International technology spillovers from trade and
FDI: a comparison between advanced and developing
countries
University of Groningen
Faculty of Economics and Business
Research paper International Economics and Business
Name Student: Qian Chen
Student ID number: 2147769
Student email: nancyfly28@gmail.com
Date Paper: 17-01-2014
Supervisor: Dr. Sorin Krammer
Contents
Abstract 1
1. Introduction 2
2. literature review 5
2.1 international trade, technology spillover and productivity growth 5
2.2 FDI, technology spillover and productivity growth 8
2.3 institutional quality and technology spillover 11
3. Empirical specification 14
4. Data and methods 21
4.1 Dependent variable 22
4.2 Independent variables 23
4.2.1 Technology spillover via FDI 23
4.2.2 Technology spillover via trade 25
4.2.3 Institutional quality 25
4.3 Control variables 26
4.3.1 Domestic R&D intensity 26
4.3.2 Human capital 27
4.3.3 Government expenditure 27
5. Empirical analyses 28
5.1 Descriptive statistics 28
5.2 Fixed and random effect estimation 30
5.3 Normality 30 5.4 Multicollinearity 5.5 Heteroskedasticity 31 33 5.5 Heteroskedasticity 33 5.6 Regression results 33
5.6.1 Technology spillover from developed and developing countries 34
5.6.2 The role of institutional quality 37
5.7 Robustness checks 39
6. Conclusion 44
Referee 48
International technology spillovers from trade and
FDI: a comparison between advanced and developing
countries
Abstract:
This paper employs panel regression analysis to investigate the effect of
technology spillover via trade and FDI on productivity growth, and considers the role
of domestic institutional quality playing in affecting their relationship. A comparison
of the spillover effects is made between advanced and developing countries. My
findings show that trade spillovers originated from developing countries has a larger
influence on productivity than that from developed countries, and this effect is
enhanced by institutional quality. Besides, although FDI spillovers do not directly
promote productivity growth, it can also benefit productivity by complementing with
domestic institutional quality.
2
1. Introduction:
It has long been recognized that technological change plays an important role in
enhancing economic growth, since it influence the ability of a country to muster and
efficiently use capital resources, accumulate human capital, and achieve technological
improvement through R&D, patent, etc (Lall, 1992). Among these three aspects,
following studies further indicates that a large share of variation in economic growth
is explained by improvement in technology, rather than development of human capital
and financial resources (Hall & Jones, 1999; Easterly & Levine, 2001).
According to Jan, Martin and Bart (2009), technology refers to knowledge about
physical processes as well as how to organize and manage these processes. Previous
studies usually use the difference in technological knowledge to explain productivity
differences across countries and industries. Advanced countries are the main engine of
creating knowledge, however, developing countries usually prefer to acquire them
from foreign countries, since innovation is costly, risky and path-dependent (Fu, et al.
2011), and developing countries have limited resources to invest in R&D and human capital.
Although technology and knowledge are tacit in nature, they could be transferred
across country through many channels, such as human mobility, patenting, licensing.
only two main channels attract the most attention of researches: international trade
and foreign direct investment. As discussed by Costantini & Liberati (2011),
developed countries usually prefer FDI to overcome problems involved in transfer
tacit technology, which requires face-to-face contacts or procedural production
standards. Meanwhile, developing countries mainly regard international trade as a
vehicle of technology transfer from advanced countries. Therefore, the international
diffusion of technology plays a key role in explaining worldwide technological
change, which in turn narrowing the productivity gap between countries. For instance,
for some countries, foreign source of technology contributes to more than 90% of
domestic productivity growth (Keller, 2009).
It is important to notice that how much the recipient country can benefit from the
technology spillover effects of FDI and international trade depends on its institutional
quality. In this study, institutional quality refers to the level of development of formal
institutions in recipient countries, since decision making relative to international trade
and FDI are significantly affected by formal institutions (Jude & Levieuge, 2013;
Kucic, 2012) which are more powerful in guaranteeing exchanges than informal ones
(Luiz, 2009 .
As addressed by previous studies, institutional quality influences the ability of
4 countries from 1984 to 2009, Jude and Levieuge (2013) find that institutional quality
do affect the spillover effects of FDI. Meanwhile, Bhattacharya et al. (2009) provide
the empirical evidence that institutional quality support the positive relationship
between international trade and economic development. However, these studies solely
focus on one channel of technology transfer. According to Krammer (2008), it is
important to regard FDI and trade as complements in the analysis of technology
spillover process.
This paper aims to conduct an empirical analysis towards the mechanism through
which R&D spillovers enhance economic growth, and considers international trade
and FDI as main channels of technology spillover. Since a large amount of studies has
proved that FDI and international trade promote productivity growth through
technology spillover effects, this paper compares the spillover effects between
developing and developed countries, because there is technology take-off in some
emerging economics, which already catch up with the developed countries (Fu, et al.
2008 and they may become new sources of innovation. Thus, the first research question of this paper is that does technology spillover from developed countries has
bigger positive influence on productivity development than that from developing
countries? In addition, with increasing importance of country-specific institution, this
effects of trade and FDI simultaneously, which has not been addressed in previous
literature. In another word, does country-specific institutional quality affect
technology spillover effects of trade and FDI?
The rest of this paper proceeds as follows: section two presents the literature
review of relevant papers and sets hypotheses; section three provides the specification
of the theoretical model; section four introduces the data and methods applied; section
five discusses the empirical results; finally, the last section provides the conclusion,
limits of this paper as well as suggestions for future research.
2. Literature review
Technology spillover can be achieved when domestic firms learn about the
multinationals’ technology. This section briefly discusses some explanations of the technology spillover effects of international trade and FDI on productivity growth, as
well as the role of institutional quality in technology spillover process that is offered
by previous literatures. These researches also provide clues for the adopted indicators
in this study.
2.1 international trade, technology spillover and productivity growth
Theoretically, international trade flows are regarded as one important vehicle of
6 process of international trade, firms in the recipient country get acquainted with the
characteristics of the imported goods, which ensure them to imitate or create a similar
technology. Secondly, international trade ensures the availability of increasing variety
of intermediate goods and capital equipments for domestic productions, which are
embodied in more advanced technologies in the production of final goods (Costantini
& Liberati, 2011; Bulent, 2008). Thirdly, trade is able to enhance transmission of
technical information, such as production processes, product design and managerial
methods, by increasing communication between domestic firms and multinationals
(Krammer, 2008).
