The Gravity of China's African Export Promise

Africa’s largest trade partner, China, criticised for exchanging resources for manufactures, has promised to increase imports and optimise the structure of trade with Africa. Using a gravity model of China’s imports for the years 1995-2009, we explore potential dynamics for this promise, uniquely accounting for market economy recognition and Taiwan recognition. The former is associated with increased imports, while the latter effect is ambiguous and statistically insignificant. Comparison of projected against actual imports across three growth-path-aligned economic geography typologies - resource-rich; landlocked and resource-poor; coastal and resource-poor – sets out China’s imports trends in an abstract framework of African export potential. We find not only ‘under’ importing across a majority of resource-poor countries. We also find that current trade policy is the least applicable to these comparatively poor exporters’ trade with China. If the latter are to serve a broader catalytic role in Africa’s regional industrial transformation as compared to the role of coastal and resource poor countries in regional economic transformation in Asia and Latin America, China-Africa trade and investment policies may need additional thinking.


Introduction
In 2009 China became Africa's largest trade partner, a symbolic turning point in extra-regional economic relations after centuries of colonial-centric ties. Bilaterally China is now the largest export destination of South Africa, sub-Saharan Africa's (SSA) largest economy, and also of Angola, Benin, Congo (Democratic Republic), Mauritania, Sudan and Zambia, among others (IMF, 2010). This growing China-Africa trade relationship is the culmination of changes that gained momentum in the mid-1990s. Our gravity model analysis of China's imports from Africa between 1995 and 2009 explores the potential for trade to generate positive development opportunities for Africa.
The visit of then Chinese President Jiang Zemin to six African nations in 1996 marked the shift in focus of China's relationship with Africa away from politics toward economics (Alden, 2007: 15). Change in both China and Africa motivated the shift. In 1991, China became a net oil importer, and rapid domestic growth prompted China to look for security of energy and other raw material supplies.
Africa was seen as potential source of still undeveloped resources. In Africa, domestic economic and political change was afoot too. Apartheid came to an end in South Africa in 1994, while the economies of most of SSA began to perform better after 1995 (Arbache and Page, 2007a). In the preceding 20 years nearly all countries south of the Sahara saw zero or negative economic growth per capita (Radelet, 2010:1).
Absolute trade volumes have risen more than 100-fold since 1990 (Brautigam, 2010:1), with a 10-fold increase since 2000. The importance of trade however differs markedly for China and African economies. The share of China in Africa's trade is now higher than for any region other than the share of China in Developing Asia's total trade (Arora and Vamvakidis, 2010), with China also the region's largest trade partner. China's trade with SSA however remains less than trade with Japan alone: 3.4 per cent of exports of $1.203trn and 4.3 per cent of imports of $1.00trn in 2009 (IMF, 2010).
These increases in trade volume and more market-driven ties between China and Africa are helping to correct undervaluation of Africa by investors (Wang, 2007).
They are also helping to generate some of the world's fastest growth rates (The Economist, 2011). At the political level these relations are fostered through ministerial-level discussions with African counterparts at the China-instigated President Xi Jinping committed China to "enlarge the scale of China-Africa trade, and optimise the trade structure". He promised that China would seek opportunities to increase its imports. 2 China's 2010 inaugural white paper on economic ties between China and Africa, China-Africa Economic and Trade Cooperation (PRC State Council, 2010), recognised the need to optimise the level and composition of trade.
Multiple factors influence the notion of 'optimal' trade, both by level and composition. In the field of economic geography trade levels and prospects have been linked to endowment level and geographic position. Empirically, 'over' and 'under' trading have been estimated using gravity models. First applied empirically to international economics by Tinbergen (1962) and Pöyhönen (1963), the gravity model links trade levels to 'distance' (costs of trade) and 'mass' (GDP, population, and so on). Theoretical foundations have since been established (Anderson, 1979;Bergstrand, 1985Bergstrand, , 1989Helpman and Krugman, 1985;Deardorff (1998).
Here we use a gravity model to scope China's promise to increase the level of trade, in this case implicitly and necessarily with reference to China's own international import norms. A global sample better estimates trade potential; it aligns China's trade promise to increase the level of imports toward African economic development, recognises the direction of trade flows, and finally avoids the effect that an Africa-only sample would largely reflect oil trade.
Variables such as common language and common colony are typically included in gravity models, but these are less applicable to China. Instead our gravity model is uniquely augmented to include a 'huaqiao' (overseas China) variable capturing high densities of Chinese outside China proper in Taiwan, Hong Kong and Singapore, alongside variables for the recognition of China as a market economy, and the recognition of Taiwan. We also remove the East Asian island economies that play a leading role in China's re-export chain from our island dummy variable, ensuring the sign presumably is negative for our 'under' and 'over' estimations of African exports to China.
Our results are interpreted to recognise the developmental hurdle that is Africa's uniquely complex economic geography. Africa has "a massive land area divided into 44 countries, with overall a low population density compared with other lowincome regions" (Collier, 2008:3). By the 1990s, only 35 percent of Africa's population was living in coastal, resource-scarce economies compared with 88 percent for the rest of the developing world, while resource rich economies accounted for 30 percent of populations in Africa compared with only 11 percent elsewhere (Collier and O'Connell, 2007:2). The most striking difference between Africa and other developing regions lies in the proportion of the population living in landlocked and resource-scarce countries -this proportion was about 30 percent of Africa's population, against 1 percent elsewhere (Collier, 2008: 5).
As a result, compared with Asia and Latin America over the last half-century where coastal and resource-poor economies had led regional industrialization transitions, in sub-Saharan Africa these countries have performed poorly (Growth Report, 2008). This trait is compared both to countries of the same grouping elsewhere and also relative to resource-rich and landlocked sub-Saharan African countries against the relevant international benchmark.
Our economic modeling and economic geography analysis serve both to give quantitative scope to China's promise to increase its imports from Africa, and to better understand the longer-term dynamism of recent trends in China's trade with Africa. Our framework enhances the analysis of contemporary Chinese trade patterns while also identifying the potential for African development. INSERT TABLE 1 ABOUT HERE   Table 1 groups sub-Saharan Africa countries into three economic geography typologies. Observers of China's rising oil imports from Africa may be surprised by the classification of former Sudan as resource-poor. Resource rich economies are defined as those generating more than 10 percent of GDP from primary commodity rents, thus reflecting the productive value of each country's relative land abundance. Our classification follows Collier and O'Connell (2007), who used a 10 percent threshold because price volatility may lead some economies to flip backwards and forwards across this threshold year-to-year for their sample period of 1960-2000. 3 No matter the classification of Sudan and Congo DRC, our results highlight the scale of the 'under' importing from non-resource rich Africa. More worryingly is that China and Africa have no policies in place to assist directly coastal and resource-poor Africa, despite their recognised development needs. 3 The classification is complicated by the fact that exploitation of land is endogenous to GDP itself, as well by the opportunities of non-land GDP.
The rest of this paper is structured as follows: the second section outlines relevant previous research; the third section presents the stylistic facts defining contemporary China-Africa trade and some of the policy variables affecting that relationship; the forth section explains the gravity model and discusses some recent relevant gravity papers; the fifth section describes the data sources and the estimation technique used for the gravity model in the present paper, while the sixth section presents our results; the seventh section discusses the implications and limitations of the results, and the eighth section concludes.

