Average                             0.00032                     0.00028                  0.00034
Standard deviation                  0.00134                     0.00089                  0.00154
Max                                 0.00498                     0.00215                  0.00498
Min                                -0.00224                     -0.00155                 -0.00224
N>0                                 25                          9                        16

Significance test
on the average
t-stat:                            1.53                         1.13                     1.17

weighted’ portfolio                0.000347                     0.000277                 0.000342
(t-stat)                           0.409                        0.249                    0.471

 Funds’ net returns are total (i.e. price returns plus dividend returns) returns net of management fees and other expenses
such as custody fees and taxes but not load fees. The word ‘excess’ refers to the fact that all returns are computed as a
difference to the risk free rate which is the 3 month T-Bill (BOT) rate.
                                        Fig.2: Distribution of monthly Jensen alpha for 41
                                          Italian mutual funds in the period 8904-9903

                         # funds    8
                                             <= -0.22% -0.10%   0.02%   0.14%   0.26% 0.38% >=

5. The ‘Jensen alpha’ in a multi-factor market model

       In the previous paragraph, the risk-adjusted performance measure was computed with

respect to the most intuitive, large and generic benchmarks we could imagine: a stock market total

return and a bond market total return where necessary.

       In this paragraph instead, we analysed changes in results when considering the factors which

Fama-French (1992) found to explain the cross section of stock returns: the question we wanted to

answer was whether mutual fund performance reflected superior stock selection skills or common

factors in stock returns explained most of the predictability in mutual fund returns.

       Thus, we estimated first an equation taking into account only the book to market equity

factor such as:

Ri=+m*Rm + v/g*Rv/g + I (i)*g*Rg                                                             (3)

       In this case, Rm is the excess return of the MSCI total return index for Italy, Rv/g is the

difference between the MSCI total return index for Italian ‘value’ stocks and the MSCI total return

index for Italian ‘growth’ stocks. In order to construct these series, the MSCI service ranks in

ascending order all securities according to their previous month end price-to-book value ratio

(P/BV). Current market capitalizations are then used to sum the security market capitalization until

approximately half of the total market capitalization has been reached. All securities above this

point are designated as ‘value’ securities and all securities below are designated as ‘growth’


        Summary statistics from equation 3 are shown in the first column of table 6.

Table 6
Summary statistics for the Jensen alpha for 41 domestic equity and balanced mutual funds on the basis of monthly data
according to the Fama-French approach.

Period                    8904-9903                  9602-9903
Model                     (Eq. 3)                    (Eq. 4)
Average                   -0.0015                    -0.00144
Standard deviation         0.00148                    0.00164
Max                        0.00193                    0.00319
Min                       -0.00453                   -0.00476
N>0                       7                          6

        The addition of the value/growth factor was useful in most cases, as its coefficient was

significantly different from zero for 28 funds.

As it is evident from the summary statistics presented above, on average mutual funds provided a

negative JA, -1.8% on annual basis, after controlling for the value/growth effect (Fig.3). Actually

on the whole period no funds offered a significantly positive JA whereas 14 funds presented a

negative one.

        In order to include the small cap effect, we also estimated equation (4) below

Ri=+m*Rm + v/g*Rv/g + s*Rs+ I (i)*g*Rg                                                                (4)

where Rm is the excess return of the Mediobanca total return stock index for Italy, R v/g is the same

as above and Rs is the difference between the Mediobanca 30 largest cap companies total return and

the smaller companies total return. These series were available since 1996 and the equation was

estimated on the related sample9. Summary statistics are reported in the second column of table 6.

   MSCI also started releasing also a small stock (price) index in 1998. We obtained this series starting in 1993 and
performed an analysis including this effect but the results are affected by a dividend bias as the MSCI small stock index
is not a total return index and there could be different dividend payment patterns between small and large firms. For this
reason they are not reported in this work.

       In both cases, controlling for ‘styles’                                 (value/growth, small/large cap factors) led to a

decrease in the measure of performance with respect to the results presented in the previous section.

