Content uploaded by Kuo-Chun Yeh
Author content
All content in this area was uploaded by Kuo-Chun Yeh on Dec 17, 2020
Content may be subject to copyright.
ORIGINAL MANUSCRIPT
Determinants of Taiwan's outward investment
in Southeast Asia: Economic factors
and institutional issues
Thu-Ha Thi An
1,2
| Kuo-Chun Yeh
1
1
Graduate Institute of National
Development, National Taiwan
University, Taipei, Taiwan
2
Faculty of Finance, Banking Academy,
Hanoi, Vietnam
Correspondence
Thu-Ha Thi An, Graduate Institute of
National Development, National Taiwan
University, No.1, Roosevelt Road,
Section 4, Taipei 10617, Taiwan.
Email: d02341008@ntu.edu.tw
Funding information
National Taiwan University; Ministry of
Science and Technology, Taiwan
Abstract
This paper investigates the economic and institutional
determinants of Taiwan's outward direct investment in
six Southeast Asian countries from 1998 to 2017, apply-
ing the panel ARDL–Pooled Mean Group estimation.
We specially examine the effects of institutional quality
with five dimensions inclusively, using the Worldwide
Governance Indicators. The results show the locational
economic factors are the primary determinants in the
long run. The tight and historic trading relation with
Southeast Asia has a long-run positive effect. On the
contrary, the institutional quality of the host countries
has strong positive effects in both the long run and short
run. Further, the paper displays the dynamics of this
investment in the last 20 years. These results are impor-
tant for Taiwan and Southeast Asia policy-makers in set-
ting up the short-run and long-run policies to sustain
their diversified economic growth.
1|INTRODUCTION
As a fast-growing area of the world, the economic potential of Southeast Asia has been targeted
by several advanced economies including Taiwan. All the same, Taiwan is not the only one
with fast-growing trade and investment relationships with ASEAN. In addition to Taiwan's
‘New Southbound Policy’, China's ‘Belt and Road Initiative (BRI)’, South Korea's ‘New South-
ern Policy’and the forthcoming EU–Vietnam free trade agreement (EUVFTA) aim to improve
economic ties with Southeast Asian countries.
In this competitive context, the most significant difference between Taiwan and other
advanced economies is its extremely high economic dependence on China, which has made
Received: 10 September 2019 Revised: 1 September 2020 Accepted: 2 September 2020
DOI: 10.1111/1468-0106.12348
Pac Econ Rev. 2020;1–38. wileyonlinelibrary.com/journal/paer ©2020 John Wiley & Sons Australia, Ltd 1
diversification a national security issue. Due to China's strong economic gravity, Taiwan's bilat-
eral trade, investment, the brain drain and the security situation might have worsened over the
past decade. In 2019, Taiwan's export dependence ratio on China was 40.2%, which is much
higher than most advanced economies including South Korea (25%) and Singapore (13%). Tai-
wan's trade balances would be negative without China. Figure 1a and b show the Chinese mar-
ket attracts almost 60% of Taiwan's outward FDI in 1991–2019, and Taiwan's industries from
electronics to the financial sectors have accumulated at a huge scale in China. However, it is
known the figures are widely perceived to be an underestimation due to official regulations and
investment via tax havens (Yeh and Ho, 2012).
Meanwhile, Southeast Asia has long been a favourable location for Taiwanese investors.
Figure 2 indicates the share of Southeast Asia in terms of total outward FDI of Taiwan is rela-
tively large and recently increasing. Additionally, the annual growth rate of Taiwan's invest-
ment in Southeast Asia is upwards, while for China it is decreasing. Taiwan is attempting to
diversify its global investment, in which Southeast Asia is the best potential choice considering
traditional economic gravity and political security. ASEAN–6 has been the second-largest export
destination of Taiwan. Taiwan's investments in Vietnam, Indonesia, Thailand, and Singapore
have made them hubs to enter into the markets of ASEAN and China, which in turn makes
risk-sharing feasible.
(a)
Taiwan’s outward FDI in major recipients (1991-2019)
(b)
Pro
p
ortions of Taiwan’s industries in China (1991-2019)
56.6%
2.3%
13.9%
1.1%
0.8%
2.8%
3.3%
0.5%
1.2% 4.3%
5.0%
1.1%
7.1% Mainland China
H.K.
British Carribean
Bermuda
Samoa
Japan
Vietnam
Indonesia
Thailand
Singapre
U.S.
Netherlands
Others
7.4%
13.5%
5.0%
18.3%
7.5%
6.1%
42.2%
Wholesale and Retail
Computers,
Electronic & Optical
Chemical Material
Electronic Parts &
Components
Financial & Insurance
Electrical Equipment
Others
FIGURE 1 Taiwan's outward
investment in major recipients and
China. Source: Statistics by
Mainland Affairs Council (2020).
Authors' compilation [Color figure
can be viewed at
wileyonlinelibrary.com]
2AN AND YEH
Taiwan's diversification implies the necessity of updating data and modelling according to
the new policy needs. Numerous papers have explored the cross-strait investment relationship
with China, yet few quantitative studies focus on Taiwan's FDI in Southeast Asia. Most of these
few studies examine Taiwan's firm-specific factors and firms' decisions of investing overseas in
general or by comparison among regions of destination. Chen and Chen (1998a, 1998b) analyze
different strategic asset linkages of Taiwanese firms in Southeast Asia, the USA and China. Lei
and Chen (2011) study the location choice behaviour of Taiwanese firms investing in China
and Vietnam. Chiu and Lo (2015) examine the impact of the firm's factors on Taiwanese enter-
prises in the choice of entry mode in Southeast Asia. Liu (2011) and Lo and Lin (2015) evaluate
the location's advantages and subsidiary's performance.
Some recent papers (mostly in traditional Chinese) have analyzed Taiwan's small and medium
enterprises (SMEs) investing in Southeast Asia by industry and service sectors. Lee and Liu (2018)
present the investment of Taiwan's banks in ASEAN. Lee, Lin, and Yang (2018) provide initial quan-
titative evidence to compare strategies of Taiwan's SMEs in ASEAN with those of China's state-
owned enterprises (SOEs). The cost factors (e.g., real interest rates and exchange rates in the host
countries) still matter for Taiwan's SMEs. In contrast, some considerations related to competitiveness
and economic potential (e.g., GDP, GDP per capita, global competition index rankings) are statisti-
cally significant in Chinese SOEs' investment in ASEAN economies. Tan and Lin (2018) describe
Taiwan's industrial layout according to the differences among the ASEAN's economic structures.
To the best of our knowledge, there is no empirical analysis focusing on determinants of Tai-
wan's aggregate outward investment in Southeast Asian countries regarding locational factors,
economic factors and institutional-political factors. This study aims to fill this gap in empirical
Shares of Taiwan's outward FDI globally
(a)
(b)
Annual
g
rowth rate of Taiwan's outward FDI in Southeast Asia and China
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
Southeast Asia Other Areas Mainland China
-1.5
-0.5
0.5
1.5
2.5
3.5
4.5
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
Southeast Asia Mainland China
Linear Trend (Southeast Asia) Linear Trend (Mainland China)
FIGURE 2 Taiwan's outward
investment in Southeast Asia and
China. Source: Statistics by Overseas
Chinese and Foreign Investment
Commission, Taiwan Ministry of
Economic Affairs (MOEAIC).
Authors' compilation [Color figure
can be viewed at
wileyonlinelibrary.com]
AN AND YEH 3
research by proposing a comprehensive model on the determinants of Taiwan's FDI in Southeast
Asia from 1998 to 2017. We study such questions as: what are the drivers of Taiwan's investment
in Southeast Asia (six host countries: Singapore, Indonesia, Malaysia, the Philippines, Thailand
and Vietnam
1
); how the investment decisions of Taiwanese firms are influenced by specific
institutional-political factors; and how Taiwan's FDI to the region has been changing over the
recent 20 years. Thus, the paper is expected to contribute to the extant literature as follows: (a) fill
the research gap with empirical research for the Taiwan-ASEAN case; (b) comprise both tradi-
tional economic factors and political issues to provide new insight into the determinants to
improve the Taiwan-ASEAN links; and (c) offer a reference for Taiwan and the host countries to
revise their outward investing and attraction of foreign investment policies.
The remainder of the study is organized as follows. Section 2 gives an overview of Taiwan's
outward FDI in Southeast Asia since the early 1980s. Section 3 presents the background of the
study, based on which the hypotheses are developed. The methodology and data for analysis
are given in Section 4. Section 5 presents and interprets the empirical results. The study con-
cludes in Section 6 with a discussion on the issues of Taiwan's outward FDI in Southeast Asia
and some implications for both Taiwan and the Southeast Asian host countries.
2|OVERVIEW ON TAIWAN'S OUTWARD FDI IN
SOUTHEAST ASIA
Taiwanese companies began investing in Southeast Asia from 1959. By the end of the 1980s,
Taiwan's investment in the region accounted for roughly 95% of the total estimated Taiwanese
investment in Asia
2
(i.e., Asian countries excluding China). In the early 1990s, Taiwan became
a major source of investment in Southeast Asia: ranked first in Vietnam, second in Malaysia
(behind Japan), third in Indonesia (behind Japan and Hong Kong), fourth in Thailand (behind
Japan, Hong Kong and the USA), fifth in the Philippines and thirteenth in Singapore. After the
initial ‘Go South’policy in 1993, investment in this region gradually increased until a slight
drop in 1998 due to the Asian financial crisis. Though fluctuating during 1998–2017, Taiwan's
investment in the six Southeast Asian countries significantly increased in the latter 10 years. In
2017, the approved amount of Taiwan's investment in ASEAN–6 was US$2.8 billion, occupying
78.9% of Taiwan's total investment to Asia. The accumulative investment stock of Taiwan in the
region reached US$32.6 billion by the end of 2017.
3
Taiwan maintained its significant share of
foreign direct investment in Southeast Asia: ranked third in Thailand, fourth in Vietnam and
Indonesia. Taiwan's FDI to Southeast Asia presents an exceptional case of offshore investing,
that is, SMEs are dominant in this activity. A series of studies looked into Taiwanese investor
behaviour and the nature of the investments. Chu, Ye, and Hsu (1999) and Chen (1998) rev-
ealed that by the early 1990s, most Taiwanese firms investing in Southeast Asia were small,
without technological superiority and organizational strength, with the top motivation being
the host country's lower wages. The second motivation was to expand the local and third-party
market, utilizing the export-oriented industrialization policy in the region. Chen and
Chen (1998a) pointed out that technology-sourcing and market-enhancing linkages are lacking
in Taiwan's FDI in Southeast Asia. Further, Chen and Chen (1998b) found Taiwanese investors
in Southeast Asia take advantage of relations to build linkages to local networks. These linkages
are mostly based on personal relations or business transactions that create trust and mutual
understandings, facilitating inter-firm cooperation. Relational linkages are a significant
determinant of the locational choice of Taiwan's outward FDI in Southeast Asia.
4AN AND YEH
Most of Taiwan's FDI in Southeast Asia is in manufacturing, where electronic parts-
components manufacturing and basic metal manufacturing are predominant. These industries
utilize labour, huge amounts of land for factories and aim to export to third markets. Service
sectors, moreover, are recently becoming the top invested sectors of Taiwanese enterprises. In
the past decade, the largest amount and the highest number of cases of Taiwanese investing in
Southeast Asia are in finance-insurance services and wholesale-retail trade (Table A1). On the
contrary, there is a decrease in investment in those industries that acquire natural resources
such as mining-quarrying, wood and bamboo products, or labour-intensive sectors such as tex-
tiles and food processing. Taiwanese enterprises in Southeast Asia are changing to exploit the
local markets and expand the domestic market other than export to third markets. This is due
to the increasing spending power and the rapid growth of Southeast Asian economies. From
the perspective of investment motivation, Taiwanese businesses are shifting from cost-targeted
investment strategies to expansive market-targeted strategies. In other words, they are shifting
from processing and manufacturing operations to developing their own brand and distribution
operations. This is also a focus of the New Southbound Policy, where the key goal is to enhance
economic and trade cooperation with Southeast Asia. Thus, the economic and investing rela-
tionship between Taiwan and Southeast Asian countries is promising and currently receiving
much support from the governments.
