by Wenn Jinn Cheah^[1], Department of Economics, University of Warwick

Abstract

Recent trade literature empirically demonstrates that service export sophistication is an alternative key driver to boost economic growth. This suggests that the upgrading of service export sophistication is a potential policy tool to reduce gaps in the economic performances of countries of different levels of development. This article adopts a 'service export sophistication' model and seeks to test if Foreign Direct Investment (FDI) as a determinant significantly affects service export sophistication. A 10-year panel data regression, controlled for year fixed-effects, is used for 86 countries, including advanced, developing and less-developed economies. Interestingly, the current study finds that FDI per capita does have a significant long-term non-linear impact on service export sophistication. The gradual diversion of FDI from traditional to modern services results in a trade-off in degrees of specialisation between these services. The impact of diversion of FDI on specialisation, however, can be influenced by domestic facilitating factors such as absorption capacity and openness of recipient countries. These factors in turn affect countries' ability to adapt to the spillovers, leading to the fall and rise in overall service export sophistication at different levels of FDI per capita.

Keywords: FDI, aggregate productivity, service export sophistication, panel regression, non-linear, services

 

Introduction

A large body of international trade literature shows that export performance is essential for economic development (Kaur and Sidhu, 2011; Sun and Heshmati, 2010). It is unsurprising that global merchandise trade has been experiencing rapid growth (Figure 1) since merchandise trade has always been a major component of international trade. What should come to our attention is the growth in global services trade which saw a jump from an average rate of 6.2% in the 1990s to 13.4% in the 2000s implying that services trade is no less important than merchandise trade. Schneider and White (2004: 16) find that many current advanced economies - including the US, Japan, Spain, Canada, the UK, France, and Australia - shifted focus from manufacturing to high-value-added services sector during the 1980s while Banga (2005: 55) observes that services often account for a considerable share of local employment. The optimistic growth in services trade somewhat reflects the emergent role of services in economic development both on the domestic and international platform.

Figure 1: The difference in average growth rate for merchandise and services trade in the 1990s and 2000s plotted based on the trade data obtained from World Trade Organisation (2013)

Figure 2: Comparison of the growth rate of traditional and modern services

Looking closely at the growth trend of services export (Figure 2) since the early 2000s, it is apparent that modern services have been growing faster than traditional services before the global financial crisis hit in the last quarter of 2008. The drastic plunge in modern services growth rate was largely due to financial services. Nonetheless, modern services have been showing signs of recovery in the recent years.

In their widely cited paper Hausmann, Hwang and Rodrik (2005) point out that, rather than volume of export per se, the diversification and sophistication of service export matter too. They argue that positive externalities generated from the 'cost-discovery' process (where many firms start to enter a sunrise industry and discover new cost-reduction means of production) could increase productivity and vary the basket of exports. On the aggregate level, this process reveals the service production possibility frontier of the economy as a whole and is reflected by the ability of firms to export services for which they are highly productive. Subsequently, countries that are capable of exporting large shares of services export associated with higher levels of productivity will have a higher implied overall service export sophistication.

Given the growing importance of services and possibly export sophistication, recent World Bank and IMF empirical papers justify that service export sophistication proves to be an 'alternative key driver' to boost economic growth. Mishra, Lundstrom and Anand (2011) find that the results hold when high-income countries are dropped by applying the dynamic system GMM model while Anand, Saurabh and Spatafora (2012) support the results in similar research using co-integrated panel data. This suggests that upgrading of service export sophistication is a potential policy tool to reduce the gap in economic performance between advanced, developing and least-developed countries in the long term. This provides great motivation to examine factors that can possibly affect service export sophistication for optimum utilisation of this potential tool. Thus far, little literature has examined determinants of service export sophistication based on Hausman, Hwang and Rodrik's export sophistication model. Therefore this article seeks to extend on the World Bank paper and to contribute new findings to the literature by studying the impact of FDI on service export sophistication and subsequently check if the results hold for separate samples of advanced and non-advanced countries.

Literature Review

Global Trends of FDI

Spurred by the fast-paced growth of international services trade, the diversion of Foreign Direct Investment (FDI; referring to long-term international investment received by a resident entity or enterprise in a host country from a foreign resident or enterprise as defined by IMF) from the manufacturing to the service sector is evident; the services sector accounts for over 70 per cent of global output (Anand et al., 2012: 5).

The global share of services FDI in total stocks increased to approximately 60% as of 2002 from a quarter in the 1970s (Banga, 2005: 55). In recent years, FDI stocks and assets continue to grow at a modest rate (UNCTAD, 2010: 2) amid uncertainties spawned by the sovereign debt and global financial crises. This literature review aims to establish a framework to analyse the potential role of FDI in enhancing service export sophistication based on classes of FDI - including efficiency, resource, strategic-asset and market-seeking FDI (Dunning, 2000) - trade theories, and eclectic paradigm. Eclectic paradigm is a widely accepted theoretical framework combining economic and business theories to explain the motivation behind foreign firms' engagement in FDI based on three sub-paradigms, namely ownership-specific, location and internalisation (OLI) advantage (Dunning, 2000).

FDI and Export Sophistication

Efficiency-Seeking FDI

Dunning (2000) finds that multinational corporations' (MNC) activities are designed to 'encourage division of labour and specialisation' to increase efficiency (efficiency-seeking), hence productivity. This suggests superior ownership-specific advantage (trademark, property rights to exclusive management know-how, marketing, innovation, entrepreneurship, technology and knowledge expertise; see Banga (2005: 60) and Twomey (2000: 8) over firms in host countries may generate positive externalities through joint ventures.