As a result, these technologies transferred through international trade contribute
to productivity growth of recipient countries, which is supported by plenty of previous
studies. The initial empirical evidence relative to OECD countries is provided by Coe
and Helpman (1995), who demonstrate that there is a positive and significant effect of
technology spillover via international trade on total factor productivity. In their study,
the technology spillover via trade is measured by import-weighted foreign R&D.
However, their estimates are mainly focus on trade among developed countries. In
1997, a similar result is found by using a larger sample, which indicates that a 1%
increase in the foreign R&D stocks of industrialized countries increases output of the
Furthermore, by using the same methodology as Coe and Helpman, Krammer
(2008 also find a positive and significant relationship between trade spillover and the total factor productivity among developed and transition countries. In addition, Lei
and Bang provide the empirical evidence that international trade remains an important
conduit for technology transfer, which is positively and significantly relative to
productivity growth.
However, the extent to which a country can benefit from international trade
depends on trade partners, who are able to exchange products and knowledge in
which the recipient country is in short supply (Coe, et al. 1997). The past decades
have witness impressively rapid economic growth in some developing countries, such
as Brazil, India and China. These countries have put rising emphasis on innovation on
different extents and with varying success (Fu, et al. 2010). But, there is no doubt that
developed countries still accounts for the largest share of R&D activities, and hold a
larger “stock of knowledge” than developing countries. Hence, developing countries seem to gain more by trading with developed countries, since advanced countries have
more knowledge and information desired by developing countries to acquire.
Based on above analyses, the first hypothesis which is to be tested in this paper
8 𝐻1: Technology spillover via international trade from developed countries has a bigger positive influence on productivity growth than that from developing countries.
2.2 FDI, technology spillover and productivity growth
An alternative and equally important channel for technology spillover is foreign
direct investment. Since governments across the world set more policies and spend
increasing amounts of resources to attract FDI, it is at least partly implies that the
production and research efforts undertaken by multinational affiliates within national
borders bring spillover benefits to domestic economy (Bransetetter, 2000).
Many studies provide some explanations about how technologies are transferred
through FDI. Initially, FDI requires establishment of international production
networks (Chang and Lee, 2009), which benefit domestic firms through backward and
forward linkages, since these linkages ensure local firms learn and imitate advanced
technologies as well as new production processes of foreign firms (Hermes and
Lensink, 2003). For instance, by taking the form of joint venture with Japanese firms,
South Korean firms are able to absorb foreign technology for a variety of consumer
goods and components, which contribute to their amazing functional upgrading of
electronic industry in the following years (Mike, 1995).
provide domestic firms opportunities to benefit from labor turnover. The studies of
Song (2000) indicates that engineers who move from the US to Korea or Taiwan play
a significant role in the technological catch-up of Korea and Taiwan, since they
facilitate spillover of tacit knowledge.
In addition, in order to work with new technologies and deal with increasing
competition from foreign firms, domestic firms are more willing to train their
employees, use alternative management practices and better organizational
capabilities, as well as hire employees from foreign firms (Alfaro, et al., 2004; Chang
and Lee, 2009; Hermes and Lensink, 2003). These trends further facilitate the process
of technology spillover via FDI.
As a result, FDI spillover seems to enhance productivity growth. According to
Borensztein, et al. (1998), technological diffusion of FDI may lead to the expansion of
domestic firms. In other words, in the processes of technological diffusion, both
production and productivity of domestic firms would have a large improvement,
which presents a higher level of productivity in the recipient country.
A large amount of studies have investigated the effects of inward FDI on the
productivity growth of host counties. Although previous researches on the micro level
found either negative or no relationship between FDI and domestic productivity,
10 which is calculated by FDI inflows-weighted foreign R&D, do promote productivity
growth. Besides, by analyzing the impact of US outward FDI on 40 host countries, Xu
(2000) finds that US multinationals contribute to the productivity growth in developed
countries. In addition, by using the FDI flows from G6 countries (the US, Japan, Italy,
the UK and Canada) to OECD countries plus Israel, Hejazi and Safarian (1999)
suggests that technology spillover has a positive relationship with productivity growth,
and this effect of FDI is larger than that of international trade.
However, FDI is not a homogenous capital flow, and some FDI projects are
better than others (Laura and Andrew, 2007), which result in different technology
spillover effects on the recipient countries. For instance, FDI inflows with better
knowledge involved will contribute to higher technology transfer to the recipient
countries (Jude & Levieuge, 2013). Since stock of knowledge in developed countries
is more advanced than in developing countries, it is reasonable to believe that FDI
spillover from developed countries can better contribute to productivity growth.
Hence, the second hypothesis of this paper is developed as follows:
𝐻2: Technology spillover via FDI from developed countries has a larger positive
2.3 institutional quality and technology spillover
Although these two mechanisms of technology transfer may influence economic
performance positively, a country with a good environment is able to benefit more
from the diffusion process. One of crucial factor is the institutional quality condition.
As suggested by North (1990), institutions are the humanly designed constraints that
structure human interactions, which set market rules to organize interaction among
actors and bound economic actions in the market. These rules affect the nature of
competition and process of knowledge acquisition (Meyer & Sinani, 2009).
Institutions can be broadly classified into formal and informal ones. This paper
aims to investigate the role of institutional quality in the technology spillover- growth
relationship by focusing on formal institutions, which includes property rights, the
expropriate risk, and enforcement of law in a country. These factors in the host
country significantly shape the decision making of firms in the process of
international trade and FDI (Jude & Levieuge, 2013; Kucic, 2012). And according to
Luiz (2009), formal institutions are more powerful in guaranteeing exchanges than
informal ones, especially when societies become increasing complex and economies
are more sophisticated.
