Global competitiveness and the diffusion of electronic commerce across diverse industries have extended across borders the functions of sourcing, producing and distributing products and services (Cohen & Mallik 1997; Turban et al. 2017). Distributing products across these diverse markets has become imperative for corporate profitability, but vast differences exist amongst countries in terms of transport time, international shipping risk, market access and logistics infrastructure (Arvis et al. 2010). Data from the World Bank Sustainable Development Goals (SDG) (World Bank 2017) report that significant proportions of the consumer population in emerging market (EM) countries still have limited access to reliable infrastructure and other basic services such as rail road, paved roads and communication services (McKinnon 2012; World Bank Group 2004; World Bank 2017). Kinra and Kotzab (2008) suggest that such differences are fundamental to the supplier selection, facility location and resource allocation problems in the context of macro-level supply chain decisions; and decision-makers are not willing to delay action in the hope that uncertainty in these decision areas and logistics inefficiencies (Havenga, Simpson & De Bod 2013) will abate. A better understanding of these infrastructure and trade interrelationships and persistent SDG trends may inform how decision-makers formulate strategies for managing and integrating the material, information and revenue flows across the global supply chain (Badenhorst-Weiss, Maurer & Brevis-Landsberg 2013).

Academic attention to EMs in the context of trade, logistics and market potential dependency has not kept pace in the face of the trends cited above in international expansion or growth (Quinn & Hilmer 1994; Tatoglu et al. 2016). Whilst researchers do not agree on the extent to which environmental, economic and supply chain infrastructure-related factors impact trade, there is little doubt that trade is influenced by transport infrastructure (Puertas, Martí & García 2014) through its extensive impact on efficiency (Gorman 2006; Takele & Buvik 2019). Recently, there has been renewed interest in advancing the understanding regarding the role of logistics or supply chain infrastructure indicators in environmental and economic context (Esfahbodi, Zhang & Watson 2016; Subramanian & Gunasekaran 2015).

The perspectives outlined above have at least two common threads: (1) given the expanding complexities of macro-level global supply chains and trade, lowering unit sourcing and manufacturing costs do not necessarily translate directly into lower total landed costs per unit (e.g. the total per unit costs associated with importing, exporting, transportation, customs clearance and processing of goods from distant markets) and (2) consumer and industrial demand create exports and import trade flows (and the logistical connections) between countries to fill the shortfall in domestic production and supply. However, the increased trade flow volume can itself directly affect supply chain costs, increase uncertainty, increase congestion and lessen the efficiency of logistics infrastructure, trade exports and imports and market potential (Kwon & Beom 2012; Puertas et al. 2014; Takele & Buvik 2019) for both EMs and developed markets (DMs). A data set comprising 89 countries that represented both groups of countries was developed for this study.

Uncertainty affects managing and integrating the material, information and revenue flows across the international supply chain. Efficiency differences derive from disparate infrastructure endowments and regional diversity across countries (Hausman, Lee & Subramanian 2005; Takele & Buvik 2019). Several studies (Bowe 2006; Havenga et al. 2013; Tirschwell 2007) have observed that domestic and global conditions promoting trade and commerce are clashing with obsolete infrastructure and uncertain market potential. We infer that patterns of trade development in EMs are influenced in part by (1) macro-level (rather than firm-level) logistics infrastructure, (2) the efficiency of such infrastructure and (3) other market potential factors. Therefore, research should include factors related to public infrastructure and country market potential for both EMs and DMs.

This study intended to explore logistics infrastructure, logistics efficiency, market potential and trade performance at the country level for a set of 89 EM and DM countries. We studied whether high market potential and efficient macro-level supply chain infrastructure lead to better trade performance. The objective was to analyse the causal relationships, if any, amongst logistics capability dimensions, market potential and trade in the context of logistics efficiency estimations carried out at the country level. Based upon reviews of the market clustering literature (which focuses on market potential) and logistics expenditure or infrastructure literature, some guiding premises are developed for this study. Efficiency is viewed as moderating key relationships between logistics and trade performance.

Two diverse but content-related streams of earlier work provided the foundation for the current research on market potential and logistics infrastructure. The first stream of work adopted the country-clustering lens and assessed market potential. Some clustering methods (Cavusgil, Kiyak & Yeniyurt 2004) and the market potential weighted-indexing methods were used by Cavusgil (1997) to develop macro-level insights to EMs. Despite strengths, there are several shortcomings of the clustering approach. These include the exclusive reliance on (1) marketing indicators (Papadopoulos & Denis 1988) and environmental macro-factors (Luqmani, Yavas & Quraeshi 1994) and (2) secondary market-oriented data (Papadopoulos & Denis 1988). Whilst indexing provides the potential for grouping countries, it is generally silent on productivity or efficiency differences between groups or members of the groups. These shortcomings provide, in subsequent sections, our motivation for proposing two ordinary least squares (OLS) regression models to incorporate efficiency considerations. Given the proliferation of new data and collection methods by governments and non-governmental organisations (NGOs), EM research remains of interest to NGOs and academics alike (Bagai & Wilson 2006; Takele & Buvik 2019). We performed relative comparisons of EM and DM countries to improve generalisability.

The second stream of work on foreign markets adopted the supply chain infrastructure lens (Bowersox 1992; Bowersox & Calantone 1998; Bowersox, Calantone & Rodrigues 2003; Rodrigues, Bowersox & Calantone 2005). Heskett, Glaskowsky and Ivie (1973) first developed a methodology for estimating country-level logistics cost, applying it to the United States of America. Total logistics cost was defined as the sum of four activities: transportation, inventory, warehousing and order processing. Annual US logistics estimates appear in the State of Logistics Report published annually for over 30 years (Wilson 2006). One major challenge in estimating global logistics and supply chain expenditure is that the data required for direct measurement are not available consistently for developed nations and generally not available for emerging and transitioning countries (Bagai & Wilson 2006). Primary and secondary data must be matched in country-level analyses, but the availability of such data often varies extensively according to country and/or region of the world. The first study of global logistics expenditure by Bowersox (1992) estimated global logistics costs based on four components: total gross domestic product (GDP), government sector product, industrial sector product and total trade ratio. Follow-up studies (Bowersox & Calantone 1998; Bowersox et al. 2003; Bowersox, Closs & Stank 1999; Rodrigues et al. 2005) refined the estimation method by using an Artificial Neural Network model or by including supply chain infrastructure variables related to cost and information or communication systems. Their estimates suggest that global logistics expenditures are 13.8% of the world GDP. Interestingly, they found that logistics efficiency increased in DMs but not in the rest of the world.

Given the literature above, it is worthwhile to incorporate comparisons with data from EM countries included. Such is our proposal in this study, that is, the stream of logistics work above highlights the necessity for logistics and supply chain infrastructure investment and efficiency improvements throughout developing nations. However, it is silent on other institutional considerations, such as customs duties and national account remittances as global trade proxies, for example. It is also silent on exploring the linkage to market potential and then comparing DMs and EMs. These gaps, which our research attempts to address, limit implications for improving trade flows and country-specific logistics infrastructure. In many African EMs, access to rural consumers both grows markets and increases logistics and supply chain complexity and cost (Savage et al. 2014). Clearly, as Clark (2003) and Lorentz et al. (2015) suggest, there is a relationship between supply chain-related infrastructure and costs and the capacity of a country to attract international trade and participate in the global economy. The current research explores more explicit dimensions of macro-level logistics infrastructure along with market potential information (Gorman 2006; Orozco 2010). An integrated conceptual framework (see Figure 1) that bridges the market clustering or ranking literature and the logistics cost or infrastructure literature is proposed in Research methods and design section. This framework proposes performance relationships between the macro input factors and the country-level trade output performance dimensions and addresses two key questions:

This kind of integrative work began widely appearing in the early 2000s (Kinra & Kotzab 2008) and continues to be of research interest today (Havenga et al. 2013). Whilst there has been some early academic interest in this area (Berman 2007; Boarnet 1997; Garrison & Souleyrette 1996; Haas, Murphy & Lancioni 2003; Ross 2002), there are no other published studies on market potential and logistics infrastructure adopted here and amongst both EM and DM countries. We compare such a set of 89 countries using the data envelopment analysis (DEA) efficiency lens and develop OLS regressions to explore the dependencies between trade factors (as dependent variables) and logistics infrastructure (independent variables). Together, these represent our primary and secondary objectives in this study, which are further detailed in subsequent sections.

The conceptual model in Figure 1 forms the basis for our methodological approach to analysing the EMs and then exploring structural differences with DMs used in the data set. As a result, EMs and developed countries alike can be efficient or inefficient based upon the unique DEA score that is computed. Thus, our novel approach fairly evaluates and compares countries on a relatively level basis given our recourse to DEA. Two research hypotheses, corresponding to two equation templates, are proposed.