China-Africa Trade: Previous Research
China-Africa trade-related research has distinguished the direct from indirect effects of China on African economies. Effects may also be complementary or competitive. Jenkins and Edwards (2005) summarise the effects of Asian drivers, China and India, on African economies in Table 2. INSERT TABLE 2 ABOUT HERE An example of an indirect effect is that China is driving up demand for resources, while its manufactured exports push down prices for would-be competitors (Kaplinsky, 2005). Resource rich countries, including Angola, Congo DRC, Guinea and Mauritania are winners (Stevens and Kennan, 2005). Mauritius and Madagascar in contrast have suffered adverse shifts in both import and export prices (Zafar, 2007). An additional indirect effect related to China's demand for some of Africa's resources are the effects described as Dutch disease, the displacement of manufacturing and agricultural activity by resource extraction. Bova (2008) estimated 2,000 jobs were lost in Zambia's horticulture sector owing to Dutch Disease. The country's largest textile mill -built with Chinese aid in the 1970s -was reputedly closed down due to clothing imports from China (Kaplinsky and Morris, 2007). Giovannetti and Sanfilippo (2009) used gravity modelling to study direct and competitive displacement effects of China's trade on African trade. The displacement effect was significant for textiles, clothing and footwear, with more limited impact on light manufactures that used medium-level technology. The absence of regional production networks in Africa, they believe, makes a replication of East Asia's 'flying geese' 4 success more difficult. Like Chen et al (2006) and Venables (2008), they found that third market trade preferences, in the traditional markets of the US and Europe, were important for preserving African market access in the face of the competitiveness of third party exporters. Brautigam (2010) argues China has recognised the role of trade and leveraging dynamic comparative advantage in its own development could provide lessons for Africa along lines of 'flying geese' industrialisation. Where Giovannetti and Sanfillippo (2009) undertake an aggregated cross-country study, Brautiguam's cites country-level cases of spillovers through partnerships with Chinese firms, within the textiles industry. Broadman (2007) used a gravity model to explore China-Africa FDI flows, finding investment from China into Africa complements rather than substitutes for bilateral export flow.
Although such papers reveal the longer-term picture may be brighter than current trends suggest, cross-country conclusions are inconsistent. For example, fears that the competitive effects of China's manufacturing exports are swamping African economies as a whole are unsupported empirically. Exaggeration of such effects, Shafaeddin (2004) argues, stem from underestimation of China's own declining cost advantages, and also estimation difficulties arising from the difference in structure of labour-intensive production in China and other countries. Similarly, calculations of the rank correlation of China's exports against those of other developing country exporters, often reach contradictory conclusions. For example, only half of such country-level results of Jenkins and Edwards (2006) agree with those of Stevens and Kennan's (2005).

