____ ____             ____               __             W PS    1X733
POLICY RESEARCH WORKING PAPER   1733
'.hinf.:' .;':   on: operating,  *
' :*    s.-,:  . .. .
The Usefulness of Private   _                                              m  o oai
c' diti'ons based on direct
and Public Information                                                :&perience in a cotuntry is
likely to bethe information
for  Foreign    Investment                         ~~~~~~~~~most credible to foreign
investors. But for investors
Decisions
new to a country, information
infehlred from observing
Yuko Kinosbita                                                        -othe*s investing in that
Ashoka Mody                                                            country may be more
- infr'ential - and may lead to
cascading investrnents in that
country, as happened in
China ard Vietram. Countries
that do not draw a critical ,,,
inass of investors are in
;  danger of being bypassed by
.investors..
The World Bank                                                              ,
East Asia and Pacific Regional Office                               A
February 1997*   ,



PoFcy RESEARCH WORKINC PAPER 1 733
Summary findings
Using a specially designed survey of Japanese firms              potential investors to view particular locations favorably.
planning investmenrs in Asia, Kinoshira and Mody                 Privatclv held information complements that public
emphasize the importance of privately held information           inforination, and ac-oulits for iniportant variations in
in making for¢ign investment decisions.                           rive.tinier plans.
Information on operating conditions hased oTn direct             Sulj¢jective perceptions about policy on foreign direct
experience in a country is likely to be rlte mnost credible      investinenit are also important.
Information, but for investors new to a c ountry,                   Pol cv designed co attract invcstors -  for exaimiple,
information inferred from  observing others investing in        .speci. , zones for foreign investors -have been
that country may he more influential. Initially, in fact,        *.icce tful in nmanv instances, cspecially in East Asia, but
observing the actions of comppetitors seenms to lead to           have ilso beell a waste of scarce investmuent resources
cascading investments in that -ountry, apparenitly               when not appropriately planned. And such policy is of
through herd behavior (as in China anid Vietnam).                lirtle  alne in artracting those already investing in a
Countries whiich do not draw a critical mass of investors        country, or those who alruady perceive rivals to be active
are in danger of being bypassed for significint perto ls.        rhere
Publicly available information is iniip(Jrtatlit in helping
shape "average" perceptions about a couLntry, leading
This paper -  a product of the East Asia and Pacific Regional Of tice -- is part of a larger effort in the department to study
trends in foreign investment in the region. Copies of the paper are av 1ilable free ftrm the World Bank, 181 8 H Street NW,
Washington, DC 20433. Please contact Rafaela Reff, room Q5- I3 9, n.luphonc 202 473-48 15, fax 202-477-0218, Internct
address rreff(&worldbank.org. February 1997. (32 pages)
The Policy Researcbh WZorking Paper Series disseminates the findings ot w irk in progress to encourage the exchange of ideas ahout
development issues. An objectiue ot the series is to get the findings out quickl,, eiven if the presentations are iess than fully polished. The
papers carr,- the narnes of the authors and should be cited accordingly. The findings interpretations, and conclusions expressed in this
paper are entirelv those o/ tie authors. Ther do not necessarily reprecotnt t?ie vitc of the World  iznk, its Exetutti'e Directors. or the
countries the- represent.
prrodluced by the l'olicy Research I i ¢se mi nation Center



The Usefulness of Private
and Public Information
for Foreign Investment Decisions
Yuko Kinoshita
New York University
Ashoka Mody
World Bank
This paper was initiated while Kinoshita was a consultant at the World Bank. The data for this paper
were obtained through a special survey designed by Susmita Dasgupta, Ashoka Mody, and Sarbajit
Sinha. Osamu Kawaguchi assisted in the preparation of the survey design and its implementation
through the Ministry of International Trade and Industry in Japan.



I I



"...either he should discover the truth about them for himself or learn it from
some one else; or if this is impossible, he should take the best and most
irrefragable of human theories and make it the raft on which he sails through
life." Plato.
Introduction
What information sources do investors use to make their decisions? Does privately-held
information play an important role in the decision-making? If so, is the private information acquired
through direct experience? Or is it based upon observations of actions undertaken by others? These
questions are especially relevant for foreign investment decisions where public, or commonly-held,
perceptions are characterized by considerable uncej tainty and, in particular, may be quite misleading
for an investor's specific requirements.
In this paper, we demonstrate the empirical importance of private information, with others'
actions being a specially important guide when "new" countries are opening up. Significant
discontinuties in investment flows are observed at such times. China has attracted a rush of
investment not only from overseas Chinese but also from U.S., Japanese, and European investors,
starting quite abruptly in the late 1980s and growing explosively into the mid-1990s. China receives
about $40 billion a year of foreign investment despite cumbersome procedures and uncertainty on
property rights and contract enforceability; in contrast, India after rolling back restrictions and a
longer tradition of a market economy chalks up less than $2 billion a year. A similar discontinuity
is now being observed for Vietnam, where competing investors are staking out their positions.
Investors do use publicly available information on market size, stocks of
infrastructure, costs of doing business (including labor costs), foreign investment policies, and other
country characteristics to make new investment decisions. However, such information is rarely a
sufficient guide to investment decisions. Moreover, public information-by definition, available to



2
all potential investors-cannot lead to cascades unless new developments occur, as when a major
policy initiative is announced. Policy initiatives do matter-however, as we show, their
consequences depend upon private information. In contrast, privately-held information-or more
accurately, private beliefs-can have a significant impact on investment flows even when no
fundamental change has occurred but when a perception of change leads to actions by a critical mass
of investors, which then has a snowballing effect. In practice, public and private information
complement each other. They also interact: certain publicly announced, though minor, events can
reinforce latent private beliefs, which may then become the primary drivers of foreign investment.
What is the content of private information? Investors seek information on a variety of
operational conditions which are not publicly available, including the functioning of labor markets,
industrial literacy of the workforce (as distinct from educational attainments), the practical
implementation of foreign investment polices, and the timely availability of inputs. Such
information may be acquired in two ways: through direct experience via past investments in a
country and through inferenences from the behavior of other investors. Where a firm's past
investment is the influence on future investment, we refer to that as a "learning" effect: presence in
an earlier period generates valuable information on market and cost conditions which forms the basis
for making new investments. Such information, this paper shows, is considered very valuable by
investors; but, by definition, it is not available to new investors in a country.
Where the behavior of other investors spurs investment, we infer the influence of reputation
and of strategic rivalry. A rivals' decision conveys the information that the rival considers
investment in a particular country to be a profitable venture, thus increasing the incentive to invest
in that country to benefit from the same opportunities. The importance of such intrinsically valuable