                                   Fig.3: Distribution of monthly Jensen alpha for 41
                                  Italian mutual funds in the period 8904-9903 (Fama-
                                      French approach with book to market equity







To understand why this was the case, it can be useful to look at three things: the funds’ ‘style’, the

relative performance of ‘value’ vs ‘growth’ stocks and the relative performance of ‘small’ vs ‘large’

stocks. In fig.4 we plotted the scatter diagram of the funds’ ‘style’: that is the coefficient of the

book-to-market and size factors for the 41 mutual funds in the period 9602-9903. A fund that was

plotted in the right side of the graph was tilted towards ‘value’ stocks, and a fund plotted in the

upper part was tilted towards ‘small cap’. Fig.5 shows instead the relative performance of ‘value’ vs

‘growth’ stocks and of ‘small’ vs ‘large’ caps. Fig 6 reports the exposure to the value/growth factor

for the whole period 8904-9903.

It resulted that, on average, during this period, fund managers were more exposed than the Comit

Index to ‘value’ stocks. Besides, ‘value’ stocks outperformed ‘growth’ stocks, on average, by

0.24% per month in that period. This biased upward funds’ performance measures when they were

compared only to the Comit Index as we did in section 4. However, this effect vanished when we

explicitly adjusted funds’ performances for this factor.

Combining the results of the multi-factor approach a la Fama-French with what we got applying

equation 2, we conclude that Italian fund managers were able to recover their fees mostly from style

rotation, that is by investing more in value stocks, which outperformed growth stocks. Should that

be taken as added value? In other words, which model conclusion should we accept: the one coming
from applying equation 2 (net JA not significantly different from zero) or the one deriving from

applying equation 3 (negative JA)?

              Fig.4: Mutual funds' style in 9602-9903                                            Fig.5: Style relative performance

                                     'small'                                            1.5
                                    0.15                                                                      'value' vs 'growth'
                      Comit Index
                                    0.05                                                1.1
       'growth'                                                        'value'          0.9
              -0.15    -0.1   -0.05 -0.05 0      0.05   0.1     0.15

                                     -0.1                                               0.7                                         'small' vs 'large'

We think that we can answer this question more appropriately after that we recall a few details of

the Italian mutual fund industry with respect to other countries. In US for example, there is a huge

variety of equity funds specialized by their style: large cap equity funds, growth equity funds, large

cap value stocks funds etc. In that context, the style choice is up to the final investor who can fairly

pretend that the chosen fund should add value with respect to the benchmark which explicitly

                                Fig. 6: Mutual fund exposure to the value/growth factor in the period


                         0       0,02          0,04     0,06     0,08            0,1   0,12   0,14    0,16      0,18          0,2

includes that specific style/factor. In Italy instead, this is not yet the case: the mutual funds we

considered did not have a declared style, so it would make sense assuming that the final investor

was delegating to the fund manager also the style choice. Hence, the Italian fund managers should

be awarded also the results (positive in this case) from actively managing the style.

6. Market timing and security selection (I)
        As Jensen (1968) and Grant (1977) showed, the JA computed as in equation 1 is a correct

measure of performance only if the portfolio managers do not try contemporaneously to time the

market. If, instead this is the case, then this simple measure of performance does not work:

according to different circumstances, the estimated alpha may over- or under-estimate the true

security selection ability of the manager.

        A very simple way to consider explicitly the market timing activity and keep it separated

from the JA which should measure only the security selection ability, was introduced by Treynor-

Mazuy (1966). They suggested to use the following equation

Rp= +1*Rm+2*Rm2 + ut                                                                                   (5)

In fact, as dRp/dRm=1+2 2*Rm, this equation implicitly says that the overall sensitivity of the

portfolio return to the market return increases (decreases) as the return of the latter increases

(decreases), which is the mathematical translation of the market timing concept. 10

To take into account the particular composition of balanced funds, the equation we used was

Ri=+m*Rm + Rm2+ I (i)*g*Rg                                                                            (6)

Results are presented in Table 7.