Comparing the effects of FDI between advanced and emerging economies, Lesher and Miroudot (2008) adopted a dynamic panel data regression using 15 OECD cross-country firm-level data from 1997 to 2006; they found that service sectors enjoy significant services 'productivity-enhancing effects' by means of technological and knowledge spillovers, particularly through backward linkages (these measure 'spillover effects on a producer industry from foreign presence in downstream sectors'; see Hamida and Gugler (2009) for Switzerland) while Terrell et al. (2007) conducted panel pooled OLS regressions using firm-level BEEPS data for 17 emerging economies and proposes that only larger and longer-established firms benefit from FDI. These evidence suggests that an improvement in service export sophistication is possible subject to there being a small technological gap between indigenous and foreign firms (Hamida and Gugler, 2009). On the contrary, Hale and Long (2007: 2) argue there is no significance of spillovers from FDI in China based on total factor productivity model because private firms refrain from adopting new technologies due to 'formal labour market restrictions and limited access to credit'. This raises doubts on the extent to which developing and less-developed economies could benefit from positive externalities of FDI given fragile institutions and inadequate skilled labour force. Bitner et al. (2010) observe that while evolution of technology altered services' medium of conduct ('Self-service or technology-assisted services'; Bitner et al., 2010: 204), the disinclination of some people (consumers or employees) for technology hinders application of advanced technology such as in healthcare services. This suggests that the influence of technological spillovers may be weaker on sophistication of services export which requires more physical proximity.

Resource-Seeking and Market-Seeking FDI

Natural resources and factor endowments are location-bound advantages unique to each country. Although labour is highly mobile, MNCs still appreciate the low-cost advantages of skilled English-speaking labours within developing countries like India. Resource-seeking MNCs capitalise on these features of a host country they find deficient in their home country to augment their competitive advantage. The capitalisation of these advantages could increase allocative and productive efficiency, move the host country towards its production frontier, simultaneously enhancing export sophistication. A case study by UNCTAD (2007) on tourism services FDI in developing countries such as Peru, Morocco and India seems to exhibit such pattern. The publication concludes MNCs has begun delving into high value-added niche markets (cultural, theme-parks, sports or other tourism features specific to the country (UNCTAD, 2007: 122)), subsequently creating 'sustainable infrastructures and organisational capabilities' through linkages with domestic firms affiliated with tourism to serve the local market (market-seeking).

Strategic Asset-Seeking FDI

Internalisation advantage suggests that MNCs internalise 'cross-border intermediate' markets through vertical or horizontal integration to utilise specific advantage within the firm at the expense of external firms (Dunning, 2000). However, one can query that positive externalities are concealed within the chain of integrated firms rather than benefiting the wider market. Consequently, service export sophistication may not be enhanced as much as it should. Rather than internalisation, some MNCs conduct service offshoring (defined by Trefler (2005: 2) as 'the use of high-skilled workers abroad to provide sophisticated services to local customers') in low-cost countries such as India and China (Trefler, 2005). Trefler explains that MNCs invest in external firms abroad to administer fragments of their operations attributable to trivial transaction costs coupled with the increase in tradability of services driven by ICT progression. This initiative helps develop host countries' service export sophistication, especially in computer and information (software-related) and business services exports.

In summary, the literature review suggest mix results but most indicate FDI positively impacts implied overall service export sophistication based on firm-level productivity and FDI linkage regression models as well as economic case studies. Moving forward, the current article will study the possible impact of FDI on aggregate 'productivity' level reflected through services export as a proxy for overall service export sophistication.

Dataset

The panel dataset is constructed manually using annual data values compiled from various reliable sources such as World Trade Organisation, World Development Indicator and United Nations Development Programme Database as there is no relevant existing dataset to assist the current research. Due to data limitations, the sample consists of only 86 countries of advanced (33), developing (40) and least-developed (13) nations (see Annex A1 for the full list of countries). The time period of study spans across one decade between 2002 and 2011, mainly because the method of recording services export values based on various mode of supply was newly introduced and adopted by 'international statistical community' beginning 2002 (see World Trade organisation, n.da.). Consequently, disaggregated modern and traditional services export data are more standardised across countries post-2002.

The Service Export Sophistication (EXPY) Index is first calculated based on the empirical export sophistication model developed by Hausmann, Hwang and Rodrik (2005). The construction of the model is explained in Annex A4. Values of service export and GDP per capita are used to generate the EXPY index. The export categories of commercial services include travel, transport, construction, communication, financial, insurance, computer and information, royalties and patents, and personal, culture and recreation services (See Annex A3 for service export definitions ). Given the intangible nature of services, there are some unfilled gaps in the reported data due to difficulties in data collection. In addition, revisions made by individual-specific countries resulted in breaks and inconsistent reporting in certain years (see World Trade organisation, n.da.). Although there are limitations to the service export data, these are nonetheless the most reliable values we can utilise to generate the index. All export values are deflated using US GDP deflator (constant at 2005).

FDI stock per capita is deflated using US GDP Deflator (constant at 2005) and controls for economy size. Francois, Pindyuk and Woerz (2009: 5) find that FDI stock per capita is better than flows for analysis since the former is 'less volatile' than the latter. Controlled variables in the dataset include real GDP per capita, total population, Human Development Index (HDI), Information Flow Index, and modern services exporter dummies. Annex A5 describes data definitions and sources while Annex A6 describes the form (binary or continuous) of these variables in detail. Before proceeding to the following sections, it is highly important to understand the construction of EXPY index.