3
information on operating conditions in a country is notably illustrated by General Motors' decision
to locate its Asian hub in Thailand: "...the fact that 11 car manufacturers already operate in Thailand
was a sign that the country's infamous physical infrastructure and labor bottlenecks could be
overcome" (Bardacke 1996). However, this example also highlights that, in addition, private beliefs
may reflect strategic considerations, which may also lead to a self-reinforcing cycle of investment
(Kuran 1995). Anecdotal evidence suggests that the strategic element of such investment decisions
is important." While the learning effect creates persistence through inducing continued investments
by existing foreign investors, inferences based on reputation of others and on strategic rivalry
considerations leads more directly to "herd" behavior.
This paper draws upon two streams of literature: the economics of private information flows
and beliefs and the determinants of foreign investment. Herd behavior parallels and reflects
"cascades" of information flows (Scharfstein and Stein 1990, Bikhchandani, Hirshleifer, and Welch
1992, and Lee 1993). The so-called "herd" behavior-actions based on others' actions-can be quite
rational in as much as it economizes on the gathering of scarce information. Arthur (1995) discusses
several examples from economics and finance where private beliefs play an important role. Kuran
(1995) explains the persistence of several social institutions as well as their abrupt breakdown on
the basis of privately-held but publicly concealed preferences.
The influence of rivalry in driving foreign investment was examined in a pioneering study
by Knickerbocker (1973). He showed that the more oligopolistic an industry, the greater was the
The rush of Japanese motorcycle investors to Vietnam has followed from a perceived
"first mover" advantage. Referring to the general interest in Vietnam, a German investor
recently summarized well the phenomenon: "We simply cannot sit back and let the
Japanese take over another market unchallenged" Financ a , March 28, 1995).



4
likelihood that foreign investments would be concentrated into a short period of time, and hence
display spikes or discontinuities in foreign investment flows. Knickerbocker did not, however, study
the influence of past presence. Recently, Head, Ries, and Swenson (1995) have shown that Japanese
investors in the United States tend to "follow-the-leader," affirning the signalling value of others'
behavior. Kogut and Chang (1996) have used firm-level data for Japanese multinationals investing
in the United States and they find past presence to be an important predictor of new
investments-however, they do not explore the influence of rivalry. The findings of this paper are
also consistent with aggregate evidence of persistence in foreign investment. Wheeler and Mody
(1992) found that U.S. investments into a country were strongly conditioned by existing stocks of
foreign investment in that country (after controlling for a variety of factors, including market size)
and speculated that agglomeration economies may be important especially where investors were
likely to be engaged in the purchase of intermediate inputs from other investors. But even where
agglomeration economies are unimportant, past presence can provide, or the presence of others can
signal, information on business operating conditions which are critical to smooth functioning but
which cannot be easily inferred from generally reported statistical indicators or from stated policy
towards foreign investors. Also, subsequent analysis shows that Japanese investors are equally
influenced by the stock of past investment (Mody and Srinivasan 1996). This study reinforces earlier
micro and aggregate evidence on the value of private information but additionally distinguishes
between the different sources of private information and demonstrates their complementarity with
public information.
The setting for the empirical examination is investment by Japanese manufacturing firms in
a number of key Asian countries in the early 1990s and the data is from a specially designed survey



5
of Japanese investors. The next section describes the questions asked in the survey, the data thus
generated, and the analysis methodology. We then present the simplest model that reveals the role
of private information in determining a fim's likelihood of investing in a country and distinguishes
between the influence of leaming and rivalry among the sampled firms. We examine the robustness
of the finding by introducing several firm characteristics in the estimated equation to determine if
the "private information" merely reflects firm attributes. The role of public information on
investment decisions is dealt with by introducing country dummies, which are assumed to embody
information available to all; also, since public information is assumed accessible to all, regressions
for individual countries help further highlight the role of private information. Successive models
add other host country features that influence the behavior of foreign investors; interaction of
learning and rivalry with these factors provides additional perspectives on foreign investment flows.
Data and methodology
The survey questionnaire was mailed by the Japanese Ministry of Trade and Industry (MrII)
to several hundred Japanese firms of which 173 returned usable responses in March 1993. The
sample thus obtained cannot be treated as representative of all Japanese firms-we do not know the
characteristics of firms who did not respond. There is, however, sufficient heterogeneity amongst
the respondents to permit a statistical analysis of their foreign investment behavior. The firms in our
sample are relatively large. The average annual sales are 330 billion yen (over $3 billion), the largest
firm in the sample has sales of $70 billion and the smallest has sales of $2 million. This is also a set
of firms that is prone to making significant foreign investments-in the three years prior to the
survey, over a fifth of their investment was undertaken outside Japan.



6
Our dependent variable is based on the following question regarding the firm's expectation
that it will invest in specific Asian countries: "In each of the following countries, how likely are you
to invest in the next three years?" Respondents were asked to check a space on a 1-7 scale provided,
ranging from "very unlikely" to "very likely".
VERY                 VERY
UNLIKELY            LIKELY
The question was answered for the following seven countries: China, Thailand, Malaysia, Indonesia,
Vietnam, Philippines, and India. These countries constitute the principal developing country
recepients of foreign investment in Asia. Their level of economic development is substantially lower
than in the so-called Asian Tigers-South Korea, Taiwan, Hong Kong, and Singapore-with
Malaysia being the closest to the Tigers by most development measures. For each of the seven
countries, we have 173 responses, potentially creating 1211 (173x7) observations (however, since
all respondents did not answer all questions, for certain estimations fewer usable observations are
available and where appropriate we have tested for selection bias).
Our two key independent variables are PAST and RIVAL. The questionnaire asked whether
the firm already had a presence in each of the seven countries being studied. For each firm and each
country, the PAST variable was coded 1 if the firm was present in the country and 0 if it was not.
Recall that we infer a learning effect if past presence leads to a high likelihood of future investment.
The other key variable allowed inference on the information obtained from competitors and the
extent of strategic rivalry. The question asked was: "Are your competitors making investments in