An interesting result was that, keeping separated the two sources of performance for the analysed

funds, security selection ability (captured by alpha in equation 6) and market timing skill (captured

by 2 in equation 6) went in opposite directions, and this was especially true for balanced funds. In

fact while the average JA was an annualised 1.16%, significantly positive at a 95% significance

level, the market timing coefficient was on average significantly negative: only one fund had a

significantly positive market timing coefficient and twelve funds had a significantly negative

coefficient. Regressing the 41 market timing coefficients on the respective 41 JAs we would get the

following result (t-stat in parenthesis) with the relative scatter plot in Fig. 7.

   Even if this equation was first introduced by Treynor-Mazuy (1966) in a sort of ad hoc way, without any structural
model supporting it, later Jensen (1972) and Battacharya-Pfleiderer (1983) derived a similar relation starting from
structural assumptions.
Tav. 7
Summary statistics for the Jensen alpha for 41 domestic equity and balanced mutual funds on the basis of monthly data
for the period 8904-9903 (equation 4).

                          Market Timing             Alpha di Jensen

Full sample
(41 funds):
Average                   -0.1388                    0.0009686
Standard deviation         0.266678                  0.0016358
Max                        0.43650                   0.004325
Min                       -0.74417                  -0.00297
N>0                        10                        28

Significancy test
on the average
t-stat:                   -3.33                     3.75

Equity Funds (13 funds):
Average                  0.0186                      0.000186
Standard deviation       0.1561                      0.001341
Max                      0.3023                      0.003032
Min                     -0.1796                     -0.001445
N>0                     4                            6

Significancy test
on the average
t-stat:                   0.4296                    0.5

Balanced Funds (28 funds):
Average                -0.2119                       0.00133
Standard deviation      0.2776                       0.00168
Max                     0.4365                       0.00432
Min                    -0.7442                      -0.00297
N>0                     6                           22

Significancy test
on the average
t-stat:                   -4.03                     4.189

MT=-0,0464-95,4058*JA                                                                                   (7)
    (-1,17) (-4,58)

7. Security selection and market timing (II)
       One of the strongest critique to the CAPM was moved by the empirical literature on the

‘predictability’ of stock returns. These studies (see for example Ferson-Harvey (1993)) showed how

stock returns depended not only on their beta but also on a set of given financial and macro

variables such as: dividend yield, interest rate levels, term structure slope and so on. Since this

literature had been around for quite some time and these factors are public information by now, it

seemed correct to adjust managed portfolios performance for the effects of these factors which

made risk premia vary over time.

       In fact, let’s consider for example a portfolio manager who wishes to keep the overall

volatility of the portfolio constant. Conditioning his/her decisions on the variables that predict risk

premia, he/she will have a lower (higher) beta when a higher (lower) volatility is expected on the

basis of those variables. In such circumstances the beta of the fund will be inversely correlated with

the expected market return and the estimated alpha will be negative. However this does not mean

that the manager has had bad skills.

       The so called conditional models allowed for beta changes explained by the arrival of public

information about risk premia and tried to adjust the performance measure for this effect. Such

models (see Ferson-Schadt, 1996) considered explicitly that the beta of a portfolio is a function of

certain exogenous variables:
Rtp= +p(Zt-1)*Rtm + ut                                                                (8)

where Zt-1 is the set of information variables

If we use a linear Taylor approximation for the beta function we have

Rtp= + b0p Rtm +BpI zt-1Rtm + ut                                                       (9)

Where zt-1 is a vector of the differences of each information variable from its mean and BpI is a

vector of the reaction coefficients to these variables.

To obtain also a measure of market timing, we can also add a quadratic terms in equation 9 to get

Rtp= + b0p Rtm +BpI zt-1Rtm + 2*Rm2 + ut                                              (10)

       Ferson-Schadt (1996) applied these models to a sample of 67 US mutual funds between

1968 and 1990 using as information variables the T-Bill interest rate levels, the stock market

dividend yield, the slope of interest rate term structure, the spread between corporate bonds and

government bonds, a dummy for the ‘January effect’.