3|RESEARCH BACKGROUND AND HYPOTHESIS
DEVELOPMENT
3.1 |Research background
FDI clearly plays an important role in the economic growth and development of a country.
Many studies, both theoretical and empirical, have been conducted to explain the causes of FDI
from different perspectives. Several major theories on FDI have been put forward, such as the
product-cycle theory (Vernon, 1966), imperfect market condition theory (Hymer, 1976;
Agarwal, 1980) and internalization theory (Buckley and Cassion, 1976; Hennart, 1982). The
eclectic paradigm (Dunning, 1973, 1977, 1980) is possibly the most comprehensive in the theo-
retical series on FDI. Dunning's OLI paradigm is a framework with three pillars: ownership
advantages, internalization advantages and location advantages. The OLI framework covers
various determinants of FDI, including firm-specific factors (ownership advantages such as
monopoly, technology, economies of scale and the internalization advantage) and location-
specific factors (economic factors, political conditions, cultural and sociological characteristics
of the host countries). Dunning (2009) identifies four major motives of outward FDI: natural
resource-seeking, market-seeking, efficiency-seeking and strategic asset-seeking. The eclectic
paradigm is not only an inclusive but also an empirical and open-ended framework. By com-
prising a range of determinants of the firm, home and host countries, the OLI framework offers
an immense, extendable and flexible context for empirical research.
This study follows Dunning's OLI framework to develop an inclusive model for Taiwan's
outward FDI in Southeast Asia, examining both the economic factors and institutional-political
factors. There are various extant studies exploring these factors of outward FDI. Researchers
commonly argue that locational determinants significantly influence the attraction of FDI
inflows, including a large and growing market (Aykut & Ratha, 2004; Buckley et al., 2007),
abundant raw materials (Buckley et al., 2007; Ramasamy, Yeung, & Laforet, 2012), cheap
AN AND YEH 5
labour cost, low operation cost (Duanmu & Guney, 2009), openness, economic-political-social
factors, institutions and incentives of the host country (Buckley et al., 2007; Dunning &
Lundan, 2008; Kang & Jiang, 2012; Tintin, 2013; Kayalvizhi & Thenmozhi, 2018).
3.2 |Determinants of Taiwan's FDI in Southeast Asia: Hypothesis
development
Since numerous theoretical and empirical studies have suggested FDI responds to a wide range
of stimuli, in this paper we include in our model the most widely acknowledged and specific
factors to keep the analysis valid and representative of Taiwan's FDI in Southeast Asia. The fol-
lowing subsections provide the hypotheses with theoretical and empirical justification for their
inclusion.
3.2.1 |Market size and market openness
Market size and market openness of the host country are the most widely acknowledged rea-
sons for FDI. Numerous empirical studies proved the strong positive relationship between the
market size of the host and FDI inflows. A large market size is attractive to foreign firms in var-
ious ways. First, a large market size is likely to offer more opportunity for foreign firms to
obtain cost effectiveness and economies of scale of domestic production, because the firms can
sell their products to the host country without cross-border transportation costs and the barriers
of tariff and import control. Second, a large market size might sustain the possibility of market
growth, which is a guarantee of the long-term benefit for investors. Third, market-seeking firms
not only aim to exploit the host country market but also target nearby or third markets, where
the host country advantages help reduce restrictions on the tariff and non-tariff trade barriers.
In this sense, the more open the host economy is, the more attractive to market-seeking firms
the location will be. Previous research indicated one main reason for Taiwan's FDI in ASEAN
is to facilitate export-oriented manufacturing to the third markets by taking advantage of the
host countries to penetrate tighter Western markets (Chang & Thornson, 1994; Chow, 1996).
Market-seeking remains the first motivation of Taiwanese investors in the region up to now
(Makino, Lau, & Yeh, 2002; Chow, 2016). Therefore, we suggest:
HYPOTHESIS 1. Taiwan's FDI in Southeast Asia is positively associated with the market size of
the host country.
HYPOTHESIS 2. Taiwan's FDI in Southeast Asia is positively associated with the openness of the
host country.
3.2.2 |Cost factors
Cost efficiency has always been a major cause for firms to do business abroad when the domes-
tic costs of production factors are rising. This was the case for Taiwan after the mid-1980s when
domestic wages doubled, and the decreasing birth rate added to the shortage of labour and fur-
ther drove up wages. Throughout these two decades, most Taiwanese FDI in Southeast Asia
was in labour-intensive manufacturing industries. Previous empirical studies showed lower rel-
ative labour costs in China and Southeast Asia were the major attraction to Taiwan's investors
6AN AND YEH
who are both cost and market oriented (Chen, Rau, & Lin, 2006). Moreover, a notable change
in the landscape of FDI in Asia in recent years is the moving of labour-intensive manufacturing
activities from higher cost locations to ASEAN and other Asian countries. Facing the issue of
an ageing population and rising domestic labour costs, while China is losing the advantage of
lower wages, Taiwanese investors have tended to move to Southeast Asia. Labour cost is a suffi-
ciently significant factor in Taiwan's outward FDI decision.
Financing cost, or the cost of capital, proxied by the real interest rate, is widely considered
to be a determinant of FDI. The impacts of real interest rate on FDI vary, depending on
whether it originates in the home or the host country. Borrowings in the host country at a rela-
tively lower interest rate might be a point for cost-oriented FDI. Though the influential signs of
the host interest rate on FDI inflows are diverse or statistically insignificant in some previous
empirics, it is commonly agreed a lower real interest rate in the host country stimulates inward
FDI. In other words, the increase in the host interest rate has a negative effect on the FDI
inflows. There is empirical evidence for the host developing countries like Southeast Asia
(Hoang, 2012; Cavallari & d'Addona, 2013).
We suggest relatively lower labour costs and favourable real interest rates in Southeast Asia
act as the points for Taiwanese cost-efficiency seekers. As a result, we propose the hypotheses
of cost-efficiency motivation:
HYPOTHESIS 3. Taiwan's FDI in Southeast Asia is negatively associated with the labour cost in
the host countries.
HYPOTHESIS 4. Taiwan's FDI in Southeast Asia is negatively associated with the host's real interest rate.
3.2.3 |Bilateral trading relationship with Taiwan
The relationship of foreign trade and foreign direct investment between the home and host
countries has received much attention from scholars in the past two decades. This relationship
can be substitutes or complements depending on whether it is in the long run or short run and
on each case of international business. In general, cross-border trade is traditionally viewed to
generate direct investment. A prior trading relationship between the home and host country
can increase the follow-up FDI, and efforts to liberalize trade can ease the barriers to FDI. It is
a current trend that more countries are involved in new generation bilateral and regional free
trade agreements, including both trading and investing relations. On one hand, freer trade can
accelerate FDI by stimulating input-imports from the home country and product-exports from
the host country, especially for export-oriented and market-seeking oriented FDI. On the other
hand, bilateral trade represents the intensity of transactional deals between the two countries,
which might help the new investors have the cognition and mindset to invest in the destination.
In practice, Taiwan is a top ranked investor in ASEAN and also a significant trading partner to
the region. Thus, we suggest:
HYPOTHESIS 5. Taiwan's FDI in Southeast Asia is positively associated with the bilateral trade
with the host countries.
3.2.4 |Institutional-political quality
The effect of institutional quality on FDI attractiveness has been explored for a long time. From
an economic perspective, institutional quality is an important driver of FDI because the related
AN AND YEH 7
factors influence the cost-related issues by which foreign firms choose one host country over
another. Supporting institutional factors certainly helps to reduce costs for firms. Host countries
with market supporting institutions and policies are more successful in attracting foreign inves-
tors. In addition, investors generally prefer the host governments who ensure a stable political
environment, reliable and predictable policies, and adequate supporting infrastructure. Further,
governments that provide an efficient and transparent mechanism for firms, encourage compe-
tition, allow foreign firms to exploit their ownership advantages, and protect intellectual prop-
erty will be favourable to foreign firms. In short, good governance quality is a basic guarantee
for foreign firms to run their cross-border economic activities effectively and profitably in the
long run. Thus, we suggest the hypotheses on institutional quality as follows:
HYPOTHESIS 6. Taiwan's FDI in Southeast Asia is positively associated with the governance qual-
ity of the host countries.
HYPOTHESIS 7. Taiwan's FDI in Southeast Asia is negatively associated with the gap in gover-
nance quality between Taiwan and the host countries.
4|METHODOLOGY AND DATA
4.1 |Empirical model and data
To investigate the determinants of outward FDI from Taiwan in Southeast Asia in 1998–2017,
we study the following empirical model based on the OLI framework and previous literature in
this line:
4
FDIit =β0+β1GDPit +β2OPENit +β3TRADEit+β4LABORit +β5INT it +β7INSit +β8Xit
+μi+ρt+εit
ð1Þ
The subscripts iand tare presented for countries and years, respectively. FDI is Taiwan's
outward direct investment in Southeast Asian countries. GDP is the real gross domestic product
of the host countries. OPEN is the trade openness of the host countries. TRADE is the real bilat-
eral trade volume between Taiwan and the host countries. LABOR is the compensation per
employee in the host countries, proxy for the cost of labour. INT is the bank real lending inter-
est rate in the host countries that usually meets the short- and medium-term financing needs of
the private sectors, proxy for the cost of finance. INS is the institutional factor of the host coun-
tries including six components. Here, we take five components into our study, that is, control of
corruption –CC, government effectiveness –GE, political stability –PS, regulatory quality –RQ
and rule of law –RL (Table A2). X
it
is a vector of control variables. μ
i
and ρ
t
are country and
year fixed effects, respectively; βare parameters to be estimated; ε
it
is the residual.
We utilize aggregate data for Taiwan and six ASEAN countries for 20 years (1998–2017)
from the World Development Indicators (WDI) for GDP,GDPpc,OPEN,INT and LABOR (statis-
tics for cost of labour in Vietnam are from Vietnam General Statistics Office –GSO); Institu-
tional statistics, INS, from the Worldwide Governance Indicators (WGI); FDI and TRADE from
annual statistics by Overseas Chinese and Foreign Investment Commission, Taiwan Ministry of
Economic Affairs (MOEAIC) and Custom Administration, Taiwan Ministry of Finance. FDI,
GDP and TRADE in the raw data are in current US$ million; GDPpc and LABOR in current US
$. OPEN in the raw data is the ratio of trade over GDP. All these variables in the model are in
8AN AND YEH
natural logarithmic transformation to stabilize data variability and for efficiency and consis-
tency. INS are aggregate governance indicators of the host countries reported in standard nor-
mal units, ranging from −2.5 to 2.5 with higher values referring to better governance. Table 1
presents the definitions of variables and sources of the data. The descriptive statistics of the data
are in Table 2.
4.2 |Methodology
Most of the previous papers discussing the impacts of economic and institutional factors on FDI
employed the standard panel regression methods, such as pooled OLS, fixed effects and random
effects estimations. A survey of the methodology, including the determinants and major conclu-
sions from other comparable research, is presented in Table A3. The common limitation in
methodology of the extant literature lies in their heavy reliance on the static panel approach,
which might not fully capture the dynamic nature of the issue. Moreover, the standard methods
might not work for data of small sample size and provide biased estimates in cases of endoge-
nous, serially correlated and non-stationary variables. In pooled OLS, a common intercept and
slope coefficients for all cross-sections are restricted disregarding individual heterogeneity.