7
the following Asian countries?" Once again, the response allowed ranged on scale of 1 (very little)
to 7 (very substantial).
The average value of the responses for the seven countries (and the standard deviations) are
reported in table 1. Respondents to our survey are most likely, by far, to invest in China, the average
measure on the 1-7 scale for China being 4.08. Only 20 percent of the firms have existing
investments in China; however, the perceived level of rivals' interests in China is high, second to
Thailand. Four countries have similar likelihoods of investment: Thailand, Malaysia, Indonesia, and
Vietnam. Of these, Malaysia and Thailand have traditionally attracted substantial Japanese interest,
with 25 and 30 percent of firms respectively reporting existing presence in those countries; and rivals
are also strongly interested. In contrast, Vietnam has low existing Japanese presence and also a
relatively low interest from rivals. The least attractive sites are the Philippines and India, with low
expected investment, low initial presence, and low rivals' activity. Thus, a simple comparison across
countries indicates a correlation between expected investrnent by the firm and its perception of the
strength of rivals' interest in the country. Since past presence is indicated only in 15 percent of the
possibilities, information provided by behavior of rivals is likely to be valuable where the firm is
entering new countries.
An ordered logit model was used to investigate these relationships more precisely. The model
is an extension of the binomial logit model: instead of two choices, we here have three ordered
choices (see Greene 1993). For the purpose of the regression, the likelihood of investment (LFDI)
variable was rescaled. As illustrated, the original data is on a scale of 1 through seven. This was
reconstructed to create three categories: 2 (highly likely to invest where the response was 6 or 7), 1
(moderately likely, where the response was 3,4, or 5), and 0 (unlikely to invest, where the response



8
was 1 or 2). As in the binomial logit model, we assume a latent regression model of the following
form:
y    PX + E
The latent variable y is not observed, but the response indicating the likelihood of investment is
observed. The observed responses are related to the latent variable in the following manner:
y =O      if y ¸gO
y = I    if O<y' ¸ p
y = 2    if M ¸ y
Then, for the logistic cumulative distribution function, the model predicts the following probabilities
for each of the responses:
Prob(y   0) = A(-Px)
Prob(y = 1) = A(V - Px) - A(-Px)
Prob(y = 2) = 1 -      - Px)
The joint probability or the likelihood function is:
L = II[Pr(Y, = O)]i°O[Pr(Y. = 1)]dIl[Pr(Y] = 2)] i2
i=1



9
where d, (k = 0,1,2) is an indicator function equal to I if y; = k and zero otherwise. "n" is the number
of observations, where the observational unit is a firm's investment plans for each country, implying
up to seven observations per firm. The parameters are estimated by maximizing the log of the
likelihood function.
For the analysis, the RIVAL response was condensed into a binary variable: RIVAL was 0
where the response was 1-4 (competitors uninterested or mildly interestd in investing in that country)
and was 1 where the response was 5-7 (competitors seriously interested). As will be evident below,
summary of RIVAL into a binary variable greatly facilitates comparison with the effects of PAST.
Regressions were also performed with the variable retained in its original 1-7 scale and with RIVAL
scaled as 0,1,2 with qualitatively similar results.
The base model: privately held information
At the risk of misspecification, we first highlight the value of privately held information by
regressing the firm's likelihood of investing in a particular country against its past presence or
absence in that country (PAST) and perceptions about competitors' interest in that country (RIVAL).
The danger of misspsecification arises on account of omitting other relevant variables, including firm
characteristics and the role of publicly available information; however, as show below, the basic
results of this section hold up even in more fully specified models. We pooled the data for all
countries thus assuming that the effects of learning and rivalry are the same in the different countries
considered.



10
The results of the ordered logit are presented in table 2. From column 1 it is clear that the
both the firm's past presence and its perception of competitors' behavior has a strong influence on
its plans to invest in a country. The coefficient on PAST (1.80) is somewhat higher than on RIVAL
(1.41). Prima facie this implies that own experience has greater value than the information inferred
from the behavior of rivals, which is a plausible finding. However, it is also likely that there exists
greater measurement error in the RIVAL variable, which would bias downwards the coefficient on
that variable.
The distinction between the two information channels becomes sharper when we interact
them and add the interaction term as an additional variable (column 2 of table 2). This model is
statistically superior to the first model discussed: inclusion of the PAST*RIVAL variable improves
the log-likelihood and from the likelihood ratio test we can conclude (at the 2.5 percent significance
level) that the interaction term belongs to the model.
The results are interesting. The two channels of private information appear primarily as
substitutes for each other, though a certain degree of complementarity in suggested. The sign of the
interaction term is negative and its size, 0.96, is large. Consider the coefficient on PAST in this
regression, which has risen from 1.80 in the previous regression to 2.30. The influence of PAST on
future investment decisions is reduced when RIVAL is equal to 1: the firm places a reduced (though
still substantial) value on its own experience (down from 2.30 to 1.34). Similarly, past experience
devalues information from rivals' behavior quite significantly: when PAST is 1, then the coefficient
on RIVAL falls from 1.63 to 0.67 (1.63-0.96). This seems a plausible result. A firm's experience
with operational conditions is likely to carry greater weight in its calculus than inferences based on
what others are doing. Moreover, any strategic considerations, such as the need for preemptive



11
positioning in the market, are also not important to those who already have a presence in the country,
limiting the need to respond with greater investment when others are perceived to be active.
However, note that when both PAST and RIVAL equal 1, then both are influential-and while past
carries greater weight, those with existing presence do watch and act on rivals' behavior.
Another perspective on these results is obtained by estimating the model's predicted
probabilities for the likelihood of foreign investment.2Y When PAST and RIVAL are both zero, the
model predicts a 93 percent probability that foreign investment will be very unlikely (y=O). With
RIVAL=O, increasing PAST from zero to one, decreases the probability of y=O by 24 percentage
points to 69 percent, with the bulk of the increase-17 percentage points-occuring for y=l
(moderately likely) and a 7 percentage point increase for y=2 (highly likely). In contrast, when
RIVAL=1, increasing PAST from zero to one has a stronger effect: the decrease in probability of y=O
is 41 percentage points, and y=2 experiences a 22 percentage point increase. Thus, though the
primary influence of PAST presence is exerted irrespective of whether strong rivalry is associated,
the positive interaction of PAST and RIVAL is clearly indicated. A similar result is obtained when
PAST is the control variable and RIVAL is varied from zero to one.
Taking the analysis one step further, we introduce firm dummies into the regression (column
3, table 2). Recall, we have multiple observations for each firm (with a maximum of seven
observations where a likelihood was reported for each country). By introducing the firm dummies
we are able to judge whether for each firm, j, certain unobserved firm reporting characteristics (nj)
are associated with the variables PAST and RIVAL. If these unobserved characterisitics, which are
The probabilities are calculated using the set of equations (3) and following Greene
(1993), pp. 675-676.