Stylistic facts
The Heckschler-Ohlin (HO) trade model emphasises the role of dynamic factor endowments as the main driver of trade, rather than the static comparative 4 Akamatsu (1961Akamatsu ( , 1962 proposeda multi-tiered hierarchical 'flying geese' model to explain the sequencing of East Asia's industrial development. The model describes how industrialization spreads from developed to developing countries: the initial 'goose' (the frontier economy) leads the second tier 'geese' (developing economies) and these in turn are followed by third-tier geese (least developed economies) through a process of gradual outsourcing. An alternative version of the model is the Vernon product cycle model. advantage model of Ricardo. Under HO assumptions, countries will export the good that uses intensively the factor with which it is most abundantly and cheaply endowed and import products that use scarce factors (WTO, 2008: 32). China reflects well these predictions, exporting labour-intensive manufactures and importing raw material resources. INSERT

Policies
The absence of diplomatic relations between two countries may be expected to increase the formal costs of trade. For the PRC, the first 30 years after 1949 was a period of relative economic autarky during which international relations and trade were shaped in part by timing of diplomatic recognition. Trade and aid throughout this period served as tools of Beijing's foreign policy goal to establish its claim as the legitimate government of China in place of the Kuomintang regime on Taiwan.
Among the first countries to recognise Beijing were Guinea and Sudan in 1959, followed by Ghana (1960), Congo DRC (1961, Kenya (1963), and Benin and Congo Republic (1964). The year of recognition also depended on when the particular African state became independent too.
Several countries have switched diplomatic ties, several times, over the decades.
Only Swaziland, a member of the Southern Africa Customs Union, has consistently recognised Taipei. Such shifts in recognition annul previous bilateral Foreign Affairs or Ministry of Commerce representations. We hypothesize that a shift to Taipei will increases trade policy distance with China, and adversely affects trade flows, while recognition of Beijing will shorten effective distance. 5 Market economy status is important when firms from non-market economies are accused of dumping goods in a market economy, in which case they can be challenged within the local trade jurisdiction of the market economy (Green, 2004).
Goods identified as selling at 'less than fair value' can be claimed to cause 'material injury' to the industry of the complainant market economy and thus lead to the imposition of counter-veiling duties (Panitchpakdi and Clifford, 2002:196).
Since the burden to disprove a dumping allegation falls on the non-market economy, political and bureaucratic whims may put at risk the business of exporters from non-market designated economies. 6 We assume recognition of China as a market economy to be positively associated with bilateral trade flows.
There is an inherent risk here that early recognition of China as a market economy is more likely from trade complementary countries. For Africa the pattern is inconsistent: among so-classified resource-rich countries, Angola had not recognised China within our survey timeframe though Nigeria had; among coastal and resource-poor East Africa, Tanzania had not, but Kenya had. That our variable is time-variant also allows the capture of potential related trade variation within and between groups across time, such that a potential for face-giving first-mover advantage is implicitly tested.
In the absence of an official public list of countries that have granted China market economy status ahead of the WTO timeframe, we compiled a list from several sources, including the Ministry of Commerce's website, reciprocal African government websites and WTO sources, double checking against local newspapers. China has also fostered the establishment of Special Economic Zones (SEZ) in four countries in Africa -Ethiopia, Mauritius, Nigeria (two) and Zambia. Drawing on the Chinese experience of SEZs, these seek to increase the value-added of exports and to promote industrial clustering. Early research questions if these African zones can emulate China's success (Brautigam et al, 2010;Brautigam, 2011). Our modeling found no significant relationship between presence of an SEZ and exports to China, at this stage.

The Model
Gravity modelling has been applied extensively to research exploring China's trade.
Benassy-Quere and Lahreche-Revil (2003)  finding that the greatest effect was for consumer goods. Greenaway, Mahabir and Milner (2006) found similarly that China displaced greatest the exports of the more advanced Asian exporters, an effect which has increased with time. Batra (2004) found China to be 'under-trading' with India, since which time China has become India's largest trading partner.
Gravity modelling has also been used to explore "natural" levels of China's bilateral trade, imports and exports. Bussiere and Schnatz (2006) found that for China's size and location it is already well integrated into the world economy. Edmonds, La Croix and Li (2008) found China had excessive orientation toward foreign trade, though this varied between trade partners. They highlighted that China's trade with Angola, Sudan and Zimbabwe was higher than expected. For this study we focus explicitly on China's imports, and apply the following model: Where: Given the distinctive differences in the economic geography, and our interest in applying the results to explore China's import dynamics and potential with Africa, we augment our model as follows: • Island ij : a dummy variable equal to one for island economies, excluding proximate trade partner island economies, including Japan and Taiwan.
• Huaqiao ij : a dummy equal to 1 for Taiwan, Hong Kong and Singapore to account for shared culture between China and these economies.
We expect the adjusted island dummy to shift from positive to negative accordingly, a result that reflects China's trade with island economies elsewhere.
We expect a positive association between the high presence of overseas Chinese and trade.
China has put in place several distinctive institutional arrangements that potentially impact trade. These include recognition of the One China policy and more recently whether a trade partner has recognised China as a market economy.
Past studies have not sought to take account of these arrangements in modeling

Estimation
Our Relative price shifts between China and a trade partner are also accounted for through use of a real exchange rate variable. We finally reduce potential endogeneity issues between trade levels and GDP by using lagged rather than current GDP.

Results
A summary of our PPML results for the augmented model is presented in Table 5.
These results broadly confirm our hypotheses. They show: • Partner GDP is positively associated with China's import levels but less than in a typical gravity result; the effect of China's GDP is negative and less significant; • The real exchange rate (RMB per unit of foreign currency) between China and a trade partner has a limited and insignificant effect on China's imports; • Distance reduces import levels; • Demand for imports on average rise with trade partner population, while the result for China's population is only significant at the 10 percent significance level; • Landlockedness is negatively associated with exports to China; • China has an atypical tendency to trade with island economies and also inversely so in trade with contiguous neighbours; • China trades more highly with those sharing similar linguistic orientation.
For our newly explored institutional variables: • Recognition of market economy status, ceteris paribus, is associated with increased trade flows; • The effect of non-recognition of the One China policy appears to have a negative effect on trade but this effect is not further explored herein.
INSERT Steps taken to reduce policy distance, in our study proxied through recognition of China as a market economy are associated with an increase in imports. Greater institutional policy distance -the non-recognition of the One China policy -is significantly associated with imports by China, but the results are inconsistent between estimation methodologies. The effect is large and positive under OLS, but relatively small and negative using PPML.

Sensitivity Checks
We perform three sensitivity checks. Firstly we adjust the island dummy to examine if the result defaults to the more usual negative coefficient estimate when the East Asian re-exporting island economies are excluded. Secondly, we explore the China-Africa import effect through use of a dummy variable SSA. Finally, we use three separate dummies to explore the effects of the different economic geographies.
For our island sensitivity check we remove proximate important Chinese trade partner island economies including Japan and Taiwan from the island classification.
The re-run of (2) as expected reverses the sign of earlier coefficient estimate (-0.516, p=0.000), a result consistent with most gravity model findings. Given our interest in trade level projections for countries including Africa's island economies, we adopted this adjustment for all onward estimations.
Re-running newly adjusted equation 2 inclusive of a dummy variable for SSA countries produced a negative but insignificant 'SSA' effect (-0.192, p=0.394).
This was the first indication of our overall under-importing hypothesis. The result however was not consistent over time.
Time period effects within the sample period were tested through division of the sample into two sub-periods, 1995-2001 (t1) and 2002-2009 (t2). The timing pivots around China's WTO ascension in 2001 and the start of its 'Going Out' policy that encouraged Chinese firms to invest abroad. In the first period, the SSA dummy is negative and significant (-0.994, p=0.00), while in the second period the dummy is positive but insignificant (0.019, 0.941). There has been a longstanding 'negative Africa dummy' attached to international growth regressions (Easterly and Levine, 1997). These results suggest that for China-Africa trade the sign of that dummy is changing, even if not yet statistically significantly.
Given signs of a changing 'Africa dummy' and the association of growth patterns across and between economic geography typologies we used dummies for each of the three typologies to explore the economic geography of China's imports from Africa. Using the entire sample period there is a positive effect for resource rich-  Results of analysis over the two mentioned time periods are presented in Table 6.
Compared to the earlier results, the size of the coefficient attached to China's population is much larger in the second period, but significant only at the 10 percent level. The effect of real exchange rate on Chinese imports is relatively small, and sensitive to change across time. Similarly, the impact of China's GDP is inconsistent, changing in sign and significance.

Over and Under Exporting
Despite the gravity model being widely used to make trade projections, this was not its intended purpose. The following results are thus best understood as relative comparisons between and within typologies, rather than absolute measures of the appropriate level of trade.
Factors directly affecting the reliability of these estimates include that our sample relies on Chinese data alone, rather than data adjusted for bilateral discrepancies, which may bias the results. Our sample is global rather than specific to the case of Africa, and scale effects may upwardly bias estimates of 'over-trading', since the model assumes trade to be a linear function of GDP. But the global sample is preferred for two reasons: it better helps to understand Africa's trade and trade potential with China, and secondly it minimises the biases arising from the dominance of trade in oil within an African sample. almost five from the resource-rich typology, even after excluding the largest of those exporters, Angola. In contrast, China 'under' imports, by on average twothirds, with resource-poor economies, whether coastal or landlocked, and similarly excluding the outliers of the biggest regional traders with China.

4.5-2 Coastal and Resource-Poor
Coastal resource-poor economies are more diversified than the other typologies.
Ten of the 17 are LDCs, a smaller proportion than among other typologies. It includes the region's largest economy, South Africa, together with smaller economies such as The Gambia. The grouping also includes Africa's island economies, many of which export nothing to China.  of it serving as a port for some exports of the landlocked countries to its north.
South Africa is an exception. As Africa's largest economy, China's largest trade partner in Africa, and with China as its largest trading partner, South Africa in contrast "over-exports" second only to Angola according to our estimates.
The worst performers, excluded from Figure 2, are the island economies, some of which export nothing to China, including Cape Verde and Comoros. Mauritius's industrial exports to China struggle to remain competitive against China's own industries (Zafar, 2007).
The density of 'under' trading among this typology is worrying because of the long-run implications for development. In particular, such a trade profile means these coastal, resource-scarce economies -with their relative "lack of agglomeration economies in labour-intensive manufactures" -require a combination of temporary protection from Asian competition in OECD markets and "ingenious" growth strategies if they are to become middle-income economies (Venables, 2008: 59;Collier, 2008, p. 8). Yet, there is every indication that these economies are not sufficiently supported by policies to facilitate the development of their trade capacity: no coastal economy plays host to a special economic zone; few coastal and resource-poor economies are LDCs, so few benefit from access to China's trade preferences for Africa. Economies of this typology are also the least likely to have recognised China as a market economy ahead of WTO requirements.

4.5-3 Landlocked and Resource-poor
Eleven of the 13 economies that Collier and O'Connell (2007)  On the basis that the success of China's own SEZs was related to their proximity to ports, it is surprising that one of the four African SEZs that China has fostered is in Ethiopia, a landlocked (and resource-poor) nation.

INSERT FIGURE 3 ABOUT HERE:
Excluding Congo DRC and the former Sudan, as illustrated in Figure 3, all countries in this group are 'under' exporting to China. Congo DRC in contrast exports more than seven times the predicted level and former Sudan thirty times the predicted level. The definition of resource-rich used by Collier and O'Connell (2007) was derived from their sample period .  Table 3).
Burundi is the worst export performer against predicted levels, realising less than 1 percent of predicted exports in 2009.

Conclusion
For decades, the diverse economies of SSA have played a peripheral role in international trade, with the exception of a few oil exporters and South Africa. there is the risk that coastal economies especially will not be able to attain a critical manufacturing threshold in the face of competition from China. China's promise to increase its imports from African economies and to optimise the structure of trade may come to naught in the absence of other policies to promote development.
Our modelling produced results typical of gravity model for GDP and distance variables. We uniquely accounted for the role of island economies in China's trade pattern. We found that steps to reduce policy distance as proxied by recognition of market economy status was, as predicted, associated with increased imports.
Steps to increase general policy distance through non-recognition of the One China policy -that is, recognizing Taipei     Note: Export dependency is the ratio of bilateral exports to one country to total exports; import dependency is calculated similarly.     *, **, *** Indicates significance at the 10%, 5% and 1% level.