Meanwhile, the fixed effects model assumes the common slope and variance but country-
specific intercepts. However, the fixed effects estimates are biased when some regressors are
endogenous and correlated with the error terms. In addition, non-stationarity in pooled time
series cross-sectional data has the probability to invalidate such traditional approaches. Recent
literature on panel regression points out the possible spurious regression and potential bias of
standard estimators when the parameters are heterogeneous across sections, or when the
regressors are serially correlated and non-stationarity as Tis large (Im, Pesaran, & Shin, 2003).
This study utilizes a heterogeneous panel of a small number of cross-sections (N= 6) and
relatively large number of periods (T= 20); the time series are not all stationary but integrated
of order one I(1), serially correlated among regressors and group-wise heteroskedasticity.
5
To
study the impacts of economic and institutional factors on outward FDI of Taiwan in Southeast
Asia over 20 years considering these features of variables and data, we adopt the dynamic
approach with a newly developed methodology, that is, the panel autoregressive distributed lag
(ARDL) framework. Specifically, we follow Pesaran, Shin, and Smith (1999, 2001) to perform
the Pooled Mean Group (PMG) estimation.
The panel ARDL–PMG estimation method was originally introduced by Pesaran et al. (1999).
This method was developed for heterogeneous panels under the assumption the intercepts,
short-run coefficients and error variances differ across sections but the long-run coefficients are
homogeneous. Accordingly, the PMG estimator is appropriate for the long panel with relatively
large Tand fixed N. The cross-sections differ in certain aspects yet there are reasons to expect
some similarity in the long run, such as a common economic development level, in the same
economic region and technology level, or having similar economic conditions and risks. In
addition, the ARDL approach provides consistent and efficient estimators because it eliminates
endogeneity problems by adding lag lengths in both endogenous and exogenous variables. This
paper selects PMG estimation because it fits the features of the research's data and meets the
aim of the study.
To justify this choice, we first perform stationarity tests of the underlying variables. The
ARDL modelling can be specified as an error correction model when the variables are inte-
grated of order one I(1), or mixture of I(1) and I(0). This method cannot be used if the variables
AN AND YEH 9
are integrated of order two I(2) or above. Then, we test the existence of the long-run relation-
ship between the dependent and independent variables according to Pesaran et al. (2001). Last,
we estimate the model using the PMG technique.
Applying the method by Pesaran et al. (1999) on time periods t=1,2,…,T, and groups
i=1,2,…,N, the ARDL (p,q,q…,q) structure for the empirical model equation 1 is as below:
FDIit =X
p
j=1
λijFDIi,t−j+X
q
j=0
δ0
ijxi,t−j+μi+εit ð2Þ
where x
it
(k×1) is the vector of regressors, λ
ij
are scalars, δ
ij
(k×1) is the coefficient vector, μ
i
is the fixed effects, ε
it
is the error term.
Equation 2 can be reparameterized as follows:
ΔFDIit =φiFDI i,t−1+X
q
j=0
δ0
ijxi,t−1+X
p−1
j=1
λ
ijΔFDIi,t−j+X
q−1
j=0
δ0
ij Δxi,t−j+μi+εit ð3Þ
where
φi=−1−P
p
j=1
λij
!
;λ
ij =−P
p
m=j+1
λim j=1,2,…,p−1ðÞ;δ
ij =−P
q
m=j+1
δim j=1,2, …,q−1ðÞ;
operator Δdenotes the first difference.
Equation 3 can be rewritten as:
ΔFDIit =φiFDI i,t−1−θ0i−θ0
1ixi,t−1
+X
p−1
j=1
λ
ijΔFDIi,t−j+X
q−1
j=0
δ0
ij Δxi,t−j+εit ð4Þ
with θ
0i
=μ
i
/φ
i
;θ1i=−P
q
j=0
δij=φi:
In equation 4, the first term on the right-hand side (represented in the level variable) reflects
the long-run cointegrated relationship; while the terms in difference show the short-run
dynamics. The parameter φ
i
measures the speed of adjustment of dependent variable FDI
it
towards the long-run equilibrium following a change in x
it
.
To determine the existence of a long-run relationship between the dependent and
independent variables, we implement panel cointegration tests. The null hypothesis of the
cointegration test for equation 4 is H
0
:φ
i
= 0 (no cointegration) and the alternative
H
1
:φ
i
≠0 (cointegration or long-run relationship). If H
0
is rejected, or there is cointegration
between dependent and independent variables, the error correction model (ECM) is
specified as:
ΔFDIit =α0i+X
p−1
j=1
β1ijΔFDIi,t−j+X
q−1
j=0
β2ij Δxi,t−j+φiECTi,t−1+εit ð5Þ
where ECT
i,t−1
is the error correction term defined by the long-run relationship.
10 AN AND YEH
The parameters are obtained by PMG estimation, which is based on the maximum likeli-
hood procedure. As stated in Pesaran et al. (1999), the PMG estimator restricts only a subset of
the long-run coefficients to be common but allows differences across groups on the ECT and
the short-run coefficients. Thereby, the PMG estimator is designed as an intermediate lying in a
continuum between one extreme of heterogeneity (the mean group –MG estimator with no
restrictions) and the other extreme of homogeneity (the dynamic fixed effects –DFE estimator,
which restricts all the coefficients and error variances to be common). The PMG estimator
involves both pooling and averaging. Meanwhile, the MG estimation is based on time series
regressions for each group, then averaging the coefficients. Thus, a test on heterogeneity of the
coefficients is equivalent to a test on the null hypothesis of no difference between the MG and
PMG estimators. Pesaran et al. (1999) show if the null hypothesis of the long-run homogeneity
cannot be rejected by a Hausman (1978)-type test, though both the PMG and MG estimators
are consistent, the former is efficient while the latter is not. Besides, the PMG estimator is
proved to be much less sensitive to outliers in the data. Dropping the outliers from the sample
does not change the PMG estimates, but substantially changes the MG. In sum, under the null
hypothesis of long-run homogeneity, the PMG estimator is preferable to the MG in the sense
the PMG is efficient and robust to small Nand Tpanel with outliers.
Further, as for the dynamics of Taiwan's outward investment in Southeast Asia during
20 years, we examine the volatility of the dependent variable in the second 10 years and the first
10 years using the time dummy D2008 as an exogenous variable in the ECM model, following
Pesaran et al. (2001). D2008 = 1, over the period 2008–2017; 0 otherwise. The conditional ECM
model becomes:
ΔFDIit =α0i+X
p−1
j=1
β1ijΔFDIi,t−j+X
q−1
j=0
β2ij Δxi,t−j+β3iD2008t+φiECTi,t−1+εit ð6Þ
where β
3i
presents a shift in the intercept, or the magnitude and direction of change in ΔFDI in
the sub-period 2008–2017 from the sub-period 1998–2007, all things being equal.
5|EMPIRICAL RESULTS
5.1 |Panel unit root tests
Like the traditional ARDL approach, the panel ARDL–PMG estimation can be applied
irrespective of stationarity I(0) or integrated of order one I(1) of variables. However, it cannot
be implemented to I(2) or higher order series. For the validity of the PMG application, it is nec-
essary to test for unit roots in the panel data. We use Im, Pesaran and Shin's W-stat
(Im et al., 2003), the ADF-Fisher Chi-square, PP-Fisher Chi-square P-stat (Choi, 2001) for the
individual unit root test; Levin, Lin and Chu t-stat (Levin, Lin, & James Chu, 2002) for the com-
mon unit root test and the Pesaran panel unit root test Zt bar-stat (Pesaran, 2007). The results
of the unit root tests are reported in Table 3. All the variables are tested for stationarity. If the
null hypothesis cannot be rejected, the tests in the first difference are implemented. The results
show FDI,GDP,OPEN,TRADE,LABOR,INS are I(1) while INT is I(0), none of the variables
are integrated of the order above I(1). These results validate the choice of ARDL approach in
this study.
AN AND YEH 11
5.2 |Panel cointegration tests
To examine the long-run relationship between the dependent and independent variables in
equation 4, we do panel cointegration tests following Pedroni (2004) and Kao (1999). The
dependent variable is FDI, while the covariates are vectors of the underlying variables. The
choice of lag lengths is based on the literature and automatically confirmed by the Akaike
Information Criterion (AIC) and Schwarz Bayesian Criterion (SBC). We use three between-
dimension tests for group statistics and four within-dimension tests for panel statistics
(including weighted statistics). We perform cointegration tests for different long-run rela-
tionships. The results of these cointegration tests are given in Table 4. For the Pedroni tests,
the panel ADF and PP statistics for within-dimension, the group ADF and PP statistics for
between-dimension are dominantly significant at 1% for all the models. The ADF statistics in
the Kao tests for all the models have the significance level at 1%. That is to say, there is
strong evidence for the existence of a long-run relationship between FDI and the economic
and institutional determinants. It is valid to run the ECM model in equation 5 to analyze
these long-run effects.
TABLE 1 Definitions of variables and sources of data
Variable Definition and measurement Source
FDI Taiwan's outward direct investment in host country, US$ million
(natural logarithm)
MOEAIC
GDP Real GDP of host country, proxy for market size, US$ million
(natural logarithm)
WDI (World Bank)
GDPpc Real GDP per capita of host country, proxy for income level, US$
(natural logarithm)
WDI (World Bank)
OPEN Ratio of trade over GDP, proxy for openness of host country
(natural logarithm)
WDI (World Bank)
TRADE Real bilateral trade volume between Taiwan and host country, US$
million (natural logarithm)
Taiwan Custom
Administration
LABOR Compensation per employee (wage and salaried workers) of host
country, US$ (natural logarithm)
Author's calculation
from WDI data
INT Real bank lending interest rate (short- and medium-term) to private
sectors
WDI (World Bank)
INS Country governance composite index WGI (World Bank)
CC Control of corruption index WGI (World Bank)
GE Government efficiency index WGI (World Bank)
PS Political stability index WGI (World Bank)
RQ Regulation quality index WGI (World Bank)
RL Rule of law index WGI (World Bank)
D2008 Time dummy variable, D2008 = 1 over the years 2008–2017; 0
otherwise
Source: Authors' compilation.
12 AN AND YEH
5.3 |Panel ARDL–PMG estimation
For panel ARDL estimation, we set the optimal lag lengths (models 1–8) by automatic selection
based on the AIC and SBC criterion, that is, the maximum lag lengths for the dependent vari-
able and dynamic regressors are 1.
6
This selection of the lag length in the panel ARDL model
can also be guided by empirical and technical reasons when dealing with economic time series
data. As noted in Wooldridge (2012), for annual data, the number of lags is typically small –
1 or 2 lags, in order not to lose degrees of freedom. Additionally, for economic time series data,
successive lags tend to be highly correlated, increasing the likelihood of multicollinearity.
Indeed, in this study, multicollinearity occurs in lag 2 and lag 1 of the variables when we try
the max lag of 2. Moreover, the optimal lag length of 1 by the AIC and SBC criterion is com-
monly confirmed in the literature, which adopts the panel ARDL methodology for panels of
annual frequency data (Goh, Wong, & Yew, 2018; Tan et al., 2018).
The parameters in the ECM model in equation 5 can be obtained by PMG and MG estima-
tion. We perform both of these estimators, then choose the favourable one using the Hausman
test. Table 5 presents the chosen PMG estimator with the Hausman p-value indicating the
restriction of homogeneity in the long run cannot be rejected at the 1% significance level in all
the models. The validity of long-run homogeneity verifies our assumption of identical long-term
effects across Southeast Asian countries. Although the six countries in the study differ in eco-
nomic and institutional levels of development, they show much less variation in percentage
change of FDI from Taiwan and the volume of bilateral trade with Taiwan. The data shows the
between variation (standard deviation between cross-sections) is rather smaller than the within
variation (standard deviation within individuals) for the variables of FDI and bilateral trade.
