---
title: "How do fund families utilize the fund performance evaluation system?"
document_role: "보고서 원문"
language: "ko"
published: "2007-11-29"
authors:
  - "한상범"
  - "원승연"
description: "This paper shows how the fund performance evaluation system influences the fund market and fund family behaviors."
canonical_url: "https://www.kcmi.re.kr/report/report_view?report_no=175"
subjects:
  - "자산운용/연금"
---

# How do fund families utilize the fund performance evaluation system?

[뷰어](https://www.kcmi.re.kr/flexer/view?fid=4789&fgu=002002&fty=004003)
[원글](https://www.kcmi.re.kr/report/report_view?report_no=175)
## Page 1

How do fund families utilize the fund
               performance evaluation system?:
     Study on the ‘Evaluation-ship bias’ in Korean equity funds


                SANG BUHM HAHNa*                           SEUNGYEON WONb
a
 Korea Securities Research Institute, 45-2 Yoido-dong, Youndeungpo-gu, Seoul,
                                150-974, Korea
    b
      Yeungnam University, Departments of Economics, , 214-1 Dae-dong,
              Gyeongsan-si, Gyeongsangbuk-do, 712-749, Korea



                                              ABSTRACT
            This paper shows how the fund performance evaluation system influences the fund
            market and fund family behaviors. Evaluation-ship Bias is defined as the out-
            performing of evaluated funds over non-evaluated funds. Regarding the Evaluation-
            ship Bias, this study confirmed two things in empirical analysis of Korean equity
            funds. First, this paper verified that there was Evaluation-ship Bias in Korea, so
            fund families’ performances were over-estimated. Second, it is also found that fund
            families did favoritism to evaluated funds in management activities. Fund families
            were more attentive to EV performance and took strategic and risk-taking measures
            to improve its ratings. This paper suggests that another factor, the performance
            evaluation system itself, is what causes fund families to favor some funds. The
            existence of too high numbers of Korean funds is an example of this evaluation
            system bias. We can expect similar phenomena to occur in other countries as well.
            We also suggest that the introduction of GIPS will diminish the evaluation system
            bias and contribute to the development of the Korean fund market.




                                     This Version: November 2007




    Key words: Evaluation-ship bias, Fund families, Fund performance evaluation system,
                 Evaluated funds, non-evaluated funds
    JEL Classification: G12, G14
_________________
* The authors’ thanks go out to Hanna Lee at Korea Securities Research Institute for research assistance.
 a
     Tel. #: +82-2-3771-0672, fax #: +82-2-3771-0811, e-mail: sbhahn@ksri.orgr
b
     Tel. #: +82-53-810-2713, e-mail: sywon@ynu.ac.kr



                                                                                                       1

## Page 2

I. Introduction

  This paper shows how the fund performance evaluation system influences the fund market
and fund family behaviors. It is commonly accepted that there is survivorship bias in the
evaluation of fund performance. Alternatively, this paper identifies Evaluation-ship Bias which,
due to the structure of the evaluation system, over-estimates fund performance.
  To evaluate fund performance, a fund must generally meet certain requirements such as
duration of survival or fund size. An evaluated fund (hereafter EV) is defined as a fund that
meets the evaluation criteria stipulated by a fund rating agency or other. EVs have a tendency to
perform better than the average fund. Thus, aggregating only EVs in performance evaluation
will bias overall performance. We coin this phenomenon Evaluation-ship Bias. Evaluation-ship
Bias is defined as the out-performing of evaluated funds over non-evaluated funds. From this
bias, investors receive mis-information about the management capabilities of fund families.
  Regarding the Evaluation-ship Bias, this study shows two things in empirical analysis of
Korean equity funds. First, this paper shows that there is Evaluation-ship Bias, which means
evaluated funds out-performed non-evaluated funds. Second, it is also found that fund families
did favor to EVs in management activities. Though these kinds of discriminating behaviors by
fund families are not always revealed to succeed, it is evident that funds were unfair treated by
fund families. Based on the results, we confirm that evaluation system can make information
biased and fund family do favoritism
  Theoretically, this study is similar to previous studies that analyze the discriminating
behaviors of fund families. For instance, Nanda et al. (2004) describes how fund families try to
create a “star” fund in the hopes of spill-over effects. Also, Gasper (2006) argues that some fund
families may subsidize a particular fund for better performance.
  In comparison to such research, this study highlights the evaluation system as the source of
the mis-information and the cause of the favoritism shown by families. In addition, this study
suggests that the evaluation principle such as the GIPS should be introduced to prevent the
information bias and favoritism of evaluated funds.
  This paper is composed as follows. Section II describes the background of Korea’s fund
market and practices. Section III & Ⅳ introduces our methodology and data. Section Ⅴ
evaluates the existence of evaluation-ship bias. Section Ⅵ assesses the existence of favoritism.
Section Ⅶ shows that evaluation-ship bias contributed to increasing fund flows into fund



                                                                                                2

## Page 3

families. Lastly, Section Ⅷ is the conclusion.