Methodology

Variables Specification

All regression specifications include variables which take the log form of service EXPY index (lexpyo), real FDI stock per capita (lfdipc), real GDP per capita (lgdppc) and population (lpop) respectively. These continuous variables are logged for better linear fit and minimise the impact of outliers on EXPY. Dummy variables include HDI (hdi), information flow index (infl) and modern services exporter (modex). The variable specification labels in the regression equation are denoted in the parentheses.

Log Linear Functional Form

Mallows (1986) show that Augmented Component-Plus-Residual Plot (ACPR) can detect non-linear relationship between variables more accurately as the smoothened curve fits the data plots better (Figure 3; see Annex B2 for further explanations on the ACPR methodology). The figure displays a slightly flat but nonetheless concave curve suggesting possible non-linear relationship between the two variables albeit the variables are already in log form. A simple scatter plot (Figure 4) shows similar non-linear pattern. Hence, the squared term of log FDI per capita (lfdipc2) is added into the regression equation to test for non-linear causal relationship.

Figure 3: ACPR plot against square of log FDI per capita

Figure 4: Scatterplot of Log Service Export Sophistication (EXPY) against Log FDI per capita

Model Specification

A formal model is first built upon this basic model specification widely used by empirical researchers.

The pooled OLS model combines independent cross-sectional data across time where to represent regression coefficients, i the country and t the time aspect for each variable. This model provides preliminary results where consistent estimates suggest lower possibility for functional form misspecification (Cameron and Trivedi, 2005). This model is often subject to omitted variable bias with no control for any fixed-effects.

Least-Square-Dummy-Variable (LSDV) Model

In this article, we run three different LSDV models. Model 2 includes both country (V_i) and year dummies (U_t).

The inclusion of country dummies controls for unobservable heterogeneous country-specific characteristics which are constant over time while time dummies control for unobservable characteristics homogeneous across all countries but vary over time to minimise bias in estimates. The LSDV model also allows for analysis of time-invariant variables which are completely eliminated in fixed-effects (FE) model. However, Baltagi (2008: 15) argues that the introduction of too many dummies can lead to inflated standard errors. In addition, presence of large number of dummies could overshadow the impact of other variables (Wooldridge, 2006) because the dummies possibly explain too much of the variation in the dependent variable.

Model 3 excludes year-fixed effects which can result in inconsistent estimates. All countries may experience a significant common event for which the impact on the dependent variable varies over time but is not accounted for. On the contrary, Model 4 includes only year fixed-effects which may also result in biased estimates due to omitted country-specific characteristics. However, this bias can be minimised by controlling for other time-invariant country-specific factors.

The regression models can be improved using dynamic system GMM panel data model. However, this will be left as potential extension for future research. The current study will only be employing the abovementioned models.

Data Analysis

Overview of Global Service Export Sophistication

Figure 5: Global Service Export Sophistication (EXPY) Pattern

The two-year group-averages of global service EXPY show increasing levels of service export sophistication in the past decade, implying that countries are getting more involved in higher value-added services, particularly between early and mid-2000s. While recovering from the dotcom crisis, the widespread use of online media, internet and advanced technologies on the global scale possibly propelled efficiency in international service delivery.

Global Mean PRODY Value of Service Exports

Figure 6: The bar chart compares average productivity value between different service export categories

In general, modern services export has higher 'productivity' value (PRODY) than traditional services export except communication services. Among traditional services export, personal, culture and recreation export has the highest PRODY value while travel export has the lowest PRODY value. Financial services have notably the highest PRODY value among modern services and in general.

Overview of Country-Specific Service Export Sophistication

Table 1: The top-five and bottom-five countries' service export sophistication ranking

Samoa's service export composition is much more concentrated in low PRODY traditional services export relative to others in the same category. The strong focus in these exports allows Samoa to rank slightly higher in sophistication.

Malta is considered a high-income nation while Thailand and Malaysia are fast-growing emerging markets but these countries rank rather low in service export sophistication. The data suggests they export relatively smaller shares of traditional services compared to Samoa. While Malta and Malaysia export more diverse modern services compared to the rest, they only constitute a miniscule share of their service export baskets resulting in lower overall service export sophistication. Moreover, they are less productive in services than in manufactured exports. These countries specialise in manufactured and primary exports (such as electric and electronic components, processed crude oil, vulcanized rubber) rather than services export (UNDATA, 2010). Hence, the development of service export sophistication may not be as advanced as other large service exporters. Papua New Guinea ranks the lowest as they export very little services to begin with.

Luxembourg has the highest overall export sophistication supported by strong specialisation in high PRODY financial services export complemented by a fair share of other modern services. Ireland, on the other hand, has rather large shares of computer and information, financial and business services export. UK also specialises in financial services export and other modern services. However, the relatively larger share of low PRODY traditional services explains the slightly lower rank in overall export sophistication relative to the previous two countries. The US ranks slightly lower than UK because they export relatively more low PRODY travel services, but maintains high overall sophistication with a significant share of other high PRODY business, financial and patent services. Japan's overall export sophistication is mainly supported by business and transport services. The export data also suggests Japan has a significant share of royalties and patent services export which possibly reflect its comparative advantage in providing environment conducive to R&D activities.