       Applying equation 1 to this sample, they found that two thirds of the funds showed a

negative JA, eight of which with statistical confidence. Using the model of equation 9, half of them

(34) showed a negative JA (with statistical confidence for 6 of them). Finally using the quadratic

version (equation 10) only 27 funds showed a negative market timing coefficient (with statistical

significance only for two). The same market timing ability measured instead with the Treynor-

Mazuy model would have implied that 44 funds showed a negative market timing coefficient (with

statistical significance for 8 of them).

       Overall, the Ferson-Schadt (1996) study showed that considering explicitly the public

information represented by the conditional variables, the performance of the funds appeared less

negative than in the unconditional estimates. This was explained by the fact that the betas of the

those portfolios were negatively correlated with market returns. So the question was why fund

managers used to reduce the beta when the market returns were expected to be higher and


        In an other study, Ferson-Warther (1996) showed that this phenomenon was the result of a

slow adjustment of new inflows into the funds to real investments. In other words, inflows used to

move into funds according to the variables and the direction the theory would suggest, but portfolio

managers were slow in investing that money coming into the fund. The authors estimated that for

each dollar of new inflows only 62 cents were invested in the same month.

        We applied the same model (equation 9 and 10) to our sample of 41 funds using as

conditional variables: the 3 month interest rate on Italian T-Bills (BOT) (Cond1), the dividend yield

on the Comit Total Return Index (Cond2)11, the difference between long term government bonds

(BTP) and short term rate (3 month BOT) (Cond3), the difference between the prime rate applied

by banks to their best clients (prime rate ABI) and the government bond (BTP) rate (Cond4), and a

dummy for the January month (Cond5). Each of these variable (except the last one), was considered

as delayed by one month.

        The result of the estimate of equation 9 and 10 (obtained according to White procedure to

account for eteroschedasticity) are showed in Table 8. From the first column of Table 8, we can see

that the average extra-performance was equal to –8.4 basis points per annum, whereas according to

the result of equation 2 showed in table 5, it was a positive 38 basis points. However in both cases it

was not different from zero. Besides, equity funds showed a significantly negative extra-

performance equal to –132 basis points per annum, while balanced funds extra-performance was a

positive 48 basis points.

        The following two columns in Table 8 allowed us to understand if the source of this result

came from market timing or security selection. We wish to remember that in this context, market

timing ability and security selection ability are meant to be as superior information and analysis

capability of the manager after considering explicitly the exogenous variables we have inserted in

our estimates.

   By dividend yield we mean a rolling yield computed over the last 12 months preceding a given one. It was derived
from the price return and total return series of the Comit Index.
Table 8
Summary statistics on market timing and risk tolerance for 41 mutual funds in the period April 1989-March 1999
(equation 9-10).
                         Jensen alpha               Jensen alpha               Market Timing
                         (eq. 9)                    (eq. 10)                    (eq. 10)

Full sample
(41 funds):
Average                  -0.00007                   0.00064                  -0.18872
Standard deviation        0.00144                   0.00150                   0.14088
Max                       0.00347                   0.00380                   0.24330
Min                      -0.00274                  -0.00242                  -0.45349
N>0                       20                        28                        4

Significance test on
the average:
t-stat:                  -0.32                      2.73                     -8.57

weighted portfolio’      0.0002                     0.000861                 -0.18143
(t-stat)                 0.278                      1.147                    -1.656

Equity funds (13 funds):
Average                  -0.0011                   -0.0002                   -0.2190
Standard deviation        0.0010                    0.0012                    0.1200
Max                       0.0009                    0.0024                    0.0549
Min                      -0.0027                   -0.0020                   -0.3710
N>0                      2                          5                         1

Significance test on
the average:
t-stat:                  -3.81                     -0.59                     -6.58
‘Equally weighted
portaolio                -0.00108                  -0.0002                   -0.2190
(t-stat)                 -1.46                     -0.227                    -1.74

Balanced funds (28 funds):
Average                  0.00040                    0.0010                   -0.1746
Standard deviation       0.00137                    0.0015                    0.1495
Max                      0.00347                    0.0038                    0.2433
Min                     -0.00197                   -0.0024                   -0.4535
N>0                      18                         23                        3

Significance test
on the average:
t-stat:                   1.538                     3.71                     -6.57

‘Equally weighted
portaolio                0.00040                   0.0010                    -0.1746
(t-stat)                 0.576                     1.47                      -1.604

         At aggregate level, the market timing contribution was negative, the security selection one

was positive; for balanced funds this pattern was even more evident.