Ⅱ. Background

    The Korean asset management market has experienced remarkable growth in the last 3 years.
In particular, the equity market shows unparalleled growth, especially due to direct and active
participation by retail investors. Equity funds in Korea grew from $8.2 billion at the end of year
2004 to $68.7 billion, recording an annual average growth rate of 83.6%.
    As the market centers on retail investors, performance of the funds is becoming an important
criteria for fund selection. However, such emphasis on fund performance is raising controversy
regarding its evaluation. Many are voicing concerns that Korea’s current evaluation and
disclosure system may not be fully objective and fair.
    Recent movement to establish GIPS (Global Investment Performance Standards) in Korea
manifests this awareness. The National Pension Fund, the largest anchor investor in Korea,
announced plans to follow these global standards for fund selection, creating momentum for
GIPS adoption by the entire asset management sector.
    The main issue on the evaluation system is the unit of performance measurement. Currently
in Korean evaluation system, each fund is an evaluated unit while GIPS requires the composite
as a unit of evaluation. It is argued that current evaluation is making problems. Especially it
makes more problems in which there are too many numbers of funds, likely to Korea1.
    In Korea, only licensed company by law can create the funds2. Therefore, though overall fund
performance in the same composite should be measured in evaluating family’s ability, only
some funds are included in evaluation and each fund is evaluated independent off the fund
family.
    Some argues that fund family utilizes only good performing funds for advertisement behind

1
  As of end of 2006, with 8,030 funds, Korea ranked 3rd in the world in terms of the number of funds, following the
US with 8,120 funds and France with 8,092. The average fund size is much smaller, recording USD 31.3 million,
compared to the US (USD 1,282.5 million) and France (USD 218.6 million). Source; Investment Company
Institute (2007), 2006 ICI Fact Book
2
  The company is called ‘”asset management company” by Korean Act. In this paper, we will call it fund
family. Compared to mutual funds that are company type funds, most of the funds in Korea are Unit Trust
type funds. In the case of Mutual Funds, investors are the shareholders and the asset management
company takes part in the BOD. However, Korean fund investors do not have ownership but only the
right to the investment returns, and they cannot participate in the actual management of the fund.
Therefore, they can not intervene in the fund operation, leaving the entire process to the asset
management companies.


                                                                                                                      3

## Page 4

bad performing fund. Fund family also substitutes new funds for bad performing funds. In
addition, favoritism towards special funds can appear in many ways. Though Gaspar et al.
(2006) points out some funds are subsidized, there are several ways for fund family to legally3
give the favor to specific funds. For example, fund family may assign top performing fund
managers to special funds or give the priority in allocation of traded amount. As a result, it often
appears that the performance of each fund in the same composite of one family is different.
    It is the main reason why GIPS is now introduced. The most anticipated impact the adoption
of GIPS will have is the change in performance measurement, from evaluating individual fund
performance to a composite. The GIPS standard includes all portfolios that have the same
investment objective and strategy into one composite. Accordingly, funds with the same
investment objective will be evaluated on overall portfolio performance.




Ⅲ. Methodology

    In general, performance evaluation of individual funds takes place only for funds that have
certain duration of survival and fund size. EVs are defined as funds that meet both these criteria
and hence are evaluated by a fund rating agency. On the other hand, Non-Evaluated fund
(hereafter NEV) is a fund that meets the minimum duration of survival but does not meet the
minimum fund size. Lastly, a fund that does not satisfy the minimum duration is called early
fund.
    Such duration of survival and fund size criteria exist because funds that are too young or too
small can distort performance results. Therefore, it is reasonable to exclude the NEVs and early
funds in performance evaluation of individual fund. However, the argument we raise in this
paper is that the exclusion can give biased information regarding the management capabilities of
fund families. EVs have a tendency to be better than NEVs because better performing funds
would get more fund flow and longer duration. We define Evaluation-ship Bias as the out-
performing of evaluated funds over non-evaluated funds, so the overall performance of fund



3
  At around the time of the Asian Financial Crisis, there were cases of illegal asset transfer among funds
in Korea. However, after experiencing the ensuing turmoil, regulators enforced more aggressive
approaches in countering such actions. Also, with the development of Information Systems to identify
such trends, it is now almost impossible for such a situation to reoccur. Even in IPO allocation, asset
management companies must allocate IPO amount to each fund in proportion to each fund’s AUM.


                                                                                                             4

## Page 5

family is over-estimated4.



                                           Duration of Survival             Fund Size


                Evaluated Fund                      meets                     Meets



              Non-Evaluated Fund                    meets                    not meet



                   Early Fund                     not meet                     N/A




     Based on the concept of evaluation-ship bias, we did empirical analysis in three way.
    First, we conformed the existence of evaluation-ship bias, comparing the EVs’ performance
with NEVs’. After composing EVs’ portfolio and NEVs’ portfolio in terms of ex-ante and ex-post,
we compared two portfolios’ risk-adjusted returns. In addition, we compared EVs’ performance
with NEVs’ by using panel data.
    Second, we evaluate whether fund family favors EVs over NEVs with two methods. We
confirmed the favoritism in terms of application of strategies. We also tested if there is
difference of management attitude in response to the previous performance.
    Third, this study analyzed that well-performing EVs contribute to the inflow of funds. We
tested whether fund flows into EVs & NEVs differently responded to the performance. We also
analyzed the spill-over effect by EVs.