In general, what the current study finds from the above data analysis based on the EXPY model is:

• Surprisingly, if countries have highly-diverse set of services export, namely spread out shares in both traditional than modern services, EXPY may not be as high as the overall export sophistication of countries that are strongly focused or specialised in traditional services.
• Countries strongly focused in modern services will certainly have EXPY higher than countries with diverse exports or strongly focused in traditional services

Key Findings and Analysis

Main Empirical Results

Regression Model	(1)	(2)	(3)	(4)
	Pooled OLS	Least Square Dummy Variable
VARIABLES	Log Service Export Sophistication (EXPY)
Log FDI per capita	-0.105***	-0.114	-0.287***	-0.112***
	(0.0347)	(0.0778)	(0.0858)	(0.0274)
Squared of Log FDI per capita	0.00775***	0.000962	0.0170***	0.00650***
	(0.00212)	(0.00447)	(0.00490)	(0.00173)
Log GDP per capita	0.0767***	0.155**	0.484***	0.0548***
	(0.0193)	(0.0637)	(0.0548)	(0.0154)
Log Population	0.0257***	0.288	1.173***	0.0158**
	(0.00711)	(0.328)	(0.276)	(0.00657)
HDI Dummy	-0.101***	-1.307	2.250***	0.00712
	(0.0307)	(1.769)	(0.836)	(0.0238)
Information Flow Dummy	0.0924***	-1.237	2.651***	0.113***
	(0.0229)	(1.003)	(0.785)	(0.0178)
Modern Services Exporter Dummy	0.173***	1.069*	1.906***	0.227***
	(0.0299)	(0.563)	(0.464)	(0.0245)
Fixed Effects	none	country year	country	year
constant	8.741***	4.742	-16.36***	8.789***
	(0.148)	(4.111)	(5.689)	(0.135)
Observations	860	860	860	860
R-squared	0.300	0.775	0.716	0.494
Number of countries	86	86	86	86

Table 2: The overall regression results
Heteroskedasticity-robust standard errors are in parentheses. Asterisk mark *** represents significance at 1%, ** at 5% and * at 10% level respectively.

Model 1 indicates that the quadratic relationship between FDI and EXPY is highly significant at 1% level controlled for other possible determinants based on previous literatures. All controlled variables are highly significant. Consistent estimates imply lower possibility for functional form misspecification of FDI per capita. The preliminary results suggest for interesting discussion on the non-linear relationship between FDI per capita and the upgrade of service export sophistication.

In Model 2, both the linear and quadratic term of FDI per capita becomes insignificant along with other variables except GDP per capita and modex dummy which has significant positive impact on EXPY in line with expectations where 1% increase in GDP per capita raises service EXPY by 0.16%. Hausmann et al. (2005) and Anand et al. (2012) find similar results for GDP per capita in their studies. Although the coefficient on modex dummy has the expected sign, it seems exaggerated as the results imply that service EXPY of major modern services exporters is 191% higher than major traditional services exporters. On the contrary, all variables are significant in Model 3 but the estimates of the controlled dummy variables are even more inconsistent.

Results from Model 4 present both consistent estimates and minimal standard errors. Both the linear and quadratic term of FDI per capita is significant at 1% level. The signs of the estimates suggest a convex relationship between FDI per capita and EXPY. All other variables are significant with the expected signs except HDI which surprisingly does not seem to matter between countries with better and lower than group-average quality of human capital. Perhaps, low-quality local labours, particularly in developing countries can always be substituted with foreign skilled-labours with suitable incentive packages.

Despite the common inclusion of country dummies, the current research finds that Model 4 with only year-fixed effects generates more reliable results compared to the other models. Model 1 could be exposed to omitted variable bias whereas the inclusion of too many country dummies in Model 2 and 3 seems to cause excessively biased estimates for the controlled dummy variables. Hence, we will base our analysis and discussion on Model 4 hereafter as HDI, information flow and modex dummies respectively do represent some of the time-invariant (or miniscule variant) country-specific characteristics.

Sensitivity Checks

Regressions are conducted to test the sensitivity of the results by comparing advanced and non-advanced economies (see table in Annex C3 for the regression results ). When non-advanced economies are dropped from the sample, the results show insignificance of both linear and quadratic term of FDI per capita at the 1% and 5% level. This suggests FDI per capita does not affect service EXPY for advanced economies possibly because advanced economies already have stable institutions, efficient information and financial systems and reliable support (patent and copyrights) for technology innovation and R&D in producing highly sophisticated services regardless of FDI.

In addition, this indicates that the overall impact of FDI per capita on service EXPY is heavily influenced by non-advanced countries. To test this implication, advanced countries are dropped from the regression. The non-linear effect of FDI per capita is indeed highly significant at the 1% level among non-advanced countries with the turning point of approximately USD 3552.

Analysis of Non-Linear Relationship between FDI per capita and Service Export Sophistication

The main empirical results suggest there is an overall convex relationship between FDI per capita and EXPY but is heavily influenced by non-advanced economies. Figure 7 depicts the relationship between these two variables of interest.

Figure 7: The non-linear relationship between service export sophistication and FDI per capita

Firstly, based on the EXPY model, higher EXPY relative to C can mean that a country’s set of services export is either strongly focused in traditional services (A) or strongly focused in modern services (B). However, a country will certainly have higher EXPY with the latter characteristic (B) than with the former (B). A country may have low level of EXPY when the services export is highly diverse because this implies they are neither specialised in modern nor traditional services (C). It follows that the movement from point A to B potentially illustrates the long-run diversification mechanism of a country’s service export basket.

The diversification mechanism works such that countries begin by being highly-concentrated in traditional services (A), then become more spread out in both modern and traditional services (C), and subsequently more highly-concentrated in modern services (B). This pattern suggests increasing levels of FDI per capita significantly brings about diversification of services export only up to a particular point as higher levels of FDI seem to promote specialisation inclined to modern services export.