In the process of international trade, institutional quality influences transaction
12 high transaction cost would exist in countries with bad institutions, since there is
increasing risk for long term trade commitments. As a result, high transaction cost
may decrease total interactions between multinationals and domestic firms, which
reduce the opportunities of technology spillover. Contrarily, well-functioned
institutional environment not only reduces transaction costs, but also decreases
information asymmetries, since good institutions are able to efficiently transfer
information to the market, which in turn enhance technology spillover.
More specifically, in order to benefit from international trade, the country must
reach a threshold level of institutional quality (Bhattacharya et al. 2009), because
formal institutions, such as the degree of control of property rights, the rules of laws,
increase gains from trade, which in turn boost development (Bhattacharya, 2009;
Costantini & Liberati, 2011). Costantini & Liberati (2011) provide the empirical
evidence, which indicate that institutions are a vehicle for facilitating technology
transfer through international trade. Therefore, the next hypothesis emphasizes on the
role of institutional quality in the relationship between trade spillover and productivity
growth.
Institutional quality also impacts the extent of technology spillover via foreign
direct investment. As addressed by Ali (2008), propriety rights, the rule of law and
expropriation risk are the most significant institutional aspects relative to FDI. For
instance, unstable institutional environment with high level of uncertainty, low
protection of investor, inefficient law enforcement, can only attract resource-seeking
FDI with low-technology involved (Jude & Levieuge, 2013), because foreign firms
manage to avoid the risk of technology leakage, and protect their core competences.
Since these invest projects do not contains valuable technology, their contributions of
technology transfer are relatively lower than others.
Furthermore, in a less developed institutional environment where property rights
are poorly protected, foreign firms are exposes to expropriation risks (Kohler, 2010).
In this case, government of the host country may engage in nationalizing foreign firms
or favor local firms at the expense of foreign firms (Ali, 2008). Obviously, the
expropriate risk reduces incentives of foreign firms to invest in the target country.
With less foreign investment, local firms do not have much access to advanced
knowledge.
In addition, institutional quality influences the entry mode of FDI. Bad
institutional quality makes FDI to entry through mergers and acquisitions (Jude &
14 relationship, acquired through mergers and acquisitions are useful for foreign firms to
substitute for the inadequacy of institutions. However, Using a sample of 84 countries
from1987 to 2001, Wang and Wong (2009) provide the empirical evidence that entry
by mergers and acquisitions has a less growth effect than green-field investment. As a
result, institutional quality influences the extent to which technologies can be
transferred from the home to host countries, which in turn moderates the relationship
between technology spillover and productivity growth.
Based on the analysis of FDI spillover, productivity growth and institutional
quality, the last hypothesizes are tested in this paper:
𝐻 Institutional quality alters the spillover effects of foreign direct investment on
productivity growth.
3. Empirical specification
The previous section has highlighted that technology could be transferred via
FDI and international trade, however, the growth effects of trade and FDI from
developed countries are different from that originated from developing countries.
productivity growth from developed and developing countries. The basic model takes
the following form:
1 2 (1
Where i presents the recipient country, t is the time index. is the logarithm of total factor productivity in the recipient country i in year t.
and are the technology spillover to country i via FDI originating from developing and developed countries, respectively. and represents the technology spillover to country i via international trade. is relative to spillovers in trade flows from developing countries, while refers to spillovers in trade flows from developed countries. is a vector of control variables, which includes domestic R&D intensity (RDI), human
capital (HC), and government expenditure (GE). is the error term.
Since there are no indicators for technology spillovers via FDI and trade, it is
computed following the method of Krammer (2008). In his analysis, the spillover
16 recipient country as a percent of total FDI (trade) outflows from the originated
country to the world, times R&D intensity of the originated country. The total
technology spillover via FDI is shown in equation (2), while the total trade spillover is
shown in equation (3).
(∑ ) (2
Where t is the time index, i and q are the index of recipient and originated
country respectively. is the logarithm of total FDI spillovers from originated countries to the recipient country. presents the share of inward FDI
of country i originated from country q in year t as a percent of the total outward FDI
of country q in that year. indicates the R&D intensity of country q in the same
year.
(∑ ) (3
country q in year t as a percent of the total export of country q in that year. Meanwhile,
are set as the same in equation (2).
In order to compare spillover effects between developing and developed
countries, this paper further develop Krammer’s method by computing spillover
variables originated from developing and developed countries respectively. The
calculation of FDI spillovers from developing countries is shown in equation (4).
∑ (4)
Where j is an index of developing countries. presents the share of inward FDI of country i originated from developing country j in year t as a percent of
the total outward FDI of country j in that year. indicates the R&D intensity of country j in the same year.
Besides, equation (5) specifies the calculation of trade spillover from developing
countries to recipient countries.
∑
18 Where j is an index of developing countries. represents the share of import of recipient country i originated from developing country j in year t as a
percent of the total exports of country j in that year. Again, indicates the R&D intensity of country j in year t.
Meanwhile, this paper also interested in technology spillovers originated from
developed countries. The level of FDI spillover from developed countries is estimated
through equation (6).
∑ (6)
Where a is an index of developed countries. presents the share of inward FDI of the recipient country i originated from developed country a in year t as
a percent of the total outward FDI of country a in that year. indicates the R&D intensity of country a in the same year.
Furthermore, equation (7) estimates technological spillovers through trade from
∑ (7)
Where a is the index of developed countries. is the share of imports of country i originated from developed country a in year t as a percent of the
total exports of country a in that year. indicates the R&D intensity of country a in year t.
In the above estimation, the coefficients 1, 2, , and are all expected to
be positive and significant. In order to test the hypothesis that technology spillover via
international trade (FDI) from developed countries has a bigger positive influence on
productivity growth than that from developing countries, this paper also expects:
1< 2 as well as < .
In order to investigate the role of domestic institutional quality in the process of
technology spillover, this paper employs two alternative models, which are presented
in equation (8) and equation (9). Equation (8) focuses on investigate the influence of
institutional quality on FDI spillover-productivity relationship. More specifically, an
20 provide insights about whether institutional quality of recipient country is able to alter
the relationship between FDI spillover and productivity. Meanwhile, equation (9)
replaces the interaction term between FDI spillover variable and institutional quality
by an interaction term between trade spillover variable and institutional quality, and
aims to find whether institutional quality also affects the trade spillover-productivity
relationship. These two equations are developed as follows:
1 2 (8) 1 2 (9)
In both equation (8) and (9), represents the institutional quality of recipient country i. and the other variables are set as the same as in equations (1) –
(3).