         The fact that the overall performance measured with conditional models was lower than the

one measured in an unconditional models such as equation 2, meant that fund managers moved the

beta of their portfolios consistently with the theory so that after adjusting their performance for the

use of this theory their performance looked worse than before the adjustment.

        To understand better this conclusion and compare it with what obtained by Ferson-Schadt

(1996) and Ferson-Warther (1996), we analysed the conditional betas, the net inflows and

conditional variables in the period January 1992- March 1999. As evidence of that, Table 9 shows

the conditional (equation 9) and unconditional betas (equation 2) and the sensitivity coefficients of

the former. We can see that the conditional bets moved consistently with explanatory variables

especially short term rates and dividend yield.

Table 9
Summary statistics on market timing and risk tolerance for 41 mutual funds in the period January 1992-March
according to the conditional (eq.6) and unconditional model (eq.2)

                                  Jensen alpha                       Conditional beta

                                  Non cond.        Cond.             b0      b1      b2       b3       b4      b5
Full sample
(41 funds):

‘Equally weighted’ portfolio      0.000357         0.000135          0.50    -2.42   7.55     -1.75    -1.66   0.04
(t-stat)                          0.34             0.155             32.9    -3.31   2.61     -0.95    -0.69   1.10

Equity funds (13 funds):

‘Equally weighted’ funds          0.000231         -0.001138         0.66    -4.67   13.17    -3.01    -3.74   0.02
(t-stat)                          0.16             -1.24             38.2    -5.35   3.86     -1.37    -1.31   0.49

Balanced funds (28 funds):

‘Equally weighted portfolio’      0.000664         0.000382          0.43    -1.33   5.35     -1.27    -0.51   0.04
(t-stat)                          0.75             0.45              29.4    -1.89   1.91     -0.72    -0.22   1.24

        The upper part of Table 10 shows that net inflows also behaved according to the theory:

investors moved flows according to the direction predicted by the theory especially with respect to

the short term rate and dividend yield. Finally, the bottom part of the same table shows that the beta

of the funds adjusted themselves to new inflows in a consistent way.

        These outcomes demonstrates that in the Italian case the experience was different from what

Ferson-Warther (1996) found for the US: Italian funds were able to adjust the risk profile almost at

the same time new funds came in or out according to what investors really wanted. Probably the

sometimes-criticized system of distribution, mostly through banks and specialized dealers, allowed

fund managers to collect information on new inflows and act in a preemptive way.

Table 10
Net inflows, conditional betas, and conditional variables

Conditional variables and net inflows:
Net inflows(t)=c0+c1Cond1(t-1)+c2Cond2(t-1)+c3Cond3(t-1)+c4Cond4(t-1)+c5Cond5(t-1)+
+c6Net inflows(t-1)+c7Net inflows (t-2)

                                            c0       c1           c2     c3      c4      c5
Full sample
(41 funds):

‘Equally weighted’ portfolio                0.006    -0.42        1.86   -0.61   -0.87   0.02
(t-stat)                                    0.29     -2.02        2.32   -1.38   -1.39   2.66

Equity funds (13 funds):

‘Equally weighted’ portfolio                0.028    -0.75        2.87   -1.70   -1.93   0.01
(t-stat)                                    0.66     -1.83        1.68   -1.77   -1.50   0.69

Balanced funds (28 funds):

‘Equally weighted’ portfolio                -0.000 -0.33          1.58   -0.33   -0.56   0.02
 (t-stat)                                   -0.021 -1.86          2.33   -0.92   -1.07   3.85

Conditional betas and net inflows:
Conditional betas(t)=c0+c1 Net inflows+ k Conditional betas(t-1)

                                            c0                    c1
Full sample
(41 funds):