12
part of the composite error term (Eij=ilj + y,,), are correlated with PAST and RIVAL, then the
coefficients will be biased.  By adding firm  dummies to the regression, the unobserved
characteristics become part of the set of regressors and the error term now has only the white noise
component, yj. Introduction of the firm dummies strengthens the result both in the size of the
coefficients and statistical significance. The increased coefficient sizes on the PAST and RIVAL
variables suggests that the composite error term is negatively correlated with these variables: in other
words, those who have past presence or perceive active rivals are generally more conservative in
their reporting their investment likelihood.
This simple model does well in predicting the outcomes. The model with firm dummies
correctly predicts 75 percent of the stated likelihoods of foreign investment (table 3).2' The "very
unlikely" statements are almost fully predicted. The prediction rate for the "very likely" category is
53 percent. The model prediction rate (at 39 percent) is the lowest for the middle group, with a
significant number of firms predicted to have a lower likelihood of foreign investment than the firms'
stated intentions. The implication is that a number of firms with PAST and RIVAL equal to zero
have moderately aggressive foreign investment plans-possibly, high production costs in Japan are
having a general effect of pushing firms to seek lower cost production locations.
Preliminary analysis in this section, therefore, affirms the importance of information cascades
but highlights the need to distinguish between the two channels through which information flows
may occur and finds, moreover, that these two channels are principally substitutes for each other.
Where past presence exists, it appears to be the more credible information source. However, only
Without the firm dummies the match between model predictions and firms' statements is
68 percent.



13
a small number of firms have past experience in this set of countries: as table 1 shows, only 15
percent of all possible firm-location combinations are occupied-less than 50 firms in the sample
operate even in Malaysia and Thailand, the countries with the highest presence. Thus the behavior
of rivals may be the more important source of information even if it is less valuable than own
experience.
Firm characterisitics: do they correlate with information flows?
If past presence and perception of rivals' behavior were correlated with firm characteristics,
we might be incorrectly attributing foreign investment decisions to particular information flows than
to specific features of the firms. To examine this possibility we introduced firm size (measured by
worldwide sales), the research and development (R&D) ratio, and export propensity into the
regression equation. Table 4 reports the results. The first column repeats the basic regression for
comparison with table 2, since we have lost some observations. The main thing to note is that the
results are essentially the same though we have fewer obserations.
The second column shows the effect of introducing the firm characteristics. Larger firms
have higher expected foreign investment. R&D has only a weak positive relationship to expected
investment; since R&D and size are correlated, it is not surprising that once the influence of size is
controlled, then any independent influence of R&D is not disernible. Finally, firms expecting to
investment significantly in Asia have a low export propensity.
For the present analysis the important observation from this set of regressions is that there
is virtually no change in the basic results that are the focus of this paper. The coefficient on the
RIVAL variable remains unchanged. The coefficient on PAST falls somewhat in magnitude but



14
remains statistically significant at the 1 percent level. Thus it appears that the sources of information
embodied in PAST and RIVAL bear no simple relationship to firm characteristics. Moreover, the
information contained in PAST and RIVAL has a much more important bearing on foreign
investment than do firm characteristics: the log-likelihood changes little when the firm
characteristics are introduced as explanatory variables.
To further explore the influence of firm characteristics, we focus on size and examine if
particular size classes have special affinity to particular information sources. Firms are divided into
three size groups or stratas. Table 5 reports the results of regressions with three dummy variables
for strata 1 (small firms, annual sales of 2-100 hundred million yen), strata 2 (medium-sized firms,
sales between 100-1000 hundred million), and the benchmark strata 3 (large firms, sales over 1000
hundred million). Note first that this set of results confirms that larger size is associated with a
greater likelihood of investment. Of interest are the interactions between size groups and PAST and
RIVAL. Small firms place a great weight on past presence, which significantly increases their
likelihood of new investment. In other words, while small firms are the least likely to make new
investments, the likelihood of continued investment in a country where they have invested in the past
is substantially higher. The data, therefore, suggest that small firms face barriers to initial
investments abroad but once having taken the leap are aggressive learners. In contrast, medium-
sized firms are much more driven by and sensitive to the actions of rival firms than are large or small
firms. Medium-sized firms have a smaller presence abroad than large firms and so have less basis
to draw on that source of information but, unlike small firms, are likely to place greater weight on
strategic considerations.



15
Country effects: the role of public information
The regressions above could be falsely attributing to privately held information a value that
does not exist if information from private sources were collinear with and equally available from
public sources. Clearly, public information sources do influence the foreign investment decision.
Firms obtain information through analytical country reports by international organizations such as
the World Bank and the International Monetary Fund and also by a growing range of private
consulting firms; and the media provides continuing reports on a range of political, economic, and
social attributes of countries. As a consequence. investors form assessments of countries from such
widely available informnation on a large number of factors including market size, growth prospects,
costs of production, the state of infrastructure, and the country's political and institutional
development.
The further analysis in this section, therefore, explores two sets of questions. First, once
firms' perceptions of different countries are controlled for, do the effects of past experience and
rivalry continue to be important? In other words, are the results of the base model robust to
differences in publicly perceived attractiveness of countries considered? Second, do PAST and
RIVAL influence investment to varying degrees in the different countries?
To answer the first question, we augmented the basic regression with country dummies which
capture in a summary forn the relative attractiveness of the different countries. Since the
coefficients on the dummy variables represent the average perception of the country, we take these
to represent the publicly available information. An alternative specification would include specific
country features, such as infrastructure, market size, and labor costs. As Head, Ries, and Swenson
(1995) have argued, a full elaboration of country characteristics is difficult, and hence a country



16
dummy which captures in summary form the country's attractiveness to the "average" investor is
preferred in this situation. In the next section, we do examine the effects of specific country features.
The regressions in table 6 leaves out Vietnam, which is consequently the reference against which the
attractiveness of other countries is measured.
The robustness of the PAST and RIVAL effects is evident (table 6). Both variables are
significant at the 1 percent level as in the base regression and the coefficient values are also similar,
with PAST coefficient somewhat larger than that of RIVAL. This finding reaffirms the value of
privately held information and its possible cascading effects. The regression also highlights the
complementary role of private and publicly available information: general perceptions of a country
based on publicly-held information are also influential in driving investment flows. The log-
likelihood when the country dummies are included is significantly different from the corresponding
regression in table 2, indicating that significant information is contained in these country dummies
and hence in the generally held view of the country. Private information complements the public
information sources, which can only describe the general state of the country and not the business
operating conditions and the actual implementation of the policy regime.
The country dummies also elicit useful information. With Vietnam as the reference, on
average, investors express a strong preference for China. Indonesia, Thailand, and Malaysia are
likely to be somewhat less sought after, though the coefficients are not statistically different from
zero, indicating about parity with Vietnam. Two countries lowest on the preference list are the
Philippines and India. Note also that the coefficients on firm characteristics remain very similar
when country dummies are added, indicating yet another influence on foreign investment flows.
In terms of goodness of fit (table 3), the predictive power of the model improves somewhat



17
overall from 75 to 79 percent. The increase is mainly in the "very likely" category and especially in
the "likely" category, where the "hits" now go up from 39 to 50 percent, implying that countries that
previously predicted as unlikely to undertake investment are being attracted by the generally
perceived attractiveness of investment sites in China and south-east Asia.
Next, given the strong differences in country preferences as indicated by the coefficients on
the country dummies, we explore whether the privately-held information is used or valued differently
depending upon the country contexts. Individual regressions are run for each of the seven countries
(table 7). Several features of these results are worth highlighting. The signs for the PAST and
RIVAL coefficients for each of the countries are the same as for all countries pooled together and
the magnitudes of the coefficients all fall in a similar range (except for the coefficient on PAST for
Vietnam, where there has been virtually no past investment and the interaction term for India which
is very large with a large standard error). Also, eight of the 14 coefficients (seven each on PAST and
RIVAL) are significant at the 1 percent level and three at the 5 percent level.
Some interesting country variations are worth highlighting. For India, Philippines, and
Vietnam, where the PAST variable is not statistically significant, the extent of past presence is also
very small, limiting the statistical predictive power of that coefficient. For Vietnam, the coefficient
on RIVAL is very large, suggesting that firms are very sensitive to perceived actions of rivals and
hence the possibility of a cascading effect. Though the effect is smaller, a similar force may well be
operative for India. At the other extreme, in Malaysia, where significant past presence exists, the
effect of RIVAL is negligible for those who are already operating in that country; however, even in
Malaysia, new entrants are signficantly guided by the actions of rivals. In this respect, Thailand is
different from Malaysia: though a signficant past presence exists there, existing investors in Thailand



18
do appear influenced by the behavior of their rivals.
The conclusion from this set of regressions, therefore, is that public information, as captured
by the country dummies, is important in influencing foreign investment plans. Such information is,
however, complementary to private information obtained through past investment in the country or
transmitted by other firms by their investment actions in that country. Country differences in the
influence of PAST and RIVAL appears to depend in large part upon the extent of the installed base
of investors in that country.
Can governments influence private information?
These findings raise important questions for policymakers. If privately held information is
important and complementary to publicly available information, then what role should policy play
in attracting foreign investment? Should it be limited to improving the enabling conditions and
ensure that these are well-publicized? That would be a legitimate task, for as we have seen publicly
available information is clearly important in guiding foreign investment flows. Or should
policymakers adopt a proactive stance at influencing private information? If the latter course is to
be pursued, then the nature of such information or at least its correlates need to be identified.
In this section we explore specific country characteristics that may be correlated with
privately held information. The survey questionnaire asked respondents to rank several country
characteristics on a scale from unfavorable to favorable. While "objective" measures may be used
to differentiate countries, the reported perceptions reveal the subjectively held views, which is our
interest. Several country features were explored; here we highlight three of them: foreign direct
investment (FDI) policy, labor costs, and market size. In doing so, we are handicapped by a reduced



19
number of observations to work with since many respondents did not provide the ratings on country
characteristics. As such, there is some danger here of a selection bias in the results. However, since
the coefficients on PAST and RIVAL remain similar to the earlier results, we can have some
confidence that the extent of the selection bias is not large. Formal tests confirned the absence of
selection bias."4
FDI policy was explained to respondents to include such elements as the ability to repatriate
earmings, restrictions on foreign ownership, and the requirements to export and source inputs locally.
Perceptions of FDI policy are strongly influential in conditioning future plans to invest in a country
(table 8, column 1). The coefficient on FDI policy is positive and significant at the 5 percent level.
Note, however, that the coefficients on PAST and RIVAL also remain positive and significant at the
1 percent level. Thus, the evidence seems to suggest that FDI policy is additional information to that
obtained by from past investment experience and actions of competitors.
Though representing additional information, perceptions of FDI policy interact in interesting
with ways with PAST and RIVAL. Notice that the coefficient on the interaction term, FDIplcy*past
is negative. Hence when past is equal to 1-i.e., when the firm has a past presence in that
country-the effect of FDI policy is more than wiped out. In other words, perceptions of FDI policy
matter little when the firm has first-hand operational experience in the country. The corollary is that
perceptions of good FDI policy are especially important in attracting new investors. In this sense,
A'    A reporting equation for the country characteristics (e.g., FDI policy) was first obtained
with a dependent variable z if that characteristic was reported equal to 1 and zero
otherwise. The variable z was regressed on firm characteristics: size, R&D, and export
propensity. The predicted values of z were used to estimate "lamda" or the inverse Mills'
ratio. The investment equation was reestimated with lamda as an additional variable.
Since the coefficient on lamda was insignificant, the absence of selection bias is
indicated.



20
the two types of information are complementary not for individual investors but rather from the
perspective of a country's policymaker: improving the working conditions is likely to be effective
for past investors and a liberal FDI policy is more relevant for first-time investors.
FDI policy has a very similar interaction with RIVAL. Where rivalry is important, it
overwhelms the influence of FDI policy: firms rely to a much greater extent on the actions of rivals
than on their perceptions of FDI policy, however favorable. This finding is consistent with General
Motors' decision to invest in Thailand, following decisions by several autombile and auto parts
manufacturers to locate in Thailand, despite very generous incentives offered by the Philippines,
which Thailand was unable to match (Bardacke 1996). The finding may also explain the rush into
China, where policies and legal operating conditions remain unclear to many Western investors.
The implications for the policymaker are complex. FDI policy does not appear to be a key
element of privately-held information that drives foreign investors. A favorable impression of FDI
policy is desirable when investors have no direct country experience or when competing investors
are not already planning to or undertaking investments in the country. Once investment begins to
occur, the role of privately-held information acquires significant importance. However, a favorable
FDI policy continues to be important to the extent that it minimizes the reliance of firms on
privately-held information, inducing them hopefully to make their investment decisions on the basis
of country fundamentals.
We also explored the possibility that privately-held information may correlate with
perceptions of labor costs in the country. While costs of production depend upon a variety of factors,
labor costs are thought of as an important attraction of this set of Asian host countries. As with FDI
policy, we added perceptions on labor cost as an additional variable to the basic regression. Where



21
labor costs are perceived as being low, the likelihood of future investments does increase
significantly (a favorable labor cost receives a high score and hence a positive sign on that
coefficient). But once again, the influence of PAST and RIVAL remains largely unchanged. Thus
the inference again is that perceptions of labor cost while important are not the essence of the
privately-held information but rather a complement to that information. The terms interacting
perceived labor costs with PAST and RIVAL are positive, though the interaction with past is not
statistically significant. In contrast, therefore, to perceived FDI policy (which becomes less
importance when private information is available), labor costs actually grow in importance when
privately-held information exists.
Finally, a larger market size also improves the improves the attractiveness of a country to
Japanese investors (the results are not presented in the interests of space but are available on request
from the authors). The interaction with RIVAL has no influence, but the interaction with PAST has
a negative and significant sign (at the 5 percent level). Thus, past presence reduces the influence of
a large market size.
Conclusions and discussion
Using a firm-level data set, we explored the empirical importance of privately-held
information in foreign investment location decisions. Though the limitations of a one-time survey
did not permit us follow an information "cascade" over successive generations, the value of private
information, which is central to the cascade phenomenon was consistently and impressively evident
The data permitted us, moreover, to distinguish between two types of private information:
one that was obtained through direct experience in the host country and the other that was inferred



22
from observing competitors. Direct experience is seen to provide the more credible inforrnation, as
may be expected. However, in the early phases of investing in a new country when few firms have
experience in the country, the actions of competitors are likely to be dominating effect, leading to
an apparent herd behavior. Such is apparently the case currently for China and Vietnam, which are
attracting large numbers of new investors. In contrast, countries, such as India and the Philippines,
that do not draw the attention of a critical mass of investors are in danger of being bypassed for
significant periods of time.
We enquired how the privately-held information could was related to publicly available
information and found it to be complementary. Thus, while firms form "average" perceptions about
a country leading them all to view particular locations favorably, considerable variation in
investment plans exists around these averages-an important element of such variation is explained
by privately-held information. We explored also whether private information was a proxy for
subjective beliefs on certain country characteristics: FDI policy, labor costs, and market size. Again,
the finding was that while such subjective perceptions are important, they represent additional
information to that obtained through either past experience or through observing others.
For policymakers, these findings represent a challenge. A generally favorable view of the
country based on its fundamentals as well as perceptions of good policy and low labor costs lead to
increased foreign investment. However, creating the right conditions for investors to directly
experience the rigors of operating in a country is empirically important, as is the opportunity to
observe competitors. This raises the controversial issue of special zones for foreign investors. While
successful in many instances, especially in East Asia, they have also been a waste of scarce
investment resources where not appropriately planned. An emerging approach is for the government



23
to take the lead in creating the policy conditions for the creation of such zones but allow private
investors to undertake the necessary investments and thus ensure greater efficiency. Mexico offers
an example. Under the maquiladora program, the policy environment has been created to attract
foreign investors. Several private initiatives have resulted in so-called "shelters" that provide early
hand-holding services to new foreign investors.



24
REFERENCES
Arthur, Brian. 1995. "Complexity in Economic and Financial Markets." Complexity, John Wiley &
Sons, Inc., 20-25.
Bardacke, Ted. 1996. "General Motors'Thai Hub will Boost Drive Into Asia." Financial Times, May
31.
Bikhchandani, S., D. Hirshleifer, and I. Welch. 1992. "A Theory of Fads, Fashion, Custom, and
Cultural Change as Informational Cascades." Journal of Political Economy 100 (5): 922-1026.
Greene, William. 1993. Econometric Analysis. Second Edition. Englewood, N.J.: Prentice Hall.
Head, Keith, John Ries, and Deborah Swenson. 1995. "Agglomeration Benefits and Location
Choice: Evidence From Japanese Manufacturing Investments in the United States." Joumal of
International Economics 38: 223-247.
Knickerbocker, Fredrick T. 1973. Oligopolistic Reaction and Multinational Enterprise. Boston:
Harvard University, Graduate School of Business Administration.
Kogut, Bruce and Sea Jin Chang. 1996. "Platform Investments and Volatile Exchange Rates: Direct
Investment in the U.S. by Japanese Electronic Companies." The Review of Economics and Statistics
78 (2): 221-231.
Kuran, Timur. 1995. Private Truths, Public Lies: The Social Consequences of Preference
Falsification. Cambridge, MA: Harvard University Press.
Lee, I. H. 1993. "On the Convergence of Informational Cascades." Journal of Economic Theory
December: 395-411.
Mody, Ashoka and Krishna Srinivasan. 1996. "Japanese and United States Firms as Foreign
Investors: Do They March to the Same Tune?" Mimeographed, The World Bank and The
International Monetary Fund, Washington D.C.
Scharfstein, D. and Jeremy Stein. 1990. "Herd Behavior and Investment." American Economic
Review 80: 465-479.
Wheeler, David and Ashoka Mody. 1992. "International Investment Location Decisions: The Case
of US Firms." Journal of International Economics 33: 57-76.



Table 1. Descriptive statistics: country averages of principal variables
Instment Plan         Past Presence         Rivals' Actiity           Investor Perceptions (Scale: 1-10)
( Scale: 1-7)      (Yes=1, No=O)            (Scale: 1-7)      FDI Iolicy   Labor Cost   Domestic Market
China                           4.08                 0.20                   3.67              5.01          8.41             7.35
(2.21)               (0.40)                 (2.32)           (1.71)        (1.48)           (2.30)
India                           1.72                 0.03                   1.94              4.00          7.18             4.32
(1.29)               (0.17)                 (1.58)           (1.44)        (2.12)           (2.22)
Indonesia                       2.84                 0.18                   3.31              5.42          7.35             6.18
(1.99)               (0.38)                 (2.21)           (1.83)        (1.71)           (2.40)
Malaysia                        2.85                 0.25                   3.53              6.22          6.25             6.11
(2.01)               (0.43)                (2.25)            (1.75)        (1.78)           (2.02)
Philippines                     2.02                 0.06                   2.54              4.52          6.77             4.70
(1.55)               (0.24)                (1.80)             (1.80)       (1.85)           (1.97)
Thailand                        3.16                 0.30                   4.10              6.05          6.66             6.56
(2.09)              (0.46)                 (2.21)             (1.65)       (1.63)           (2.08)
Vietnam                         2.53                 0.01                   1.92              4.32          7.90             4.18
(1.85)               (0.11)                (1.56)             (1.90)       (1.78)           (2.35)
Al                              2.75                 0.15                   3.02              5.16          7.22             5.74
(2.01)               (0.35)                (2.17)             (1.89)       (1.89)            (2.46)
Note: Parentheses are standard errors.



Table 2. The base model :value of private infornation
dependent variable = LFDI (likeihood of FDI)
[1]                  [2]               [3]
intercept                         -2.61**              -2.66**           -5.52**
(0.12)              (0.12)             (0.32)
past                              1.80**               2.30**             2.78**
(0.20)              (0.28)             (0.38)
rival                             1.41**               1.63**             2.03**
(0.16)              (0.18)             (0.26)
past*rival                                             -0.96*            -1.44**
(0.50)
-. t1.50                                                1.51               2.09
firm dumm-ies                       no                   no                yes
country dummies                     no                   no                 no
n                                  997                  997                997
logliklihood                     -780.75               -777.70           -589.75
Note:
(1) Parentheses are standard errors. ** and * indicate 1% and 5% significance level, respectively.
(2) p. is the second intercept defining the threshold for the transition from LFDI equal to 1 to 2.



Table 3: Model predictdictions: "hits and misses"
A: [lfdi=f(past, rival, past*rival, and firm dummies)]
predict=2      predict=l       predict=O         total
observe=2             80              36             33              149
(.53)
observe=1             38              83             93              214
F(.39)
observe=0             8               38             588             634
I     (.93)    I
total           126             157            714             997
1     (.75)    l
B: [Ifdi=f(past, rival, past*rival, firm dummies, and country dummies)]
predict=2      predict=1       predict=O         total
observe=2             88             45              16              149
(.59)
observe=1             34             108             72              214
F     (.50)    1
observe=0             4               39             591             634
F     (.93)    I
total           126             192            679             997
I     (.79)    I
Note:
In parentheses are the percentage of observations that are correctly predicted. For example
in the first panel for predict=2, 80 out 149 (53%) of the observations are predicted correctly.
For the model with only firm dummies (80+83+588) out of 997 or 75% are correctly predicted.



Table 4: Do firm attributes correlate with private information?
Dependent variable=LFDI(likelihood of FDI)
[1]                        [2]
intercept                                -2.59**                     -2.65**
(0.13)                     (0.27)
past                                      2.23**                     2.04**
(0.31)                     (0.32)
rival                                     1.49**                     1.49**
(0.20)                     (0.20)
past*rival                                O0.90*                     -0.80*
(0.42)                     (0.43)
size                                                                0.00001 *
(8.753E-6)
R&D                                                                   0.04
(0.07)
export propensity                                                    -1.10*
(0.56)
P.                                         1.54                       1.56
n                                         840                         840
logliklihood                             -678.68                     -676.11
Note:
(1) Parentheses are standard errors. ** and * indicate 1 % and 5% significance level, respectively.
(2) P is the second intercept defining the threshold for the transition from LFDI equal to I to 2.



Table 5. Private information and firm size
[1]               [2]               [3]
intercept                         -2.23**            -2.21**           -2.18**
(0.14)            (0.15)            (0.15)
past                               1.66**            2.29**            1.75**
(0.36)            (0.29)            (0.36)
rival                              1.54**            1.28**             1.33**
(0.19)            (0.24)            (0.25)
past*rival                         -0.68*            -1.00*             -0.65
(0.41)            (0.39)            (0.41)
stratal                           -1.25**            -0.93**           -1.22**
(0.24)            (0.23)            (0.25)
strata2                           -0.63**            -0.76**           -0.77**
(0.16)            (0.18)            (0.18)
past*stratal                       2.68**                              2.62**
(0.68)                              (0.68)
past*strata2                        0.33                                0.11
(0.42)                              (0.44)
rival*stratal                                         0.17              -0.21
(0.51)           (0.57)
rival*strata2                                        0.68*              0.62*
(0.34)            (0.36)
p..                                 1.55              1.53              1.56
n                                   997               '997              997
logliklihood                      -756.69            -763.86           -754.83
Note:
(1) Parentheses are standard errors. ** and * indicate 1% and 5 % significance level, respectively.
(2) p. is the second intercept defining the threshold for the transition from LFDI equal to 1 to 2.
(3) Stratal=small firms, strata2=medium-sized firms, strata3=large firms: strata3 was used as the base.



Table 6. Public information is complementary to private information
Dependent variable=LFDI(likellhood of FDI)
111i                            [21                                   [31
intercept           -2.47**       intercept          -6.07**       intercept               -2.43**
(0.20)                          (0.42)                                (0.32)
past                 2.27**       past               2.87**        past                     2.02**
(0.29)                          (0.41)                                (0.33)
rival                1.59**       rival               1.84**       rival                    1.46**
(0.19)                          (0.29)                                (0.21)
past*rival           -0.93*       past*-rival        -1.45**       past*rival              -0.84**
(0.40)                          (0.53)                                (0.43)
China                0.94**       China               1.85**       Chiina                   0.88**
(0.24)                          (0.31)                                (0.26)
India               -I.17**       India              -1.89**       India                   -1.28**
(0.29)                          (0.37)                                (0.31)
Indonesia            -0.17        Indonesia           0.02         Indonesia                -0.24
(0.25)                          (0.31)                                (0.27)
Malaysia             -0.47*       Malaysia            -0.55        Malaysia                 -0.41
(0.26)                          (0.34)                                (0.27)
Philippines         -0.94**       Philippines        -1.24**       Philippinies            -1.04**
(0.28)                          (0.34)                                (0.29)
Thailand             -0.42        Thailand            -0.20        Thailand                 -0.40
(0.26)                          (0.33)                                (0.28)
size                   0.00002*
(9. 1OE-6)
export propensity        -1.13
(0.58)
R&D                      0.04
(0.08)
Firm dummies          No                               Yes                                   No
Country dummies       Yes                              Yes                                   Yes
IL                  1.61           .                2.46          tp                       1.67
n                   997            n                 997           n                       840
logliklihood        -736.78       loglikelihood      -515.29       loglikelihood           -637.98
Note:
(1) Parentheses are standard errors. ** and * indicate 1% and 5% significance level, respectively.
(2) p is the second intercept defining the threshiold for the transition from LFDI equal to I to 2.   (3) Vietniam was used as a hase.



Table 7: Country-specific value of private information
Dependent variable=LFDI(likelihood of FDI)
China          India          Indo            Mal            Phil           Thai           Viet
intercept             -1.25**        -4.37**        -2.73**        -3.24**        -3.64**        -2.45**        -2.89**
(0.24)         (0.63)         (0.33)         (0.38)         (0.49)         (0.33)         (0.37)
past                  2.60**           2.01          2.13**         3.00**         2.66*           1.37*        -2.7E-16
(0.82)         (1.35)         (0.66)         (0.66)         (1.16)         (0.57)         (1.45)
rival                  1.24**        2.38**          1.75**         1.82**          0.82           1.02*         3.10**
(0.42)         (0.70)         (0.47)         (0.49)         (0.58)         (0.46)         (0.73)
past*rival             -0.98          24.17          -1.02          -1.95*          0.94           0.05            0
(1.09)       (127372)         (0.91)         (0.85)         (1.74)         (0.79)          (0)
[1              ~~~~     ~~1.22  2.19          1.71           1.89           1.95            1.45           1.96
n                       141            152            143            141            i45            133            142
logliklihood          -133.32        -61.46         -117.83        -108.80         -79.37        -116.69         -106.01
Note:
(1) Parentheses are standard errors. ** and * indicate I % and 5 % significance level, respectively.
(2) p1 is the second intercept defining the threshold for the transition from LFDI equal to 1 to 2.



Table 8. The role of FDI policy and labor costs
Dependent variable=LFDI (likelihood of FDI)
[1]       [2]        [3]       [4]                          [5]       [6]        [7]       [8]
intercept        -2.82**    -3.58**   -4.05**   -4.16**      intercept       -2.75**   -5.81**   -5.60**    -4.77**
(0.33)    (0.52)    (0.56)    (0.58)                        (0.26)    (0.67)    (0.72)      (0.71)
past             2.31**     2.01**    4.11**    1.85**       past             2.14**    2.51**      1.71     2.41**
(0.41)     (0.44)    (1.00)    (0.45)                       (0.38)    (0.40)    (1.16)      (0.40)
rival            1.87**     1.69**    1.60**    3.99**      rival             1.73**    1.98**    1.96**      -1.55
(0.34)     (0.35)    (0.35)    (0.98)                       (0.32)    (0.33)    (0.33)      (1.13)
pa_ri            -0.99*     -0.86      -0.49     -0.35       pa_ri           -0.88*    -1.34*    -1.37*      - 1.32*
(0.57)    (0.58)    (0.59)    (0.61)                        (0.53)    (0.55)    (0.55)      (0.55)
FDlplcy                     0.17*     0.27**    0.31**      laborcost                   0.39**    0.37**     0.25**
(0.08)    (0.09)    (0.10)                                 (0.07)    (0.08)      (0.08)
FDlplcy*past                          -0.38*                 Icost*past                            0.11
(0.16)                                                     (0.15)
FDIplcy*rival                                    -0.44*     lcost*rival                                       0.50*
(0.17)                                                      (0.15)
p.                2.03       2.05       2.07      2.06       p.                1.88      1.96       1.97      2.01
firm dummies       yes        yes       yes       yes       firm dummies       yes       yes        yes        yes
n                 427        427        427       427        n                 480       480        480       480
logliklihood     -336.02    -334.12   -331.36   -330.88      logliklihood    -385.28   -370.71   -370.44    -365.17
Note:
(1) Parentheses are standard errors. ** and * indicate 1% and 5% significance level respectively.
(2) IL is the second intercept defining the threshold for the transition from LFDI equal to 1 to 2.
(3) 67 firm dummies were included in the regressions.









Policy Research Working Paper Series
Contact
Title                            Author                   Date              for paper
WPS1719 Shifting Responsibility for Social  David Seweil           February 1997       C. Lawrence
Services As Enterprises Privatize in                                         32216
Belarus
WPS1720 The Distribution of Foreign Direct    Harry G. Broadman    February 1997      J. Grigsby
Investment in China              Xiaolun Sun                                 82423
WPS1721 EU Accession of Central and        Luca Barbone            February 1997       L. Barbone
Eastern Europe. Bridging the     Juan Zalduendo                              32556
Income Gap
WPS1722 Uncertainty, Instability, and      Luis Serven             February 1997       E. Khine
Irreversible Investmnent Theory,                                             37471
Evidence, and Lessons for Africa
WPS1723 The Productivity Effects of        Lixin Colin Xu          February 1997      P. Sintim-Aboagye
of Decentralized Reforms: An                                                 38526
Analysis of the Chinese
industrial Reforms
WPS1724 Debt Maturity and Firm Performance Fabio Schiantarelli     February 1997       B. Moore
A Panel Study of Indian Companies  Vivek Srivastava                          38526
WPS1725 Access to Long-Term Debt and       Fidel Jaramillo         February 1997      B. Moore
Effects on Firms' Performance    Fabio Schiantarelli                         38526
Lessons form Ecuador
WPS1 126 Roads, Population Pressures, and    Maureen Cropper       February 1997      A. Maranon
Deforestation in Thailand,       Charles Griffiths                           39074
1976-89                          Muthukumara Mani
WPS1727 The Economics ot the Informal      Norman A. Loayza        February 1997      R. Martin
Sector: A Simple Model and Some                                              39026
Empirical Evidence from L atin
America
WPS1728 Regional Labor Markets during      Chris Manning           February 1997      J. Israel
Deregulation in Indonesia Have the                                           85117
Outer Islands Been Left Behind?
WPS1729 Does Mercosur's Trade Performance Alexander Yeats          February 1997      J. Ngaine
Raise Concerns about the Effects                                             37947
of Regional Trade Arrangements?
WPS1730 Have Trade Policy Reforms Led      Shuby Andriamananjara   February 1997      J. Ngaine
to Greater Openness in Developing   John Nash                                37959
Countries?
WPS1731 Pension Reform, Growth, and the    Michelle Riboud         February 1997      R. Hablero
Labor Market in Ukraine          Hoaquan Chu                                 33971



Policy Research Working Paper Series
Contact
Title                           Author                  Date              for paper
WPS1732 Agricultural Trade and Rural      Dean A. DeRosa          February 1997     J. Ngaine
Development in the Middle East                                             37959
and North Africa: Recent Developments
and Prospects
WPS1733 The Usefulness of Private and Public Yuko Konoshita       February 1997     R. Reff
Information for Foreign Investment  Ashoka Mody                            34815
Decisions