The current data suggests countries with low levels of FDI per capita are mostly developing or less-developed countries with relatively poor infrastructure quality and weak institutions. They still export services but are highly focused in traditional services export, particularly in travel and transport. There are, however, exceptions for certain countries such as India, Bangladesh and Philippines (See Annex C4 for scatter plot). Despite low levels of FDI given the size of their economies, they have high EXPY possibly because of active modern services offshoring investment activities. For example, a survey of several literatures by Lewin, Massini and Peeters (2009: 4) indicates that India was one of the earliest among developing Asian countries to attract these FDIs in the early 1980s due to 'favourable' political environment and presence of highly-skilled English-speaking labour force with low-cost advantages. Contrary to the aforesaid countries, for a majority of countries which begin with low levels of FDI per capita (F1), FDI is more likely to be highly-concentrated in traditional services such as tourism.

Increasing levels of FDI per capita from F1 indicates more active integration of advanced technology and specific know-how which induces diversification from traditional services export towards increasing modern services export. One possible explanation for the occurrence of diversification is the diversion of FDI from traditional services to modern services. This results in decreasing and increasing specialisation in traditional and modern services respectively. Certainly, we would expect diversion of FDI to modern services to result in increasing EXPY. Interestingly, the results from the current study show that increasing FDI per capita leads to decreasing EXPY at a slower rate up to a particular threshold level (or minimum level of EXPY), FT. This may be because countries with FDI below the threshold level may be at the stage where they have yet to have sufficient skilled-labours and efficient support system to adapt to the latest services production techniques and technology for modern services. In addition, countries may also impose restrictions on foreign companies to protect potential sun-rise exporting industries which possibly cause stronger internalisation of positive externalities that supposedly should benefit the modern export industry. Subsequently, while the country becomes less specialised in traditional services due to the diversion of FDI away from these services, the trickle-down effect from the diversion of FDI to modern services is simultaneously impeded and cannot be realised to the full extent immediately, resulting in decreasing EXPY up to the threshold level. UNCTAD (2006: 9) finds FDI activities in telecommunications or financial-related services are often heavily restricted to prevent natural monopoly and protect 'infant' entrepreneurs. These restrictions are found to be more apparent in 'East, South-East and West Asia' which typically have relatively lower FDI per capita. Almeida (2009: 14) finds that technological innovation indeed requires greater skilled human capacity especially when the technology is 'skill-bias'.

Figure 8: The impact of FDI per capita on service export sophistication

The threshold level, FT of approximately USD 5516 indicates the point at which the services export basket is most diverse. This means countries are neither specialised in traditional nor modern services which results in the minimum level of EXPY. When countries are not (or no longer) interested in specialising in a particular sector, they have lesser incentive to bring innovation to develop the particular service. This is not necessarily bad as this serves as an 'indicator' or benchmark level after which countries become more inclined to specialisation in high PRODY modern services export. Pass FT, as FDI increases, the country possibly have the capacity to better adapt to the spillovers indicated by their capability in exporting larger shares of high PRODY modern services export. Thus, service EXPY rises at an increasing rate when FDI becomes increasingly concentrated in modern services.

Limitations and Extension of Research

In certain years, unreported service export values may result in lower EXPY index rather than revealing the actual 'sophistication'. There may also be unaccounted unobserved heterogeneities since country fixed-effects were excluded due to inconsistencies. The current research did not account
for dynamic impact of FDI on service EXPY. This suggests for interesting extension to this research by using dynamic panel data models. Given very limited literature on this topic, there may be other determinants which can better explain the model while other methodologies such as time-series regression can be adopted for country-specific empirical studies.

Policy Implications and Conclusion

The global level of service export sophistication has been on the rise in the past decade. The EXPY model suggests that countries that are able to export larger shares of high PRODY services are said to have higher implied overall export sophistication.

The current study finds that FDI per capita has a significant convex relationship with service EXPY, heavily influenced by non-advanced countries. This relationship is less evident in the sub-sample of advanced countries. At lower levels of FDI, the impact of FDI is highly concentrated in traditional services. As FDI increases, FDI gradually diverts to modern services but its impact is possibly limited by the lack of capacity and openness of countries to adapt to the spillovers, thus leading to the fall in overall EXPY at a decreasing rate up to the threshold level. The threshold level of FDI per capita at which EXPY is at its lowest represents the most diverse basket of services export. Thereafter, as countries begin to adapt to the spillovers, increasing levels of FDI per capita leads to inclination towards high PRODY modern services which raises overall EXPY.

This suggests possible long-term policy implications whereby governments can consider using the benchmark level of FDI per capita to determine where their countries are positioned in terms of service export sophistication level and whether this particular level of export sophistication is due to high or low PRODY services export. Consequently, governments can determine the long-term policy target level of FDI per capita in order to raise EXPY. Since the benchmark level acts as an 'indicator' of countries' most diverse set of services export, when this level of FDI is achieved, governments can decide to keep the services export highly-diversified or specialise in higher PRODY modern services to raise overall service EXPY or strike a balance between these two options.

In conclusion, the current study finds that FDI per capita does have a significant long-term non-linear impact on service export sophistication. The gradual diversion of FDI from traditional to modern services results in trade-off in degrees of specialisation between modern and traditional services. The impact of diversion of FDI on specialisation, however, can be influenced by domestic facilitating factors like absorption capacity and openness of recipient countries. These factors in turn affect countries' ability to adapt to the spillovers, leading to the fall and rise in overall service export sophistication at different levels of FDI per capita.

---

Acknowledgements

Firstly, I would like to thank my supervisor, Professor Peter Hammond, particularly at the initial stage of the research for his kind advice and feedback.

Secondly, I would like to thank the British Conference for Undergraduate Research and the University of Warwick Reinvention Journal Team for providing me a wonderful opportunity to share my research findings with others.

