1 Introduction

The search for sound knowledge on the drivers of export product diversification is important for the developing world as it is associated with economic growth (Dadush et al. 2020; Van den Berg and Lewer 2007) and mitigates the risks associated with commodity price volatility and macroeconomic shocks (Berthélemy 2005). Diversification has been shown to be relevant for developing countries as an engine of economic growth through technological spillovers to other sectors and as a source of jobs creation, structural transformation, and sustainable development (Freire 2019). Diversification of the productive structure of a country’s export basket is considered as an important source of resilience to external macroeconomic shocks and development for low-income countries (Berthélemy 2005; Caselli et al. 2020).

Economic diversification is equated with the exports structure of a country. For Berthélemy (2005), an economy is said to be diversified if its productive structure is dispersed into a large number of activities that differ from one another like goods and services produced. Subsequently, export product diversification refers to diversification in the basket of product export on international market by a given country. Empirical and analytical studies on diversification have long focused on the structure and dynamics of trade of tangible goods. Empirical findings (Agosin et al. 2012; Berthélemy 2005; Cadot et al. 2011; Elhiraika and Mbate 2014; Fosu and Abass 2019) show that export product diversification is determined by a multitude of factors that influence the long run behavior of a country or region’s export structure. Services, despite their increasingly dominant place in international trade, according to recent statistics (Footnote 1; Footnote 2), are almost invisible in the exploration of the determinants of economic diversification. Yet today, both the developed and developing worlds are undergoing structural changes that bring services to the forefront. The WTO’s 2019 report estimates that global trade in services is growing faster than trade in goods, with the value of trade in services reaching US$13.3 trillion in 2017. The share of developing economies in world trade in commercial services was 34% in 2017 (WTO 2018).

The premises of the theoretical foundations on the contribution of services to real activity can be found in classical economic thought which, centered on the role of the manufacturing industry, has contributed to forging an image of deficient services from the performance view (Faïz 2007). Smith (1776) contrasts the productive work of manufacturing with the unproductive work of services, which vanish the moment they are produced. For Smith (1776), services are immaterial and do not create value identified with material production. The debate on services resurfaced in the 1960s with Baumol (1967) and Fuchs (1968) attributing to services the status of Cost Disease.Footnote 3 In the 1980s, much of the research was devoted to trade in services, drawing on WTO’s publications and regional trade agreements. However, as Francois and Hoekman (2010) figure out, the majority of research on services focuses more on liberalisation in the services sector as well as the literature gives credit to trade and Foreign direct investments (FDI) in services. Meanwhile studies on the contributions of international services flows to world output and export patterns are even less visible. Therefore, we are in a phase of questioning the effect of the international diffusion of trade in services on export product diversification. What are the dimensions of trade in services that are conducive to export product diversification in Sub-Saharan Africa (SSA)? The main hypothesis in this paper is that commercial services exert positive effect on export product diversification in SSA. Therefore, the objective of this research is to identify the dimensions of trade in services that promote exports product diversification in SSA.

The relationship between services trade dynamism and export product diversification still needs empirical studies to strengthen the so far insufficient debate. The search for directions in the relationship between trade openness policies and economic diversification has guided the research prism for a long time. In view of the importance of diversification in transforming economies and achieving the goals set out in recent national, regionalFootnote 4 and global development initiatives,Footnote 5 an assessment of the determinants, with particular attention to the role of services on export product diversification in developing countries, especially in SSA, is needed.

Exploring services as trade policy strategy remains an alternative in Africa, since natural resources account for the bulk of African exports. Moreover, given the growing role of services in Africa, the implementation of the African Continental Free Trade Area (AfCFTA) will be difficult if services are relegated to a secondary position in favor only of trade in goods.

A better understanding of the relationship between the development of international trade in services and export product diversification could better guide African policy makers in terms of policy adoption and implementation. In this sense, this paper contributes to better inform decision makers. The existence of a fairly extensive literature on economic diversification demonstrates the importance of the interest in analysing the vectors of export product diversification. However, the existing studies do not consider theoretically and empirically the role of exporting commercial services by an economy on its export product diversification. Therefore, this study would be, according to our understanding, one of the first paper to empirically address this gap.

The rest of the paper is structured as follows. The second section presents the literature review. The third section presents the stylized facts on trade in services and export product diversification in SSA. The fourth section presents the data and the methodology. The fifth section presents and discusses the empirical results on the effects of trade in services on export product diversification in SSA. The sixth and last section concludes.

2 Litterature Review

The premises of export product diversification debate according to some authors (Berthélemy 2005; Cadot et al. 2011; Cadot and De Melo 2016; Hausmann et al. 2007; Imbs and Wacziarg 2003) can be traced back to the classical, neoclassical and the debate is non-exhaustive in the literature. While Ricardo supports specialisation, Heckscher-Ohlin models argue instead that export dynamics are largely determined by endowments, so that, if anything, we should be concerned with factors accumulation, not diversification (Cadot et al. 2011). Yet export product diversification remains a constant concern for policy makers in developing countries. Even more naive is the idea of explaining export dynamics primarily by endowments. Indeed, according to Cadot et al. (2011), the relationship between endowments, trade and growth is complex and imperfectly understood. Models of intra-industry trade have long shown that many factors other than endowments, including market failures and policies, can affect trade patterns. This idea is supported by the work of Hausmann et al. (2007) who find that export patterns can exhibit path dependence in the presence of externalities.

While some argue that export product diversification allows a country to better manage the risks associated with commodity price volatility (Berthélemy 2005; Hammouda and Ben 2006), others argue that it allows countries to hedge country-specific demand fluctuations and insure against downturns at home (Romer 1990). Faïz (2007) argues that pioneers’ skepticism (Baumol 1967; Bladen 1960; Smith 1776) about the role of services in economic development has for a long time dominated economic literature, so that services have remained in a non-tradable and unproductive sector consideration. However, in recent years, services have become increasingly important in international trade and investment. This is accompanied by a growing interest in services trade in the diversification literature (Caselli et al. 2020; Feng et al. 2021; Gnangnon 2020a; Nieminen 2020).

Moreover, despite the proven role of services in production (Arndt and Kierzkowski 2001; Fisher 1939) and growth (Baer and Samuelson 1981; Baumol 1967; Lee and McKibbin 2018; Romão 2020), there is less focus on the role of trade services in export product diversification. Yet, some theoretical and empirical evidences show the importance of commercial services in manufacturing production dynamism (Amiti and Wei 2005; Arnold et al. 2008; Beverelli et al. 2017; Chand and Sen 2002; UNCTAD 2022; Vogel 2022). The focus has been more relevant on services imports as inputs into production processes, but the main question here is to question the role of commercial services exports in the export product diversification. That said, our hypothesis is that among different dimensions of trade in services, commercial services exports exert favorable effects on export product diversification in Sub-Saharan Africa. In this vein, commercial services exports like tourism, travel, transport, financial and insurance services which are linked to exports of physical products can be catalysts to diversification of export product (Eichengreen and Gupta 2013).

But some voices emphasize the limited effects of some aspects of commercial services. For instance, although the increase in tourism revenues, Lejárraga and Walkenhorst (2013) argue that it does not automatically translate into large-scale economic development. According to them, the fact is that this sector is once again at the heart of the major development and economic diversification strategies of several developing countries. Tourism demand has induced effects on other economic sectors through indirect and direct effects generated by tourism spending on non-tourism sectors in host economies (Lejárraga and Walkenhorst 2013; Romão 2020; Sharpley 2002). Another phenomenon that affects export product diversification is servitization, which occurs when firms begin to produce and export services in addition to physical exports. This phenomenon allows not only productivity growth, but also an increasing capacity to firms for resilience and products diversification.

Trade in financial and banking services in the multilateral system affects not only the volume of exports (Baldwin and Krugman 1989; Chan and Manova 2015; Dixit 1989; Kletzer and Bardhan 1987; Memanova and Mylonidis 2020) but also the dynamics of a country’s export basket structure (Bose et al. 2020; Foley and Manova 2015; Nieminen 2020). Foley and Manova (2015) argue that the ability to access financial capital to pay for fixed and variable costs affects firms’ choices about entry and export operations and, therefore, influences the overall structure of trade. Financial frictions and the use of internal capital markets influence the decisions made by multinationals regarding production locations, integration, and corporate governance.

The fast development of Information and Communication Technologies (ICT) has a significant impact on the performance of companies (Chari et al. 2007; León et al. 2016; Ravichandran et al. 2009), in the production of new goods (Chari et al. 2007), as well as on their access to new markets (León et al. 2016). León et al. (2016), in their study, analyse the impact of ICT use on the degree and type of diversification of small and medium-sized enterprises (SME). From a sample of 95 companies in the Autonomous Community of the Basque Country, they realize that diversified companies show a higher level of ICT use and this resource positively affects the degree of international diversification and the intensity of the company’s activities.

Some findings have empirically shown the effect of commercial services export on diversification. The relationship between the tourism sector and export product diversification is controversial in the empirical literature. On the demand side, Lin et Sung (1984) find that tourism export growth in Hong Kong is more stable than that of major commodity exports, partly because tourism is less subject to import protectionism. They believe that tourism is therefore considered a prime choice in Hong Kong’s economic diversification. Lejárraga et Walkenhorst (2013) in their study of a large sample of developing countries with cross-sectional data, find that the area most amenable to short-term policy interventions, such as the business environment or trade regulations, are the most important in fostering productive linkages between tourism and the general economy. In contrast, fixed factors, such as land availability, or longer-term objectives, such as progress in development levels, have less influence on the productive and export structure. Using a panel data model for 2006–2017, Romão (2020) finds that specialisation patterns combining tourism and agriculture have positive effects in both cases. Diversification strategies that include unrelated sectors contribute to increasing the resilience of European regions, while a focus on construction reduces regional resilience.

Nieminen (2020) using data from the Exporter Dynamics Database (EDD) finds that access to domestic financial services contributes positively to export product diversification by increasing the number of small exporters, as financial services alleviate the credit constraints faced by these exporters. Nguyen et al. (2020) mobilise nine financial development indices and three patent variables to identify the main determinants of the captured economic complexity index. Nguyen et al. (2020) find that an overly large financial sector does not contribute to the diversification and sophistication of a national economy, but the efficiency of financial markets seems to have a positive influence on these processes, probably because financial markets provide alternative ways of financing patents and knowledge.

Unger (2016) conducts an empirical analysis on the role of financial intermediation in international trade. Combining Melitz’s (2003) firm heterogeneity with Holmstrom and Tirole’s (1997) credit frictions, Unger (2016) observes a selection of larger firms towards exporting and unsupervised financing, such as government debt or corporate bonds. Smaller producers only serve the domestic market and have to resort to costlier financial intermediation. He also finds that producers respond to financial shocks by switching to other types of financing. Furthermore, his model highlights a new source of gains from trade: average productivity increases when lower trade costs allow some exporters to select cheaper unguarded financing.

Chan and Manova (2015) empirically show that financial market imperfections affect the number and identity of exporters’ destinations. Their results reveal that large economies with lower trade costs are more attractive markets because they offer higher export profits. They show that financially advanced nations therefore have more trading partners and move down the hierarchy, particularly in sectors that are highly dependent on the financial system. Gani and Clemes (2016) find a statistically significant positive correlation between the rule of law and regulatory quality and export and import of insurance and financial services in OECDFootnote 6 countries and in some developing countries. In contrast, their empirical results reveal a negative and statistically significant relationship of contract enforcement with exports and imports of insurance and financial services.

Chari et al. (2007) develop and empirically test the hypothesis that investment in information technology helps to leverage the firm’s specific assets across national borders and thus contributes to improving international diversification performance. They show that the impact of international diversification on performance is a positive function of the level of ICT investment. For the latter, the impact on performance can be significantly positive (for firms with high ICT investment), significantly negative (for firms with low ICT investment), or neutral (for the average internationally diversified firm, i.e. firms with an average level of ICT investment). Ravichandran et al. (2009) find in their study of US firms that while ICT spending interacts with tied diversification to have a positive effect on firm performance, similar interactions with untied diversification have no effect on firm performance. Moreover, the interaction between ICT spending and geographic diversification is only positively associated with performance when the level of geographic diversification is low.

3 Trade in services and export product diversification: stylized facts

This section presents the stylized facts of the relation between trade in services and export product diversification in SSA.

3.1 Dynamics of trade in services in Sub Saharan Africa

Figure 1 shows the evolution of trade in services in percentage of GDP in some regions of the world, namely Sub-Saharan Africa, the European Union (EU), South Asia and North America. The EU is the region that emphasized strong increase in trade in services over the period 2005–2019. The second region where trade in services accounts significantly in GDP is South Asia, representing 11.24 per cent of GDP on average over the same period. Sub-Saharan Africa is the third region with 11.10 per cent of GDP over 2005–2019. In North America, trade in services represents 7.08 percent of GDP.

Fig. 1
figure 1

Trade in services (% of GDP) dynamics in some regions over 2005–2019. Source: Authors construction, UNCTAD data (2021)

Full size image

Trade in services in SSA is experiencing significant dynamism in view of its remarkable performance. SSA countries are net importers of services. Over the period 2005–2019, total services imports accounted for 66.61% of total services trade on average against 33.38% for services exports. However, due to the COVID-19 pandemic, exports and imports of services fell significantly in 2020, with a 15.55% drop in total trade in services in 2020 compared to 2019.

Figures 2 and 3 provide a set of information on the evolution of trade in services in the most important dimensions of this sector in SSA during the period 2005–2019. The Figures 2 and 8 in appendix show a strong expansion of the tourism sector in SSA since 2005. With an average annual growth rate of 3.16% over the period 2005–2019, SSA is becoming an important destination for tourism. The number of arriving visitors in SSA has increased from 25.92 million in 2005 to over 48 million annual visitors in 2016 before slowing down to 38.3 million in 2019. Over the same period, the top tourist destinations in SSA are respectively South Africa followed by Nigeria, Botswana, Mozambique, Kenya, Eswatini, Namibia, Rwanda, and Senegal (see Fig. 8 in appendix). This surge in tourism demand is accompanied by a significant increase in local consumer products, with important knock-on effects on other sectors.

Fig. 2
figure 2

Commercial services exports share dynamics in SSA over 2005–2019. Source: Authors’ construction, data from UNCTAD (2021)

Full size image
Fig. 3
figure 3

Commercial services imports share dynamics in SSA over 2005–2019. Source: Authors’ construction, data from UNCTAD (2021)

Full size image

The Fig. 2 shows that travel services occupy the largest share of commercial services exports in SSA from 2005 to 2019. Indeed, this sector accounts for an average of 43.97 percent of total services exports in SSA (with 4.65% average growth) while travel services imports are down with an average of 9.93% growth over the same period representing 16.94% of total commercial services imports. Imports of transports services lead total services imports with an average growth of 0.19% over the period 2005–2019 and averaging about 41%. Exports of transport services represent 23.78% on average over the same period and an average growth of 6.66%. Trade in ICT services experienced positive trends, both in export and import and representing respectively an average growth of 7.81 and 9.68% over 2005–2019. Over the same period, financial and insurance services experienced an increase in exports (9.14 and 6.61% respectively) and in imports (11.50 and 8.27% respectively).