    Ⅳ. Data

    This study utilizes Korean equity fund data, attained from the Korean Fund Rating Company,
for empirical analysis. This time series data includes basic information regarding funds on a
monthly basis. The history of equity funds begins from mid 1970s in Korea. However, since
data until the 1990s are not available, we use the data starting in year 2000. The Asset

4
  “Survival Bias” refers to when poor performing funds disappear from the market and performance is
evaluated only for the surviving funds. Meanwhile, the “Evaluation Bias” refers to the situation where
even if the funds have survived; some are excluded from the performance evaluation.


                                                                                                         5

## Page 6

Management Association of Korea keeps all data and information regard funds and distributes it
to fund rating agencies and others. Although the data originates with the Association, data from
the fund rating agencies have been cleaned; thus we use agency’s data.


                              Table 1. Trends of Equity Funds in Korea
                                                   (unit: as of the end of year, number, Million US$)

                Type         Yr 2001      Yr.2002       Yr.2003      Yr.2004       Yr.2005      Yr.2006

             EVs                     9           47            59           38            61          124
    No. of   NEVs                    6           35            65           75            72           88
    Funds    Early Funds          116            86            25           48          105            90
             Total Funds          131           168          149           161          238           302
             EVs                272.8       2,083.3      2,168.8       1,740.5      8,739.9     26,385.6
             NEVs                  5.0         42.3        141.0         121.1          90.3        188.0
    AUM
             Early Funds      1,946.8       1,268.9        145.8         647.8      4,650.2       2,134.4
             Total Funds      2,224.6       3,394.5      2,455.6       2,509.3     13,480.3     28,708.0
             EVs                  30.3         44.3          36.8         45.8        143.3         212.8
    AUM
             NEVs                  0.8           1.2          2.2          1.6           1.3          2.1
     Per
             Early Funds          16.8           6.3          5.8         13.5          44.3         23.7
    Funds
             Total Funds          17.0         20.2          16.5         15.6          56.6         95.1


    We select equity funds according to the following:
    1. To avoid Survivorship Bias, we select all equity funds that existed in year 2000. Since the
definition of an EV requires that the fund last a specified duration, the actual period for the
empirical test is 6 years, from year 2001 to 2006.
    2. Among equity funds that existed since 2000, for performance comparison, we chose funds
according to the following conditions. Active, domestic, pure equity type,5 and public placing
funds were included. Index, foreign, mixed and private placing funds were excluded. Some
sector funds were included.6
    3. After checking the data of each fund, we excluded funds that had errors or missing data.

5
  In Korea, an equity fund is defined to have an equity ratio higher than 60% and no bonds. This
classification criteria was adopted in 2000.
6
  Several kinds of sector funds are being developed in Korea. However, instead of using a sector specific
benchmark, sector funds use the KOSPI as a benchmark like other ordinary equity funds. In reality, there
are many cases where sector funds are managed as ordinary funds. For this reason, we include sector
funds.


                                                                                                        6

## Page 7

After selection, 518 funds were selected for analysis.
EVs have the following criteria. EVs last no shorter than 1 year and are no less than KRW 10bn7
in AUM. This criterion is commonly accepted in the market, though it varies more or less by
agency.8 Funds with less than 1 year duration are early funds and funds with less than KRW
10bn AUM and duration longer than 1 year are NEVs.
    For whole period, the number of EVs is about a third of whole sample. Looking just at AUM
amount, EVs handle a majority of the funds invested, while NEVs handle only about 5% of the
funds. (Table 1) The percentage of EVs is very different each year. In terms of AUM, while the
fraction of EVs in 2003 & 2006 is around 90%, in 2004 & 2005 its percentage was around 60%.
This implies that performance evaluation only with EVs will bias the evaluation of fund family
capabilities.




    Ⅴ. Evaluation-ship Bias

1. Performance Comparison (1): EV Portfolio vs. NEV Portfolio


    We compose an EV portfolio and a NEV portfolio in the following way and then compare
their performances.
    First, ex post, we analyzed the performance of the EVs & NEVs by forming each portfolio in
the following way. Based on the AUM at the end of every year, EVs & NEVs are classified.
Then, assuming that each fund is invested with equal weight, we compose EV portfolio and NEV
portfolio. Return of each portfolio is calculated on monthly base.
    Risk-adjusted return based on CAPM Model and Fama & French Model (1993)(hereafter “FF
Model”) (Refer to formula <1> & <2>) will be used as the performance measurement.9 KOSPI
return will be the proxy for market return ( Rmt ) and the 3 month CD rate will be the proxy for
risk-free return ( R ft ).10 The size effect in the FF Model is the difference between large capital
returns and small & mid capital returns from the KOPSI Index. Also, due to a lack of accounting


7
   As of the end of 2006, it amounts to US$ 10.8Million.
8
   These criteria are used by fund rating agencies as well as the distributor and other financial institutions
to choose funds and asset management companies.
9
   Hereafter, without specific comment, return is FF Model based.
10
    TB rate may be more appropriate as the proxy of risk-free rate. But, considering the character of bond
market, CD is more liquid and regarded as risk-free rate in the short term market. CD rate is also used for
calculation of KTB future price.