List of Illustrations

Figure 1: 10-years-average Growth Rate of Merchandise and Services Trade

Figure 2: Growth Rate of Modern and Traditional Services (2002 - 2011)

Figure 3: ACPR Plot against Square of Log FDI per capita

Figure 4: Scatterplot of Log Service Export Sophistication (EXPY) against Log FDI per capita

Figure 5: Global Service Export Sophistication (EXPY) Pattern

Figure 6: Mean PRODY Values of Different Service Export Categories (2002 - 2011)

Figure 7: Graph of Non-Linear Relationship of Log EXPY against Log FDI per capita

Figure 8: Mechanism of Impact of FDI on Service Export Sophistication (EXPY)

List of Tables

Table 1: Top and Bottom-Five Countries' Service EXPY Ranking

Table 2: Overall Econometric Regression Results

Annex A: Dataset

A1. List of Countries

Advanced Economies (33)		Developing Economies (40)		Least-Developed Economies (13)
Australia	Malta	Argentina	Nicaragua	Bangladesh
Austria	Netherlands	Belarus	Nigeria	Cambodia
Belgium	New Zealand	Brazil	Pakistan	Ethiopia
Canada	Norway	Bulgaria	Panama	Gambia
Croatia	Portugal	Chile	Papua New Guinea	Mozambique
Cyprus	Singapore	China	Paraguay	Rwanda
Czech Republic	Slovak Republic	Egypt	Peru	Samoa
Denmark	Slovenia	FYR Macedonia	Philippines	Solomon Islands
Estonia	Spain	Georgia	Poland	Sudan
Finland	Sweden	Hungary	Romania	Tanzania
France	Switzerland	India	Russian Federation	Uganda
Germany	United Kingdom	Indonesia	South Africa	Vanuatu
Greece	United States	Israel	Sri Lanka	Zambia
Hong Kong, China		Jamaica	Tajikistan
Iceland		Latvia	Thailand
Ireland		Malaysia	Tunisia
Italy		Mauritius	Turkey
Japan		Mexico	Ukraine
Korea, Republic of		Moldova	Uruguay
Luxembourg		Mongolia	Venezuela

Note: The countries are categorised by UN and IMF's definition of level of development. See World Economic Outlook (2012).

A2. Modes of Services

Mode	Description and Examples	Presence of Exporter (A) in Foreign Country (B)
Mode 1 Cross-Border	Exporting country, A provides services to users in foreign country, B through B's domestic telecommunications or postal infrastructure. Such supplies may include consultancy or market research reports, tele-medical advice, distance training, or architectural drawings	No
Mode 2 Domestic Provision	Nationals of B have come from abroad as tourists, students, or patients to consume the respective services provided in A	No
Mode 3 Commercial Presence Abroad	Services are provided in country B by a locally-established affiliate, subsidiary, or representative office of a foreign-owned and — controlled company (bank, hotel group, construction company, etc.) from A	Yes
Mode 4 Movement of Natural Persons Abroad	A national from A provides a service within B as an independent supplier (e.g., consultant, health worker) or employee of a service supplier (e.g. consultancy firm, hospital, construction company)	Yes

Note: The table of description for examples of services of different modes is adopted from Chapter 1 of World Trade Organisation's GATS training module but from an exporter's perspective. See World Trade Organisation (n.db).

A3. Definition of Service Export Categories

Type	Export Categories	Description
Traditional Services	Transportation	Transportation covers all transportation (sea, air, land, internal waterway, space, pipeline, etc.) services that are performed by residents of one economy for those of another and that involve the carriage of passengers, goods (freight), rentals of carriers with crew, and related supporting services
	Travel	Travel differs from the other categories in the sense that the consumer (traveler) moves to the location of the provider (a resident of the economy) and what is covered by the travel category are those goods and services acquired during the visit (less than a year). The international carriage of the traveler is covered under transportation. The measure can be interpreted as a proxy for international tourism, business travelling and international student (even if staying longer than one year)
	Construction	These are construction services performed by employees outside the country of the location of the enterprise. It also includes the goods the employees bring with them abroad to perform the task. Expenditures for local good though are recorded under Other business services
	Personal, Culture & Recreation	These services are divided into (i) audiovisual (services and fees for motion pictures—including to actors and producers, radio and television programs and musical recordings) and (ii) other (services related to museums, libraries, sporting, correspondence courses, etc.)
Modern Services	Communication	Communication covers (i) telecommunication and (ii) postal and courier between residents and nonresidents international transactions
	Insurance	Insurance services cover insurance provided by a resident to a nonresident and vice versa. It would often be freight insurances but also other direct services
	Financial	Financial services cover financial intermediary and auxiliary services (except those of insurance enterprises and pension funds) between residents and nonresidents. This could be fees related to letters of credit, lines of credit, financial leasing, foreign exchange transaction, transaction in securities, asset management, etc.
	Computer Information	Computer data and new-related service transactions between residents and nonresidents. These could be data bases, data processing, hardware consultancy, software implementation, maintenance and repair of computers, new agency services, etc.
	Royalties & license fees	These are exchange of payments between residents and nonresidents for the use of intangible and nonfinancial assets or property rights such as patents, copy rights, franchising, manuscripts, films, etc.)
	Other business	Other businesses include (i) Merchanting (the purchase of goods by a resident from a nonresident and the subsequent resale to another nonresident, during which the good does not leave the compiling country), (ii) Operational leasing without operators covers residentnonresident leasing, and charter without crew, (iii) Miscellaneous services, including (a) legal, accounting, management consulting, public relation services, (b) advertising and market research services, (c) research and development services, (d) architectural, engineering and other technical services, (d) agricultural, mining and on-site processing services, and (e) other services between residents and nonresidents

Note: The definitions of different service export categories above are adopted directly from page 17 of Anand, Mishra and Spatafora (2012) with kind permission from the authors. The descriptions are based on the 5th Edition of the IMF Balance of Payment Manual (1993).

A4. The Construction of the EXPY Index

The Ranking of Service Export 'Productivity' (PRODY)

i represents a particular service export (e.g. Travel)
j represents a particular country

GDP per capita is the measure of income/productivity level of a country while the Revealed Comparative Advantage (RCA) value acts as the weight and allows us to rank the productivity level of a particular service export independent of countries' size.

For example, say US has GDP per capita of $50000 while Bangladesh, $1000. However, the value-share of Bangladesh travel service export to the aggregate value-share is 0.60 while for the US, 0.002. Hence, Bangladesh is ranked higher in terms of productivity (0.60 x $1000 = $600) because it specialises much more in travel service export than the US (0.002 x $50000 = $100) as implied by the trade figures.

It follows that PRODY is the collective 'productivity' level (or weighted average of income level) of a particular service export across all countries in the sample. The number of countries with high, moderate and low income levels and the magnitude of their share of a particular export to the sample aggregate value-share affect magnitude of PRODY. For instance, in the case where many high-income countries (and less of low-income countries) export a large share of a particular service relative to the aggregate value-share (for instance, financial services), the collective productivity level (PRODY) of financial services export tends to be higher than services exported in large shares by many low-income countries.

This implies that considerable technological skills, knowledge and quality infrastructure are needed for those high-income countries to be involved in that particular service export. Thus, financial services is an export deemed to be of higher 'sophistication' associated with high level of productivity compared to other services exported by mostly lower-income countries associated with lower level of productivity such as construction services.

The Overall Service Export Sophistication (EXPY) Index

The overall service export sophistication index (EXPY) of a country is simply the mean of collective 'productivity' of each service export weighted by the country's relevant share of exports. The weight also indicates the composition of the service export basket. Since service export sophistication is a variable that cannot be measured directly, it is important to note that here, mean collective 'productivity' is used as a sort of proxy for overall service export sophistication based on the intuition mentioned before.

Countries tend to have higher overall export sophistication if they are capable of exporting large shares of more sophisticated services (high PRODY). For example, Singapore has slightly lower income level compared to Finland, but in the case where Singapore has larger shares of different high PRODY service exports relative to Finland, then Singapore will still be ranked higher in terms of overall service export sophistication. This applies to any comparison between low-low, low-high and high-high income countries.

In a sense, the overall service export sophistication of a country depends on the type of services it exports, the 'productivity' value (PRODY) of the particular service export (i.e. whether associated with high or low income level) and the composition of service export basket. The unit of EXPY index is the US Dollars since all the service export and GDP per capita values are denominated in USD.

Limitations of EXPY Index
Since the index is derived from service export values, the index could sway according to external shock rather than an actual upgrade or decrease in export sophistication or 'productivity'. The 2009 global financial crisis shock lead to the fall in service export values for all countries. Nonetheless, the growth of service export has been portraying an upward trend after the shock, showing slow but positive recovery. This suggests that one-time shock may not necessarily inhibit countries' ability to raise productivity again.

A5. Data Definitions and Data Source

Indicator	Description	Data Source
Service EXPY	The mean of collective 'productivity' of each service export weighted by the country's relevant share of exports	Author's calculation
FDI stock per capita (constant 2005)	Annual amount of capital and retained earnings accumulated within a country as result of a foreign investment	UNCTAD
GDP per capita (constant 2005)	Gross domestic product divided by midyear population. GDP is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation	World Development Indicator
Population	Total population counts all residents regardless of legal status or citizenship, except for refugees not permanently settled in the country of asylum and generally considered part of the population of their country of origin	World Development Indicator
Human Development Index	HDI is built on the basis of: Life expectancy at birthas a proxy for population health and longevity Education Indexwhich include adult literacy rate, and combined primary, secondary and tertiary gross enrolment ratio (with 1/3 weightage each) as a proxy for knowledge and education GDP per capita at PPPas a proxy for standard of living	United Nations Development Programme (UNDP)
Information Flow Index	The information flow index is defined by the combination of: Number of internet users (per 1000 people) with worldwide internet network Television users(per 100 people) - share of households with a television set Trade in newspaper as a percentage of GDP	KOF Globalisation Index
Modern Services Exporter	Countries with mean share of modern services larger than mean share of traditional services export	Generated by Author

A6. Variables Specification and Justification

GDP per capita is used as a proxy for country's level of development controlled for country size. It is found to have a significant impact on EXPY by Hausmann et al. (2005) and Anand et al. (2012). Current GDP per capita is deflated using US GDP deflator constant at 2005. Population controls for size of the economy.

The Human Development Index (HDI) is used as a proxy for quality of human capital. In addition to education index, HDI also captures the aspects of life quality and economic situation in shaping human capital. Kwon (2009) suggests that HDI can be used as a new form of human capital measurement as International Labour Office (ILO) uses similar index to judge quality aspects of labour. The data is only available from 2005 onwards.

The Information Flow Index could possibly capture communication efficiency which is essential in delivering services. The data is available only up to 2009.

HDI and Information Flow Index are made as dummy variables by grouping countries into two groups. Countries with ten-years mean HDI and information flow index above group-average takes the value 1 and 0, otherwise.

The 'Modern Services Exporter' dummy variable accounts for countries with a larger share of modern relative to traditional services export. Country with mean share of modern services larger than mean share of traditional services export takes the value 1 and 0, otherwise.

