i
.M‘afia

m

 

.3. 3 1!:
92......le
lrfxr‘ 1....
16 It.

 

. «1. xi!
[5 or: it
s; 5":

 

 

 

.1
1:. v3.2: .

$03.37. . ‘
I . . Qlkrnkl:
‘! ‘ $1. I

 

THESlS

J .
Ml HIGAN STATE UNIVE BRARIES L

Mllllzl l l filll’llillllllllllll)!ll

3 93 01029 6592

ll

 

 

 

 

 

 

 

 

 

This is to certify that the
dissertation entitled
A Study of the Price and Volatility of Closed-End
Country Fund Shares and Net Asset Value
presented by

Woojin Hahn

has been accepted towards fulfillment
of the requirements for

Ph . D . degree in Bus . Adm .

 

 

7<LJCKJ¢4

Major professor

Date SQ£T 277. ’993

 

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

 

LIBRARY
Mlchlgan State
Unlverslty

 

 

 

PLACE IN RETURN BOX to roman this checkout from your record.
TO AVOID FINES Mum on or baton date duo.

DATE DUE DATE DUE DATE DUE

       

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU I. An Affirmative Adlai/Equal Opportunity InotItqun
Wm:

A Study of the Price and Volatility of Closed-End
Country Fund Shares and Net Asset Value

BY

Woojin Hahn

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
Graduate School of Business

1993

ABSTRACT

A Study of the Price and Volatility of Closed-End
Country Fund Shares and Net Asset Value

BY

Woojin Hahn

Two factors are known to affect closed-end country fund
premiums/discounts: international investment ownership
constraints and investor sentiment. This is an analytical and
empirical investigation into which really explains changes in
premiums/discounts to net asset value, the variance of fund share

prices, and the variance of net asset values.

Copyright by
WOO JIN HAHN

1993

To my family

iv

ACKNOWLEDGEMENTS

I would like to thank all the members of the Finance
and Insurance Department for their strong support and
intellectual help during the process of my study.

Dr. Richard Simonds, the Chairperson, and Dr. John
O'donnell will be always remembered for their considerate
guidance and encouragement.

Dr. Kirt Butler’s respectable intelligence and
dedication made this dissertation come to an end.

Special thanks to Dr. Chang Ahn for his direction for
the theoretical part of this thesis, and to Dr. James
Wiggins for his helpful comments for the empirical part. I
also wish to express my deep thanks to Dr. Raoul Lepage for
his guidance.

I appreciate Dr. Simon Wheatley and the Korea Fund for
providing indispensable data for this research and Dr.
Charles Lee for his valuable comments.

table of Contents
Introduction and Literature Review

Explanations of closed-end fund discounts/premiums.

International Asset Pricing which incorporates the effects
of investment restrictions

Investor sentiment hypothesis
ESSAY I..Eun and Janakiramanan’s model and its extension
1. Eun and Janakiramanan’s theoretical model

2. Extension of Eun and Janakiramanan’s model
K

2.1. Extension on premiums/discounts
2.2. Extension on the volatility of CECF share price and NAV
Appendix I-A Volatility of CECF share price and NAV and how

the volatility difference changes with change
in 6

ESSAY II. Lucas-tng.mode1 (A consggption-based asset
pricing.modell

1. Introduction
2. Theoretical reasoning
3. Model

4. Implications and Analysis

Appendix II-A Exponential Utility

Appendix II-B Log Utility

vi

Appendix II-C Power Utility

ESSAY 111. An gygirical Investigation into the Determinants
of Closed—End Cogntry Fund PremiumsLDiscounts and

Vblatilities

Introduction

1. CECF premiums/discounts and ownership restrictions

1.1. Long-run evidence : The case of Korea Fund
1.2. Short-run evidence
1.3. Critique of BBNW study

1.4. Testable hypotheses

2. Volatility ratios of CECF share prices over underlying
NAVs

3. The relation between premiums/discounts and
premium/discount volatilities
Appendix III-A. Events and data update

Appendix III-B. Average annual premiums/discounts and
institutional ownership of CECFs

Summary and conclusion

Bibliography

vii

III-1:

III-2:

III-3:

III-4:

III-5:

III-6:

III-7:

III-8:

III-9:

III-10:

III-11:

List of Tables

Korea Fund’s premium and ownership
constraints (1984.9-1987.3)

Dummy variable regressions after
differentiating capital inflow and outflow
restrictions

Regression of dummy variables over
World-adjusted changes in premium

Regression of NASDAQ index return over World
CECF premium change

Korea fund’s premium and Korean stock
exchange index

Change in premiums/discounts and NAV return
(a proxy for foreign market return)

Relationship between premium and
institutional ownership

Correlation test between CECF premiums and
variance ratios

Logit Analysis

Correlation Table based on monthly
premiums/discounts of CECFs

Correlation Table based on weekly
premiums/discounts of CECFs

viii

104

105

109

111

112

113

115

116
117

118

120

II-l:

III-1A:

III-1B:

III-2:

III-3:

III-4:

III-5:

III-6:

List of Figures

The coordinates of risk aversions which
makes foreign assets to sell at a premium
(6 = 0.1)

The coordinates of risk aversions which
makes foreign assets to sell at a premium
(6 = 0.5)

Variance ratio (Variance ratio of Korea fund
price over variance of Korea fund NAV)

Korea Fund Premium

The relationship between regression
standardized residual and regression
standardized predicted value :

The case of KF

Scatterplot of residuals and predicted
values based on BBNW regression

CECF premium changes during the 3-day
event periods

CECF premium changes around announcements of
loosening restrictions

CECF premium changes around announcements of
tightening restrictions

CECF premiums/discounts and variance ratios

ix

36

37

50
97

98

99

100

101

102
103

1
Introduction and Literature Review

A closed-end investment company neither issues new
shares nor redeems outstanding ones. Hence investors must
buy or sell existing shares at market price on an open
market such as the NYSE. Closed-End Country Funds (CECFs)
are closed-end mutual funds invested in the assets
(typically common stocks) of a foreign country.

CECFs trade like ordinary stocks in the domestic stock
market. Fund share price is determined not only by net
asset value (NAV), as is the case with open-end country
funds, but also by the supply and demand for equity
ownership in the foreign country. Closed-end country funds»
can provide domestic investors with access to foreign
markets. If the foreign and domestic markets are segmented
such as through an investment restriction'on domestic
ownership of foreign assets, then CECFs can trade at
substantial premiums or discounts to NAV.

There are two principal schools of thought regarding
CECF premiums and discounts. One school has focused on the
effect of international asset restrictions on the actions of
foreign and domestic investors using formal utility-based
models. The other school has chosen to emphasis changes and
differences in investors’ expectations.

Agency costs, tax liabilities, and asset illiquidities

have been used to explain closed-end fund discounts.

2
Closed-end fund premiums are often explained as arising from
an investment restriction as in Eun and Janakiramanan
[1986]. --hereafter EJ-- EJ derive a closed-form solution
for international asset prices when there is an investment
restriction. Bonser-Neal, et, a1. [1990] --hereafter BBNW--
find empirical evidence consistent with EJ's model.

Lee, Shleifer, and Thaler [1991] --hereafter LST--
argue that closed-end fund discounts are a proxy for small
investor sentiment. Considering the fact that the major
owners of closed-end funds are small investors, it is
plausible that this sentiment systematically affects all
closed-end fund share prices. Therefore, they assert that
closed-end fund premiums move together following investor
sentiment changes.

This dissertation contains three essays. In the first
essay, I extend EJ's international asset pricing model by
imposing randomness on the supply side in order to find the
volatility implication. In the second essay, based on
Lucas’ [1974] asset pricing model, I derive closed-form
solutions for international asset prices when there are
investment ownership constraints. In the third essay, I
investigate the factors affecting Closed-End Country Fund
premiums/discounts and premium/discount volatilities by
adopting BBNW’s [1990] dummy regression and LST's [1991]
investor sentiment hypothesis. J

In the remainder of this introductory chapter, I review the

3
theories and explanations of closed-end fund premiums and

discounts.

4

Explanations of closed-end fund discountstremiums.

The market price of closed-end fund shares is largely
determined by the demand and supply of shares at a point in
time. Therefore, the closed-end fund share price is not
necessarily the same as its net asset value based on the
underlying securities held by the fund and fund shares can
sell at a discount or premium to net asset value. The
premium(discount) is usually defined as positive(negative)
value of (P-NAV)/NAV or ln(P/NAV) where P is CECF share
price and NAV is net asset value.

Several traditional theories have been put forth to
explain the closed-end fund discount/premium puzzle.

These include

1) Agency costs: Boudreaux [1973] argues that if
management fees are too high or future portfolio management
is expected to be bad, then agency costs could result in a
closed-end fund discount. '

2) Unrealized capital gains of the fund portfolio:
According to Malkiel[1977], this is a reason for the
existence of closed- end fund discounts since unrealized
gains must ultimately be taxed at the rate of a purchaser of
the fund shares.

3) Illiquidity of assets: There are two hypotheses to
explain observed fund discounts based on asset illiquidity.

The ’restricted stock’ hypothesis says that the fund with

5
the letter stock1 has a market value which is lower than its
complement.
The ’block discount' hypothesis says that the achievable
gains are lower in a closed-end fund because of the block
trading necessary to sell off the assets upon liquidation.

4) Manager’s ability to predict security prices:
Thompson [1978] argues that the expected productivity of
fund management is responsible for the discount/premium.

5) Lack of aggressive advertising: Wisenberger Service
cites a lack of aggressive advertising and sponsorship as a
possible reason for the discounts.

None of these explanations appear capable of fully
explaining the existence of closed-end fund
discounts/premiums,
especially for Closed-end Country Funds (CECFs) which often
sell at substantial premiums to Net Asset Value.

Some recent studies seem to provide more plausible
explanations. One representative study is Lee, Shleifer,
and Thaler’s [1991] ”Investor sentiment and the closed-end
fund puzzle”. These authors use a model in which markets
are not semi-strong form efficient to explain the average
discount on domestic closed-end funds. In their model,
irrational noise traders are accountable for a larger

fraction of fund share trades than of trades on assets

 

1Letter stock is a restricted stock which can sell only
with a letter attached.

6
underlying the shares. These noise traders impose
additional risk on fund shares relative to their underlying
assets. Thus, funds trade on average at discounts. Also,
as noise traders become excessively optimistic, closed-end
funds sometimes sell at premiums, especially at inception.
Delong et al [1990] and LST [1991] find that domestic
closed-end fund premiums occur only at IPO and suggest that
the CECF discount reflects the noise trader risk. LST
[1991] conclude that closed-end fund discounts are a measure
of the changing sentiment of individual investors toward
closed-end funds.

Another interesting study is Bonser-Neal, Brauer, Neal,
and Wheatley’s [1990] ”International Investment Restrictions
and Closed-End Country Fund Prices". The purpose of BBNW’s
study is to examine whether changes in closed-end country
fund premiums are related to announcements of changes in
international investment restrictions. In particular, they
test whether an announcement of liberalization of a
restriction is associated with decline in the premium and
announcement of tightening of a restriction with a rise in
the premium.

In the real world, however, the existence of investment
restrictions does not always create a premium. Some closed-
end funds are sold at premiums and some funds are sold at
discounts. Therefore, BBNW [1990] looked at the relationship

between changes in the restriction and changes in the

7
price/(net asset value) ratio. BBNW's conclusion is
thatgovernment regulation can be effective in segmenting

markets.

8
International asset pricing theory with investment

restrictions

Several CAPM-based international asset pricing models
have incorporated the effects of investment restrictions on
foreign and domestic asset prices.

Black [1974] & Stulz [1981] studied the effects of
international investment barriers that take the form of
taxes on holding foreign assets. In Black’s model,
investors are taxed on long positions in foreign assets but
are subsidized in short positions. In Stulz’s model,
investors are taxed on both long and short positions. Both
models foresee that the expected return on a foreign asset
held long will exceed the expected return on a domestic
asset of equal risk by the rate at which domestic holdings
of the asset are taxed. This assures that the after-tax
returns are the same. Closed-end country funds represent
long positions taken indirectly by domestic investors in
foreign assets. While the underlying shares of these funds
are foreign assets, the funds’ shares are domestic assets.
Thus, these models imply that barriers that increase the
cost for a domestic investor of holding foreign assets will
raise the return required on their shares.

Hietala [1989] studied the Finnish stock market in
which Finnish investors can hold only Finnish shares while

foreign investors can hold both Finnish and foreign shares.

9
Finnish stocks are divided into restricted (for Finnish
investors only) and unrestricted shares. By using an
equilibrium asset pricing model, Hietala explains the
premium on the unrestricted shares. The premium arises if
and only if the price of the unrestricted stock is decided
by foreign investors. This is true if the foreign investors
require a lower premium on this stock than do domestic
investors.

Errunza 8 Losq [1985] derive a closed-form solution for
the case when domestic investors are legally prohibited from
investing in foreign assets while foreign investors are not
prohibited from investing in the domestic market. They show
that the required return on the foreign securities is higher
than the return required under no such restrictions.

Recently, Bergstrom et. al. [working paper] studied
asset pricing with inflow and outflow constraints in the
Swedish market. In Sweden, there exists a switch-currency
constraint which forces Swedish investors to pay a premium
for foreign assets. The analysis of the switch-currency
constraint is extended by adding Eun and Janakiramanan’s
delta constraint on foreign holdings of domestic (i.e.
Swedish) assets. Consistent with the model, Swedish
investors pay a premium for foreign assets and foreign
investors pay a premium for Swedish assets.

The existence of both types of premiums suggest,

however, that markets are segmented as a result of the

10
ownership constraints.

Eun 8 Janakiramanan’s [1986] model defines a specific
form of barrier to international investment. Their ’delta
constraint’ represents a binding limit on the proportion of
foreign shares which can be held by domestic investors.
When delta constraint is binding in the foreign market,
domestic investors must buy stocks from other domestic
investors. This constraint is applied only to domestic
investors and not to foreign investors. In other words,
domestic investors cannot hold foreign securities beyond a
maximum value so that the ownership constraint is effective.
In contrast, foreign investors are allowed unrestricted
access to the domestic market. This investment restriction
proxy is measurable. They derive a closed-form solution for
international asset pricing in a two-country, two-agent
(Domestic and Foreign investors) setting when the market is
imperfect due to an investment restriction on foreign
ownership. The restriction on domestic investment in
foreign assets results in domestic investors paying a
premium for closed-end country fund shares. Conversely,
foreign investors price these shares at a discount relative
to unrestricted equilibrium prices.

In other words, when there is an ownership constraint
on the fraction of foreign assets which can be held by
domestic investors, there are two groups of foreign shares:

restricted shares that only foreigners can buy and

11
unrestricted shares that both domestic and foreign investors
can buy. The foreign government establishes limits on the
fraction of shares that are unrestricted, and when these
limits are securing, the unrestricted shares sell at a
premium relative to the restricted shares. This premium
depends on the covariance matrix of returns and on investor
preferences. They show that the prices of foreign assets for
the domestic and for the foreign investors can be expressed
as a function of the proportion of foreign equity ownership
allowed.

In their model, two different prices rule for foreign
assets reflecting a premium offered by domestic investors
above the full integration prices and a discount below the
full integration price required by the foreign investors.
Both the premium and the discount are determined by the
severity of the delta constraint. The tighter the
restriction, the greater the domestic premium and the
foreign discount on unrestricted shares.

According to this international asset pricing theory,
closed-end country fund (CECF) premiums/discounts seem to be
due to investment ownership constraints. CECF share prices
can be viewed as foreign asset prices for domestic investors
and net asset values can be viewed as foreign asset prices
for foreign investors. If 6 constraint is binding, CECF
should sell at a premium over NAV and as delta constraints

become less binding, the CECF premium over NAV should fall.