In this estimation, if coefficients of the interaction term 2 as well as 2 are
spillovers on productivity are larger for a higher level of domestic institutional quality,
which is stand in line with hypothesis 3a and hypothesis 3b.
4. Data and methods
This paper tends to involve a panel dataset of 37 countries in the period of 1990
to 2010. One reason this time period is chosen lies in that many developing countries’
R&D activities began to surge in the 1990s, thus it becomes interesting to observe
spillover effects originated from developing countries. In this dataset, 21 developed
countries and 16 developing countries are involved: 24 from Central and Eastern
Europe, 6 from Latin America, 5 from Asia. Meanwhile, the United States is from
North America, while New Zealand is from Oceania.
Refers to technology spillovers from developing countries, 16 developing
countries are set as the source for technology spillovers. Besides, when technology
spillover from developed countries is taken into consideration, 21 developed countries
are used for source of spillovers. In these two cases, both the developing and
developed countries in my sample are the recipient of these spillovers. In addition, for
equations (8) and (9) which investigate total technology spillover, each country is
considered as the recipient country, while the other 36 countries are the source of
22
4.1 Dependent variable:
The dependent variable of this paper is the logarithm of total factor productivity
in the recipient country ( ), which is frequently used in previous studies (Coe
& Helpman, 1995; Coe, et al. 1997; Lei & Bang, 2007). Subject to a Cobb–Douglas
production function, the log of TFP is usually defined as the residual from the
aggregated output production function using the country’s stock of capital, labor force and output, as shown in equation (10).
(10)
Where Y is the total output, K is capital stock, and L is total employment. The
parameters and represents the share of capital and labor in country i at year t. In this equation, the capital and labor shares are set as 0.35 and 0.65, which are
widely used in the literature and empirically validate (Krammer, 2008). The data on
GDP and employment comes from the Total Economy Database of the Conference
Board1. Besides, the data on capital stocks are cited from the World Bank’s World
4.2 Independent variables
As discussed above, this study interests in the effects of four variables on
productivity growth: technology spillover of FDI from developing countries
( ), technology spillover of FDI from developed countries ( ),
trade spillover from developing countries ( ), as well as trade spillover from
developed countries ( ). Since the measurement of these variables is
explained in equations (4) – (7), I only introduce the data sources of these variables
here.
4.2.1 Technology spillover via FDI
Equation (4) and (6) are adopted to compute FDI spillovers for each country by
using following sources. Initially, this paper takes Bilateral FDI flows (US dollar,
millions) from the OECD International Direct Investment statistics3 for OECD
countries. However, the data do not cover most of the developing countries. As a
result, the UNCTAD World investment Directory4 and IMF’s Coordinated Direct
Investment Survey5 are adopted as a source of bilateral FDI data for developing
countries. In the case of China, the data are from National Bureau of statistics of
China6. Besides, the data of total outward FDI of each country comes from the
24 UNCTAD Statistics7. In addition, the data of R&D intensity (gross R&D expenditure
as a percent of GDP) are cited from the World Bank’s World Development Indicators8,
supplemented by the Eurostat9 and national statistics.
However, not all of the bilateral FDI inflows are positive between partners. In
this case, this paper only allows for positive spillovers in country i from country j or
country , which is shown in formula (11) and (12). This is due to that if the value of bilateral FDI equals or less than zero, there is zero spillovers caused by disinvestment
4.2.2 Technology spillover via trade
The calculation of trade spillovers is in a similar manner as the FDI spillovers,
which is based on equations (5) and (7). The data of bilateral trade (in US dollar) are
taken from IMF’s Direction of Trade Statistics10. Moreover, the data of total export
(in US dollar, millions) of all the countries comes from the UNCTAD Statistics11.
Before using of these data, they are transferred into the same unit.
4.2.3 Institutional quality
For the alternative models which focus on the role of domestic institutions, the
data of institutional quality is adopted from Institutional Quality database12. In this
database, institutions are classified into legal, political and economic ones. According
to Kucic (2012), legal institutions explain a large part of the formal institutions. As a
result, the absolute value of legal institution ( ) are taken as proxy for institutional quality of the recipient countries.
However, the other two kinds of institutions are also playing important roles in
explaining formal institutions. For instance, political institutions has a wide range of
influences on the rules and limits of the country, while economic institutions are
needed to secure legal system and enforcement of property rights (Kucic, 2012). Thus,
10
http://elibrary-data.imf.org/
11
26 political institution ( ) and economic institution ( ) are adopted as proxies for the institutional quality in the robust check in section 5.7.
4.3 Control variables
This paper includes a series of control variables, which are widely used in
previous studies.
4.3.1 Domestic R&D intensity (RDI)
Domestic R&D expenditure plays an important role in explaining productivity
growth (Gianfranco, et al, 2012). Therefore, this paper adopts domestic R&D intensity
as one of the control variables to correctly estimate the spillover effects from abroad.
As specified before, the data of R&D intensity comes from the World Bank’s World
Development Indicators13, supplemented by the Eurostat14 and national statistics.
4.3.2 Human capital (HC)
Countries differ among themselves in terms of human capital. Human capital is
able to directly affect productivity as a factor of production, and indirectly influence
productivity by facilitating technology transfer (Bulent, 2008). As a consequence, this
13
http://data.worldbank.org/indicator/GB.XPD.RSDV.GD.ZS
14
paper includes human capital as the second control variable. In this case, the tertiary
enrollment as a percent of the gross is used as a proxy of human capital, which is from
the World Bank15.
4. . Government expenditure (GE)
Government expenditure seems to have a significant relationship with economic
growth. Kojo and Yemane (2013) find robust evidence of a long-run positive
relationship between government expenditure and GDP. However, James and (2011)
address that government expenditures are estimated to have negative growth effects
for some developed countries. Although the relationship between government
expenditure and economic growth is still a subject of debate, this paper would like to
control for the possible impact of government expenditure on total factor productivity.