We propose that there is a direct relationship to trade performance from country macro factors concerning logistics infrastructure and development (Kinra & Kotzab 2008; Weber & Weber 2004). This is our baseline proposal.

Hypothesis 1 (H1): Macro input factors focusing on logistics infrastructure and development are related to trade performance.

More specifically, H1a– H1g concern the following trade-related dependent variables:

All variables in the model are defined in Table 1. Thus, seven linear models of the direct relationships between country-level infrastructure dimensions and trade performance factors will be estimated. Each of these seven serves as a dependent variable in Equation 1 to test Hypothesis 1, as illustrated for H1a: Customs duty (Cduty):

CDutyj=β0+β1*Remitj+β2 *RurlDensj+β3 *RoadDensj+    β4 *RailDensj+β5 *Lnxdysj+β6 *Telemoblj+εj[Eqn 1]

In Equation 1, there are six independent variables (Table 1). Workers’ remittances (Remit), as defined by the International Monetary Fund (IMF), is the transfer of disposable income mainly from foreign nationals and migrant workers in developed countries to family members in their home countries. The IMF views such financial transfers as drivers of consumption in the receiving country (Orozco 2010) because recipients spend domestically more than 60% of remittances on consumer goods and services. Rural density (RurlDens), road density (RoadDens) and rail density (RailDens) capture the concentrations of potential consumers, road infrastructure and rail infrastructure, respectively. Lnxdys is the natural log of average export clearance time for containers, whilst Telemobl captures the concentration of cell phone and landline usage. These are used as indicators of logistics infrastructure capability.

H1: Equation 1 can be extended to include structural efficiency and market potential. We propose that structural differences between developed and EM countries could be reflected in the relative efficiency with which country input resources are combined to generate trade activity through various logistics infrastructure capabilities. These differences impact the proposed direct relationships in H1. Countries consume different levels of the macro input factors considered earlier, but EMs have their own unique, diverse constraints (Tatoglu et al. 2016; World Bank 2017). For example, some are very mountainous and have little in the way of navigable road networks to reach rural markets, whilst others have very limited railroad or communications infrastructure (Savage et al. 2014). This results in varying reliance on available modes of distribution from country to country. Therefore, we propose that there are differences in trade performance because of structural differences in efficiency and market potential. Thus, our second research hypothesis is formulated.

H2: There are structural differences in efficiency and differences in market potential between countries that cause variance in trade performance (in addition to the logistics infrastructure and development factors included in H1).

This results in H2a–H2g for the same trade performance dependent variables as before:

A linear model is proposed regarding the relationship amongst country-level macro factors, MPIs and the extent of moderation by efficiency. Moderation is modelled by the interaction of DEA efficiency scores with each of the six independent variables in Equation 1; MPIs, representing independent control variables, are added to Equation 1 as main effects. Thus, Equation 2, illustrated for customs duties (CDuty), has the following terms (Table 1):

CDutyj=β0+β1*Remitj+β2 *RurlDensj+β3 *RoadDensj+β4    *RailDensj+β5 *LNxDysj+β6 *Telemoblj+β7 *DEAj    *Remitj+β8 *DEAj *RurlDensj+β9 *DEAj*RoadDensj+β10    DEAj *RailDensj+β11 *DEAj *LNxDysj+β12 *DEAj    *Telemoblj+β13 *CRiskj+β14 *Mktszj+β15 *MktGrthj+    β16 *MktRecptj+β17 *MktIntenj+β18 *Mktstruc+εj[Eqn 2]

There are six logistics infrastructure and development dimensions (β₁ – β₆), the same six as in Equation 1; six corresponding efficiency interaction effects DEA*{β₇ *DEA_j *Remit_j + β₈ *DEA_j *RurlDens_j + β₉ *DEA_j *RoadDens_j + β₁₀ *DEA_j *RailDens_j + β₁₁ *DEA_j *LNxDys_j + β₁₂ *DEA_j *Telemobl_j} (β₇ – β₁₂); and six MPIs (β₁₃ – β₁₈). For the efficiency interaction effects, a measure of relative efficiency was calculated using DEA for each country, based on the macro factors and trade performance factors as input. A country is inefficient if its score is less than 1.0; scores are between 0.00 and 1.00. A country is classified as efficient and it is on the efficient frontier if its score is 1.0; this implies that it is not possible to increase any of its outputs without increasing any of the inputs (or, alternately, it is not possible to decrease any inputs without decreasing some outputs). For inefficient countries, the higher the DEA score, the closer that country is to the efficient frontier. Finally, the six MPIs in Equation 2 include country risk (Crisk), market size (Mktsz), market growth (MktGrth), market receptivity (MktRecpt), market intensity (MktInten) and market structure (MktStruc).

The results are presented in three sections below. Firstly, the results from the DEA efficiency analysis are presented; these efficiency scores are necessary in the estimation of the regressions in Equation 2. Next are the results for two proposed models: H1 – Equation 1 is presented first, followed by H2 – Equation 2.

This study proposed relationships between trade performance and logistics infrastructure, structural efficiencies and market potential; it explored these relationships using secondary data from 89 economies. Data envelopment analysis (for efficiency scores) and OLS regression were used to demonstrate the importance of both logistics infrastructure and market potential in determining trade performance. We have offered insights into the connection between macro logistics capability, efficiency and the performance of national environments or countries.

Equation 1 had six macro-logistics variables serving as the independent variables in six regressions. Overall, the results of Equation 1 (H1) show that (1) five of the seven independent variables studied were significantly predicted by at least one of the six independent variables; (2) of the five significant models, two were predicted by exactly one independent variable; and (3) in several instances, negative relationships were unexpectedly found. These mixed results were perhaps partly because of country-level differences in MPIs or because of differences in structural efficiencies, none of which were included in Equation 1. Thus, our motivation for examining Equation 2 (H2) seems validated.

Equation 2 had six macro-logistics variables, six corresponding efficiency interaction effects and six MPIs, for a total of 18 independent variables in each of the seven regressions. Overall, inclusion of efficiency considerations and market potential improved these models over the baselines estimated in Equation 1. The results of Equation 2 show that all seven regression models were significant overall (in terms of model p-values). In addition, each of the seven dependent variables was predicted by at least three of the independent variables. Overall, our proposed dependent and independent variables as modelled in the hypothesised Equation 2 seem to account for a substantial proportion of the variance in the data.

In four of seven significant models, at least one of the efficiency interaction components was significant, supporting the idea that structural efficiency is important. In particular, there are significant country-level differences in the ability to move air freight efficiently, and these seem to be partly because of structural differences, as calculated using DEA-derived efficiency scores and modelled as moderation effects. In all seven significant models, at least one of the MPIs was significant, supporting the notion that country differences in market potential make a difference. In particular, amongst MPI variables, the assembly or production of high-technology products for export, as well as airfreight, seems highly dependent on market size and receptivity. Assuming these specific types of high-tech products are frequently moved via airfreight, it is interesting to observe the consistency in direction and significance of these two MPIs.

We presented a discussion of logistics and supply chain infrastructure indicators, market potential and trade performance in the context of efficiency estimations. It is important to observe that our analysis shows there are several resource-capacitated countries that were determined to be DEA efficient in our modelling. Thus, the designation as a ‘good performing’ country is not (and should not be) reserved for large-sized countries or markets with high levels of macro-logistics capability as reported in our data set. Improving macro-logistics capability has become an important development policy objective for many governments and industry-based organisations because supply chain infrastructure is considered critical in attracting and sustaining business activity in developed and developing economies alike. Decision-makers must also understand and monitor the country-specific and regional environment context as they design supply chain strategies and execute market selection or entry plans; overall market potential is important. This study offers insights and motivates the need for more academic research on other perspectives related to measuring and comparing macro-logistics infrastructure, market potential, efficiency and trade performance. This might set the stage for bringing together policymakers and leaders in the private sector for collaborative investments aimed at further facilitating trade and improving the scope and efficiency of logistics and supply chains.

This stream of work could be extended by examining (1) other dimensions related to infrastructure or trade (i.e. other variables within the categories established in this research) or (2) relevant categories other than those considered here, e.g., this research focused on country as the unit of analysis, but these countries could be divided into regions or clusters; or they could be classified into those that are landlocked, that are island nations (e.g. Indonesia, New Zealand and others) or that have large coastal regions. As another example, consider such variables as terrain size, the location of shipping ports and the access to population centres via air or land transport. Improving roadway and railway network densities can influence the distribution of imports to local consumers as well as producers’ abilities to export their goods. Density is not the same as efficiency (we considered only efficiency); for some countries, density per square mile or per population may be an informative metric. Finally, our findings are not restricted to a given industry. Some industries may have specific logistics requirements that are not modelled in this research. Future studies should include an investigation that segments the countries into groups (e.g. geographic regions or World Bank income classifications, or World Trade Organization (WTO) developed vs. EMs). Together, these limitations represent future research opportunities.