Overall, when considering services merged as traditional and modern servicesFootnote 7 (Katouzian 1970; Eichengreen and Gupta 2013), Fig. 9 in appendix shows a clear superiority of traditional services exports over modern services exports. The export trends on the Fig. 9 confirm those illustrated in Fig. 2. The weight of traditional services in international trade in services is partly related to the economic characteristics of countries in SSA that are heavily dependent on service sub-sectors such as transport, travel and tourism. Modern services are described as knowledge-intensive services (Nelson and Winter 1985), the intensity of their export by a country depends on its competitive human capital endowment. Therefore, the low level of human capital in most SSA countries explains the low level of exports of modern services.

3.2 Dynamics of export product diversification in Sub Saharan Africa

Figure 4 traces the evolution of the export concentration index in five regions of the world, namely SSA, North America, Western Europe, East and Southeast Asia, and North Africa over the period 2005–2019. This figure reveals a shift in Herfindahl-Hirschman index (HHI)Footnote 8 into two groups. The first group includes regions with low exports concentration (North America, Western Europe, and East and Southeast Asia) and the second group includes regions with high exports concentration index (SSA and North Africa). The export concentration of East and South Asian countries evolves on average around 0.1 while North American and Western European countries show an evolution of their export concentration below 0.1 over the period 2005–2019. The trend in the export concentration index for SSA and North Africa shows growth over the period 2005–2008. Over this period, North Africa shows higher levels of concentration than the other parts of the world, with a peak of 0.48 in 2006 and 2007. The figure shows a downward trend in the export concentration index in SSA in recent years, but the level is still high compared to other regions. Indeed, over the period 2005–2019, SSA has an average concentration index level of 0.374 while the index is on average 0.370 in North Africa, 0.124 in East and South Asia, 0.084 in North America and 0.069 in western Europe.

Fig. 4
figure 4

Evolution of exports concentration (HHI) in some regions, 2005–2019. Source: Authors’ construction, data from UNCTAD (2021)

Full size image

However, starting in 2008, there was a sharp decline in the export concentration index in both regions of Africa (SSA and North Africa). The efforts undertaken in SSA countries have resulted in an average decrease of 2.14% in the HHI over 2005–2019. The financial crisis of 2007–2009 can explain this situation. Indeed, the financial crisis during these years led to a contraction in global demand for raw materials, of which African countries are the largest suppliers. Faced with low demand for non-value-added exports from developed countries, African countries are forced to increase their local processing capacity for their raw materials.

3.3 Correlation between interest variables

Figure 5a, b provides an overview of the correlation between export concentration (HHI) and the explanatory variables of interest (tourism, export of transport, travel, financial, insurance, ICT, licence, creative services and other services). Its shows a negative relationship between export of tourism, transport, travel, financial, ICT, licence, creative services and other services and export concentration index (HHI). This means that an increase (decrease) in exports volume of these services leads to a decrease (increase) in export concentration in SSA between 2005 and 2019.Footnote 9 The correlation coefficients in appendix Table 2 illustrate the negative correlations between the services trade dimensions and the export concentration index.

Fig. 5
figure 5

a Correlation between export concentration and exports of total services and disaggregated traditional services SSA, 2005–2019. Source: Authors’ construction, data from UNCTAD (2021). b Correlation between export product concentration and exports of disaggregated modern services SSA, 2005–2019. Source: Authors’ construction, data from UNCTAD (2021)

Full size image

Moreover, Fig. 6 shows a strictly negative correlation between the export concentration index and traditional and modern services exports (See Table 1 in appendix for more details on the coefficients of correlations). That is, it suggests that an increase in the level of exports of traditional and modern services is accompanied by an increase in export product diversification in SSA. The figure shows that countries with high levels of export diversification are major exporters of traditional and modern services. South Africa with the highest level of export diversification is also by far the largest exporter of traditional and modern services. This is due to South Africa’s technological endowments and its level of development. The country is not only a tourist magnet but also its level of development allows it to export highly competitive services. Countries such as Kenya, Mauritius, Tanzania, Uganda, Senegal also have similarly high levels of export diversification and exports of traditional and modern services. Ethiopia with a high level of exports of traditional services is a concentrated economy in terms of product exports. Countries such as Angola, Burundi, Nigeria, Guinea Bissau, Chad, Equatorial Guinea are the least diversified and also most have low exports of modern and traditional services.

Fig. 6
figure 6

Correlation between export product concentration and aggregated traditional and modern services export in SSA, 2005–2019. Source: Authors’ construction, data from UNCTAD (2021)

Full size image

The Fig. 7a, b shows a positive relationship between import of transport, travel, financial, ICT, license, creative services and other services and export concentration index (HHI) in SSA. That is, an increase (decrease) in import of those services dimensions imply an increase (decrease) in export concentration index (HHI) or a decrease in export product diversification in SSA over the period 2005 to 2019. Moreover, Table 3 in the Appendix shows that all dimensions of services imports in SSA are not only positively correlated with the export concentration index, but also have fairly low correlation coefficients.

Fig. 7
figure 7

a Correlation between export concentration and import of total services and disaggregated traditional services SSA, 2005–2019. Source: Authors’ construction, data from UNCTAD (2021). b Correlation between export product concentration and exports of disaggregated modern services SSA, 2005–2019. Source: Authors’ construction, data from UNCTAD (2021)

Full size image

Given that export in services have more negative relationship with export concentration more than import in services, that means the focus should be put on the relationship between export in services and export product diversification. However, it is important to notice that correlation does not automatically imply a causal relationship between the variables, so econometric regressions will allow us to verify the true nature of the relationship between the variables.

4 Data and methodology

This section presents the data and the methodology of the study.

4.1 Data

In the literature, we find several indexes expressing the degree of diversification of export product, such as the warhead index (Attaran and Zwick 1987; Hammouda and Ben 2006), the entropy index (Attaran and Zwick 1987; Berthélemy 2005), the Herfindahl-Hirschman index (Agosin et al. 2012; Berthélemy 2005; Hammouda and Ben 2006), the aggregated specialization index (Berthélemy 2005; Dadush et al. 2020; Hammouda and Ben 2006) and the export ubiquity index developed by Hausmann and Hidalgo (2012). In this study, we use the Herfindahl-Hirschman index (HHI) as a relative measure of export product diversification by expressing its value between 0 and 1. HHI is preferred because it is both the simplest to program and the most frequently used in the literature on export product diversification.Footnote 10 The normalized HHI index reflects the degree of concentration of a country’s export, expressed as a value between 0 and 1. When the value of the index is close to 0, it reflects a lower concentration of export and a high degree of economic diversification; however, a value close to 1 reflects a high concentration of exports and therefore a low degree of economic diversification. Data on HHI are collected from the UNCTAD (2021) database.

Data on trade services (all services dimensions) are also collected from the UNCTAD (2021) database. These data are of two types: (i) those corresponding to the concepts and definitions of the fifth edition of the IMF Balance of Payments Manual (BPM5) edited in 1993 with data from 1980 to 2013 and (ii) the sixth edition of the IMF Balance of Payments and International Investment Position Manual (BPM6) edition in 2009 which provides a new definition of services with a classification of eighteen (18) categories and sub-categories of services by breaking the first six categories according to the BPM5. The BPM5 categorizes services as follow: (1) all services, (2) transport, (3) travel, (4) other services, (5) all commercial services, and (6) other commercial services. Given that the BPM6’s classification gives a wider range of subcategories of services than the BPM5’s classification, we therefore consider in this study only data collected from BPM6’s classification over the period 2005–2019. There are several reasons for not combining the two databases. First, they do not have the same classification methodology, combining them can lead to errors in the analysis. Second, the fact that the number of service categories differs from each database can be a potential source of missing in combined databases.

Data on human capital (number of years spent in secondary school) are from the Penn World Tables (PWT). Data on macroeconomic, physical, and institutional variables are from the World Bank (WDIFootnote 11 and WGIFootnote 12). The study covers an unbalanced panel of 48 countries in SSA (Table 3, Appendix) over the period 2005–2019 (15 years). While Table 5 (Appendix) presents the variables, their definitions and their sources, Table 6 (Appendix), presents the descriptive statistics of those variables.

4.2 Model specification

According to the literature on export dynamics and structure, the theoretical foundation formulation (Agosin et al. 2012; Benbouziane 2018; Cadot et al. 2011; Dadush et al. 2020; Elhiraika and Mbate 2014; Hammouda and Ben 2006; Imbs and Wacziarg 2003; Nieminen 2020) of export product diversification can be summarized as follow:

$${{{\boldsymbol{ED}}}} = {{{\boldsymbol{f}}}}\left( {{{{\boldsymbol{MV}}}},\,{{{\boldsymbol{PV}}}},\,{{{\boldsymbol{PhV}}}},\,{{{\boldsymbol{IV}}}}} \right)$$
(1)

where ED is export product diversification, MV are macroeconomic variables, PV are policy variables, PhV are physical variables and IV are institutional variables.

While some authors use simple linear models (Berthélemy 2005; Hammouda and Ben 2006; Klinger and Lederman 2006), others use non-parametric models (Imbs and Wacziarg 2003), general equilibrium models (Hausmann and Rodrik 2003). In this empirical exercise, we estimate the following equation:

$${{{\boldsymbol{EXCON}}}}_{{{{\boldsymbol{it}}}}} = \lambda _0 + \lambda _1{{{\boldsymbol{EXCON}}}}_{{{{\boldsymbol{it}}}} - 1} + \lambda _2{{{\boldsymbol{TS}}}}_{{{{\boldsymbol{it}}}}} + \lambda _{{{\boldsymbol{j}}}}{{{\boldsymbol{X}}}}_{{{{\boldsymbol{it}}}}} + \eta _{{{\boldsymbol{t}}}} + {{{\boldsymbol{u}}}}_{{{\boldsymbol{i}}}} + \varepsilon _{{{{\boldsymbol{it}}}}}$$
(2)

with i = 1……N, t = 1……T; where EXCONit represents the dependent variable reflecting export concentration, EXCONit−1 represents the lagged variable of the dependent variable, the use of this autoregressive lagged variable (AR1) is motivated by the fact that export product diversification is a slow and dynamic process. TSit represents the set of our variables of interest (the dimensions of services trade). ∑Xit gathers our control variables, ηt represents the period fixed effects, ui a vector represents the country fixed effects, εit represents the error terms capturing all unobserved variables and likely to influence the dependent variable, with E (εit) = 0, i represents the individual (country) and t time period.

4.3 Stationarity and diagnostic tests

Among the different tests that can be used to determine the existence or not of unit roots in panel data, we have the Levin-Lin-Chou (Levin et al. 2002), Harris - Tzavalis (1999), Breitung and Pesaran (2008), Im-Pesaran-Shin (Im et al. 2003) and Fisher-type (Choi 2001) tests with null hypothesis, all panels contain a unit root. The Fisher ADF test is preferred because it does not require highly balanced data and also accepts cross sectional that have deviations. Table 7 in appendix shows that except variables such as the logarithm of tourism, construction services export, natural resources endowment, credit to private sector and merchandise trade openness that are stationary in first difference, all the other variables contain unit roots that is they are stationary at level.

Table 8 in the appendix shows the results of the Wald test for the heteroscedasticity of the errors, the Pesaran and Yamagata (2008) test for the slope heterogeneity and Wooldridge test for autocorrelation of the panel residuals. The significance of the tests performed implies that firstly the errors are heteroscedastic. In other words, the variance of the residuals from the regressions is not constant. The consequence is that in the presence of non-homoscedastic errors, the Ordinary Least Squares (OLS) estimators become inefficient. Second, with Pesaran and Yamagata (2008) test, the significance of all the tests performed imply the rejection of the null hypothesis of homogeneity of the model slope coefficients. Third, the Wooldridge tests show the presence of first order autocorrelation in the residuals. Thus, all these show the need for an estimation method that can take into account the heterogeneity of the slopes and solve the problems of heteroscedasticity and autocorrelation.

4.4 Estimation strategy

Our model is a multiple linear model, and several estimation methods are available to estimate such model. The most common basic method of estimation is the OLS method with fixed effects or random effects estimation. Among the widely and sustained methods of estimation, there is the Two Stage Least Squares (2SLS) method. However, due to the heteroscedastic and autocorrelation issues, the OLS method can lead to biased estimates due to the violation of certain assumptions such as the autocorrelation of errors, the heteroscedasticity of errors, and the endogeneity of certain variables.Footnote 13 Also, the pairwise correlation test shows a significant correlation between our explanatory variables that makes also OLS and within estimators inconsistent. In such situations, some alternative and consistent estimators are recommended such as the Generalized Method of Moments (GMM).

The Generalized Method of Moments (GMM) is based on the conditions of orthogonality between the lagged endogenous variables and the error terms, i.e., absence of correlation between the lagged endogenous variable and the error terms. There are two types of GMM methods. The first one is the difference GMM developed by Arellano and Bond (1991) which can face the problem of over-identification. The second type is the system GMM proposed by Arellano and Bover (1995). This method combines first difference equations and level equations. In the two-step system GMM method, the instruments in the first difference equation are expressed in level and the instruments in the level equation are expressed in first difference.

As highlighted by Roodman (2009), GMM is appropriate in panels with small T (15) and large N (48), meaning few periods and many individuals (N > T), a linear functional relationship, a dynamic left-hand side variable (dependent variable), dependent on its own past realizations, and independent variables that are not strictly exogenous, meaning that they are correlated with past and possibly current realizations of the error. For all these reasons, the two-step system Generalized Methtod of Moments (GMM) is the appropriate method to estimate Eq. (2).

5 Results and discussion

This section presents and discusses the empirical results on the effects of trade in services on export product diversification in Sub-Saharan Africa.

5.1 Heterogeneous effects of trade in services on export product diversification in SSA

Using the two-step system Generalized Methtod of Moments (GMM) and performed several calibrations, the empirical results are presented in Table 9. The results of the diagnostic tests show that all models are well specified. The Hansen test does not reject the validity of instruments (Hansen test p values ≥ 10), and the absence of second-order serial correlation is also not rejected (AR (2) p values ≥ 10). Too many instruments can severely weaken and bias the Hansen over-identifying restrictions test and, therefore, the rule of thumb is that the number of instruments should be less than the number of countries (Roodman 2009). In all tables, the number of countries is more than the number of instruments, indicating that there is no problem of instruments proliferation.

Realizing the presence of outliers that can bias the results since they have extreme values for some variables, we performed regressions by excluding the outliers to check the sensibility of trade in services on export product diversification. The results found from regressions without outliers are the same to the first results with outliers. We first ran the regressions with services export (dimensions) as explanatory variables and then with services import. From the different regressions, only the services export present convincing results. All the dimensions as well as total services imports show positive and insignificant coefficients, which confirms Fig. 6 on the correlation between services imports and export product diversification. Thus, our main focus has been made on the effect of services exports on export product diversification.

In the analysis of the results, an independent variable with a negative sign implies that this variable leads to a decrease in the concentration of exports and an increase in exports product diversification. An explanatory variable with a positive coefficient leads to an increase in exports concentration and to a decrease in exports product diversification.