Especially, the variation in FDI is mostly within individuals –time series, not between
TABLE 2 Descriptive statistics
Variable Obs. Mean Std. Dev. Min Max
FDI 120 448.907 1,188.589 4.670 11,818.330
GDP 120 244,322.000 208,617.200 27,209.600 1,015,539.000
GDPpc 120 9,221.386 14,733.790 346.827 57,714.300
OPEN 120 158.848 105.230 37.439 441.604
TRADE 120 9,797.350 5,532.762 1,556.290 28,912.640
LABOR 113 1,131.503 1,238.797 36.042 4,732.316
INT 116 8.943 4.740 4.334 32.154
INS 108 0.047 0.738 −0.943 1.615
CC 108 0.029 1.022 −1.176 2.326
GE 108 0.492 0.880 −0.609 2.437
PS 108 −0.248 0.979 −2.095 1.586
RQ 108 0.291 0.853 −0.796 2.261
RL 108 0.104 0.791 −0.914 1.825
Note: Values reported are the statistics of the variables in the raw data (as defined in Table 1). The variables in
the models are in natural logarithmic transformation.
Source: Authors' calculations.
AN AND YEH 13
TABLE 3 Panel unit root tests
Variables
ADF - Fisher Chi-Square P-stat PP - Fisher Chi-Square P-stat
Im, Pesaran and
Shin W-stat Levin, Lin & Chu t- stat
Pesaran CADF Zt-
bar - stat
No
Intercept Intercept
Intercept
and trend
No
Intercept Intercept
Intercept
and trend Intercept
Intercept
and trend
No
Intercept Intercept
Intercept
and trend Intercept
Intercept
and trend
Level
FDI 5.092 22.395 36.889 5.941 24.025 34.320 −1.853 −4.086 1.951 −1.592 −4.071 −0.270 −1.141
(0.955) (0.033) (0.000) (0.919) (0.020) (0.000) (0.032) (0.000) (0.975) (0.056) (0.000) (0.394) (0.127)
GDP 0.042 6.853 5.041 0.003 6.856 2.266 1.294 1.646 10.185 −1.673 1.561 0.970 1.186
(1.000) (0.867) (0.957) (1.000) (0.867) (0.999) (0.902) (0.950) (1.000) (0.047) (0.941) (0.834) (0.882)
OPEN 19.623 9.112 13.948 23.723 14.487 33.460 0.669 −0.589 −0.296 −0.412 −1.226 2.948 3.651
(0.075) (0.693) (0.304) (0.022) (0.271) (0.001) (0.748) (0.278) (0.384) (0.340) (0.110) (0.998) (1.000)
TRADE 0.354 11.501 17.125 0.189 18.259 17.896 −0.638 −0.754 4.645 −3.402 −1.514 0.060 1.298
(1.000) (0.487) (0.145) (1.000) (0.108) (0.119) (0.262) (0.225) (1.000) (0.000) (0.065) (0.524) (0.903)
LABOR 1.102 13.017 17.044 0.309 23.859 10.143 −0.155 −0.004 5.521 −2.836 −1.122 −0.328 −0.780
(1.000) (0.368) (0.148) (1.000) (0.021) (0.603) (0.262) (0.498) (1.000) (0.002) (0.131) (0.372) (0.218)
INT 49.330 54.705 36.439 50.884 265.82 108.75 −5.651 −4.031 −2.735 −5.370 5.0046 −0.663 −3.478
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.003) (0.000) (1.000) (0.254) (0.000)
INS 17.238 12.408 19.825 17.925 15.684 14.146 0.548 −1.576 0.572 −1.394 −1.983 −0.389 0.141
(0.141) (0.414) (0.071) (0.118) (0.206) (0.292) (0.708) (0.058) (0.717) (0.082) (0.024) (0.349) (0.556)
First difference
ΔFDI 113.297 75.689 41.537 129.322 162.209 112.654 −7.603 −4.509 −14.730 −6.759 0.471
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.681)
ΔGDP 39.701 43.314 30.468 39.907 43.2309 30.551 −4.783 −3.368 −4.260 −6.588 −5.895
(0.000) (0.000) (0.002) (0.000) (0.000) (0.002) (0.000) (0.000) (0.000) (0.000) (0.000)
ΔOPEN 76.985 73.628 58.523 100.711 325.906 90.346 −8.030 −6.765 −8.225 −7.426 −6.347
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
14 AN AND YEH
TABLE 3 (Continued)
Variables
ADF - Fisher Chi-Square P-stat PP - Fisher Chi-Square P-stat
Im, Pesaran and
Shin W-stat Levin, Lin & Chu t- stat
Pesaran CADF Zt-
bar - stat
No
Intercept Intercept
Intercept
and trend
No
Intercept Intercept
Intercept
and trend Intercept
Intercept
and trend
No
Intercept Intercept
Intercept
and trend Intercept
Intercept
and trend
ΔTRADE 82.545 70.319 52.956 89.593 117.19 91.563 −7.651 −6.117 −8.503 −8.616 −6.836
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
ΔLABOR 48.702 49.332 37.967 74.319 91.366 50.848 −4.564 −2.220 −6.750 −1.172 −4.975
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.013) (0.000) (0.121) (0.000)
ΔINS 76.851 47.435 49.806 88.277 78.078 64.422 −5.073 −5.886 −8.360 −3.967 −6.237
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Note: p-values in parentheses. The probabilities for the Fisher tests are computed using asymptotic chi-square distribution, whereas the others assume asymptotic nor-
mality. Pesaran CADF test (2007) is the second-generation panel unit root test assuming cross-section dependence.
Source: Authors' calculations.
AN AND YEH 15
TABLE 4 Panel cointegration tests
(1) (2) (3) (4) (5) (6) (7) (8)
Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob.
Pedroni test
Common AR coefs. (within-dimension)
Panel v-
statistic
0.1861 0.4262 −0.7854 0.7839 0.1813 0.4280 −0.470 0.6798 −0.9749 0.8352 −0.2054 0.5814 −0.1439 0.5572 −0.4976 0.6906
Panel rho-
statistic
0.4556 0.6757 1.2249 0.8897 1.0175 0.8455 1.2997 0.9032 1.1570 0.8764 1.1282 0.8704 1.1317 0.8711 1.1828 0.8816
Panel PP-
statistic
−2.0963 0.0180 −1.5047 0.0662 −3.7367 0.0001 −4.1592 0.0000 −1.4178 0.0781 −3.5311 0.0002 −3.7340 0.0001 −2.8411 0.0022
Panel ADF-
statistic
−3.0963 0.0010 −2.2515 0.0122 −4.3591 0.0000 −4.6925 0.0000 −0.0300 0.4880 −5.2652 0.0000 −4.4925 0.0000 −3.7247 0.0001
Common AR coefs. (weighted statistic)
Panel v-
statistic
−0.0948 0.5378 −0.8358 0.7984 −0.8756 0.8094 −1.8623 0.9687 −1.5421 0.9385 −0.8761 0.8095 −0.6905 0.7551 −0.9374 0.8257
Panel rho-
statistic
0.7807 0.7825 1.2969 0.9027 1.4790 0.9304 2.4351 0.9926 1.8082 0.9647 1.6822 0.9537 1.6831 0.9538 1.8280 0.9662
Panel PP-
statistic
−3.5145 0.0002 −3.5334 0.0002 −4.5213 0.0000 −2.5479 0.0054 −2.5454 0.0055 −5.2573 0.0000 −4.7081 0.0000 −2.0671 0.0194
Panel ADF-
statistic
−4.1210 0.0000 −3.7991 0.0001 −4.6561 0.0000 −4.1144 0.0000 −2.4192 0.0078 −5.1691 0.0000 −4.6837 0.0000 −3.0733 0.0011
Individual AR coefs. (between-dimension)
Group rho-
statistic
1.4479 0.9262 2.1615 0.9847 2.2425 0.9875 2.3039 0.9894 2.5214 0.9942 2.1457 0.9841 2.2947 0.9891 1.4293 0.9924
Group PP-
statistic
−10.8264 0.0000 −6.4464 0.0000 −6.6877 0.0000 −7.9640 0.0000 −5.8954 0.0000 −13.2746 0.0000 −9.8225 0.0000 −4.3387 0.0000
Group ADF-
statistic
−6.1824 0.0000 −5.1459 0.0000 −6.4233 0.0000 −5.9594 0.0000 −2.9677 0.0015 −8.1739 0.0000 −6.5491 0.0000 −5.0152 0.0000
16 AN AND YEH
TABLE 4 (Continued)
(1) (2) (3) (4) (5) (6) (7) (8)
Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob. Statistic Prob.
Kao Test
(ADF - stat)
−3.0750 0.0011 −3.2200 0.0006 −3.5374 0.0002 −3.6239 0.0001 −3.4322 0.0003 −3.4190 0.0003 −35,650 0.0002 −3.5616 0.0002
Included
observations
120 120 120 120 120 120 120 120
Note: For the Pedroni tests, automatic lag length selection is based on the AIC and SBC criterion with lags from 1 to 2, assuming individual intercept without trend. For
the Kao tests, automatic lag length selection is based on the AIC and SBC criterion with a max lag of 2. The numbers in bold are statistically significant.
Note: (1) f(FDI jGDP OPEN TRADE LABOR), (2) f(FDI jGDP OPEN TRADE LABOR INT), (3) f(FDI jGDP OPEN TRADE LABOR INS), (4) f(FDI jGDP OPEN TRADE
LABOR CC), (5) f(FDI jGDP OPEN TRADE LABOR GE), (6) f(FDI jGDP OPEN TRADE LABOR PS), (7) f(FDI jGDP OPEN TRADE LABOR RL), (8) f(FDI jGDP OPEN
TRADE LABOR RQ).
Source: Authors' calculations.