The data of government consumption as a share of GDP is adopted from the Penn
World Table16 8.0.
. Empirical analyses
This section provides descriptive of the variables, which provide the information
on the characteristics for all the data in the sample. Moreover, the diagnostic checks
are conducted in order to check the relationship between variables and the validity of
the model. In addition the regression results of the estimation as well as robust check
are presented.
15
28 5.1 Descriptive statistics
Table 1 Descriptive Statistics
Variables Description Obs. Mean Minimum Maximum Std. Dev
Log total factor productivity 777 5.355 3.678 7.412 0.683
Log FDI spillover from developing countries 269 -2.082 -13.435 4.814 2.620
Log FDI spillover from developed countries 415 -0.489 -6.267 4.892 1.595
Log trade spillover from developing countries 410 -2.338 -6.448 0.403 1.264
Log trade spillover from developed countries 656 -0.771 -4.451 14.633 2.691
Log total trade spillover 411 -0.666 -3.796 2.245 1.250
Log total FDI spillover 273 -0.102 -6.266 4.950 1.685
Legal institution quality 795 0.708 0.288 1 0.172
Political institution quality 766 0.687 0.172 0.928 0.166
Economic institution quality 768 0.617 0.034 0.893 0.173
Log total trade spillover weighted legal institution quality 411 -1.053 -3.925 2.010 1.320
Log total trade spillover weighted polotical institution quality 411 -0.074 -3.979 1.981 1.300
Log total trade spillover weighted economic institution quality 411 -1.107 -4.230 2.020 1.312
Log total FDI spillover weighted legal institution quality 273 -0.414 -6.389 4.637 1.679
Log total FDI spillover weighted political institution quality 273 -0.435 -6.431 4.657 1.677
Log total FDI spillover weighted economic institution quality 273 -0.481 -6.644 4.603 1.699
Human capital 716 46.732 2.847 103.873 21.301
Government expenditure share of GDP 777 0.188 0.059 0.459 0.063
Domestic R&D intensity 739 1.410 0.010 5.887 1.023
Source: Author’s own calculation, the World Bank, the Penne World table 8.0, the Eurostat and national statistics.
As shown in table 1, the variation of the dependent variable is not obvious,
which might due to that the logarithm form of total factor productivity is used in this
analysis. Among the sample countries, the United States outperforms other countries
with the highest number of (7.412) in 2010, however, Romania shows the
In terms of spillover variables, the number of observations falls a lot, especially
relative to FDI spillover from developing countries (269 observations of ).
This is due to the limitation of availability of bilateral FDI and R&D intensity data.
However, there is relatively large level of variation of the spillover variables. For
instance, the number of varies from -13.435 in Iceland at 2000 to 4.814 in
Spain at the year 2010. Besides, Iceland also received the lowest level of FDI
spillover from developed countries ( -6.267), while the largest number of
FDI spillovers originated from developed countries goes to Japan in 2005
( 4.892). As to trade spillovers, ranges from -6.448 (Iceland,
2008) to 0.403 (US, 2003), whereas Ukraine received the lowest level of trade spillover from developed countries in 1994 ( -4.451), and much of this term is distributed to France in 2000 ( 14.633). It is obvious that the largest numbers of spillovers are all distributed to developed countries, which might
contribute to their higher than developing countries. In addition, the interaction terms of institutional quality and spillover variables presents relatively
large variation.
5.2 Fixed and random effect estimation
The panel estimation can be conducted via either random or fixed effects. The
30 estimation assume that all the differences within countries are captured by the
intercept parameters, but regard the differences between countries as random; while
the fixed effect model assume that the intercept captures all the differences between
countries overtime. In this paper, both random and fixed effect estimation are adopted.
5.3 Normality
Hypothesis test and estimation for the coefficients rely on the assumption that the
dependent variable is normally distributed (Hill, et al. 2011). In order to estimate the
distribution of dependent variable, the Kolmogorov-Smirnov Test is used in this paper.
In this test, if the p-value is larger than 0.05, then the dependent variable is normally
distributed. By conducting the Kolmogorov-Smirnov Test in the Stata, the p-value is
observed as 0.128, which implies that the distribution of the dependent variable is
approximately normal. The normality distribution of can also be observed in figure 1. Figure 1 Distribution of lnTFP 0 20 40 60 80 1 0 0 F re q u e n cy 4 5 6 7 8
log total factor productivity
5.4 Multicollinearity
Before the regression, another concern is the multicollinearity problem, which
happens when two or more explanatory variable have a linear relationship (Hill, et al.
2011). It is a major obstacle in the estimation, since this problem will cause bias of
coefficients. In this case, the correlation matrix is used to investigate whether
variables are potentially correlated with each other. More specifically, if the absolute
value of the correlation between each pair of explanatory variables are larger than 0.8,
the multicollinearity problem occurs (Hill, et al 2011).
Table 2 Correlation matrix
variables 1.000 0.310 1.000 0.381 0.627 1.000 0.645 0.545 0.474 1.000 0.530 0.268 0.343 0.429 1.000 0.142 -0.229 -0.173 -0.147 -0.134 1.000 -0.471 0.089 0.141 -0.086 0.046 -0.144 1.000 0.375 -0.117 0.038 0.073 0.201 0.553 -0.050 1.000 0.736 0.494 0.558 0.853 0.535 -0.254 -0.049 0.170 1.000 0.390 0.797 0.935 0.508 0.321 -0.180 0.136 -0.005 0.558 1.000 0.191 -0.163 -0.069 -0.264 0.063 0.508 -0.069 0.632 -0.099 -0.080 1.000 0.053 -0.144 -0.099 -0.307 -0.025 0.574 -0.003 0.471 -0.259 -0.096 0.908 1.000 0.191 -0.067 0.038 -0.192 0.065 0.492 0.037 0.447 -0.057 0.025 0.880 0.880 1.000
32 As shown in the correlation matrix (table 2), in the basic model and the
alternative models, none of the explanatory variables are significantly correlated to
each other.