Table 9 shows that export concentration index (HHI) initial variable is positive and significant at the 1 and 5 % levels in all specified models. This is not only supporting the findings of Agosin et al. (2012), Elhiraika and Mbate (2014), Fosu and Abass (2019), but also the economic theory arguing that export dynamics is a long-run implication than in the short-run. These results support the idea of the dependence of African countries on their export product diversification trajectories developed by Elhiraika and Mbate (2014) and supported then by Fosu and Abass (2019).

All our variables of interest show the expected signs according to economic theory. Indeed, total services exports, transport services, travel services, insurance services, financial services, other business services (services to enterprises), creative economy services have coefficients with negative sign and are all significant at 5 or 10%. The exceptions are exports in ICT and in creative economy that show coefficients with negative sign but are not significant while export in construction services shows coefficient with positive sign but not significant. In that regard, there is a negative relationship between total services export and export concentration in SSA. In addition, the results show that an increase in transport and travel services exports leads to a decrease in SSA export concentration. The coefficients of transport and travel services are negative and significant at 10% levels. That is an increase in transport and travel services exports leads to a decrease in export concentration. This positive relationship between transport services and export product diversification confirms the thesis that the efficiency of transport services determines the ability of firms to compete in foreign markets (Casas 1983; Francois and Wooton 2001; Strandenes 2021). To a situation where transport costs is high, exporting firms must pay lower wages to workers in order to remain competitive or either accept lower returns to capital or must be more productive. For a country whose exports are made possible by imported transport services, facing exorbitant transport costs not only reduces the competitiveness of its firms but also affects the productive capacity and the number of products dedicated to export.

The negative relationship between exports in insurance services, financial services and export concentration reinforces Foley and Manova (2015) idea that financial frictions and the use of domestic capital markets influence multinationals decisions about where and how to produce. These results further support Nieminen (2020) findings that argue that access to banking and financial services through the development of the financial sector and banking structure positively affects the microstructures of the export sector as well as the behavior of exporters which will have effects on export product diversification at the macro level, through the number of active export lines and the concentration among active export lines. Indeed, any increase in exports of financial services and insurance services leads to a decrease of export concentration in SSA. The negative relationship between other business services (services to enterprises) exports and export concentration highlights the important role that services provide to enterprises play as intermediaries in the production process and productivity of firms (Arnold et al. 2008, 2011; Jones and Kierzkowski 1990; Malchow-Møller et al. 2015; Su et al. 2020).

There is a negative relationship between tourism and export concentration due to the negative sign of the coefficient of the variable which is also significant at the 10% level. The increase in the number of tourists on arrival leads to a decrease in export concentration in SSA. This implies that the development of tourism sector leads to a greater diversification of export product in SSA. A theoretical foundation of these findings can be tied to Lejárraga and Walkenhorst (2013), Lin and Sung (1984) and Romão (2020) who show that tourism demand is always accompanied by significant spillover effects on other activity sectors due to the growing demand for consumption goods complementary to tourism. In the same line, Lejárraga and Walkenhorst (2013) argue that since tourism services are consumed locally, tourists will demand a variety of products and services to satisfy their needs, which encourages the visited country to increase the supply of consumer goods.

However, despite the substantial literature (Biryukova and Matiukhina 2019; Hausmann et al. 2007; Luong and Nguyen 2021; Xing 2018) on the catalytic role of ICT in the growth of firms’ productive performance, the non-significance of the coefficients of exports of ICT services, construction services and creative economy services can be explained in part by the fact that the exports of these services in SSA, even if they are increasing over time, are still low compare to the other commercial services. For instance, Fink et al. (2005) in their estimates using disaggregated data reveal that communication costs are more important for trade in differentiated products than for trade in homogeneous products.

The control variables show mostly the expected signs, while some give rather mixed signs depending on the specification. The negative and significant relationship between inflation, FDI, credit to the private sector and export concentration is valid in the theoretical field insofar as it confirms some previous findings. The results on inflation are in line with Balavac and Pugh (2016) who find that permanent instability in the price level is unfavorable for export product diversification. The negative relationship between FDI, credit to the private sector and export concentration confirms the findings of Agosin et al. (2012), Balavac and Pugh (2016), Elhiraika and Mbate (2014) and of Fosu and Abass (2019) who in their work find favorable effects of these variables to the diversification of a country’s export product basket. An environment conducive to FDI and access to credit by the private sector increases the productive structure and competitiveness of firms. Human capital formation is a strong lever for economic diversification because it increases the workforce skills and productivity. In fact, human capital is a strong determinant of export product diversification since countries where population show higher levels of education are more likely to boost export product diversification (Elhiraika and Mbate 2014). However, in this study, despite having negative relation with export concentration, human capital is not significant in all our regressions. That can be explained by the low level of education in most of SSA countries. In addition, most of African manufacturing industries are driven by imported expertise meanwhile local workforce is dedicated to low qualification work.

Natural resources and GDP per capita encourage export concentration in SSA. A country with a large endowment of natural resources has a high propensity to export more raw materials than manufacturing goods, which explains the positive relationship between natural resources and export concentration (Agosin et al. 2012; Ansu et al. 2016; Elhiraika and Mbate 2014). Political stability is very important for export product diversification as argued by Fosu and Abass (2019). Political stability shows rather mixed results as in some specifications it shows positive signs and others negative signs, but all are not significant. This is the result of the stabilisation and pacification efforts observed in some countries over the past decades, although in most cases in SSA political stability remains an ongoing quest. Openness expresses here as the openness in merchandises trade appears to be an important determinant of exports product diversification since in most of the specifications there is a negative relationship with export concentration, but the coefficients are not significant. As so far countries are open to international markets, they are more likely to diversify their exports. The non-significant of the coefficient can be related to the fact that SSA exports are dominated by commodities. This is in line with some findings like Agosin et al. (2012), Elhiraika and Mbate (2014), Feng et al. (2021), Khalil (2019) and Makhlouf et al. (2015).

The time effect incorporated in the regressions reveals a rather interesting feature. The negative and significant relationship of years 2007 to 2009 shows that the period of the international financial crisis has a favorable effect on exports product diversification in SSA. The reason is that during the financial crisis, most of African countries faced difficulties in selling their production due to the global demand contraction for commodities. Faced with such a situation, efforts to transform commodities domestically have emerged in several countries in SSA that are heavily dependent on commodities, such as Nigeria, South Africa, Kenya, Ethiopia, Ghana and Angola.

5.2 Robustness checks of the results

The checking of the robustness of our results is done by three main types of distinct procedures. First, we perform additional estimations by changing the dependent variable. We use the Theil index of export products concentration from the International Monetary Fund (IMF) database instead of the UNCTAD HHI index. Theil index calculated from the seminal work of Cadot et al. (2011) is used in several empirical works (Agosin et al. 2012; Fosu and Abass 2019; Gnangnon 2020a; Nieminen 2020) as part of the work on export product diversification.

Secondly, in the same vein, we classify the main types of commercial services into two broad categories, traditional services (transport, travel and tourism) and modern services (construction, insurance, finance, use of license rights, ICT, business services and creative services). In contrast to Eichengreen and Gupta (2013),Footnote 14 we classify financial and insurance services as modern services following the classification used by Sahoo and Dash (2017) for two main reasons.Footnote 15 The following equation models the relationship between trade in services and export product diversification captured by the Theil index. With THEILit the Theil index, THEILit−1 the lagged variable of the dependent variable, TiSit capturing the dimensions of trade in services, Xit a set of control variables, ϑt and τi the temporal and individual effects respectively and εit the error terms.

$${{{\boldsymbol{TI}}}}_{{{{\boldsymbol{it}}}}} = \delta _0 + \delta _1{{{\boldsymbol{TI}}}}_{{{{\boldsymbol{it}}}} - 1} + \delta _2{{{\boldsymbol{TiS}}}}_{{{{\boldsymbol{it}}}}} + \delta _{{{\boldsymbol{j}}}}{{{\boldsymbol{X}}}}_{{{{\boldsymbol{it}}}}} + \vartheta _{{{\boldsymbol{t}}}} + \tau _{{{\boldsymbol{i}}}} + \varepsilon _{{{{\boldsymbol{it}}}}}$$
(3)

with i = 1……N, t = 1……T

Tables 10 and 11 show that the estimates parameters are stable regarding their signs and amplitudes by using Theil index. This implies a confirmation of the robustness of the results, in the sense that most of our previous results are confirmed. In contrast to the results in Table 9, with the Theil Index, only export in creative economy services is not significant. However, if most of our interest variables become significant and with the expected sign (negative), the other control variables appear in Table 9 with more heterogeneous effect on export product diversification compared to the results with HHI in Table 8. Moreover, Table 11 shows that traditional and modern services export promote export product diversification either with Herfindahl-Hirschman index or Theil index. In both cases, these results are in line with the findings of Eichengreen and Gupta (2013). The results suggest that exports of modern services have greater effects than traditional services on export product diversification.

Third, to have a broad database over the period 1996 to 2019, we reprocess the data from two UNCTAD datasets by combining the bases of the fifth and sixth editions of the Balance of Payments Manual.Footnote 16 Thus, with the new database, we re-estimated Eqs. (2) and (3) whose results can be founded in Tables 12 and 13. The latest results show that there are no major variations in the magnitude of the coefficients of the different models’ parameters as well as the signs with respect to the previous results. Therefore, we can assert that the results are stable and robust weather we use data from BPM6 or from the combined BPM5 and BPM6.

6 Conclusion

This study has identified the dimensions of trade in services that are conducive to export product diversification in SSA. The analysis of the theoretical foundations of the relationship between trade in services through its different components and export product diversification led us to resort to empirical method to identify the dimensions of services that promote export product diversification. The two-step system Generalized Methtod of Moments (GMM) is used as an estimation method. The results show that total export of services, export of traditional and modern services, export of transport, travel, insurance, financial, licences serivces and tourism promote export product diversification in SSA.

Given these results, a number of economic policies can be suggested to the SSA policy makers to make commercial services export an important lever for export product diversification. In this perspective, policy makers can adopt policies and strategies to develop and orient the services sector towards more efficient and high value-added services. In this sense, as export of modern services exert more favorable effects, policy makers can focus more on increasing a stock of highly skilled human capital. This will allow for a better reallocation of labor from the services sector to be more productive and capable of providing more skilled and competitive services as is the case in South East Asia. They can also strengthen national strategies for the development and modernization of the tourism sector as a lever for export product diversification.