Annex B: Methodology

B1. Calculating the Turning Point of FDI per capita

The turning point of FDI per capita, X is given by the formula:

X = e [(coefficient of log fdipc)/ (2*(coefficient of squared term of log fdipc)]

B2. The Augmented Component-Plus-Residual Plot

Mallows (1986) explains that the lowess curve is a smoothened curve which passes through the fitted values more closely to illustrate any non-linear relationship. Hamilton (2009) argues that lowess curve may divert towards outlier values rather than displaying actual non-linearity, especially if the 'central part of the lowess curve' does not show a 'systematically curved pattern'. In the case of the current study, looking closely at Diagram 3, the central part does have a curved shape albeit very minimal. In addition, the upward curve on both end of the lowess line is not exactly caused by very significant extreme values as Hamilton suggests. As a result, these probably indicate that there may be a (weak) non-linear relationship between the variables of interest.

Annex C: Data presentation and empirical results

C1. Annual PRODY of Different Service Export Categories

Year	Traditional Services Export				Modern Services Export
	Travel Services	Construction Services	Transportation Services	Personal, Culture and Recreation	Communication Services	Business Services	Computer and Information Services	Royalties and Patent Services	Financial Services
2002	8012.68	12108.45	10276.90	13377.01	7353.02	16740.88	16864.12	17921.68	25883.72
2003	9366.91	11016.38	11518.84	15441.12	8024.73	18265.18	20001.80	22604.29	30962.73
2004	10260.98	12834.82	13266.66	16111.07	8716.94	20259.87	22665.83	24069.42	34025.99
2005	10551.73	12910.13	13464.11	17404.27	9206.50	21925.39	22082.33	22529.00	35091.71
2006	10795.74	14051.65	13788.05	19555.65	9424.31	22841.47	21847.16	25174.79	35737.56
2007	11786.66	15242.74	15849.93	20948.08	10787.53	25113.91	23251.87	25435.87	39095.17
2008	12529.06	17990.47	17197.59	24637.69	11996.71	25664.46	24522.17	27428.55	40167.04
2009	11163.96	14154.05	14425.79	23368.96	10906.75	22577.32	20746.76	23877.73	37193.53
2010	11988.33	12019.16	15063.45	24088.09	9728.27	22617.88	24078.34	24719.00	38660.49
2011	12885.78	15167.06	15658.92	22029.98	12044.78	23199.75	24144.31	24709.04	41511.07
Mean	10934.18	13749.49	14051.02	19696.19	9818.95	21920.61	22020.47	23846.94	35832.90
Standard Deviation	1485.23	2032.47	2070.23	3949.31	1594.16	2800.10	2339.42	2517.93	4690.93

Source: Author's calculation based on Hausmann, Hwang and Rodrik's export sophistication model

C2. Overall Mean and Standard Deviation

Variables	Mean	Std. Dev.
Service EXPY (USD)	14581.10	4264.03
FDI per capita (USD)	10538.52	24086.17
GDP per capita (USD)	15231.21	18332.12
Population (per head)	6.42e+07	1.90e+08
Human Development Index	0.72	0.16
Information Flow Index	68.09	19.68

C3. Regression Results of Sub-Sample Countries: Advanced and Non-Advanced Countries

Region	Advanced Countries		Non-Advanced Countries
Regression Model	OLS	LSDV	OLS	LSDV
VARIABLES	Log Service Export Sophistication (EXPY)
Log FDI per capita	-0.259	-0.267*	-0.181***	-0.138***
	(0.165)	(0.143)	(0.0626)	(0.0474)
Squared of Log FDI per capita	0.0160*	0.0143*	0.0145***	0.00844**
	(0.00934)	(0.00820)	(0.00465)	(0.00361)
Log GDP per capita	0.145***	0.0872***	0.0837***	0.0593***
	(0.0369)	(0.0331)	(0.0273)	(0.0214)
Log Population	0.0328**	0.0153	0.0212**	0.0157**
	(0.0139)	(0.0120)	(0.00837)	(0.00783)
HDI Dummy			-0.117***	-0.0154
			(0.0337)	(0.0259)
Information Flow Dummy	-0.0257	-0.0183	0.148***	0.161***
	(0.0558)	(0.0406)	(0.0266)	(0.0206)
Modern Services Exporter Dummy	0.155***	0.236***	0.189***	0.190***
	(0.0438)	(0.0349)	(0.0569)	(0.0480)
Fixed Effects	year	year	year	year
constant	8.627***	9.384***	8.962***	8.812***
	(0.711)	(0.625)	(0.219)	(0.178)
Observations	330	330	530	530
R-squared	0.406	0.610	0.136	0.351
Number of countries	33	33	53	53

Heteroskedasticity-robust standard errors are in parentheses. Asterisk mark *** represents significance at 1%, ** at 5% and * at 10% level respectively.

C4. Scatter Plot of Log Service Export Sophistication (EXPY) against Log FDI per capita

Notes

[1] Wenn Jinn Cheah has recently completed her degree in BSc (Hons) in Economics with second-upper class honours at the University of Warwick. For the future, she will be pursuing her career at the Central Bank of Malaysia.

 

References

 

 

To cite this paper please use the following details: Cheah, W. J. (2013), 'Does Foreign Direct Investment (FDI) Affect Service Export Sophistication?,' Reinvention: an International Journal of Undergraduate Research, BCUR/ICUR 2013 Special Issue, http://www.warwick.ac.uk/reinventionjournal/archive/bcur2013specialissue/cheah/. Date accessed [insert date]. If you cite this article or use it in any teaching or other related activities please let us know by e-mailing us at Reinventionjournal at warwick dot ac dot uk.