AN AND YEH 17
TABLE 5 PMG estimation results
Independent
variable: ΔFDI
Base
model
Lending
rate
Institutional
quality
Control of
corruption
Government
efficiency
Political
stability
Regulation
quality
Rule
of law
(1) (2) (3) (4) (5) (6) (7) (8)
Dependent variables
Long-run effects
GDP 0.504 1.376*** 2.179*** 2.522*** 1.099 1.505* 1.253 −0.315
(0.423) (0.534) (0.821) (0.719) (0.678) (0.797) (0.843) (0.308)
OPEN 0.452 1.595* 4.038*** 4.156*** −2.582* 5.060*** 1.080 4.222***
(0.666) (0.835) (1.267) (1.046) (1.348) (1.319) (1.082) (0.494)
TRADE 2.360*** 1.510** 0.319 −0.00481 5.034*** 0.517 1.040 0.371
(0.499) (0.693) (0.798) (0.722) (1.083) (0.794) (0.894) (0.295)
LABOR −1.904*** −2.369*** −2.078*** −2.574*** −4.873*** −1.430*** −1.820*** −0.482***
(0.467) (0.436) (0.486) (0.395) (0.817) (0.512) (0.506) (0.182)
INT −0.0252
(0.0496)
INS 1.990*** 1.402* −6.287*** 0.528*** 1.730** 1.680***
(0.477) (0.759) (1.560) (0.202) (0.683) (0.189)
Short-run effects
ECT −0.999*** −1.039*** −0.928*** −0.589** −0.708*** −0.924*** −1.062*** −0.697**
(0.160) (0.139) (0.251) (0.256) (0.234) (0.240) (0.276) (0.305)
ΔGDP 1.749 1.834 3.613 3.046 5.950* 0.948 5.206 −3.302
(2.327) (2.713) (3.568) (3.371) (3.119) (3.567) (4.235) (5.564)
ΔOPEN 1.476 0.639 6.021 11.82 7.743 8.439** 4.204 7.208
(4.121) (4.532) (4.329) (7.844) (4.816) (4.294) (2.706) (5.121)
ΔTRADE 0.00874 −0.284 −1.366 −3.326 −3.265 −2.406 −1.052 −0.834
(1.389) (1.519) (1.776) (3.399) (2.355) (2.136) (1.588) (1.934)
18 AN AND YEH
TABLE 5 (Continued)
Independent
variable: ΔFDI
Base
model
Lending
rate
Institutional
quality
Control of
corruption
Government
efficiency
Political
stability
Regulation
quality
Rule
of law
(1) (2) (3) (4) (5) (6) (7) (8)
ΔLABOR −0.376 −0.835*** 0.902 1.841 0.347 4.519* −0.474 7.804*
(0.342) (0.216) (1.470) (1.830) (1.815) (2.367) (1.000) (4.473)
ΔINT −0.187
(0.248)
ΔINS 5.515* 2.013 0.190 1.168 1.687** 0.783
(3.072) (4.450) (1.513) (1.293) (0.855) (4.609)
Constant −12.46*** −18.29*** −28.67*** −17.25** −6.526 −30.66*** −14.93*** −9.835**
(2.144) (2.635) (7.142) (7.399) (4.493) (7.499) (3.552) (4.146)
Akaike info criterion 2.6518 2.5802 2.1114 2.0799 2.2067 2.0341 2.2819 2.0273
Schwarz criterion 3.6173 3.7341 3.3209 3.2895 3.4163 3.2437 3.4915 3.2369
Observations 105 101 84 84 84 84 84 84
Number of Countries 6 6 6 6 6 6 6 6
Hausman Test (p-
value)
0.9249 0.1351 0.9917 0.676 0.9505 0.9981 0.5774 0.9583
Note: Standard errors in parentheses. The INS variables in models 3–8 are institutional quality (INS), control of corruption (CC), government efficiency (GE), political sta-
bility (PS), regulation quality (RQ) and rule of law (RL), respectively. Maximum lag selections for dependent variable and dynamic regressors are 1, based on the AIC and
SBC criterion. *, **, *** denote significance levels at 10%, 5%, 1% respectively. Numbers in bold are statistically significant.
Source: Authors' calculations.
AN AND YEH 19
individuals –cross sections. These stylized facts might imply the trend of investment decisions
of Taiwanese investors in these host countries during the 20 years does not differ much across
the countries, and neither does the trading partnership. Further, the six countries in the sample
also share similarities in Taiwanese investors’cognition of distance advantages, cultural prox-
imity, potential economic growth and investment promoting incentives. Therefore, the long-
run homogeneity can be attributed to these common factors in the investment decisions of
Taiwan in Southeast Asian countries.
7
Note, the error correction term (ECT) is significant at the 1% level with a negative sign,
which is less than, or about unity. This confirms the existence of the cointegrating vector that
defines the long-run relationship between FDI and its determinants. The coefficients of ECT
referring to the speed of adjustment, or how quickly the dependent variable converges to its
long-run equilibrium, fall in the range from 0.928 to 1.039 (models 1–3).
Regarding the long-run relationships, Taiwan's outward FDI in Southeast Asia for the
20 years are explained by economic factors such as: the size of the host market (GDP), the trade
openness of ASEAN economies (OPEN), tight economic and trading relationship with Taiwan
(TRADE) and preferable cost of labour (LABOR). In particular, the institutional factors (INS)
have significantly high impacts on the location decision of Taiwanese investors in
Southeast Asia.
The coefficients of GDP are significant at the 1% significance level and vary from 1.376 to
2.522, meaning in the long run, the size of the host country positively impacts the volume of
inward FDI from Taiwan. This result supports our Hypothesis 1 and is consistent with the the-
ory and previous empirical studies.
The long-run elasticities with respect to OPEN are significantly positive in the eight models,
confirming Hypothesis 2. The results show a significant positive effect of the openness of the
host countries on the investment inflows from Taiwan. Precisely, the more open the host coun-
tries, the greater the attractiveness of the countries to Taiwan's investors. This result is consis-
tent with both theory and practice in the location choice of foreign investment with market-
seeking and export-oriented motives.
The large and significant positive coefficients of TRADE indicate that the location decisions
of Taiwan's investors strongly depend on the bilateral trading relationship between the home
and the host countries. This verifies Hypothesis 5. This result can be traced to the theory that
the prior trading relationship can increase the follow-up FDI, and efforts in liberalizing trade
can ease the barriers to FDI. In fact, Taiwan and Southeast Asian countries have a long and
tight relationship in both trading and investing. As most of Taiwan's firms investing in South-
east Asia are SMEs, it is rational for them to locate their business in this region, where trading
barriers are comparatively looser than in other places, along with distance advantages, cultural
similarity and promoting incentives of the host countries. These results are also consistent with
previous studies concerning the impacts of trade on FDI in emerging and developing economies
(Tintin, 2013; Kayalvizhi & Thenmozhi, 2018).
The significantly negative long-run coefficients of LABOR in the eight models fully approve
Hypothesis 3, meaning the relatively lower cost of labour in Southeast Asian countries explains
the higher investment inflows from Taiwan. The estimates suggest a high long-run cost elastic-
ity of approximately −2.0 (in the range of −1.904 to −2.369 in models 1–3). In other words, the
cost of labour is an apparently important factor of Taiwan's FDI in Southeast Asia.
In regard to institutional factors, the estimation results show the long-run coefficients of INS
are significantly positive with large effects, affirming Hypothesis 6, and consistent with literature
(Bailey, 2018; Kayalvizhi & Thenmozhi, 2018; Peres, Ameer, & Xu, 2018; Sabir, Rafique, &
20 AN AND YEH
Abbas, 2019). The INS variable in model 3, presenting the composite index of country gover-
nance indicators of Southeast Asia, has a significantly positive effect of 1.990 at the 1% signifi-
cance level. This indicates a high semi-elasticity of Taiwan's FDI outflows on the institutional-
political quality of the host countries. Specifically, the investment inflows from Taiwan to South-
east Asia increase corresponding to the increase in institutional-political quality in this region.
In addition, the host countries with a higher governance level are significantly better in FDI
attractiveness for Taiwanese investors. The results are consistent across the four dimensions of
country governance, which are verified by significantly positive coefficients of CC,PS,RQ,RL in
models 4, 6, 7 and 8, presenting control of corruption, political stability, regulation quality and
rule of law, respectively.
The coefficient of GE in model 5 corresponds to government efficiency; however, it indicates
a significantly negative effect. The large negative effect (−6.287) of government efficiency along
with the equivalent large positive effect (5.034) of bilateral trade might reflect two aspects. First,
if the host countries have a poor-quality policy formulation and implementation as well as a
low degree of government independence from political pressures, investment inflows should lie
largely on the bilateral trade linkages, or the cognitive understandings and informal relation-
ships between Taiwanese firms and their counterparts in the region. Second, when the host
countries have less government efficiency, the investment inflows from Taiwan show a signifi-
cant increase, all other things being equal. The negative effect of government efficiency on the
investment inflows is also confirmed in previous studies (Kayalvizhi & Thenmozhi, 2018;
Kamal, Shah, Jing, & Hasnat, 2019). Thus, this evidence requires further understanding of how
institutional configurations may drive FDI location decisions.
The estimates for the short run provide interesting results. These coefficients account for
short-run fluctuations not due to the deviations in the long run. Note, the short-run coefficients
are in the expected sign but are not statistically significant. This result reveals a lack of short-
run relationships between foreign investment inflows and the determinants. In other words,
the investment inflows do not react (or react less) to short-run shocks. In the short run, the cost
of labour has a significantly negative effect at the 1% level (−0.835), meaning the investment
decision of Taiwan is contemporaneously affected by the cost of labour in the host countries.
The estimates for INT in model 2 have the expected sign in both the long run and short run.
Accordingly, the real bank lending interest rate of the host countries has a negative effect on
the investment inflows from Taiwan, supporting Hypothesis 4. The relatively larger negative
effect in the short run (−0.187) than in the long run (−0.0252) indicates financing cost has a
contemporaneous unfavourable impact on the investment decision. However, the coefficients
are statistically insignificant, showing the support is weak and the data might be inconclusive.
The advantage of the PMG technique is it restricts the homogeneity in the long run, but
allows different short-run dynamic specifications across countries. Table 6 reports the heteroge-
neous individual error correction terms for six Southeast Asian countries. The coefficients for
individual countries are significantly negative at the 1% level. The average speed of adjustment
for the six countries to restore the long-run equilibrium differ, which are higher for Thailand
and Indonesia (at a high speed of convergence of over 130%), followed by Singapore, Philippines
(around the unity speed of 90%) and lower for Malaysia and Vietnam (at a moderate speed of
60%). This means any deviation from the long-run equilibrium is fully corrected instanta-
neously for Thailand and Indonesia. Meanwhile, for Singapore and Philippines, the disequilib-
rium is almost dissipated after one period. Similar but a bit slower are Malaysia and Vietnam,
where the restoration is approximately 60% after one period, or say, it takes more than one
period to converge back to the long-run equilibrium. It is further noted that changes in
AN AND YEH 21
TABLE 6 Country-wise error correction terms
Country
Base
model
Lending
rate
Institutional
quality
Control of
corruption
Government
efficiency
Political
stability
Regulation
quality Rule of law
(1) (2) (3) (4) (5) (6) (7) (8)
ECT
Indonesia −1.606 −1.527 −1. 862 0.320 −1.427 −1.689 −2.193 0.007
(0.0000) (0.0000) (0.0005) (0.645) (0.0001) (0.0003) (0.0016) (0.9817)
Malaysia −0.636 −0.685 −0.671 −0.678 −0.156 −0.716 −0.571 −0.0625
(0.0001) (0.0001) (0.0001) (0.0001) (0.0085) (0.0000) (0.0008) (0.0000)
Philippines −0.922 −0.961 −0.654 −0.728 −0.498 −0.398 −1.002 −0.237
(0.0001) (0.0001) (0.0026) (0.0001) (0.0006) (0.0048) (0.0009) (0.0082)
Singapore −0.840 −0.910 −0.952 −0.779 −0.606 −1.270 −0.931 −1.51
(0.0004) (0.0003) (0.0004) (0.0011) (0.0012) (0.0001) (0.0006) (0.0029)
Thailand −1.337 −1.383 −1.339 −1.527 −1.394 −1.295 −1.409 −1.722
(0.0000) (0.0001) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
Vietnam −0.654 −0.768 −0.086 −0.139 −0.159 −0.165 −0.259 −0.080
(0.0025) (0.0019) (0.0977) (0.0350) (0.0380) (0.0083) (0.0358) (0.3706)
Note: p-values in parentheses. Numbers in bold are statistically significant.
Source: Authors' calculations.