Another approach is to use the Variance Inflation Factor (VIF). In this estimation,
if the value of VIF is larger than 10, there is a linear relationship within explanatory
variables and multicollinearity exists (UCLA, 2010). The results are shown in the last
row of table 3 and table 4 in section 5.6. The value of VIF is less than 10 in the basic
model, but extremely high in the alternative models (column [6] and [7]). However,
this is due to that an interaction term is added in these two estimations. Hence, the
results of this study will not be interrupted by the multicollinearity problem.
5.5 Heteroskedasticity
Heteroskedasticity exist when the variances for all observations are not the same.
It would result in biased standard errors, which in turn lead to that the confidence
intervals and hypothesis tests that use these standard errors are misleading.
The White test is employed to provide some statistical evidence about
which means that the null hypothesis of this test is rejected and the estimations of this
paper are biased by heteroskedasticity. In this case, the significance level of the
explanatory variables will be under- or over-estimated, which result in incorrect
understanding of the relationship between explanatory variables and the dependent
variable.
In order to provide solution for the heteroskedasticity problem, the White robust
standard errors will be used, which is helpful to correct the P-values and standard
errors in the regression (Hill, et al. 2011)
. Regression results
The regression results are presented in table 3 and table 4. Table 3 presents the
empirical results by using random effect estimation, while table 4 shows the results of
fixed effect estimation. In both table 3 and table 4, the values of R-square are all
larger than 0.60, which means that these models have good explanatory power, since
more than 60% variation of the dependent variable is explained by variables in the
right side of the equations.
. .1 Technology spillover from developed and developing countries
Firstly, columns [1]-[4] show the results of the basic model (equation 1), which
34 Specifically, column [1] emphasizes on the effect of technology spillover from
developed countries on productivity. The coefficient of is positive and
significant relative to total factor productivity in both random and fixed effect
estimation. Meanwhile, when technology spillover from developing countries is
considered instead of that from advanced countries, as shown in column [2], the
coefficient of ln is positive and significant at the 1 percent level. In addition,
when technology spillovers originated from developing and developed country are
pooled together (column [4]), the coefficient of and ln still
remain positive and significant. This is the evidence that there is a strong positive
relationship between trade spillovers and productivity.
However, it is surprise to find that the coefficients of ln is larger than that of , as shown in column [4], which means that the influence of trade
spillover originated from developing countries are larger than that from developed
Table 3 Estimation results (Random Effects)
Basic model Alternative model 1 Alternative model 2
Equation (1) Equation (8) Equation (9)
Variable 1 2 3 4 5 6] 7 ln 0.002 0.003 (0.004 (0.004 ln 0.158 0.162 (0.053 (0.055 ln 0.010 0.007 (0.004 (0.007 ln 0.010 0.011 (0.006 (0.005 0.004 0.008 0.005 0.005 0.004 0.004 0.004 (0.002 (0.001 (0.001 (0.001 (0.001 (0.001 (0.001 0.034 0.093 0.182 0.181 0.158 0.152 0.152 (0.030 (0.040 (0.060 (0.059 (0.046 (0.043 (0.043 -0.033 -1.15 -1.343 -1.553 -1.172 -1.135 -1.135 (0.028 (0.766 (0.731 (0.727 (0.609 (0.624 (0.624 0.006 -2.118 0.007 (0.007 (1.025 (0.006 0.245 0.229 -1.896 (0.039 (0.037 (1.014 -3.077 -3.077 (1.609 (1.609 2.125 (1.025 2.125 (1.025 constant 5.074 5.181 5.601 5.680 5.437 8.403 8.403 (0.129 (0.152 (0.210 (0.205 (0.134 (1.566 (1.566 Obs. 690 355 215 206 217 217 217 2 0.685 0.649 0.622 0.625 0.613 0.651 0.651 VIF 1.000 1.120 1.330 1.460 1.300 1516.200 739.350
Note: The dependent variable is logarithm of total factor productivity. Figures in parentheses are standard deviations. ***, ** and * indicate significant at the 1 percent, 5 percent and 10 percent levels, respectively.
36
Table 4 Estimation results (Fixed Effects)
Note: The dependent variable is logarithm of total factor productivity. Figures in parentheses are standard deviations. ***, ** and * indicate significant at the 1 percent, 5 percent and 10 percent levels, respectively.
Source: Author’s own calculation, the World Bank, the Penne World table 8.0, the Eurostat and national statistics.
Basic Model Alternative model 1 Alternative model 2
Equation (1) Equation (8) Equation (9)
Besides, relative to FDI spillovers of the basic model, the results are ambiguity.
In column 2 of table 3 the coefficient of ln is observed to be positive and
significant, however, the coefficients of other FDI spillover variables are all
statistically insignificant. As a result, this paper does not find strong empirical
evidence to prove that technology spillover via FDI from developed countries has a
larger positive influence on productivity growth than that from developing countries,
and hypothesis 2 is not supported in general.
5.6.2 The role of institutional quality
Column [5]-[7] present the empirical evidence relative to how institutional
quality of recipient countries influence the relationship between technology spillover
and productivity. Columns [6] are relative to the alternative model 1 (equation 8),
which is interested in the role of institutional quality in FDI spillover-productivity
relationship. In addition, columns [7] investigate the influence of institutional quality
on the relationship between trade spillover and productivity, as described in the
alternative model2 (equation 9).
Initially, according to column [5] of both tables, without considering the role of
institutional quality, the coefficients of total FDI spillovers ( ) are statistically
insignificant relative to total factor productivity. However, as shown in columns [6],
38 ( ) is positive and significant, which indicates that foreign FDI becomes beneficial for recipient countries when the level of domestic institutional
quality is high. Thus, hypothesis 3a is confirmed.
Moreover, column [5] also presents that regardless of institutional quality, the
coefficient of equals to 0.245 in the random effect estimation and 0.157 in
the fixed effect estimation. Both of them are statistically significant. However, as
shown in column [7], the coefficient of interaction term between trade spillover and
institutional quality ( ) is significant and increase to 2.125 in table 3 and 2.526 in table 4. This is the evidence that the effects of trade spillover on productivity will be larger when recipient country has a higher level of institutional
quality. Hence, hypothesis 3b is confirmed.