                                                                                                             7

## Page 8

data for Korean firms, the difference in return between value stocks and growth stocks was
imputed with the difference between the KOSPI and KOSDAQ returns. The KOSDAQ market
is regarded as the representative of growth stocks being composed of mainly IT sector stocks, in
contrast with KOSPI market.


     Rit - R ft = α i1 + β mi
                           1
                              × ( Rmt - R ft )           ---------------------------------------------- (1)
     Rit - R ft = α i2 + β mi2 ×( Rmt - R ft )+ β si2 × ( Rst - Rlt ) + β vi2 × ( Rqt - Rmt ) ------- ---- (2)
           Rit : fund return (month to month), Rmt : KOSPI Return (monthly),
           R ft : 3 month CD rate, Rst : Returns of small & mid cap. stocks (monthly)
           Rlt : Returns of large capital stocks (monthly),
           Rqt : KOSDAQ Return (monthly),


     The Seemingly Unrelated Regression method was used to compare EV portfolio with NEV
portfolio (refer to <Table 2>). Both CAPM and FF models significantly explained fund
performance. Among the explanatory variables, market risk premiums and KOSDAQ and
KOSPI return difference have significant influences on performance.
     The results show that the EV portfolio out-performs the NEV portfolio. Although not
significant, the CAPM model shows that EV portfolio has an excess return amounting to
0.398% which is higher than NEV portfolio which amounts to 0.298% by 10bps. In the FF
model, EV portfolio out-performs NEV’ at a 10% significant level, having a return that is
0.349%, opposed to the 0.235% of NEV portfolio. These results demonstrate the existence of
evaluation-ship bias in Korean equity funds.
     We reclassified the funds ex-ante. EV portfolio and NEV portfolio are composed, based on the
AUM at the first of every year. Also, we did the same analysis with a monthly rebalancing of
the portfolios.
     The results11 show that there is no significant difference between the performance of EVs &
NEVs in both CAPM and FF model with the yearly classification. It is likewise with monthly
rebalancing. As it were, it is not useful for better fund investment whether a fund is EV or not.




11
  On yearly base, EVs out-perform by 6.5bps and 8.1bps in CAPM and FF models, respectively. On a
monthly base, EVs out-performs NEVs by 3.3bps and 1.9bps in the CAPM and FF models, respectively.


                                                                                                                 8

## Page 9

Table 2. Comparison of EVs’and NEVs’performances
                                              CAPM Model                                                FF Model
                              Coef.          Std. Err.        t-stat.   p-value      Coef.          Std. Err.      t-stat.   p-value

        (EVs)

      Consant              .3982981***      .1286499          3.10       0.002    .3494297***      .1289168         2.71     0.007
     ( R mt − R   ft   )   .8751151***      .0172813          50.64      0.000    .8871518***      .0182245        48.68     0.000

      ( R st − R lt )                                                             -.0185027        .0234996        -0.79     0.431


     ( R qt − R mt )                                                              -.0346511*       .0198413        -1.75     0.081

                                      R-Square: 0.9641           n=72                        R-Square: 0.9738         n=72
       (NEVs)
      Consant              .2980095**       .1431264          2.08       0.037    .2346252*        .142331          1.65     0.099
     ( R mt − R ft )       .8204706***      .0192259          42.68      0.000    .8360063***      .0201208        41.55     0.000

      ( R st − R lt )                                                             -.018401         .0259448        -0.71     0.478

     ( R qt − R mt )                                                              -.0453778**      .0219059        -2.07     0.038
                                      R-Square: 0.9620           n=72                        R-Square: 0.9727         n=72
     α iE = α iNE              Chi-Square: 2.16             P-Value: 0.1420            Chi-Square: 2.74         P-Value: 0.0980

***, **, * is significant in 1%, 5%, 10% confidence level




                                                                                                                                       9

## Page 10

2. Performance Comparison (2): Panel Data Analysis


     The above analysis failed to consider two important factors. Some funds can switch from EV
to NEV in one year, vice versa. In these cases, fund family is not expected to have consistent
attitude toward those funds. Also the above test did not control fund size. Since the
classification into EVs and NEVs is based on differing AUM sizes, the above results may come
from a fund size effect
     To take these considerations into account, we compare fund performance using panel data.
For considering the attitude of fund family, the switched funds were excluded from the selection
for samples. Only the funds which have continually survived as EV or NEV for one year were
selected.
     In result, we set up year t return for fund i as a basic unit. If the AUM of every month at year t
for fund i is not less than KRW 10bn, fund i at year t is an EV. On the other hand, if the AUM of
every month is less than KRW 10bn for year t, fund i is a NEV. We calculated the excess returns
for the EVs and NEVs at each year t by using formula <1> & <2>12.
     The calculated excess returns, ait, is the dependent variable. Then we compare EV with NEV
performance using estimation formula <3>. As suggested above, we include fund size as a
control variable in the model. Performance difference is estimated by a dummy variable which
identifies EVs and NEVs. As the influence of fund size may differ according to whether a fund
is an EV or not, we include both an EV dummy variables ( DitE ) and a NEV dummy variables
( DitNE ) multiplied by fund size in model. In addition, the net fund flow change and risk of fund
return are also included as explanatory variables.