‘Equally weighted’ portfolio                0.66                  0.08
 (t-stat)                                   12.4                  4.73

Equità funds (13 funds):

‘Equally weighted’ funds                    0.77                  0.05
(t-stat)                                    10.8                  2.23

Balanced funds (28 funds):

‘Equally weighted’ funds                    0.57                  0.05
 (t-stat)                                   12.2                  4.76

8. Conclusions
       Mutual fund performance evaluation is one of the most interesting type of study in finance

both for market efficiency implications and to assess the ability of the mutual fund industry to add

value. In this paper we assessed this ability for the Italian mutual fund industry after 15 years since

its inception with special regard to the two possible source of extra-performance: security selection

and market timing.

       In aggregate, the evidence showed that Italian mutual funds offered a net extra-performance

not significantly different from zero, implying that the information and analysis skills went to cover

costs (mainly management fees).

       Though, this neutrality outcome hides two opposite patterns: on one side the security

selection and style rotation produced very positive results, on the other side the market timing

contribution was mostly negative. This could be the result of the participation of mutual funds in

IPOs which usually offered positive abnormal returns at security level, whereas it showed how

predicting the general direction of the market could be a fruitless exercise.

       Finally mutual funds have been an efficient way to invest according the rational models

which explains stock market risk premia. In fact fund managers have been able to accommodate the

risk profile of their portfolios according to the rational patterns of inflows.

       The success of the Italian mutual fund industry is terrific: the increase in the asset under

management is the result not only of the recent very good returns of the stock market but also of a

huge amount of net inflows (which could have been affected of course by the exceptional market


       Looking forward, it is very difficult that these exceptional market returns can repeat

themselves in the future. Moreover, the increasing level of competition by foreign funds and index

funds will put more and more pressure on the existing environment so that only players that can add

value in the most efficient way will enter the maturity phase of this industry.

Appendix I
List of the 41 Italian mutual funds which as of March 1999 had been existing for at least 10 years.
(Funds with performance fees in italics)

                 Category* Starting date     Management fee        NAV                Average
                                              (as of Feb 99)      (bn ITL)         yearly return**
Arca BB                  B   18/09/84             1.60%             5,676             10.71%
Aureo                    B   01/07/85             1.00%             1,518             10.72%
Aureo Previdenza         A   02/05/88             1.50%                798            12.80%
Azzurro                  A   19/03/85             1.50%             1,432             12.84%
BN Bilanciato            B   04/11/85             1.30%             1,345               7.46%
Capitalcredit            B   07/04/86             1.25%             3,668               9.93%
Capitalgest Bilanciato   B   08/07/85             1.50%                196            10.08%
Capitalgest Italia       A   02/01/89             1.62%                405            12.91%
Carifondo Libra          B   25/03/85             1.20%             3,899             12.46%
Centrale Capital         A   01/02/88             1.80%                395            15.72%
Cisalpino Bilanciato     B   09/06/88             1.40%             1,882             12.42%
Eptacapital              B   26/05/86             1.60%                926              9.28%
Euromob. Capitalfit      B   20/12/85             1.45%                549            12.26%
F&F Lagest Az,Italia     A   15/02/88             1.98%                992            14.38%
F&F Professionale        B   26/11/84             1.50%                 831           11.77%
Fdnvest Piazza Affari    A   13/02/89             1.92%              1,207            13.47%
Fondersel                B   24/08/84             1.20%              1,057            11.35%
Fondicri Bilanciato      B   04/08/86             1.25%                 344             8.18%
Fondinvest Futuro        B   20/06/85             1.40%              1,824            10.96%
GenerComit               B   27/12/84             1.00%              2,367            11.77%
GepoCapital              A   26/10/88             1.20%                 482           12.10%
GepoReinvest             B   20/10/86             1.00%                 535           10.65%
Giallo                   B   13/06/88             1.00%              1,632            11.81%
Grifocapital             B   31/10/88             1.44%                 436           11.80%
Imicapital               B   02/07/84             1.00%                 505             8.31%
ING Portfolio            B   03/02/86             1.45%                 362           14.90%
Interbancaria Az.        A   21/01/85             1.50%                 321           10.73%
Intermobiliare Fondo     B   18/04/88             0.30%                 105           11.39%
Investire Azionario      A   01/04/88             1.80%              1,167            12.24%
Investire Bilanciato     B   01/04/88             1.50%              1,601              9.34%
Multiras                 B   05/04/85             1.50%              1,048            10.22%
Nagracapital             B   01/07/85             1.32%                 919             8.84%
Nordcapital              B   06/10/86             1.80%                 603             9.90%
Prime Capital            A   22/10/84             1.20%                 528           13.38%
PrimeClub Azion.Ita.     A   25/06/87             1.80%                 270           13.88%
PrimeRend                B   22/10/84             1.20%                 189           11.86%
Rolomix                  B   22/06/87             1.20%                 954             8.36%
Spaolo Aldebaran It.     A   15/09/88             1.60%              2,644            13.17%
Visconteo                B   20/05/85             1.20%              1,100            12.39%
Zeta Azionario           A   04/01/89             1.30%                 552           13.63%
Zeta Bilanciato          B   04/05/87             1.10%                 632           12.32%