12

In this research, I test the implications of Eun and
Janakiramanan’s model for CECF price and volatility for the
long run and short run with and without controlling for
investor sentiment and differential risk aversion.

Both change in the delta constraint and change in
investor sentiment is likely to affect the CECF premium.
Therefore, it is difficult to believe that CECF premium
behavior can be fully explained with only one of these
factors.

There is empirical evidence that CECF prices respond
to information in announcements of changes in investment
restrictions, presumably because such information is
relevant to diversification-oriented investors. This
implication was tested by BBNW [1990]. Based on the
hypothesis that premiums decrease as investment restrictions
are loosened, they tested how premiums/discounts change
around 23 announcements of changes in investment
restrictions. They report significant correlations between
premiums/discounts and tightenings/loosenings over the 3-
week period surrounding such announcements. Tightening of
investment restrictions causes an increase in CECF premiums
and loosening of investment restrictions does the opposite.

I have several concerns regarding this study. First,
BBNW use only 23 events and many of them are questionable in
terms of their influence over CECF prices. Second, it is

not clear whether a loosening of capital outflow constraints

13
decreases or increases the premium. 2 Third, even after
adjusting for expected dividends and missing variables BBNW
still found insignificant results for two of the five funds
they tested.

After adding eight more events in Essay III, I
calculate a world CECF premium change index based on 14
country funds and run a regression of the world-adjusted
premium change over the dummy variables as was done by BBNW
to see if the excess premium/discount change can be
explained by investment restriction changes’. I report
results which do not support BBNW and the empirical evidence
strongly suggests that premiums/discounts changes are
subject more to 0.8. investor sentiment than to changes in
the delta constraint.

If a foreign securities market is segmented due to
investment restrictions, domestic investors have an
incentive to pay a premium over unrestricted equilibrium
prices in order to capture the diversification benefit
offered by foreign securities. However, a distinction
should be made between direct investment in foreign

securities markets and indirect investment through CECFs.

 

’Bergstrom et al [1990] find that a binding switch
currency constraint (capital outflow constraint) makes
foreign securities less attractive. This constraint and the
delta constraint (capital inflow constraint) operate in
different directions.

3Lee et al [1992] find that CECF premiums move together
with investor sentiment changes.

14
With both direct and indirect investments, changes in
investment restrictions are likely to affect CECF premiums
over NAV. With indirect investment, NAVs are determined by
foreign market price changes while CECF prices are more
highly correlated with domestic share prices.

Bailey and Lim [1992] find that, while foreign market
index returns are not correlated with U.S. returns, CECF
returns are strongly correlated with the U.S. market
returns. Similarly, Lee et a1 [1993] find that after
controlling for foreign market fundamentals, changes in CECF
prices co-move with U.S. market returns. This suggests that
investors must buy foreign equity to achieve the full
diversification benefit. CECFs are not effective

diversification vehicle into foreign markets.

15

Investor sentiment hypothesis

In an early example of the expectational school,
Malkiel [1977] suggests that closed-end country fund
premiums or discounts result not from any market
imperfections but rather from investor fascination or
disappointment with foreign securities. In this view, the
effects of investment restrictions on the premiums of the
closed-end country funds may be small. Some recent evidence
(e.g. BBNW [1990]) suggests that changes in investment
restrictions affect closed-end country fund premiums
significantly. LST [1991] use the term ’investor sentiment’
in a modern version of Malkiel’s behavioral explanation.

In a recent study, Lee, Kim, and Bodurtha [working
paper] showed that after controlling for foreign market
fundamentals, the stock prices of country funds co-move with

U.S. market returns, but their net asset values do not.

What they found are :

l. CECF premiums/discounts move together.

2. CECF premiums/discounts are not correlated with U.S.
market returns while CECFs are.

3. Changes in NAV are strongly positively correlated with

foreign market returns and the NAVs are not correlated

16
with U.S. market returns. This is consistent with
Bailey and Lim’s [1992] finding that CECFs are not
useful for diversification purposes. Investors should
directly participate in local markets rather than use

CECFs to achieve the diversification benefit.

Under Lee et al’s view, some of the time noise traders
are optimistic about returns on these securities and drive
up the prices relative to fundamental values. Unlike other
assets, in case of CECFs the fundamental values are
observed. Therefore, noise traders’ sentiment can be
captured by the magnitude of premiums/discounts. In this
way, investors with changing return expectations create
stochastic changes in CECF premiums/discounts. If the same
investors are investing in both the underlying securities
and in fund shares, then any changes in sentiment affects
both the NAV and share price resulting in no change in
premiums/discounts. Therefore, changes in premiums/discounts
reflect not the aggregate effect of investor sentiment
changes but the differential effect of the sentiment of the
CECF-investing clientele relative to the investing clientele
of the underlying assets. An index of CECF
premiums/discounts provide a measure of the differential
sentiment of U.S. investors relative to their foreign
counterparts.

When domestic investors are bullish about stocks in

17
general, they bid up the price of foreign country funds
relative to their NAV, as well as the price of other assets
in which they hold large positions. If this sentiment is
wide-spread, it will be priced in equilibrium. {Arbitrage
across market is costly, so local sentiments persist and are

priced in equilibrium.

18

Essay I..Eun and Janakiramanan’s model and its extension

In this essay, I briefly review Eun and Janakiramanan’s
[1986] theoretical model of investment restrictions and CECF
share prices. I then extend this model of CECF
premiums/discounts to derive the relation between the

volatility of CECF prices and NAVs.
l. Eun and Janakiramanan’s theoretical model

Eun and Janakiramanan’s [1986] model has two basic

assumptions.

1. There is a domestic country with a domestic agent and a
foreign country with a foreign agent.

2. There is an investment restriction on domestic ownership
of the foreign market and no restriction on foreign

ownership of the domestic market.

Pratt-Arrow’s absolute risk aversion measure is given
by A" - -U"'(w")/U"'(w"), where k represents either a foreign
investor (f) or a domestic investor (d). Similarly, Let
foreign and domestic securities represented by k - F and

k - D, respectively. Other variables include :

Uk : Utility of agent k.

CEQk

wk

LR

EJ adopt a

19
: Certainty equivalent of end-of-period wealth
for investor k.

The investible wealth of investor k at time 0.

: The random and normally distributed end-of-

period wealth of investor k.

Amount invested in the risk-free asset by
investor k.
: l + risk-free rate (same risk-free rate assumed
for both investors).
: Vector of number of shares of the domestic
(K - D) and foreign (K - F) securities
outstanding.
: Vector of number of shares of domestic (K - D)
and foreign (K I F) securities outstanding

owned by agent k. ( k e (f, d))

Current price vector distributed as

multivariate normal.

End-of-period Price vector covariance matrix
distributed as multivariate normal.

Expected price vector.

negative exponential utility function as follows.

U*(w*)=-exp[—A*W*], “>0. (1)

Investors maximize their expected utility of end-of-

period wealth E[U"(W")]. Since W" is normal and the prices of

20
the securities F and D are joint-normally distributed,

expected utility can be expressed as,

.. — k ..
Elvwwkn=-exp[-A*[W"-i2-var(wk)ll ., (2)
Therefore, agent k’s optimal investment is to maximize the

certainty equivalent of end-of-period wealth,

_. k ~
Maximize CEQ" (Certainty Equivalent) =Wk-A7var(W*) . (3)

wk=n,§"PD+n:’P,+L 1‘. (4)

Subject to the budget constraint,

Expected end-of-period wealth is given by

I I
”tan: up”): PP+L k1

wimp-Par) +n:’<u.-P.r) +w*r- <5)
The variance of end-of-period wealth is given by
Var(W") =nD“’VDn,f+2n,§"VD,n:+n}'V,n}. (6)
Therefore,

CEQ*=n;’(pD-P0r) +n,f’(p,—P,.r) +Wkr

21

k I I I
—% In;c Von§+2nz§ Vppnfi‘mi Vpnil - <7)

Maximization yields the following first-order conditions

 

 

k .
dCECki =(pD-PDI) _Ak[VDnDk+VDrn;] =0 (8)
an); I
and
k
5:132 = (pfppr) "AkIVzIJank+Vp-n:] =0- (9’
n, '

Rewriting equations (8) and (9) with respect to the demand

vectors n: yields

 

 

 

 

 

 

 

120k 1 VD VDF- I‘D-Par
k=__k o (10)
n, A V!» V. Fr'Prl’
By defining the above inverse matrix as
_ VD VDP-1 _ I‘D PD? (11)
VII)? VP P1,)? PP ’
we can rewrite the aggregate demand functions as
123‘ u -Ppr
“its " (12)
n? A "errr

 

 

 

 

The market clearing conditions are required to arrive at

22

equilibrium asset prices. They are,

nf+nf€N3 iED. (13)

HidSDNi and n1f=N1—n1d 16F. (14)

With a short sale restriction,

nfzo iEF (15)

Since the 6 constraint is assumed to be binding, the foreign
securities will be held long, in aggregate, by the domestic
investors. Therefore, the short-sale restriction need not be
considered.

The foreign asset price for the domestic investor
equals the full integration price plus the premium offered
by the domestic investor due to the restriction. In this
case the demand exceeds the unconstrained equilibrium

supply. That is to say,

P}=P}+tt for 100. (16)

Conversely, the foreign asset price for the foreign investor
equals the full integration price minus the discount
demanded by the foreign investor. Here, the unconstrained
equilibrium demand is less than the unconstrained
equilibrium supply.

Plugging these into the aggregate demand functions in (12)

23

and solving for the price vectors results in

_ __95 .A ()_,q 0]
(my) 74)}

where world risk tolerance is given by

F5 PD

Fr

 

 

 

 

 

1
o=_{
P} I

i=i+i, (18)

First, we need to find equilibrium asset prices with no
restrictions. With no foreign ownership constraint (delta81)
the last two terms of equation (17) go to zero.

Therefore,
D
«=31-l. (19)

By using the market clearing conditions in (13) and (14), we

get the following two equations:
ADON,= {Emma-Pg] +I‘,[p.,-P;r] I -I‘,:tr. (20)

AF(1-5)N,= {rpm-p011 +r,[p,-p;r] I +l",.1r. (21)

Substituting (21) from (20) gives

[A’-(A"+A’)6]NP=I‘(n+l)r. (22)

24

Substituting (19) for n in (22) and rearranging provides

An:— [AF(1-b) -A "1 I: where, 2:: [VF-V£FV51VDF]NF. (23)

Substituting (19) for 1 in (22) and rearranging provides

um; [A"-A"b] Z. (24)

Consequently, equilibrium asset foreign price (P?) under no

restriction (delta - l) is given by
pie; minimums»! }. (25)

Foreign asset prices for each agent are then

P:=P;+1t=% {p,—A"[V,’,,ND+V,N,] + (A"-A”6) 2}. (26)
and

Pi=P;- =%{u,-A"IV£pND+V,N,] -(A’(1-b)-A")2} . (27)

25

2. Extension of Eun and Janakiramanan’s model

In this section, I extend Eun and Janakiramanan’s [1986]

model and derive implications.

2.1. Extension on Premiums/Discounts
From (26) and (27), we can derive the domestic

investor’s premium on foreign assets as follows.

Rf—P;=u+l (28)

=%[A"(1-6)-A”6]2. (29)

From equation (2), we find that whether closed-end
country funds are selling at a premium or a discount is

determined by :

1. If AP > (1/6 - 1) AF , the fund is selling at

a discount. (30)
2. If A” I (1/6 - l) A’ , the fund is selling at NAV. (31)
3. If API< (1/6 - 1) AF , the fund is selling at
a premium. (32)

Where A? is Absolute Risk Aversion of agent k. (k e D, F)

For example, let A” and A" take on values in an

26
arbitrary range from 0 to 2, and investigate the pricing of
CECFs for two distinct values of 6 in (29). The following
examples use 6 I .1 (Figure I-lA) and 6 I .5 (Figure I-lB).
In these cases, if the coordinate (A',A") falls in the
shaded area, the fund sells at a premium. If (A',A") falls
on the line, it sells neither a discount nor a premium. If
(A’,A°) falls below the line, it sells at a discount. A
discount is allowed by not restricting the binding condition
to 0<6<A"/A". If we assume 6 is still binding in the range
between A"/AD and 1, then a discount is allowed.

We know that (1/6 - 1) is the slope of the line and it
decreases as 6 increases. This means that as 6 increases
the shaded area becomes smaller. This can be interpreted as
meaning that the probability that a CECF sells at discount
(premium) becomes higher as 6 increases (decreases). From
the figures, we can see that as delta changes (loosening of
restrictions), at the same risk aversion coordinates the
chance of the foreign asset selling at a premium is lower at
higher levels of delta, 6. Let M” and M’ represent the
wealth of investors in each market and Wand WA:—
represent the mean per unit absolute risk aversion of the
investors. Then

1 _ _I-
“A—B‘MD*(—A-P) (33)

and

27

1
“DIF'1‘ 1‘34

Plugging these into equation (30) yields the following

result:

 

6 > 1 = 1 (35)
A_:+1 M7x(—._r;)
A1
———+1
mull?)
A:

Assuming equal average risk aversion for foreign and

domestic investors, then for the fund to sell at a discount,

-T_;—T'

-—---—-. (36)
111’ A,”
and
1
_, 7
6> m1 (3 )

where M is the size of the foreign market relative to the
domestic market. Conversely, for the fund to sell at a

premium,

1
MM”. (38)

 

This means that as the ratio M of the size of the foreign

market to that of the domestic market grows, the threshold

28
delta which determines whether a fund sells at a premium or

a discount decreases.
2.2. Extension on the volatility of CECF share price and NAV

If the premium/discount was a constant fraction of NAV,
investors would not care about the volatility of the
premium/discount. However, fluctuations in the
premium/discount appear to be mean reverting (Sharpe and
Sosin (1975)). Others document significant positive
abnormal returns from assuming long positions on funds with
large discounts. Does this hold true for CECFs? If so,
what are the driving forces and are there any differences in
volatilities between premium funds and discount funds?
Unlike domestic CEFs, there are several CECFs which sell at
high premiums.

Another interesting motivation for a study of the
volatility of CECF prices and NAV is based on the noise
trader effect. Lee, Schleifer and Thaler [1990] assume that
individual investors are irrational noise traders and
institutional investors are rational. If this is true, the
institutional ownership of CECFs selling at a premium should
be lower than funds selling at a discount. Assuming that the
volatility of NAV reflects the fundamental volatility, we
can conjecture that the variance ratio (Var(P§)/Var(P§)) of

funds selling at a premium should be higher than funds

29
selling at a discount.

The theoretical motivation for a study of price
volatility is based on equations (3) and (6) in this paper.
We can see that the variance of wealth is a function of
price volatility and that CEO is a function of the variance
of wealth. Therefore, the variance of the CECF price affects
the utility of investors. EJ’s model looks only at the
demand side assuming that the supply (endowment) is
exogenous. However, we can impose randomness on the supply
as suggested by Lucas [1978]. Let, ND and N, be the
supply vectors such that ND~N(]JD, ED) and N,~N(p,, 2,) . This
will allow us to test the relative volatility of CECF prices
and NAVs with respect to changes in delta.