22 AN AND YEH
TABLE 7 PMG estimation results (with GDPpc as a replacement for the GDP variable)
Independent
variable: ΔFDI
Base
model
Lending
rate
Institutional
quality
Control of
corruption
Government
efficiency
Political
stability
Regulation
quality Rule of law
(1) (2) (3) (4) (5) (6) (7) (8)
Dependent variables
Long-run effects
GDPpc 0.734 1.564*** 2.203*** 2.563*** 1.122 1.646** 1.620* −0.391
(0.498) (0.601) (0.832) (0.750) (0.729) (0.784) (0.882) (0.326)
OPEN 0.468 1.429* 3.539*** 3.673*** −2.720** 4.519*** 1.243 4.177***
(0.648) (0.807) (1.151) (0.958) (1.328) (1.131) (0.973) (0.476)
TRADE 2.270*** 1.544** 0.549 0.296 5.130*** 0.689 0.850 0.407
(0.495) (0.683) (0.733) (0.676) (1.095) (0.683) (0.864) (0.285)
LABOR −2.051*** −2.479*** −2.133*** −2.623*** −4.910*** −1.599*** −1.972*** −0.451**
(0.496) (0.467) (0.501) (0.406) (0.837) (0.527) (0.518) (0.188)
INT −0.0232
(0.0502)
INS 1.932*** 1.563** −6.359*** 0.459** 1.777*** 1.686***
(0.475) (0.778) (1.590) (0.195) (0.683) (0.190)
Short-run effects
ECT −1.002*** −1.035*** −0.901*** −0.634*** −0.688*** −0.929*** −1.061*** −0.738**
(0.159) (0.139) (0.239) (0.229) (0.232) (0.238) (0.279) (0.327)
ΔGDPpc 1.457 1.362 3.996 2.306 4.660* 1.063 4.951 −4.922
(2.238) (2.433) (3.801) (3.570) (2.644) (3.549) (4.061) (6.927)
ΔOPEN 1.373 0.483 5.658 10.70 7.005 8.125* 3.860 7.391
(4.108) (4.444) (4.349) (7.434) (5.136) (4.318) (2.776) (5.143)
ΔTRADE 0.0757 −0.139 −1.469 −2.707 −2.554 −2.399 −0.873 −0.577
(1.343) (1.424) −1.776 (3.267) (2.606) (2.144) (1.456) (1.993)
(Continues)
AN AND YEH 23
TABLE 7 (Continued)
Independent
variable: ΔFDI
Base
model
Lending
rate
Institutional
quality
Control of
corruption
Government
efficiency
Political
stability
Regulation
quality Rule of law
(1) (2) (3) (4) (5) (6) (7) (8)
ΔLABOR −0.344 −0.701*** 0.628 1.856 0.408 4.324* −0.682 8.599*
(0.436) (0.266) (1.708) (1.625) (1.829) (2.359) (1.001) (5.198)
ΔINT −0.123
(0.198)
ΔINS 5.909* 2.444 0.547 1.122 1.806* 0.691
(3.487) (4.215) (1.629) (1.306) (0.940) (4.693)
Constant −10.59*** −12.76*** −19.33*** −13.11*** −3.423 −24.54*** −10.30*** −11.03**
(1.605) (1.449) (4.216) (4.506) (3.262) (5.929) (1.933) (4.821)
Akaike info criterion 2.6511 2.5905 2.0993 2.0763 2.2168 2.0274 2.2826 2.0141
Schwarz criterion 3.6165 3.7444 3.3089 3.2859 3.4263 3.2369 3.4922 3.2236
Observations 105 101 84 84 84 84 84 84
Number of Countries 6 6 6 6 6 6 6 6
Note: Standard errors in parentheses. The INS variables in models 3–8 are institutional quality (INS), control of corruption (CC), government efficiency (GE), political sta-
bility (PS), regulation quality (RQ) and rule of law (RL), respectively. Maximum lag selections for dependent variable and dynamic regressors are 1, based on the AIC and
SBC criterion. *, **, *** denote significance levels at 10%, 5%, 1% respectively. Numbers in bold are statistically significant.
Source: Authors' calculations.
24 AN AND YEH
TABLE 8 PMG estimation results (with INSg)
Independent
variable: ΔFDI
Institutional
quality
Control of
corruption
Government
efficiency
Political
stability
Regulation
quality Rule of law
(1) (2) (3) (4) (5) (6)
Dependent variables
Long-run effects
GDP 3.118*** 0.291 1.487 2.127** 1.734 2.841***
(0.875) (0.227) (1.068) (0.904) (1.152) (1.070)
OPEN 3.078*** 2.017*** 1.339 5.666*** 0.781 3.245***
(1.075) (0.593) (1.254) (1.257) (1.244) (1.165)
TRADE −0.469 0.138 1.357 0.315 1.435 −0.187
(0.805) (0.272) (0.957) (0.804) (1.115) (0.856)
LABOR −2.404*** −0.858*** −2.011*** −1.824*** −2.385*** −2.448***
(0.454) (0.122) (0.598) (0.575) (0.591) (0.845)
INSg −1.517*** 1.551*** −1.579** −0.692*** −0.499 −1.223
(0.371) (0.217) (0.661) (0.197) (0.472) (0.801)
Short-run effects
ECT −1.116*** −0.616*** −0.935*** −0.882*** −0.969*** −1.049***
(0.272) (0.236) (0.226) (0.243) (0.261) (0.284)
ΔGDP 2.723 −0.791 2.474 −0.647 5.941 2.570
(3.363) (3.593) (3.812) (2.527) (4.112) (3.440)
ΔOPEN 5.909 8.752 8.137*** 8.003* 8.238 6.015
(4.245) (5.474) (3.140) (4.702) (5.073) (4.800)
ΔTRADE −1.504 −1.886 −2.193 −1.418 −2.970 −1.707
(2.044) (1.945) (1.689) (1.868) (2.189) (1.797)
ΔLABOR 1.027 4.156** 1.006 5.287** 0.763 −0.379
(1.673) (1.760) (1.237) (2.315) (1.380) (0.895)
(Continues)
AN AND YEH 25
TABLE 8 (Continued)
Independent
variable: ΔFDI
Institutional
quality
Control of
corruption
Government
efficiency
Political
stability
Regulation
quality Rule of law
(1) (2) (3) (4) (5) (6)
ΔINSg −3.825 −0.878 −2.537** −1.632 0.687 −1.235
(3.857) (1.914) (1.141) (1.692) (0.793) (3.918)
Constant −30.30*** −2.684** −16.92*** −34.35*** −17.38*** −28.30***
(7.305) (1.087) (3.854) (9.366) (4.714) (7.401)
Observations 84 84 84 84 84 84
Number of Countries 6 6 6 6 6 6
Note: Standard errors in parentheses. The INSg variables in models 1–6 are institutional quality (INSg), control of corruption (CCg), government efficiency (GEg), political
stability (PSg), regulation quality (RQg) and rule of law (RLg), respectively. Maximum lag selections for dependent variable and dynamic regressors are 1, based on the
AIC and SBC criterion. *, **, *** denote significant levels at 10%, 5%, 1% respectively. Numbers in bold are statistically significant.
Source: Authors' calculations.
26 AN AND YEH
economic and institutional factors affect Taiwan's outward investment decision more speedily
and strongly in Thailand and Indonesia than in Malaysia and Vietnam.
5.4 |Estimations with alternative proxies
First, we replicate the estimation procedure taking GDPpc as a proxy for income levels. Table 7
displays the estimates consistent with the previous results, with slightly higher value coeffi-
cients for GDPpc. The income level of the host countries has a significantly positive effect on
Taiwan's outward investment in the long run.
Second, to further investigate the institution issues, we replace the INS variable with an
alternative variable, INSg, measuring the absolute gap in the institution index between Taiwan
and the host countries. For the component indicators, we generate new variables CCg,GEg,
PSg,RQg,RLg in the same manner. As the original institutional variables take values from −2.5
to 2.5, these new variables range from −5.0 to 5.0. The higher value corresponds to the larger
gap in institutional quality between Taiwan and the host countries. The positive value reflects a
higher level of governance quality in Taiwan, while the negative value presents a lower level.
By bringing these alternative variables into the models, we examine the relative institutional
figures considering both sides, the home and the host countries.
As the unit root tests and panel cointegration tests are justified, we perform the PMG esti-
mation. The results are reported in Table 8. The speed of adjustment, as well as the long-run
and short-run coefficients for economic factors, are close to the previous estimates. On the
contrary, the coefficients of the institutional factors are statistically significant and negative.
This can be interpreted in two ways. On one hand, if the institutional variable is negative,
meaning the institutional quality of Taiwan is lower than that of the host countries, then the
larger the gap, the greater the increase in the outward FDI of Taiwan to these economies. In
other words, Taiwan prefers to invest in the country with better governance standards in
comparison with Taiwan. On the other hand, if the institutional variable is positive, meaning
the institutional quality of Taiwan is higher than that of the hosts; then the larger the gap,
the greater the decrease in the outward FDI of Taiwan to these countries. Or we can say, Tai-
wan prefers to invest in the country that has similar institutional standards to theirs. These
two cases clearly confirm Hypothesis 7. Moreover, the negative effects of the gaps in institu-
tional quality are fairly high, indicating the important role of institutional factors in Taiwan-
ese investors' locational decisions. This is consistent with the previous result that Taiwan
aimstolocatetheirbusinessinSoutheastAsiancountrieswhichhavebettergovernance
quality.
Last, we re-estimate the models adding the time dummy variable D2008 to see if there is
any significant change in the dynamics of Taiwan's investment in Southeast Asia during the
past 20 years. The result in Table 9 shows changes in the investment have a significant decrease
in the second 10-year period compared to the first 10-year, other things being equal. The plausi-
ble explanation for this decrease lies in the effects of the 2008 global crisis and, more probably,
the changes in the economic-political environment of Taiwan, such as the shift in government
administration, adjustment of policy and, especially, investment strategy.
We also estimate the models by the same token with the time dummy variable D2016
(D2016 = 1 over the year 2016–2017; 0 otherwise) to see if there is a significant effect of
unobserved institutional-political issues, such as the probable impacts of the New Southbound
AN AND YEH 27
TABLE 9 PMG estimation results (with Time dummy variable)
Independent
variable: ΔFDI
Base
model
Lending
rate
Institutional
quality
Government
efficiency
Political
stability
Regulation
quality Rule of law
(1) (2) (3) (4) (5) (6) (7)
Dependent variables
Long-run effects
GDP 0.677 1.560** 1.050 −0.171 1.698 2.001* −1.088*
(0.429) (0.731) (1.201) (1.591) (1.100) (1.129) (0.632)
OPEN −0.164 1.546 2.626 −4.670 3.507** 1.194 3.957***
(1.147) (1.316) (1.699) (3.224) (1.547) (1.575) (0.535)
TRADE 2.244*** 1.123 1.875 7.686** 0.687 −0.162 1.179*
(0.729) (1.144) (1.410) (3.303) (1.273) (1.399) (0.656)
LABOR −1.979*** −1.831*** −2.463*** −5.959*** −1.573*** −1.375** −0.494***
(0.650) (0.666) (0.694) (1.435) (0.594) (0.604) (0.178)
INT −0.0157
(0.0505)
INS 1.995** −7.931*** 0.234 1.664* 1.897***
(0.811) (2.599) (0.278) (0.862) (0.232)
Short-run effects
ECT −1.013*** −1.050*** −0.798*** −0.620*** −0.927*** −1.096*** −0.890***
(0.171) (0.154) (0.243) (0.223) (0.264) (0.398) (0.286)
ΔGDP 2.112 2.096 0.675 3.653 −1.114 2.409 −6.114
(3.658) (2.924) (3.271) (4.491) (3.899) (4.998) (5.526)
ΔOPEN 0.760 0.591 5.101 5.089 5.931 1.127 7.593
(4.584) (4.388) (5.210) (7.433) (4.552) (4.846) (5.643)
ΔTRADE −0.284 −0.975 −0.285 −1.724 −1.038 0.781 −0.00349
(1.474) (1.462) (1.583) (3.584) (1.814) (2.808) (1.934)
28 AN AND YEH
TABLE 9 (Continued)
Independent
variable: ΔFDI
Base
model
Lending
rate
Institutional
quality
Government
efficiency
Political
stability
Regulation
quality Rule of law
(1) (2) (3) (4) (5) (6) (7)
ΔLABOR −1.174 −1.162 2.669 1.300 4.608** 0.449 8.880**
(1.243) (0.866) (2.279) (2.510) (2.057) (2.114) (4.433)
ΔINT −0.114
(0.193)
ΔINS 5.973* −0.581 1.126 1.423 −0.675
(3.208) (1.518) (1.409) (1.243) (4.536)
D2008 −0.566* −0.723*** −0.565* −0.0207 −0.399 −0.244 0.789*
(0.321) (0.249) (0.341) (0.365) (0.310) (0.325) (0.450)
Constant −9.993*** −20.20*** −17.48*** −0.139 −26.41*** −17.39*** −8.482***
(2.287) (3.290) (4.735) (3.617) (7.137) (5.967) (3.082)
Observations 105 101 84 84 84 84 84
Number of Countries 6 6 6 6 6 6 6
Note: Standard errors in parentheses. The INS variables in models 3–7 are institutional quality (INS), government efficiency (GE), political stability (PS), regulation qual-
ity (RQ) and rule of law (RL), respectively. *, **, *** denote significant levels at 10%, 5%, 1% respectively. Numbers in bold are statistically significant.
Source: Authors' calculations.
AN AND YEH 29
Policy, on Taiwan's investment in Southeast Asia. The statistically insignificant estimates,
however, show there might be ambiguous results for the current policy so far.
6|CONCLUDING REMARKS
The study addresses the direction and magnitude of the key determinants of Taiwan's outward
FDI in Southeast Asia by employing a panel data of 6 ASEAN countries in 1998–2017. It is sta-
tistically proved economic factors are the primary drivers of Taiwan's FDI in the region.
Besides, this study verifies other drivers inclusively, that is, the institutional factors. We contrib-
ute to the extant literature on FDI determinants by exploring a dynamic panel model with
ARDL–PMG methodology that restricts the homogeneity in the long run for a small group of
countries in the same economic region, but allows for country-specific heterogeneity in the
short run. Compared to the standard static panel model, our dynamic model is robust in dealing
with endogeneity, group-wise heteroskedasticity and autocorrelation for the panel of a small
number of countries and relatively long period coverage.
In addition, the study makes contributions to cover different dimensions of institutional
quality as crucial determinants in the locational decision of Taiwan's investors in Southeast
Asia. Our findings reveal the composite governance indicator as well as five components (con-
trol of corruption, government efficiency, political stability, regulation quality and rule of law)
are significant in the long run. The empirical results present remarkable evidence that institu-
tional quality has as large effects as economic factors. Precisely, Taiwan's investment in South-
east Asia is not only for the large size of the economy, advantages of openness, rapid growth
and low labour cost but also for improvements in the institutional system, a stable economic-
political environment, sound policy formation and implementation, and corruption controlla-
bility. This is consistent with the current trend of global investment flows, where more capital
flows to emerging countries for their better institutional quality and promotion incentives
besides the rich natural sources and cheap labour. Further, the study confirms the rationale of
Taiwanese enterprises to preferably run businesses in ASEAN for their primary understandings
through long-time trading relationships. Growing and large volumes of bilateral trade between
Taiwan and Southeast Asian counterparts explicitly explain the increase in Taiwan's investment
in the region.
This study therefore suggests some policy implications for both the home and host
countries. As for ASEAN countries, Taiwan has been an important trader and investor for a
long time. Taiwanese firms not only make up a large share of foreign direct investment but also
possess advanced technology and managerial skills that encourage the local modern
manufacturing and supporting industries. The recent tendency of Taiwanese firms entering
Southeast Asia show more and more capital is running to electronic parts and component
manufacturing, financial insurance, wholesale and retail trade (Table A1). This opens up oppor-
tunities for ASEAN to attract and exploit the spillover effects of Taiwan's outward FDI. At the
same time, the host countries have to take on the environmental, intellectual, economic and
technological challenges of investment absorption. Apparently, ASEAN governments need to
develop a proactive strategy of improving institutional quality to create a more favourable envi-
ronment for Taiwanese investors. As for the home country, Taiwan has advantages to grasp
investment possibilities in this growing, dynamic and competitive region. Launching the third
wave of the Go South policy in 2016 reflects the incumbent government's strategy to forge eco-
nomic ties with Southeast Asia, where trade and investment are two vital pillars. However, due
30 AN AND YEH
to its political state, Taiwan's government has a free trade agreement with one ASEAN nation
only, that is Singapore, and a bilateral investment agreement re-signing with the Philippines in
December 2017. Accelerating bilateral trade and facilitating private cooperation are thus strate-
gic programmes of Taiwan to develop comprehensive and closer partnerships with ASEAN.
The empirical result confirms bilateral trade is a key driver for Taiwan's outward investment in
the region. Last but not least, to position itself in Southeast Asia regardless of the political con-
straints of China, Taiwan needs to pursue plans to boost Taiwan-ASEAN economic cooperation
through focusing on private sectors and a people-centric approach in the long term.
ACKNOWLEDGMENTS
We are thankful to two anonymous referees for the insightful comments and suggestions. An
earlier version of this paper also benefitted from Professor Chin-Ho Lin (Feng Chia University)
and the participants at the 2019 Annual Conference of the Taiwan Economic Association. This
paper was subsidized by Ministry of Science and Technology and National Taiwan University
(NTU), Taiwan.
ORCID
Thu-Ha Thi An https://orcid.org/0000-0002-2650-5942
Kuo-Chun Yeh https://orcid.org/0000-0002-2430-3802
ENDNOTES
1
The selection of countries in this study is limited due to the availability of data. If more data are available in
the future, we can expand the number of observations under this empirical framework, which will certainly
provide better results illustrating Taiwan's outward FDI in ASEAN-10. Besides, the relatively small data sample
in the study suggests caution in interpreting the results. More and better data from other Southeast Asian
countries are plainly needed.
2
Yearly Reports of Taiwan's outward foreign investment, Statistics by the Investment Commission, Taiwan
Ministry of Economic Affairs.
3
2017 Yearly Report of Taiwan's outward foreign investment, Statistics by the Investment Commission, Taiwan
Ministry of Economic Affairs.
4
With the aim of investigating the determinants of Taiwan's outward FDI in Southeast Asian countries over
20 years, we re-examine the conventional linear panel model to provide evidence of the existence and the esti-
mated magnitude of the effects of the determinants in the short run and long run. Further, it would be worth
considering a nonlinear model to explore the asymmetric responses of FDI to shocks and the relationship
between FDI and other variables that might exhibit nonlinearity. We are grateful to an anonymous referee for
this suggestion and save these potential ideas for future research.
5
Diagnostic tests for the panel are done. The Woolridge test for autocorrelation in panel data rejects the null
hypothesis of no first-order autocorrelation at the 1% significance level. The Modified Wald test for group-wise
heteroskedasticity rejects the null hypothesis of homoskedasticity at the 1% significance level.
6
When the maximum number of lag length is set to 2 for dependent and independent variables, a singular
matrix occurs.
7
We are thankful to an anonymous referee for raising this insight.
REFERENCES
Agarwal, J. P. (1980). Determinants of foreign direct investment: A survey. Weltwirtschaftliches Archiv,116(4),
739–773.
AN AND YEH 31
Asiedu, E. (2006). Foreign direct investment in Africa: The role of natural resources, market size, government
policy, institutions and political instability. The World Economy,29(1), 63–77.
Aykut, D., & Ratha, D. (2004). South-south FDI flows: How big are they? Transnational Corporations,13(1),
149–176.
Bailey, N. (2018). Exploring the relationship between institutional factors and FDI attractiveness: A meta-
analytic review. International Business Review,27(1), 139–148.
Bevan, A. A., & Estrin, S. (2004). The determinants of foreign direct investment into European transition econo-
mies. Journal of Comparative Economics,32(4), 775–787.
Buchanan, B. G., Le, Q. V., & Rishi, M. (2012). Foreign direct investment and institutional quality: Some empiri-
cal evidence. International Review of Financial Analysis,21,81–89.
Buckley, P. J., & Cassion, M. (1976). Future of the multinational Enterprise. London, England: Palgrave
Macmillan.
Buckley, P. J., Clegg, L. J., Cross, A. R., Liu, X., Voss, H., & Zheng, P. (2007). The determinants of Chinese out-
ward foreign direct investment. Journal of International Business Studies,38(4), 499–518.
Cavallari, L., & d'Addona, S. (2013). Nominal and real volatility as determinants of FDI. Applied Economics,45
(18), 2603–2610.
Chang, K.-T., & Thornson, C. N. (1994). Taiwanese foreign direct investment and trade with Thailand. Singapore
Journal of Tropical Geography,15(2), 112–127.
Chen, H., & Chen, T.-J. (1998a). Network linkages and location choice in foreign direct investment. Journal of
International Business Studies,29(3), 445–467.
Chen, H., & Chen, T.-J. (1998b). Foreign direct investment as a strategic linkage. Thunderbird International Busi-
ness Review,40(1), 13–30.
Chen, K.-M., Rau, H.-H., & Lin, C.-C. (2006). The impact of exchange rate movements on foreign direct invest-
ment: Market-oriented versus cost-oriented. The Developing Economies,44(3), 269–287.
Chen, T. J. (1998). Taiwanese firms in Southeast Asia: Networking across Borders. Gloucestershire, UK: Edward
Elgar Publishing Ltd.
Chiu, S.-K., & Lo, F.-Y. (2015). Taiwanese foreign direct investment in Southeast Asia: An empirical investigation
of the OLI framework. Journal of Economics and Management,11(2), 127–141.
Choi, I. (2001). Unit root tests for panel data. Journal of International Money and Finance,20(2), 249–272.
Chow, P. C. Y. (1996). Outward foreign investment and export structure: A case study of Taiwan's FDI in Malay-
sia. The Journal of International Trade & Economic Development,5(2), 183–205.
Chow, P. C. Y. (2016). Taiwan in international economic relations. Orbis,60(4), 531–549.
Chu, W., Ye, K., & Hsu, M. (1999). Taiwanese investment in Southeast Asia. Global Economic Review,28(1),
77–90.
Duanmu, J., & Guney, Y. (2009). A panel data analysis of locational determinants of Chinese and Indian outward
foreign direct investment. Journal of Asia Business Studies,3(2), 1–15.
Dunning, J. H. (1973). The determinants of international production. Oxford Economic Papers,25(3), 289–336.
Dunning, J. H. (1977). Trade, location of economic activity and the MNE: A search for an eclectic approach. In
B. Ohlin, P.-O. Hesselborn, & P. M. Wijkman (Eds.), The international allocation of economic activity: Pro-
ceedings of a Nobel symposium held at Stockholm (pp. 395–418). London, England: Palgrave Macmillan.
Dunning, J. H. (1980). Toward an eclectic theory of international production: Some empirical tests. Journal of
International Business Studies,11(1), 9–31.
Dunning, J. H. (2009). Location and the multinational enterprise: A neglected factor? Journal of International
Business Studies,40(1), 5–19.
Dunning, J. H., & Lundan, S. M. (2008). Institutions and the OLI paradigm of the multinational enterprise. Asia
Pacific Journal of Management,25(4), 573–593.
Fukumi, A., & Nishijima, S. (2010). Institutional quality and foreign direct investment in Latin America and the
Caribbean. Applied Economics,42(14), 1857–1864.
Goh, S., Wong, K., & Yew, C. (2018). Are merchandise exports and FDI inflows an engine of growth for ASEAN-
10 countries? New evidence using the pooled mean group estimation method. Margin: The Journal of
Applied Economic Research,12, 414–430.
Hausman, J. A. (1978). Specification tests in econometrics. Econometrica,46(6), 1251–1271.
32 AN AND YEH
Hennart, J.-F. (1982). Theories of the multinational Enterprise (2nd ed.). Ann Arbor, MI: University of Michigan
Press.
Hoang, H. H. (2012). Foreign direct investment in Southeast Asia: Determinants and spatial distribution (No. 30;
Working papers). Development and Policies Research Center (DEPOCEN), Hanoi, Vietnam
Hymer, S. (1976). The international operations of national firms: A study of direct foreign investment. Cambridge,
MA: MIT Press.
Im, K. S., Pesaran, M. H., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Economet-
rics,115(1), 53–74.
Kamal, M. A., Shah, S. H., Jing, W., & Hasnat, H. (2019). Does the quality of institutions in host countries affect
the location choice of Chinese OFDI: Evidence from Asia and Africa. Emerging Markets Finance and Trade,
56(1), 208–227.
Kang, Y., & Jiang, F. (2012). FDI location choice of Chinese multinationals in east and Southeast Asia: Tradi-
tional economic factors and institutional perspective. Journal of World Business,47(1), 45–53.
Kao, C. (1999). Spurious regression and residual-based tests for cointegration in panel data. Journal of Economet-
rics,90(1), 1–44.
Kayalvizhi, P. N., & Thenmozhi, M. (2018). Does quality of innovation, culture and governance drive FDI? Evi-
dence from emerging markets. Emerging Markets Review,34, 175–191.
Lee, H. F., & Liu, M. Y. (2018). Taiwanese banks in ASEAN. In H. L. Lin & H. F. Lee (Eds.), Trade and FDI in
AEC: Cooperation and challenge among Taiwan, Japan and China (pp. 145–178). Taipei, Taiwan: National
Taiwan University Press.
Lee, H. Y., Lin, H. L., & Yang, C. H. (2018). Taiwan's and China's investment in ASEAN: A comparison. In H. L.
Lin & H. F. Lee (Eds.), Trade and FDI in AEC: Cooperation and challenge among Taiwan, Japan and China
(pp. 179–228). Taipei: National Taiwan University Press.
Lei, H.-S., & Chen, Y.-S. (2011). The right tree for the right bird: Location choice decision of Taiwanese firms'
FDI in China and Vietnam. International Business Review,20(3), 338–352.
Levin, A., Lin, C.-F., & James Chu, C.-S. (2002). Unit root tests in panel data: Asymptotic and finite-sample prop-
erties. Journal of Econometrics,108(1), 1–24.
Liu, B. J. (2011). MNEs and local linkages: Evidence from Taiwanese affiliates. World Development,39(4),
633–647.
Lo, F.-Y., & Lin, F.-J. (2015). Advantage transfer on location choice and subsidiary performance. Journal of Busi-
ness Research,68(7), 1527–1531.
Mainland Affairs Council (2020) Cross Strait Economic Statistics Monthly 324. Retrieved from https://www.mac.
gov.tw/News.aspx?n=2C28D363038C300F&sms=231F60B3498BBB19
Makino, S., Lau, C.-M., & Yeh, R.-S. (2002). Asset-exploitation versus asset-seeking: Implications for location
choice of foreign direct investment from newly industrialized economies. Journal of International Business
Studies,33(3), 403–421.
Pedroni, P. (2004). Panel cointegration: Asymptotic and finite sample properties of pooled time series tests with
an application to the PPP hypothesis. Econometric Theory,20(3), 597–625.
Peres, M., Ameer, W., & Xu, H. (2018). The impact of institutional quality on foreign direct investment inflows:
Evidence for developed and developing countries. Economic Research-Ekonomska Istraživanja,31(1),
626–644.
Pesaran, M. H. (2007). A simple panel unit root test in the presence of cross-section dependence. Journal of
Applied Econometrics,22(2), 265–312.
Pesaran, M. H., Shin, Y., & Smith, R. J. (2001). Bounds testing approaches to the analysis of level relationships.
Journal of Applied Econometrics,16(3), 289–326.
Pesaran, M. H., Shin, Y., & Smith, R. P. (1999). Pooled mean group estimation of dynamic heterogeneous panels.
Journal of the American Statistical Association,94(446), 621–634.
Ramasamy, B., Yeung, M., & Laforet, S. (2012). China's outward foreign direct investment: Location choice and
firm ownership. Journal of World Business,47(1), 17–25.
Sabir, S., Rafique, A., & Abbas, K. (2019). Institutions and FDI: Evidence from developed and developing coun-
tries. Financial Innovation,5(1), 8.
AN AND YEH 33
Tan, B. W., Wong, K. N., & Goh, S. K. (2018). The surge in intra-ASEAN outwards foreign direct investment and
its key determinants: Evidence using pooled mean group approach. International Journal of Business and
Society,19(2), 347–362.
Tan, C. Y., & Lin, C. Y. (2018). Taiwan's “new southbound policy:”trade and industrial strategies. In H. L. Lin &
H. F. Lee (Eds.), Trade and FDI in AEC: Cooperation and challenge among Taiwan, Japan and China
(pp. 257–304). Taipei, Taiwan: National Taiwan University Press.
Tintin, C. (2013). The determinants of foreign direct investment inflows in the central and eastern European
countries: The importance of institutions. Communist and Post-Communist Studies,46(2), 287–298.
Vernon, R. (1966). International investment and international trade in the product cycle. The Quarterly Journal
of Economics,80(2), 190–207.
Wooldridge, J. M. (2012). Introductory econometrics: A modern approach (Vol., 5th ed.(658). Mason, OH: South-
Western, Cengage Learning.
Yeh, K., & Ho, T. (2012). Magnitude and volatility of Taiwan's net foreign assets against mainland China:
1981–2009. China Economic Review,23(3), 720–728.
How to cite this article: An T-HT, Yeh K-C. Determinants of Taiwan's outward
investment in Southeast Asia: Economic factors and institutional issues. Pac Econ Rev.
2020;1–38. https://doi.org/10.1111/1468-0106.12348
APPENDIX A
34 AN AND YEH
TABLE A1 Top 10 industries attracting Taiwanese investment in Southeast Asia
Rank
Accumulative up to 2007 2008–2012 2013–2017
By amount By case By amount By case By amount By case
1 Electronic parts and
components
manufacturing
Wholesale and retail trade Financial and
Insurance
Wholesale and retail
trade
Basic metal manufacturing Wholesale and retail
trade
2 Financial and insurance Textiles mills Basic metal
manufacturing
Financial and
insurance
Financial and insurance Financial and
insurance
3 Textiles mills Wood and bamboo
products manufacturing
Electronic parts and
components
manufacturing
Electronic parts and
components
manufacturing
Electronic parts and
components
manufacturing
Electronic parts and
components
manufacturing
4 Computers, electronic and
optical products
manufacturing
Mining and quarrying Chemical material
manufacturing
Chemical material
manufacturing
Wholesale and retail trade Electrical equipment
manufacturing
5 Wood and bamboo
products manufacturing
Electronic parts and
components
manufacturing
Transportation and
storage
Fabricated metal
products
manufacturing
Transportation and storage Machinery and
equipment
manufacturing
6 Wholesale and retail trade Financial and Insurance Non-metallic mineral
products
manufacturing
Professional, scientific
and technical
services
Textiles mills Fabricated metal
products
manufacturing
7 Mining and quarrying Chemical material
manufacturing
Wholesale and retail
trade
Construction Chemical material
manufacturing
Textiles mills
8 Chemical material
manufacturing
Computers, electronic and
optical products
manufacturing
Fabricated metal
products
manufacturing
Real estate Computers, electronic and
optical products
manufacturing
Transportation and
storage
9 Food manufacturing Food manufacturing Textiles mills Information and
communication
Electrical equipment
manufacturing
Basic metal
manufacturing
10 Electrical equipment
manufacturing
Fabricated metal products
manufacturing
Information and
communication
Electrical equipment
manufacturing
Plastic products
manufacturing
Information and
communication
Source: Authors' compilation from statistics by MOEAIC, Taiwan.
AN AND YEH 35
TABLE A2 Definition of institutional quality
Variable Name Definition
INS Country governance
composite index
The composite indicator of six components: control of corruption,
government efficiency, political stability, regulation quality, rule of
law, voice and accountability.
CC Control of corruption Control of corruption captures perceptions of the extent to which
public power is exercised for private gain, including both petty and
grand forms of corruption, as well as ‘capture’of the state by elites
and private interests.
PS Government efficiency Government effectiveness captures perceptions of the quality of
public services, the quality of the civil service and the degree of its
independence from political pressures, the quality of policy
formulation and implementation, and the credibility of the
government's commitment to such policies.
GE Political stability Political Stability and Absence of Violence/Terrorism measures
perceptions of the likelihood of political instability and/or
politically motivated violence, including terrorism.
RQ Regulation quality Regulatory quality captures perceptions of the ability of the
government to formulate and implement sound policies and
regulations that permit and promote private sector development.
RL Rule of law Rule of law captures perceptions of the extent to which agents have
confidence in and abide by the rules of society, and in particular
the quality of contract enforcement, property rights, the police,
and the courts, as well as the likelihood of crime and violence.
Source: The Worldwide Governance Indicators (WGI), The World Bank.
36 AN AND YEH
TABLE A3 Survey of recent comparable studies on FDI determinants
Study Country and Time Methodology FDI determinants Main findings
Peres, Ameer, and
Xu (2018)
110 developed and
developing countries in
2002–2012
Panel OLS
estimation, IV
estimation
Institutional quality Governance has a positive effect on FDI in
developed countries but is not good enough to
have positive impacts in developing countries.
Kayalvizhi and
Thenmozhi (2018)
22 emerging countries in
1996–2014
Fixed effects
estimation
Governance, culture, technology Country governance has a weakening effect,
while technology is a crucial factor that
attracts FDI
Tan, Wong, and
Goh (2018)
Intra-regional outward
FDI in 10 ASEAN
countries in 1995–2012
Panel PMG
estimation
Market size, political stability, trade
openness
There are positive long-run relationships
between intra-regional FDI and the
determinants.
Tintin (2013) 6 Central and Eastern
European countries in
1996–2009
Panel OLS
estimation with
fixed effects
GDP size, trade openness, institutional
quality
GDP size, trade openness and institutional
quality have positive effects on FDI inflows
Kang and
Jiang (2012)
Chinese FDI in 8
countries in East and
Southeast Asia in
1995–2007
Random effects
estimation
GDP per capita, GDP growth, market
openness, unit labour cost, patent,
economic freedom, FDI restriction,
cultural distance, bilateral trade
GDP growth and GDP per capita are
insignificant. Labour cost has a significant
negative impact. Institutional factors have
significant and diverse effects
Buchanan, Le, and
Rishi (2012)
165 countries in
1996–2006
OLS, IV, Fixed
effects and
Random effects
estimation
Institutional quality Institutional quality has significant positive
effects on FDI levels
Fukumi and
Nishijima (2010)
19 countries in Latin
America and the
Caribbean in 1983–2000
Random effects
estimation, 2SLS
estimation with
fixed effects
Institutional quality, exports Institutional quality and exports have
significant positive effects on FDI
Duanmu and
Guney (2009)
FDI inflows to 30
countries from China
(1999–2002) and India
(2001–2004)
Fixed effects
estimation
Market size, openness, natural resource,
bureaucracy, corruption and political
index
Market size, volume of imports, institutional
environment are important for FDI inflows
from China and India
(Continues)
AN AND YEH 37
TABLE A3 (Continued)
Study Country and Time Methodology FDI determinants Main findings
Buckley et al. (2007) Chinese outward FDI to
22 OECD and 27 other
countries in 1984–2001
Pooled OLS,
Random effects
estimation
Political risk, cultural proximity, natural
resource, GDP, GDP per capita, exports to
the host country, patent.
All the factors have conventional effects on FDI
except for political risk. These effects are not
similar between sub-periods.
Asiedu (2006) 22 African countries in
1984–2000
Fixed effects
estimation
Natural resource, market size, political risk,
institutional quality
Large local markets, natural resource
endowments, good infrastructure, efficient
investment environment have positive effects
on FDI.
Bevan and
Estrin (2004)
FDI from 14 countries to
15 Central and Eastern
European countries in
1992–2000
Random effects
estimation
Gravity factors including distance, unit
labour cost, bilateral trade, country risk
The determinants have expected effects except
for country risk.
Source: Authors' compilation.
38 AN AND YEH