     α it = γ1 Flowit + γ 2 DitE × LNSizeit + γ3 DitNE × LNSizeit + γ 4 STDit + γ5 DitE ------- (3)
             LNSizeit : Log value of fund size
              Flowit = ( Sizeit - Sizeit −1 (1+ Rit )) / Sizeit −1 : Net fund flow change,
              Sizeit : fund size, Rit : return of fund (year to year)
              STDit = risk of fund return (= standard deviation of monthly returns)
              DitE : EV Dummy (value is 1 if a fund is EV, 0 if NEV)
              DitNE : NEV Dummy (value is 1 if a fund is NEV, 0 if EV)


12
   To decrease estimation error, the coefficients of CAPM and FF Model are estimated by using monthly
returns and then excess returns are calculated.


                                                                                                      10

## Page 11

Table 3. Determinants of Performances


                                           CAPM Model                                               FM Model
                           Coef.         Std. Err.          t-stat.   p-value      Coef.          Std. Err.    t-stat.   p-value
Consant                -0.2505              0.1013            -2.47     0.013   -0.1964*            0.1098       -1.79     0.074
Flow it                 0.0020              0.0025             0.81     0.419   -0.0002             0.0029       -0.06     0.949
            E
LNSize     it          -0.0073              0.0052            -1.40     0.160   -0.0089             0.0056       -1.57     0.116
            NE
LNSize     it           0.0085              0.0051             1.64     0.101   0.0061              0.0055        1.11     0.267
STD   it                5.7052***           1.1429             4.99     0.001   0.4893***           0.9147        6.00     0.000
Dummyit                 0.3619**            0.1545             2.34     0.019   0.3583**            0.1647        2.18     0.030
            E
LNSize     it
                  NE
 = LNSize        it         Chi-Square: 5.46         P-Value: 0.0195                      Chi-Square: 4.39     P-Value: 0.0361

***, **, * is significant in 1%, 5%, 10% confidence level




                                                                                                                                 11

## Page 12

Because the panel data is unbalanced and the dummy variables are time-constant covariates,
the fixed-effect method is not applicable in such cases. Instead, we use random-effect GLS
Panel Estimation Method.
     The estimation result (table 3) shows that EVs outperformed NEVs. The coefficient of the EV
dummy is 0.358 which is significant at a 5% confidence level13. The better EV performance
cannot be attributed to a fund size effect. The coefficients of the fund size effects, γ 2 and γ3 ,
are not significant. Rather, the signs of γ 2 and γ3 are interesting because they are opposite.
EVs increase in performance as the fund size decreases while NEV increase in performance as
the fund size grows.




     Ⅵ. Favoritism: Fund families

     Irrespective of EV performance, we will show that fund families favor EVs in their
management process. There are two ways fund families favor EVs. First, fund families prioritize
EVs when they can apply strategies to their funds. Second, fund families are more sensitive and
responsive to EV performance, especially when a fund’s performance ranks very low.


     1. Favoritism in the Application of Strategies


     Hypothesis 1: The better a fund family’s overall performance, the larger the difference
between EV and NEV performance for the same fund family.


     <Hypothesis1> means that fund families favor EVs when applying their investment strategies.
If a fund family behaves in this way, EV performance would be better than NEV performance
when an fund family’s strategy succeeds, while the probability of high performance decreases
when the strategy fails. In this context, <Hypothesis 1> can demonstrate this favoritism.



13
   The out-performance may come from sample bias. It can be interpreted that, among EV, the funds
which have persisted in EV character for 1 year would not have worse performance for the cash to go out
of. In short, the sample can exclude worst cases in performance. On the contrary, from NEV sample, the
best performing funds may be excluded. To test the sample bias, after composing the portfolios using two
samples, we tested the performance difference between two portfolios. The result is that EVs
outperformed NEVs by 4.3bps and 2.3bps, which is not significant. This result is not different from the
previous result using all the samples. Therefore, the sample bias does not matter.


                                                                                                     12

## Page 13

α tk, Diff = γ1 α tk + γ 2 ( RHt - RLt ) + γ3 SizeRatioitk         ----------------------- (4)


        α tk : performance (weighted average of returns) of fund family k at t year
        α tk,e : weighted average of returns of EVs by fund family k
        α tk,u : weighted average of returns of NEVs by fund family k
        α tk, Diff (= α tk,e - α tk,u )
        SizeRatioitk : the ratio of NEV size to EV size managed by fund family k
                               (as of the end of year t)


  To test <Hypothesis 1>, we use estimation model <formula 4>. The overall performance of
each fund family k at year t, ( α tk ) is the weighted average of returns of all funds managed by
fund family k. α tk, Diff is defined as the performance gap between the EV & NEV of fund family
k. We focused on whether the overall performance ( α tk ) explains the performance gap ( α tk, Diff ).
In the model, we also include two explanatory variables, momentum effect and the ratio of NEV
size to EV size. The momentum effect ( RHt - R Lt ) is based on Carhart (1997). The ratio
( SizeRatioitk ) is included to control the fund families’ different responses to EV weight
compared to the total AUM.
  Based on the sign of excess return α tk , we divided fund families into good companies that
have positive returns and bad companies that have negative returns. Among 38 companies in
total, there were 23 good companies and 15 bad companies(refer to <table 4>. EV return was
higher than NEV return by 2.5% in good company samples while EV returns were higher than
NEVs by only 0.2% in bad company samples. The difference between the EV and NEV returns
is significant at a 10% confidence level.




         Table 4. Good Companies’ and Bad Companies’ Performances (FF Model)
                         N.of obs.           Average        Std. Err.        minimum           maximum
     Good                     23            0.0251547      0.0502972         0.0701495         0.1273517
  Comapny
     Bad                      15            0.0022552      0.0473546        -0.0904863         0.1016852
  Company




                                                                                                         13

## Page 14

The results of estimation <formula 4> confirms hypothesis 1 (refer to <table 5>. The overall
performance of fund families explains the better performance of EVs at a 1% confidence level.
As hypothesized, the higher the overall performance of all the fund families, the better the
performance of EVs.


        Table 5. fund families’ performances & EVs’ out-performances (FF Model)
                            Coef.            Robust s.e.          t-statistics       p-value
    Consant               -.005134            .0206216              -0.25             0.803
            k
        α   t             .314373***          .0759449                4.14            0.000
     RHt - RLt            -.040563            .0628657              -0.65             0.519
   SizeRatiotk            -.026067            .0987822              -0.26             0.792
***, **, * is significant in 1%, 5%, 10% confidence level


2. Differentiated Response to Performance


Hypothesis 2: For all EV and NEV funds with the lowest performance in previous period, the
portfolios of EV funds are adjusted to take higher risk than the portfolios of NEV funds.


  We set up the hypothesis 2 to confirm that fund families are more active in composing EV
portfolios. Previous studies show that fund managers will take more risk if performance of
previous period is poor. We assume that the attitudes towards EVs and NEVs differ. The
portfolio risk in EVs should be changed higher than the portfolio risk in NEVs, when previous
performances are poor. This should be the case if fund families are more responsive to EV
performance then to NEV performance.
  To test hypothesis 2, we use the following method. The copy ratio reflects the attitude
towards market risk. The higher the risk-taking, the lesser the copy ratio. Therefore, we tested
how much the copy ratio of EVs and NEVs changes, especially for funds that perform the
lowest. We can infer that when an fund family is more attentive to a particular EV fund, the
portfolio of the EV fund is changed to take risk more than its NEV counterpart.
  We could not calculate the copy ratio directly; thus, we define the non-copy ratio to be the
monthly averaging the absolute values of the relative returns to a benchmark and used it a proxy
(refer to <formula 5>).


                                                                                               14

## Page 15

12
                     NonCopy it =      ∑( R
                                       m =1
                                               m
                                              it   − Rmtm ) / 12     ------------------------------ (5)


  Fund rankings are grouped in quintiles. Using the method used by Sirri & Tufano (1998), we
ranked each fund i, based on the performance of the whole funds at year (t-1) (refer to <table
6>). Then, by <formula 6>, we analyzed the relation between the changing rate of the non-copy
ratio and the previous year’s ranked value.


∆NonCopy it = γ1 LowPerf it −1 + γ 2 LoMidPerf it −1 + γ3 MidPerf it −1 + γ 4 UpMidPerf it −1
                     + γ5 HighPerf it −1 + γ 6 LNSizeit −1         -------------------------------- (6)
              where ∆NonCopy it = NonCopy it - NonCopy it −1


  The estimation result shows that the response to performance is different between EVs and
NEVs (refer to <table 6>). In case of EV, the non-copy ratio of the lowest performance funds
significantly increased. The coefficients of other ranked groups were negative. The better the
performance, the less risk is taken.
  On the contrary, we can not find any relation between ranking and the change of portfolio in
NEVs. Even with the lowest performing NEVs, there is no evidence that fund families change
these portfolios actively.




Ⅶ. Evaluation-ship bias and fund flow

   Why do fund families favor Evaluated Funds? We show that, in terms of fund flows, there
are incentives. First, in the same conditions, the inflows into EVs are larger than the inflows into
NEVs. Second, the better EVs’ performance contributes to the total inflow of funds into the asset
management sector.
  To test the difference of fund flows between EVs and NEVs, we used <formula 7> as
estimation model.


               Flowit = γ1 LNSizeit −1 + γ 2 Feei + γ3 ait −1 + γ 4 DitE ------------------- (7)




                                                                                                          15

## Page 16

Table 6. Ranking of previous year’s performances and the change of portfolios


                                               EVs                                                        NEVs

                        CAPM Model                          FF Model                    CAPM Model                  FF Model

Consant                .0269384(0.76)                .0350211(0.91)                   .0050543( 0.42)            -.0004577(-0.04)

LowPerf it −1          .1184856***(3.63)             .132022***(2.72)             -.0295745(-0.96)               .0087809( 0.35)
LoMidPerf it −1       -.0026302(-0.11)               -.0292267(-0.69)                 .0405581( 2.06)**          .0018513( 0.10)
MidPerfit −1          -.0060681(-0.27)               -.008175(-0.23)              -.0108691(-0.57)               .0113179( 0.64)
UpMidPerfit −1        -.055256**(-2.13)              -.0056055(-0.18)                 .0017020( 0.09)            -.0050274(-0.28)
HighPerfit −1         -.0612191(-0.93)               -.00705(-0.14)               -.0089754(-0.54)               -.0091600(-0.48)
LNSizeit −1           -.0019616(-1.46)               -.0023132(-1.55)             -.0002881(-0.53)               -.0001830(-0.34)

R-Square                      0.2970                           0.1990                      0.0863                      0.0315

  Low Performance Quintile(“LowPerf”) = min {Rankt-1, 0.2}

  Lower-Middle Performance Quintile(“LoMidPerf”) = min {0.2, Rankt-1- LowPerf}

  Middle Performance Quintile(“MidPerf”) = min {0.2, Rankt-1- (LowPerf+ LoMidPerf)}

  Upper-Middle Performance Quintile(“UpMidPerf”)

    = min {0.2, Rankt-1- (LowPerf + LoMidPerf + MidPerf)}

  High Performance Quintile(“HighPerf”)

     = {Rankt-1- (LowPerf + LoMidPerf + MidPerf+ UpMidPerf)}

***, **, * is significant in 1%, 5%, 10% confidence level



                                                                                                                                    16

## Page 17

Results are shown in <table 7>. The coefficient of the EV dummy variable is significantly
positive at a 5% confidence level. This suggests that, ceteris paribus, inflows into EVs are
higher than inflows into NEVs. The main reason is because EVs benefit from advertisement, as
Jain & Wu(2000) pointed out – an advertisement effect.


                      Table 7. Comparison between EVs’and NEVs’Fund Flows
                                     CAPM Model                                       FF Model
                              Coef.               t-statistics            Coef.                 t-statistics
      Consant               3.434354                  1.51             3.423491                     1.39
     LNSizeit −1            -.135433*                -1.67             -.1374652                  -1.53
       Feei                 -.5111077                -1.21             -.4987589                  -1.20
       ait −1               -.5454873                -0.29             1.53838                      0.17
       DitE                 1.573465**               2.07              1.460536**                   2.02
      R-Square                           0.0941                                        0.0935
  ***, **, * is significant in 1%, 5%, 10% confidence level


  In addition, if a spill-over effect also exists, there is a higher incentive for favoritism. For this
test, we used model <formula 8> to analyze the relation between EV performance and net fund
inflow into fund family.


       Flowtk = γ1 LNSizeitk −1 + γ 2 α tk−1, Diff     ------------------------ (8)
       Flow =k
             t         ∑ Flow
                        i
                                k
                                it   :   total net fund inflow of fund family k at year t
                                         (including below 1 year fund)
                 k
        LNSize   it   : Log value of AUM of fund family k at year t


  The estimation result of Table 8 implies that the performance of EVs contributed to the
increase in total AUM. The positive performance of the EVs is positively related with the total
net cash flows, which is significant at 10% confidence level in the FF model.




                                                                                                               17

## Page 18

Table 8. Fund families’ Fund Flows and EVs’ Out-performance
                                     CAPM Model                              FF Model
                                 Coef.             t-statistics      Coef.             t-statistics
      Consant                 -48.60068**            -2.15        -51.64524**            -2.08
                        k
      LNSize            i      1.918516**              2.17        2.037922**              2.08
           k
       a   t −1, diff           4.53976               1.07        26.23179*               1.75
      R-Square                            0.4385                              0.4689
   ***, **, * is significant in 1%, 5%, 10% confidence level




Ⅷ. Conclusions

  This study analyzes evaluation-ship bias caused by the fund evaluation system. This research
 shows how fund families give favoritism to evaluated funds. Our empirical results of Korean
 equity funds are summarized as follows:
   First, we verified the existence of evaluation-ship bias. In the empirical test, EV portfolios
 performed better than NEV portfolios. Thus it is possible that an fund family’s performance
 may be over-estimated.
   Second, there are incentives for fund families to favor evaluated funds. They are more
 attentive to EV performance and take strategic and risk-taking measures to improve its ratings.
  This paper contributes to recent studies that focus on the discriminating behaviors of fund
families. Adding to previous literature, this paper suggests that another factor, the performance
evaluation system itself, is what causes fund families to favor some funds. The existence of such
a high number of Korean funds is an example of this evaluation system bias. We can expect
similar phenomena to occur in other countries as well.
  Our results may not be decisive, owing to the lack of time series data. However, the empirical
results we have demonstrate the need to avoid information bias in Korea. We hope that the
introduction of GIPS will diminish the evaluation system bias and contribute to the development
of the Korean fund market.




                                                                                                      18

## Page 19

<References>

Brown Keith C., W.V. Harlow, Laura T. Starks(1996), “Of Tournaments and Temptations: An
         Analysis of Managerial Incentives in the Mutual Fund Industry”, Journal of Finance,
         Vol.51, No.1, pp. 85-110
Brown, Stephen J., William Goetzmann, Roger G. Ibbotson, Stephen A. Ross(1992),
         “Survivorship Bias in Performance Studies”,    Review of Financial Studies, Vol. 5, No.
         5, pp. 553-580
Carhart, Mark M.(1997), “On Persistence in Mutual Fund Performance”, Journal of Finance,
         Vol.52, No.1, pp. 57-82
Carhart, Mark M., Jennifer N. Carpenter, Anthony W. Lynch, David K. Musto(2002), “Mutual
         Fund Survivorship”, Review of   Financial Studies, Vol.15, No.5, pp. 1439-1463
Chen, Joseph, Harrison Hong, Ming Huang, Jeffrey D. Kubik(2004), “Does Fund Size Erode
         Mutual Fund Performance? The Role of Liquidity and Organization”, American
         Economic Review, Vol.94, Np.5, pp.1276-1302
Chevalier, Judith, Glenn Ellison(1997), “Risk Taking by Mutual Funds as a Response to
         Incentive”, Journal of Political Economy, Vol.105, No.6, pp.1167-1200
Elton, Edwin J., Martin J. Gruber, Christopher R. Blake(1996), “Survivorship Bias and Mutual
         Fund Performance”, Review of    Financial Studies, Vol.9, No.4, pp. 1097-1120
Elton, Edwin J., Martin J. Gruber, Jeffrey A. Busse(2004), “Are Investors Rational? Choices
         among Index Funds”, Journal of Finance, Vol.59, No.1, pp. 261-288
Frye, Melissa B.(2001), "The Performance of Bank-Managed Mutual Funds", The Journal of
         Financial Research, Vol.24, No.3, pp.419-442
Fama, Eugene F., Kenneth R. French(1993), “Common Risk Factors in the Returns on Stocks
         and Bonds”, Journal of Financial Economics, 33, pp.3-56
Fama, Eugene F., Kenneth R. French(1996), “Multifactor Explanations of Asset Pricing
         Anomalies”, Journal of Finance, Vol.51, No.1, pp.55-84
Gallaher, Stephen, Ron Kaniel, Laura Starks (2006), “Madison Meets Wall Streets: Mutual Fund
         Families, Competition and Advertising”, Working Paper
Gaspar, José-Miguel, Massimo Massa, Pedro Matos(2006), “Favoritism in Mutual Fund
         Families? Evidence on Strategic Cross-Fund Subsidization”, Journal of Finance,
         Vol.61, No.1, pp. 73-104
Gruber, Martin J.(1996), “Another Puzzle: The Growth in Actively Managed Mutual Funds”,
         Journal of Finance, Vol.51, No.3, pp.783-810
Hendricks, Darryll, Jayendu Patel, Richard Zeckhauser(1993), “Hot hands in Mutual Fund:
         Short-run Persistence of Relative Performance, 1974-1988”, Journal of Finance,



                                                                                            19

## Page 20

Vol.48, No.1, pp.93-130
Huang, Jennifer, Kesley D. Wei, Hong Yan(2007), “Participation Costs and the Sensitivity of
         Fund Flows to Past Performance”, Journal of Finance, Vol.62, No.3, pp.1273-1311
Jain Prem C., Joanna Shuang Wu(2000), “Truth in Mutual Fund Advertising: Evidence on
         Future Performance and Fund Flow”, Journal of Finance, Vol.55, No.2, pp. 937-958
Jensen, Michael C.(1967), “ The Performance of Mutual Funds in the Period 1945-1964”,
         Journal of Finance, Vol.23, No.2, pp. 389-416
Kempf, Alexander, Stefan Ruenzi (2004), “Family Matters: The Performance Flow Relationship
         in the Mutual Fund Industry”, Working Paper
Lynch Anthony W., David K. Musto(2003), “How Investors Interpreter Past Fund Returns”,
         Journal of Finance, Vol.58, No.5, pp. 2033-2058
Massa, Massimo (2003), “How do Family Strategies affect fund performance? When
         Performance-Maximization in not the only game in town”, Journal of Financial
         Economics, 67, pp.249-304
Malkiel, Burton G.(1995), “Returns from Investing in Equity Mutual Funds 1971 to 1991”,
         Journal of Finance, Vol.50, No.2, pp. 549-572
Nanda, Vikram, Z.Jay Wang, Lu Zheng(2004), “Family Values and the Star Phenomenon:
         Strategies of Mutual Fund Families”, The Review of Financial Studies, Vol.17, No.3,
         pp.667-698
Sirri, Erik R., Peter Tufano(1998), “ Costly Research and Mutual Fund Flows”, Journal of
         Finance, Vol.53, No.5, pp. 1589-1622
Wermers, Russ(2000), “Mutual Fund Performance: An Empirical Decomposition into Stock-
         Picking Talent, Style, Transactions Costs, and Expense”, Journal of Finance, Vol.55,
         No.4, pp. 1655-1695




                                                                                           20