With stochastic supply, the prices in EJ’s model become :

P155135”: =—:- {MFA '[VI’FNN VrNP] + (A ”-A ”6) [V,- Wyn-1%,] NF} (39)

and
Pi=P£~A=§ {p,-A"[v’mND+V,N.J -(AF(1-a) -A ") tv.-V£FV5‘VD,1N,} (40)
Rearranging these prices with respect to ND and NF yields

Pia; {MFA "VIDFND- [A "Vivi/51 VDP+A D6 “’1" VIIJPVDI Var] ] NF} ' (4 1)

and

30

P15; {up-A "Vépzvp— [A .(1 -6) (v): Vii/51v”) +4 "Vépvslvmi N.) . (42)

For notational convenience, let

MJL =A "V3,, (43)
M.=A "vspvalvppm ”6 [Vp‘Vz'aerSlep] . (44)

and
M3 =A "my; VDF+A 1’ ( 1 -b) [v,- V3,.V51 vm] . (4 5)

Now we can obtain a closed-form solution for the variance of

CECF prices P‘,1 and NAV Pfi.

Var (P1?) = Var ( % I -A " up-ND- [A "Vz'arVfii VDHA ”5 [Vp- Vz’apVE 1Vap] ] Np} )

“:15 [MIEDM1’+M22PM;+2M12DPMZI] ° (46)

Var(pr) =Var(-:'-_ {-A" DpND- [A’(1-6) (Vp-VfngISIVDp) +11" DPVDIVDp] Np})

:ai%(fiaznufufifihfig+ZMyfimfig]. (47)

Changes in delta result in changes in CECF price
variance and NAV variance according to this extension of
EJ’s model. In appendix I-A, we show that there are two

cases which yield different results.

31

Let n20 and 120.

Case I : If ADZA' and Cov(P; , n)>0 (i.e. Cov(N°, N')<0), then
Var(P€)/Var(P§).is greater than 1, and an increase in delta
should result in a decrease in the variance ratio
((Var(P:)/Var(P§)) by decreasing the volatility of P‘,’ and

increasing that of P‘.

Case II: If A”<A’ and Cov(P; , n)<0 (i.e. Cov(N”, N’)>0), then
Var(P§)fVar(P§).is less than 1, and an increase in delta
should result in a increase in the variance ratio
((Var(P:)/Var(P§)) by increasing the volatility of P: and

decreasing that of P‘.

If we let n<0 and l<0, then variance ratios are reversed in
each case, that is, Var(P€)/Var(P§) is less than 1 in case
I and Var(P§)/Var(P§).is greater than 1 in case II. Also,

Q[Var(P§) - Var(P§)] is positive in case I and negative in
66

case II.

32

Appendix I-A.Volatility of CECF share price and NAV and

how the volatility difference changee‘with

change in 6

From (26) and (27), P: I P; + n and Pf. I P; - 1
Var(P§) I Var(P;) + 2Cov(P; , n) + Var(n).

Var(P§) I Var(P;) - 2Cov(P; , l) + var(l).

when

Vanp‘) =Var(p‘) -2cOv(p‘ £3) + (EPVarht)
r P FIAP AD -

=Var(P‘) -2 AFCovU" 1:) +( AP)3Var(1t)
’ F2 1". F2 -

Therefore,
Var(P§) - Var(P§)
- 2(1 + A’lA")*Cov(P; , n) + (1 - (A’/A8)’)*Var(n')

= (1 + A’lA”){2Cov(P; , n) + (1 - (A’/A°))*Var(n)}

(A'll

(3‘2)

(A-3)

(A4)

The relative size of Var(P:)iand Var(P§) is determined by

the following conditions.

33

Let n20 and 120.

Case I : If ADZA’ and Cov(P; , n)>0 (i.e. Cov(N”, N')<0‘)I

then Var(P‘,‘) z Var(Pf.)

Case II: If A”<A’ and Cov(P; , n)<0 (i.e. Cov(N°, N’)>0), then
Var(P:) < Var(Pf,)
If we let n<0 and 1<0, then the above relationship is

reversed.

Sign of change in [Var(P‘,‘) - Var(P§)] with respect to

change in 6 is also determined by these conditions.

Let n20 and 120.

Case I : A°2A’ and Cov(P; , n)>0 (i.e. Cov(N", N’)<0)
Var(P§') - Var(P,‘,) - (l/r’)[2(1 + A’lA") Cov{(VD,’ND + V,N,)
(4)") I (A' " 195) [Vt-Vor'Vo'JVonNp}
+ (l-(A'/A”)’)*Var((A' - A06)

[VP-VD? I VD- 1Vin! ] Nr ) ]

" (1/r')[2(1 + AP/AD) C°V{(vor'No T err)

("A"): A'[V,-VD,’VD'1VD,]N,} 7' 2(1 + A’IAD)

 

‘C°V(P;: n) ' C°V{(Vpp’No + VFNP)(-A')I

. (A. " PD?) [Vr'vorvo-1vor ] Nr}
For this to be positive, [V’WCov(ND,N,) + V,Var(N,)]
should be negative. Then Cov(ND,N,)<-V’,,,"V,Var(N,)<0.

34
Cov{(VD,’ND + V,N,)A". A°6[V.-Vnp’Vo”Vo.]N.} +
(1-(A’/A”)’)*(A' " AD5)2[Vp‘Vop’Vo'1Vop]
Var (11,) [V.-V.. ’Vo'1er] ’ I

(A'5)

6[Var(P§) - Var(P,‘)]
66

=_ (1/r’)[2(1 + Ar/ADHXD COV{ (vor'Nn + V,N,)A',
[Vr'vor’vn-lvorlnr} " 2AD(1'(A'/AD)2)(ALA06)*

[VP-VD! ' VD-1vDF ] Var (NP ) [VF-VD! ' Via-Ivor J ' 1

= (l/r’)[2(l + A’/A”)A° [Cov{(VD,’ND+ V,N,)A",
[Vy'er'Vn'IVnrJNd ' (1'(A'/AD) ) (A' "' A05“
[Vr‘Vor ' vn-lvor ] Var (N, ) [Vr‘Vnr ' Vo'lvnr] ' ] 1 <0
(18-5)
where, (l + A’/A°)A” is positive, Cov{(VD,’ND + V,N,)A",
[V,- ”VD'1VD,]N,} is negative and (l-(A’/A”)) is positive.
(A' - AD6) is positive since binding 6 has the range of
0<6<(A"/A°) . Finally, [V,-VD,’VD"VD,]Var(N,) [V,-VD,’VD"VD,] ’ is
positive. Therefore, if ADZA' and Cov(P; , n)>0
(i.e. Cov(N", N’)<0), then

QIVagtPSl - Var(P,‘)] <0.
66

Case II: A°<A’ and Cov(P; , n)<0 (i.e. Cov(N”, N')>0)

§[Var(P§) - Vermin >0.
66

35
Let n<0 and 1<0. Then, by the same analysis as above

Case I : A”.>.A' and Cov(P; , rt)>0 (i.e. Cov(N°, N’)<0)

6 Var P" - Var Pf >0.
66

Case II: A”<A’ and Cov(P; , n)<0 (i.e. Cov(N°, N’)>0)

6IVarIP2) - Var(P,‘)] <0.
66

36

Figure I-lA

The coordinates of absolute risk aversion measures

A’(foreign) and A°(domestic) which make foreign assets sell
at a premium (6 I 0.1) .

 

 

 

IDELTA==J
'35 2"
3831.5“
fl .-
I13 5 ‘1”
E29
33(05
2 o

 

 

 

37

Figure I-lB

The cgordinates of absolute risk aversion measures
A’(foreign) and A°(domestic) which make foreign assets sell
at a premium (6 I 0.51 .

 

 

[DELTAH=.5
or: 2.)-
alif§15*
02? I'm
EMS g
8 34:05
2 o

 

 

 

 

 

38

ESSAY II. Lucas-tng.mode1 LA consggption-based asset
pricing.model)

1. Introduction.

Few investors are more susceptible to turbulent times
than the owners of closed-end funds (CECFs). They often
react in panic to news headlines. This anxiety comes from
the fact that the predominant holders of closed-end country
funds are small non-institutional investors who normally are
less informed about foreign assets than the local people.
The sentiments of these investors drive CECF share prices
and this makes CECF share prices Ram,mpre volatile than
that of the underlying net asset values. Currency
fluctuations and political risk could be sources which make
fund share volatility higher than net asset value
volatility. As an example, Figure II-l presents the ratio
Var(Price)/Var(NAV) for the Korea Fund over the period 1985-
1991 using monthly observations. The Korea Fund’s share
volatility is much higher than its NAV volatility.

While I was working on the Eun and Janakiramanan
extension, I became interested in the theoretical variance
of CECF and its NAV. However, in the Eun & Janakiramanan’s
model, variances couldn’t be obtained without adding

randomness, since there is no random variable in price

39
functions. In particular, how can the volatility of fund
share price and its net asset value he obtained in a
theoretical framework, when there is an investment
restriction like that in Eun and Janakiramanan [1986]? I
decided to impose the investment constraint in a Lucas-type
[1974] consumption-based general equilibrium asset pricing

model.
2. Theoretical reasoning.

Eun and Janakiramanan’s [1986] asset pricing model
basically follows the CAPM approach along with two agents
and a restriction
on foreign ownership (a ’delta’ constraint). As such, it is
a one-period pricing model with constant parameters. Under
these circumstances, we can not find the variances of both
net asset value and fund share value. A different approach
is required to obtain the variance of net asset value and
fund share price.

Lucas’s [1974] stochastic equilibrium asset pricing
model is the basic framework for this study. This model
assumes a one-good, pure exchange economy with identical
consumers. An asset is a claim to all or part of the output
of one of these units. The Lucas model basically assumes
that the endowment is stochastic over the period following a

Markov process defined by its transition function. In this

40
study, due to an assumption of time-additive preference, a
constant discount rate is assumed. By using the utility
maximization rule, the asset price can be endogenously
determined as a function of y (endowment).

I introduce this one-period asset pricing model to
include a random variable into the pricing function. In so
doing, I obtain not only the price but also the volatility
of the price.

Suppose that consumption is smooth and endowment (y) is
stochastic over time. As long as the endowment is generous
(endowment is greater than consumption), people will
postpone consumption into the future by increasing their
asset holdings. This will increase asset prices.

On the other hand, if the endowment is poor (endowment
is less than consumption), people will not be able to invest
in assets for future consumption. Thus, the asset price will
not be much affected. From this, we can infer the fact that
the volatility of the fund share price should be higher than
that of the net asset value if and only if the ownership
constraints are binding.

The model applied in this study is based on Lucas’s
[1974] asset pricing model. Michener [1982] (Appendix 1)
used a log utility function which is unbounded and myopic.
This model is a good example of the Lucas-type derivation of

asset price and its volatility. However, the closed-form

41
solution of the asset price does not include either 2
(beginning-of-the period number of shares) or x (end-of—the
period number of shares). This is due to the
characteristics of the log utility function which makes it
impossible to impose a restriction on the price function.
Therefore, the log utility function can not be used for a
volatility test of this kind.

The other utility function tried was the power utility
function which also failed to get the appropriate price
function. I include derivations based on the log and power
utility functions as Appendx II-B and II-C, respectively.

Finally, I used the negative exponential utility
function, which resulted in the price function with x
(number of shares at the end-of-the period) and y
(production or endowment) variables in it. This makes it
possible to compare the variances of fund share price and
the net asset value under the imposition of the restriction.
As is shown in the main text, the price function is not
simple enough for a direct comparison. Therefore, we resort

to numerical analysis to investigate the differences.

3 . Model

Following Lucas [1978], assume a pure exchange economy

with two agents f (foreign) and d (domestic) possessing

42

identical time-additive preferences of the form :
U(Cé") =-%exp(-nC.") (1)

where G; denotes investor k’s consumption at time t and eta
is a constant. Agents f and d can be interpreted as
representing a large number of identical consumers in a
foreign and in a domestic country, respectively. Their
exponential utility function exhibits constant absolute risk

aversion :

‘ARA =-———=-n (2)

and decreasing relative risk aversion :

RRA = «(cf (3)
Each agent maximizes his expected utility across all future

periods where exponential utility is given by

Eli B‘Uth)1=Et-(l)i B‘expt-ncm (4)
c-o I] t-O

and B is a subjective discount factor such that 0<B<1.
There is a single infinitely-divisible asset produced and
consumed in the economy. Production is generated

independently each period according to a Markov process:

43

tit-«(11.02) . (5)

In this economy, each period will be identical with respect
to all variables except output It. Since the Markov
production process Y is strong-form stationary, all
information necessary for summarizing the present and future
state of the economy is contained in the current output
realization Y;.IFor this reason, we’ll drop the t subscript
when it is not necessary for clarity and for the stream of
outputs from the stochastic production process y. The
production outputs y are represented by the distribution
function G(y) or, more simply, G.

Ownership of production y is determined in a
competitive stock market between agents f and d. For
convenience, assume that ownership in the productive process
is represented by a single infinitely-divisible equity

share. Let

P;:Domestic asset price for investor k

P}:Foreign asset price for investor k

X}: Agent k’s end-of-period number of domestic shares
x}: Agent k’s end-of-period number of foreign shares

2:: Agent k’s begirming-of—period number of domestic shares

44

Z}: Agent k’s beginning-of—period number of foreign shares

Agent f holds beginning-of-period proportions 2%, z: and
agent d holds beginning-of-period proportions 23, z: of this
equity share which entitle them to Y(zf, + 2;) and
Y(zg + z§)«output units, respectively. Shares are traded ex-
dividend in a competitive auction at the beginning of each
period. Let x}; and x; denote end-of period share holdings
after the next competitive auction. Allowing the asset price
to differ across agents k e (d, f), perhaps due to
restrictions on share holdings zgiand 2;, results in prices
P}; and P5.

In this economy, all shares are held by one of the

agents, so that
z:+z,,‘= 219w; =1. (6)

Each agent’s beginning-of-period resources include a claim
(2; + 2:) on production Y (a random.variable) as well as
beginning-of—period equity value [23% + 259:].

These resources are either consumed (C*)<or used to purchase
end-of-period equity in an amount [x395 + xtPfi].

Consequently, each agent faces a constraint

C"+P,§‘X,§‘+P}‘X}‘s¥(z;+z:) +P§ZJ+P§Z§ (7)

Now, the equilibrium condition is

45

xpd+xf=xg+x§=1. (8)

Suppose there is a binding ’delta constraint’ (6) on

domestic ownership of the foreign asset,

Xf=b and Xi=1-b. (9)

and a binding ’gamma constraint’ (1) on foreign ownership of

the domestic asset.
XDf=y and Xg=1-y. (10)

A stationarity condition is imposed on asset ownership to

simplify the problem:
I:,§‘=z,§c and X}=Z}, (11)

k E (d'f) e
Let y’ denote next period’s production represented by the

distribution function of G.

Now we define the indirect utility function for agent k.
k k _ _ 1 __ k I k k I
U(szbozr)"max[ (filexm 1'IC' )+BIU(YIXDIxr)dG(YH

=max [- (fi) exp(—n (y(z:+z:) +956!) (25—26) m: (y) (zf-xi) 1)

46
+BE{U(Y’. 2:15.295) }] (12)

where n(eta) is the absolute risk aversion for both agents.
Following Lucas [1974], Appendix II-A derives the closed-
form.solution for the volatilities of P: Ck e d, f) assuming
exponential utility as follows:

For kIf,this means that the variance of net asset value is

given by
Var(Ppf =(ng)2(u-noz(xp‘+xif))2x

exp(2n’az(xp"+x})’) [expmza2 (xlfmifi) ~11 . (13)

For kId,this means that the variance of the foreign fund

share price is given by
Var(Pif') = (3215) 2 (ts-n02 (xpd'rxif') )2x

exp (211202 (x§+x,?) 2) [exp ('qzo2 (de+x,§’) 2) -1] . (14)

4. Implications and Analysis

When the delta constraint on domestic ownership of the
foreign asset is binding, domestic investors’ unconstrained

equilibrium quantity demanded of the foreign asset is

47
greater than the constraint delta. Similarly, there is a
gamma constraint on foreign ownership of the domestic asset.
Even though the domestic market is open for foreign
investors, if the foreign government restricts capital
outflow, there is a gamma constraint. Using the binding
conditions in (9) and (10) would yield prices

pfl= B(p-no’(1+(b-y)) )eXfln(1+(6ér)) {Y‘IJ}+']202(1+(5‘Y))2/2)
1..

and (15)

2,543 (is-1102(1-(6-y) ) )exgm (1-(6ivn {y-tt} +(Wan-(6*!) )3/2)
1..

(16)
The closed-end country fund share price premium over net

asset value is then

Premium P3 - P; I

1% I (u-na’(1+T) )expm (1+T) {y-p} +n’a’(1+T)’/2)

-(p-n02(1-T))exp(n(1-T) {y-p}+n202(1-T)’/2)] (17)

where T is the difference between 6 and 1. The domestic

investor’s ownership claim on the foreign asset is (1+T)

48
while that of the foreign investor is (l-T). When the

ownership restrictions are binding,

Var(pi) =(Tg—)3(|1-t]02(1-T) )2x

B
exp(2n202(1-1)2)[expin202(1-I)2)-1] (18)

and

Var(Pé’) = (Tea): (p-naz (1+:r) )2x

exp(2n203(1+1)’)[exn(n303(1+I)’)-1l (19)

When T I 6 - y > 0, under smooth consumption domestic
investors have more y to invest in the foreign asset than
foreign investors do.

The higher the y, the higher the asset price. Under an
assumption of smooth consumption, as y increases the portion
of y to invest in the foreign asset increases, which
increases asset price P‘,l more than P3. If y is small, both
investors may not be able to invest in the foreign asset.
Therefore, the price will not be much affected. This implies
that the variance of P€.is greater than the variance of 2;.
It is difficult to show analytically the relative size of

var(P§) over var(P§). However, in some special cases, we

49
can prove it. For instance, if or2 is very small, then 0‘

approaches 0. In that case, the variance ratio reduces to

Var(Pf) = (1+T)2

. 20
var(2{) (1-T)2 ( )

 

This is greater than 1 as long as T is positive.

The limitation of this model lies in the assumption of
a single asset. Lucas’s model assumes a single asset which
is shared by the domestic and foreign investor. Therefore,
there is no diversification possibility in this economy. In
EJ’s model, investors pay a premium for diversification. In
Lucas’s model, endowment y determines the level of the
foreign asset premium. The higher the claim on y, the
higher the asset price and hence the premium. Single-asset
models such as this are unable to capture the
diversification effect such as in Eun and Janakiramanan

[1986].

50
Figure II-l Variance ratio (variance of Korea fund price

over variance of Korea fund NAV)

This figure shows the actual annual variance ratio {var(fund
share price)/var(net asset value)} changes based on monthly
observations. Variance ratio is always far greater than 1,
which means that the variance of fund share is greater than
the variance of net asset value. In contrast to what the
theoretical model predicts, the variance ratio does not
show monotonic increases as investment restrictions are
loosened. Therefore, other factors should also be

considered.

 

 

 

 

 

 

51

APPENDIX IIIA Expgnential Utilit!

This appendix derives equations (13) and (14) using a negative
exponential utility function. Define the indirect utility

function for agent k as:

U(y. 25. 2}) =max[- (31;) exm-nc“) +BfU(y’.xz§‘.x}‘) dG(y’)]

=max[-(-%)exP(-II{Y(Z;+Z;) +P5(Z§'x6)+1’i(zi'x:)})

+(32w(y'.x:.x:) 11 (A-l)

where n is the absolute risk aversion for both agents and.E is
an expectation operator. Imposing the stationary condition
that all shares are held by investors throughout the period
(2,," I xp" and z,k I x,") results in the indirect utility

function5

U(y.xlf.x}‘) =- (fi) exp(-ny(x)§+x}‘) ) +h(x1§‘.x}‘) (A-2)

Substituting (A-2) into (A-l) yields,
WY. 29“. 2})

=max [- (l) exp (-qy{z,§+z:} +P,f(z,f-x§) +P}5(z}-x}) })

+9 { - (34-) exp ( «m (xg‘mfi‘) +1120: (xfixfi) 2/2) +12 (x:.x}‘) )1 (2)—3)

 

5The existence and uniqueness of the indirect utility
function have been proved by Lucas [1978].

52

Using the stationary condition 20" I x082," I x,", let

Bur}. Xi) =- (fi) exp (mp (x;+x}‘) +11202 (xgfld‘) 2/2) .

A(x1§‘.x}‘) =B(x1§‘.x}‘) +h(x0"+x}) .

Then

U(y. x15“. Xi) =- (7%)) exp(-ny(x1f+x}) ) +BA(X$‘. Xi)

=- (3‘17) exp(-qy(xg‘+x}‘) ) +h(x1§.x}‘) .
Then from the above relationship,
B [B(xzf.x;) +h(xzf.x:)] =h(x)§.x:-‘) .

Solving for h(x}‘,, xii) yields

_ BB(x;.x:)

h(kalx:) ——TB—_'

Therefore,

B(x" x")
AU‘DkI X:) = 10"B P

 

= -exp(-nu(x1§‘+x}‘) +11202 (Xi'an?) 2/ 2)

I1(1-I5)

 

(ii-4)

(A-S)

(A'G)

(Ii-7)

(ii-8)

(A-9)

The first order condition for (A-3) with respect to x’,‘ is

53

-P}‘exp(-ny(x,§+x}‘) ) +8A’(x,§,x:) =0 (A-lO)

where

- l-nwnza’ Mimi) 1 en) ( mt (2:52:15) “1202 (xzf'rxi) 2/ 2)

(HI-ll)

A’(kaIka) =

 

(A-ll)

Plugging equation (A-ll) into (A-10) and rearranging yields

Pfexr) (-ny(xx§+x}‘) )

 

(nu-aza’ (duh )eXD ( mu (xzfmf) +1120“ (aim?) 2/2)

=3 n(1-B)

(A-12)

Therefore, the equilibrium price of the foreign asset to agent

kis

p“- B (Is-n02 (2:159:95) ) exp (1) (2:sz95) {Y‘Il} +1120? (xfi‘mi) 2/2)
F" 1"6
=C(x5.x}‘) exp [1) (X§+x}‘)y] - (A-13)

This implies that the asset price for agent k is a function of

xD" , x," and y, where

54

(3(u-na’(x1f+x}‘) )eXP ( -n (xé‘mi‘) nmzoz (xzfmf) 2/ 2)

 

C(x:.x:)= H,
(A-l4)
Since C(x§,x',‘) is a constant term,
var(P*(y))=(C(x:,xfi))zxvar(exp[n(x§+x§)y1). (A-15)
The stoChastic component of equation (A-15) is
var(exp[n(x§tx;)y])
=E(exP [2n (x:+x}‘)y] ) - {E(exr> [1) (ka+x1’k)Y] ) I2 (It-16)

Now, the second term of equation (A-16) is

E'(exp [1) (xg+x})y] ) = 1 [exp (I) (Icing?) y) exp(—'—[-t§]—2) dy
J2n 20

 

(A-17)
We can expand the term in the integral of equation (A-17)

as follows:

exp (1) (xé‘flri) y) exr> ( 1%)?)

=exP [-3167 {ya-2 (n+ozn (xlfflrif) )Yfllzil

0211’ (2:33:95) 2
2

 

=exp II") (x1529?) + leXD tuft; (y- (“+0211 (like?) ) )2]

(A-18)

55

Therefore,

Elexp (n (XJ+X}‘)y)] =

k
021]: (295x, ) 3

2 1"

fem-3%; [y- (Nazi) Mimi) ] 2dy-

 

eXP [M (xé‘mi) +
(250
(A-19)

Now, y~N(u+ozn (xgncf) , 02) .

since

 

 

1 epr- 1 (y-(p+02n(xg+x})))2] (It-20)

f(y) = Ma 202

and

[£00 dy=1. (A—21)

then plugging (A-20) into (A-21) yields

 

fexp [' 2:2 0" (Wozn (XJ+xp'-‘) ) )2] dye/HO (A-22)

56

k k
021) (xD +x,)

2 )1. (A—23)

 

.-., E(exp [1| (x§+x}-‘)y] ) =exp [1| (X19105) (IN

By using the same method we obtain,

E(exp [2n (XA‘+X}‘)y] ) =exp [211 (Xbxif) (wozn (x;+x)5) )] (It-24)

:., Var(exp [n (X;+X;).Y] ) =

0’1) (Kim?)

2 )1

exp [21) (Kim?) (swan (XE’rxil‘) ) ] -exp[2n(x1§+x}‘) (w

 

(A-25)

As Var(Pf) =C(x§,x}‘) 2Var(eXp [1| (X;+X:)Y] ) I

Var ( P15)

 

=[ l3 (II-1102 (Kiwi) ) exp ( -'(| (xzffld‘) umza’ (gag!) 2/2) 12

1-B

x exp (2n (xfixfi) (n+1)2 (x;+x:) 2«22) (exp (n2 (x:+x:) W) -1)

= (395V (is-n02 (xg‘mfi‘) )zx

57

exp (211202 (xg’rxfi‘) 1') [exp (1)202 (x§+x}‘) 2) —1] (A-26)

These simplify as equations (13) and (14):

Var(pr) = (1%): (ti-no2 (foer'f) )2x

exp(21|202 (fo-I-xffl) [exp(1|202 (x,f+x;)2) -1] . (13)

Var(Pfi) = (3216)2 (p—na= (sea ,2x

exp(2112(12 (x,,d+xfi')z) [exp(11202(x,§'+x:)2) -1] . (14)

58
APPENDIX II-B Log Utility

Michener [1982] derived a closed-form solution of an asset
price in a pure exchange economy based on a log utility

function. Assume a Lucas’ [1974] economy and utility function

EB‘logCt where 0<B<1 (B—l)
t-O

where B : discount rate

Ct: consumption at time t.
As in Lucas, there is only one asset in this world. Michener
uses an example of shares in a fruit tree which produces its

nonstorable consumption good y, following the Markov process:

lny’=alny+¢-:c where 0sa<1, et“N(0,oz) . (B-2)

As there is no storage or investment, c=y*z in equilibrium and

the budget constraint is

wealtthz+P(y)z=c+P(y)x (B-3)

where z is beginning-of-the period number of shares and x is
end-of-period number of shares.

Let the value function be

WY. 2) If; [U(c) +B V(y’.X) dF(y’|y)] - (B-4)

59
Conjecturing
P(y)=a1y and V(y,z) =ké +k1’log (Wealth) +kz’logy (B-S)

yields

V(y. 2) =ké +k1’log [yz+P(y) z] +k2’1ogy

=k,J +kllogz+k21 ogy. (B-6)
Provided that V(y,z) has the conjectured form,

k0 +kllogz+k21 ogy=max {log [yz+a1y ( z-x) ]
X
+8 £ (k0 +kllogx+kzlogy’) dF( y’ | y) }

=max{log [yz+a1y(z-x) ] +03 (1% +k11093+k2109Y’) }
X

=m3xtlog [yz+a1y(z-x) ]
+9 (ko+kllogx+k2alogy) } . (B-7)

The first order condition with respect to x is

av_ -a1y + Bk.

6?fyz+a1y(z-x) x =0 (8-8)

 

60

Under the stationary condition (z=x), this gives the

relationship
-a1+8k5=0 (B—9)
and
k0+kllogx+kzlogy=logy+logx+fl (k0 +kllogx+k2alogy) . (B-lO)
From this, we can get
Pm airway???» (B-n)

since

= = 1 = 1 —
k0 0,k1 fi,k2 m- (B 12)
Note that x is not a factor affecting P(y). Therefore, under

the log utility function, a delta constraint can not be

imposed in this economy.

61

APPENDIX II-C Power Utility

In this appendix, I derive a closed-form price solution for
asset y using a power utility function in Lucas’[1974]
economy. If x (end-of-period number of. shares) is in the
price function we can impose an ownership restriction since,
when the ownership restriction is binding, xI6. If we don’t
have x in the price function, we can not impose a delta
constraint on the end-of-period number of shares and the power
utility function will not help us impose an ownership

restriction in this economy.

Let,Y I e’ where y is the stochastic endowment.

“Hilda:

lnY~N(u,02), E’(Y‘)=Ee‘y=e 2 =A. (C—l)
People maximize the following utility function :
U(Y.z*) =maxt-(1/r) [(Yz*+P*(Y) (zk-x*)]‘+BE[U(Y’.x*)]} (C-z)

Subject to

C+P“(Y)x“<Yz"+P"(Y)z". (C-3)

Conjecture
UIYIZ")=-(1/r) (YZ")‘+Bf(Z")- (C-4)

Substituting (C-4) into (C-2) yields,

62

-(1/r)(Y2*)‘+Bf(Z*)

=max{-(1/r)[(Y2*+P*(Y)(Zk-X*)]‘+flE[-(l/r)(YX*)‘+BfLX*)]}

=max{-(1/r) [(YZ“+P"(Y) (zk-xk)]’+(3E[-(1/r)A(xk)‘+flf(xk)lI-
By using x" I z"(stationary condition), we obtain

B (-1/r)A(z") r+132f(z") =Bf(z*) . (c-6)

Rearranging (C-6) yields

f(z*)=fi(-1/r)4(z*)r. (c—v)
Now the utility function (C-5) can be rewritten as

U(Y. z“) =max{-(1/r) (Y2“+P*(Y) (zk-X“) 1‘

+T%(-1/r)4(zk)r1. (c-e)

Therefore, from (C-l),

1::
+—
ru 2rd

U(Y.z*) =-(1/r) (Yz*+P*m (xx-meg); t—(1/r) (e )(z*)

(C‘9)

Taking the derivative with respect to x" yields

63

122

max=Pk(Y) [(sz+Pk(Y) (Zk'xk)’] t-1+_fi£6 {-(l/I) (6111* 21' O ) (xk) r-

 

6xk

(c-10)

Solving for P*(Y) yields

+311.202
MY) 4396) um (e"‘ 2 1y”. (c-n)

Therefore, imposition of a delta restriction is not possible

since the x variable does not appear in (C-ll).

64

ESSAY’III..An Empirical Investigation into the Deteggination
of ClosedIEnd country FUnd.Premiums[Discounts and

volatilities

Introduction.

The key assumption of Eun and Janakiramanan’s[1986]
international asset pricing model with a constraint on
foreign investment is that investors have homogeneous
expectations regarding asset returns and variances. The
ownership restriction then affects foreign and domestic
prices through a diversification effect. In contrast, Lee et
al’s [1990] investor sentiment hypothesis is based on an
assumption of heterogeneous expectations. This essay is an
empirical investigation into which of these effects is
predominant in Closed-End Country Fund (CECF) premiums,

discounts and premium/discount volatilities.

1. CECF premiums/discounts and ownership restrictions

1.1 Long-run evidence : The case of the Korea Fund

There was a time period during which the only available

investment tool in the Korean stock market for non-Korean

investors was the Korea fund (NYSE symbol ’KF’). During this

time, the Korea Fund represented a proxy for Eun and

65
Janakiramanan’s foreign equity ownership constraint (6) in
the Korean market. This section documents and interprets the
Korea Fund’s premium over net asset value which has existed

since its inception in 1984.
The Korea Fund provided the following information :

a. The daily per share market price of the Korea fund and

its underlying net asset value between 1984-1991.

b. Korea fund shares outstanding during the period on a

daily basis.

c. Korean stock market index during the period on a daily

basis.
From this data, the following two indices were constructed :

a. The total Korean stock market value on a daily basis.
This index is based on the market value as of 1/1/80 when
we put the market value at that time as 100

(2,609,400,000,000 Korean won).

b. The ratio of the NAV of the Korea Fund over the total
Korean stock market value. This ratio is used as a proxy

for the proportion of non-Korean ownership of Korean

66
stocks allowed by the Korean government. This was an
appropriate proxy during the period when the Korea fund
was the only available investment instrument. The ratio
increased throughout the 1980’s as the Korean government
allowed progressively more non-Korean investment in
Korean equities. This ratio is our proxy for Eun and

Janakiramanan’s (EJ) delta constraint 6.

Figure III-1A plots the Korea Fund premium-to-NAV alongside
the ratio of Korea Fund NAV to the Korean Stock Exchange
market value. According to EJ, premiums should be an
inverse function of delta. As we can see from figure III-1,
movement in the Korea Fund premium relative to changes in
the delta restriction shows that EJ’s model is not
sufficient to explain the observed behavior of CECF
premiums. The premium rose and then fell during these
period even though the delta constraint on foreign ownership
of Korean stocks was increasingly less restrictive
throughout this period.

The following regression was run using daily
observations over the period (1984.9-1987.3) when the Korea
fund was the only fund available for non-Korean investors.
(i.e. before other Korean funds emerged, such as the Korea
Europe fund, Korea Asia Fund, Korea Investment Fund, and so

forth). (See Table III-1)

67
PREMIUM I A + B*A6 + C*NASDAQ
where A6 (Change in the ratio of the NAV of the Korea Fund
over the total Korean stock market value) is a proxy for the
change in foreign ownership constraint and NASDAQ (NASDAQ
index return) is a proxy for investor sentiment.
Ho : B I 0
H‘ : B < 0
The regression results in Table III-1 show that coefficient
of 6 is significantly positive.‘ 'This is inconsistent with
EJ’s model. Increase in delta should have resulted in a
decrease in the premium according to their model. However,

the opposite happened.

1.2 Short-run evidence

Bonser-Neal et al [1990] (BBNW) find that changes in
fund premiums are related to announcements of changes in
investment restrictions. In particular, they test whether
announcements of liberalization of foreign or domestic
restrictions are associated with declines in premiums and
announcements of tightening of restrictions with rises in
the premiums. Ceterus paribus, constraining the foreign
ownership restriction will raise the price/net asset value
ratio by approximately the amount the marginal domestic

investor is going to pay to avoid the restriction. BBNW’s

 

‘Figure III-1B illustrates that residuals are
relatively randomly distributed.

68
conclusion is that government regulation can be effective in

segmenting markets.

BBNW use the following dummy variable regression model:

where

rtit : change in the jth fund’s premium in week t.

In” I l(-1) if t is between two and seven weeks before the
announcement of a loosening (tightening) of the jth

country’s investment restrictions.

[5“ I l(-1) if t is between one week before and one week
after the announcement of a loosening (tightening) of the

jth country’s investment restrictions.

[5” I 1(-l) if t is between two and seven weeks after the
announcement of a loosening (tightening) of the jth

country’s investment restrictions.

6“,: The average weekly effect before the announcement of
regulatory changes on the jth fund’s premiums/discounts.
62,: The average weekly change in the premium or discount
during the three-week period surrounding the announcement.
5x): The average weekly effect after the announcement of

regulatory changes on the jth fund’s premiums/discounts.

69
The regression coefficients measure the average weekly
effects of announced changes in investment restrictions on
closed-end country fund premiums/discounts.

After adjusting for dividends and missing data, BBNW
ran the dummy variable regression above. Their results show
that 61j and 6t,are significantly negative at the 1% level
when an event-weighted test was performeda7 The negative
signs on these coefficients mean that, on average, the
premium (discount) decreased (increased) over the 5-week
period before the announcements and over the 3-week period
around the announcements. However, among the five CECFs
tested, Japan Fund and Taiwan Fund yielded insignificant
results.

Following Bergstr6m et al [working paper], we
differentiate between capital inflow and outflow
restrictions. By adding a capital outflow constraint to
EJ’s model, Bergstrfim et. al. derive closed-form solutions
for P: and PL An outflow constraint and an inflow or 6
constraint operate in opposite directions. With a capital
outflow constraint (such as Bergstr6m’s switch currency
constraint), the discount demanded by the domestic investors
for holding domestic assets is lower, since foreign assets
are not so attractive with a binding capital outflow

constraint. Therefore, unlike the BBNW model, dummy

 

7Figure III-2 shows that residuals for BBNW regression
are randomly distributed.

70

variables are -1 if the announcement involves tightening of
an inflow restriction or loosening of an outflow
restriction, and 1 if loosening of an inflow restriction or
tightening of an outflow restriction. Results are reported
in Table III-2. No coefficients are significant at a 5%
significance level.8 The results imply that the
announcement effect of a change in an investment ownership
restriction cannot by itself explain the changes in CECF

premiums surrounding these events.

1.3. Critique of BBNW study

There are several points that concern us regarding the

Bonser-Neal et a1 [1990] study. These include the following.

a. Theoretical basis:

Following EJ’s theoretical framework, Bergstrfim et al
[working paper] show that with in-and-outflow constraints
foreign investors pay a premium for domestic assets and
domestic investors pay a premium for foreign assets. They
used restricted shares which are available for foreign(i.e.

home country) investors only, and unrestricted shares which

 

“We also test with the events related to changes in
capital inflow restrictions only. With BBNW data, only 5n
is significantly negative at 5% significance level. With
updated data, no coefficients are significant at 5%
significance level.

71
are available for both foreign and domestic investors. BBNW
did not differentiate between inflow and outflow
constraints. They treated them the same and only looked at

whether the restrictions were tightened or loosened.

b. Number of events:

BBNW used only 23 events. Only two of these are
announcements of tightening in an investment restriction. In
the empirical analysis which follows, we were able to add 8
more events. Hopefully, this will lead to a more robust
test of the relation between changes in CECF premiums and

changes in investment restrictions.

1.4 Controlling for investor sentiment

In EJ’s model, the price of fund shares and NAVs are a
function of the absolute risk aversion of the foreign and

the domestic agents and the delta constraint.

Pf-P;=f(AD,A’, a)

Lee’s study suggests that differential sentiment of U.S.
investors relative to their foreign counterparts affects the
closed-end country fund premiums. While EJ’s model restricts

the ’sentiment’ to risk aversion, LST use the sentiment to

72
indicate either higher/lower expected returns or
lower/higher risk aversion between domestic and foreign
investors.

Assuming that A6 (announcement of changes in investment
restrictions) is linearly related to the Apremium, changes
in investor sentiment can be ’backed out’ with ZSLS:

l. APremium I a + bAS + e

2. e I a + BA6 + u

where AS is a proxy for investor sentiment changes. Just as
Lee et. al. used changes in value weighted discounts on
closed-end funds (CEFs) as a proxy for investor sentiment
changes, we can form and use a CECF premium index as a proxy
for investor sentiment regarding foreign investment. This is
equivalent to controlling the relation between the Apremium
and A6 for changes in investor sentiment with a multiple

regression.

The following hypothesis is tested here.

1% : Change in 6 is independent of change in the CECF
premium.

HA : Loosening 6 decreases the premium.and tightening 6

increases the premium.

As a proxy for change in investor sentiment, AS, across all

73
CECFs, we construct an index of all CECF premium changes and
use the difference e I Apremium (world-adjusted) I Apremium
in jth fund - AAverage premium across all other funds in the
second stage regression. This assumes aIO and bIl in the

first stage regression.

Apremium (world-adjusted) I Apremium in jth fund -

I 1 i4 ' 1
mm Pram“

The relation between Apremium and A6 (announcement of
changes in investment restrictions) after controlling for

changes across all CECF’s is then
Apremium I a-6A6.

where Apremium is the change in the jth CECF premium.minus
the average premium change in other funds over the period.
This regression equation controls for U.S. investor
sentiment regarding foreign investment and hence looks at
the pure announcement effect of an investment restriction
change. BBNW’s study is based on 23 events. When we looked
up the announcement dates in the WSJ index, we found eight
additional announcement dates which are clustered for
several particular funds. Considering the fact that

differential risk aversion and investor sentiment can affect

74
premiums/discounts, we think an update of the data may yield
more conclusive results than BBNW’s study. We ran the test
with their original data and with our updated data. If this
test yields an insignificant result, the announcement effect
of investment restriction changes may not be a significant
factor in CECF premium/discount changes. Results are
reported in Table III-3. After controlling for investor
sentiment, the announcements do not seem to affect CECF
premium changes.

According to LST, stocks and Closed-End Funds (CEFs)
with low institutional ownership are more sensitive to
changes in investor sentiment than CEFs with high
institutional ownership. In other words, CEFs and small
stock prices in which institutional ownership is usually low
tend to move together with changes in the sentiments of
individual investors. If the ownership structure of CECFs
resembles that of CEFs and small stocks, changes in U.S.
investor sentiment regarding foreign investment should
affect CECF prices. Foreign investor sentiment change should
directly affect only the NAV. Therefore, the ACECF premium
reflects change in differential investor sentiment. One item
to note is that the underlying assets of CECFs are usually
the largest stocks in foreign countries with high
institutional ownership, while the fund shares are largely

owned by small U.S. investors.

75

Two hypotheses are tested. First,

Ho : U.S. small stock market returns are independent of
ACECF premiums

ACECF premiums widen (narrow) when ULS. small

In
3‘
as

stock market returns are high (low) and vice versa.

NASDAQ index returns were regressed on the CECF premium
change index (WPINX). Table III-4 reports that they are
significantly positively correlated at 1% level. This
tells us that small investor sentiment, which is represented
by the NASDAQ index return, is an important factor affecting
CECF premiums/discounts. NASDAQ returns mostly affect CECF
share prices, not the NAV. In other words, domestic
investor sentiment affects CECF premiums/discounts by

influencing the Fund share prices.

The following hypothesis can also be tested :

Ho : Foreign market returns are independent of CECF premiums

:1:
b

CECF premiums narrow (widen) when foreign market

returns are high (low) and vice versa.

In Table III-5, the Korean stock market index return at time

t-l is regressed on the change in the Korea Fund premium at

76
time t. They are strongly positively correlated. This
means that the premium is larger when the foreign market is
performing well. We view this as consistent with the
investor sentiment hypothesis. Small investors (major
shareholders of CECFs) seem to respond very sensitively to
country-specific news. In other words, they overreact and
thus increase premium volatilities. Due to the limitation
of our dataP,‘we confine the test only to the Korea Fund.
However, tests of other funds are also done by using NAV
return as the proxy for foreign market return. Net assets
are usually the blue chip stocks in the foreign markets. If
the investor sentiment hypothesis holds, the foreign market
return should have explanatory power for changes in CECF
premiums/discounts. The results (Table III-6) show that
some funds’ changes in premiums at week t are positively
affected by foreign market returns at week t-l. Changes in
premiums at week t are negatively related to NAV returns at
week t. Therefore, investors seem to respond to some
foreign market returns, but U.S. market returns seem to be a
more significant factor affecting CECF premiums than the

foreign market return.

2. volatility ratio : CECFs selling at a preaium.have higher

VRs(variance ratios) than those selling at a discount.

 

’We do not have other foreign stock market returns
data.

77

There is a difference between the variance of CECF
prices and the variance of their NAVs. In Essay I, we
derive the theoretical implication of the volatility ratio
(Var(P§)fVar(P§)). The ratio is expected to decrease if
cov(ND, NF) is negative as ownership constraint is loosened.
In most cases, the ownership constraint is loosened as time
passes. Therefore, the variance ratios are expected to
decrease as time goes on. However, the variance ratios are
not monotonically decreasing as the theoretical model
predicts.

According to our test results, CECFs selling at
premiums usually have higher variances than those selling at
discounts (see Table III-7 and III-8). We view this as
consistent with Lee’s [1990] investor sentiment hypothesis.
If institutional investors are rational and small investors
are irrational noise traders, the lower the institutional
ownership, the higher the noise trader effect on CECF share
price volatility. Discounts of CECFs are eliminated as
CECFs are open-ended as were Japan fund and France fund.10
This makes funds selling at discounts attractive to
institutional investors since we can expect open-ending at
some time in the future. Also, the fact that discounts

narrow in a bullish market and widen in a bearish market

 

1°Brickley and Schallheim [1985] find that CEF
discounts are real since when they go open-ending share
prices increase sharply converge to NAV.

78
offers rational investors a winning strategy“.

Premiums on CECFs can be explained by the investor
sentiment hypothesis as follows. Unlike domestic CEFs, the
underlying assets of CECFs can not be easily duplicated
especially for funds with investment restrictions. And
individual investors usually can not invest directly in
foreign countries mainly due to lack of information.
Therefore, CECFs are favored and these could be possible
explanations for the existence of CECF premiums. However,
institutional investors who can buy the same underlying
assets at a lower cost than the CECF share price, do not
have to pay a premium for CECFs. In that sense, the
institutional ownership of CECFs selling at a premium should
be lower than for those funds selling at a discount.

If this is true, assuming that volatility of NAV
reflects fundamental volatility, we can conjecture that the
variance ratio (Var(P:)/Var(P§)) of funds selling at a
premium should be higher than the variance ratio of funds
selling at a discount. To support this hypothesis, we
gathered annual institutional ownership data on CECFs from
the Spectrum database. Average annual institutional
ownership data for premium funds and discount funds are
described in Appendix III-B. Consistent with our

expectation, premium funds seem to have lower institutional

 

11Richards et al [1980] used a filter rule to generate
excess returns for CEFs across bull and bear markets.

79
ownership than discount funds.12
One important finding of this research is that CECFs
selling at a premium tend to have higher variance ratios.
This finding was statistically significant at 1% based on
both a Pearson correlation test and a Logit analysis. (Table

III-8 and III-9)

3. The relation between premiums/discounts and
premium/discount volatilities

Figure III-5 shows that premium funds’ volatility ratio
is greater than discount funds’ volatility ratio. This
means that investors react more sensitively on premium funds
than on discount funds. Tables III-8 and III-9 show that
funds with higher premiums usually have higher share
price/NAV volatilities as well.” Share price/NAV
volatilities are especially high when premiums are at their
highest, but the relation holds at lower premiums and

discounts as well.“ Investor sentiment is presumably

 

12Regression results (Table III-9) show that there is a
negative correlation between CECF premium and institutional
ownership.

13The relation between fund premiums and the variance
of the premium (rather than the share price-to-net asset
value ratio) is also positive and significant using either
regression or logit analysis

1‘The relationship between Var(CECF share price
return)/Var(NAV return) and premiums/discounts is also
tested. However, it is not significant at 5% significance
level.

80
responsible for both the premiums (and discounts) and the

share price/NAV volatilities.

81

Summary and Conclusion

These essays investigate factors affecting CECF
premiums/discounts and CECF share price/NAV volatilities.
CECFs are international assets. Therefore, we studied
international asset pricing models with investment barriers
which may explain the difference between CECF share price
and NAV. Among others, EJ’s investment ownership
constraints seem to affect premiums paid over foreign
asset’s NAV. Essay I extends EJ’s CAPM-based model to
extract the theoretical implication on foreign asset’s
premiume/discounts and on the volatility ratio. Essay II,
based on Lucas’s general asset pricing framework, derives
closed-form solutions for foreign asset prices and variances
when there are ownership constraints. BBNW adopt EJ’s
theoretical implications in their study of CECF
premiums/discounts. However, it is difficult for us to
believe that announcements of investment restriction changes
alone can explain the CECF premium/discount movement because
there are many other factors affecting CECF
premiums/discounts such as U.S. investor sentiment or
foreign market returns, etc. Essay III, empirically
investigates these factors affecting CECF premiums/discounts
and premium/discount volatilities.

BBNW’s study has at least two shortcomings. First,

they did not differentiate between capital inflow and

82
outflow constraints. Second, they used only 23 events.
This study shows that the investment ownership constraint is
not sufficient to explain changes in CECF premiums/discounts
after differentiating capital inflow and outflow
restrictions and updating the sample. Results are similar
after controlling for the world CECF premium index. There
are two other problems associated with using CECFs in a
BBNW-type study. First, premiums and discounts tend to move
with U.S. investor sentiment and with a foreign market
index. A strong correlation was found between a World CECF
premium index and the NASDAQ index. Almost all individual
fund premium changes are strongly positively correlated both
with each other and with the NASDAQ index. (Table III-10 and
III-ll) On the other hand, the Korea Fund’s premium is
affected by Korean stock market returns. This means that
investors are responding to country-specific news. Out of
the 14 funds tested, only 5 funds’ NAVs are related to CECF
premiums. This could be due to the use of weekly rather
than daily data. Also, NAV returns may not be a good proxy
for some foreign market returns. But if the proxy is
reasonable, foreign market returns may have less of an
influence on CECF premiums than U.S. investor sentiment.
For a more powerful test, correlations between foreign
market returns and CECF premiums/discounts should be
considered.

Second, Bailey and Lim [1992] showed that CECF returns

83
move with U.S. maghet returns, while foreign stock market
returns do not move with the U.S. market returns. This was
also documented in Lee, Kim and Bodurtha’s [working paper]
recent study. Lee et. a1. conclude that CECF
premiums/discounts are affected by U.S. investor sentiments.
For this reason, CECFs may not be the best choice for
diversification purposes. It would be better to test the
impact of ownership restrictions with the restricted foreign
stocks available for domestic investors in the foreign stock
market. For instance, Korea Mobile Telecom is now selling
at more than a 100% premium for non-Korean investors while
the Korea Fund is selling close to NAV. It may be fruitful
to use individual foreign stocks other than CECFs in tests
of factors affecting CECF premiums/discounts.

Finally, we find that the variance ratio (VAR of Share
price over VAR of NAV) of CECFs is usually greater than 1.
This means that CECF share prices are more volatile than
NAVs. Funds selling at large premiums tend to have much
higher variance ratios than discount funds. I interpret
this as meaning that investors react more sensitively toward
premium funds. These results could be due to differential
investor sentiments on premium funds and on discount funds
since premium funds, on average, have a lower percentage of

institutional ownership.

‘/

 

84

Appendix III-A

Events and data update

France:

7/23/86 - The French Parliament amends the government’s
denationalization bill by raising the limit on foreign
participation in state-owned companies at the time of sale
from 15% to 20% (IMF, 1986; and WSJ, 8/1/86, p. 21): Direct

investment, loosening, capital inflows.

9/30/86 - The Government plans to announce a move that will
allow residents to hold foreign currency bank accounts (WSJ,
10/1/86, p. 37): Foreign exchange accounts, loosening,

capital outflows.

Japan:

8/24/81 - Banking sources say that Japanese banking
authorities are planning to raise the ceiling on yen-
denominated loans outside Japan (WSJ, 8/25/81, p. 34):
Commercial banks’ international transactions, loosening,

capital outflows.

5/17/82 - Japan’s Finance Ministry relaxes rules covering

yen loans to foreign borrowers (WSJ, 5/18/82, p. 35):

85
Commercial banks’ international transactions, loosening,

capital outflows.

1/30/83 - Japan’s Finance Ministry announces that it plans
to lift the ban on domestic sales of foreign zero-coupon
bonds (WSJ, 1/31/83, p. 34): Portfolio investment,

loosening, capital outflows.

9/22/83 - The Japanese Finance Ministry says that it is
considering easing restrictions on investments in Japanese
real estate and stocks by foreigners (WSJ, 9/23/83, p. 34):

Direct and portfolio investment, loosening, capital inflows.

4/17/84 - Beryl Sprinkel, under secretary of the U.S.
Treasury for monetary affairs, announces an agreement with
Japan which gives U.S. and other foreign banks greater
access to Japanese financial markets (WSJ, 4/18/84, p. 35):
Commercial banks’ international transactions, loosening,

capital outflows and inflows.

3/28/85 - The Finance Ministry announces that it will
increase the number of foreign companies allowed to issue
yen-denominated bonds in Japan. In addition, the Ministry
announces that it will permit medium to long-term Euroyen
loans to foreign companies. Previously, such loans were

restricted to maturities of less than one year (WSJ,

86
3/29/85, p. 32): Portfolio investment, loosening, capital

outflows.

6/23/85 - A Ministry of Finance official announces that
Japan will open trust banking business and pension funds to
9 foreign banks. Foreigners had anticipated only 8 banks
(WSJ, 6/24/85, p. 22): Commercial banks’ international

transactions, loosening, capital outflows.

10/27/85 - At least 7 U.S. and other non-Japanese brokerage
firms have decided to apply for membership on the Tokyo
Stock Exchange. This implies that foreign members can make
their own trades and avoid paying 27% of their commissions
to Japanese members and may facilitate the sale of Japanese
stocks abroad (WSJ, 10/28/85, p. 27): Portfolio investment,

loosening, capital inflows.
South Korea

12/28/84 - The Ministry of Finance announces that foreign
investment will be permitted in 19 additional industries

(IMF, 1984): Direct investment, loosening, capital inflows.

6/6/85 - The Finance Ministry announces plans to allow
foreign investment in 230 more business areas by 1988 (WSJ,

6/7/85, p. 7): Direct investment, loosening, capital

87

inflows.

10/15/85 - The Finance Ministry announces that it will open
an additional 102 business areas to foreign investment (FT,
10/15/85, p. 5; and WSJ, 10/17/85, p. 32): Direct

investment, loosening, capital inflows.

12/2/87 - The South Korean government announces that it will
allow foreigners to exchange Korean convertible bonds for
stock (WSJ, 12/3/87, p. 48): Portfolio investment,

loosening, capital inflows.

3/28/88 - South Korean officials said that the government is
likely to allow Korean securities, insurance, and investment
trust companies to buy foreign stocks (WSJ, 3/29/88, p. 27):

Portfolio investment, loosening, capital outflows.
Mexico:

8/13/82 - The Government closes the foreign exchange markets
and imposes restrictions on money in foreign currency
accounts (WSJ, 8/17/82. p. 33): Foreign exchange accounts,

tightening, capital outflows.

6/19/83 - The Government announces that it is relaxing rules

governing foreign investment (WSJ, 6/20/83, p. 31): Direct

88

investment, loosening, capital inflows.

2/17/84 - The Government announces a list of priority
sectors where majority foreign ownership would be authorized

(IMF, 1984): Direct investment, loosening, capital inflows.

1/18/85 - The Government announces that it is rejecting
IBM’s application to build a 100% IBM-owned microcomputer
factory. The IBM project has been ”viewed by many foreign
investors as a test of Mexico’s stated policy of welcoming
more 100% foreign-owned projects” (WSJ, 1/21/85, p. 27):

Direct investment, tightening, capital inflows.

7/23/85 - The Government reverses its position and accepts a
proposal to allow IBM to build its factory (WSJ, 7/24/85, p.

10): Direct investment, loosening, capital inflows.

11/6/85 - The Government places new restrictions on Mexican
bank accounts held by foreign bank offices. Most ”bankers
interpreted the new measures as a means of slowing capital
flight" (WSJ, 11/7/85, p. 34): Commercial banks’

international transactions, tightening, capital outflows.

Taiwan:

6/14/87 - The Review committees to Taiwan’s Parliament

89
approve the lifting of controls on outward foreign exchange
movements (FT, 6/15/87, p. 5): Foreign exchange accounts,

loosening, capital outflows.

5/5/88 - A Government official announces that Taiwan’s
cabinet has approved a proposal to allow foreigners to
invest in local securities houses. The cabinet also
approved plans to increase foreign direct investment in
Taiwan (WSJ, 5/6/88, p. 16): Portfolio and direct

investment, loosening, capital inflows.

Events update

The limit on foreign equity holding in India will be raised
to 51% from 40%, as part of the month-old government’s push
to open up the nation’s economy. New Delhi also said it
would remove bureaucratic obstacles to industrial growth.

(S)Jl 25/91 - A, 5:4

Mexico is loosening its regulations on foreign investment,
opening up its economy to 100% foreign ownership of many
enterprises with assets of as much as $100 million; the new
regulations are a bid to attract foreign capital at a time

when bank credit is dwindling. 5/16/89-All:3

The Mexican government is promoting new investment

90
instruments that allow foreigners to buy into publicly
traded Mexican companies through a trust administered by the
state development bank. The procedure grants investors all
the privileges of shareholders except voting rights; table.

(L)Mr 26/90 - c, 12:1

South Korea is allowing foreign investment of up to 100% in

three manufacturing sectors and 99% in advertising agencies.

(S)Ja 2/90 - A, 8:2

Seoul plans by the end of 1991 to restrict to about $1
billion annually South Korean firms’ total sale of

securities on foreign markets. (S)s 25/91 - A, 6:1

Taiwan is about to open its door a little wider to global
financial markets by allowing the first local sales of

foreign securities. (M)D 17/90 - B, 6C:3

The Taiwan government approved regulations opening Taiwan’s
stock exchange to direct investment by foreigners, and fund
managers may be able to apply in Jan. 1991, (M)D 28/90 -A,
27:5

Taiwan’s SEC has proposed easing restrictions on foreign
capital in the stock market and other investments. The plan

would permit foreign institutions’ capital gains to be taken

91
out of the country after three months, instead of a year.

(5)119 28/91 - A,5:4

91
out of the country after three months, instead of a year.

(S)Ag 28/91 - A,5:4

92

Appendix III-B

Average annual premiums/discounts and institutional
ownership of CECFs (111189-12131191)

Premium is the annual average of weekly
ln(Fund share price/NAV) and the source of institutional
ownership data is Spectrum database. Beginning-of-the year
institutional ownership was used.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

: YEAR

‘ FUND 89 90 91

Brazil _premium -0.381 0.1228 0.1103
inst. 34% 27% 20%

. ownership

3 Germany __premium 0.4594 0.049 0.1775
inst. 3.6% 4% 1%
ownership

9 Swiss _premium 0.1475 -0.103 -0.055
inst. 7.3% 7% 4%
ownership

f India _premium -0.258 -0.05
inst. 32% 33%
ownership

Italy _premium 0.1114 -0.152 -0.217

inst. 13.3 16 13
ownership

f Korea _premium 0.6169 0.1096 0.226
inst. 11.7% 10% 14%
ownership

 

93

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

fl YEAR

FUND 89 90 91

Malaysia _premium 0.308 -0.l97 -0.162
inst. 17.2% 13% 9%
ownership

Mexico premium -0.151 -0.216 -O.127
inst. 12.8% 18% 28%
ownership

Spain _premium 0.8318 0.0313 0.0341
inst. 1.5% 1% 0%
ownership

Taiwan _premium 0.2252 0.2524 0.1147
inst. 2.2% 2% 1%
ownership

UK premium -0.082
r—
inst. 7%
ownership

Average institutional 8.1%

ownership of premium

funds

Average institutional 18%

ownership of discount

 

*As we can see from the lower part of the Appendix III-B,

the average institutional ownership of a premium fund is

 

 

lower than that of a discount fund.

 

 

94

Bibliggraphy

Adler, Michael and Bernard Dumas, 1983, International
Portfolio Choice and Corporation Finance : A Synthesis,
Journal of Finance 38, 923-984.

Bailey, Warren and Lim, Joseph "Evaluating the
Diversification Benefits of the New Country Funds." The
Journal of Portfolio Management (Spring 1992) : 74-80.

Boudreaux, K.J., 1973, ”Discounts and Premiums on Closed-end
Mutual Funds: A Study in Valuation", Journal of Finance 28,
515-522.

Bergman, Y. "Time Preference and Capital Asset Pricing
Models.” Journal of Financial Economics 14 (March 1985) :
145-59.

Bergstr6m, Rydgvist, Sellin, Asset Pricing with in-and
outflow constraints, working paper, Stockholm school of
Economics.

Black, Fisher,1974, International capital market equilibrium
with investment barriers, Journal of Financial Economics 1,
337-352.

Bonser-Neal, Catherine, Greggory Brauer, Robert Neal, Simon
Wheatley, 1990, International investment restrictions and
closed-end country fund prices, Journal of Finance 45, 523-
548.

Brauer, G.A., 1984, Open-ending closed-end funds, Journal of
Finance 41, 897-913.

Brauer, G.A., 1988, ”Closed-End Fund Shares’ Abnormal
Returns and the Information Content of Discounts and
Premiums," Journal of Finance, 43, 113-127.

Brickley, James A. and James S. Schallheim, 1985, ”Lifting
the Lid on Closed-end Investment Companies: A Case of
Abnormal Returns", Journal of Financial and Quantitative
Analysis 20, 107-118

Chen Nai-fu, Raymond Kan and Merton H. Miller, l993a,"Are
the discounts on closed-end funds a sentiment index?",
Journal of Finance, forthcoming.

Chopra, Navin, Charles M.C. Lee, Andrei Schleifer and
Richard Thaler”, Journal of Finance, forthcoming.

95

Delong,J.Bradford;Schleifer,Andrei;Summers,Lawrence H.;and
Waldmann,Robert J. ”Noise Trader Risk in Financial Markets."

Errunza, Vihang and Etienne Losq, 1985, International Asset
Pricing Under Mild Segmentation: Theory and Test, Journal of
Finance 40, 105-123.

Eun, C. and Janakiramanan, 1986, A model of international
asset pricing with a constraint on the foreign equity
ownership, Journal of Finance 41, 897-913.

Friend, Irwin and Marshall E. Blume, 1975, The Demand For
Risky Assets, The American Economic Review 65, 900-922.

Hansen, L., and Singleton, K. ”Stochastic consumption, Risk
Aversion and the Temporal Behavior of Asset Returns.”
Journal of Political Economy 91 (April 1983) :249-65.

HIETALA, 1989, Asset Pricing in Partially Segmented Markets:
Evidence from the Finnish Market, Journal of Finance 44,
697-718. -

Lee, Charles, Andrei Schleifer, and Richard Thaler, 1991,
Investor sentiment and the closed-end fund puzzle, Journal
of

Finance 46, 75-110.

Lee, C., Kim, D., Bodurtha, J., Closed-end Country Funds and
U.S. Market Sentiment, 1993, Univ. of Michigan working
paper.

Lucas, Jr., R. "Asset Prices in an Exchange Economy.”
Econometrica 46 (November 1978) : 1429-46.

Malkiel, Burton, 1977, The valuation of closed-end
investment company shares, Journal of Finance 32, 847-859.

Peavy, John W., 1990, ”Returns on initial public offerings
of closed-end funds”, Review of Financial Studies, 3, 695-
708

Pratt, E, 1966, Myths associated with closed-end investment
company discounts, Financial Analysts Journal, Jul-Aug., 79-

Richards, R., D. Fraser and J. Groth, 1979, Premiums,
Discounts, and the volatility of closed-end mutual funds,
Financial Review, Fall, 26-33.

96

Roenfeldt, Rodney L. and Donald L. Tuttle, 1973,”An
examnation of the Discounts and Premiums of Closed-end
Investment Companies", Journal of Business research, 1, 129-
140.

Stulz, Rene M., "On the Effects of Barriers to International
Investment" Journal of Finance 36 (September 1981) : 923-
933.

Thompson, Rex, 1978, The information content of discounts
and premiums on closed-end fund shares, Journal of Financial
Economics 6, 151-186.

Weisenberger,A., various years, Investment companies
(Warrren, Gorham 8 Lamont, New York)

Zweig, Martin E., 1973,”An Investor Expectations Stock Price
Predictive Model Using Closed-end Fund Premiums”, Journal of
Finance, 28, 67-87.

97
Figure III-1A

Korea Fund premium.(l984.9-1991.12)

The changes in Korea fund premium between 1984.9-1991.12 are
shown in this figure. The premium is measured as the
logarithm of the ratio of the fund’s price to its net asset
value. If premium is 1, that means KF is selling at 100%

 

 

premium.
Korea Fund Premium
1
03
3'?
50::
'g 05
933
a
02 ,
0d
0
-0.1
v-NNGDIDTNNvmv-LDQMDOFPP
sszaasseeeaesSSSSSS
vmmmmmunnwammmscpu:
CDCDGDCDCDQIDIDDIDGDCDQCDOQDOF
Deb:

 

 

 

98

Figure III-la

mmmmmmmmmmm
SEQ—WW

PREMIUM I A + 8*A6 + C'NASDAQ

 

Scatterplot
Dependent Variable: Change in KF premium

 

 

 

 

j I
'20 -I0 0 I0

—Dcn-—ee1) new-needy...“

newsman Md Padded Vdue

 

 

 

 

99

Figure III-2

Scattepplot for pesiduals ppd predicted valpes based op BBNW
regressiop

 

Scatterplot
Dependent Variable: Change in premium

6: A
o

6

 

"If I I
-3

-2 -l

O.
-‘s
N-
can)

-scn—sem new-nqeoaeam
O N L
omo—ooo
oo—
0
oooo
o orb—coo
—.
000.
0-0

Regression Standsdzed Preciotsd Value

 

.J

 

 

100
rignre III-3
Premium.change during the 3-day event pgriod
(25 events'with no missing observations * 3 days per event I

75 observations)

 

PREMIUM CHANGE

 

SIist-m
Meat-m
N-TSII)

 

 

 

 

 

 

*Normality test was rejected at 1% level as follows.

Statistic df Significance

 

x-s (Lilliefors) .2029 75 .0000

The distribution is skewed to the right and clustered around
Mean.

Mean .005 A Std Dev .082
Kurtosis 13.265 S.E. Kurt .548
Skewness 2.703 S.E. Skew .277

101
Figure III-4

Premium changes around loosening restrictions

(23 events * 3 days per event I 69 observations)

According to BBNW, with the announcement of loosening
restrictions, the distribution of premiums are supposed to
be skewed to the left (the mean is close to 0, -0.0034).
However, it is skewed to the right.

 

 

 

 

 

 

 

DEJA

 

 

*Normality test was rejected at 1% level as follows.

Statistic df Significance

 

K-S (Lilliefors) .1561 69 .0003

The distribution is skewed to the right and clustered around
Mean.

Mean -0.0034 Std Err .0077
Skewness 1.3364 8 E Skew .2887
Kurtosis 7.0582 8 E Kurt .5701

102

Figure 111-5

Prenigpzchapges around tightening restrictions

(2 events * 3 days per event I 6 observations)

This figure shows CECF premium changes around the
announcement of tightening investment restrictions. However,
Two events (six observations) would seem to be insufficient
for conclusive results.

 

CI'I

 

Std.Dev-.18
Meat-.10
N-6.00

 

   

DELTA

 

 

 

103

Figure III-6

CECF remi discounts nd variance ratios

* This chart shows the difference in the annual variance
ratio {variance (fund share price)/var(net asset value)}
between premium funds and discount funds. The reference
line indicates 0% premium. Therefore, points above the line
show variance ratios of premium funds and points below the
line show variance ratios of discount funds. Premium funds
have higher variance ratios and discount funds have lower
variance ratios.

 

 

 

 

 

 

 

Premiums/Discounts and Variance Ratio
p 8
R
E 9 o
M 6'. °°o
'.
3 AI. 0
I o o
D 2). ° °o oo 0
l ° 0% 8 o
S o. o O
C -0 o e
o u % o o
U o
N -.2-I- o
T .33?
-41. O
{6 2 r
-20 0 20 40 60 80 100
VARIANCE RATIO

 

 

104

Table III-1

Korea Fund’s premium and ownership constraints
(1984.9-1987.3)

Regression was run during the period when the Korea Fund was

the only investment instrument available for foreign

investors in the Korean stock market.

Change in KF premium I a + b*delta + c*NASDAQ +e

Delta: Changes in the proxy for ownership constraint

NASDAQ: Daily NASDAQ index return

Variable Coefficient SE(Coefficient) T
Delta 16.696691 2.125615 7.855
NASDAQ 1.002232 .318173 3.150

(Constant) 4.738013-04 .001378 .344

Sig T

.0000
.0017
.7311

105

Table III-2

Dgppy variable regressions after differentiating capital

inflow and outflow restrictions

Dummy variable regressions for each fund with the BBNW data
(original data) and with the updated data are presented

below.

where
j : Initial of each fund.

rt3t : Change in the jth fund’s premium in week t.

In“ I -1 if t is between two and seven weeks before the
announcement of tightening an inflow restriction or
loosening an outflow restriction, and 1 if loosening an
inflow restriction or tightening an outflow restriction of

the jth country.

Iyfi_= -1 if t is between one week before and one week after
the announcement of tightening an inflow restriction or
loosening an outflow restriction, and 1 if loosening an
inflow restriction or tightening an outflow restriction of

the jth country.

In” I -1 if t is between two and seven weeks after the

announcement of tightening an inflow restriction or

106
loosening an outflow restriction, and 1 if loosening an
inflow restriction or tightening an outflow restriction of

the jth country.

6t,: The average weekly effect before the announcement of
regulatory changes on the jth fund’s premiums/discounts.
5m): The average weekly change in the premium or discount
during the three-week period surrounding the announcement.
5t): The average weekly effect after the announcement of

regulatory changes on the jth fund’s premiums/discounts.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

e wi or' 1 dat
Name France Japan Korea Mexico Taiwan Overall
4 of 2 7 5 6 2 22
events
I of obs 132 292 208 314 80 1026
constant 7.93-4 6.68-4 .002 -.001 -.011 -2.0lE-4
Std. .004 .002 .005 .005 .012 .002
error
T-stat .2 .368 .48 -.216 -.97 -.104
6u .001 -3.71 -.13 -.026 -.003 -.01
Std. .022 .0052 .014 .019 .039 .007
error
T-stat .055 -.007 -.9 -1.377 -.079 -1.33
6,3 -.041 .002 -.015 .048 -.037 ‘ 5.42-4
Std. .021 .007 .017 .024 .051 .009
error
T-stat -.189 .268 -.88 1.97 -.73 .061
6’1 .009 -.003 .019 .005 -.038 8.248-4
Std. .0183 .005 .014 .022 .036 .007
error
T-stat .49 -.59 1.31 .235 -1.05 .12

 

 

 

 

 

 

 

 

108

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Re re s'o with ated

Name France Japan Korea Mexico Taiwan India Overal l
i of 2 8 7 8 5 1 31
events

4 of obs 132 292 379 486 223 172 f 1664
constant 7.9E-4 6.63-4 5.4E-4 7.88-4 -.006 -.0014 -3.3E-4
Std. .004 .002 .003 .004 .0067 .0051 .0017
error

T-stat .2 .368 .17 .185 -.85 -.271 -.l94
6u .001 -3.71 -.14 -.022 .0116 -.02 -.0085
Std. .022 .0052 .0107 .0165 .025 .038 .0064
error

T-stat .055 -.007 -l.306 -1.343 .459 -.476 -l.318
6gp -.041 .002 -.013 .031 -.0108 .0424 .0015
Std. .021 .007 .0132 .021 .0304 .038 .008
error .

T-stat -.189 .268 -.971 1.478 -.355 1.119 .182
631 .009 -.003 .007 -.005 -.0222 -.0159 -.0033
Std. .0183 .005 .0109 .018 .026 .0295 .0065
error

T-stat .49 -.59 .631 -.303 -.857 -.54 -.513

 

 

 

 

 

 

 

 

 

*The results show that after differentiating inflow and outflow
constraints, there is no direct relationship between the changes
in premium and the announcement of restriction changes.

 

109

Tabl. III-3

Regression of d9!!! variables over Werld-adjusted premium
change

jth fund Premium-AWOrld CECF index‘without jth fund

Ia+b1D1+b,D,+b,D,-t-e

1% I -1 if t is between two and seven weeks before the
announcement of tightening an inflow restriction or
loosening an outflow restriction, and 1 if loosening an
inflow restriction or tightening an outflow restriction of
the jth country.

In I -1 if t is between one week before and one week after
the announcement of tightening an inflow restriction or
loosening an outflow restriction, and 1 if loosening an
inflow restriction or tightening an outflow restriction of
the jth country.

1% I -1 if t is between two and seven weeks after the
announcement of tightening an inflow restriction or
loosening an outflow restriction, and 1 if loosening an
inflow restriction or tightening an outflow restriction of
the jth country.

b1: The average weekly effect before the announcement of
regulatory changes on the event-weighted premiums/discounts.
b,: The average weekly change in the event-weighted premium
or discount during the three-week period surrounding the
announcement.

b,: The average weekly effect after the announcement of
regulatory changes on the event-weighted premiums/discounts.

Test with BBNW data

Variable Coefficient SE(Coef) T Sig T
b1 -.035983 .028432 -1.266 .2059
b, -.008485 .035972 -.236 .8136
b3 .018737 .028587 .655 .5123
(Constant) -.014319 .008252 -1.735 .0830

Test with updated data

Variable Coefficient
b1 -.038668
b2 .022157
b -.001582

3
(Constant) -.002084

110

SE(Coef)
.031542
.039450

.031657
.008495

T
-1.226
.562

-e050
-e245

Sig T
.2204
.5744

.9602
.8062

111

Table III-4

Regression of NASDAQ index return over Herld CECF Premium
Change index

AWPINK I a + b * NASDAQ index return + e,

DEPENDENT VARIABLE : World CECF Premium change Index

Variable Coefficient SE(Coef) T Sig T
NASDAQ INDEX .468490 .058612 7.993 .0000
(Constant) -.001792 .001080 -l.659 .0976

* Change in World CECF premium index is significantly
positively correlated with NASDAQ index return at 1%
dependence level. This is consistent with Lee’s hypothesis
in that small investor sentiment commonly affects both.

112

Table III-5

Korea fund's premium and Korean stock exchange index

(Daily observations)

Prem = a + b*KMR

Prem: Changes in KF premium at time t
KMR: Korean stock market return at time t-l

Variable Coefficient SE(Coeff) T Sig T
KMR .034846 .006598 5.281 .0000
(Constant) -3.21879E-04 8.5139E-04 -.378 .7054

Result shows that Korean stock market return affects changes
in KP premium. 1607 daily observations are used from 9/84
through 12/91.

113
Table 111-6
Chan e in re-iu-s discounts and NAV return a ro for
to at etu

(Weekly observations)

APREMIUMt- A.+ B*NAV return...1 + et

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Fund Name B(Coeff) S.E.(Coeff) T-stat(signif)
BRAZIL .005 .077 .064(.9489)
FRANCE .3679 .149 2.463(.0147)
GERMANY .4007 .109 3.68(.0003)
HELVETIA(SWISS) .3409 . 116 2 .95( .0035)
INDIA .1135 .103 1.1(.273)
ITALY .576 .133 4.33(.000)
JAPAN .0513 .058 .887(.376)
KOREA .0255 .068 .377(.706)
MALAYSIA .178 .088 2.033(.043)
MEXICO .0154 .065 .235(.814)
SPAIN .242 .177 1.37(.1725)
TAIWAN .0229 .108 .212(.832)
THAI .0096 .099 .098(.922)
UK .00115 .0048 .242(.809)

 

 

 

 

 

APREMIUM. - A + B*NAV return. + et

114

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Fund Name B(Coeff) S.E.(Coeff) T-stat(signifi)
BRAZIL -.49 .067 -7.34(.000)
FRANCE -.36 .15 -2.39(.018)
GERMANY -.803 .1 ' -7.94(.000)
EELVETIA(SWISS) -.212 .117 -1.81(.0713)
INDIA -.37 .099 -3.74(.000)

ITALY -.603 .133 -4.53(.000)

JAPAN -.165 .058 -2.89(.004)

KOREA -.206 .067 -3.09(.002)
MALAYSIA -.202 .087 —2.31(.022)
MEXICO -.338 .064 -5.32(.000)

SPAIN -.544 .174 -3.13(.002) I
TAIWAN -.731 .095 -7.71(.000)

THAI -.399 .095 -4.2(.000) I
0x .002 .005 .433(.667) I

 

 

 

 

115

Table III-7

Relationship between premium and institutional

ownership

Average annual premium/discount = a + b * Inst. Ownership

Variable Coefficient SE(coeff) T Sig T
Inst. ownership -.009317 .004300 -2.167 .0383
(Constant) .151850 .065675 2.312 .0278

The result is significant at 5% level. There should be some
other reasons but there seems to be some institutional
preferences on discount funds than premium funds.

116

Thhl. III-8

Correlation test between premium and variance ratio

Correlation between average annual premium/discount and
ratio of annual CECF share price variance over NAV variance
is tested in this table. They are strongly positively
correlated at 1% significance level. .

Approximate
Statistic Value T-value Si nificance
Pearson’s R .44366 4.25862 .00006

*14 country specific country funds are used for the test
between 1981-1991. They are Brazil, France, Germany, Swiss,
India, Italy, Japan, Korea, Malaysia, Mexico, Spain, Taiwan,
Thailand, and UK fund.

 

 

 

117

Table III-9

Logit analysis

Logit analysis was done to test whether variance ratios are
really different between premium funds and discount funds.
The result is significant at 1% level that premium funds
seem to have higher variance ratios than discount funds.
This is consistent with the hypothesis that the ownership
structure could possibly affect the variance ratios of
CECFs.

1 if premium fund, 0 if discount fund
Annual variance of CECF SP over that

Dependent Variable
Independent variable

of NAV
Iterations Taken 7
Usable Observations 76 Degrees of Freedom 74
Cases Correct 57
Log Likelihood -41.561662
Average Likelihood 0.5787620
Variable Coefficient Std Error T-Stat Signif
1. Constant -1.362 0.364 -3.745 0.00018

2. Variance ratio 0.197 0.076 2.594 0.00947

*14 country funds are used for the test over the period
1981-1991. They are Brazil, France, Germany, Swiss, India,
Italy, Japan, Korea, Malaysia, Mexico, Spain, Taiwan, Thai,
and UK fund.

Table III-10

Correlation T le based 0 monthl

Variable

ERAzIL
FRANCE
GERMANY
BELVETIA
INDIA_GR
ITALY
JAPAN
KOREA
MALAYSIA
MEXICO
SPAIN
TAIWAN
TEAI

UK

Correlations:
BRAZIL
FRANCE
GERMANY
HELVETIA
INDIApGR
ITALY
JAPAN
KOREA
MALAYSIA
MEXICO
SPAIN
TAIWAN
THAI
UK

Correlations:

‘BRAZIL
FRANCE
GERMANY
HELVETIA
INDIA_GR
ITALY
JAPAN
KOREA
MALAYSIA
MEXICO
SPAIN
TAIWAN
THAI
UK

Cases

40
42
65
53
37
68
67
87
56
107
42
53
47
51

BRAZIL
1.0000
-.4674

.1194

.2860

.0135

.0120

-.8005**
.0536
.3581

-.1422
.2913

-.4680*
.2242

JAPAN

1.0000

-.2445
.6242
.4109*

- e 6905

Mean

-.1928
-.1760
.0226
-.0700
-.0836
-.1364
-.1022
.4247
-.0316
-.1480
.1460
.2325
.1931
-.l778

FRANCE

1.0000
.1293
.3320
.5561
.4909**
.7170*
.4848**
.0381
.7042**
.5691*

-.2075
.2526
.2267

KOREA

1.0000
.2160

-.1273
.2648
.4184*
.5770**

-.1576

118

remium discount

Std Dev
.2901
.0936
.1505
.0784
.1912
.1390
.0972
.1994
.1768
.2866
.2728
.2972
.1542
.0519
GERMANY HELVETIA
1.0000
.7725** 1.0000
.6245** .6019**
.6642** .6342**
.3204 .
-.0927 -.0344
.7200** .7243**
.5060** .4345**
.7581** .7416**
-.0436 .2900
.3796* .3910*
.6435** .6388**
MALAYSIA MEXICO
1.0000
.3211* 1.0000
.7301** .3709*
.2620 -.2112
.3379 .0461
.6469** .3261*

INDIApcR

1.0000
.5317**

.2953
.7878**
.5615**
.7299**
.0749
.3207
.7367**

SPAIN

1.0000

-.0678
.5644**
.5979**

ITALY

1.0000
.7719**
.1386
.5914**
.3365*
.7412**

-.0813
.2762
.5660**

TAIWAN

1.0000
.3997*
-.1678

119

Correlations: THAI UK
THAI 1.0000
UK -.0818 1.0000

2-tai1ed Signif: * - .01 ** - .001
' . ” is printed it a coefficient cannot be computed
*This table shows that changes in many funds’ premiums/discounts are

significantly positively correlated with each other. No particularly
strong regional correlations are detected.

120

Table III-ll

orrelati t ble based n weekl chan es n r d cou s

l

o: CECE;

This table shows that changes in CECF premiums are strongly correlated with
each other and with NASDAQ index returns. Except for a few number of Funds,
most funds' premium changes are strongly positively correlated with NASDAQ
index return. Small investor sentiment which affects NASDAQ index return also
seems to affect CECF premium changes. Table values represent pairwise
Spearman correlation coefficients. The number of observations is in
parentheses. P-values are also reported.

Variable definition

CLEM : change in Clemente global fund premium

FAUS : change in First Australia fund premium

FIBE : change in First Iberian fund premium

MAL : change in Malaysia fund premium

NASD : NASDAQ index return

OVER : change in Overseas securities fund premium

SCAN : change in Scandanavian fund premium

SCUD : change in Scudder New Asia fund premium

TEMP : change in Templeton emerging market fund premium
WORLD : change in Worldwide V fund premium

WPI : change in world CECF premiums

ASIA

BRAZIL

CLEM'

FAUS

FIBE

FRANCE

GERMANY

INDIA

ITALY

JAPAN

KOREA

ASA
1.0000
( 389)
P- .
-.1806
( 30)
P- .109
-e0761
( 34)
P. .669
.2583
( 74)
P- .026
( 144)
P- .034
.2753
( 33)
P. .121
.3471
( 129)
P- .000
.0907
( 123)
P- .319
.0710
( 17)
P. .787
.1400
( 141)
P- .098
.0524
( 311)
P. .357
.0569
( 210)
P- .412

ASIA

1.0000
( 254)

P' .
.1260
( 194)
P. .080
- e 0057
( 71)
P- .962
.4790
( 80)
P- .000
.0214
( 31)
P- .909
.0616
( 124)
P- .497
.3648
( 249)
P. .000
.1300
( 179)
P. .083
.1343
( 246)
P- .035
.0641
( 12)
P. .843
.1831
( 247)
P- .004

121

BRAZIL
1.0000
( 204)
P. .
.0371
( 35)
P- .833
.0644
( 35)
P- .713
( 32)
P. .325
.0457
( 79)
P. .689
.1255
( 201)
P- .076
.0425
( 175)
P- .577
.0707
( 196)
P- .325
( 200)
P- .477

CLEM

1.0000

-.0544
( 73)
P- .648
.2035
P- .248
.2308
76)
P- .045
.0429
P- .716
-.3574
P- .159
.1775
P- .128
.5668
P- .319
.3083

( 73)
P- .008

FAUS

1.0000
( 149)
P- .

.2977
( 34)
P- .087

-eO“7
( 127)
P- .618

.1593
( 121)
P- .081

( 17)
P- .031

.2099
( 137)
P- .014

( 75)
P- .855

.1154
( 145)
P- .167

FIBE
1.0000
( 34)
P- .
.1670
( 34)
P- .345
.4043
( 32)
P- .022
-.3113
( 15)
P- .259
.2304
( 33)
P- .197
-.0103
( 33)

P- .955

MEXICO
NASD
OVER
SCAN
SCUD
SPAIN
SWISS
TAIWAN

TEMP

THAI

UK

ASIA

.2604
( 250)
P- .000

( 249)
P. .524

.1974
( 229)
P8 .003

P- .608

.0931
( 237)
P- .153

.1436
( 193)
P- .046

.2830
( 235)
P- .000

.0994
( 228)
P- .134

( 252)
P- .876

.2843
( 213)
P- .000

.2513
( 233)
P- .000

122

BRAZIL

.0480

( 203)
P- .497
.0153

( 199)
P- .830
.0714

( 179)
P- .342
.0569

( 35)
P- .746
.1730

( 196)
P- .015
.0523

( 189)
P' .474
.1135

( 202)
P- .108
.1435

( 186)
P- .051
.0080

( 203)
P- .909
.0546

( 202)
P- .441
-.0499

( 199)
P. .484

CLEM

.2322

( 75)
P- .044
.2857
76)

P- .012
.3961

( 76)
P- .000
.1676

( 75)
P- .151
( 74)
P- .749
-.0038
( 25)
P- .986
.1916.

( 67)
P- .120
.0634

( 69)
P- .605
.2046

( 75)
P- .078
.0363

( 45)
P- .813
.0633

( 69)
P- .605

PADS
.1908

82)
.086

-.0387

149)
.639

1760
149)
.032

124)
.559

.2791

75)
.015

.0594

25)
.778

1545
67)
.212

.0368

86)
.736

91)
.492

1217
45)
.426

.3747

69)
.002

'P“

FIBE

34)
.669

-.1630

34)
.357

.2043

34)
.246

2438

34)
.165

3809

34)
.026

1383

22)
.539

3423

34)
.048

0073

31)
.969

4550

34)
.007

0902

34)
.612

3682

32)
.038

US

WORLD

FRANCE

GERMANY

INDIA

ITALY

JAPAN

KOREA

MEXICO

ASA
-.0223
( 157)
P- .782
( 114)
P. .157
.4152
( 389)
P. .000
FRANCE
1.0000
( 180)
P. .
.0506
( 172)
P- .510
.1341
( 59)
P- .311
.2164
( 177)
P- .004
.2166
( 57)
P- .106
.2597
( 175)
P- .001
.0614
( 132)
P. .484
.0723
( 180)
P- .335

ASIA
.3545
( 32)
P- .001
.4691
( 254)
P. .000
GERMANY
1.0000
( 303)
P- .
( 183)
P- .416
.0029
( 295)
P' .961
.0452
( 53)
P- .748
.1322
( 296)
P- .023
.2542
( 258)
P. .000
.0957
x 298)
P- .099

123

BRAZIL

.0299
P- .864
.3604
P- .000

INDIA

1.0000
( 186)
P- .

.1663
( 181)
P- .025

.1531
( 182)
P- .039

.2029
( 186)
P- .005

.0886
( 182)
P- .234

CLEM

-.0036
P- .975

.3759
P- .001

ITALY

1.0000
( 317)

.3307
( 69)
P- .006

.0917
( 310)
p- .107

.0296
( 256)
P- .637

.0369
( 313)
P- .515

FAUS
.3164
( 111)
P- .001
.3655'
( 149)
P- .000
JAPAN
1.0000
( 319)
P- .
.0762
( 142)
P- .367
.1886
( 10)
P- .602
.0295
( 264)

P- .634

FIBE

.1876
P- .288
.4982
P- .003

KOREA

1.0000
( 393)
.1913
P- .002
.0693

( 370)
P- .183

NASD

OVER

SCAN

SCUD

SPAIN

SWISS

TAIWAN

TEMP

THAI

0?

WORLD

FRANCE
.4740

( 180)
P- .000
.2274

( 129)
P- .010
.1769

( 120)
P- .053
.0435

( 71)
P- .719
.1678

( 115)
P- .073
.2267

( 133)
P- .009
.0999

( 140)
P- .240
.0404
93)

P- .701
.2465

( 116)
P- .008
.1630

( 117)
P- .079
.4382

( 180)
P- .000

GERMANY
.0710
( 279)
P- .237
.1145
( 124)
P- .205
.1251
( 244)
P- .051
.3830
( 198)
P- .000
.2519
( 242)
P- .000
.0165
( 247)
P- .797
.0682
( 267)
P- .267
.2960
( 218)
P- .000
.1610
( 240)
P- .013
.0884
( 115)
P- .348
.4806
( 303)
P- .000

124

INDIA
.3369
( 162)
P- .000
.0694
( 17)’
P- .791
.1043
( 178)
P- .166
.0333
( 184)
P- .653
.0474
( 184)
P- .523
.0704
( 166)
P- .367
.0933
( 185)
P- .207
“.1090
( 184)
P- .141
.2062
( 179)
P- .006
-.2082
( 17)
P- .423
.3309
( 186)
P- .000

ITALY
.2140

( 292)

P- .000

.2040
( 129)

P. .020 -

.4016
( 241)
P- .000

.0904
( 195)
P- .209

.0750
( 239)
P- .248

.0349
P- .588

.0807
( 264)
P- .191

-.0408
( 215)
P- .552

.0977
( 237)
P- .134

.0290
( 116)
P- .757

.4055
( 317)
P- .000

JAPAN
.1692
( 319)
P- .002
-.2378
( 219)
P- .000
.2356
( 55)
P- .083
-.1802
( 5)
P- .772
-e2546
22)
P- .253
-.0602
( 19)
P- .807
-.0015
( 156)
P- .985
.2468
( 42)
P. .115
( 319)

P- .000

KOREA
.3175

( 368)
P- .000
-.0245

( 43)
P- .869
.1104

( 126)
P- .218
.1173

( 245)
P- .067
.2096

( 200)
P- .003
.1760

( 240)
P- .006
.3482

( 244)
P- .000
.1248

( 265)
P- .042
.0932

( 218)
P- .170
.2667

( 238)
P- .000
.0817

( 113)
P- .390
.4936

( 393)
P- .000

ASA

ASIA

BRAZIL

CLEM

FAUS

FIBE

MEXICO

NASD

OVER

SCAN

SCUD

MAL
.1497
( 33)
P- .177
.2604
( 250)
P- .000
.0480
( 203)
P- .497
.2322
( 75)
P- .044
.1908
( 82)
P- .086
-.0761
( 34)
P- .669
1.0000
( 263)
P- .
.1981
( 259)
P- .001
.2412
( 238)
P- .000
.2766
( 84)
P- .011
.1096
( 248)
P- .085

MEXICO
.0365
( 334)
P- .506
( 249)
P- .524
.0153
( 199)
P- .830
.2857
( 75)
P- .012
'.0387
( 149)
P- .639
-.1630
( 34)
P. .357
1.0000
( 516)
P- .
.1372
( 495)
P- .002
-e 1107
( 170)
P. .151
-e0708
( 130)
P- .423
.1233
( 244)
P- .054

125

NASD
.1768
( 389)
P- .000
.1974
( 229)
P- .003
.0714
( 179)
P- .342
.3961
( 75)
P- .000
.1760
( 149)
P- .032
.2043
( 34)
P- .246
1.0000
( 553)
P- .
-.1055
( 221)
P- .118
.2035
( 130)
P- .020
.2269
( 224)
P- .001

OVER

.0382
( 217)
P- .576

1.0000
( 221)
P- .

SCAN

.3550
( 127)
PI .000
81)
P- .608
.0569
( 35)
P. .746
.1676
( 75)
P- .151
( 124)
P- .559
.2438
( 34)
P- .165
1.0000
( 130)

P- .
.1460
( 77)

P- .205

SCUD
.0011
( 75)
P- .992
.0931
( 237)
P- .153
.1730
( 196)
P- .015
-.0379
( 74)
P- .749
.2791
( 75)
P- .015
.3809
( 34)
P- .026
1.0000
( 248)
P- .

SPAIN

SWISS

TAIWAN

US

WORLD

SPAIN

SWISS

.2040
( 202)
P- .004

.0797
( 246)
P- .213

.2337
( 237)
P- .000

.0856
( 261)
P3 .168

.4555
( 222)
P- .000

.1354

( 244)
P- .035

.0638
P- .562
.5420

( 263)
P- .000

SPAIN

1.0000
( 202)
p-

.2233
( 200)
P- .001

MEXICO
-e0033
( 198)
P' .963
.0024

( 242)
P. .971
.1762

( 247)
P- .005
.3051

( 267)
P. .000
.0678

( 219)
P- .318
-e0117
( 240)
P. .856
.0363

( 123)
P- .691
-.0155
( 117)
P- .868
.5530

( 516)
P- .000
SWISS
1.0000
( 247)

126

NASD
.1924
( 177)
P. .010
.1818
( 222)
P- .007
.3122
( 228)
P- .000
.1864
( 247)
P- .003
“.0727
( 199)
P- .307
.3371
( 220)
P- .000
-.0203
( 157)
P. .801
.1393
( 117)
P. .134
.3223
( 553)
P' .000
TAIWAN

OVER

-e0478
( 153)
P3 .557

.3607
( 221)
P- .000

SCAN
-.0917
( 25)
P- .663
.0899
( 70)
P- .459
.1308
( 89)
P- .222
.3868
( 92)
P. .000
.0450
( 45)
P. e766
‘.1307
( 71)
P. .277
-.2740
( 114)
P. .003
.3357
( 130)
P- .000
THAI

SCUD
-e0‘37
( 195)
P. .544
.1083
( 236)
P- .097
.0863
( 225)
P- .197
.1213
( 246)
P. .057
.0758
( 214)
P- .269
.1416
( 235)
P. .030
.3701
( 73)
P- .001
.4465
( 248)
P- .000

UK

TAIWAN
TEMP

THAI

WORLD

US

WORLD

SPAIN
.0978
( 182)
P- .189
.0008
( 201)
P' .991
.2054
( 200)
P- .004
.2181
( 195)
P- .002
( 25)
P- .037
.4407
( 202)
P. .000
US
1.0000
( 157)
P. .
.1911
( 157)
P'I .017

SWISS

.0480

( 223)

P- .476

.0896

( 246)

P- .161

.2134

( 221)

P- .001

.3307

( 239)

P- .000

.1340

( 70)

P- .269

.4045

( 247)

P- .000

WORLD

1.0000

( 117)
P- .

.2408

( 117)

P- .009

127

TAIWAN

1.0000
( 251)
P. O

.1306
( 244)
P- .042

.0534
( 201)
P- .452

.2020
( 220)
P- .003

-.0839
( 90)
P- .432

.4773
( 251)
P- .000

WPI

1.0000
( 578)
P- .

TEMP

1.0000
( 272)
P- .

-.0034
( 222)
P- .960

.0043
( 244)
P- .947

-e2181
( 94)
P- .035

.3337
( 272)
P- .000

(Coefficient / (Cases) / 2-tailed Significance)

” . ' is printed if a coefficient cannot be computed

THAI

1.0000
( 223)

.0572
( 215)
P- .404

.1451
P- .336
.4190

( 223)
P- .000

UK

1.0000
( 245)

.4587
( 72)
P- .000

.3993
( 245)
P- .000

“1))))))))))))))r