*: A: domestic equity fund, B: domestic balanced fund
**: Yearly compounded


-Bhattacharya S.-Pfliederer P. 1983 ‘A Note on Performance Evaluation’ Technical Report 714,

       Stanford University, Graduate School of Business.

-Carhart M. 1997 ‘On Persistence in Mutual Fund Performance’ Journal of Finance 52.

-Elton, E.J.-Gruber M.J.-Das S.-Hlavka M. 1993 ‘Efficiency with Costly Information: A

       Reinterpretation of Evidence from Managed Portfolios’ Review of Financial Studies 6.

-Fama E.F.-French K.R. 1992 ‘The Cross-section of Expected Stock Returns’ Journal of

       Finance 47.

-Ferson W. E.-Harvey R. 1993 ‘The Risk and Predictability of International Equity Returns’ Review

       of Financial Studies 6.

-Ferson W. E.-Schadt R.W. 1996 ‘Measuring Fund Strategy and Performance in Changing

       Economic Conditions’ Journal of Finance 51.

-Ferson W. E.- Warther V. A. 1996 ‘Evaluating Fund Performance in a Dynamic Market’

       Financial Analysts Journal Nov./Dec. 1996

-Grant D. 1977 ‘Portfolio Performance and the ‘Cost’ of Timing Decisions’ Journal of Finance 32.

-Grossman S.-Stiglitz J. 1980 ‘On the Impossibility of Informationally Efficient Markets’

       American Economic Review 70.

-Huang C- Litzemberger R. 1988 ‘Foundations for Financial Economics’ Amsterdam, North


-Ippolito R.A. 1989 ‘Efficiency with Costly Information: A Study of Mutual Fund

       Performance, 1965-1984’ Quarterly Journal of Economics 104;

-Jensen M. 1968 ‘The Performance of Mutual Funds in the Period 1945-64’ Journal of Finance


-Jensen M. 1972 ‘Optimal Utilization of Market Forecasts and the Evolution of Investment

       Performance’ Mathematical Methods in Investments and Finance, ed. Szego P. e Shell K.

-Kothari S.P – Shanken J. – Sloan R.G. 1995 ‘Another Look at the Cross-section of Expected Stock

       Returns’ Journal of Finance 50.

-Lintner J. 1965 ‘The Valuation of Risk Assets and the Selection of Risky Investments in Stock

       Portfolios and Capital Budgets’ Review of Economic and Statistics 47.

-Sharpe W. F. 1964 ‘Capital Asset Prices: A Theory of Market Equilibrium under Condition of

       Risk’ Journal of Finance 19.

-Treynor J. L.-Mazuy K.K. 1966 ‘Can Mutual Funds Outguess the Market?’ Harvard Business

       Review 44.

-Warther V. A. 1995 ‘Aggregate Mutual Fund Flows and Security Returns’ Journal of Financial

       Economics 39


Shared By: