"N v“ I
.3 13‘ ,0‘
39%;?

3

mix"?

. .V‘
,, -mr
..
rzmvw

“WW

p-vl w..."
M-~I

 

. »,

FEM;

i ,

"‘}..:1t:‘ “' f
.x ., . 5,; 2w. Q -7. .,
3‘3; ”i "1' '2‘ ‘2‘ 1 $31 ' :43

 

"mt-x 5.3:. ' .
‘ ‘ 4 .J
.
V W .
M
‘Jé'
..
J35 $1..
“273.;

-rl-
.. 4n-

4
.,
A . ,.
3:."
_.
‘ *9.
.... :- .2»

m
‘32-. .
in
- .2.
2:)?

.¢
w

.

. z in g \.
I -' 1:; w; " .1 :h.‘ ‘ -..=,=:='le’.z. n:
“I? "37“ " '- "“" *i“ “ ‘7“S'EiE2‘-*:‘-iu§z$‘zn‘:’ .
.; . 'I “a - k M (Ma; 1am ' try . ‘ x a“ ‘ ., 41
,g ; Am, 4.4”)”“413 :14. 1, ‘ , , . gam'fi‘ a:
Wing-J“. 1 ‘ ‘ "x" 9.; 3 , . , 2r, €n’3?$;;.%
‘ ‘ I“ W ' " ' yr: " > ‘ 1233:!

$

 

mes-s

G1 llllllllll\Illllllllllllllllllllllll‘lllllllll

3 1293 0141

This is to certify that the

dissertation entitled
3/5 (ff/3f «7 /v’(“/ 0 f /'/L/ r (VA/~17, art/xi L,

m a 7'0 4 z, fumJS 7i" c m/MC / {I} A /'~ vi)" fl (”W/"3U
presented by
")7 . /?/9 M34 4’9”? )6 fag/Q

has been accepted towards fulfillment
of the requirements for

P; . /) v degree in ///U’,4/v CC:

 

112 719— - évAzzfl

Major aprofesZarl/

Date W 7’, /¢%

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

LIBRARY
Mlchlgan State
Unlverslty

 

 

 

PLACE IN RETURN BOX to remove thle checkout from your record.

TO AVOID FINES return on or bdore date due.

DATE DUE DATE DUE DATE DU

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU leAn Affirmative ActioNEmel Opportunity Inetltwon
W

m1

 

THE EFFICIENCY OF INTERNATIONAL MUTUAL FUNDS: AN EMPIRICAL
INVESTIGATION

By

Miranda Lam Detzler

A DISSERTATION

Submitted to Michigan State University in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Finance and Insurance

1996

ABSTRACT

THE EFFICIENCY OF INTERNATIONAL MUTUAL FUNDS: AN EMPIRICAL
INVESTIGATION

By

Miranda Lam Detzler

This dissertation tests the performance of actively managed international
mutual funds against several global benchmarks that represent viable alternative
investment strategies. Since open-end mutual funds cannot generally be sold
short, investors can only profit from superior performance. Two-sided joint tests
used in previous studies reject the null hypothesis too often because both
superior and inferior performance contribute to rejection. I apply a one-sided
multivariate test which recognizes the short-sale restriction to evaluate joint fund

performance.

I find that the MSCI world index is not mean-variance efficient compared
to 12 unmanaged country equity indices during the January 1985-March 1994
period. Even though the 35 international equity funds jointly outperform the

MSCI world index, beating an inefficient benchmark does not imply managers

have superior investment skills. I use a 3-region equity benchmark to test their
abilities to select securities and / or identify outperforming countries outside the
US. and Japan. Then I use a 12-country equity benchmark to isolate managers’
security selectivity ability. I do not find any evidence of security or country
selectivity abilities. However, the funds outperform the Wilshire 5000 index,
confirming that a passive US. investor can benefit by adding international equity
funds to their domestic portfolio. I use the SB World Government Bond index as
the benchmark for evaluating 18 international bond funds from January 1989-
March 1994. The SB world bond index is mean-variance efficient versus 11
unmanaged country bond indices and again I find no evidence of superior
performance among fund managers. Surprisingly, the international bond funds
cannot beat the SB BroadTM Index, implying that adding these funds to a
diversified domestic bond portfolio over my sample period does not lead to a

higher Sharpe ratio.

I examine the effects of currency hedging in fund performance using three
forward contracts, Deutschemark, Japanese Yen, and Canadian Dollar. My
qualitative conclusions remain the same when the forward contracts are added to
the benchmarks to control for hedging activities. I also compute the robust

Positive Period Weight and the Treynor-Mazuy measures and they are very

similar to the Jensen measures for the same benchmark, confirming that my
conclusions are not affected by nonlinearity in fund returns or timing activities of

managers.

To my parents for their love, trust and support.

ACKNOWLEDGMENTS

I wish to thank Dr. James B. Wiggins, Chair of my Dissertation Guidance
Committee, for his guidance and valuable suggestions in the preparation of this
dissertation. I also wish to thank the other members of the committee, Dr. Kirt C.
Butler and Dr. Ching-fan Chung for their assistance. In addition, I wish to thank
Dr. Richard R. Simonds and the Center for International Business Education and
Research for their support in acquiring the data for this study. Finally I wish to
thank my husband, Roger, for his loving support and technical assistance.

vi

TABLE OF CONTENTS

1. Introduction ................................................................................................................... 1
2. Literature Review .......................................................................................................... 6
2.1 International Capital Asset Pricing Models ........................................................ 6
2.2 Mutual Fund Performance Measures .................................................................. 9
2.3 Performance Persistence ...................................................................................... 14
2.4 International Mutual Fund Performance Evaluation ..... I ................................. 15
3. Methodology ............................................................................................................... 17
3.1 A Mean-variance Spanning Test with a Short-sale Constraint ...................... 17
3.2 Measuring Timing Ability ................................................................................... 24
3.3 The Positive Period Weight Measure ................................................................. 26
3.4 Measuring Performance Persistence .................................................................. 30
4. Data ............................................................................................................................... 31
4.1 Monthly Returns on Mutual Funds and Indices .............................................. 31
4.2 Currency Hedging ................................................................................................ 33
5. Performance Tests of International Equity Funds .................................................. 36
5.1 Hypotheses and Benchmarks .............................................................................. 36

5.1.1 Hypothesis 1 - International mutual funds do not exhibit security
selectivity ability. .......................................................................................... 37

5.1.2 Hypothesis 2 - International mutual funds do not provide efficient
global diversification. ................................................................................... 38

5.1.3 Hypothesis 3 - International mutual funds do not exhibit country
selectivity ability. .......................................................................................... 39

vii

viii

5.2 Transaction Costs .................................................................................................. 41
5.3 Empirical Results of International Equity Mutual Funds ............................... 43
5.3.1 Mean-variance Performance Tests ............................................................... 43
5.3.2 Positive Period Weight Measures ................................................................ 46
5.3.3 Nonlinearity Between Fund and Benchmark Returns .............................. 48
5.4 Performance Tests Results with Currency Hedged Equity Benchmarks ..... 49
5.5 Performance Persistence ...................................................................................... 51
5.6 Survivorship .......................................................................................................... 52
5.7 Summary of Empirical Tests on Equity Funds ................................................. 53
6. Performance Tests of International Bond Mutual Funds ...................................... 54
6.1 Hypotheses and Benchmarks for International Bond Funds ......................... 56

6.1.1 Hypothesis 4 - International bond fund managers do not have security
selectivity ability. .......................................................................................... 56

6.1.2 Hypothesis 5 - International bond funds do not provide efficient global
diversification. ............................................................................................... 58
6.2 Empirical Results of International Bond Funds ............................................... 58
6.2.1 Mean-variance Tests ...................................................................................... 58
6.2.2 PPW Tests ........................................................................................................ 60
6.2.3 Market Timing and Overall Performance .................................................. 61
6.3 Performance Tests Results with Currency Hedged Bond Benchmarks ........ 62
6.4 Performance Persistence ...................................................................................... 66
6.5 Discontinued Funds and Survivorship Bias ..................................................... 67
6.6 Summary of Empirical Tests on Bond Fund Performance ............................. 68
6.7 Balanced Benchmarks ........................................................................................... 70

7. Conclusions and Future Research ............................................................................ 73

ix
8. Appendices ................................................................................................................ 108

9. List of References ...................................................................................................... 116

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 11

Table 12

Table 13

Table 14

LIST OF TABLES

Summary Statistics: Monthly Excess Returns in US$ (%)
Tests of the Mean-variance Efficiency of the Equity Benchmarks
Transaction Costs Estimation

International Equity Mutual Fund Performance: Mean-variance
Tests

International Equity Mutual Fund Performance: PPW Measures

International Equity Mutual Fund Performance: Treynor-Mazuy
Measures

International Equity Mutual Fund Performance: Simultaneous
Hedging Strategy

International Equity Mutual Fund Performance: Unitary Hedging
Strategy

International Equity Mutual Fund Performance Persistence

Tests of the Mean-variance Efficiency of the World Bond Index
International Bond Mutual Fund Performance: Mean-variance Tests
Determinants of International Bond Fund Returns

International Bond Mutual Fund Performance: PPW Measures

International Bond Mutual Funds: Comparison of Performance
Measures

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

xi
International Bond Mutual Fund Performance: Simultaneous
Hedging Strategy

International Bond Mutual Fund Performance: Unitary Hedged
World Bond Index

International Bond Mutual Fund Performance Persistence
Tests of the Mean-variance Efficiency of the Balanced Benchmarks
International Mutual Fund Performance: Mean-variance Tests

International Mutual Fund Performance: PPW Measures

1. Introduction

Many researchers (50le (1974), Eun and Resnick (1984) and Grauer and
Hakansson (1987)) have documented the benefits of a globally diversified
portfolio, yet only recently have US. investors become interested in investing
overseas. International mutual funds provide US. investors easy access to
investment opportunities abroad without direct dealings with foreign markets.
With total assets increasing from $3.5 billion in 1985 to $108.9 billion in 1993,
international equity funds represent one of the fastest growing segments of the
industry (Wiesenberger Investment Companies Yearbook). Since global investment
has gained popularity only in recent times, relatively few international funds
operated long enough to be included in previous performance studies. Cumby
and Glen (1990) and Eun, Kolodny and Resnick (1991) evaluate 15 and 13 equity
funds respectively and there exists no rigorous study on the performance of

international bond funds.

This dissertation fills this gap by empirically estimating the value
provided by active international fund management. In addition, it makes several
contributions to the mutual fund evaluation literature. First, I take into account
the practical prohibition against short selling open-end funds when testing joint

fund performance. Traditionally, mutual fund studies (Ippolito 1992) focus on
1

2

the performance of individual funds. However, simply measuring individual
fund performance does not provide any insight on the optimal investment
strategy for a passive investor. Out of a sample of 30 to 50 funds, one would
expect to find a few funds exhibiting significant performance purely by chance.
To conclude that active management adds value to a passive indexing strategy,
the funds must jointly outperform the benchmark index. Cumby and Glen (1990)
test the joint performance of their funds using a chi-square test. Their approach
is asymptotically equivalent to the mean-variance test of Gibbons, Ross and
Shanken (GRS, 1989). The GRS test rejects benchmark efficiency if adding
mutual funds in unrestricted proportions to the benchmark produces a
significant increase in the Sharpe (1966) ratio. Since open-end mutual funds
cannot generally be sold short by investors, I incorporate a short-sale constraint
into the GRS test. The short-sale constrained test rejects benchmark efficiency
less frequently than the GRS test because underperforming funds no longer
contribute to a higher Sharpe ratio. Investors cannot profit from inferior
performance under the short-sale restriction because 1) portfolio holdings of
mutual funds are not available in a timely manner, and 2) inferior performance

may result from excessive operating expenses or management fees.

3

Secondly, I estimate transaction costs for the indexing strategies and
account for them explicitly in the performance tests. Mutual fund returns are
reported net of operating expenses. When evaluating fund performance,
previous studies use total return on a benchmark which do not include the
transaction costs involved to invest in the benchmark. Ignoring transaction costs

of the indexing strategy creates a bias against the mutual funds.

Even if funds do not jointly outperform the benchmark index, investors
can still profit if they can identify funds that beat the benchmark consistently. I

test for performance persistence using the methodology of Grinblatt and Titman

(1992)

The choice of a benchmark for international mutual funds remains a
controversial issue. The world equity index, the adopted benchmark in previous
studies (Cumby and Glen (1990), Eun, Kolodny and Resnick (1991), Droms and
Walker (1994)), does not fully satisfy the optimality conditions postulated by
asset pricing models. At the same time, researchers (Cooper and Kaplanis
(1994)) document strong home bias by investors in various countries. If investors
in one country prefer their home security, the optimal portfolio allocation for all

other investors will deviate from market value weights and the world equity

4
index will be inefficient, and thus not an appropriate benchmark for evaluating

fund performance. To test fund managers' investment skills requires a
benchmark to be efficient relative to the opportunity set available to managers.
However, constructing an efficient portfolio from the universe of world
securities is not an easy task and such a strategy is impractical for a passive
investor. Solnik (1994) discusses difficulties in deriving a theoretically optimal
benchmark for international funds. Sharpe (1992) recommends that a feasible
performance benchmark should be ”1) a viable alternative, 2) not easily beaten,
3) low in cost, and 4) identifiable before the fact." I evaluate equity and bond
funds against several passive global investment strategies that can be executed
inexpensively by an US. investor. Each strategy is associated with a benchmark
portfolio designed to test a specific service provided by international fund
managers. To test the efficiency of these benchmark portfolios, I use national
indices as surrogates for individual securities, an approach commonly used in
testing international asset pricing models (Harvey (1990), Cumby and Glen

(1990)).

Mean-variance performance statistics, including both the Jensen measure
(1968) and the Sharpe ratio, may be biased if fund and index returns are not

linearly related (Dybvig and Ross (1985)). N onlinearity could result from

5

manager’s market timing strategies or inherently nonlinear relations between
individual stocks and benchmarks. Grinblatt and Titman (1989) develop the
Positive Period Weight (PPW) measure which is not subject to this bias. As a
check, I evaluate performance using both the Jensen and PPW measures.

Another important consequence of a nonlinear return relation is that fund and
benchmark returns are not jointly normal; therefore I estimate the PPW measures

using the generalized method of moments.

The results of this study have several practical implications. Foremost, if
the addition of actively managed international mutual funds does not improve
the efficiency of passive strategies, an investor is better off purchasing index
funds because an indexing strategy tends to have lower operating and
management expenses. Today there are only a limited number of funds that
track international equity indices and there is no international bond index fund.
Secondly, performance measures are very sensitive to the benchmark used
(Grinblatt and Titman, 1994 and Lehman and Modest 1987). Mutual fund rating
services and academic researchers frequently use the world index or the S&P 500
index as benchmarks for international equity funds. If the world index is

inefficient, it will not be appropriate for evaluating managerial ability.

6

Designing an efficient benchmark that isolates the manager’s various investment

skills is especially important in an international setting.

In section 2, I review studies on international asset pricing models, mutual
fund performance measures and international mutual fund evaluation. I discuss
estimation procedures and performance test statistics in section 3. Section 4
describes the data used in this study. Performance evaluation of international
equity and bond funds appears in sections 5 and 6 respectively. Conclusions of

this study and topics for future research are presented in Section 7.

2. Literature Review

2.1 International Capital Asset Pricing Models

Solnik (1974) is the first to extend the closed economy partial—equilibrium
capital asset pricing model (CAPM) to an international framework. Assuming
frictionless capital markets, perfect competition, riskless borrowing and lending,
homogeneous expectations, and non-random domestic inflation, Solnik shows
that in equilibrium an investor holds a portfolio of risky assets hedged against
exchange rate risk and the risk-free asset denominated in the home currency.
Since all investors hold risky assets in the same proportions in equilibrium, the

world equity portfolio hedged against exchange risk is the optimal risky

7
portfolio. Adler and Dumas (1983) generalize the Solnik (1974) model by

allowing domestic inflation to be stochastic. They show that in equilibrium, the

optimal portfolio has two components: the universal log portfolio and the hedge
portfolio. If inflation for the home country is random, investors hold risky assets
in both portfolios and the equilibrium condition no longer implies that investors

hold risky assets in the same proportion as the world market equity portfolio.

Glen and Jorion (1993) show that an ex post optimized portfolio
containing national stock and bond indices and forward contracts is more
efficient than a portfolio containing only stock and bond indices. Comparing the
world index against a portfolio containing the world index and forward currency
contracts, they conclude that hedging the value weighted world index does not
improve its efficiency. They also do not find improvement in efficiency when
forward contracts are added to the Salomon Brothers World Bond Index or a

market value weighted index of world equity and bonds.

Cooper and Kaplanis (1994) documented substantial differences between
actual portfolio holdings and market capitalization weights. The motivation for
home bias is still unresolved. Cooper and Kaplanis estimate the Adler and

Dumas (1983) model and conclude that hedging for inflation risk cannot explain

8

the observed home bias at "reasonable" risk aversion level. Including other
institutional restrictions, such as withholding tax and transaction costs, the

model still cannot sufficiently explain the observed home bias.

International asset pricing models do not prescribe an optimal benchmark
for evaluating international mutual funds. Empirically, Cumby and Glen (1990)
and Harvey (1990) cannot reject the mean-variance efficiency of the unhedged
MSCI World Index using national indices. Cumby and Glen also cannot reject the
efficiency of a benchmark containing the MSCI World Index and a portfolio of
currency contracts. Harvey applies the mean-variance efficiency test developed
by Gibbons, Ross and Shanken (GRS, 1989). Cumby and Glen use a large sample
version of the GRS test and the Positive Period Weight (PPW) test of Grinblatt
and Titman (1989). Even though the MSCI World Index does not satisfy the
optimality conditions postulated by a theoretical asset pricing model, it possesses
desirable attributes as a feasible benchmark and results of past empirical studies
indicates that it is not easy to beat over the sample period of these studies.
Section 4 discusses the benchmarks used in this study and how each benchmark

is related to the hypothesis tested.

9

2.2 Mutual Fund Performance Measures

The Sharpe ratio (1966), the Treynor measure (1965) and the Jensen
measure (1968, 1969) have been widely used in measuring mutual fund
performance. The Sharpe ratio is defined as the mean divided by the standard
deviation of fund excess returns. The Jensen measure is the intercept from
regressing fund excess returns on a reference portfolio. The Treynor measure is
the ratio of the mean excess return to the regression coefficient (beta) on a
reference portfolio. It is well known that the Sharpe ratio is the appropriate
mean-variance performance measure when the asset is the only investment held
by an investor. The Jensen measure can be interpreted as the marginal
improvement, in a mean-variance sense, to the reference portfolio when the asset
examined is added to the investor’s investment set. All three measures assume

the existence of a risk-free asset.

Jobson and Korkie (1982) are the first to analyze the statistical properties
of the sample estimators for Sharpe ratio and Treynor measure. They assume
that asset returns are jointly normally distributed and derive approximations for
the mean and variances of the Sharpe ratio, along with the asymptotic

distribution and test statistic for comparing two or more Sharpe ratios. Gibbons,

10
Ross and Shanken (1989) propose a multivariate test to determine the relative

mean-variance efficiency of two sets of portfolios. They also assume joint
normality for asset returns and the existence of a risk-free asset. They show that
their multivariate test statistic is a monotonic linear transformation of the
difference between the maximized squared Sharpe ratios of two portfolios.
Kandel and Stambaugh (1989) develop likelihood ratio tests to determine the
mean-variance efficiency of a set of portfolios, including the Gibbons, Ross and
Shanken (1989) test as a special case. They investigate more general cases when a
risk-free asset does not exist and when the alternative hypothesis is not specified.
In addition to tests for portfolios, they also derive an efficiency test for a set of

factors.

All of the above statistical tests are developed based on the unconditional
distribution of asset returns. When the model assumes the existence of a risk-
free asset and asset returns are expressed as excess returns, the tests should be
interpreted as conditional on the observed returns of the risk-free asset. I
evaluate fund performance using the Jensen measure and a mean-variance test
that takes into account the short sale restriction on open-end mutual funds. If
fund managers have timing ability, the distribution of the benchmarks given the

timing information will be different from the distribution observed by an

1 l
uninformed investor. Dybvig and Ross (1985) show that if distribution of the

benchmarks conditional on the fund manager’s information set differs from the

unconditional distribution, the performance measures discussed above may be

biased.

Treynor and Mazuy (1966) propose a quadratic regression of mutual
funds on the market portfolio to estimate the timing and selectivity ability of
mutual fund managers. They argue that if managers have timing ability, they
will earn higher returns when the market is volatile. The regression intercept
measures selectivity ability and the coefficient for the quadratic term measures
timing ability. The Treynor-Mazuy measure is the sum of the regression
intercept and the product of the coefficient for the quadratic term and the

variance of the excess return on the market portfolio.

Admati, Bhattacharya, Pfleiderer and Ross (1986) propose two models for
measuring timing information when fund manager’s information set is not
directly observable. The factor approach assumes asset returns are generated by
a factor model. Distribution of the factors conditional on the information set of a
manager with timing ability is different from the unconditional distribution.

When manager has selectivity ability, distribution of risky assets conditional on

12

his decision information set is also different. The fund manager responds to the
timing information by shifting investment among the timing portfolios. The
number of timing portfolios equals the number of factors in the model.
However, the composition of the timing portfolios depends on the selectivity
ability of fund manager. Therefore, the timing ability and selectivity ability of

the fund manager must be determined simultaneously.

The second model proposed by Admati, et a1. is the portfolio approach. In
this model, selectivity information is defined to be independent of timing
information. Therefore, composition of the timing portfolio is invariant to the
selectivity ability of fund managers. The number of parameters to be estimated
using the portfolio approach is substantially reduced. The authors acquiesce that
despite the conceptual appeal of the factor approach, the portfolio approach is
more feasible to implement given the limited number of time series observations.
In the case when the reference portfolio contains only one asset and when the
response function of the informed investor is linearl, the portfolio approach is

equivalent to the quadratic regression of Treynor and Mazuy (1966).

 

1 A linear response function to the timing signal is consistent with a utility maximizing investor with

constant absolute risk aversion.

13
Grinblatt and Titman (1989) propose the Positive Period Weight (PPW)

measure for evaluating performance of mutual funds. Unlike other performance
measures, the PPW measure is not biased by timing ability of fund managers.

Grinblatt and Titman define the PPW measures as:

PPW = i wtat

t = l
where

W = f(pt,T),

T
plim[z w‘p,] = 0,

ts]

lplim[th]| < oo,

1
2w, =1,wt >0for all t.

[=1

p. is the excess returns on the benchmark portfolios and a: is the excess returns
on mutual funds. The period weights, Wt, can be interpreted as normalized

marginal utilities and the PPW measure represents change in the uninformed

14

investor’s expected marginal utility from adding mutual fund to his existing

portfolio.

Grinblatt and Titman (1994) examine the performance of US. mutual
funds using the Jensen measure, the PPW measure and the Treynor-Mazuy Total
Performance measure. They use four benchmarks: the equally-weighted US.
equity index, the value-weighted US. equity index, a 10 factor portfolio, and an 8
characteristic-based portfolio. They find that all the performance measures are
sensitive to the choice of benchmarks. All three measures produce similar

results for the same benchmark.

2.3 Performance Persistence

The efficient market hypothesis posits that any superior mutual fund
performance is distributed randomly among managers and over time. The
ability to predict future performance using historic performance represents a
Violation of the efficient market hypothesis. For passive investors, studying fund
ranking and past return statistics will not add value to their investment unless
Superior performance persists. Grinblatt and Titman (1992) test for performance
persistence and find positive performance persistence among their sample of 279

funds from 1975 to 1984. Hendricks, Patel and Zeckhauser (1993) use a sample of

15
no-load growth-oriented equity funds from 1975-1988. They find that funds that

outperform their peers for the last four quarters continue to earn returns higher
than the average fund in the following quarter. However, these funds do not
exhibit superior performance against the benchmarks. They also find that poor
performers continue to earn below average returns. Their overall result is
consistent with the positive performance persistence observed by Grinblatt and
Titman (1992). Bauman and Miller (1994) compare fund performance over
complete stock market cycles and find that funds that outperform over one cycle
tend to continue to perform well. Goetzmann and Ibbotson (1994) also find

positive persistence after controlling for cross-correlations among funds.

2.4 International Mutual Fund Performance Evaluation

Cumby and Glen (1990) compare the performance of fifteen international
mutual funds against two benchmarks from January 1982 through June 1988.
One of the benchmarks is the Morgan Stanley Capital International (MSCI)
World Index and the other contains the MSCI World Index and an equally
weighted portfolio of Eurocurrency deposits. They apply the Jensen (1968, 1969)
measure and the PPW measure of Grinblatt and Titman (1989) to their data.

Using MSCI national indices as dependent variables, Cumby and Glen do not

l6
reject the efficiency of the benchmarks in their study. Univariate tests for the

international mutual funds do not reject zero performance using either the
Jensen measure or the PPW measure. A large number of the estimated
performance measures are negative. Cumby and Glen use an asymptotic two-
sided test for the joint performance of all fifteen funds. With the Jensen measure,
they reject the joint hypothesis just above the five percent significance level. With
the PPW measure, they reject the joint hypothesis at the five percent significance
level. The test statistic for the PPW measure used by Cumby and Glen (1990) and
Grinblatt and Titman (1994) assumes that fund and benchmark returns are jointly
normal. If fund managers have timing ability, fund and benchmark returns will
not be linearly related and will not satisfy the joint nomarlity assumption. I
estimate the PPW measures for my funds using the generalized method of
moments (GMM), which provides consistent estimates even when returns are

not jointly normal.

Eun, Kolodny and Resnick (1991) examine the performance of thirteen
international mutual funds from 1977 through 1986 against three benchmarks:
the S&P 500 Index, the MSCI World Index and a portfolio of U. S. multinational
firms. They use the Sharpe ratio, the Treynor measure and the Jensen measure.

Five of the Jensen measures are significant when the S&P 500 Index is used as the

l7
benchmark. In all the other tests the null hypothesis cannot be rejected using the

Jensen measure. Eun, Kolodny and Resnick do not report a significance level for
the Sharpe ratio or the Treynor measure. Contrary to the Cumby and Glen (1990)
study, most of the estimated Jensen measures are positive using the MSCI World
Index as the benchmark even though these measures are not statistically

significant in either study.

Droms and Walker (1994) examine the performance of international
mutual funds using an error component model. They investigate the cross-
sectional relationship between fund performance and four fund characteristics:
total assets, expense ratio to average net assets, turnover rate and load fee versus
no-load fee. Their sample includes 108 funds from 1971 to 1990. They do not
find a statistically significant relationship between fund performance and fund

characteristics.

3. Methodology

3.1 A Mean-variance Spanning Test with a Short-sale Constraint

In this study, I evaluate individual and joint fund performance against
several passive investment strategies. Sections 5.1 and 6.1 contain detailed

descriptions of the benchmarks used in each passive strategy. Gibbons, Ross,

18
and Shanken (GRS, 1989) develop a multivariate test for comparing the mean-

variance efficiency of two sets of assets. They show that their test statistic can be
interpreted as the difference between the maximized squared Sharpe ratios of the
two portfolios. Let pk: denote the excess return on passive asset k in month t and
let an denote the excess return on mutual fund i in month t. A passive
benchmark, {pi}, includes only the passive assets while the combined investment
set, {phat}, includes the passive assets and the mutual funds. To determine the
performance of a single mutual fund, a fund investment set, {pean}, can be
formed. The null hypothesis of the GRS test is that {p} spans the mean-variance
space of {peat}. In other words, adding mutual funds to the passive benchmark
does not increase its Sharpe ratio. The necessary and sufficient condition for the

null is that the intercepts are jointly zero, Bo = 0, in the following regression:

at=B0+PtB+et (1)

where t = 1,2,..T, and en, en, i,j=1,2..,N, are jointly normal with zero mean,
covariance matrix 21, and E(eu esj) = 0 for t¢S. The regression intercept of a single
fund, Boil is the Jensen measure. The alternative hypothesis of the GRS test is that

the Jensen measures are not jointly zero, Bo #3 0. The GRS test statistic is:

(2)

FMV has an F distribution with N and (T-N-K) degrees of freedom under the null
and is equivalent to the Jensen measure t statistic when evaluating a single fund.
The large sample version of the GRS test is a Wald test, with the test statistic

WALDMv distributed as x2 with N degrees of freedom.

Using the GRS test to evaluate fund performance has one major
disadvantage. Since the alternative hypothesis is defined as Bo ¢ 0, both positive
and negative Jensen measures can trigger rejection of the null. This criterion is
appropriate only if investors can exploit both inferior and superior fund
performance. In practice, investors can increase the Sharpe ratio of their
portfolio by buying outperforming funds, but not by shorting underperforming
funds because open-end mutual funds cannot be sold short.2 One might argue
that if the underperformance results from inferior security selection, and
investors could observe the portfolio composition of the fund, they could get

around the short sale restriction on the fund by shorting or underweighting the

 

2 Two brokerages, Jack White and Co. and Fidelity, offer limited opportunities to short sell open—end funds.
Fidelity limits its offerings to a few of its own sector funds, and charges a brokerage commission to sell
and then repurchase the funds. The number of funds offered by Jack White depends on what funds are

internally available in customer margin accounts.

20

individual stocks held by the fund. However, portfolio compositions are not
publicly available in a timely manner. Moreover, underperformance resulting
from excessive management fees or transaction costs cannot be exploited by
trading on individual stocks. I show in Appendix B that given the short-sale
constraint, a zero vector of Jensen measures is no longer the necessary and

sufficient condition for the null, and the GRS test rejects too often.

As an example of the GRS test rejecting too frequently, Cumby and Glen
(1990) test the joint performance of international funds using a chi-square test
that is asymptotically equivalent to the GRS test. Their joint test rejects the
efficiency of the world index with a p-value of 0.055, yet 12 of their 15 funds have
negative Jensen measures. While Cumby and Glen correctly conclude their
funds underperform overall, their joint test erroneously suggests the world index

is inefficient.

Since investors cannot short sell open-end funds, a fund adds value only if

the optimal weight on the fund is positive in a portfolio containing the

benchmark and the fund. Let

[:lrrjwrmw) (3)

21

where

pt is an lxK vector of benchmark excess returns in month t;

at is an lxN vector of mutual fund excess returns in month t;

rfl is the risk-free rate in month t;

u is an (K+N) x 1 vector of expected excess returns. p. can be partitioned into p.’
= [up' u,’] where up is an le vector of benchmark expected excess returns
and u, is an le vector of mutual fund expected excess returns.

V is the covariance matrix of the excess returns on all assets. The (K+N) x

Vp V...

V...’ V.:l where Vp IS the

(K+N) matrix, V, can be partitioned into V =[

KxK covariance matrix of benchmark excess returns; V1| is the NxN
covariance matrix of mutual fund excess returns; and Vpa is the KxN

covariance matrix of benchmarks and mutual funds.

The optimal portfolio weights for the combined investment set are computed by

solving the following problem:

w = Argmin w'Vw (4)
{W}

subject to: W}; = c
where w' = {wp' w..'} are the optimal weights on {pg' at'} and c is a scalar constant.

The solutions to (4) are:

22

 

 

WP = w'r‘lw VP-ih’lp —vpe 2" BO]
W t1
(5)
w. = “'"vw z—l B0
w u

where 22" is the covariance matrix of at conditional on pg. Since the first term,
(w'Vw)/ (w'u), is a positive constant, the null and alternative hypotheses can be
written in terms of the conditional moments of at and p, Ho: Z'lflo = 0 versus Ha:
2" Bo 2 0. When evaluating a single fund, the correct null and alternative
hypotheses are [30 = 0 and 80 > 0 respectively because 2‘." is a positive scalar.
When evaluating joint performance of multiple funds, the null is the same as in
the GRS test, Bo = 0. If 2" is diagonal, the appropriate alternative is Ha: Bo Z 0. If
2'1 has non-zero off-diagonal elements, there are two cases when [30 2 0 does not
equal 2" Bo 2 0. The first case is when the optimal weight on a fund is positive
even though its Jensen measure is negative. The second case is when the optimal
weight on a fund with a positive Jensen measure is negative. These cases occur
only if the diversification benefit outweighs the reduction in expected returns
from including the underperforming fund or excluding the outperforming fund.
To fully account for these diversification effects, the residual covariance matrix

and the regression coefficients must be estimated simultaneously in a nonlinear

23

programming problem.3 Given the limited number of time series observations
and the high degree polynomial in the problem, estimating all the parameters
simultaneously is impractical. To make estimation feasible, I use Bo Z 0 as the
alternative. Intuitively, when the Jensen measure for one mutual fund is
positive, an investor can improve the mean-variance efficiency of his investment
set by purchasing the fund. Therefore even if the short-sale constraint is binding
on some of the funds, the investor can still be better off by purchasing at least

one of the funds which have positive intercepts.

Gourieroux, Holly and Monfort (1982), henceforth GHM, outline a two-
stage estimation procedure for the inequality-constrained model. In stage one, I

estimate (1) subject to 80 = 0, and compute the null restricted regression

residual é , and covariance matrix ‘2‘. . In the second stage, I estimate:

é, =70 +pty +v:
subject to y 0 2 0 (6)

fort =1,2,..T

 

3 Consider the case for testing performance of equity mutual funds using the world index. With 35 funds

and one index, the total number of parameters to be estimated is 700.

24

where V; is assumed normally distributed with zero mean and covariance matrix

f1 computed in stage one under the null restriction. The determination

coefficient R2 is defined as:

 

I—ZT (é: —70 "'ptrf):

t=l

, (7)

where 7 o and 7 are computed in stage two. GHM define the Kuhn-Tucker

statistic as:
KTMV = (TN)R2 (8)

where T is the number of time series observations and N is the number of funds
in the sample. 1(va is asymptotically distributed as a weighted mixture of x2

under Ho.4

3.2 Measuring Timing Ability

A number of researchers, including Dybvig and Ross (1985), Admati,

Bhattacharya, Pfleiderer and Ross (1986), and Grinblatt and Titman (1989), have

 

4 Wolak (1991) shows that the upper and lower bounds for the critical value of KTMV computed by Kodde
and Palm (1986) are tight for linear inequality constrained models such as ours. If the test statistic KT MV
falls between the upper and lower bounds, the bounds test is inconclusive. For such cases, I use a Monte

Carlo simulation method suggested by Wolak (1989b) to approximate the critical values.

25

shown that market timing strategies of fund managers can bias the Jensen
measure. Admati, et. al (1986) provide two models which disentangle the timing
portion and the selectivity portion of a performance measure. I define a fund
manager as an index timer if his investment set includes only the risk-free asset
and the benchmark and his investment decision is conditional on the timing
signal he receives. This definition of index timing is consistent with the portfolio
model by Admati, et. a1 (1986). They show that the covariance matrix of the
noise of the timing signal and the risk aversion coefficient of the informed
investor can be estimated by a quadratic regression of the mutual funds on the
timing portfolios. If the coefficients of the quadratic terms are significantly
different from zero, the expected returns of mutual funds conditional on the
benchmark are not linear and the Jensen measures may be biased. I estimate the

following quadratic regression to test for a linear return relationship between

mutual funds and the benchmarks:

In R k k
ait = who + ijtmij + Zpitmijflt + zzpltpmtmilluhm + Dit (9)
j=l

j.| I=I Isl
Incl

where i = 1,2,..N,

t = 1,2,..T,

26

mm is the coefficient for fund i on index m for m <=k,

mm is the quadratic coefficient for fund i on index m for m > k,

m = 1,2.. (1+K+K*(I<+1)/2),

vn, vii are distributed jointly normal for i,j=1,2..,N with zero mean and

variance-covariance matrix ‘P. E(vm v54) = 0 for tats.

When the benchmark contains a single index, (9) is the quadratic regression used
in Treynor and Mazuy (1966) who show that mutual fund performance
comprises two components. The timing component is the product of the
quadratic coefficient and the variance of the benchmark, and the security
selectivity component is the regression intercept in (9). The Treynor-Mazuy
performance measure is the sum of the timing and the security selectivity

components:

Treynor-Mazuy measurei = mo + ma * var(pt) (10)

3.3 The Positive Period Weight Measure

Grinblatt and Titman (1989) develop the Positive Period Weight (PPW)

measure for evaluating performance of mutual funds. Unlike other performance

27

measures, the PPW measure is robust to nonlinearity between mutual fund and
benchmark returns. To compute the PPW measures, I assume that the investor

has a power utility function with risk aversion parameter 0,
13[U(w.)] = EL—l—O Wt""] (11)

where W: is wealth at time t with W. = Wo*(1 + m + p (p). The investor selects
optimal weights (p on the benchmarks p. to maximize expected utility given the

risk-free rate m. Maximizing (10) yields:
E[(1 + m + pt<p)‘9pt'] = o. (12)

I choose 9 to be five to ensure that the optimal weights will sum close to one.5
The PPW measure represents the change in expected marginal utility from

adding the actively managed fund to the benchmark strategy:
E[(1 + rt. + p.¢)-9a.'] = PPW. (13)

The null hypothesis is that investing in mutual funds does not improve a passive

investor’s expected utility, PPW = 0. Since (13) implies that buying a fund with a

 

5 Cumby and Glen (1990) use a power utility function with a coefficient of relative risk aversion of six when

computing the PPW measure.

28

positive PPW measure improves utility, the short-sale constraint is not binding

for funds with positive measures.

I use the Generalized Method of Moments (GMM) to estimate (12) and
(13). The GMM procedure provides consistent estimates even when the active
funds and indices are not jointly normal. It is important to use a consistent
estimation method because if fund managers have timing ability, fund and
benchmark returns will not be jointly normal. Given K benchmarks and N
funds, there are K+N moment conditions. With K parameters to estimate, there
are N over-identified restrictions that can be used to test the null. Using the
separability result for GMM estimation by Ahn and Schmidt (1994), I estimate
(12) and (13) in a two-step procedure. In step one, I solve for (p in the following

system of nonlinear equations:

1 T _
g(<p)=;2[(1+r. +p.<p) Gp.']=0. (14)
tnl
The PPW estimates are:

PPW = %Z[(1+ r, + p.(p)'9a,'], (15)

29
where if) is the solution to (14).6 The vector PPW is asymptotically normal with

mean PPW and variance (2.7 The joint test statistic is:

WALDppw = PPW'f)" Piw. (16)

WALDppw is distributed asymptotically as x2 with N degrees of freedom.

Equation (13) implies that the utility of the passive investor can be
improved by positive investment in the fund when the PPW measure is greater
than zero. Therefore the short-sale constraint will not be binding for funds with
positive PPW measures. To impose the short-sale constraint, the parameter space
for WALDppw must be restricted. Andrews (1994) derives the Directed Wald
statistic which restricts the parameter space for WALDppw. However, the
distributional properties and robustness of the Directed Wald statistic for
nonnormal variables are not well known. Since I use the PPW measure primarily

as a check for whether the Jensen measures are biased by market timing or

 

6 To be consistent with Grinblatt and Titman (1989), we normalize (9) and (10) by dividing through by
; §[(1+ m + pot]-

7 fl is the lower right sub-matrix of A = [D’B-‘DJ-l, where D is the gradient of (13) and (14) with respect to (p
and PPW. B is the asymptotic covariance matrix of (13) and (14). The diagonal elements of Q is equivalent
to the heteroskedastic variances used by Cumby and Glen (1990) and Grinblatt and Titman (1994) only if

p and a. are jointly normal.

30

nonlinearity, I did not compute the Directed Wald statistic in the performance

test.

3.4 Measuring Performance Persistence

I apply Grinblatt and Titman's (1992) method to test for performance
persistence. First, I divide the sample into two subperiods and compute the
Jensen measures for each subperiod using (1). Secondly, I regress the Jensen
measures from the second subperiod on the Jensen measures from the first
subperiod. If performance persists, the regression coefficient should be positive.
Since the fund return residuals are likely to be cross-correlated, I use the
correction method suggested by Grinblatt and Titman to compute t statistics. Let
on be the difference between the Jensen measure of fund i and the average Jensen
measure for all funds from the first subperiod. By construction, the (11's sum to
zero. Next, I create a portfolio using the (11's as weights. The excess return on

this (at-weighted portfolio over the second subperiod is:

N

aiait
V2375 <17)

31

N
where ai = [30, - 2 E04 , t = T1 + 1, .., T, and var(a) is N times the cross-sectional

i=1
variance of the Jensen measures computed from the first sub-period. The
intercept from regressing apt on the excess returns of the benchmarks,pt , is
algebraically identical to the least squared slope coefficient from the cross-
sectional regression of the sub-period Jensen measures. However, the t statistic
of the intercept will not be biased by cross-correlations among the Jensen

measures.

4. Data

4.1 Monthly Returns on Mutual Funds and Indices

Monthly total returns for mutual funds are obtained from Morningstar
OnDisc. Monthly total returns on international and national equity and bond
indices, spot and forward exchange rates and 30-Day US. Treasury Bill are from
Ibbotson and Associates. Appendix A contains Ibbotson’s and Morningstar’s
documentation on the characteristics and methods of calculation for each index.
Monthly excess returns on indices and mutual funds are computed by
subtracting the 30-Day Treasury Bill return from total returns. Table 1 contains

summary statistics for the indices and mutual funds.

32 ,.
The sample period for equity funds is from January 1985 through March

1994. I look for funds with inception dates earlier than December 1984 and
”International Equity” or ” World Equity” as the investment objective as of
March 1994. There are 38 funds satisfying this criteria. Three of these 38 funds
were deleted from the sample. Vontobel EuroPacific was founded in December
1984 but does not have returns information for the first few months in 1985. Two
funds, Lexington Worldwide Emerging Markets and Phoenix Worldwide
Opportunities changed names and investment policies in 1991 and 1990
respectively. Thus, there are 35 international equity funds remaining in the

sample.

The bond mutual fund sample contains funds with ”Worldwide Bond” as
the investment objective in March 1994 and an inception date earlier than
December 1988.8 There are 24 funds which satisfies this criteria and have
continuous returns from January 1989 through March 1994. Six of these 24 funds
were deleted from the sample. Two funds, Bull & Bear Global Income and API
Global Income changed investment objective during the sample period.

Franklin / Templeton Global Currency only invests in short-term instruments.

 

3 When I set the criteria for the fund inception date to be earlier than December 1984, only one bond fund is

available.

33

Merrill Lynch Global Convertible invests mostly in convertible preferred stocks
and convertible bonds. Prudential Intermediate Global Income was named
Prudential Intermediate Income before 1990.9 Franklin Partners Tax-Advantage
International is open only to non-US. investors. There are 18 international bond
funds with continuous returns from January 1989 through March 1994 in the
sample. Issues relating to survivorship bias are discussed in section 5.6 and

section 6.5 for equity and bond funds respectively.

4.2 Currency Hedging

The international asset pricing model by Solnik (1976) states that hedging
against exchange rate risk is necessary for the world equity market portfolio to
be efficient. Glen and Jorion (1993) study various hedging strategies and show
that adding forward contracts improves the efficiency of a portfolio containing
national equity and bond indices in unrestricted proportions. However, they do
not find improvement in efficiency when hedging a market value weighted

portfolio. If mutual fund managers outperform an unhedged passive

 

9 I cannot obtain the investment objective of the Prudential Intermediate Income fund prior to 1990. This

fund was not covered by either Wiesenberger or Morningstar prior to 1993.

34

benchmark, the performance may be due to currency hedging and not

necessarily security or country selectivity.

I apply two currency hedging strategies to the passive benchmarks. The
first strategy chooses the optimal weights on the equity / bond indices and
currency forward contracts simultaneously. To implement this strategy, I add
three forward contracts: deutschemark, Japanese Yen and Canadian Dollar, to
each of the benchmark portfolios.10 Monthly return fit on a forward contract in
currency i is defined as the U. S. dollar payoff at time t from buying (1 / 811-1)

contracts at time t-1:

ft: = (51.1 - Fi,t-1)/Si,t-1 (18)

where Sm is the spot exchange rate in U. S. dollars per currency i at time t and
Fig-1 is the one month forward rate for currency i at time t-1 for delivery at time t.
The forward contract returns, fit, are added as independent variables to the
benchmark portfolio, p, when estimating performance using (1). The exchange
rate component of the dollar return on a fund is now captured by the forward

contracts, and no longer contributes to the fund's Jensen measure.

 

1° The other European currencies are highly collinear with the Deutschemark and are not used.

35
The second strategy is often referred to as unitary hedge because the

hedge ratio for each currency is one. The unitary hedged return on a country
index (uhpn) is computed as the US. dollar excess return plus the return from
selling Pm forward contracts at time t-1, where Pm is the price of index i

denominated in currency i,

Uhpm = pm - fit. (19)

Unitary hedged regional indices are constructed using beginning of month
market value weights and hedged returns on 14 national equity indices.11
Forward contract prices for the 14 countries begin in December 1985. Therefore
the sample period for performance tests applying the unitary hedge strategy is
from January 1986 through March 1994 for equity funds. Ibbotson and
Associates provide hedged regional and country bond index returns computed
by Salomon Brothers. The unitary hedged excess returns computed using (19)

are very similar to the Salomon Brothers hedged returns.

 

1‘ The 14 countries are: Australia, Belgium, Canada, Denmark, France, Germany, Italy, Japan, the
Netherlands, Spain, Sweden, Switzerland, the United Kingdom and the United States. The total market
value of these 14 countries averages 96% of the world market during the sample period from January

1986 through March 1994.

36
The two hedging strategies have important practical implications. The

unitary hedge strategy is easy to implement but does not take into account
correlations between currency and index returns. Furthermore, only the initial
investment is hedged. On the other hand, the simultaneous hedging strategy
significantly increases the number of assets to be managed. To assess the role of
currency hedging on fund performance, I repeat the mean-variance and PPW
tests using both hedging strategies. If funds outperform the unhedged
benchmarks but fail to outperform the hedged benchmarks, currency hedging is

an important factor in fund performance.

5. Performance Tests of International Equity Funds

5.1 Hypotheses and Benchmarks

I test for three attributes of international equity fund performance. First, I
test for security selectivity, the ability of managers to choose securities that
outperform their respective country indices. Second, I examine whether
international funds provide effective global diversification for US. investors.
Third, I examine country selectivity, the ability of fund managers to identify

superior performing countries in the world.

37
5.1.1 Hypothesis 1 - International mutual funds do not exhibit security

selectivity ability.

Security selectivity has been the focus of numerous performance
evaluation studies (Ippolito (1992)). Since international fund managers engage in
country as well as security selection, I use a 12-country benchmark to isolate their
security selectivity ability. Estimating individual sensitivities to each of the 12
countries removes country selection related performance from the Jensen
measure. For example, a fund that earns high returns by overweighting a
country that beats the world index will exhibit a high sensitivity to that country,
but will not obtain a positive Jensen measure against the 12-country benchmark.

Managers must select securities that outperform their respective country indices

for H1 to be rejected. 12

The 12-country benchmark includes Australia, Belgium, Canada, France,
Germany, Hong Kong, Italy, Japan, the Netherlands, Switzerland, the United

Kingdom and the United States. These 12 markets make up approximately 96%

 

‘2 I assume that each of the 12 country indices is efficient relative to securities within a country. Individual
securities from all 12 countries are needed to test the efficiency of the 12-country benchmark. At present,
I do not have data on individual securities from all 12 countries. If the funds reject the 12—country
benchmark, a formal efficiency test on the 12-country benchmark must be performed before I can

conclude that managers have security selectivity ability.

38
of the world equity market over our sample period. The GHM test rejects H2 if

the Sharpe ratio of the optimal combination of the 12 country indices is
significantly improved by purchasing the 35 mutual funds.

5.1.2 Hypothesis 2 - International mutual funds do not provide efficient global
diversification.

Global diversification is an important service provided by international
funds. Capital market theory suggests restricting investment to domestic assets
is inefficient a priori. In Table 1, all countries except Canada have return
correlations with the US. market below .61, indicating a potential for risk
reduction from investing in foreign securities. Nonetheless, I may fail to reject
H2 if international funds incur excessive expenses, offsetting their diversification

benefits.

I use the broadly-based Wilshire 5000 as a proxy for the US. equity
market. Rejecting H2 using the GHM test implies that a portfolio containing both
international funds and the Wilshire 5000 index has a better risk-return tradeoff
than the Wilshire 5000 alone. If fund managers do not exhibit superior security
selection skills but outperform the Wilshire 5000 index, I attribute their

performance to providing effective global diversification.

39
5.1.3 Hypothesis 3 - International mutual funds do not exhibit country

selectivity ability.

Even if fund managers do not have security selectivity ability, they may
be able to identify countries that outperform the world index. If the world index
is efficient vis-a-vis unmanaged country indices, and a fund has a positive Jensen
measure against the world index but not against the 12-country benchmark, then
country selectivity must be a factor in the fund’s performance. If the world index

is inefficient, it will not be an appropriate benchmark for testing H3.

Table 2 contains results of the GRS mean-variance efficiency tests on the
MSCI World Index using monthly dollar excess returns on national indices as
dependent variables. The national indices are unmanaged and their returns are
not affected by active security selection. I use the GRS test because the national
indices can be easily underweighted or shorted using futures contracts.13 Both
the GRS and PPW tests reject the efficiency of the world index at 5%. This is the
first study to reject the efficiency of the unhedged world index using an

unconditional test. Harvey (1991) cannot reject the mean-variance efficiency of

 

‘3 Futures contracts exist for most of the major stock exchanges in the countries of my sample except for
Belgium, Italy and Switzerland. Negative investment in the national indices can be accomplished easily

using futures contracts.

'31, uopoas u; passnasip .ffiajeris Buifipaq 15.11} at“ s; anbiuqaai snu ‘Alsnoaueilnurts uasoqa

aq oi xapui leOM our pue spenuoa promo; 9 an no siqBiaM [9111de our smone xraunpuaq Jesse-1; an ,.

 

uo siaerruoa promo; 9 pue xaput pjrom our jo pasoduroo ereunpuaq 13532-1;

12 Buisn rsar 539 am unorrad 1 ’iuapgja aq or asp arer aBueqaxa rsuiefie paBpaq
3‘1 35m” X919“! PIJOM 3‘11 19111 531919 WcIVD (9.461) >IIUIOS 3111331118 43113131113

sit SBAOJdLLII Afiarerrs Butxapui pIrom aAieu e or spun; [euoueurarut Butppe

ieqi rajui 1([uo uea I use; mug) aqr Buisn xaput plrom ruapgjaut ue urmjradmo
spun} our 51 '91.; Bunsen r0; xreunpuaq aieirdordde ue aq rou [[IM 1! ’potrad
aldures Aur Butrnp iuapijjaui st xaput pIrom aqr aauis °srdaararui uotssarfiar am
or renuns AraA are sarnseaur McIcI emu '[aA31 %g aqr re rueagiufits pue aAprsod
are spuepaqra N aqr pue urntBIag 10jsidaararuiuotssarfiar aql 'uotsnpuoa [fur
raije iou saop user: 2861 raqorao our 10; [onuoo or aquIIBA Aurump e Butppv
[900 st 1531 ADUQIDIHB our 10; an[eA-d ’potradqns 17661 pram-0661 Krenuef our
raAQ 9900 )0 anpeA-d e an Kauapgja xaput pIJOM133I31 1 '17661 qorew anorqr
0A6I Arenuef or potrad aldures aqi puaixa 1 uaqM 'sairrunoa u Aux Bursn potrad
aldures slfiaareH raAo rsar Aauapgja 31.11 .10} 17610 jo BHIBA-CI e uterqo 1 °xaput
pIrom aqr uo rsar Aauapgja jo sijnsar poiradqns suieiuoo z anBL jo [aued uronoq
an. '6861 Kew 113mm 0A6I Alemqaa 11101; 9! pouad atdums 91H use: SHE) am

Butsn sajqeirea iuapuadap se seatput [euoueu a Buisn xepuI pIIO M 133w our
017

41
Deutschemark, Japanese Yen and Canadian Dollars. Adding the 3 forward

contracts as independent variables removes the exchange rate component in the
regression intercept. The GRS test also rejects the efficiency of the 4-asset
benchmark, indicating that exchange rate movement is not a major factor in

rejecting the world index over my sample period.

Since the MSCI World Index is inefficient, I use a 3-region equity
benchmark to test for country selectivity ability of fund managers outside the
US. and Japan. The 3 regions are US, Japan and the rest of the world. In Table
2, both the GRS and the PPW tests do not reject the efficiency of the 3-region
benchmark. To outperform the 3-region equity benchmark, fund managers must
select countries other than the US. and Japan, or individual securities that have
superior performance. If the funds outperform the 3-region benchmark but not
the 12—country benchmark, the most likely source of performance is from

identifying outperforming countries outside of the US. and Japan.

5.2 Transaction Costs

Even passive investment strategies incur transaction costs. Since mutual
fund returns are reported net of expenses, comparing the performance of active

funds against an index which does not reflect the transaction costs of a passive

42
strategy is biased against the funds. To attenuate this bias, I estimate the

transaction costs for the indexing strategies by regressing returns on index
mutual funds against the indices they imitate. I select the index mutual funds
using the secondary investment objective by Morningstar. Composition of these
index funds either replicates the market value weights of an index or is
determined by a random sampling method to maximize the correlation between
index return and fund return. I use the Vanguard Index Total Stock Market fund
which tracks the Wilshire 5000 index, the Vanguard International Equity Index
European which tracks the MSCI EAFE Europe index, and the Vanguard
International Index Pacific which tracks the MSCI EAFE Pacific index. The

regression equation for estimating these transaction costs is:

mu = 80" + 511' Ijt + 8,} (20)

where j = 1,2,..,N, t = 1,2,~,T, 811'” NID“), sz)»
mjt = monthly return on index mutual fund j,
IJ-t = monthly return on the index tracked by mutual fund j,

601-: estimator of transactions costs for index j.

If an index fund perfectly tracks the index Without expense, the R2 in (20) will be
unity and intercept zero. Fund expenses give rise to a negative intercept,

assuming tracking error averages to zero. The absolute value of the intercept

43
plus an amortized Vanguard transaction fee is the estimate of the expenses of a

passive indexing strategy. In Table 3, all R2's exceed 0.989, indicating little
tracking error. The estimated monthly transaction costs are 0.036% for the
Wilshire 5000 index, 0.042% for the Europe index and 0.031 % for the Pacific
index. These estimates exceed the expense ratios reported by the index funds for
1993, because expense ratios include administrative expenses but not
commissions or bid-ask spreads. When evaluating performance of actively
managed equity funds, the estimated transaction costs are subtracted from

excess returns on the benchmarks.15

5.3 Empirical Results of International Equity Mutual Funds
5.3.1 Mean-variance Performance Tests

Univariate and joint performance test results for the 35 international
mutual funds appear in Table 4. The intercepts in the univariate regressions are
the traditional Jensen performance measures. Only two of the 35 funds have

positive Jensen measures against the 12-country benchmark, and 15 are

 

'5 Transaction costs for the world index and the 3-region benchmark are market value weighted averages of
the costs for the Wilshire 5(XJO, the Europe index and the Pacific index. The weights are computed using
market capitalization values of the corresponding PTA international indices as of March 1994. For the 12-
country benchmark, individual country returns are adjusted using transach'on costs from their respective

regions.

44
significantly negative at 10%. The GHM test does not reject the efficiency of the

12—country benchmark. Finding no evidence of security selectivity ability among
international fund managers is consistent with results from domestic studies

(Jensen (1968), Lehman and Modest (1987), Grinblatt and Titman (1994)).

In contrast, most of the Jensen measures are positive when the Wilshire
5000 Index is used as the benchmark with one significant at 5% and five
significant at 10%. These results are consistent with the international
diversification benefits reported by Eun and Resnick (1984) and Grauer and
Hakansson (1987). The GHM test rejects the Wilshire 5000 at slightly above 5%,
implying that international funds provide effective global diversification to US.

investors, even after accounting for their expenses.

Even without security selectivity ability, managers can still be rewarded
by overweighting countries that outperform the world index. Out of 35 funds, 27
Jensen measures are positive against the world index but only one is significant
at 10%. In previous tests using the world index, Eun, Kolodny and Resnick
(1991) find that 11 of 13 Jensen measures over 1977-86 are positive with none
significant, while Cumby and Glen (1990) find that 12 of 15 Jensen measures are

negative over January 1982-June 1988 with none significant. My results are more

45
in agreement with the findings of Eun, Kolodny and Resnick. The GHM test

reject the efficiency of the world index at a p-value of 0.0530. This is the first
paper to find that international funds as a group outperform the world index.16
Cumby and Glen (1990) also reject the world index using their funds but they use
a two-sided joint test and attribute inferior rather than superior performance as
the cause of rejection. Since the world index is inefficient during my sample
period, the superior performance of the funds only implies that managers have

successfully exploited this inefficiency.

To access the investment skills of fund managers, I use the efficient
3-region benchmark. In Table 4, 22 Jensen measures are negative, with one
significant at the 1 % level, one at 5%, and one at 10% against this benchmark.
Only one fund has a positive and significant Jensen measure. The GHM test does
not reject the 3-region benchmark, suggesting that managers are not able to select
outperforming countries outside the US. and Japan nor do they have security

selectivity ability.

 

1‘ Since my sample period include the 1987 stock market crash, I introduce a dummy variable for October
1987 and repeat the performance tests. More Jensen measures (31 of 35) are positive from the dummy

regression and the overall conclusions remain the same after controlling for the October 1987 crash.

46

In summary, I find that international equity funds provide diversification
benefits for US. investors. Since the MSCI World Index is inefficient during my
sample period, it is not an appropriate benchmark for testing manager’s
investment skills. My results suggest that international fund managers
successfully exploited the observed inefficiency in the world index. Using the
12-country and the 3—region equity benchmarks, I find no evidence of security or
country selectivity ability. For all 4 benchmarks, the p-values of the GRS test are
lower than those of the GHM test. The difference is most pronounced when a
large portion of the funds has negative performance, giving rise to low p-values
in the GRS test and high p-values in the GHM test. The null hypothesis of this
study is that actively managed international funds do not add value to a passive
investment strategy. Therefore, the null should not be rejected by inferior
performance, because investors cannot profit from inferior performance due to
the short-sale constraint on open-end mutual funds. The rejection region of the
GHM test is consistent with the null hypothesis whereas the GRS test rejects the

null too frequently.

5.3.2 Positive Period Weight Measures
As discussed in section 3.2, the Jensen measure and the mean-variance

tests may be biased if fund and benchmark returns are nonlinearly related. To

47

examine whether this potential source of bias affects the results in section 5.3.1, I
compute the robust Positive Period Weight (PPW) measures. Consistent with the
mean-variance tests, I subtract transaction costs from the excess returns on the
benchmarks when computing the PPW measures. Table 5 contains the estimated
PPW measures for individual funds. The univariate PPW measures are similar to
the Jensen measures. With the 12-country benchmark, 21 PPW measures are
significantly negative at 5%. For the Wilshire 5000 Index, the PPW measure
estimates are positive for 32 out of 35 funds with one significant at 5% and two at
10%. For the MSCI World Index, 23 of 35 funds have positive PPW measures but
only one is significant at 10%. For the 3-region benchmark, 24 of 35 funds have

negative PPW measures with one significant at 1 %, three at 5% and four at 10%.

The PPW measures and the GMM estimation procedure are robust to
nonlinearity between fund and benchmark returns and nonnormal residuals.
The similarity between the PPW and Jensen measures suggests that any
downward bias in the Jensen measures introduced by market timing strategies

(Grinblatt and Titman (1989)) is not economically significant in my sample.

48
5.3.3 Nonlinearity Between Fund and Benchmark Returns

Studies of US. mutual funds find that managers have significant timing
ability.17 Eun, Kolodny and Resnick (1991) use the S&P 500 Index and the market
timing model by Henriksson and Merton (1981) and find some evidence of
negative timing ability among international fund managers but they do not
provide statistical significance. Cumby and Glen (1990) use the Treynor-Mazuy
model and find significant negative timing coefficients against the world index

for most of the their international funds.

As a comparison to existing literature, I estimate the timing and security
selectivity performance of the 35 international funds using (9). Table 6 contains
the quadratic coefficients and the Treynor-Mazuy total performance measures.

With the MSCI World Index as the benchmark, 34 out of 35 funds have negative

quadratic coefficients and 30 are significant at 5%.18 The quadratic coefficients

 

‘7 Lehmann and Modest (1988) use an APT model and report significant regression coefficients on the
squared terms of the factors. Chang and Lewellen (1984) and Henriksson (1984) use the Merton model
and find significant negative timing ability among fund managers. Jagannathan and Korajczyk (1986)
demonstrate that the nonlinear pricing relationship can be due to the option like feature of common

stocks and not an indication of the timing ability of fund managers.

‘8 Since the MSCI World Index is inefficient during my sample period, it may not be a good benchmark for

testing timing ability. I estimate (9) using unmanaged national indices as dependent variables. The

49
using the Wilshire 5000 Index are all negative and significant at 5% for 28 out of

35 funds. My results are consistent with Cumby and Glen (1990) and Eun,
Kolondy and Resnick (1991). Even though the quadratic coefficients are
significantly negative, the Jensen measures in section 5.3.1 are very similar to the
robust PPW measures in section 5.3.2, suggesting that my conclusions from the
GHM tests are not materially affected by the observed nonlinearity. To further
examine the implications of the observed nonlinearity, I compute the Treynor-
Mazuy Total Performance Measure defined in section 3.2 equation (10). All three
measures are in agreement, providing evidence that any potential bias in the

Jensen measures and the mean-variance tests are not economically significant.

5.4 Performance Tests Results with Currency Hedged Equity Benchmarks

Table 7 contains the mean-variance and PPW test results when the
simultaneous hedging strategy is applied using three forward contracts: the
deutschemark, Japanese Yen and Canadian Dollar. All but one Jensen measures
are negative against the 12-country benchmark and forward contracts and 8 are

significant. Using the world index and forward contracts, five funds have

 

quadratic coefficient are negative for 10 counties and positive for Japan and Italy, with 2 significant at
1 % and 3 significant at 5%. Considering the results for the national indices, the mutual fund results could

be driven by an inherent nonlinearity with respect to the benchmarks, and not timing ability per se.

50

positive and significant Jensen measures at 5%. For the 3—region equity
benchmark, hedged versus unhedged univariate results are also similar,
especially for the 4 funds with significant Jensen measures. The GHM tests for
the simultaneously hedged benchmarks have slightly higher p-values but the
overall conclusions are the same. Since the results using the simultaneous
hedging strategy are consistent with those using the unhedged benchmarks, it
does not appear that currency hedging makes a significant aggregate

contribution to international equity fund performance.

Test results using unitary hedged benchmarks are in Table 8. As noted in
Section 4.2, the sample period for the unitary hedge is from January 1986
through March 1994.19 Results of the performance tests using the unitary hedge
method are quite different. Only two funds have negative Jensen measures
against the 11-country benchmark20 and 15 measures are positive and significant.
For the world index, 32 Jensen measures are positive with 5 significant at 5%.

Using the 3-region equity benchmark, 32 Jensen measures are positive and 8 are

 

1"I repeat performance tests using the unhedged benchmarks over the January 1986-March 1994 subperiod

and the results are the same as the entire sample period.

20 There is no forward contract for the Hong Kong Dollar. Therefore the Hong Kong Index cannot be

hedged directly using the unitary hedge method, leaving 11 countries in the benchmark.

51
significant. The GHM tests reject all three unitary hedged benchmarks at 5%.

During most of my sample period, the US. Dollar depreciated against the major
currencies except the Canadian Dollar. If the future spot rate Sm is greater than
the forward rate F114, an unitary hedged portfolio will have a lower return than
an unhedged portfolio. Since the future spot rates were higher than the forward
rates for all 10 currencies21 for most of the period, it is not surprising that funds
outperform the unitary hedged benchmarks. The PPW measures are similar to
the Jensen measures, confirming that any bias in the mean-variance tests is not

economically significant.

5.5 Performance Persistence

Even if mutual funds as a group do not exhibit superior performance,
investors can still benefit if an individual fund can consistently outperform the
benchmark. Several domestic studies find evidence of persistence (Grinblatt and
Titman (1992), Hendricks, Patel and Zeckhauser (1993), and Goetzmann and
Ibbotson (1994)). I divide the equity fund sample into subperiods of 56 and 55

observations respectively, and estimate Jensen measures 001 and [302 for each

 

21The 10 currencies are: Australian Dollar, Belgium Franc, Canadian Dollar, French Franc, Deutschemark,

Lira, Yen, Gilder, Swiss Franc and British Pound.

 

 

 

I:

to

52

subperiod. A positive slope coefficient from regressing [301 on B02 implies
persistence. I use the method of Grinblatt and Titman (1992) to compute the
slope coefficient, which corrects the t-statistic for correlations in the Jensen
measures. In Table 9, the funds exhibit persistence only against the 3-region
benchmark. As indicated in Table 4, most funds have negative Jensen measures
against the 3-region benchmark. If the observed persistence orginates from
inferior performance, the ability to predict poor results is not valuable to

investors. I find no evidence of performance persistence using the other three

benchmarks.

5.6 Survivorship

Since I require the funds to have continuous return histories through out
the entire period, my sample is subject to survivorship bias. There were 3 funds
in existence in 1985 which merged with other funds or changed their investment
objective during my sample period. Templeton Global I merged with Templeton
Global 11 in 1988 and later became Templeton Smaller Company Growth.

Sci/ Tech Holdings changed to Merrill Lynch Health Care A in 1992. World of
Technology was acquired by Financial Strategic Portfolios in 1988. The monthly

total return on the discontinued funds during their existence averaged 0.4%

53

below the average for all international equity funds listed in Wiesenberger’s
Investment Companies Yearbooks. Therefore, the joint performance of the surviving

funds may be overstated.

5.7 Summary of Empirical Tests on Equity Funds

This section examines the performance of international equity mutual
funds relative to several passive global investment strategies. Using individual
national indices as active assets, I reject the unconditional efficiency of the world
index, indicating that the world index is not appropriate for evaluating
managers’ investment skills. I use the 12-country benchmark to test for security
selectivity ability and the 3—region benchmark to test for country selectivity
ability outside the US. and Japan. Since open-end mutual funds cannot
generally be sold short, I use the GHM test which takes into account the short-
sale constraint. I allow for transaction costs on the indexing strategies when
evaluating fund performance. My results demonstrate that joint tests which
ignore the short-sale constraint, such as the GRS test, reject the benchmarks too

frequently.

The GHM test rejects the Wilshire 5000 index, implying that actively

managed international equity funds provide global diversification to US.

54

investors. Both the 12-country and the 3-region benchmarks are not rejected,
indicating fund managers do not have security or country selectivity ability.
However, international funds outperform the world index as a group, suggesting
that managers successfully exploited an inefficiency observed during my sample

period.

These conclusions do not change substantially when the passive strategy
incorporates currency hedging using forward contracts in unrestricted
proportions. Results from quadratic regressions suggest that the return
relationship between international funds and the benchmarks is nonlinear. I
estimate the Positive Period Weight (PPlN) measure by Grinblatt and Titman
(1989) which is not biased by the observed nonlinearity. I use the generalized
method of moments in my estimation which is robust to departure from joint
normality. The PPW results are similar to the Jensen measure results, indicating

the observed nonlinearity does not affect the overall qualitative conclusions.

6. Performance Tests of International Bond Mutual Funds

International bond funds are a relatively new investment vehicle for US.
investors. The Morningstar Mutual Fund Source Book did not have a category for

international bond funds until Fall 1988. This is the first study of their

55

performance. In addition to testing whether fund managers have superior
investment skills, I also examine whether actively managed international bond
funds provide effective global diversification to US. investors. The benefits of a
global equity portfolio have been advocated by several studies (Solnik (1974), Eun
and Resnick (1984) and Grauer and Hakansson (1987)), but only recently have
US. investors shown interest in international bonds. In Table 1, foreign country
bond indices have return correlations with the US. market below .58, indicating
a strong potential for risk reduction from adding foreign bonds to a domestic
portfolio. My observations are consistent with the findings in Odier and Solnik
(1993). Despite the low correlations among international bond indices, fund
managers may fail to outperform the US. domestic bond index if they incur

excessive expenses.

Odier and Solnik (1993) show that currency fluctuation has a greater
impact on the volatility of foreign bonds than foreign stocks. A recent Business
Week (1995) article reports that currency hedging reduced returns on
international bond funds. Section 6.3 examines the role of currency hedging in

international bond fund performance.

56
6.1 Hypotheses and Benchmarks for International Bond Funds

6.1.1 Hypothesis 4 - International bond fund managers do not have security

selectivity ability.

I use the Salomon Brothers (SB) World Government Bond Index as a
proxy for the market value weighted world bond index. As shown in Table 10,
the GRS test does not reject the efficiency of the SB world bond index against 11
unmanaged country government bond indices. The countries are Australia,
Belgium, Canada, France, Germany, Italy, Japan, The Netherlands, Switzerland,
the United Kingdom and the United States. Five intercepts are positive but none
is significant. Since the country bond indices are not actively managed, they
should not exhibit timing ability. I use equation 9 to test whether excess returns
on the SB world bond index are linearly related to excess returns on the country
bond indices. In Table 10, 7 of 11 coefficients on the squared world index are
positive but none is significant and the joint test does not reject a linear return
relationship. As expected, the PPW measures are very similar to the regression
intercepts from equation 1. The empirical results support the SB world bond
index as an appropriate benchmark for evaluating managers' investment ability

during my sample period. To outperform the SB world bond index, fund

57

managers must successfully select countries or individual government bond

issues within each country.

Since corporate bonds usually have higher default risk and higher return
than government bonds, using the world government bond index as a
benchmark does not capture performance related to the risk-return
characterisitcs of corporate bonds. If international bond funds invest in foreign
or domestic corporate bonds, they may outperform the SB world bond index
even if the managers do not have investment abilities. To examine whether the
risk-return factor of coporate bonds is a an important component in international
bond fund returns, I construct a 3—asset benchmark containing the SB non-US.
government bond index, the SB US. government bond index and the regression
residuals of the Lehman Brothers (LB) Corporate Bond index on the US.
government bond index.22 This 3-asset benchmark will provide some insight on

the determinents of international bond fund returns.

 

22 Ibbotson and Associates only provides annual returns on foreign corporate bond indices. By regressing
the Lehman Brothers Corporate Bond Index on the US. government bond index, I remove the U5. bond
market influence from the residuals. I then use the regression residuals as a proxy for the risk-return

factor of domestic as well as foreign corporate bonds.

58
6.1.2 Hypothesis 5 - International bond funds do not provide efficient global

diversification.

I use the Salomon Brothers (SB) BroadTM Index as the benchmark for
testing H5. The bonds comprising the BroadTM Index include SB high-grade
corporate bonds and 7, 10 and 30 year US. Treasury bonds. Rejecting the
benchmark using the GHM test implies that adding international bond funds to

the SB BroadTM Index produces a higher Sharpe ratio.

6.2 Empirical Results of International Bond Funds

6.2.1 Mean-variance Tests

Table 11 contains results of the mean-variance test and Jensen measures
for 18 international bond funds. As of March 1994, there is no index fund which
tracks international bond indicesl‘. Hence, I cannot estimate transaction costs for
passive bond indexing strategies. When interpreting performance test results of
bond funds, transaction costs remain a factor if the funds underperform the
benchmarks. With the SB world bond index, 11 funds have positive Jensen

measures but only one is significant at 10%. Of the 7 funds with negative

 

13 There are only three US. bond indices, the LB Aggregate, the LB U.S. Govemment/ Corporate and the

Salomon Brothers Broad”, that are currently tracked by index funds.

59

measures, one is significant at 5%. The Jensen measures range from -0.1823% to
0.2557%, averaging 0.0025% per month. The average monthly expense ratio is
0.1235% in 1993 for the 18 funds in the sample. It appears that international bond
fund managers generate enough return to pay for their expenses. Unfortunately,
there is no index fund that tracks the world bond index to allow a direct
comparison between the actively managed funds and an indexing strategy net of
all expenses. The GHM test does not reject the SB world bond index, indicating
that fund managers jointly do not exhibit superior performance. The p-value of
the GHM test is 0.8788 versus the p-value of the GRS test is 0.0972,
demonstrating once again that the GRS test rejects the null too often. The
difference between the GRS test and the Wald test is most likely due to
adjustment for degrees of freedom because there are only 63 monthly

observations and 18 funds to be tested.

Even though international bond funds do not outperform the world bond
index, they may still add value to an US. investor's domestic portfolio through
diversification. As of today, there is no international index bond fund, leaving
direct foreign investment the only other alternative available to a US. investor
desiring global diversification in the bond market. Using the SB BroadTM Index

as benchmark, only 2 funds have positive Jensen measures but none is

60
significant. The GHM test does not reject the SB BroadTM Index, implying that

adding international bond funds does not increase its Sharpe ratio despite the
low correlation between US. and foreign bond indices. Given that the SB
BroadTM Index in Table 1 has the lowest standard deviation among all bond
indices and an average excess return higher than the world bond index during
my sample period, it is not surprising that international fund managers cannot

outperform the US. bond market.

Next I estimate the funds’ sensitivities to each of the 3 factors in the 3-asset
benchmark. In Table 12 all the funds have positive and significant coefficients
for both the non-US. and US. government bond indices. Coefficients on the
non-US. index range from 0.1747 to 0.8831, averaging 0.4717 while coefficients
on the US. index range from 0.1884 to 1.4425, averaging 0.5505. All the funds
have a positive coefficient on the LB corporate bond residuals and 8 are
statistically significant, suggesting that corporate bonds remain an important
factor in international bond fund returns even after controlling for foreign and

US. government bonds.

6.2.2 PPW Tests

If bond fund managers have timing ability, the Jensen measures may be

biased downwards. To determine whether market timing explains the poor

61
performance found in section 6.2.1, I compute the Positive Period Weight (PPW)

measures which is robust to bias due to market timing. Table 13 shows that the
PPW measures are very similar to the Jensen measures reported in Table 11. For
the world bond index, 11 of 18 funds have positive PPW measures. Only two
funds have significant PPW measures, one positive and one negative. For the SB
BroadTM Index, the PPW measures are also consistent with the results in section
6.2.1, suggesting that any downward bias in the Jensen measures induced by

managers’ timing strategies is not economically significant.

6.2.3 Market Timing and Overall Performance

I use the SB World Bond Index as the benchmark in equation 9 to test
whether international bond fund managers have market timing ability. The SB
World Bond Index is efficient and linearly related to unmanaged country bond
indices. Results in Table 14 suggest that fund managers' may practice some
timing strategies. Out of 18 funds, 14 have positive quadratic coefficients and 5
are significant. The Wald test rejects the null that all funds have zero quadratic

coefficients.

To examine the role of market timing in the funds’ overall performance, I

compute the timing and selectivity components of the Treynor-Mazuy

62

performance measure. Security selectivity related performance is captured by
the intercept in equation 9 and timing related performance is the product of the
quadratic coefficient and the variance of the world bond index. As expected, the
timing components are relatively large for funds with significant quadratic
coefficients. However, the total performance measures combining timing and
security selectivity are similar in magnitude to the PPW and the Jensen measures.
Averaging over 18 funds, the timing-related performance is 0.0892%, the security
selection performance is -0.0882%, and the total performance is 0.0010% per
month. The Treynor-Mazuy measures further corroborate that manager's timing
activities do not materially affect the qualitative conclusions of the mean-

variance tests in 6.2.1..

6.3 Performance Tests Results with Currency Hedged Bond Benchmarks

There is no data available on the actual hedging activities of fund
managers. To examine whether currency hedging is a factor in fund
performance, I compare returns on international bond funds against the world
bond index hedged using forward contracts. 1 apply two hedging strategies, the

first computes the optimal weights on the forward contracts and on the world

63

bond index simultaneously, the second uses a hedge ratio of one. I refer to the

first strategy as simultaneous hedge and the second strategy as unitary hedge.

To apply the simultaneous hedging strategy in performance evaluation, I
increase the number of independent variables in equation 1 by three forward
contracts: the deutschemark, Japanese Yen and Canadian Dollar. Adding the
forward contracts removes the currency component in the Jensen measures.
Against a benchmark containing the SB world bond index and three forward
contracts, 12 of 18 funds have negative Jensen measures, ranging from ~0.1974%
to 0.2505%, averaging 0.0265% per month. All funds have lower Jensen
measures against the simultaneously hedged world bond benchmark than the
unhedged world bond index. The coefficients on the Canadian Dollar are
positive for all funds with 13 significant and the coefficients on the
Deutschemark and Japanese Yen are negative for most funds, with 8 and 10
significant respectively. The uniformity of the signs on the coefficients across
funds is rather striking. The observed coefficients are consistent with short

positions (hedging) on the deutschemark and Yen, and long positions on the

64

Canadian Dollar.24 Even though returns on international bond funds are
sensitive to exchange rate movements, the Jensen measures in Table 15 are
essentially comparable to those in Table 11, especially for funds with statistically
significant Jensen measures. Again, only two Jensen measures are significant,
one positive and one negative. Interestingly, Scudder International Bond, the
only fund with a significant positive Jensen measure, is the least sensitive to
forward contract returns. The GHM test does not reject the simultaneously
hedged world bond index, confirming that fund managers do not have security
selectivity ability even after controlling for the effects of changes in exchange

rates.

I use the US. Dollar Hedged Salomon Brothers World Government Bond
Index as the unitary hedged world bond index. Salomon Brothers constructs
US. Dollar Hedged returns for international and country bond indices using
one-month forward contracts with the contract amount set to the bond price plus

the accrued and expected coupon payment. I compute the unitary hedged

 

24 Even if managers do not make use of any currency hedging strategy, the funds may still obtain a positive
coefficient on the Canada Dollar if they overweight Canadian government bonds. Similarily the negative

coefficients could result from the funds underweighting German and Japanese government bonds.

65

returns for 7 countries25 using equation 19 and my results are very similar to the
returns from Salomon Brothers, with an average difference of 0.012% and
average correlation over 0.98. If the future spot rate, S“, is higher than the
forward rate, F114, hedging using the unitary strategy will result in lower returns
than an unhedged strategy. At the same time, hedged returns often have lower
volatility than unhedged returns. If a fund employs currency hedging, its
performance may be reduced if the decrease in the hedged return is not off set by
a corresponding decrease in volatility. Since the average future spot rates were
higher than forward rates during my sample period for 10 currenciesZ", currency
hedging may be an explanation for their poor performance in Table 12. If fund
managers have security and / or country selectivity ability but their performance
is reduced by a unitary hedging strategy, they will outperform the SB US. Dollar
Hedged World Bond Index but not the unhedged SB world bond index. Table 16
contains the performance results of the international bond funds. Ten funds
have negative Jensen measures but none is significant. Most funds have worse

performance against the unitary hedged world bond index than the unhedged

 

25 The 7 countries are: Australia, Canada, France, Germany, Japan, Switzerland and the United Kingdom.

26 The 10 currencies are: Australia Dollar, Belgium Franc, Canadian Dollar, French Franc, Deutschemark,

Lira, Japanese Yen, Gilder, Swiss Franc and British Pound.

66

index, with CT. Global Strategic Income experiencing the largest decrease in its
Jensen measure, from 0.052% to -0.1476%. A notable exception is T. Rowe Price
International Bond whose Jensen measure increase from -0.1366% to -0.0072%
when the world bond index is hedged. The GHM test does not reject the unitary
hedged world bond index. The PPW and Treynor-Mazuy measures also
demonstrate that the funds do not have exceptional performance. Interestingly,
only one fund has a significant timing coefficient compared to 5 funds when the
unhedged world bond index is used. Given that the funds underperform both
the unitary hedged and the unhedged world index, currency hedging does not

appear to be an explanation for the funds’ lack of superior performance.

6.4 Performance Persistence

I divide the bond fund sample into subperiods of 32 and 31 observations
respectively, and then estimate Jensen measures Bo1 and [302 for each subperiod.
A significantly positive slope coefficient from regressing Bo1 on [302 implies
persistence. I use the method of Grinblatt and Titman (1992) to compute the
slope coefficient, which corrects the t-statistic for correlations in fund return
residuals. I find no evidence of performance persistence against all three

benchmarks in Table 17. The relatively small number of observations available

67
on international bond funds limits the power of the persistence test. Blake, Elton,

and Gruber (1993) also do not observe any performance persistence in their

sample of domestic bond mutual funds.

6.5 Discontinued Funds and Survivorship Bias

Morningstar first began a category for international bond funds in the Fall
1988 edition of Mutual Fund Source Book. Another popular mutual fund
publication, Investment Companies Yearbook by Wiesenberger, did not have a
classification for international bond funds until 1991. I look for funds with
international bond as the investment objective in the Mutual Fund Source Book
(Fall 1988) and funds with a primary objective as ”1” (Income) and an investment
policy of ”C&I” (Canadian and International) and funds with an investment
policy of ”Bond” and names containing the words: "international", "world",
"global", "foreign" plus "bond", "fixed income" in the 1989 Investment Companies
Yearbook. I exclude funds that invest primarily in short-term instruments or
stocks. There are 25 international bond funds listed in either the Mutual Fund

Source Book or the Investment Companies as of December 1988.27 Two funds:

 

1’7 One fund, Keystone American World Bond, was first listed in Wiesenberger in 1990 even though the fund

was found in January 1987.

68

Fennimore International Fund - Fixed Income, and Transatlantic Income Fund
(later became Kleinwort Benson Global Income), were liquidated in 1990. Fund
Source Global Bond, Hancock World Trust - World Fixed Income, and Pilgrim
Foreign Invest International Bond were listed in the 1989 Investment Companies
Yearbook but disappeared since the 1990 edition. Therefore return data on these
three funds is only available until 1988. Meeschaert International Bond Trust
changed name to Anchor International Bond Trust in 1990 and coverage on the
new fund stopped after 1992. AMA Income Global Income and DF A Fixed
Income Portfolio also disappeared since the 1992 edition. Monthly returns on the
discontinued funds during their existence average 0.24% below the funds in my

sample.

6.6 Summary of Empirical Tests on Bond Fund Performance

I use the SB World Government Bond Index as the benchmark to evaluate
performance of international bond funds. Using 11 unmanaged country bond
indices, the GRS test do not reject the efficiency of the SB World Government
Bond Index and I did not find any evidence of nonlinearity. The international
bond funds in my sample do not outperform the SB world bond index. The

GHM test has a p-value of 0.8788 and the average Jensen measure is 0.0025% per

69

month. Using a 3-asset benchmark I find that international fund returns are
sensitive to a corporate bond factor in addition to US. and non-US. government
bonds. Surprisingly, the international bond funds do not outperform a domestic
benchmark, the SB BroadTM Index. The robust PPW and Treynor-Mazuy
performance measures are similar to the Jensen measures, confirming that
conclusions from the mean-variance tests are not biased by fund managers’
market timing strategies. Returns on international bond funds are sensitive to
exchange rate movements. However, the funds do not outperform the SB Dollar
Hedged world bond index, therefore their prosaic performance against the

unhedged world bond index cannot be explained by a unitary hedge strategy.

The lack of superior performance against the world bond index prompts
the question: why are there no international or regional bond index funds? In
fact, there are only three domestic index bond funds, the Portico Bond Immdex
which tracks the LB Aggregate Index, the SE1 Index Bond Index which tracks the
LB Corp/ Gov’t Index, and the Vanguard Bond Index Total Bond which tracks

the SB BroadTM Index.

70
6.7 Balanced Benchmarks

Since a well diversified portfolio contains both stocks and bonds, a
balanced benchmark is useful for evaluating mutual funds investing in both
instruments. A market value-weighted world balanced index is a desirable
candidate because it is identifiable ex ante and can be implemented by an
indexing strategy. As discussed in section 2, an appropriate benchmark must be
efficient relative to the investment opportunity set available to fund managers.
Table 18 presents the mean-variance efficiency test of three balanced benchmark
candidates. The GRS test rejects the world balanced index at just above 5% using
12 equity and 11 bond country indices. Twenty of 23 intercepts are positive and
four are significant. Since the world balanced index is not efficient over the
sample period, it is not an appropriate benchmark for performance evaluation.
The 2-asset balanced benchmark containing the MSCI world equity and the SB
world bond indices has a p—value of 0.1097 in the GRS test and is only marginally
acceptable as a benchmark. Lastly I examine the 4-asset balanced benchmark
containing three regional equity indices (US, Japan, and the rest of the world)
and the SB world bond index. The GRS test for the 4-asset benchmark has a

p-value of 0.2104 against 10 equity and 11 bond country indices. I control for the

71
effects of exchange rate changes using three forward contracts (Deutschemark,

Japanese Yen, and Canadian Dollar) and the results remain essentially the same.

I compare the performance of 53 international equity and bond funds
against these three benchmarks. Since the efficiency of the first two benchmarks
are rejected by unmanaged country indices, superior performance by mutual
funds against these benchmarks does not imply exceptional investment skills. In
order to beat the efficient 4-asset balanced benchmark, fund managers must be
able to identify outperforming countries, instruments (bond versus stocks), or
securities within each country. In addition to the 3 international balanced
benchmarks, I also compare the funds to a value-weighted U.S. balanced
benchmark. Outperforming the US. balanced benchmark implies that the
international funds provide global diversification to aU.S. investor holding a

balanced domestic portfolio.

The sample period is from January 1989 through March 1994. Excess
returns on international funds are computed net of expenses but excess returns
on the benchmarks do not include transaction costs needed to implement an
indexing strategy. Therefore, the empirical results in Table 19 may slightly

understate fund performance. Since there are only 63 monthly observations and

72
there are 53 funds to be tested, inference based on the GRS test, a finite sample

test, is more reliable and I do not report the Wald statistic from the large sample
test. Both the GHM and GRS tests do not reject any of the benchmarks, implying
that international fund managers do not exhibit superior performance. More
than half of the funds have negative Jensen measures against the US. balanced
benchmark. Only two funds have significant Jensen measures and they are both
negative. The funds perform relatively well against the world balanced index,
with 41 positive Jensen measures and 6 are significant. However, the GHM test
indicates that jointly the funds do not outperform the world balanced index even
though it is inefficient during my sample period. The funds perform
substantially worse against the efficient 4-asset balanced benchmark, with 35
negative Jensen measures and 5 are significant. Only 2 funds have positive and
significant Jensen measures. Table 20 contains the PPW measures for the 53
funds. The overall results of the PPW tests are consistent with the Jensen
measures, indicating that any potential bias in the mean-variance tests

introduced by nonlinearity or market timing is not economically important.

73

7. Conclusions and Future Research

This dissertation examines the value provided by actively managed
international mutual funds. Within a sample of funds, a few will exhibit superior
performance simply by chance. To conclude that active management adds value,
the funds must jointly outperform a passive benchmark. With few exceptions,
open-end mutual funds cannot be sold short. Therefore, investors can only profit
from superior, not inferior performance. I compute the Jensen measures for
individual funds and apply the GHM test, which accounts for the short-sale
restriction, to determine if the funds jointly outperform the benchmark. As a
robustness check, I also compute the PPW and Treynor-Mazuy performance

measures.

Domestic mutual fund studies (Lehman and Modest 1987 and Grinblatt
and Titman 1994) find that performance measures are sensitive to benchmarks.
Selecting an appropriate benchmark for international equity and bond funds is
not a trivial task. International asset pricing models do not identify an easily
observable optimal portfolio. I use the 12-country equity benchmark, the
3-region equity benchmark, and the Wilshire 5000 Index to test the value

provided by international equity funds. The two benchmarks for evaluating

74
international bond funds are the SB World Government Bond Index, and the SB

BroadTM Index. With the exception of the world bond index, each of these
benchmarks is identifiable ex ante, not easily beaten, and can be implemented

through an indexing strategy.

The MSCI world index is not efficient during my sample period and
therefore not an appropriate benchmark for evaluating managers’ investment
ability. Nonetheless, international equity fund managers successfully exploited
this inefficiency and they outperform the MSCI world index during my sample
period. Compared to the efficient benchmarks, equity and bond fund managers
do not exhibit exceptional security or country selectivity abilities. The lack of
superior investment skills among fund managers is consistent with evidence
from domestic mutual fund studies. International equity funds provide effective
global diversification to a US. investor holding a domestic equity portfolio.
However, international bond funds do not outperform the domestic SB BroadTM
Index during my sample period. I verify results of the mean-variance tests and
the Jensen measures against the robust PPW and Treynor-Mazuy performance
measures. All three performance measures are very similar for the same
benchmark, confirming that my conclusions are not materially affected by any

observed nonlinearity or timing strategies.

75
Results from this dissertation and studies on domestic funds suggest that

active management cannot outperform indexing strategies. At the same time,
growth in actively managed funds greatly exceeds index funds. The 1995
Business Week Guide to Mutual Funds contains ratings of over 1,800 mutual funds,
but only 32 index funds are listed and the 10 largest equity funds and the 10
largest bond funds are all actively managed. Since fund managers do not
produce superior risk-adjusted returns, their popularity must be due to other
factors. There is very little research on the determinants of individual investors’
choice of funds. Most studies (Chevalier and Ellison (1995), Sirri and Tufano
(1993), Ippolito (1992)) focus on the relationship between historic performance
and future cash flow into mutual funds and they find that funds with the highest
returns attract more new investment. Ippolito (1992) identifies an asyrmnetry in
investors’ response to historic performance. He shows that growth rate for the
best funds is larger than the rate of decrease for underperforming funds. These
studies are the first few attempts to examine the characteristics of the individual
investor’s demand for mutual funds. As suggested by Brennan (1995), a fruitful
area for future research is to gain better understanding of how the individual
investor formulates his investment decision. For example, what information
source does the individual investor rely on when making investment decisions?

Can information with low search cost lead to a profitable investment strategy?

76

Do investors value non-financial factors when choosing an investment?
Traditionally, academic research only emphasizes on the financial factors and
rule out non-financial factors a priori. How do investors save? What are the time
series characteristics of cash flows to mutual funds? The answers to these
questions can provide important insights to the role of mutual funds as a

financial intermediary.

77
Table 1

Summary Statistics: Monthly Excess Returns in US$ (%)

 

Sample

Sample Standard
Average Deviation

 

Equity Regional Indices: lanuariLI985 to March 1994

 

 

30-day Treasury Bill a 0.4781
Wilshire 5000 Index 0.7695
MSCI World Equity Index 0.8887
World Equity excluding US. and Japan 1.2258
US$ Hedge World Equity Index 0.4146
Forward Contracts b: January 1985 to March 1994
Deutschmark 0.6267
Japanese Yen 0.7405
Canadian Dollar 0.1714
MSCI Country Equity Indices: January 1985 to March 1994
Australia 1.1391
Belgium 1.6707
Canada 0.3350
France 1.4862
Germany 1.2641
Hong Kong 2.1156
Italy 1.3858
Japan 1.2137
Netherlands 1.3246
Switzerland 1.4125
United Kingdom 1.1541
United States 0.7903

0.1544
4.5502
4.5380
5.0118
4.4713

3.6980
3.4685
1.3023

7.9917
6.2652
4.6498
6.8134
6.9575
8.4229
8.1230
8.0674
4.6339
5.7150
6.3923
4.5132

 

a Statistics for the 30-day Treasury Bill are based on total return,

not excess return.

b Since forward contracts do not require initial investment, we

standarize the forward contract returns by the initial spot price:

f“ = (Sm - Fi,.-1)/Si,.-1 where 51,: is the spot exchange rate in U. S.

dollars per currency i at time t and Fig-1 is the one month forward

rate for currency i at time t-1 for delivery at time t.

78

Table 1 (cont'd).

 

Sample
Investment Sample Standard
Oll'ective Average Deviation

 

Efiauitufi Funds: January 1985 to March 1994

Alliance Canadian Foreign 0.1639 5.1920
Alliance Global Small Cap A World 0.5124 6.2953
Alliance International A Foreign 0.9450 5.3507
Bailard, Biehl International Equity Foreign 0.6990 5.2541
Centerland Kleinwort International Equity B World 1.0357 5.1152
Dean Witter WorldWide Investment Foreign 0.7413 4.2240
EuroPacific Growth Foreign 1.0744 4.2992
Fidelity Overseas Foreign 1.2877 5.7635
First Invest Global World 0.8814 5.5050
FT International Equity A Foreign 0.9669 5.1026
G.T. Global New Pacific Growth A Pacific 1.0146 5.5248
IDS International World 0.8654 5.0865
Invesco Pacific Basin Pacific 0.9306 6.2257
Japan Pacific 1.1103 6.6873
Kemper International Foreign 0.9274 4.7528
Keystone International Foreign 0.6699 4.8447
Merrill Lynch Global Holding A World 0.7641 3.9580
Merrill Lynch Pacific A Pacific 1.3291 6.2727
New Perspective World 0.9045 4.0611
Oppenheimer Global A World 1.1874 5.4000
PaineWebber Atlas Global Growth A World 0.9304 4.7760
Princor World Foreign 0.6747 5.2529
Prudential Global 8 World 0.7610 4.8699
Putnam Global Growth A World 0.9993 4.5220
RSI Retirement International Equity World 0.8483 5.0018
Scudder International Foreign 1.0101 4.8447
Smith Barney Shearson Global Opportunity A World 0.4566 4.7167
T. Rowe Price International Stock World 1.1341 4.9524
Templeton Foreign Foreign 1.0784 3.9158
Templeton Growth World 0.8620 4.2586
Templeton Smaller Company Growth World 0.7329 4.7092
Templeton World World 0.7500 4.2494
United International Growth Foreign 0.8638 4.5028
Vanguard International Growth Foreign 1.0704 5.1442
Vanguard / Trustees' Equity International Foreign 1.0117 4.4145

 

79

Table 1 (cont'd).

 

Sample

Investment Sample Standard
Objective Average Deviation

 

Bond Regional Indices: January 1989 to March 1994

30-day Treasury Bill ‘

SB World Government Bond Index

SB Non—US. Government Bond Index

SB Broad Index
LB Corporate Bond Index

SB USS World Goverment Bond Index

SB CountryBond Indices: Jaunary1989 to March 1994

Australia
Belgium

Canada

France

Germany

Italy

Japan

The Netherlands
Switzerland
United Kingdom
United States

Bond Funds: January 1989 to March 1994

Capital World Bond

Fidelity Global Bond

Franklin Global Govt Income
G.T. Global Govt Income A
CT Global Strategic Inc A
Hancock Freedom Global Inc B
Keystone Amer World Bond A
Lord Abbett Global Income
Merrill Lynch Global Bond A
MFS World Governments A
PaineWebber Global Income B
Putnam Global Govt] Income A
Scudder International Bond
Smith Barney Shear Glob Bd B
T. Rowe Price Intl Bond
Templeton Income

TNE Global Government A
Van Eck World Income

Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond
Worldwide Bond

0.4462
0.3373
0.3363
0.3724
0.4114
0.2379

0.4893
0.5182
0.2751
0.5565
0.3411
0.4449
0.4451
0.3763
0.1687
0.3095
0.3595

0.2579
0.2528
0.1766
0.3290
0.4000
0.0576
0.1801
0.3370
0.4074
0.3295
0.2179
0.3406
0.6138
0.2117
0.3231
0.2189
0.1216
0.2295

0.1827
1.8952
2.9557
1.2095
1.3691
1.0442

3.2824
3.3762
2.6388
3.5071
3.6784
3.9599
3.6068
3.6234
3.6902
4.2298
1.3271

1.6859
1.8361
1.8250
2.0958
3.1271
1.6272
1.9072
1.9202
2.0139
2.1588
1.6302
1.9194
2.3004
1.6548
2.7489
1.6670
1.8351
2.1929

 

80

 

 

 

 

 

 

 

 

dddd. m.dvd Ehmd ,dmnmd wound, mend ~.mmd Send mound dvmmd dn.md 53d. .m.D
dddd. ~.mcd mnwcd Sand owned Ndmod coded .mbmd named dmdmd Bond 2.:
dddd. 33d «and ~53... emdwd Send 58d :23 womdd gowd 0:0..053m
dddd. 35nd Emnd meddd omddd 22d Sad .Vddd mmdwd 5:0...0502
dddd. dmwmd dnwmd amend amid ..mnd ~dddd 035d 502
dddd. mdmnd dcmhd 32d .dd..d mtdd- waded b0:
dddd. Shad amid mwmdd dnndd 39:. 505.00
dddd. n...d..d dcvdd Sodd .vvwd 00:00”.
dddd. amid 80nd mummd 50:00
dddd. engd medwd E:.m.0m
dddd. . mode 0..050=<
.. dddd. x050. .Eom. 28>» mm

0.: 2.: 05. as. 5%. be. 0.250 85E «850 520.00 0.803 505050

L08.3m 3.0502 283 mm
0850354353503
A33 5:02.83 E0555 00035 com no 2553— 0008a..— uo 3530.06
dddd. Seed 3.de waded dd.~d nvnmddnbvd ”End Send Send domed Nedvd ddFd dd
dddd. added 32.... dzvd mmvmd .mned Bond mmwnd Sand Emmd NBdmd Fahd .V..D
dddd. mwnbd mwdvd devd dwwmd .vddd Smdd 53d vwsd amend mdded 2.0.53.3m
dddd. chd vad waned emdbd nmced nde added ddnvd ddvnd 55.00502
dddd. .Svd 32d mmdmd dmmvd dvnmd 33d mutd mnwnd 59:.
dddd. :NNd .m.nd dwNnd ..mmd 33d .wd.d .NNnd b0:
88.. m~.md ndmmd mound c.mmd d.dnd Shed wcov. mcom
dddd. N..hd nd.md mendd .2de Nnnnd E0050
88.. 2de momnd wdmmd fixed 00:00“.
dddd. ~m~vd .wnnd domed 000000
88.. 25. 33.0 52000
dddd. 33d 0..0=m=<
I dddd. x00... 2.53 .09).
.m... 2.3 0:0. 5:0. :09: b0: mcov. E0500 00:00.... 000500 E:.w.0m 0..050=< x03.
405...; -3502 mac... 283 .Ums.

 

 

egg 582.000“ .505... 80.05 be.

.6008. a «Ea...

 

cm no 0533— 0008.."— uo 9530.00.80

81

Table 2
Tests of the Mean-variance Efficiency of the Equity Benchmarks

The sample period is January 1985-March 1994. The model is a. = Bo + ptB + e.
where a. are excess returns on country indices and p. are excess returns on the world
index and the 3-region benchmark. I30 is the regression intercept and PPWi is the
positive period weight measure for country i. The null and alternative hypotheses of
the GRS test are Bo=0 and 00 #0 respectively. The GRS statistic Fm has an F distribution
and the Wald test statistics, WALDMV and WALDPPw, are distributed as )6.

 

 

 

 

 

 

 

Benchmarks MSCI World Index

Countries [30 t statistics“ PPW z statisticsb
Australia 0.4476 0.5600 0.3124 0.3818
Belgium 0.8784 2.0285 ** 0.8581 1.9661 **
Canada -0.2684 -0.7634 -0.3026 08403
France 0.5683 1.2295 0.5553 1.2080
Germany 0.5075 0.9215 0.4533 0.8269
Hong Kong 1.3330 1.6570 1.1953 1.4752
Italy 0.5554 0.8552 0.5669 0.8718
Japan -0.0312 -0.0602 0.0345 0.0652
The Netherlands 0.6422 2.1421 ** 0.6145 2.0457 **
Switzerland 0.6290 1.6238 0.6025 1.5682
United Kingdom 0.2040 0.5123 0.1949 0.4922
United States 0.1539 0.4684 0.1203 0.3604
Joint tests on all countries Statistic P value Statistic P value
GRS Test: FMV (12,98) 2.2142 0.0165 **

WALDMV, WALDPPW (12) 27.9902 0.0056 *** 27.3102 0.007 ***
Subperiod GRS Test of the World Index

FMv: (12,281) Jan 70-Mar 94 1.7253 0.0611 *

FMV: (12,219) Feb 70-May 89 1.3469 0.1937 C

Fmv: (12,38) Jan 90-Mar 94 2.1485 0.0366 **

 

Significance levels: *** indicates 1 %, ** indicates 5%, * indicates 10%.

" t statistic computed using heteroscedastic consistent variance.

"’ 2 statistic computed using the GMM estimated variance.

c Harvey (1990) reported their p value of the GRS test to be 0.304 for 17 countries.

82

005...? 00.05.30 $.20 05 we? 50.2.58 0:505 N

..

0000...? 2.0.0.0000 owe-0500808.. mam: 00.2.58 0.3.0.0 . a

as... 8.8.0:. .. a... 8.8.0:. r. a... 3.8.0:. ..,: are. cascade

 

 

 

 

8.005505: 530.74”

3.00555: «00.00-..

32... 02.3. .02... 080.... t... 8...... 080.00 3...... 82...... 8.0.. sac-23.20.23
Nam... 9.2.. i 00...... 8.0.0 $0.31.... see my...

0:.0> n. 0.3.05 0:.0> A. 0.6.05 0:.0> n. 33.0% 0:.0> A. 0.3.05 00.5580 ..0 ..o 300. 2......
a: a. s. a: 8.0m XE... .. 83. 0.3.... mesm- 8......
$8...- St... «E...- 88? .08...- 0.8...- 0%.... 8...... 580:... use:
0%.... 230... 23... 080... 200.. 3...... 03.0.. 43.... 0.3.55.3
2%.. 80...... 00%.. EA... .. an... 82.... : 3.0.0 8...... €522.02 2...
s. s. a. a. 8.....- RS...- 33? $5....- 5%.
3...... men... can... 05... a: .... 3...... 3%... 29.... ..s.
an... 2%... 28.. 8...... .. on... as... r .800 83.. M33. 0.5..
38...- 3.8...- 008? 2......- 083 22... 40...... $0.... .5530
SR... 83... 82... EN... 89... $8... Es... 08...... ooze“.
.. 80...- 8......- .. ENE. 03.....- r 033- 3.....- WE...- oa....- 0850
00%.. 8%... 28.. .30... 0.2.. 2%... $8.. 28... 520.00
2.3...- 3......- 880- $8...- $3.... 38... 09...... 83... 0.05:...
.8550... N 3.... a 00.3.0.0 . on .8956 N 3.... a00.3.0.0. om. 00.5500
0.00333 250553.332 2.0.5.03.—

 

..0..=8. a 2.3

83

 

some... N35... coco. 0000.0 ommmd- 9.5...- 0....00.. b.0am...0:..00w00> 0....00...Um2

 

$8... .02.... 8...... 2.0.... 0......- SS..- 2.950 03.... .0. 220.; B... 2.2.... .00.).
5.0.0 008... 003.0 0.00... .02... 3 m0...- .00..0.2 ..00.m .0.0..- x000. 0.0:w00> 0000 0.20:3
000% .000 00 .. NM. 0..0..0.0. can. 00...... 00.800. 4.0.5.003.
0000....m 00..00000..-.- 00000000....- 0.00.0.0.

.08... .00..w00> 00.0000w0y.

00.00.05

 

003.0010... 0.09m .50..- x000. 0000m00> 0...
..0. 000. £000.). ..w00.... N02 .03. 0.0... .000 0.000. 0.000.. .000 000.0... 3.0.5 .000..00..0.0.
.0000w0> 0... ..0. 000. 00.0.). .3000... O00. b0. .00.. 0.0 000.00.. 030.00 0.....- .0000.
0.0... 0. 00.0..0 .0 0000000.. .0. 00. 02.00000... 0 00w00..0 000w00> .00N...0.. w50.0.. 000.. 0.... 0
0 ..0>0 00N....00.0 00. 00000000.. 0.00w00> 0... 00... an .0 00.0> 0.0.0000 0... 0. .000 00000000..
.0.0. 00.00000 0....- .v.-.0.0..000.. 0.000. 0... 00 000.0. 0008.0 3.0.0.0 0... 0. 0.. .000 . 000.

x000. 00 000.0. 0008.0 3....000. 0... 0. ..0. 0.0..3 ..0 + ~... 0... + aw u ..0. 0. .0000. 00.00000 0......

0030.020...— 0.0OU 00300000....
m 0.00...

Table 4

84

International Equity Mutual Fund Performance: Mean-variance Tests

The sample period is January 1985-March 1994. The model is a, = [30 + p.13 + et where a. are
excess returns on mutual funds and p, are excess returns on the benchmarks. [30. is the Jensen measure
for fund i. The null and alternative hypotheses of the GHM test are [30 = 0 and [30 2 0 respectively.
The GHM statistic, KTMV is asymptotically distributed as a weighted mixture of 12.

The null hypothesis of the GRS test is Bo=0 versus the alternative hypothesis that 13., $0. The
GRS test statistic Fm, has an F distribution. The Wald statistics WALDMV is distributed as )8.

 

 

 

 

Erasmus; 12W il
lntemational Mutual Funds Jensen tstatistics‘ Jensen tstatistics"
Alliance Canadian -0.1429 -0.7958 -0.4490 -1.3351
Alliance Global Small Cap A -0.4245 -1.5422 -0.4228 -1 .8884 ‘
Alliance lntemational A -0.3948 -1.7308 ‘ 0.4192 0.9778
Bailard, Biehl lntl Equity -0.6014 -3.1215 '" 0.3280 0.7070
Centerland Kleinwrt lntl EqB -0.2742 -1.4852 0.5639 1.2972
Dean Witter WorldWide lnvmnt -0.4079 -3.7868 "s 0.2451 0.8386
EuroPacific Growth -0.0454 —0.3289 0.6164 1.8461 ‘
Fidelity Overseas -0.0035 -0.0135 0.8300 1.7103 '
First lnvest Global -0.4175 -1.6140 0.2879 0.7095
FT lntemational Equity A -0.4473 -2.5249 " 0.5072 1.1567
G.T. Global New Pacific Gr A -0.2636 -0.7876 0.6119 1.2031
IDS lntemational -0.4754 -1.9538 ‘ 0.3995 0.8769
lnvesco Pacific Basin -0.5798 -2.0346 ” 0.4127 0.7067
Japan -0.1 168 -0.4097 0.8778 1.3619
Kemper lntemational -0.3314 -1.8106 " 0.4880 1.2193
Keystone lntemational -0.5857 -3.4935 *” 0.2733 0.6519
Merrill Lynch Global Holdg A -0.2441 -2.3907 " 0.2613 1.1244
Merrill Lynch Pacific A 0.0254 0.0581 0.9176 1.5297
New Perspective -0.0430 -0.3874 0.3575 1.6436
Oppenheimer Global A -0. 1623 -0.6810 0.591 1 1.4294
PaineWebber Atlas Global GrA -O.2591 -1.3815 0.3952 1.0900
Princor World -O.5593 -2.0635 " 0.1219 0.2957
Prudential Global B -0.4168 -1.9244 * 0.2518 0.6538
Putnam Global Growth A -0.1106 -0.6510 0.4469 1.5758
RSI Retrmnt lntl Equity -0.4051 -2.5390 " 0.4404 1.0195
Scudder lntemational -0.3002 -1.8660 ' 0.5230 1.3065
Smith Barney Shear Glob OppA -0.7335 -4.2570 "s -0.1269 -0.3945
T. Rowe Price lntl Stock -0.2282 -1.6530 0.6742 1.6298
Templeton Foreign 0.1653 1.3342 0.6239 2.2808 "
Templeton Growth -0.0653 -0.5140 0.2608 1.2417
Templeton Smaller Comp Grth -0.2759 -1.3061 0.0806 0.3264
Templeton World -0.1126 -0.8167 0.1252 0.7238
United lntemational Growth -0.2602 -l.6864 * 0.4180 1.1919
Vanguard lntl Growth -0.2606 -1 .5313 0.6447 1.4595
Vanguard/Trustees eqy lntl -0.0974 -0.6772 0.6125 1.6620 ‘
Joint performance test on all funds Statistic P value Statistic P value
GHM Test: KTMV (l to 35) 27.22 0.6700 46.87 0.0633 "
GRS Test: Fm (35;64,75,75,73) 1.84 0.0174 " 1.80 0.0176 ”
'Wald Test: WALDMV (35) 120.24 0.0000 *“ 96.67 0.0000 ‘“

 

Significance levels: "" indicates 1%, " indicates 5%, " indicates 10%.

"t statistic computed using heteroscedastic consistent variance.

‘5! ll: III I I I I II II

 

 

 

85

 

 

 

 

Table 4 (cont'd).
Benchmarks W W
lntemational Mutual Funds Jensen t statistics' Jensen tstatistics'
Alliance Canadian -0.4271 -1.0178 -0.6122 -1.9157 ‘
Alliance Global Small Cap A -0.2664 -0.5462 -0.4501 -2.0608 ”
Alliance lntemational A 0.11 16 0.3731 -0.1629 -0.7091
Bailard, Biehl Intl Equity -0.1480 -0.5961 -0.2539 -1.2326
Centerland Kleinwrt lntl EqB 0.1915 0.7448 -0.0146 -0.0688
Dean Witter WorldWide lnvmnt 0.0169 0.0937 -0.1599 -1.2503
EuroPacific Growth 0.4075 1.5534 0.1403 0.8332
Fidelity Overseas 0.3773 1.2485 0.2524 0.9616
First Invest Global 0.0004 0.0015 -0.1887 -0.7635
FT lntemational Equity A 0.1440 0.5361 -0.0713 -0.3184
G.T. Global New Pacific Gr A 0.2943 0.7528 0.1162 0.3017
IDS lntemational 0.0381 0.1323 -0.1671 -0.6636
Invesco Pacific Basin -0.0502 -0.1379 -0.1145 -0.2869
Japan 0.1908 0.4532 0.3198 1.0913
Kemper lntemational 0.1 867 0.6824 -0.0634 -0.2983
Keystone lntemational -0.1465 -0.7086 -0.2592 -1 .4246
Merrill Lynch Global Holdg A 0.0800 0.5097 -0.0751 -0.8235
Merrill Lynch Pacific A 0.3928 1.0196 0.4383 1.0313
New Perspective 0.2592 1.1161 0.0301 0.2850
Oppenheimer Global A 0.3723 1.0477 0.0375 0.1629
PaineWebber Atlas Global GrA 0.1386 0.5303 -0.0651 -0.3060
Princor World -0.0493 -0.1294 —0.3842 -1 .3906
Prudential Global B -0.0432 -0.1715 -0.1881 -0.8132
Putnam Global Growth A 0.2756 1.1347 0.0122 0.0866
RSI Retrmnt lntl Equity 0.0066 0.0315 -0.1463 -0.9481
Scudder lntemational 0.2 1 85 0.8269 -0.0487 -0.2772
Smith Barney Shear Glob OppA -0.3191 -1.1719 -0.5453 -2.8776 ”‘
T. Rowe Price lntl Stock 0.3098 1.3457 0.0740 0.4832
Templeton Foreign 0.4724 1.9835 “ 0.2337 1.6774 "
Templeton Growth 0.2566 0.8466 0.0527 0.3103
Templeton Smaller Comp Grth 0.1 128 0.3014 -0.0630 -0.2754
Templeton World 0.1360 0.4835 -0.0488 -0.3561
United lntemational Growth 0.1335 0.6079 -0.0636 -0.3 702
Vanguard Intl Growth 0.2338 0.9700 0.0000 -0.0003
Vanguard/Trustees eqy lntl 0.2906 1.3557 0.0961 0.6236
Joint performance test on all funds Statistic P value Statistic P value
GHM Test: KTMV (l to 35) 46.53 0.0534 ‘ 37.51 0.2128
GRS Test: FMV (35;64,75,75,73) 1.71 0.0263 " 1.84 0.0147 "
Wald Test: WALDMV (35) 94.47 0.0000 ‘” 109.40 0.0000 "‘

 

Significance levels: *" indicates 1%, " indicates 5%, ‘1' indicates 10%.

' t statistic computed using heteroscedastic consistent variance.

86
Table 5

International Equity Mutual Fund Performance: PPW Measures

The sample period is January 1985-March 1994. PPWi is the Positive Period Weight
Measure for fund i. The null hypothesis is PPW = 0 versus the alternative hypothesis that

PPW ¢ 0. The Wald statistic WALDPPW is distributed as x2-

 

 

 

 

Intemational Mutual Funds PPW z statisticsa PPW 2 statistics‘
Alliance Canadian -0.1923 -1.1338 -0.4713 -1.3977 ‘
Alliance Global Small Cap A 04732 -1.8248 ” -0.4301 -1.9216 ”
Alliance lntemational A -0.4477 -2.0255 " 0.3620 0.8632
Bailard, Biehl IntI Equity -0.5894 -3.2234 ‘" 0.2905 0.6330
Centerland Kleinwrt Intl EqB -0.3371 -2.0016 " 0.5122 1.1998
Dean Witter WorldWide lnvmnt -0.4077 -3.9218 ‘“ 0.2179 0.7576
EuroPacific Growth -0.0582 -0.4386 0.5683 1.7417 ”
Fidelity Overseas -0.0609 -0.2712 0.7915 1.6409 ‘
First Invest Global -0.4407 -1.7998 ” 0.2632 0.6546
FT lntemational Equity A -0.5043 -3.1415 "' 0.4485 1.0462
G.T. Global New Pacific Gr A -0.2430 -0.7943 0.5509 1.1012
IDS lntemational -0.5385 -2.6864 ‘" 0.3317 0.7415
Invesco Pacific Basin -0.6570 -2.7661 ‘" 0.3399 0.5913
Japan -0.1992 -0.7370 0.8518 1.3380 "
Kemper lntemational -0.3854 -2.3295 "" 0.4359 1.1089
Keystone lntemational -0.6213 -3.8635 ‘” 0.2389 0.5780
Merrill Lynch Global Holdg A -0.2229 -2.2099 " 0.2481 1.0758
Merrill Lynch Pacific A -0.0864 -0.2420 0.8473 1.4284 '
New Perspective -0.0193 -0.1869 0.3390 1.5740 ‘
Oppenheimer Global A -0.1787 -0.8274 0.5369 1.3188 ‘
PaineWebber Atlas Global GrA -0.2987 -1.8070 ” 0.3438 0.9648
Princor World -0.6413 -2.6464 '” 0.0522 0.1300
Prudential Global B -0.4295 -2.2930 " 0.2119 0.5539
Putnam Global Growth A -0.0699 -0.4225 0.4226 1.5095 '
RSI Retrmnt lntl Equity -0.4066 -2.7729 "‘ 0.3995 0.9360
Scudder lntemational -0.3539 -2.4766 ‘" 0.4664 1.1927
Smith Barney Shear Glob OppA -0.7827 -5.0104 "* -0.1776 -0.5631
T. Rowe Price lntl Stock -0.2529 -2.0601 " 0.6282 1.5463 *
Templeton Foreign 0.1588 1.3632 ’ 0.5936 2.2011 “
Templeton Growth -0.0514 -0.4538 0.2389 1.1510
Templeton Smaller Comp Grth -0.3252 -1.6044 ‘ 0.0499 0.2054
Templeton World -0.0828 -0.6704 0.1174 0.6835
United lntemational Growth -0.2559 -1.6498 ” 0.3911 1.1305
Vanguard lntl Growth -0.2793 -1.7620 ” 0.5962 1.3709 ‘
Vanguard/Trustees eqy Int] -0. 1034 -0.7651 0.5805 1.6026 ‘
Joint performance test on all funds Statistic P value Statistic P value
Wald Test: WALDMV (35) 100.77 0.0000 t" 95.65 0.0000 ‘"

 

Significance levels: ""‘ indicates 1%, ““" indicates 5%, "‘ indicates 10%.
' PPW coefficients and 2 statistics are computed using the generalized method of moments.

87

 

 

 

 

Table 5 (cont'd).
Benchmarks mm W
lntemational Mutual Funds PPW 2 statistics” PPW z statistics8|
Alliance Canadian -0.4726 -1.1001 -0.6399 -1.9997 "
Alliance Global Small Cap A -0.3330 -0.6787 -0.4713 -2.1570 "
Alliance lntemational A 0.05 34 0.1805 -0.2076 -0.9406
Bailard, Biehl Intl Equity -0.1629 -0.6585 -0.2735 -1.3527 '
Centerland Kleinwrt Intl EqB 0.1499 0.5853 -0.0623 -0.3086
Dean Witter WorldWide lnvmnt -0.0142 -0.0782 -0.1769 -1.4114 *
EuroPacific Growth 0.3590 1.3615 * 0.0994 0.6226
Fidelity Overseas 0.3634 1.2033 0.2120 0.8200
First Invest Global -0.0303 -0. 1079 -0.2101 -0.8622
FT lntemational Equity A 0.0968 0.3609 -0.1225 -0.5775
G.T. Global New Pacific Gr A 0.2442 0.6255 0.0644 0.1733
IDS lntemational -0.0166 -0.0565 -0.2444 -1.0340
Invesco Pacific Basin -0.1136 -0.3126 -0.2181 -0.5840
Japan 0.2049 0.4807 0.2891 0.9972
Kemper lntemational 0.1410 0.5138 -0.1034 -0.5039
Keystone lntemational -0.1644 -0.7991 -0.2851 -1.5980 *
Merrill Lynch Global Holdg A 0.0613 0.3874 -0.0719 -0.7760
Merrill Lynch Pacific A 0.3603 0.9195 0.3311 0.8149
New Perspective 0.2262 0.9627 0.0216 0.2059
Oppenheimer Global A 0.3109 0.8653 -0.0122 -0.0549
PaineWebber Atlas Global GrA 0.0902 0.3359 -0.1 165 -0.5707
Princor World -0.1231 -0.3185 -0.4460 -1.6917 "
Prudential Global B -0.0776 -0.2994 -0.2355 -1.0409
Putnam Global Growth A 0.2398 0.9954 0.0088 0.0627
RSI Retrrnnt Intl Equity -0.0178 -0.0854 -0.1793 -1.1834
Scudder lntemational 0.1660 0.6279 -0.1000 -0.6155
Smith Barney Shear Glob OppA -0.3730 -1.3226 * -0.5936 -3.2297 "t
T. Rowe Price Int1 Stock 0.2746 1.1958 0.0408 0.2808
Templeton Foreign 0.4360 1.8174 " 0.2104 1.5413 *
Templeton Growth 0.2064 0.6758 0.0325 0.1964
Templeton Smaller Comp Grth 0.0495 0.1311 -0.1046 -0.4740
Templeton World 0.0961 0.3413 -0.0518 03838
United lntemational Growth 0.1084 0.4972 -0.0736 -0.4333
Vanguard Intl Growth 0.2004 0.8323 -0.0326 -0.2032
Vanguard/Trustees' eqy Intl 0.2632 1.2501 0.0765 0.5081
Joint performance test on all funds Statistic P value Statistic P value
Wald Test: WALDMV (35) 92.97 0.0000 ***" 104.82 0.0000 'm

 

Significance levels: *** indicates 1%, " indicates 5%, "‘ indicates 10%.

a PPW coefficients and 2 statistics are computed using the generalized method of moments.

88

Table 6

lntemational Equity Mutual Funds: Comparison of Performance Measures

The sample period is January 1985-March 1994. The Treynor-Mazuy model is a. = m. + p.13. + pfm, + e, where a.
are excess returns on mutual funds and p. is excess returns on the world index. am is fund i's regression coefficient for
the squared excess returns on the world index. The Jensen measure model is a. = 130 + p.13 + q where [30; is the Jensen
measure for fund i. PPWi is the Positive Period Measure for fund i.

 

 

Security Treynor- PPW Jensen

WW1! Timing Selectivity Mazuy Measure Measure
lntemational Mutual Funds m, t statistics' to, * var(p.) tn. mo+m;‘var(p.) PPW Jensen
Alliance Canadian -0.0262 -l.8196 ' -0.5396 0.1471 -0.3925 -0.4726 -0.4271

Alliance Global Small Cap A -0.0393 -3.I813 "‘ -0.8093 0.5938 -0.2155 -0.3330 -0.2664

Alliance lntemational A 00364 -5.4382 '" -0.7496 0.9095 0.1599 0.0534 0.1116

Bailard, Biehl lntl Equity -0.0097 -1.4891 -0.1998 0.0652 -0.1346 -0.l629 -0.1480

Centerland Kleinwrt lntl EqB -0.0251 4.9734 "‘ -0.5169 0.7413 0.2244 0.1499 0.1915

Dean Witter WorldWide lnvmnt -0.0199 -6.3501 "‘ -0.4098 0.4539 0.0441 -0.0142 0.0169

EuroPacific Growth -0.0297 -5.0517 "‘ -0.6116 1.0580 0.4464 0.3590 0.4075

Fidelity Overseas -0.0060 -0.8047 -0.1236 0.5092 0.3856 0.3634 0.3773

First Invest Global -0.0175 .25454 " -0.3604 0.3841 0.0237 -0.0303 0.0004

FT lntemational Equity A 00294 -5.9528 '“ -0.6055 0.7876 0.1821 0.0968 0.1440

G.T. Global New Pacific Gr A -0.0313 4.2858 “' -0.6446 0.9793 0.3347 0.2442 0.2943

IDS lntemational -0.0329 4.4688 "" -0.6775 0.7579 0.0804 -0.0166 0.0381

Invesco Pacific Basin -0.0408 -6.8023 '” -0.8402 0.8431 0.0029 -0.1136 -0.0502

Japan 0.0056 0.4035 0.1 153 0.0687 0.1840 0.2049 0.1908
Kemper lntemational -0.0286 -5.2405 "‘ -0.5890 0.8122 0.2232 0.1410 0.1867

Keystone International -0.0137 -3.2261 "' -0.2821 0.1536 -0.1285 -0.1644 -0.1465

Merrill Lynch Global Holdg A -0.0112 -3.3428 "' -0.2307 0.3248 0.0941 0.0613 0.0800

Merrill Lynch Pacific A 00185 -1.5534 -0.3810 0.7988 0.4178 0.3603 0.3928
New Perspective -0.0191 -3.3037 "" -0.3933 0.6765 0.2832 0.2262 0.2592
Oppenheimer Global A -0.0345 4.2366 "' -0.7105 1.1288 0.4183 0.3109 0.3723
PaineWebber Atlas Global GrA -0.0281 -3.8960 "‘ -0.5787 0.7549 0.1762 0.0902 0.1386
Princor World -0.0469 4.5228 ”‘ -0.9658 0.9774 0.0116 -0.1231 -0.0493
Prudential Global B -0.0204 -2.2780 " -0.4201 0.4039 -0.0162 -0.0776 -0.0432

Putnam Global Growth A -0.0213 -5.6017 "" -0.4386 0.7412 0.3026 0.2398 0.2756

RSI Retrmnt lntl Equity -0.0158 -2.9238 "" -0.3254 0.3531 0.0277 -0.0178 0.0066

Scudder lntemational -0.0318 -7.1088 '” -0.6549 0.9157 0.2608 0.1660 0.2185

Smith Barney Shear Glob OppA -0.0313 -3.6325 "' -0.6446 0.3663 -0.2783 -0.3730 0319]

T. Rowe Price lntl Stock -0.0221 4.5539 '" -0.4551 0.7933 0.3382 0.2746 0.3098

Templeton Foreign -0.0219 4.0042 '" -0.4510 0.9526 0.5016 0.4360 0.4724

Templeton Grth -0.0304 4.1704 "" -0.6261 0.9230 0.2969 0.2064 0.2566

Templeton Smaller Comp Grth -0.0388 4.4024 '" -0.7990 0.9623 0.1633 0.0495 0.1128

Templeton World -0.0246 4.1187 "' -0.5066 0.6750 0.1684 0.0961 0.1360

United lntemational Growth -0.0184 -5.2473 "' -0.3789 0.5372 0.1583 0.1084 0.1335

Vanguard Intl Growth -0.0212 -3.1448 "' -0.4366 0.6984 0.2618 0.2004 0.2338

Vanguard/Trustees' eqy lntl -0.0175 4.6920 ”' -0.3604 0.6730 0.3126 0.2632 0.2906

 

Significance levels: "‘ indicates 1%, " indicates 5%, " indicates 10%.

' t statistic computed using heteroscedastic consistent variance.

89

Table 6 (cont'd).

 

 

Security Treynor- PPW Jensen
WWW Timing Selectivity Mazuy Measure Measure
lntemational Mutual Funds tau2 t statistics' us, * var(p,) tn0 m0+mz‘var(p.) PPW Jensen
Alliance Canadian -0.0109 -2.0483 “ -0.2257 -0.1843 -0.4100 -0.4713 -0.4490
Alliance Global Small Cap A .00051 -1.5084 -0. 1056 -0.2986 -0.4042 -0.4301 -0.4228
Alliance lntemational A -0.0280 -5.3479 "' -0.5797 1.1002 0.5205 0.3620 0.4192
Bailard, Biehl Intl Equity -0.0173 -3.7629 '“ -0.3582 0.7482 0.3900 0.2905 0.3280
Centerland Kleinwrt lntl EqB -0.0242 -5.9691 m -0.501 1 1.1519 0.6508 0.5122 0.5639
Dean Witter WorldWide lnvmnt -0.0135 4.9642 '“ -0.2795 0.5728 0.2933 0.2179 0.2451
EuroPacific Growth -0.0235 -6.9835 '“ -0.4866 1.1881 0.7015 0.5683 0.6164
Fidelity Overseas -0.0160 -2.1589 " -0.3313 1.2184 0.8871 0.7915 0.8300
First Invest Global -0.0104 -1.7580 ‘ -0.2153 0.5417 0.3264 0.2632 0.2879
FT lntemational Equity A 00282 —7.5190 "" -0.5839 1.1929 0.6090 0.4485 0.5072
G.T. Global New Pacific Gr A -0.0297 -6.6078 ... -0.6149 1.3350 0.7201 0.5509 0.6119
IDS lntemational .00317 -7.7060 ... 06563 1.1704 0.5141 0.3317 0.3995
lnvesco Pacific Basin -0.0353 -5.2962 ... -0.7309 1.2714 0.5405 0.3399 0.4127
Japan 0.0125 -1.9707 ' -0.2588 1.1818 0.9230 0.8518 0.8778
Kemper lntemational -0.0250 -6.7444 ... .05176 1.0965 0.5789 0.4359 0.4880
Keystone lntemational ~0.0169 -3.8272 '“ ~0.3499 0.6838 0.3339 0.2389 0.2733
Merrill Lynch Global Holdg A 00066 -2.6808 ”‘ -0.1367 0.4217 0.2850 0.2481 0.2613
Merrill Lynch Pacific A -0.0310 -2.8933 '“ -0.6418 1.6708 1.0290 0.8473 0.9176
New Perspective -0.0088 4.5339 ”‘ -0.1822 0.5724 0.3902 0.3390 0.3575
Oppenheimer Global A 00239 -5.6742 ”" -0.4948 1.1725 0.6777 0.5369 0.5911
PaineWebber Atlas Global GrA -0.0234 -7.3749 ”‘ -0.4845 0.9656 0.4811 0.3438 0.3952
Princor World -0.0365 -5.1782 ... -0.7557 1.0105 0.2548 0.0522 0.1219
Prudential Global B -0.0183 -2.6340 ”‘ -0.3789 0.6963 0.3174 0.2119 0.2518
Putnam Global Growth A -0.01 17 -3.6006 ”‘ -0.2422 0.7310 0.4888 0.4226 0.4469
RSI Retrrnnt lntl Equity -0.0189 4.4168 "‘ -0.3913 0.9012 0.5099 0.3995 0.4404
Scudder lntemational -0.0267 -8.4078 ... -O.5528 1.1733 0.6205 0.4664 0.5230
Smith Barney Shear Glob OppA -0.0236 -8.6212 ... —0.4886 0.4483 -0.0403 -0.1776 01269
T. Rowe Price lntl Stock 00221 -6.0155 '“ -0.4576 1.2120 0.7544 0.6282 0.6742
Templeton Foreign -0.0149 -6.3246 ’” -0.3085 0.9871 0.6786 0.5936 0.6239
Templeton Growth 00120 4.8896 ... -0.2485 0.5525 0.3040 0.2389 0.2608
Templeton Smaller Comp Grth -0.0169 -7.3768 ”‘ -0.3499 0.4925 0.1426 0.0499 0.0806
Templeton World -0.0057 -2.5109 “ 0.1180 0.2632 0.1452 0.1174 0.1252
United lntemational Growth -0.0151 4.9520 M 03126 0.7859 0.4733 0.3911 0.4180
Vanguard lntl Growth -0.0230 -5.3324 “' -0.4762 1.2037 0.7275 0.5962 0.6447
Vanguard/Trustees eqy lntl -0.0152 4.5267 ”' -0.3I47 0.9816 0.6669 0.5805 0.6125

 

Significance levels: ‘“ indicates 1%, " indicates 5%, ‘ indicates 10%.

' t statistic computed using heteroscedastic consistent variance.

90

Table 7
lntemational Equity Mutual Fund Performance: Simultaneous Hedging Strategy

The sample period is January 1985-March 1994. The model is at = [30 + p,B + e, where a. are excess returns on
mutual funds and p. are excess returns on the benchmarks and three forward contracts: the Deutschmark, Japanese Yen
and the Canadian Dollar. 00. is the Jensen measure for fund i. The null and alternative hypotheses of the GHM test

are 0. = 0 and [30 2 0 respectively. The GIIM statistic. KTMV is asymptotically distributed as a weighted mixture of xz.

The null hypothesis of the GRS test is 150:0 versus the alternative hypothesis that 130 $0. The GRS test statistic FMV has
F distribution. The Wald statistics WALDMV is distributed as 12.

-_

 

 

 

 

lntemational Mutual Funds Jensen t statistics' Jensen t statistics' Jensen tstatistics'
Alliance Canadian -0.1690 -0.9596 -0.3345 -1.0681 -0.7138 -2.5554 ”
Alliance Global Small Cap A -0.2048 -0.7875 0.2049 0.6074 -0.3573 -1.7141 ‘
Alliance lntemational A 02966 -1.3212 0.1522 0.5645 -0.0779 -0.3570
Bailard, Biehl lntl Equity -0.6916 -3.3823 ‘“ -0.2826 -1.1817 -0.2831 -1.3912
Centerland Kleinwrt lntl EqB -0.3040 -1.6555 0.1465 0.6035 -0.0161 -0.0818
Dean Witter WorldWide lnvmnt -0.4245 -3.6411 '" 0.0571 0.3581 -0.I715 -1.3377
EuroPacific Growth 00230 -0.1605 0.4496 2.0185 ” 0.1571 1.0088
Fidelity Overseas -0.0662 -0.2618 0.2055 0.6897 0.1349 0.5133
First Invest Global -0.3036 -1. 1931 0.0728 0.2799 -0.I411 -0.5808
FT lntemational Equity A ~0.4494 -2.4367 “ 0.1168 0.4791 -0.0535 -0.2560
G.T. Global New Pacific Gr A ~0.3376 -0.9700 0.2989 0.8221 0.1896 0.5383
IDS lntemational -0.4029 -1.8037 ‘ 0.0345 0.1409 -0.1026 -0.4556
Invesco Pacific Basin -0.5020 -1.9112 ' -0.0222 -0.0725 0.0229 0.0683
Japan -0.1657 -0.5705 -0.0624 -0. 1719 0.3431 1.2085
Kemper lntemational -0.3293 4.8376 ° 0.1872 0.7359 -0.0422 -0.2219
Keystone lntemational -0.6276 -3.6685 ”° -0.2448 -1.2611 -0.2799 -1.5845
Merrill Lynch Global Holdg A 02366 -2.2814 " 0.1541 1.1050 -0.0780 081%
Merrill Lynch Pacific A -0.1307 -0.3280 0.1429 0.4303 0.2358 0.6076
New Perspective -0.0092 -0.0806 0.3861 2.0611 “ 0.0316 0.2893
Oppenheimer Global A 00512 -0.2234 0.4700 1.5630 0.0973 0.4533
PaineWebber Atlas Global GrA -0.2554 -1.3594 0.1885 0.8467 -0.0592 -0.3045
Princor World -0.3806 -1.4858 0.1716 0.6003 -0.1610 -0.6954
Prudential Global B 04323 -2.0293 “ -0.0748 -0.3308 -0.2596 -l.1902
Putnam Global Growth A 01448 -0.8506 0.3371 1.5084 -0.0340 -0.2371
RSI Retrrnnt Intl Equity -0.4221 -2.5884 °° -0.0997 -0.5149 -0.1535 -1.0630
Scudder lntemational o0.2751 -l.7172 ° 0.2139 0.9467 -0.0177 -0.1119
Smith Barney Shear Glob OppA -0.7006 4.2669 ”° -0.2254 -1.1107 -0.5304 -3.1001 '”
T. Rowe Price lntl Stock 02544 -1.9605 ‘ 0.2623 1.2254 0.0708 0.5224
Templeton Foreign 0.1782 1.4341 0.5487 2.7012 ‘” 0.2693 1.9748 ‘
Templeton Growth 00033 -0.0280 0.5172 2.5334 “ 0.1439 0.9484
Templeton Smaller Comp Grth —0.1382 -0.6812 0.4411 1.6983 ° 0.0569 0.2726
Templeton World -0.0203 -0.1443 0.4031 2.0455 ” 0.0218 0.1545
United lntemational Growth -0.2000 -1.1797 0.1609 0.7407 0.0072 0.0394
Vanguard lntl Growth -0.2904 -1.7112 ° 0.1457 0.6540 -0.0060 -0.0408
Vanguardffrustees' eqy lntl -0. 1046 -0.6954 0.2238 1.1370 0.0986 0.6510
Joint performance test on all funds Statistic P value Statistic P value Statistic P value
GHM Test: KTMV (1 to 35) 26.78 0.6901 46.35 0.0573 ‘ 35.06 0.2978
GRS Test: Fm (35:61,72.70) 1.61 0.0504 ‘ 1.66 0.0360 " 1.67 0.0340 ”
Wald Test: WALDMV (35) 102.55 0.0000 ”° 99.92 0.0000 “‘ 105.66 0.0000 "‘

 

Significance levels: ‘” indicates 1%, " indicates 5%. ‘ indicates 10%.

' t statistic computed using heteroscedastic consistent variance.

Table 7 (cont'd).

91

The sample period is January 1985-March 1994. PPWi is the Positive Period Weight Measure for fund i.

The null hypothesis is PPW = 0 versus the alternative hypothesis that PPW at 0. The Wald statistic

WALDPPW is distributed as x2.

 

 

 

 

lntemational Mutual Funds PPW 2 statistics' PPW 2 statistics' PPW 2 statistics’
Alliance Canadian -0.1372 -0.8524 -0.3132 -1.0064 -0.7009 -2.5141 "*
Alliance Global Small Cap A -0.2079 -0.8522 0.2336 0.6964 -0.3404 -1.6680 "
Alliance lntemational A -0.1994 -0.9924 0.1442 0.5392 -0.0752 -0.3573
Bailard, Biehl Intl Equity -0.5723 -3.1692 *” -0.2801 -1.1665 -0.2619 -1.3365 '
Centerland Kleinwrt lntl EqB -0.2666 -1.6l43 ' 0.1481 0.6054 -0.0141 -0.0735
Dean Witter WorldWide lnvmnt -0.4023 -3.7293 "" 0.0544 0.3438 -0.1805 -1.4414 '
EuroPacific Growth 0.0203 0.1470 0.4459 2.0023 *' 0.1629 1.0771
Fidelity Overseas -0.0602 -0.2870 0.2141 0.7110 0.1184 0.4732
First Invest Global -0.3465 -1.5415 ' 0.0873 0.3413 -0.1168 -0.4951
FT lntemational Equity A -0.4323 -2.5992 “' 0.11 10 0.4568 -0.0577 -0.2876
G.T. Global New Pacific Gr A -0.1584 -0.5048 0.2998 0.8244 0.2134 0.6378
IDS lntemational -0.3701 -1.9301 "’ 0.0188 0.0775 -0.1371 v0.6447
Invesco Pacific Basin -0.5035 -2.2037 " -0.0264 -0.0879 -0.0108 -0.0338
Japan -0.1695 -0.6417 -0.0692 .0.1915 0.3606 1.3035
Kemper lntemational -0.3502 -2.1902 "' 0.1753 0.6874 -0.0338 -0.1859
Keystone lntemational -0.6382 -3.9985 '“ -0.2630 -1.3428 ‘ -0.2985 -1.6972
Merrill Lynch Global Holdg A -0.1990 -2.0501 "' 0.1559 1.1255 -0.0710 -0.7447
Merrill Lynch Pacific A -0.1279 -0.4021 0.1544 0.4699 0.1767 0.4889
New Perspective 0.0253 0.2305 0.3829 2.0486 ” 0.0284 0.2594
Oppenheimer Global A -0.0249 -0.1252 0.4652 1.5524 " 0.0908 0.4320
PaineWebber Atlas Global GrA -0.2306 -1.4366 ' 0.1756 0.7949 -0.1005 -0.5463
Princor World -0.4233 -1.7931 “ 0.1450 0.5069 -0.2059 -0.9358
Prudential Global B -0.3973 -2.2159 " -0.0673 -0.3006 -0.2906 -1.3912
Putnam Global Growth A -0.0861 -0.5637 0.3356 1.5184 " -0.0295 -0.2124
RSI Retrmnt Intl Equity -0.3515 -2.3952 *" -0.1189 -0.6164 -0.1622 -1.1516
Scudder lntemational -0.2399 -1.6243 " 0.2114 0.9440 -0.0240 -0.1628
Smith Barney Shear Glob OppA -0.6774 4.61 17 '" -0.2268 -1.1250 -0.5361 -3.2852
T. Rowe Price Intl Stock 02282 -2.0301 “ 0.2580 1.2023 0.0749 0.5833
Templeton Foreign 0.1827 1.5544 ' 0.5448 2.6797 '” 0.2711 1.9794
Templeton Growth -0.0049 -0.0457 0.5137 2.5178 ”" 0.1396 0.9055
Templeton Smaller Comp Grth -0.1484 -0.7381 0.4474 1.7303 "' 0.0541 0.2603
Templeton World -0.0031 -0.0241 0.4095 2.0907 " 0.0371 0.2610
United lntemational Growth -0.2480 -1.5321 ' 0.1402 0.6494 -0.0014 -0.0078
Vanguard Intl Growth -0.2286 -1.4655 " 0.1333 0.5964 0.0070 0.0491
Vanguard/Trustees eqy Intl -0.0909 -0.6414 0.2189 1.1211 0.1021 0.6896
Joint performance test on all funds Statistic P value Statistic P value Statistic P value
Wald Test: WALDMV (35) 95.24 0.0000 *"‘ 97.75 0.0000 "" 97.97 0.0000 '”

 

Significance levels: “" indicates 1%, " indicates 5%, ‘ indicates 10%.

' PPW coefficients and 2 statistics are computed using the generalized method of moments.

92

Table 8
International Equity Mutual Fund Performance: Unitary Hedging Strategy

The sample period is January 1986-March 1994. The model is a, = Bo + p.13 + e. where a. are excess returns on
mutual funds and p. are unitary hedged excess returns on the benchmarks. B.» is the Jensen measure for fund 1. The null

and altemative hypotheses of the GHM test are I}. = 0 and B, 2 0 respectively. The GHM statistic, KTMV is

asymptotically distributed as a weighted mixture of xi.
The null hypothesis of the GRS test is 130:0 versus the alternative hypothesis that 130 :0. The GRS test statistic Fmv

has F distribution. The Wald statistics WALDMV is distributed as 12.

 

 

 

 

mm W‘ Minder; W
International Mutual Funds Jensen t statistics" Jensen t statisticsb Jensen t statisticsb
Alliance Canadian 0.0230 0.1122 -0.1919 -0.4734 -0.3315 -0.9378
Alliance Global Small Cap A -0.2279 -0.7763 -0.0412 -0.0989 -0.3517 -1.4751
Alliance lntemational A 0.3999 1.4742 0.2450 0.7503 0.3318 1.1840
Bailard. Biehl lntl Equity 0.2023 0.6367 0.0333 0.0968 0.1722 0.5573
Centerland Kleinwrt lntl EqB 0.5366 2.0315 " 0.4210 1.3969 0.5241 1.9468 '
Dean Witter WorldWide lnvmnt 0.2856 1.3774 0.2546 1.1207 0.2314 1.0690
EuroPacific Growth 0.7754 3.5728 "‘ 0.6724 2.3628 " 0.7282 3.0299 ""
Fidelity Overseas 0.6791 1.8577 ‘ 0.5531 1.3622 0.6154 1.6100
First Invest Global 0.6635 2.5436 " 0.5551 1.8409 ' 0.5488 1.8564 '
FT lntemational Equity A 0.4396 1.5620 0.3122 0.9667 0.4065 1.3573
G.T. Global New Pacific Gr A 0.8721 2.2144 " 0.7806 1.8620 ' 0.9400 2.3871 "
IDS lntemational 0.4371 1.5768 0.3023 0.8908 0.3807 1.1987
lnvesco Pacific Basin 0.5748 1.7312 ' 0.4188 1.0603 0.5106 1.3413
Japan 0.8543 2.0381 " 0.6115 1.1265 0.7748 1.8910 "
Kemper lntemational 0.4746 1.8550 ‘ 0.3328 1.0899 0.4441 1.7343 ‘
Keystone lntemational 0.2869 1.0370 0.1477 0.4816 0.2332 0.8075
Merrill Lynch Global Holdg A 0.2726 1.5889 0.3089 1.5950 0.2554 1.4979
Merrill Lynch Pacific A 0.9424 2.0297 " 0.8378 1.7186 ' 0.8566 1.8048 '
New Perspective 0.4999 3.5156 "' 0.4873 2.1278 " 0.4263 2.5420 "
Oppenheimer Global A 0.7444 2.6902 "' 0.6221 1.7378 ' 0.6292 1.9756 '
PaineWebber Atlas Global GrA 0.3059 1.2354 0.2486 0.8807 0.2608 0.9859
Princor World 0.4063 1.3832 0.2729 0.7225 0.3767 1.2308
Prudential Global B 0.3380 1.3317 0.1986 0.6666 0.1793 0.6210
Putnam Global Growth A 0.3584 1.6791 ' 0.3580 1.3584 0.3389 1.6274
RSI Retrmnt lntl Equity 0.4961 1.7888 ' 0.2957 0.9464 0.3960 1.4130
Scudder International 0.6077 2.4527 " 0.4517 1.4797 0.5299 1.9771 '
Smith Barney Shear Glob OppA -0. 1833 -0.9319 -0.1728 -0.6482 -0.2106 ~0.9343

T. Rowe Price Intl Stock 0.7677 2.7783 "‘ 0.6106 1.9842 " 0.7126 2.6286 "'
Templeton Foreign 0.8324 4.6948 "" 0.7623 3.0689 "‘ 0.7829 3.7692 "‘
Templeton Growth 0.3176 2.1651 " 0.4778 1.8734 ' 0.3785 2.2300 "
Templeton Smaller Comp Grth 0.1549 0.6371 0.2963 0.9344 0.1559 0.6581
Templeton World 0.2305 1.5036 0.3302 1.3547 0.1942 1.2072
United lntemational Growth 0.5074 2.0557 " 0.3837 1.4272 0.4362 1.7044 "
Vanguard Intl Growth 0.6812 2.2587 " 0.4654 1.3793 0.6013 2.0355 "
Vanguard/Trustees' eqy lntl 0.6644 2.8591 "‘ 0.5461 2.0720 " 0.5895 2.3738 "'
Joint performance test on all funds Statistic P value Statistic P value Statistic P value
GHM Test; KTMV (1 to 35) 53.18 0.0151 "‘ 49.66 0.0330 " 49.99 0.0299 ”
GRS Test: Fm (35;53.63.61) 1.86 0.0201 " 1.85 0.0165 " 1.99 0.0090 ”"
Wald Test: WALDMV (35) 121.51 0.0000 "‘ 105.53 0.0000 "‘ 122.60 0.0000 ‘”

 

Significance levels: "‘ indicates 1%, "' indicates 5%. ‘ indicates 10%.

' There is no forward contract for the Hong Kong Dollar. leaving only 11 countries in the unitary hedged benchmark.

b t statistic computed using heteroscedastic consistent variance.

93
Table 8 (cont'd).

The sample period is January 1986-March 1994. PPWi is the Positive Period Weight Measure for fund i.
The null hypothesis is PPW = 0 versus the alternative hypothesis that PPW at 0. The Wald statistic WALDppw

 

 

 

 

is distributed as x’-

mm nacttumntaenshmatk‘ weddindea mansenmtnak
lntemational Mutual Funds PPW 2 statistics” PPW 2 statistics” PPW 2 statistics”
Alliance Canadian 0.0162 0.0810 -0.1971 -0.4849 -0.3351 -0.9467
Alliance Global Small Cap A -0.2125 -0.7358 -0.0498 -0.1198 -0.3623 -1.5259 '
Alliance lntemational A 0.3766 1.3983 " 0.2337 0.7182 0.3099 1.1136
Bailard, Biehl lntl Equity 0.2085 0.6665 0.0319 0.0928 0.1744 0.5666
Centerland Kleinwrt lntl EqB 0.5246 2.0132 " 0.4134 1.3774 " 0.5189 1.9441 “
Dean Witter WorldWide lnvmnt 0.2864 1.3866 * 0.2503 1.1045 0.2318 1.0731
EuroPacific Growth 0.7548 3.4708 “‘ 0.6625 2.3373 “‘ 0.7168 2.9931 ***
Fidelity Overseas 0.6535 1.8407 ” 0.5504 1.3570 " 0.6224 1.6318 "
First Invest Global 0.6524 2.5569 ”" 0.5472 1.8212 *‘ 0.5576 1.9031 '*
FT lntemational Equity A 0.4266 1.5393 ' 0.3037 0.9448 0.3991 1.3451 '
G.T. Global New Pacific Gr A 0.8666 2.2254 " 0.7695 1.8364 *" 0.9305 2.3622 ""
IDS lntemational 0.4212 1.5633 ' 0.2912 0.8593 0.3668 1.1656
Invesco Pacific Basin 0.5727 1.7909 "’ 0.4103 1.0432 0.5025 1.3346 '
Japan 0.8282 1.9636 "' 0.6168 1.1359 0.7604 1.8630 "'
Kemper lntemational 0.4660 1.8435 "' 0.3247 1.0670 0.4393 1.7261 "
Keystone lntemational 0.2758 1 .0061 0.1477 0.4839 0.2378 0.8264
Merrill Lynch Global Holdg A 0.2599 1.5383 ' 0.3071 1.5853 * 0.2576 1.5037 "
Merrill Lynch Pacific A 0.9219 2.0717 " 0.8303 1.7064 “ 0.8459 1.8025 "'
New Perspective 0.4854 3.3950 m 0.4813 2.1038 *" 0.4280 2.5482 ***
Oppenheimer Global A 0.7249 2.7270 '“ 0.6075 1.7025 '* 0.6279 1.9638 "
PaineWebber Atlas Global GrA 0.2946 1.2074 0.2400 0.8487 0.2521 0.9599
Princor World 0.4242 1.4691 ' 0.2593 0.6877 0.3530 1.1599
Prudential Global B 0.3134 1.2689 0.1952 0.6531 0.1869 0.6497
Putnam Global Growth A 0.3537 1.7003 "' 0.3515 1.3406 " 0.3336 1.6132 '
RSI Retrmnt Intl Equity 0.4736 1.7295 ” 0.2923 0.9373 0.3920 1.4042 "
Scudder lntemational 0.5809 2.3836 *** 0.4409 1.4518 ' 0.5206 1.9583 "
Smith Barney Shear Glob OppA -0.2036 -1.0571 -0.1825 -0.6839 -0.2224 -0.9932

T. Rowe Price Intl Stock 0.7546 2.7503 "" 0.6039 1.9703 “ 0.7094 2.6278 “"
Templeton Foreign 0.8330 4.7477 “‘ 0.7560 3.0521 "" 0.7845 3.7743 ""
Templeton Growth 0.3420 2.3763 '" 0.4711 1.8551 "' 0.3778 2.2289 “
Templeton Smaller Comp Grth 0.1800 0.7595 0.2874 0.91 13 0.1491 0.6329
Templeton World 0.2409 1.6124 ' 0.3264 1.3444 " 0.1941 1.2041
United lntemational Growth 0.5025 2.0676 “ 0.3836 1.4278 " 0.4418 1.7302 “
Vanguard Intl Growth 0.6737 2.2167 *’ 0.4589 1.3626 ‘ 0.5940 2.0217 “
Vanguard/Trustees' eqy lntl 0.6565 2.8196 ”‘ 0.5417 2.0633 *" 0.5855 2.3729 ”'*
Joint performance test on all funds Statistic P value Statistic P value Statistic P value
Wald Test: WALDMV (35) 146.24 0.0000 *** 105.39 0.0000 ... 121.33 0.0000 ”"

 

Significance levels: *" indicates 1%, *‘1' indicates 5%, '1 indicates 10%.
' There is no forward contract for the Hong Kong Dollar, leaving only 11 countries in the unitary hedged benchma

b PPW coefficients and 2 statistics are computed using the generalized method of moments.

94

Table 9
International Equity Mutual Fund Performance Persistence

The sample period is January 1985-March 1994. The
model is apt = 130 + ptB + er where apt is excess return on a
portfolio of mutual funds weighted by at, the normalized
first sub-period Jensen measures. 80 is the estimated
performance persistence parameter.

 

 

Benchmarks
Persistence
Parameter t statisticsa
12-country Benchmark 0.0981 0.8362
Wilshire 5000 0.1399 0.4346
World Index 0.1743 0.5519
3-region Benchmark 0.2455 1.8562 *

 

Significance levels: ‘1'" indicates 1%, ** indicates 5%, '1' indicates 10%.

95

Table 10
Tests of the Mean-variance Efficiency of the World Bond Index

The sample period is January 1989-March 1994. The mean-variance model is a. = 130 + p.13 + et where
a. are excess returns on country bond indices and p, are excess returns on the Salomon Brothers world bond
index. 130, is the regression intercept and PPW, is positive period weight measure for country i. The null
and alternative hypotheses of the GRS test are 80:0 and Bo :0 respectively. The GRS statistic F My has
an F distribution and the Wald test statistics, WALDMV and WALDPPW are distributed as )8.

The quadratic model is a. = 13° + p.13. + p.213; + et where m2, is country i's regression coefficient on the
squared world bond index. The null and alternative hypotheses of the Wald test are urz=0 and m2 #0
respectively. The Wald test statistic WALDQUAD is distributed as x2.

 

 

 

 

Benchmarks WWW

Countries 00 t statistics' PPW z statisticsb m, t statisticsa
Australia 0.3667 . 0.8983 0.3570 0.8766 -0.0933 -1.1453
Belgium 0.0349 0.1303 0.0231 0.0863 -0.0595 -1.1895
Canada 0.1073 0.3337 0.1076 0.3342 0.0157 0.2781
France 0.0297 0.1 199 0.0215 0.0870 -0.0265 05571
Germany 02213 -0.9090 -0.2327 -0.9589 -0.0625 -1.3590
Italy -0.0243 -0.0663 -0.0319 -0.0873 -0.0018 -0.0325
Japan -0.0655 -0.2359 -0.0566 -0.2043 0.0147 0.3035
The Netherlands -0.1817 -0.7666 -0.1934 -0.8165 -0.0659 -1.3943
Switzerland -0.3610 -1.2762 -0.3705 -1.3154 —0.0656 -1.4297
United Kingdom -0.2865 -0.8883 -0.2867 -0.8930 0.0275 0.5004
United States 0.2063 1.4759 0.2079 1.4842 0.0190 0.6792
Joint tests on all countries Statistic P value Statistic P value Statistic P value
GRS Test: Fm (11,51) 1.17 0.3305

WALDMV, WALDPPW, WALDQUAD (11 15.53 0.1597 15.54 0.1589 8.51 0.6666

 

Significance levels: *" indicates 1%, '1‘" indicates 5%, " indicates 10%.

‘t statistic computed using heteroscedastic consistent variance.

h 2 statistic computed using the GMM estimated variance.

96

Table 11
lntemational Bond Mutual Fund Performance: Mean-variance Tests

The sample period is January 1989-March 1994. The model is a, = 130 + p,B + e, where at are

excess returns on mutual funds and p, are excess returns on the benchmarks. [30, is the Jensen
measure for fund i. The null and alternative hypotheses of the GHM test are 80 = 0 and 80 Z 0

respectively. The GHM statistic, KTMV is asymptotically distributed as a weighted mixture of 12.
The null hypothesis of the GRS test is 80:0 versus the alternative hypothesis that 80 $0. The GRS

test statistic Fmv has an F distribution. The Wald statistics WALDMV is distributed as )8.

 

Benchmarks

lntemational Mutual Funds

Waders

Jensen

t statistics‘

Waders

Jensen

t statistics'

 

 

 

Capital World Bond -0.0070 -0.0705 -0.0877 -0.6193
Fidelity Global Bond 0.0454 0.2415 -0.1030 -0.4785
Franklin Global Govt Income 0.0274 0.1314 -0.0826 -0.3943
G.T. Global Govt Income A 0.1066 0.4838 -0.1131 -0.5432
G.T. Global Strategic Inc A 0.0520 0.1600 -0.2848 -0.8997
Hancock Freedom Global Inc B -0.1569 -1.0931 -0.2396 -1.4492
Keystone Amer World Bond A 0.121? -1.1277 -0.2093 -1.2853
Lord Abbett Global Income 0.0303 0.2691 -0.0993 -0.6154
Merrill Lynch Global Bond A 0.1 106 0.7199 0.0621 0.3223
MF S World Governments A 0.0197 0.1 170 -0.0117 -0.0535
PaineWebber Global Income B 0.0013 0.0091 -0.0361 -0.2104
Putnam Global Govtl Income A 0.0797 0.4855 -0.0153 -0.0783
Scudder lntemational Bond 0.2557 1.7976 ‘ 0.2326 1.0714
Smith Barney Shear Glob Bd B ~0.0525 -0.5662 -0.1358 -1.0111
T. Rowe Price Intl Bond ~0.1366 -1.0291 -0.0629 -0.2290
Templeton Income 0.0299 0.1778 -0.0781 -0.4680
TNE Global Government A -0.1823 -2.0503 "' -0.2408 -1.4789
Van Eck World Income -0.0573 -0.2929 -0.1058 -0.4380
Joint performance test on all funds Statistic P value Statistic P value
GHM Test: KTMV (1 to 18) 10.01 0.8788 5.73 0.9935
GRS Test: FMV (18;44,43,44) 1.62 0.0972 " 1.90 0.0419 "
Wald Test: WALDMV (18) 39.63 0.0023 ‘" 46.35 0.0003 *"

 

Significance levels: ”1' indicates 1%, 1" indicates 5%, "' indicates 10%.

' t statistic computed using heteroscedastic consistent variance.

97

0050.03 0:06.300 030300000000... 3.3 .005an0 0.2.5”. 0 a

.e\eo. 00002.02 _.. .e\em 0.0.3.2.. I. .3. 5.3.2.. I... 6.05. 00:00c.:w.m

 

:8. N03... 2. ...N.N 08¢... a: NE...” :3... 0:53. 25>» 0.0m. =0>
350... 82... 2:. mowed gov... a... omhmd. vwmvd < E0EEo>o0 .320 m2...

1... o.o¢.m Nmmm.. a: $36 cmmvd a: $86 3.3... 0:505 50.050...
2:. Nhnmd 003.... a 23.. 3.3.0 as. 2&de .3»... 25m. .3. 02.5 030M. ....
mNVm. can... a: wmwwd omvmd :3 monN. Chm... m pm. 320 000—5 05%. 5.8m

_. $0.... 220... a: 32.... :3... a: 33mm. 030... .Eom .3203005. 0020:0m
031... 33... a: 33.... 2%... 2:. mmvwd 09%... < 0800:. .300 .320 £052.".
080.. Sun... a... 38:0. Eton... :3 mmvcfi moon... m. 0500:. .320 03303025.
80.... move... a... cmmwd 29V... 2... 33.... 53... < 0.5055050 2.53 m“..2
omen. 8mm... a: mom... :9... a: 3:0 0.3»... < .Eom .320 :09... ...t0—).
mnovd come... :3 .03.... 3.3.... 2:. mums... .oomd 0500:. .320 8357.. .05..

t. aged .Smd a... in... Sun... 2: mmm...m. 33... < 25m. 2.53 008< 0:90.02
a. mnmmd 030... a: $006 33.... :5 2.02m .N.m.o m 0:. .320 E20000... 0.0032.
van»... ommhd a: void mmvv. a: mvvoé oewmd < 0:. 2w003m .320 .....0
2.3.. 2.5.... a: 39.6 83... a: ammoN 3...... < 0:53. 300 .320 ....0

a... Nmomfi SS. 2. mwnmd Bow... .2. vovoN 33... 0:52.. “>00 .320 2.0.52".
_. 030.. who... a: .m. ..m 33... a: 02m.» 23... .05.. .320 5:02...
a... .98..V 0.3.0 a: $3.0 3.3... a: 30.6. n. .0... 25m 25>» .8500

 

 

.0226; 0 d emotmcfim . "m. .3322: d
22.030405335000313 2.342001%... €341.02

20:2“. .35.). .3203002.

2055.30“—

.502 253 3085050 .m... 05 5 08.02 253 0350.50 m... 05 3.80.30 :5... 20.200.
new x02: .33 ~=0EEo>00 .m.D 05 £0.02 .33 E0EE0>00 .m.D-:oZ 05 5 0530. $0008 03 x. .05 3.5.. .032:
5 0520.. 30008 0.8 .a 0.33 J + m... + em. u .0 m. .255 0.... .30. £202.03. 03:2. m. 02.00 0.953 0;...

0:30.. .05.". .35.. 16022—0005 .0 35:30.09
NH 0.5.0...

98

Table 13

International Bond Mutual Fund Performance: PPW Measures

The sample period is January 1989-March 1994. PPWi is the Positive Period
Weight Measure for fund i. The null hypothesis is PPW = 0 versus PPW at 0. The

Wald statistic WALDPPW is distributed as x2.

 

 

 

 

Benchmarks SW SBBLQagLIndex
lntemational Mutual Funds PPW 2 statistics3 PPW z statisticsal
Capital World Bond -0.0060 -0.0608 -0.0940 -0.6678
Fidelity Global Bond 0.0457 0.2442 -0.l452 -0.6582
Franklin Global Govt Income 0.0302 0.1449 -0.0987 -0.4689
G.T. Global Govt Income A 0.1 128 0.5132 -0.1207 -0.5730
G.T. Global Strategic Inc A 0.0630 0.1948 -0.3241 -1.0034
Hancock Freedom Global Inc B -0.1592 -1.1098 -O.2448 -l .4904
Keystone Amer World Bond A -0.1 181 -l .0991 -0.2084 -1.2936
Lord Abbett Global Income 0.0279 0.2491 -0.0945 -0.5900
Merrill Lynch Global Bond A 0.1087 0.7007 0.0595 0.3113
MFS World Governments A 0.0199 0.1178 -0.0148 -0.0675
PaineWebber Global Income B 0.0006 0.0040 -0.0444 -0.2586
Putnam Global Govtl Income A 0.0819 0.5002 -0.0280 -0.1431
Scudder lntemational Bond 0.2615 1.8386 " 0.2277 1.0485
Smith Barney Shear Glob Bd B -0.0472 -0.5101 -0.1369 -1 .0163
T. Rowe Price lntl Bond -0.1385 -1 .0361 -0.0620 -0.2255
Templeton Income 0.0376 0.2238 -0.0833 -0.4877
TNE Global Government A -0.1853 -2.0825 ** -0.2457 -l.5181
Van Eek World Income -0.0552 -0.2837 -0.1098 -0.4516
Joint performance test on all funds Statistic P value Statistic P value
Wald Test: WALDMV (18) 38.60 0.0000 *" 44.12 0.0006 ***

 

Significance levels: *** indicates 1%, ** indicates 5%, "' indicates 10%.

’ PPW coefficients and 2 statistics are computed using the generalized method of moments

99

.oo:m_8> 88288 2388888; 87.: 389:8 03283. H .

$2 8:365 .. :5 8.865 : ..x; 83065 ..,: "222 8535:»;

 

 

 

 

:. 88... $6 €123.45 use 235
33> m oumzfim 8:3 =m 8 $3 88..
2.36. unwed- 236- 923.0. mmmmd 3 mthd 386 0:82: 28>» 20m. 5>
me _ .o- mmm _.o- 3w _ .o. «E _.c. moved. 3%. T 356- < 285080 1320 mZH
oomcd owned nomad 3.36. 336 3 $mm.m 886 0:82: 88883‘
3m _.o- mam _.o. 22.? $86.. N836- 336.. mm 86. 28m :5 out; 039: F
33.? N356- 335.? Sand. maid 2. 83“” 336 m .5 :20 82m 55.5 28m
nmmmd m Gmd m Gmd thod gmmd .3. Sid vmccd 28m 3838885 8288
336 Swod mmwod aomod. Sue—d now: named < 0:88:— 380 .320 :85:
m _ cod good Coed on 5d- 5 56 _ 3 _ .o mmood m 2885 .320 83030592
356 356 28.0 sowed- good 336 386 < $865080 28>» mm.)—
2: ..c Sofia oo. _.o $.36 38.? cwwmd- 386. < 28m 7320 883 5.52
836 38.0 386 386 886 386. 886 0:82: 1320 32:2 n84
:36- 5. _.o- :2 _.o- 336. moo—.0 mm 52 2.86 < 28m 28>» 88< 28532
$26. Now—d- 926. 9.2.0- ovood- 38.? :86. m 2: .320 5808.2 2828:
owned omcod wooed cmwmd- 3Vde : $2: aged < 2: 29.85 .820 9.0
woo—d mm. ..o N. _ ..o 386. $26 2 $.— 236 < «:80:— 300 .320 9.0
2.8.0 886 vwmod $8.0. $36 893 ode 0:82: 300 .320 £28..”—
vaod $35 oovod 886 ammod Eomd Ngqo 28m .820 2502.;
286. 88.9- vcocd- 886- 28.0 0823 c256 28m 28>» :3qu
828—. 3.5 Ergmeta ..B at: . fla .moumtfim 8 N.__. 88..— EBEZ 383888:—
08802 08822 >322 bano—om wEEwH gagging—Aqua
882. 3%. Leger... btaoom

 

..o + a... + an n .a 2 E88 0832: 5.82. 2:. .52: 2.3 283 05 8 888.. 3088 38:9. 05
8 820880 868th mm 25“. 2 RB .82: 2.8 283 05 8 mEBo. 8088 mm .a 28 85c 28p 8 2:38 3088 2a
.a 08:3 .0 + "B“... + .6... + ..B n .a 2 .088 2522-888; 2:. 63— nos—2-32 98:2. 2 2288 29:8 2F

88822 «285829.— 8 6818580 6655 33:2 98: 2:832:85

wa mink.

100

60:00.... .5220... 008008220: wEm: 0050800 002...... . .
...\..c_ 8.0205 .. .o\..m 8.0305 .5 ..\.._ 8.3.0:. .2... .203. 00:00..=:w_m

 

 

 

 

 

... 8...... and. ... 8...... ~29 ...: $3.; .8: 22..

2.... an. ...-0...}. ”.8: 20

.....a... m: .... 2 .. ES. .8: 2:0

030> m 28:05 03:.» m oflmzfim mvca :0 no “no. gauche-tom ~50.-

.. ~...m.~ 2...... ....2..~- .32.... : 82~ 5.... :2... 3...... 2:...- wvvo... 2:...- ..mm.....- 028... 253 ...-n. §>
2.3 3..~.. ~00...- m..-....- ~90...- ~$o..- ... ...-02 «...-.... ... 5.00 .....~..- : 8.3.0 :2...- < 2.6520 .220 02:
E-~ ....n... : “$.2- .8..... : 9:2- 22...- ..n... :3... .82... :8... 2...... 2......- ES... 5.23....
- 8...... .3... E... 2...... . £0... :8... ... ~82 9...... 3......- .-....... 2......- ~m...... 28:28.... 2.3. :
...-2... ...... .. ......~.- ”8...... m2“..- 3......- omn... 3...... 80....- ..2...... 3......- m.~....- m ..m ..20 a2: .25: 5.5

. 3...... ..~..~... 2......- onm..... 2...... 5...... 8:... 0...... .. ~22... 82... . 3...... 82... 2...: 38.350.23.25;
.. .....o.~ 2.2.. .503. .32.... .23...- ~......... ~83. 8. ... .....m... 88... 0.x... 3...... < 2:82 :80 .220 .55....
~23. as... $8...- owvo... 0.8...- 38... SS... 3:... $3.... 3......- o........ 3......- m 2:8... 3.6.0.3....3022.
:- ~....m :3... - 3:..- ~o~....- :2... 5.... .2...“ Ea... ~..~... :2... 8...... 3...... <2coaeo>o0 253 m"...
$8... .30.. 23..- R...... 5......- ....n..... 22... 82... 3:2. .3... ......h... 22.... <25: .....20 .8... ...Es.
...... .... ..-.~... ... an... :3..- .EL.... 38.... 82.... ..m-.. 28... .....o... ...-.... 2...... 2:82 .220 23.2 ......
2...... .8... . .8:- 3....- ~82- 2.2...- ... 5...... ~...~.. . ~83. 82.... 3...... Sn...- <..=om 25>. .25.. 82:3.
82.: 3.2.. .. ..~....~. ......- ~.$..- 8......- :. .5... 9...... . .2...- X...... .-...- ...-......- m u... .220 888.... .085...
... .0. ......2. .....2- 35... 0.33- :5...- 3~§ 8n- :2... 8...... .~n~...- 28.... <8. 2.28: .820 :0

: .mn~.~ .30.. $.3- o..~..- :- 32...- ~..~.m... 3...... RS. :2... ~m~.... 2.2.. :8... < 2:8... .80 .820 :0
:5... .23... - 82$- 3:... ~82- 2.0..- SS: .8... 2.....- 8...... 3......- n-....- 2:82.50 .820 2:5...
2:3 2%.. : Sn~.- 2:...- : m...m.- m8...- E... 2...... 88.... ...-......- 2........ ~........- 28: .220 2.2.:
-n..~ ...R... 83..- -3..- 3......- 32..- $8... 2...... .3...- .mm....- :2..- $3...- 25.. 2.2.. .230
amatmcwum a d .mOthSm a an amozmcgm a an amotmfifim a .n amotmmgm N Eh amotmflgm a Goa-3% mficzn— 33:: 1.502an3—

Eadfiaaau 3.33.02. E 52.. 95: 25>. 3 g

 

 

.... a 3:25... 2 2.0.23 2.2.2... 2.5 2: ... u 3.... a... 2.352.. 2.. was; .. u 3.... 2 2858.... ...... 2: .. .25. .2 2:802 22-2: .25.. 228,.

2.. 2 5.... .... 8 .5528... 2 2.0.3... 8.2.8. 2.3 2: .8..=..§_.. : s. a. E: 22.5. .8. 9.0 2: ..a ... ...... 2.352... 2.. 85> ..n... m. .8. 9.0 2.. ...

2852.... ...E 2: .. .05. .2 2. ...... ... .... .... 28.8... ...... x32. . x a s. .... .— .... .... 22...... 3.203 a a 8325... 2.8.82.5? .... 52:0. .22....” 2:0 2:

5300008.. e N on 0:: e u on 0.3 «.6. 2:0 20 ..0 mow-0500.9. 02.95:: 0:: =0: 2:- ._ 0S... 8.. 2:80.: 08:3. 20 ...: an 98.28 033.0. 3:250 0:0 .600. 0:0:
0:03 0:. :0 85.2 $8.... 2: ... 0:: «.05... 0.0.0:. :0 ...EEE 38x... 2: ... 20:3 .0 + a... + on u .0 m. .0008 0.:- éoo. £802-30— ES... 2 00:00 0.083 0.:-

3335 mama—.0: 50033—335 80:03.85...— 055 .0532 0:0: ...—03:53:—
n— 03:-

101

60:22.; 3:22.88 22338220: 38: “.2388 22:23 3 .

.o\oo_ 8:365 3 {on m.2855 .5 .o\e_ 32865 ...: H.05— 3:85:9m

 

 

 

 

... 2333 £33 33333 33.2 a: 23$: :32 55:
33:3 32 :32: >23 22 3:0
:333 33 a: 9 : >25 an; 2::

031; n— otmzflm 03—9» A oflngm mace =m :0 “no“ 00:58.5.“th “EC—-

:23 3R. .3 333. .3 .3: 3. 2:3 323 $83 3233 :83 2:85 253 3.8m 5”>
323 3333 33333 3323. 223. 28.7 523- 3232- 5: 3- < 3855230 .220 m5
2: .3- 3:33. 3:33- 8.83 3333 3833 2333 3383 3833 2:85 8.23:5
: 332 .N 3.333 293 3383- 283- 3333 ~23 3.83. 2333. 25: :5 8F. 26: H
2%.. 2.: 3 5: .3 £33- 283- 223. 3383- 5333- 533- m 3: 85 325 :25: 5:5
332 283 $23 2333. 3333 322 3: m3 382 3333 25: 32.52525 .2252
RE 3 383 2:3 283 2333 3323 3233 8:3 283 < 2:85 :80 .320 535::
333: 32.3 $2 3 .32 3- 2333 3333 3:33 2:3 2:3 m 2385 .320 22.8323:
2 333 2.: .3 82 3 3323. 3383 3323 :33 333 $23 < 2555230 253 2.2
3.33 2.83 33333 383 2.3.3 323 mm. _3 383-3 2: .3 < 25: .350 8:3 5222
£33 5333 $333 ~233- 3333 2333 3:33 3233 533 2:85 .320 :35- 23
3:2 3: .3 3:: .3 823. 3323. £333. 8. .3. 3233. $2 3- < 28: 283 58-1 2.223.
82 .3- 3:33- «.233- RS 3- 3323. 3.3..- :23- NMS..- £2 3. m 05 .335 5282: 3.82::
$23- one .3- z :3- 3:33 3323. 33333- 223. 323- 3323. < 25 23222 .83.: Ho
23. .3- 883. 283- 33333- 533- 233. 3333. $23. 333. < 2:85 so: .22: Ho
82... 523. 3E3. 223 3383. :23. 32.33. 223- 32.33. 2:85 :50 .350 53.2:
$3. _. 33:3. 32:3. 3:3 2:3. :2 .3. 3333. 33. _3. 283. 28: .220 2:25:
22.. 33. _3 3323 mm: 3. 3:33. 3333. 2.333. 2333. 3233. 25m 283 3538
.mocmzfim. N... :3...» . "a .9 Air-3:539 .82233 N 3.: M23323 3 :35: 23:..— EBEA 3:28:22:—
Eeoeoou use: 5.22 23.822 :52 288.2 2:: 23282 32:2 was?

3:5... .852..-
I gum-«Bu... + .8... + 39 u ... 2 3:2: x:~:2.._o:»2._- 2E- .Nx mm @2323: 2 3.30435 22:23
233 2F .e u 3.: .2: 52.58»: 03322—3 2: 3:22, a u 3.: 5 2858.3 =:: 2E. ._ .33 :8 2:802 E203 vote: 328: 05 2 SE.—
.Nx m: @2355: 2 220:3 8:32.: 23>» 2:- .:o_~:£.:m_: "— :3 ma: >2"— ucmzsm .8. mac 2:- 62 on 85 285092 03:83:: 2: 3:22, a u an
5 332 mMO 2: .3 232:8? =:: 2:- .Nx be 2:25.: :2nm63 : m: @2332: bfiuzoasxm: 5 52.5— .03232 2:0 2:- 503:892 e N on :5
c .I. on 2: 3.2 2:0 2: .«o «2058»: 03852.: :5 =:: 2:- ._ :5... :8 2:82: :2:o_. 05 2 5a 38:5 go: 233 3332. ESE: 05 :o E32
2838 2: ... :5 23:8 38::— :o 8:82 3888 2: .: 20:3 6 + m... + 3n u .a 2 .25:— of- .32 £222.32 3:5:- 2 veto: 29:8 of.

€503

38::— cgm 3.33 eon—.0: 22:93:53.5“..— e::..— ...-3:2 33: ...—23:52:—
: 93-h

102

Table 17
lntemational Bond Mutual Fund Performance Persistence

The sample period is January 1989-March 1994. The model is
apt = [30 + p-B + at where apt is excess return on a portfolio of
mutual funds weighted by on, the normalized first sub—period
Jensen measures. Do is the performance persistence parameter.

 

 

 

Benchmarks
Persistence
Parameter t statistics“
SB Broad Index -O.1676 -0.597O
SB World Bond Index —O.2232 -O.7356

 

Significance levels: *** indicates 1 %, ** indicates 5%, * indicates 10%.
a t statistic computed using heteroscedastic consistent variance.

.3533: 8333:3330 .220 2: w:_3: 83:33:38 233.3333 N 3
.8532? 3:23.28 2333:8820: min: 83:33:38 3:333:33 3 .
<32 3232::3 3 .333 3232::3 .. .33— 3338_::_ ...: H3.25. 8:8...Ewa

103

 

 

 

 

H3.-32:52.0: 333:...

33.32-32.35 308:."

3.3.33 33.. 33.52.3332 33.: 333.3 33.. 33.33.2333 33.: .8333 .3.. .3333323383 33.:
o:_m>n— oflmtfim o:_w>n— omamcflm o=_:>m ozmufim mom-5:38 :33 co mum“: 350—.
.3333 3333.3 3333.. .323 3333.. 33.3.3 3333.. 33.3.3 .333; 3.3333 333.3 3333.3 383.3233 32.3:
3333..- 5.3.3- 233..- 3333.3- 33.3.3. 33333- 33333. 3333.3. 3333.3 333.3 3333.3 333.3 253 23333.3. 32.3:
3333.3- 3333.3. 3333... 33.333. .3333... 33333. 3333... 3333.3. 3.3.3.3 3333.3 33.3.3 .333 253. 333.3233
3333.3- .333- 3333..- 3333.3. 33:... 33333. 3333... 3.333. 333.3 3333.3 :33 3333 33.33.335.882 2:
3333.3 333.3 3333.3 33:3 33:3. 3333- 3.33.3- 3.333- 3333.3 3333.3 3333.3 3333.3 2.33 3332
33.3.3 333.3 333.3 ...-33.3 332.3. 3333.3. 333.3. .3333. 35.3 3333.3 3333.3 .3333 3.8: 3.3..
3333.3- .3333- 3333..- 3.333- 333.3... 33333- 333.... 33333. 3333.3 3333.3 3333.3 3.3333 3333 33253:
33.3.3 33.3.3 33333- 3.33. 3333.3. 3.33. 3333. 33333. 333... 3333.3 333... 3333.3 283. 8:23
33333- 3333. 3:3. 3333.3. 3333.3 333.3 33:3 332.3 3333.3 33.3.3 3:33 33.33 3.53 2.25
.3333- 3.33.3. 3333.3. 33.3.3. 3333.3. 3:33- 3:33. 33.333. 333... 3333.3 3.33... 3333.3 3333 35.3.2.
3333.3 3.3..3 3333.3 3333.3 3333.3 3333.3 3.33.3 33.333 33.... 333.3 33.... 333.3 383 2.2.22
2 2 2 2 1.2.3.3 3333.3 .3333 3333.3 3333.. 33333 3333.. 3.3333 33.333.33.333 32...:
53.3. 33333- 33333- 3333. 33333 333.33 .3333 33333 333.3 3333.3 .3333 .3333 33.333. 353353. 32.3:
.33.. 3333.3 3333.. 33.3.3 .333: 332.3 .3333. .3333 ...-333.3 .3333 3.33.3 3333.3 33333 35.85.33
3.33.3 .3333 3333.. 3333.3 .3333. 3333.3 3.333.. 2.33.3 .3333 3333.3 3333.3 3.33 35.33.335.552 2:
2 2 2 2 3333... 333.3. 3333... 3.3333. 3333..- 33333. :33..- .3333- 33.333. 532
3.33.3 33:3 :23 333.3 33:3 333.3 33:3 :33 3333.3. 3333.3- .333. 3333.3- 33.33.”. 3.3..
3.33.. .33.... .53.. 3.33.. .3332 33.2.3 ...-3.33.3 33:3 ...3333 3.33.. 3333.3 3.33.. 33.333.383.38:
3333.3- 333.3- 323. 3.33.3. 3:33 3333 3333.3 3333.3 3333.3 333.3 3333.3 3333 3.333. 333:5:
33.3.3. 333.3. 3333.3. .333- 3333.3 .3333 .3333 33333 33.3.3 3333.3 3323 3.333 3.533 832”.
333.... 3333.3- 3333... .3333- 3.3.3 .3333 .323 333.33 3.33.3- 333.3. 3333.3- 333.3. 3.5333335:
3.333. 3333.3. 33333. 33.33- 3323 3.3.3 3333.3 333.3 3333.3 3:33 3.33.3 33.3.3 333.2. E23.33
333.3. 3333.3. 3333.3- 3333.3. 33333 332.3 3333.3 3333.3 3333 3333.3 3333.3 :33 33.333 2.2.32
3323333333 N 3%. . 3233:3333 3 3n 3323333333 N 33: . 3233:3333 3 on 3323233333 N 3.5 33232323 3 3: 3233::35

.33. 33 85353.3 23 32:33.3 333 >235: 32.3333 .32 3.33 2.. 2.3 8.53.3.3... 3 =3 32. >23 2.333.... 33.0 2:

8.0233233“: as on ::: cue: 2: 332 9.0 2: 3o 3032:3392 2:33:32? ::: ::: 2:. ._ 23:28 .30.. 2:82: 3:325 :33»: 933.30: 3. .33: ::: 3338325
:2332w2 2: 3. .3: . 3333:3332. 30333-3. 2: ::: 30333-~ 2: 38:5 8:333: 23:3 33:22: 0:33.» 2: ”33:85:23 80:23: m 2: :o 3:332 338383 23
... ::: 383:5 :::: ::: 333:8 3:28 :o 36:32 33838 2: .- 2233 3o + a... + on u 3: 33 3:9: 2: .232 33232-33332 23:53.- 33 :23: 29:33 2F

an 033—.

8:25.23...— :u23a_um 2.3 a: .32—223m— oucnmugéoz 23 3 33h

104

62.3.3.3 3:23.88 33:33:33,332»: 23.3.. :03an0 2.3.33 3 .
.336. 328.3. .3 6\..n 328.9: .3 3x.— 328.::. 3: U3.333.. 8:8c.:w.m

 

eoomd. 386. came. 89.: 03.5.— Ngmd amend. 2: _.¢. < 5390 .320 3:33:53.
:36. ...-55. whcmd 2.2.: 3&— 6. 886. omen-o. cvmmé- m .320 3.26:3;
owned @056 N53... Nave: nomad vmond 38.? 33.? 2.53 .85....
no.0. T nmmmd. Nmamd one. 6 536. on 5.? 3.36. 2.3.6. <30 .3320 3.3.33 333.33.523.33.
.3. cm. Eamd .. avwoN no.5: foo.— mmond $33.: vnNNd < .320 88.2.5350
932.. 33.: 3. :::.N nave-o : amid Nmmmd 3.3m: Nnood 0262.383. 302
:83... vocmd SUN-31o emaod 333.: .52.: 3.3. .o nNN. .o < 23.233. 3.0:».— ...tuS.
33mm... :98... : echN 533.0 2.93.. mm . m6 omwmd- Nawod- < :20! .3305 20:33 ...:o—z
.. moo—N- 23.6.. 39.... 93de- omQ..- ..de. 3:... «wood. 3:28:32... 2.8333—
33: m6- 85¢. 53.: momwd Nommd cmnod M33336- _ . .Nd. 3:233:32... .2.on
wvmm. Nmovd wound- omomd- ~036- cecflo. Sofie- 32...... S393.-
moo..c- 236- 3.2.: 33.: 9333.0. 30. .o- N335:- ohwvd. £25 0....233. 83>...
omwvd- «3 _ .56. Sand .3 _ d omowto _ .86 3336- ”N _ Nd- 73:33:23.2:— mm:

. Sod mowmd mmoc. .wond 33.33.: wmmvd 33.: one. .o < .0 3....233. 3oz .320 ...-.0
meme:- nwz .o- 3.33.: v. _ ..o 2.86. mmmod- honed- ommmd- < 5.3.3“. 3:232:25 E
Nmmmd- 3%de- mmoo. mmomd mvwvd vmv. .o Soho- 3.3.6. .3320 .302: 33......
.026. 0336. .vad vow. .c 2.de mmaod nmvmd- mom _ .o. mac-£930 3:023.
... .wooN omomd ... Nm_m.m ems-d 3 ~m.n.~ 3:2. 52.6 0333...: 5320 35.83.23”.
v. 9.6. amped- wmmm. .mnmd mmmnd own... Nwwvd- v33. .o- .552. 0232.03 323.3 53:9
:3. ..o- ammo..- SN... cam: mam—d vmvcd cmcmd. mowmd- mam ...: 333:6; :::—32:00

. .mnw..- £33.... £2..- n~_m.o. inn... 3309.? 9233.... 3.30.:- b.:¢m ::— _..o.m .::3_.3m
vmemd- maxed. ”new: m3; .3. an _ .o. ommod. .83....- owbmd. < 3:28:32:— 8:33:33
movm. _- woovd. .nvwd .53.: fixed. ~03... . moon... 2.2.:- < 3.30 :::.w .320 853:2
N. 3... ~39... 03.3.3.6. 3:6. awn-ad- 33.3%.:- 3 3.3.3... 2.3.:- §.:3.:.U 853:?

 

333.3-:83 . 53:9:

 

38.3.8.3 . :85...

 

332.393.3- . 53:9:

.33..3.§3 . 535—.

2:33:33

3:35... .3222 3:25:32:—

«Eu—Hag

.3 33 3233.33... 3. $333.33 8.3.3.3 3.33 233 33.53.33...
-. :3 33.. >2... 2.3.5.3 .32 me 2.... .cu 3.. .2: 3.32.8.3... 23:25.... 2: 3:38.. an... 3. .32 5.0 2: ..o 3.32.8933. ::: 2:

.3x .0 2233.3: 3.2.3.03 3 33 3.3.3.3.: 3.33.30.38.33: 3. >2h¥ 6:3..83 2:0 2.... 3.3.82.3“: : N 3: ::: c n on
:3 .32 2:0 2: 3: 382.593.. u>..3E2.3 :53 ::: 2:. .. :::. 5.. 95332: 53:33.. 2: 3. .33. 34.36553. 2: :o 3E82 338.8 2.3 .3.
:5. 3:5... .322: :o 33:32 33088 23 .- 22.; .o + m..— + 3a u .- 3. .25.: 2:. .32 5:32.33. 3032:... 3. 3.32. 0.9533 2...-

Bauh 023.23.75.32 Nougat».— :§m 12:2 3:322:85
o3 finch

105

.859; .5388 2.838232 38: 3.2.88 0.3.3... . .

...x-o. £865 . 3% 8:86... : ex... 8.86:. ... 222 8505:»;

 

 

 

come... NN.. 36m... 3... 6536 v. .. Nmmn6 N6. Gdddmmn. >2... :8..- 9.0
.6056 M56... 6306 nv..n 336 3.3 6036 n56v Ann 3 3 >25! 33.5 2:0
03.3 A. 23.3% 2...; A. 23.3% o=.:> A. 26.8% 03.9, An. ouncflm 3:3,. :5 :o .8. cop—agate: .50..
N636- 5mm . 6- NowN6- 336- 3606 53.6 «mm... 5866 088:. 28>» 30m. ::>
3 wmmvN- V5.N6- . goo..- $5.6- 566m... «366 3666- 59.66- < 2.255.300 .520 m2..-
mm .N6- $.86- 6wam6 5.066 3006 63.6 m5eN6 N366 058:. :90...th
vmw...- N5N.6- 66.36- 66m..6- eva6 5MNN6 3.506 63.6 6:3. .2. our... 033. ....
6336- 66666- vav6- .366- 336 NNw.6 95m... 3666 m cm no.0 303m mop—mm 5.8m
5mmv. mm.N6 3 Nam-6N 35N6 3 em.v.N mONn6 33.. 634.6 «Eon 3:029:85 325%
33666- 6656- N636 w 366 336v. 605N6 vom56 "Ac-5.6 < 088:. ..>o0 .330 8:53.
35...... .6666- 5.3.66 wN.66 @5606 $6. 6 3.26 N66. 6 m. 088:. .326 83333.9.
3.36- 5N566- m5. .6 60.66 :3... 9336 66.56 3.26 < ScuEEo>oO 233 mm:
6026 .MN66 6:36 3N... . 356.. .526 5.66.. mnvN6 < 95m .320 p.093 ..E02
6.36- 6536- 6.6m... ”$66 «mix. 556N6 MNocd wow. 6 2:85 .320 32.3. 6.5.
M556. .- 5wN. 6- 5366- moo. 6- 3.66 N66. 6 N366 3.66 < .33. 25>» 38¢. 0Sam-m3.

. 5Nn5..- oNMN6- Neoa6- mam—6- 6.8.6 9666 336- 6566- m. o:. .320 80.52". v.ooocs...
v.5N6- 62.66- m .oN6 ”.666 oNNo6 m56m6 MNcN6 N366 < o:. 23.8% .320 .....O
N5566- M536- 5636 «mm... 6N6N.. omoN6 5m5v6 3.. .6 < 088:. 3:0 .320 .....O
.55m6- £666- VVNvd 3566 6636 66. .6 N266 5.5666 2:85 300 .320 52:5,.
:36- ...36- 666:6 N.566 5.306 86.6 . .5m6 N366 ::om .320 5:36.".
5:36- @366- 506N6 $.66 .voN. .ow.6 60.566 3.666 :55 253 .5295
MEN..- Nvmo6- 505m... @626 nmmN6 2666 V636. NNa. 6- .3. 5.7m. .moo.mESB:w::>
no. .6 3.66 6356 63.6 @636 Nm. .6 oMMN6- 5nN.6- 530.5 3:. Baas;
.56. 6 6686 nN5N.. mwam... 9.36 mn5N6 6.0.6- $566- 5320 3:222:85 USED
6636 N5N66 3 anmN .266 3:. 569.6 6626- 6366- 283 :OBEEQ...
n. .66 OVNN6 3 3.va 2&5... onm. 6626 336 3:66 5.0 anU 3.23% 5.9.5.89...
59.6... 35.6 .3 N62...” V5m56 . 3.9.. «new... 5woN6 65566 5380 :22:th
3 wovoN Nwmm6 .3 mvwoN mwN56 3 nacmN 3666 6556 N306 5.2:”. 5.0350...
5.36 .Nv. 6 . n55... «v.96 cm. ... ocoN6 66666 .866 385 2:. our... 033. .-
3. avN5.N- .cNm6- 059N6- 6vvc6- N536- v3.6- 35n... ”53.6- <aa0 no.0 32% 505mm 5.8m
N5N56 N... .6 556m. 656m... 6036 w5nN6 v. «66- MVN66- 120.8585 henna-um
.0566- 6366- N5NN6- .366- cxocd- Nwm .6- .636- .omm6- .953. 6:. £8.53. .9.

 

$65.33 . eon—3..

.855; . com-......

.528. 2 «Ed.

.8595. . 302:...

.8636...” . non-3..

 

macs“. .352 EccenEoE.

gang—s

 

106

3:052: ..0 0050:: 03:30:00 05 0:3: 00:09:00 0:. 8.3.0.3.: N 0:: 300.0508 3%. .

$2 8:805 . {on 8.805 .... ...x... 8:86... ...: ”232 852.02%

 

w. . ..0- .050. .- vcmw. ”03.0 5.0.. 3000 0520- 0020- < 530.0 300.0 505::
$00.0- hm0N.0- 0m .N.0 0.3.0 N. . ..0- KN00- «V0.0- .mmmd- m 300.0 33:00.5.
vm00.0- V0000- - 3.5: 3.3.0 0NN00 0300 30.0- 2000- 2:03 :00:..:.
0V0..- m . mN.0- 0.500 0000.0 0.00.0- 00.0.0- 0N.0.0- 0000? <:0 300.0 00.0.4. 0.503050“.

- 00.0. mm.m.0 .- mm00.~ 00Nh.0 . N000. hand M030 wvm~0 < 300.0 08.0.30an
Com. 93.0 :- 0vww.m 0.00.0 .- .nm..~ .m~m.0 3.de .300 03000039. 302
0000.. N030 0mmN0 .3000 0. 3.0 050.0 N020 ”N. ..0 < 00.00.”. .353 ...::0$.
Sum-.00 030.0 .- owamd 003.0 . 009.. mm.m.0 0.000- 0000.0- < wEo... 300.0 F.093 ...::0$.

: 0.0..N- $3.0- - N.mv..- vvmmd- - $00..- wo.m.0- . Dun..- 0m.0.0- 3:002:05. 0:0.mx0v.
m0w~.0- 3.00.0- vmvwd h0g0 .wmmd a 2.0.0 3%.? .wmm0- 3:0.3E0E. :0QE0V.

- Nvmm. v0mv0 mmhmd- MNON0- 0.3.0- :30- .00m.0- momma? :32
33:0. 00. ..0- 0mm..0 N200 33.0- m00..0- 300.0- Nvmmd- 50:5 0.0.0.». 00m0>:.
R300- 380- .med ..Nv..0 .030 800.0 03.0.0- Sum-10- 3:0.3E0E. m0.
w.mw.0 N300 . wk. .0. 0m .50 .800 009.0 30.0 8000 < :0 00.00“. 302 300.0 ...-.0
mmmed- mom—.0. m .mm.0 900.0 .0000- 0m~0.0- 2.00.0- m0mm.0- < 5.3%. 3:033:35 ...-n.
$00.0- 3.00.0- 00. ... 000m.0 vnovd mum . .0 052.0- eohmd- 300.0 83:. 0.0:...

w . 3.0- 0000.0- Nwhm0 ~.N . .0 .030 ”000.0 0m . m0- mm0..0- 38:05 3:02....
:- homod m.wm.0 2: .53...” N850 :- 0n0m.N 3000 080.0 mwvmd 530:0 05003.05”.
.N.v.0- N0500- . momm. :30 095.0 mom—.0 053.0- Nm0m.0- .:E>:. 00.30.33 83.3 50A.
N210- .0m0.0- N000. mmnmd 22.0 030.0 Emcd- 3000- mom. 2:. 03503. 0:018:00

.. mmm0... 0.5.0- 30..- 0.2.0- . 33.7 509.0- . N.0N..- 00:..0- 5.3.0”. .E. 205 0:0.3m
N0900- m0~0.0- mvmnd 0030 N020- 020.0- 002.0- wm.v.0- < 3:002:02. 00:0...<
omen. .- 3.3.0.. 3.00.0 N300 530.0- mhvod- .- mug. .- 33.0- < :00 .3Em 300.0 005:2

: 0mm..- N.N.-.0- 503:0- mmm..0- 003.0- 330- : mmmmN- 3.0.0- 5.00000 00:0...<
.8537. N 3.: .3500; n 3...— .mozmzfim N 3...— .m00m28m N 3.... 00:0,... 333). 3:002:02.

 

.u 8 33:20 m. 32033 22.3“ 255 2: ... a 3.... .2. $8583 0:352; 2.. 33> a u 3:: 0.: £8582
:2 2F ._ 95: a: 2382 2903 85: 02:8: 2.. a Si: .32 632-32 025. a 3:2: 29:9. 2:

00.50003. 3.... "00525030.. 0:5... 33:3. 3:030:03:—
0~ 030,—.

107

9:258 go 352: BEN—05w 2: m5? 359:8 2: 8333.: N ::: 356598 3.5 .
652 8:865 .. .o\om 8:865 .. ..x: mofiofifi ...: ”232 358555

 

 

wooed- ”mm—d- 38.0. :36- named 32 .c 3.36 mac—d 2:8:— 283 *5 5“>
2; Send. Smmd- 3 awed- 53.0- 020° 33.6 Embed- 35.0. < EoEEoSO Ego—O mZF
o _ cud- vm mod- ”coed mmhcd good 30 _ .o 2.de $36 2:00:- :QoEEoH
inc. 7 on. ..o- COS. _. 03” _.o- 336 OmNNd $86 $26 20m :5 corn. 953— F
3%.? 386.. szd- :86- Soad own _.c vwwmd 28.0 m um :20 .325 onam :::—m

. 2mm; vummd 2. mmwcd mcwud t. 3.: Quad ~85..— Snvd :55 3:23:32: ..ocnzom
vwwod- _m 5.9- ”and Eocd .. 25v.— 336 2:6 own—d < 2:85 =>oO .320 5:53;

N. Evd- good- 386 a. 5d uneod coo—d m End 2 _ _.o m 2:8:— _:no_0 ..onnoaofimm
vmovd- good. of _.o vied «on _._ wmomd ocefio 3.96 < 3585030 2.53 mm:
owe—d ofod :Nwd 3N2: t 3%.. ecmmd Nuke.— oovmd < 28m .320 £083 Etc—2
chhd- vmhod- ohmmd Named . 3.3..— ocomd nmocd «av—d 2:8:— Eno—O 32:2 You
wmvo..- com—d- 335. good- .NNod one—d VE ..o 886 < 9.5m 2.53 ..oE< ocoamxox

t mama. T Smmd- :36- mum ..o- mcood 286 fimvd- wmwod- m 05 .320 Setup—m x085:
mmmmd- «N. _.o. 33.6 Ono—d on 5.0 chomd ewgd mmwod < 2: 2»ng Ego—O 9.0
och _ .o- 535- w 506 5.3.6 ”ma... mnemd $36 95.6 < 3:8:— 260 .320 9.0
aawmd- 25.0- 33.6 unwed $36 .2 _.o 356.. 386. 2:8:— §oO 3920 535.5
ehmmd- good. crowd :86 id ”mm—d Bmmd Shed “Eon Boo—O >502.”—
mcavd- 836. mwomd 55.9 . Nam; van—d _vmm.o owed “Eam— 253 133:0
wan — .o- 38.9- 33.: 0mm _ .o Svmd Namcd wvnvd- hm _ ad- :5 bum .m08m3$3:w§>
m2 _.o mm 56 nacho «on _.o chvd o. _ _.c fwmd- mom _ .o- 5380 :5 :::—ES;
honed no 5.0 mmomg 236 336 whhmd ENNd- moo—.0- 5380 3:28:32: 3ch
mhmod 356 9: “bond awed .. mam..— ostd 33d. woo—d- 2:03 :82:th
336 0. Sue to ahmmd ovowd . 3mm; acid _Nmo.o- @256. 5:0 9:00 .5::Em :92:th
2: ~._ ave—d 2... $86 33.6 t «3%. 336 3.36 mmood 5390 :92:th
.... moved Nemmd .... memo.N 23.6 : ooomd aoood Onvwd mmaNd :wBSm SEES—Pr
Rhod maid 3 ago; govd N_ _ _._ hmamd saved- 286- xooum :5 SE 033— ....
g: 9‘38. ownmd- hohmd- moved- 336. 33.6- . Nmmo. T 0. Ed- <30 :20 Roam aogm .EEm
9.3.6 on. _.o .. mmmm; awomd mmvad finmd m2 ..c- 336. 3:28:82: :ouusom
good- omood- wmwmd- 936. Ne _ ed- on m _ .o- Nacho- fiend- bmscm =5 8:on 5M
.3538: N 3.: .mozmcflm N 3...— .moumzsm N 3.... .8338: N B...— mvgm 3822 3:23:55—
aaafiauflgum gag mung 3% gag-55

 

$28. 8 «Eu...

APPENDICES

Appendix A Data Description.
Mutual Fund Monthly Total Returns (Morningtar)

The monthly total return data on mutual funds are obtained from
Morningstar On Disc database. Total return is computed by taking the change in
monthly net asset value, reinvesting all income and capital-gains distributions
during that month, and dividing by the starting net asset value. Reinvestments
are made on the reinvestment date. The total returns are not adjusted for sales
charges (such as front-end and deferred charges and redemption fees). The total
returns are net of management administrative, and 12b-1 fees and other costs

automatically taken out of fund assets.

Wilshire 5000 Equity Index (Ibbotson & Associates)

The Wilshire 5000 Equity Index is comprised of over 6000 capitalization-
weighted security returns and is designed to measure the performance of all US.
common equity securities with readily available price data. The Wilshire 5000 is
about 86% NYSE, 3% AMEX and 11 % OTC. The returns are computed by

Wilshire Associates.

108

109
US. 30-day Treasury Bills (Ibbotson 8: Associates)

Updates are from The Wall Street Journal. A one-bond portfolio is
constructed with the shortest-term bill of not less than one month to maturity. A
one-month holding period return is measured for each month. Total return is

calculated as the change in beginning and ending flat prices.

Morgan Stanley Capital lntemational IndicesiMSCILflbbotson & Associates)

The monthly total returns are computed with gross dividend reinvested.
The MSCI regional and national indices are based on approximately 1,500
companies listed on stock exchanges in twenty-two countries. The combined
market capitalization of companies in these indices represents approximately
60% of the aggregate market value of the covered stock exchanges. The monthly
returns for the regional indices are reported in US. dollars. The monthly returns
for the national indices are reported in local currency. Exchange rates for
converting currencies are taken at 4:00 pm Central European Time each day. The
regional indices used in this study are the MSCI World Index, the MSCI World x
US Index, the MSCI World x Japan Index, the MSCI EAFE Index, the MSCI
Europe Index and the MSCI Pacific Index. The national indices used in this

study are Australia, Belgium, Canada, France, West Germany, Hong Kong, Italy,

l 10
Japan, the Netherlands, Switzerland, the United Kingdom and the United States.

The total return MSCI Japan Index in US. dollars is computed as the weighted
difference between the MSCI World Index and the MSCI World x ]apan Index.
The MSCI World x Japan and US Index is computed as the weighted difference
between the MSCI World Index and the MSCI Japan and MSCI US. Indices. The

market value weights are computed using the FTA market capitalization values.

FT-Actuaries World Indices (Ibbotson & Associates)

These indices are compiled jointly by The Financial Times Limited,
Goldman Sachs & Co., County NatWest/ Wood Mackenzie & Co. and the
Institute of Actuaries and the Faculty of Actuaries. The monthly total returns are
calculated with the gross dividend reinvested. The FT -Actuaries World Indices
aims for at least 70% coverage of the aggregate market value of all domestic
exchange-listed companies. Markets, companies and securities are only included
where direct holdings of capital by foreign nationals is permissible. The monthly
returns and market capitalization values for the regional indices are reported in
US. dollars. The monthly returns and market capitalization values for the

national indices are reported in local currency.

111

 

Exchange Rates (Ibbotson & Associates)

From 1960 through 1987, exchange rate data are obtained from OECD
Main Economic Indicators Historical Statistics. From 1988 on, exchange rate data are

provided by The Wall Street Journal reported at 3 pm New York time.
Salomon Brothers SB BroadTM IndexTM (Ibbotson & Associates)

The total returns on this index is reported in Salomon Brothers’ Mortgage
Research, Mortgage Pass-Through Security Total Rate of Return Index. The bonds
comprising this series include Salomon Brothers High-Grade Corporate bonds

and 7, 10 and 30 year Treasury bonds.

Salomon Brothers Currency-Hedged World Government Bonds Lbbotson 8:

Associates)

The currency-hedged indices are constructed by using rolling one-month
forward exchange contracts as hedging instruments. The face value is the
principal amount plus the interest that has already accrued and the interest that
is expected to accrue during the one-month investment period. This will leave

the intra-month changes in bond prices unhedged. However, the residual

112

currency exposure resulting from over- or under-hedging is limited only to

changes in bond prices, multiplied by changes in currency values.

Salomon Brothers World Government Bonds (Ibbotson & Associates)

Total return represents the one month percentage change in the bond.
Maturities of bonds in this index is at least one year. Regional index returns are

stated in US. dollars. Country index returns are stated in local currencies.

Lehman Brothers Global Bond IndexesTM (Ibbotson & Associates)

All issues in the Global Bond Index have a minimum of one year to
maturity. The index does not include securities that have floating rates,
convertibles, warrants or linked bonds. All country components are weighted
according to market capitalization excluding Japan. The Japanese bond index is
weighted according to the market capitalization of the largest and most actively
traded Japanese government bonds. All regional indexes are expressed in US.

dollar currency.

Forward Contract Rates (Ibbotson 8: Associates)

Rates are reported on the last business day of each month. The data

source from 1986-1993 is JP. Morgan with rates reported at 5:00pm London

113

close. The data source from 1994 to present is the WM Companies with rates

reported at 4:00pm London close.

Appendix B Mean-variance Test with Short Sale Constraint on Mutual Funds
Here I extend the results of Gibbons, Ross and Shanken (GRS 1989) to the

case where a subset of the assets, the mutual funds, is subject to a short sale
constraint. The maximized squared Sharpe ratio of the passive investment set,
{pt}, remains the same and is up’Vp'lpp. The maximized squared Sharpe ratio of
the combined investment set, {ab p¢}, becomes

[pp'vp-lpp + (BO + A‘s YET] BO].

[up'Vfiup +(Bo +K.)'Z"(Bo +?».)]"l (A1)

[H.‘W'F, +(Bo + 101490]
where B0 = pa-Vpa’V '1, 2‘. = V. - V,,.-.'V'1 VP. and A... is the vector of slack variables
associated with the no-short-sale constraint. The difference between the

maximized squared Sharpe ratio between the two investment sets is

[(90 + 10'2" [3012 +[(Hp'V;1Hp)(l30 + RYE-'03.) - 1.)] (A2)

Since the slack variables are restricted to be non-positive, the difference will be
greater than zero if Bo is greater than k... The null hypothesis that the difference
is equal to zero will hold when Bo = 0 and A... = 0 or when Bo = 4“. Therefore the

GRS test rejects the null hypothesis too often.

114

Appendix C Optimal Portfolio Weights.

The optimal portfolio weights for the combined investment set is

computed by solving the following minimization problem:

w = Argmin w'Vw
m

subject to (A3)
w'p = c

The solutions are:

WP = w\;:,w VP'lD‘l'p —Vpa 24 BO]

.V“ (A4)
wa : 17324 B0

w it

Since the first term, (w’Vw)/ (w’u), is a positive constant, the null and alternative

hypotheses can be written in terms of the conditional moments of at and Pt.

115

LIST OF REFERENCES

9. List of References

Adler, Michael and Bernard Dumas, 1983, ”International Portfolio Choice and
Corporation Finance: A Synthesis.” Journal of Finance 38, 925-984.

__, and Bhaskar Prasad, 1992, ”On Universal Currency Hedges.” Journal of
Financial and Quantitative analysis 27, 19-37.

Admati, Anat, Sudipto Bhattacharya, Paul Pfleiderer and Stephen Ross, 1986,
”On Timing and Selectivity.” Journal of Finance 41, 715-730.

Bauman, Scott and Robert Miller, 1994, ”Can Managed Portfolio Performance Be
Predicted?.” Journal of Portfolio Management Summer 1994, 31-40.

Blake, Christopher, Edwin Elton and Martin Gruber, 1993, ”The Performance of
Bond Mutual Funds.” Journal of Business 66, 371-403.

Brennan, Michael J., 1995, ”The Individual Investor.” The Journal of Financial
Research 18, 59-74.

Brown, Stephen, William Goetzmann, Roger Ibbotson and Stephen Ross, 1992,
”Survivorship Bias in Performance Studies.” Review of Financial Studies 5,
553-580.

Business Week, ”Bond Fund’s Dark Secret." April 03, 1995, 114.

Business Week: Guide to Mutual Funds. 1995 Edition. McGraw Hill.

Chang, Eric and Wilbur Lewellen, 1984, ”Market Timing And Mutual Fund

Investment Performance.” Journal of Business 57, 57-72.

116

1 17
Chevalier, Judith and Glenn Ellison, 1995, ”Risk Taking by Mutual Funds as a

Response to Incentives.” NBER Working Paper No. 5234.

Cumby, Robert and Jack Glen, 1990, ” Evaluating the Performance of
International Mutual Funds.” Journal of Finance 45, 497-521.

Droms, William and David Walker, 1994, ” Investment Performance of
International Mutual Funds.” Journal of Financial Research, 1-14.

Dumas, Bernard, 1993, ”Partial vs. General Equilibrium Models of the
International Capital Market.” NBER Working Paper #4446.

Dybvig, Philip and Stephen Ross, 1985, ” Differential Information and
Performance Measurement Using a Security Market Line.” Journal of
Finance 40, 383-399.

Eun, Cheol, and Bruce Resnick, 1984, ”Estimating the Correlation Structure of
lntemational Share Prices.” Journal of Finance 39, 1311-1324.

Eun, Cheol, Richard Kolodny, and Bruce Resnick, 1991, ”US-based International
Mutual Funds: A Performance Evaluation.” The Journal of Portfolio
Management 17, 88-94.

Glen, Jack and Philippe Jorion, 1993, ”Currency Hedging for lntemational
Portfolios.” Journal of Finance 48, 1865-1886.

Gibbons, Michael, Stephen Ross and Jay Shanken, 1989, ”A Test of the Efficiency

of a Given Portfolio.” Econometrica 57, 1121-1152.

118
Gourieroux, Christian, Alberto Holly, and Alain Monfort, 1982, ”Likelihood

Ratio Test, Wald Test, and Kuhn-Tucker Test in Linear Models with
Inequality Constraints on the Regression Parameters.” Econometrica 50, 63-
79.

Grauer, R. and Hakansson, N ., 1987, ”Gains from International Diversification:
1968-1985,". Journal of Finance 42, 721-739.

Grinblatt, Mark and Sheridan Titman, 1989, ”Portfolio Performance Evaluation:
Old Issues and New Insights.” Review of Financial Studies 2, 393-421.

_ and ____, 1992, ”The Persistence of Mutual Fund Performance.” Journal of

Finance 47, 1977-1984.
and _, 1994, ”A Study of Monthly Mutual Fund Returns and

Performance Evaluation Techniques.” Journal of Financial and Quantitative
Analysis 29, 419-444.

Henriksson, Roy, 1984, ”Market Timing And Mutual Fund Performance: An
Empirical Investigation.” Journal of Business 57, 73-96.

Ibbotson, Roger and Gary Brinson, 1993, Global Investing. McGraw-Hill Inc.

Ippolito, Richard, 1992, ”Consumer Reaction to Measures of Poor Quality:
Evidence from the Mutual Fund Industry.” Journal of Law 6’ Economics 35,
45-70.

Jagannathan, Ravi and Robert Korajczyk, 1986, ” Assessing the market timing

performance of managed portfolios.” Journal of Business 57, 217-235.

119
Jensen, Michael, 1968, ”The Performance of Mutual Funds in the Period 1945-

1964.” Journal of Finance 23, 389-416.

__, 1969, ”Risk, the Pricing of Capital Assets and the Evaluation of Investment
Portfolios.” Journal of Business 42, 167-247.

]obson, J. and Bob Korkie, 1981, ” Performance Hypothesis Testing with the
Sharpe and Treynor Measures.” Journal of Finance 36, 889-908.

___, and ____, 1989, ”A Performance Interpretation of Multivariate Tests
of Asset Set Intersection, Spanning, and Mean-Variance Efficiency.”
Journal of Financial and Quantitative Analysis 24, 185-204.

Kodde, David and Franz Palm, 1986, ”Wald Criteria for Jointly Testing Equality
and Inequality Restrictions.” Econometrica 54, 1243-1248.

Lehmann, Bruce and David Modest, 1987, ”Mutual fund performance evaluation:
A comparison of benchmarks and benchmark comparisons". Journal of
Finance 42, 233-265.

Odier, Patrick and Bruno Solnik, 1993, ” Lessons for lntemational Asset
Allocation.” Financial Analysts Journal March-April 1993, 63-77.

Sharpe, William, 1966, ”Mutual Fund Performance.” Journal of Business 39, 119-
138.

, 1992, ” Asset Allocation: Management Style and Performance
Measurement,” Journal of Portfolio Management 18, Winter, 7-19.
Sirri, Eirk and Peter Tufano, 1993, ” Buying and Selling Mutual Funds: Flows,

Performance, Fees and Services”, Harvard Business School Working Paper.

 

120
Solnik, Bruno, 1974, ” An Equilibrium Model of the International Capital

Market.” Journal of Economic Theory August 1974, pp. 500-524.

__J 1974, ”The International Pricing of Risk: An Empirical Investigation of The
World Capital Market Structure.” Journal of Finance 29, 365-378.

___, 1991,1nternational Investment, 2nd edition. Addison-Wesley Publishing
Company.

Treynor, J., 1965, ”How to Rate Management of Investment Funds.” Harvard
Business Review 53, February 1965, 63-75

___, and F. Mazuy, 1966, ”Can Mutual Funds Outguess the Market”. Harvard
Business Review 54, 131-136

Wolak, Frank, 1987, ”An Exact Test for Multiple Inequality and Equality
Constraints in the Linear Regression Model.” Journal of the American
Statistical Association, September 1987, 782-793.

1989a, ”Local and Global Testing of Linear and Nonlinear Inequality

 

Constraints in Nonlinear Econometric Models.” Econometric Theory 5, 1-35.

 

1989b, ”Testing Inequality Constraints in Linear Econometric Models.”
Journal of Econometrics 89, 205-235.

1991, ”The Local Nature of Hypothesis Tests Involving Inequality

 

Constraints in Nonlinear Models.” Econometrica 59, 981-995

 

 

nICHIGnN STATE UNIV. LIBRRRIES
llllulllllllllllllllllllllllllllllllll
31293014215408