Refers to control variables of this paper, human capital has a significant and
positive influence on productivity. Moreover, as demonstrated by previous studies,
domestic R&D intensity remains an important factor for growth, since most of the
coefficients of are positive and significant. In addition, this paper observes an ambiguity relationship between government expenditure and productivity, since in the
random effect estimation, the coefficients of are significant and negative, while that in the fixed effect estimation is statistically insignificant.
The next step is to check the robustness of above result. For the basic model, this
paper tends to do this by adding a one-year lag to variables in the right-side of the
equations. There are mainly three reasons.
Firstly, relative to trade spillover, domestic firms may need a period of time to
imitate and learn foreign technologies and apply it into production; in the process of
FDI spillover, there is a lag of anywhere from one to four years before new ventures
become profitable (Kentor 1998). In another word, new investment projects do not
usually generate profits immediately. Secondly, according to Borensztein, et al.
(1998 human capital and FDI is considered to be complementary in the process of
technology diffusion, thus human capital needs to be taken in analysis within the same
time as FDI, and this mechanism can also be applied in domestic R&D intensity and
government expenditure. Thus, all control variables are applied a one-year lag.
Thirdly, lagged values can be used to deal with the potential endogeneity problem
(Fritsch, 2002).
With a one-year lag, the basic model shift to equation (13).
1 1 2 1 1
1 1 (13)
40 The results of random effects estimation for equation (13) are shown in columns
[1]-[4] of table 5. In this estimation, the effects of FDI and trade spillovers as well as
control variables remain the same signs and similar significant level as the results
observed without using lagged variables.
Table 5 Robustness check estimation results (Random Effects)
Basic model Alternative model 1 Alternative model 2
Equation (13) Equation (8) Equation (9)
1.537 (0.780 1.625 (0.768 constant 5.027*** 5.157*** 5.582*** 5.651*** 5.437 7.930 8.076 7.930 8.076 (0.136) (0.156) (0.227) (0.222) (0.134 (1.294 (1.162 (1.294 (1.162 Obs. 690 355 215 206 217 217 217 217 217 2 0.651 0.677 0.635 0.640 0.613 0.628 0.657 0.628 0.657 VIF 1.000 1.120 1.330 1.460 1.300 604.740 2029.920 284.900 971.360
Note: The dependent variable is logarithm of total factor productivity. Figures in parentheses are standard deviations. ***, ** and * indicate significant at the 1 percent, 5 percent and 10 percent levels, respectively.
Source: Author’s own calculation, the World Bank, the Penne World table 8.0, the Eurostat and national statistics.
However, there are significant differences in the fixed effect estimation relative
to the effect of trade spillover on productivity, which presents some clues for why the
impact of trade spillover from developing countries is larger than that from developed
countries, as obtained in table 4.
According to columns [2]-[4] of table 6, when a one-year lag is applied in
regression, the coefficient of trade spillover from developing countries (ln ) becomes insignificant, while that from advanced countries (ln ) becomes positive and statistically significant. According to Timothy, et al. (2012) in the
process of fragmentation, developing countries concentrate on low value-added and
low-technological activities, while developed countries specialized in high-skill and
42 developing countries are relatively easier to adopted and imitate in the short term,
while technology originated from advanced countries requires more complex learning
process and significant adaptation. Thus, when this study does not consider lag values,
trade spillover from developing countries seems to benefit productivity more,
however, when a lag value is involved in this study, trade spillover from developed
countries are demonstrated to have a significant and positive influence on productivity
growth.
The estimated results of FDI spillovers variables are similar as in table 3 and
table 4. These findings are quite different from previous studies which demonstrate
that FDI is a significant channel for technology spillovers. This might be due to that
distinct country sample and time period are used in the analysis. Another explanation
lies in that the benefit effect of FDI is influenced by other factors, such as the entry
mode of FDI (Wang and Wong, 2009) and the development of financial system in the
recipient countries (Hermes and Lensink, 2003), which is beyond the scope of this
research.
Table 6 Robustness check estimation results (Fixed Effects)
Basic Model Alternative model 1 Alternative model 2
Equation (13) Equation (8) Equation (9)
Variable 1 2 3 4 5 6 7 8 9
ln -0.000 0.000
(0.006) (0.004
Note: The dependent variable is logarithm of total factor productivity. Figures in parentheses are standard deviations. ***, ** and * indicate significant at the 1 percent, 5 percent and 10 percent levels, respectively.
44 In the robustness check for the alternative models, this paper uses alternative
indicators for institutional quality, as specified in section 4.2.3. In this case, legal
institution ( ), is replaced by political institution ( ) and economic institution ( ). The results are shown in columns [6]-[9] of
table 5 and table 6.
The results for the estimated effects of FDI and trade spillovers are robust and
similar to the ones obtained by using legal institution ( ). More specifically, when and are adopted in the estimation, their interaction term with FDI spillovers ( ,
) are positive and significant. Besides, the coefficients of
interaction term with trade spillovers ( ,
) are also significant and become larger.
6. Conclusion
During the past few decades, many researchers have paid a lot of attention on the
relationship between technology spillover and productivity development, since in the
trend of globalization, foreign advanced technologies become increasing important,
Since there are significant differences between countries in term of economic
development, quantity and quality of FDI and international trade as well as
institutional quality, this paper focus on investigating whether technology spillover
originated from developed countries has bigger positive influence on productivity
development than that from developing countries. Besides, this paper also aims to
provide some empirical insight about how domestic institutional quality influences the
spillover effects of trade and FDI.
Using a panel dataset of 37 countries from 1990 to 2010, this paper
observed that trade spillover have a robust positive relationship with productivity,
which is similar with previous studies (Coe & Helpman, 1997; Krammer, 2008).
However, in our estimation, trade from developing countries has a larger spillover
effects than that from developed countries, which might be due to that among trade
(FDI) partners in my sample, the technologies originated from developing countries
are easier for other countries to adapt in production than that from advanced countries,
especially in a strong trend of fragmentation.
Furthermore, this paper does not find a strong positive and significant
relationship between FDI spillover and productivity, which is different from previous
studies (Hejazi & Safarian, 1999; Xu, 2000; Krammer, 2008). This might be caused
46 important factors, such as the entry mode of FDI (Wang and Wong, 2009) and the
development of financial system (Hermes and Lensink, 2003).
Moreover, institutional quality of the recipient country is a critical factor in
spillover process since it is able to enhance technology spillover of trade and FDI. In
addition, my findings also confirmed that human capital and domestic R&D
expenditure plays an important role in explaining productivity level.
However, this paper is marked my several important limitations. A considerable
limitation is the lack of data source for bilateral FDI and R&D intensity, especially for
developing countries. Under this situation, our sample size is limited, which may also
prevent this study to observe a significant relationship between FDI spillover and
productivity. Another limitation is about the measurement of technology spillover. In
this study, technology spillover is measured by FDI- and trade-weighted R&D
intensity. However, since there is no standard measurement of spillover, it is hard to
judge whether this measurement can fully capture the amount of technology
spillovers.
Future studies can contribute to technology spillover issue by further distinguish
types of countries to advanced countries, emerging countries and poor countries,
because as addressed by Gary and Karina (2011), advanced countries and emerging
poor countries. Furthermore, another way to distinguish countries is to group them
within or out of a trade agreement, since the flow of trade and FDI seems to have
different pattern towards countries in and out of a trade agreement.
Referee:
Alfaro, L., Chanda, A., Ozcan, K. S., & Sayek, S. 2004. FDI and economic growth: the role of local financial markets. Journal of International Economics, 64: 89-112.
48 Bhattacharya, S., Dowrick, S., Golley, J. 2009. Institutions and trade: competitors or complements in economic development? Economic Record, 85: 318-330.
Borensztein, E., Gregorio, D. J., & Lee, W. J. 1998. How does foreign direct investment affect economic growth? Journal of International Economics, 45: 115-135.
Bransetetter, L. 2000. Is foreign direct investment a channel of knowledge spillovers?
Evidence from Japan’s FDI in the United States. NBER working papers, no. 8015,
Cambridge.
Bulent, U. 2008. R&D spillovers through trade in a panel of OECD countries. The
journal of international trade and economic development, 17(1 : 105-133.
Chang, C. P., & Lee, C. C. 2009. FDI financial development and economic growth: international evidence. Journal of Applied Economics, 12(2 : 249-271.
Coe, D., & Helpman, E. 1995. International R&D spillovers. European economic
review, 39: 859-887..
Coe, D., Helpman, E., & Hoffmaister, A. 1997. North-South R&D spillovers. The
economic journal, 107(440 : 134-149.
Costantini, V., & Liberati, P. 2011. Technology transfer, institutions and
development. Working paper no.135, Roma.
Easterly W., & Levine, R. 2001. It is not factor accumulation: stylized facts and growth models. World bank economic review, 15 (2 :177-219.
Fritsch, M. 2002. Measuring the quality of regional innovation systems: a knowledge production function approach. International regional science review, 25: 86-101.
Fu, X. 2008. Foreign direct investment, absorptive capacity and regional innovation capabilities: evidence from China. Oxford development studies, 36(1 : 89-110.
Fu, X., Pietrobelli, C., & Soete, L. 2011. The role of foreign technology and indigenous innovation in emerging economies: technological change and catching up.
Gary, G., & Karina, F. 2011. Global value chain analysis: a primer. CGGC, Duke university, Durham.
Gianfranco, p., Paolo, P., & Patricio, A. 2012. Specialization, R&D and productivity growth: evidence from EU regions. Ann Reg Sci, 49: 35-51.
Hall, R., & Jones, C. 1999. Why do some countries produce so much more output
per worker than others? NBER working papers, no. 6564.
Hejazi, W., Safarian, A. 1999. Trade, foreign direct investment and R&D spillovers.
Journal of international business studies, 30(3 : 491-511.
Hermes, N., & Lensink, R. 2003. Foreign direct investment, financial development and economic growth. Journal of Development Studies, 40(1 : 142-163.
Hill, R., Griffiths, W., & Lim, G. 2011. Principles of econometrics 4th edition, New
York: John Wiley & Sons, Inc.
James, B., Halit, Y. 2011. Institutions and the impact of government spending on growth. Journal of applied economics, 14(2 : 319-341.
Jan, F., Martin, S., & Bart, V. 2009. Innovation and economic development. TIK working papers, no. 20090723, Norway.
Jude, C., & Levieuge, G. 2013. Growth effect of FDI in developing economies: the
role of institutional quality.
Keller, W. 2009. International trade, foreign direct investment, and technology
spillovers. NBER working paper, no. 15442, Cambridge.
Kentor, J. 1998. The long-term effects of foreign investment dependence on economic growth, 1940-1990. American Journal of Sociology, 103:1024-1046.
50 Kohler, P. 2010. Foreign direct investment in countries with weak institutions. http://www.ecipe.org/media/publication_pdfs/Kohler_foreign_direct_investment03 2010.pdf
Kojo, M., & Yemane, R. 2013. Government expenditure and economic growth: the Ethiopian experience, 1950-2007. The journal of developing areas, 47(1 : 263-280.
Kuncic, A. 2012. Institutional quality database. Working paper no. 457, Kiel institute for the world economy, Germany.
Lall, S. 1992. Technological capabilities and industrialization. World development, 20 (2 : 165-186.
Laura, A., Andrew, C. 2007. Growth and quality of foreign direct investment: is all
FDI equal? Working paper, no. 07-072.
Lei, Z., & Bang, J. 2007. International R&D spillovers: trade, FDI and information technology as spillover channels. Review of international economics, 15(5 : 955-976.
Luiz, J. 2009. Institutions and economic performance: implications for African development. Journal of international development, 21: 58-75.
Meyer, K., & Sinani, E. 2009. Where and when does foreign direct investment generate positive spillovers? A meta analysis. Journal of international business
studies, 40: 1075-1094.
Mike, H. 1995. East Asia latecomer firms: learning the technology of electronics.
World development, 23(7 : 1171-1193.
North, D. 1990. Institutions, institutional change and economic performance. Cambridge University press.
Timothy, S., Peter, N., Greg, L., Gary, G., & Clair, B. 2012. Direct measurement of
global value chains: collecting product- and firm- level statistics on value added and business function outsourcing and offshoring. Chapter 9, World bank volume: