Search
Close
Search
Advanced Search
Search Menu

Abstract

Sub-Saharan Africa's (SSA) physical geography is often blamed for its poor economic performance. A country's geographical location does, however, not only determine its agricultural conditions or disease environment. It also pins down a country's relative position vis-à-vis other countries, affecting its ease of access to foreign markets. This paper assesses the importance of market access for manufactures in explaining the observed income differences between SSA countries over the period 1993–2009. We construct yearly, theory-based measures of each SSA country's market access using the information contained in bilateral manufacturing trade flows. Using these measures, we find a robust positive effect of market access on economic development that has increased in importance during the last decade. Interestingly, when further unraveling this finding, access to other SSA markets in particular turns out to be important.

Sub-Saharan Africa (SSA) is home to the world's poorest countries. Alongside factors such as poor institutional quality, low (labour) productivity and low levels of human capital, the region's geographical disadvantages are often viewed as an important determinant of its dismal economic performance. A country's geography directly affects economic development through its effect on disease burden, agricultural productivity, the availability of natural resources, or its accessibility (see Gallup and others 1999; Collier and Gunning, 1999; Ndulu, 2007; Nunn and Puga, 2011).

Recently, the new economic geography (NEG) literature (see Krugman, 1991; Fujita and others 1999; World Bank 2008) has highlighed another mechanism through which geography affects a country's prosperity. A country's location not only determines its physical geography; it also pins down its position on the globe vis-à-vis all other countries (its relative geography). The NEG literature in particular emphasizes the important role of relative geography in determining a country's access to international markets. It predicts that the better this market access, the higher a country's level of income.

Redding and Venables (2004) were the first to establish empirically that market access indeed matters for economic development. Based on the estimation results for a worldwide sample of 101 countries, they find for example that if Zimbabwe were located in central Europe, the resulting improvement in its market access would ceteris paribus increase its per capita income by almost 80 percent. Subsequently, several other studies confirmed the positive effect of market access on economic development (e.g. Knaap 2006; Breinlich 2006; or Mayer, 2008). More recently, it has also been found to hold for developing countries (see Deichmann, Lall, Redding and Venables, 2008 for a good overview). Amiti and Cameron (2007) show that wages are higher in Indonesian districts that enjoy better market access, Hering and Poncet (2010) and Bosker and others (2010) find similar evidence in case of Chinese cities, and Fally and others (2010) do so for Brazilian states.1 The importance of relative geography in shaping global and regional patterns of economic development has also not gone unnoticed in policy circles: it was the main topic of the World Bank's 2009 World Development Report (World Bank, 2008).

Despite the attention given to the role of economic geography in shaping patterns of economic development in both the developing and developed world, we are unaware of a study that empirically establishes whether, and if so, to what extent, it can help to explain the observed differences in economic development between SSA countries.2

SSA is only a marginal player in the world's export and import markets. Since 1970, the region's share in global trade has declined from about 4 percent to a mere 2 percent in 2005 (IMF, 2007). Through their detrimental effect on market access, high trade costs are generally viewed as one of the main causes for SSA's poor trade performance (see Freund and Rocha, 2011; Collier, 2002; Foroutan and Pritchett, 1993; Coe and Hoffmaister, 1999; Limao and Venables, 2001; Amjadi and Yeats, 1995; Portugal-Perez and Wilson, 2008). Increasing SSA participation in world markets, as well as stimulating trade relations between SSA countries, is viewed as very important to its future economic success (IMF, 2007; World Bank 2007; or UNCTAD 2010). Especially, expanding the (exporting) manufacturing sector is seen as crucial to the region's chances on future economic success (Collier and Venables, 2007; IMF, 2007; World Bank, 2007). It has been one of the key ingredients of the sustained growth witnessed in the rapidly developing Asian countries (see e.g. Johnson et al., 2006; or Jones and Olken, 2008). Developing an exporting manufacturing sector will not only help to diversify SSA countries' export portfolio, making them less vulnerable to price fluctuations on world commodity markets, it is also expected to increase overall productivity through increased knowledge spillovers and learning by doing (Van Biesebroeck, 2005; Bigsten and Söderbom, 2006).

As a result, improving the region's market access by investing in infrastructure, stimulating regional integration, or providing preferential access to European and U.S. markets are all seen as a vital ingredients for improving SSA's trade potential and its overall economic performance (Buys et al., 2010; UNCTAD, 2009, 2010; Frazer and van Biesebroeck, 2010; Freund and Rocha, 2011).

Against this background the main contribution of our paper is to empirically establish the importance of SSA market access, and market access for manufactures in particular, for its economic development over the last two decades.3 To do this, we follow the empirical strategy introduced by Redding and Venables (2004) that is firmly based in the theoretical new economic geography (NEG) literature. We first construct yearly measure(s) of each SSA country's market access over the period 1993–2009, making use of bilateral manufacturing export data involving at least one SSA country. Next, and by using our constructed measure(s) of market access, we estimate the impact of market access on GDP per worker. We do adapt the Redding and Venables (2004) strategy in three different ways. First, when constructing our market access measure(s) using the information contained in bilateral trade flows, we take explicit account of the fact that most SSA countries only trade with a fraction of their possible trade partners. Second, our 17-year sample period allows us to use panel data methods and control for time-invariant unobserved heterogeneity in all our estimations (hereby most notably capturing all possible differences in SSA countries' physical geography). Finally, we distinguish explicitly between the importance of access to other SSA markets and to markets in the rest of the world (ROW).

Overall, our main finding is that economic geography is an important determinant of SSA's economic development. Even after controlling for many other posited explanations of SSA's poor economic performance such as its physical geography, education levels, or institutional quality, market access for manufactures has a significant positive effect on GDP per worker. The effect of market access that we find for SSA countries is, however, significantly lower than that found in comparable studies looking at Brazil (Fally and others 2010), Indonesia (Amiti and Cameron, 2007), or China (Hering and Poncet, 2010). But, although lower than those found in other parts of the world, our results do show that the effect of market access has increased markedly in SSA over the last two decades: the positive relationship between market access and economic development is strongest in the second half of our sample period.

Interestingly, when further unraveling this finding by distinguishing between the importance of access to other SSA markets and to markets in the rest of the world (ROW), we find that it is the variation in access to other SSA markets in particular that drives our findings. ROW market access loses its significance after controlling for other (more standard) explanations for SSA's poor economic performance. It confirms the (increased) importance of SSA markets for SSA's own economic development (see also Easterly and Reshef, 2010, footnote 2; UNCTAD, 2009, 2010).

Finally, based on our estimation results, we tentatively ‘decompose’ the contribution of policy-relevant variables to overall market access, and look at the predicted effect of several (policy induced) changes aimed at improving SSA market access. This shows that improving SSA infrastructure (see also Buys and others, 2010), alleviating the burden of landlocked countries, and increasing regional economic integration, all positively affect a country's market access in varying degrees, carrying important benefits for its economic development.

I. Economic Development and Market Access: Theoretical Framework and Empirical Strategy

At the heart of our analysis lies the theoretical relationship between market access in manufactures and income levels that follows from standard economic geography theory. Referring to Appendix C for a more formal exposition of the NEG model that underlies our analyses,4 this relationship is shown in log-linear form in equation (1) below (corresponding to equation (C5) in Appendix C, with an added subscript t to denote years):
(1)

Equation (1) is the so-called wage equation that lies at the heart of virtually all empirical NEG studies (see e.g. Hanson, 2005; Redding and Venables, 2004; Knaap, 2006 and Amiti and Cameron, 2007; Hering and Poncet, 2010). It predicts that wages in country i in year t, wit, are higher the better a country's production efficiency, cit, and, most importantly for our present purposes, the better its so-called real market access MAit.5

This market access is a trade cost (Tij) weighted sum of all countries' market capacities (mj). Each country j's contribution to country i's market access consists of country j's market capacity, a reflection of its real spending power, weighted by the level of trade costs incurred when shipping goods from country i to country j, i.e. MAij = mj /Tijσ−1. The closer (or better connected) a country is to world markets, the better its market access. It is equation (1) that constitutes the backbone of our empirical analysis into the relevance of market access for SSA economic development (see section III).

Estimating the Wage Equation: The Redding and Venables (2004) Approach

Estimating equation (1) is not as straightforward as it may seem. The difficulty comes from the fact that the market access term, MAit, is not directly observed. Two estimation strategies have been proposed to deal with this issue.

The first strategy follows Hanson (2005) and estimates equation (1) directly. A drawback of this method is that it requires additional assumptions on how to proxy each country's market capacity, mj. Particularly problematic in this respect is the fact that a country's price index of manufacturing varieties, one of the two main components of mj (see Appendix C) is not directly observed.6

This is why we base our empirical analysis on the second proposed strategy to estimate (1). This method does not face the problem of having to make ad-hoc assumptions on how to proxy a country's market capacity. It was first introduced by Redding and Venables (2004)7 and involves a two-step procedure. In a first step, additionally collected information contained in (bilateral) trade data is used to provide estimates of the (relative) role of trade costs, Tij, and market capacity, mj, in determining a country's market access. The way this is done is firmly based on NEG theory. As derived in Appendix C, equation (C6) shows that the connection between bilateral exports and market access follows directly from a standard NEG model (where we have again added a subscript t to denote years).
(2)
Exports EXij from country i to country j depend on the ‘supplier capacity’ of the exporting country, si (see (C6) for its definition), the market capacity of the importing country, mj, and the magnitude of bilateral trade costs Tij between the two countries. Comparing MAij in (1) and (2) immediately shows that we can estimate (2), and use the resulting predicted values of MAijt to construct yearly measures of each country's market access. More formally:
A. Estimate the bilateral export equation (2) in log-linear form using information on bilateral export flows and trade costs, capturing each country's supplier and market capacity by a full set of importer-year and exporter-year dummies:
(3)
B. Use formula and formula⁠, the estimated parameters on the included importer-year dummies and on trade costs respectively, to construct each country's market access based on the direct relationship between bilateral exports and market access [again compare (1) to (2)]:
(4)
The constructed measure of each country's market access shown in (4) is then used in the second step to get an estimate of the impact of market access on income levels:
(5)
where the error term ηit in (5) captures a country's level of technological efficiency [cit in (1)]. The estimated value of χ2, together with its standard deviation, is the most important parameter for the purpose of our paper. It provides us with an indication of the size, sign, and significance of the effect of market access on income levels.

In the next two sections we implement the above-described two-step method in order to verify the importance of market access for SSA economic development over the last two decades (1993–2009). In section II, we focus on estimating the trade equation and constructing our measure(s) of market access. Next, in section III, we estimate the wage equation making use of our constructed measures of market access and show that market access is of increasing importance in understanding the observed differences in economic development between SSA countries. Moreover, we decompose each country's market access into access to other SSA markets and to markets in the ROW, and show that having good access to other SSA markets has become particularly important over the last two decades.

II. Estimating the Trade Equation and Constructing Market Access

The starting point of our empirical analysis is the trade equation (3) capturing each country's supplier and market capacity by an exporter-year and importer-year dummy respectively. In order to estimate (3), we collected information on yearly bilateral manufacturing exports to and from SSA countries over the period 1993–2009.8 We take this data from the UN COMTRADE database, focusing on manufacturing goods as defined by the Standard International Trade Classification (SITC Rev.3). It contains information on bilateral manufacturing export flows from each SSA country to and from 47 other SSA countries and 153 countries in the rest of the world.

We think a particular focus on manufacturing exports is warranted for two important reasons. First, it most closely follows the NEG theory that underlies our analysis. NEG theory only predicts a relationship between market access in the manufacturing sector and income levels (see Appendix C). It is not evident from theory that a similar relationship should hold for primary goods' market access (trade patterns of which are more likely to be dominated by more standard comparative advantage or Heckscher-Ohlin type forces). Taking total export flows when estimating (3) is likely to give biased estimates of the parameters needed to build our measures of market access, particularly in SSA where overall exports are dominated by exports of natural resources and/or agriculture (although this dominance varies substantially between SSA countries, see Figure B2 and Table B2 in Appendix B).9

Second, developing the (exporting) manufacturing sector is viewed by many as crucial to the region's chances on future economic success (Collier and Venables, 2007; IMF, 2007; World Bank, 2007). Previous spells of sustained growth (mostly experienced by Asian countries) were all accompanied by a rapid expansion of international trade, and trade in manufacturing goods in particular (see e.g. Johnson and others 2006 or Jones and Olken, 2008). In this respect it is interesting to note that manufacturing goods already dominate SSA exports to the rest of Africa (see UNCTAD, 2009; 2010 [see Annex 4.5 for a country-by-country overview]). They constitute an average 40 percent of total intra-SSA exports. When disregarding primary exports (fuel, ores, minerals, etc.), that account for roughly 75 percent of total exports to the ROW, the same is true of SSA exports to the ROW.

Next, we need to decide on how to measure trade costs, Tijt. The NEG-model does not specify trade costs in any way (except that they are of the iceberg type). In the absence of actual trade cost data and following the modern empirical trade and economic geography literature (see e.g. Anderson and van Wincoop, 2004; Limao and Venables, 2001; Redding and Venables, 2004; Bosker and Garretsen, 2010), we specify Tijt to be a multiplicative function10 of the following observable variables that are commonly used in the literature11: bilateral distance (Dij), sharing a common border (Bij), a common language (CLij), or a common colonial heritage [distinguishing between sharing a common colonizer (CCij) and having had a colony-colonizer relationship (CRij)], and finally a dummy variable indicating membership of the same African regional or free trade agreement (RFTAijt) in year t (see Appendix A for a full list of the RFTAs included in this definition). In loglinear form this amounts to substituting the following trade costs specification for β ln Tijt in (3):
(6)
Overall, this results in the following bilateral export equation that we estimate for each year of our sample period separately (which explains the added subscript t to all coefficients):
(7)

Equation (7) forms the basis for constructing our yearly measures of each SSA country's market access over the 1993–2009 period.

Estimating the Trade Equation. Dealing with the ‘Zeroes’ in Bilateral SSA Trade

The actual estimation of (7) raises a number of issues of its own. In particular, the presence of zero trade flows complicates matters. The average SSA country exports manufacturing goods to only 20 percent of possible partner countries, so that about 80 percent of bilateral SSA manufacturing export flows are zeroes. And, although the number of zeroes drops over our sample period (from 89 percent in 1993 to 77 percent in 2003), it does complicate matters when estimating the parameters of (7) that we need to construct our measures of market access. Failing to adequately take account of these zeroes results in inconsistent estimates of these parameters, and thus in wrongly constructed market access measure(s).

To deal with these zero observations, several estimation strategies have been proposed that each have their (dis)advantages. We follow Helpman and others (2008) and use a Heckman 2-step estimation strategy to estimate the parameters of (7). This method has the virtue of not having to impose exogenous sample selection, that is, assuming that there is no unobserved variable related to both the probability to trade and the amount of trade [as e.g. discarding the zero observations and applying OLS on the non-zeroes only, or applying zero-inflated Poisson or negative binomial methods do]. Nor do we have to assume a priori that the exact same model explains both the zero and the non-zero bilateral trade flows [as e.g. using Tobit, NLS or pseudo-Poisson (PPML) techniques would imply].12

The Heckman 2-step procedure amounts to first estimating, using probit, how each of the included explanatory variables affects the probability to trade. Next, in the second stage, the effect of each variable on the amount of trade is estimated, including the inverse Mills ratio (constructed using the results from the first step) to control for the endogenous selection bias that would plague the results when simply discarding the non-zero observations (see for instance ch. 17 in Wooldridge, 2003). However, using the Heckman 2-step procedure is also not free of assumptions: results are only convincing when one can rely on a valid exclusion restriction: a valid exclusion restriction: that is, having at least one variable that determines the probability to trade but not the amount of trade (see Wooldridge, 2003, p. 589).13

The choice of such a variable is generally quite difficult. However, in our case we can build on a recent paper by Helpman and others (2008), and use their suggested measure of the religious similarity of two countries as the variable explaining the probability to trade but not the amount of trade conditional upon trading. The economic rationale behind the use of this variable is firmly based on recent trade models that show that in order to trade at all, exporters have to be able to cover the fixed costs of exporting. The higher these costs between two countries, the higher the probability of not observing any bilateral trade between them. Helpman and others (2008, p. 466) show that religious (dis)similarity serves as a useful proxy of these fixed costs,14 and they moreover show convincing evidence that, econometrically, it can not be rejected as a valid ‘instrument’.15

Using Helpman and others's (2008) religious similarity variable to fulfill the (necessary) exclusion restriction, we estimate (7) for each year in our sample period separately.16 To explicitly allow intra-SSA trade to be differently influenced by our included variables, we interact all variables, including all importer- and exporter-dummies, with an intra-SSA trade dummy-variable. Table 1 shows the results. In Table 1, the postfix “- SSA” denotes that a variable is interacted with this intra-SSA trade dummy. Significance of an “SSA”-variable indicates a significantly different effect of that particular variable on intra-SSA trade than on SSA trade with the ROW. The coefficients give the overall effects of each of the included variables on the amount of trade (after taking the first stage into account) and the results for 0/1 trade refer to the estimated coefficients in the first stage probit estimations.

Table 1.
Open in new tab

Trade Equation with Importer and Exporter Dummies

Dep: ln Manuf Exports1993–2009
Coefficients eq. (7)0/1 trade
ln dist−1.96***−0.97***
[0.00][0.00]
ln dist - SSA−0.44*−0.38
[0.24][0.14]
Contiguity0.921.16*
[0.38][0.18]
Contiguity - SSA0.33−0.56
[0.57][0.36]
Com. lang.0.81***0.45***
[0.00][0.00]
Com. lang. - SSA0.230.02
[0.42][0.50]
Com. col.0.76***0.30**
[0.003][0.03]
Com. col. – SSA−0.180.42
[0.51][0.11]
Colonizer1.55***0.80
[0.00][0.19]
RFTA0.73*−0.25
[0.32][0.43]
RFTA - SSA0.620.44
 [0.32][0.60]
Religion HMR−0.613**
−[0.034]
P-value Mills' ratio[0.000]***−
nr. obs.16560
Censored13344 (80%)
Uncensored3216 (20%)
Dep: ln Manuf Exports1993–2009
Coefficients eq. (7)0/1 trade
ln dist−1.96***−0.97***
[0.00][0.00]
ln dist - SSA−0.44*−0.38
[0.24][0.14]
Contiguity0.921.16*
[0.38][0.18]
Contiguity - SSA0.33−0.56
[0.57][0.36]
Com. lang.0.81***0.45***
[0.00][0.00]
Com. lang. - SSA0.230.02
[0.42][0.50]
Com. col.0.76***0.30**
[0.003][0.03]
Com. col. – SSA−0.180.42
[0.51][0.11]
Colonizer1.55***0.80
[0.00][0.19]
RFTA0.73*−0.25
[0.32][0.43]
RFTA - SSA0.620.44
 [0.32][0.60]
Religion HMR−0.613**
−[0.034]
P-value Mills' ratio[0.000]***−
nr. obs.16560
Censored13344 (80%)
Uncensored3216 (20%)

Notes: We report estimated coefficients and not marginal effects. Marginal effects are available upon request. The coefficients are used as input in the construction of our market access measures. These coefficients actually differ by year, the numbers reported in the Table are the mean coefficients, mean p-values (in brackets), and mean number of observations over the period specified. *, **, *** denotes significant at the 5 percent level in at least 50, 80, or 100 percent of the years. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 1.
Open in new tab

Trade Equation with Importer and Exporter Dummies

Dep: ln Manuf Exports1993–2009
Coefficients eq. (7)0/1 trade
ln dist−1.96***−0.97***
[0.00][0.00]
ln dist - SSA−0.44*−0.38
[0.24][0.14]
Contiguity0.921.16*
[0.38][0.18]
Contiguity - SSA0.33−0.56
[0.57][0.36]
Com. lang.0.81***0.45***
[0.00][0.00]
Com. lang. - SSA0.230.02
[0.42][0.50]
Com. col.0.76***0.30**
[0.003][0.03]
Com. col. – SSA−0.180.42
[0.51][0.11]
Colonizer1.55***0.80
[0.00][0.19]
RFTA0.73*−0.25
[0.32][0.43]
RFTA - SSA0.620.44
 [0.32][0.60]
Religion HMR−0.613**
−[0.034]
P-value Mills' ratio[0.000]***−
nr. obs.16560
Censored13344 (80%)
Uncensored3216 (20%)
Dep: ln Manuf Exports1993–2009
Coefficients eq. (7)0/1 trade
ln dist−1.96***−0.97***
[0.00][0.00]
ln dist - SSA−0.44*−0.38
[0.24][0.14]
Contiguity0.921.16*
[0.38][0.18]
Contiguity - SSA0.33−0.56
[0.57][0.36]
Com. lang.0.81***0.45***
[0.00][0.00]
Com. lang. - SSA0.230.02
[0.42][0.50]
Com. col.0.76***0.30**
[0.003][0.03]
Com. col. – SSA−0.180.42
[0.51][0.11]
Colonizer1.55***0.80
[0.00][0.19]
RFTA0.73*−0.25
[0.32][0.43]
RFTA - SSA0.620.44
 [0.32][0.60]
Religion HMR−0.613**
−[0.034]
P-value Mills' ratio[0.000]***−
nr. obs.16560
Censored13344 (80%)
Uncensored3216 (20%)

Notes: We report estimated coefficients and not marginal effects. Marginal effects are available upon request. The coefficients are used as input in the construction of our market access measures. These coefficients actually differ by year, the numbers reported in the Table are the mean coefficients, mean p-values (in brackets), and mean number of observations over the period specified. *, **, *** denotes significant at the 5 percent level in at least 50, 80, or 100 percent of the years. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 2.
Open in new tab

Market Access and Economic Development in SSA

Dep: ln GDP Per Worker12345
ln MA0.067***0.031***0.021***0.0110.031**
[0.000][0.003][0.006][0.129][0.039]
      
Polity IV−−−0.0020.0020.0004
−−[0.559][0.633][0.964]
Urbanization rate−−−0.0090.01−0.005
−−[0.333][0.473][0.812]
Gr. prim. enrollment−−−0.002−0.0010.0004
−−[0.179][0.659][0.815]
% Oil in GDP−−0.018***0.026**0.001
−−[0.000][0.018][0.887]
Civil war−−−0.159**−0.218**0.015
−−[0.015][0.044][0.809]
Civil conflict−−−0.055−0.03−0.053
−−[0.112][0.374][0.163]
% Agriculture in gdp−−−0.014***−0.012**−0.02***
−−[0.001][0.012][0.001]
ln working pop. dens.−−−0.022−0.2751.043
−−[0.943][0.623][0.114]
Nr. obs775775583268315
Time-period1993–20091993–20091993–20091993–20002001–2009
P-value country FE−[0.000][0.000][0.000][0.000]
P-value year FE−[0.000][0.000][0.000][0.001]
Dep: ln GDP Per Worker12345
ln MA0.067***0.031***0.021***0.0110.031**
[0.000][0.003][0.006][0.129][0.039]
      
Polity IV−−−0.0020.0020.0004
−−[0.559][0.633][0.964]
Urbanization rate−−−0.0090.01−0.005
−−[0.333][0.473][0.812]
Gr. prim. enrollment−−−0.002−0.0010.0004
−−[0.179][0.659][0.815]
% Oil in GDP−−0.018***0.026**0.001
−−[0.000][0.018][0.887]
Civil war−−−0.159**−0.218**0.015
−−[0.015][0.044][0.809]
Civil conflict−−−0.055−0.03−0.053
−−[0.112][0.374][0.163]
% Agriculture in gdp−−−0.014***−0.012**−0.02***
−−[0.001][0.012][0.001]
ln working pop. dens.−−−0.022−0.2751.043
−−[0.943][0.623][0.114]
Nr. obs775775583268315
Time-period1993–20091993–20091993–20091993–20002001–2009
P-value country FE−[0.000][0.000][0.000][0.000]
P-value year FE−[0.000][0.000][0.000][0.001]

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent respectively. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 2.
Open in new tab

Market Access and Economic Development in SSA

Dep: ln GDP Per Worker12345
ln MA0.067***0.031***0.021***0.0110.031**
[0.000][0.003][0.006][0.129][0.039]
      
Polity IV−−−0.0020.0020.0004
−−[0.559][0.633][0.964]
Urbanization rate−−−0.0090.01−0.005
−−[0.333][0.473][0.812]
Gr. prim. enrollment−−−0.002−0.0010.0004
−−[0.179][0.659][0.815]
% Oil in GDP−−0.018***0.026**0.001
−−[0.000][0.018][0.887]
Civil war−−−0.159**−0.218**0.015
−−[0.015][0.044][0.809]
Civil conflict−−−0.055−0.03−0.053
−−[0.112][0.374][0.163]
% Agriculture in gdp−−−0.014***−0.012**−0.02***
−−[0.001][0.012][0.001]
ln working pop. dens.−−−0.022−0.2751.043
−−[0.943][0.623][0.114]
Nr. obs775775583268315
Time-period1993–20091993–20091993–20091993–20002001–2009
P-value country FE−[0.000][0.000][0.000][0.000]
P-value year FE−[0.000][0.000][0.000][0.001]
Dep: ln GDP Per Worker12345
ln MA0.067***0.031***0.021***0.0110.031**
[0.000][0.003][0.006][0.129][0.039]
      
Polity IV−−−0.0020.0020.0004
−−[0.559][0.633][0.964]
Urbanization rate−−−0.0090.01−0.005
−−[0.333][0.473][0.812]
Gr. prim. enrollment−−−0.002−0.0010.0004
−−[0.179][0.659][0.815]
% Oil in GDP−−0.018***0.026**0.001
−−[0.000][0.018][0.887]
Civil war−−−0.159**−0.218**0.015
−−[0.015][0.044][0.809]
Civil conflict−−−0.055−0.03−0.053
−−[0.112][0.374][0.163]
% Agriculture in gdp−−−0.014***−0.012**−0.02***
−−[0.001][0.012][0.001]
ln working pop. dens.−−−0.022−0.2751.043
−−[0.943][0.623][0.114]
Nr. obs775775583268315
Time-period1993–20091993–20091993–20091993–20002001–2009
P-value country FE−[0.000][0.000][0.000][0.000]
P-value year FE−[0.000][0.000][0.000][0.001]

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent respectively. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 3.
Open in new tab

Foreign Market Access and Economic Development in SSA

Dep: ln GDP Per Worker12345678
ln DMA0.004*0.004*0.020.02−−−−
[0.087][0.087][0.151][0.195]−−−−
ln FMA0.009−0.048**−0.011−0.053**−
[0.679]−[0.036][0.606]−[0.019]−
ln FMA - ROW−−0.087−−0.039−−0.075−−0.039
−[0.396]−[0.655]−[0.435]−[0.665]
ln FMA - SSA−0.014−0.049**−0.016−0.054***
−[0.370]−[0.023]−[0.311]−[0.004]
     
Controls:see Table 2
    
nr. obs268268315315268268315315
time-period1993–20001993–20002001–20092001–20091993–20001993–20002001–20092001–2009
p-value
country FE[0.000][0.000][0.000][0.000][0.000][0.000][0.000][0.000]
year FE[0.000][0.000][0.001][0.001][0.000][0.000][0.001][0.001]
Dep: ln GDP Per Worker12345678
ln DMA0.004*0.004*0.020.02−−−−
[0.087][0.087][0.151][0.195]−−−−
ln FMA0.009−0.048**−0.011−0.053**−
[0.679]−[0.036][0.606]−[0.019]−
ln FMA - ROW−−0.087−−0.039−−0.075−−0.039
−[0.396]−[0.655]−[0.435]−[0.665]
ln FMA - SSA−0.014−0.049**−0.016−0.054***
−[0.370]−[0.023]−[0.311]−[0.004]
     
Controls:see Table 2
    
nr. obs268268315315268268315315
time-period1993–20001993–20002001–20092001–20091993–20001993–20002001–20092001–2009
p-value
country FE[0.000][0.000][0.000][0.000][0.000][0.000][0.000][0.000]
year FE[0.000][0.000][0.001][0.001][0.000][0.000][0.001][0.001]

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent, respectively.Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 3.
Open in new tab

Foreign Market Access and Economic Development in SSA

Dep: ln GDP Per Worker12345678
ln DMA0.004*0.004*0.020.02−−−−
[0.087][0.087][0.151][0.195]−−−−
ln FMA0.009−0.048**−0.011−0.053**−
[0.679]−[0.036][0.606]−[0.019]−
ln FMA - ROW−−0.087−−0.039−−0.075−−0.039
−[0.396]−[0.655]−[0.435]−[0.665]
ln FMA - SSA−0.014−0.049**−0.016−0.054***
−[0.370]−[0.023]−[0.311]−[0.004]
     
Controls:see Table 2
    
nr. obs268268315315268268315315
time-period1993–20001993–20002001–20092001–20091993–20001993–20002001–20092001–2009
p-value
country FE[0.000][0.000][0.000][0.000][0.000][0.000][0.000][0.000]
year FE[0.000][0.000][0.001][0.001][0.000][0.000][0.001][0.001]
Dep: ln GDP Per Worker12345678
ln DMA0.004*0.004*0.020.02−−−−
[0.087][0.087][0.151][0.195]−−−−
ln FMA0.009−0.048**−0.011−0.053**−
[0.679]−[0.036][0.606]−[0.019]−
ln FMA - ROW−−0.087−−0.039−−0.075−−0.039
−[0.396]−[0.655]−[0.435]−[0.665]
ln FMA - SSA−0.014−0.049**−0.016−0.054***
−[0.370]−[0.023]−[0.311]−[0.004]
     
Controls:see Table 2
    
nr. obs268268315315268268315315
time-period1993–20001993–20002001–20092001–20091993–20001993–20002001–20092001–2009
p-value
country FE[0.000][0.000][0.000][0.000][0.000][0.000][0.000][0.000]
year FE[0.000][0.000][0.001][0.001][0.000][0.000][0.001][0.001]

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent, respectively.Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 4a.
Open in new tab

IV-Results

dep: ln GDP per worker1234
ln FMA0.091−0.313***−
[0.316]−[0.008]−
ln FMA - ROW−−0.059−−0.273*
−[0.533]−[0.065]
ln FMA - SSA−0.064−0.176***
−[0.249]−[0.001]
Controlssee Table 2 (no country FE)
nr. obs268268315315
time-period1993–20001993–20002001–20092001–2009
F-stat. instrument14.02−7.44−
FMA - ROW−198.1−143.94
FMA - SSA−17.62−18.41
p-value over ID-test[0.415][0.269][0.501][0.822]
dep: ln GDP per worker1234
ln FMA0.091−0.313***−
[0.316]−[0.008]−
ln FMA - ROW−−0.059−−0.273*
−[0.533]−[0.065]
ln FMA - SSA−0.064−0.176***
−[0.249]−[0.001]
Controlssee Table 2 (no country FE)
nr. obs268268315315
time-period1993–20001993–20002001–20092001–2009
F-stat. instrument14.02−7.44−
FMA - ROW−198.1−143.94
FMA - SSA−17.62−18.41
p-value over ID-test[0.415][0.269][0.501][0.822]

Notes: P-values, based on robust standard errors, in brackets. ***, **, * denotes significance at 1, 5, or 10 percent, respectively.

Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 4a.
Open in new tab

IV-Results

dep: ln GDP per worker1234
ln FMA0.091−0.313***−
[0.316]−[0.008]−
ln FMA - ROW−−0.059−−0.273*
−[0.533]−[0.065]
ln FMA - SSA−0.064−0.176***
−[0.249]−[0.001]
Controlssee Table 2 (no country FE)
nr. obs268268315315
time-period1993–20001993–20002001–20092001–2009
F-stat. instrument14.02−7.44−
FMA - ROW−198.1−143.94
FMA - SSA−17.62−18.41
p-value over ID-test[0.415][0.269][0.501][0.822]
dep: ln GDP per worker1234
ln FMA0.091−0.313***−
[0.316]−[0.008]−
ln FMA - ROW−−0.059−−0.273*
−[0.533]−[0.065]
ln FMA - SSA−0.064−0.176***
−[0.249]−[0.001]
Controlssee Table 2 (no country FE)
nr. obs268268315315
time-period1993–20001993–20002001–20092001–2009
F-stat. instrument14.02−7.44−
FMA - ROW−198.1−143.94
FMA - SSA−17.62−18.41
p-value over ID-test[0.415][0.269][0.501][0.822]

Notes: P-values, based on robust standard errors, in brackets. ***, **, * denotes significance at 1, 5, or 10 percent, respectively.

Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 4b.
Open in new tab

Lagged Market Access

Dep: ln GDP Per Worker1234
ln FMA−0.006−0.037*−
[0.695]−[0.072]−
ln FMA - ROW−0.005−−0.01
−[0.951]−[0.908]
ln FMA - SSA−−0.007−0.042**
−[0.540]−[0.024]
Controlssee Table 2 (also lagged)
nr. obs203203299299
time-period1993–20001993–20002001–20092001–2009
Dep: ln GDP Per Worker1234
ln FMA−0.006−0.037*−
[0.695]−[0.072]−
ln FMA - ROW−0.005−−0.01
−[0.951]−[0.908]
ln FMA - SSA−−0.007−0.042**
−[0.540]−[0.024]
Controlssee Table 2 (also lagged)
nr. obs203203299299
time-period1993–20001993–20002001–20092001–2009

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent, respectively. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 4b.
Open in new tab

Lagged Market Access

Dep: ln GDP Per Worker1234
ln FMA−0.006−0.037*−
[0.695]−[0.072]−
ln FMA - ROW−0.005−−0.01
−[0.951]−[0.908]
ln FMA - SSA−−0.007−0.042**
−[0.540]−[0.024]
Controlssee Table 2 (also lagged)
nr. obs203203299299
time-period1993–20001993–20002001–20092001–2009
Dep: ln GDP Per Worker1234
ln FMA−0.006−0.037*−
[0.695]−[0.072]−
ln FMA - ROW−0.005−−0.01
−[0.951]−[0.908]
ln FMA - SSA−−0.007−0.042**
−[0.540]−[0.024]
Controlssee Table 2 (also lagged)
nr. obs203203299299
time-period1993–20001993–20002001–20092001–2009

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent, respectively. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 4c.
Open in new tab

Supplier Access

dep: ln GDP Per Worker123456
ln MA0.011*−−0.031**−−
[0.096]−−[0.029]−−
ln FMA−0.010−−0.050**−
−[0.646]−−[0.018]−
ln FMA - ROW−−−0.056−−−0.042
−−[0.597]−−[0.581]
ln FMA - SSA−−0.031*−−0.050**
−−[0.080]−−[0.016]
ln DMA−−−−−−
−−−−−−
ln SA0.001−−0.002−−
[0.806]−−[0.729]−−
ln FSA−−0.003−−0.007−
−[0.851]−−[0.536]−
ln FSA - ROW−−0.091−−0.042
−−[0.247]−−[0.451]
ln FSA - SSA−−−0.023**−−0.034***
−−[0.044]−−[0.006]
Controlssee Table 2
Nr. obs268268268315315315
Time-period1993–20001993–20001993–20002001–20092001–20092001–2009
dep: ln GDP Per Worker123456
ln MA0.011*−−0.031**−−
[0.096]−−[0.029]−−
ln FMA−0.010−−0.050**−
−[0.646]−−[0.018]−
ln FMA - ROW−−−0.056−−−0.042
−−[0.597]−−[0.581]
ln FMA - SSA−−0.031*−−0.050**
−−[0.080]−−[0.016]
ln DMA−−−−−−
−−−−−−
ln SA0.001−−0.002−−
[0.806]−−[0.729]−−
ln FSA−−0.003−−0.007−
−[0.851]−−[0.536]−
ln FSA - ROW−−0.091−−0.042
−−[0.247]−−[0.451]
ln FSA - SSA−−−0.023**−−0.034***
−−[0.044]−−[0.006]
Controlssee Table 2
Nr. obs268268268315315315
Time-period1993–20001993–20001993–20002001–20092001–20092001–2009

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent, respectively. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 4c.
Open in new tab

Supplier Access

dep: ln GDP Per Worker123456
ln MA0.011*−−0.031**−−
[0.096]−−[0.029]−−
ln FMA−0.010−−0.050**−
−[0.646]−−[0.018]−
ln FMA - ROW−−−0.056−−−0.042
−−[0.597]−−[0.581]
ln FMA - SSA−−0.031*−−0.050**
−−[0.080]−−[0.016]
ln DMA−−−−−−
−−−−−−
ln SA0.001−−0.002−−
[0.806]−−[0.729]−−
ln FSA−−0.003−−0.007−
−[0.851]−−[0.536]−
ln FSA - ROW−−0.091−−0.042
−−[0.247]−−[0.451]
ln FSA - SSA−−−0.023**−−0.034***
−−[0.044]−−[0.006]
Controlssee Table 2
Nr. obs268268268315315315
Time-period1993–20001993–20001993–20002001–20092001–20092001–2009
dep: ln GDP Per Worker123456
ln MA0.011*−−0.031**−−
[0.096]−−[0.029]−−
ln FMA−0.010−−0.050**−
−[0.646]−−[0.018]−
ln FMA - ROW−−−0.056−−−0.042
−−[0.597]−−[0.581]
ln FMA - SSA−−0.031*−−0.050**
−−[0.080]−−[0.016]
ln DMA−−−−−−
−−−−−−
ln SA0.001−−0.002−−
[0.806]−−[0.729]−−
ln FSA−−0.003−−0.007−
−[0.851]−−[0.536]−
ln FSA - ROW−−0.091−−0.042
−−[0.247]−−[0.451]
ln FSA - SSA−−−0.023**−−0.034***
−−[0.044]−−[0.006]
Controlssee Table 2
Nr. obs268268268315315315
Time-period1993–20001993–20001993–20002001–20092001–20092001–2009

Notes: P-values in brackets. Bootstrapped p-values on the basis of 200 replications. ***, **, * denotes significance at 1, 5, or 10 percent, respectively. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 5.
Open in new tab

Policy Experiments – Increase in Market Access and GDP Per worker

123456
CountryNo Longer LandlockedNo Longer IslandInfrastructure +1 s.d.All distances HalvedRFTA with South AfricaSSA - Wide Free Trade Zone
Cape Verde
FMA−71.8920.6763.720.342.68
FMA - SSA−51.1076.57112.092.9421.11
FMA - ROW−74.3110.3955.26−−
GDP per worker−2.764.146.050.161.14
Botswana
FMA85.89−34.6175.43−0.41
FMA - SSA90.29−76.57112.09−1.41
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.08
Central African Republic
FMA84.60−18.5561.960.060.35
FMA - SSA90.29−76.57112.090.693.83
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.050.040.21
Ethiopia
FMA85.01−23.8166.34−0.27
FMA - SSA90.29−76.57112.09−1.76
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.10
Sudan
FMA−−15.8459.73−0.18
FMA - SSA−−76.57112.09−2.99
FMA – ROW−−10.3955.26−−
GDP per worker−−4.146.050.16
123456
CountryNo Longer LandlockedNo Longer IslandInfrastructure +1 s.d.All distances HalvedRFTA with South AfricaSSA - Wide Free Trade Zone
Cape Verde
FMA−71.8920.6763.720.342.68
FMA - SSA−51.1076.57112.092.9421.11
FMA - ROW−74.3110.3955.26−−
GDP per worker−2.764.146.050.161.14
Botswana
FMA85.89−34.6175.43−0.41
FMA - SSA90.29−76.57112.09−1.41
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.08
Central African Republic
FMA84.60−18.5561.960.060.35
FMA - SSA90.29−76.57112.090.693.83
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.050.040.21
Ethiopia
FMA85.01−23.8166.34−0.27
FMA - SSA90.29−76.57112.09−1.76
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.10
Sudan
FMA−−15.8459.73−0.18
FMA - SSA−−76.57112.09−2.99
FMA – ROW−−10.3955.26−−
GDP per worker−−4.146.050.16

Notes: All numbers are percentages. The effect on GDP per worker is calculated by multiplying the change in SSA MA by the coefficient on SSA MA reported in column 8 in Table 3. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table 5.
Open in new tab

Policy Experiments – Increase in Market Access and GDP Per worker

123456
CountryNo Longer LandlockedNo Longer IslandInfrastructure +1 s.d.All distances HalvedRFTA with South AfricaSSA - Wide Free Trade Zone
Cape Verde
FMA−71.8920.6763.720.342.68
FMA - SSA−51.1076.57112.092.9421.11
FMA - ROW−74.3110.3955.26−−
GDP per worker−2.764.146.050.161.14
Botswana
FMA85.89−34.6175.43−0.41
FMA - SSA90.29−76.57112.09−1.41
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.08
Central African Republic
FMA84.60−18.5561.960.060.35
FMA - SSA90.29−76.57112.090.693.83
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.050.040.21
Ethiopia
FMA85.01−23.8166.34−0.27
FMA - SSA90.29−76.57112.09−1.76
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.10
Sudan
FMA−−15.8459.73−0.18
FMA - SSA−−76.57112.09−2.99
FMA – ROW−−10.3955.26−−
GDP per worker−−4.146.050.16
123456
CountryNo Longer LandlockedNo Longer IslandInfrastructure +1 s.d.All distances HalvedRFTA with South AfricaSSA - Wide Free Trade Zone
Cape Verde
FMA−71.8920.6763.720.342.68
FMA - SSA−51.1076.57112.092.9421.11
FMA - ROW−74.3110.3955.26−−
GDP per worker−2.764.146.050.161.14
Botswana
FMA85.89−34.6175.43−0.41
FMA - SSA90.29−76.57112.09−1.41
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.08
Central African Republic
FMA84.60−18.5561.960.060.35
FMA - SSA90.29−76.57112.090.693.83
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.050.040.21
Ethiopia
FMA85.01−23.8166.34−0.27
FMA - SSA90.29−76.57112.09−1.76
FMA – ROW84.02−10.3955.26−−
GDP per worker4.88−4.146.05−0.10
Sudan
FMA−−15.8459.73−0.18
FMA - SSA−−76.57112.09−2.99
FMA – ROW−−10.3955.26−−
GDP per worker−−4.146.050.16

Notes: All numbers are percentages. The effect on GDP per worker is calculated by multiplying the change in SSA MA by the coefficient on SSA MA reported in column 8 in Table 3. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table B1.
Open in new tab

Trade Equation Without Importer and Exporter Dummies. Year = 2008

Dep: ln Manuf ExportsCoefficients
Coefficients
VariableROWExtra SSAVariableROWExtra SSA
ln dist−0.797***−0.82***ln GDP exp1.405***0.087
[0.00][0.003] [0.00][0.472]
contiguity−1.944***4.182***ln GDP imp0.782***−0.326***
[0.055][0.00] [0.00][0.001]
com. lang.0.528***−0.378civil war exp0.255−
[0.00][0.272] [0.207]−
com. col.1.20***−0.258civil war imp−0.03−
[0.00][0.49] [0.888]−
colonizer1.611***−civil conflict exp0.234−
[0.00]− [0.205]−
RFTA1.151***−0.165civil conflict imp0.113−
[0.00][0.717] [0.498]−
infra exp0.416***2.647***   
[0.00][0.00]p-value Mills' ratio[0.066]
infra imp0.192*0.733p-value religion[0.000]
[0.098][0.233] 
ll exp−0.84***−0.063nr. obs14946
[0.00][0.86]censored11184
ll imp−1.252***0.189uncensored3762
[0.00][0.56] 
isl exp−0.797***0.284 
[0.00][0.616] 
isl imp−0.342**−0.41 
 [0.044][0.443]   
Dep: ln Manuf ExportsCoefficients
Coefficients
VariableROWExtra SSAVariableROWExtra SSA
ln dist−0.797***−0.82***ln GDP exp1.405***0.087
[0.00][0.003] [0.00][0.472]
contiguity−1.944***4.182***ln GDP imp0.782***−0.326***
[0.055][0.00] [0.00][0.001]
com. lang.0.528***−0.378civil war exp0.255−
[0.00][0.272] [0.207]−
com. col.1.20***−0.258civil war imp−0.03−
[0.00][0.49] [0.888]−
colonizer1.611***−civil conflict exp0.234−
[0.00]− [0.205]−
RFTA1.151***−0.165civil conflict imp0.113−
[0.00][0.717] [0.498]−
infra exp0.416***2.647***   
[0.00][0.00]p-value Mills' ratio[0.066]
infra imp0.192*0.733p-value religion[0.000]
[0.098][0.233] 
ll exp−0.84***−0.063nr. obs14946
[0.00][0.86]censored11184
ll imp−1.252***0.189uncensored3762
[0.00][0.56] 
isl exp−0.797***0.284 
[0.00][0.616] 
isl imp−0.342**−0.41 
 [0.044][0.443]   

Notes: We report estimated coefficients and not marginal effects. Marginal effects, and the results for the 1st stage probit are available upon request. The coefficients are used as input in the construction of our market access measures. *, **, *** denotes significant at the 5 percent level in at least 50, 80, or 100 percent of the years. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table B1.
Open in new tab

Trade Equation Without Importer and Exporter Dummies. Year = 2008

Dep: ln Manuf ExportsCoefficients
Coefficients
VariableROWExtra SSAVariableROWExtra SSA
ln dist−0.797***−0.82***ln GDP exp1.405***0.087
[0.00][0.003] [0.00][0.472]
contiguity−1.944***4.182***ln GDP imp0.782***−0.326***
[0.055][0.00] [0.00][0.001]
com. lang.0.528***−0.378civil war exp0.255−
[0.00][0.272] [0.207]−
com. col.1.20***−0.258civil war imp−0.03−
[0.00][0.49] [0.888]−
colonizer1.611***−civil conflict exp0.234−
[0.00]− [0.205]−
RFTA1.151***−0.165civil conflict imp0.113−
[0.00][0.717] [0.498]−
infra exp0.416***2.647***   
[0.00][0.00]p-value Mills' ratio[0.066]
infra imp0.192*0.733p-value religion[0.000]
[0.098][0.233] 
ll exp−0.84***−0.063nr. obs14946
[0.00][0.86]censored11184
ll imp−1.252***0.189uncensored3762
[0.00][0.56] 
isl exp−0.797***0.284 
[0.00][0.616] 
isl imp−0.342**−0.41 
 [0.044][0.443]   
Dep: ln Manuf ExportsCoefficients
Coefficients
VariableROWExtra SSAVariableROWExtra SSA
ln dist−0.797***−0.82***ln GDP exp1.405***0.087
[0.00][0.003] [0.00][0.472]
contiguity−1.944***4.182***ln GDP imp0.782***−0.326***
[0.055][0.00] [0.00][0.001]
com. lang.0.528***−0.378civil war exp0.255−
[0.00][0.272] [0.207]−
com. col.1.20***−0.258civil war imp−0.03−
[0.00][0.49] [0.888]−
colonizer1.611***−civil conflict exp0.234−
[0.00]− [0.205]−
RFTA1.151***−0.165civil conflict imp0.113−
[0.00][0.717] [0.498]−
infra exp0.416***2.647***   
[0.00][0.00]p-value Mills' ratio[0.066]
infra imp0.192*0.733p-value religion[0.000]
[0.098][0.233] 
ll exp−0.84***−0.063nr. obs14946
[0.00][0.86]censored11184
ll imp−1.252***0.189uncensored3762
[0.00][0.56] 
isl exp−0.797***0.284 
[0.00][0.616] 
isl imp−0.342**−0.41 
 [0.044][0.443]   

Notes: We report estimated coefficients and not marginal effects. Marginal effects, and the results for the 1st stage probit are available upon request. The coefficients are used as input in the construction of our market access measures. *, **, *** denotes significant at the 5 percent level in at least 50, 80, or 100 percent of the years. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table B2.
Open in new tab

Average Share of Manufacturing in Merchandise Exports

Avg. % Manufacturing in Total Merchandise Exports
Country1993–20002001–20091993–2009
Sub-Saharan Africa29.432.431.1
Djibouti−90.790.7
Lesotho94.988.890.0
Botswana89.682.983.6
Cape Verde81.368.273.8
Mauritius72.266.068.9
Swaziland54.465.263.9
South Africa49.654.752.3
Namibia55.847.148.0
Central African Republic49.037.444.5
Senegal43.141.742.3
Madagascar28.446.538.0
Eritrea28.040.536.3
Zimbabwe31.433.532.5
Togo12.853.131.4
Kenya25.430.628.1
Gambia, The19.021.920.7
Guinea22.818.420.4
Ghana14.722.318.9
Tanzania16.619.518.6
Cote d'Ivoire14.117.216.0
Comoros23.66.314.3
Burkina Faso12.711.512.1
Uganda6.914.511.0
Malawi8.411.910.4
Zambia10.710.110.4
Mali4.113.69.9
Mozambique11.66.98.8
Ethiopia7.59.48.6
Sierra Leone9.77.58.6
Benin5.88.77.0
Niger2.89.36.6
Burundi2.210.26.4
Rwanda5.85.95.9
Cameroon6.84.05.1
Gabon2.77.85.0
Sao Tome and Principe2.04.33.9
Congo, Rep.2.4−2.4
Seychelles1.13.72.4
Nigeria1.62.92.3
Sudan3.70.62.1
Guinea-Bissau0.21.21.0
Mauritania0.20.00.1
Avg. % Manufacturing in Total Merchandise Exports
Country1993–20002001–20091993–2009
Sub-Saharan Africa29.432.431.1
Djibouti−90.790.7
Lesotho94.988.890.0
Botswana89.682.983.6
Cape Verde81.368.273.8
Mauritius72.266.068.9
Swaziland54.465.263.9
South Africa49.654.752.3
Namibia55.847.148.0
Central African Republic49.037.444.5
Senegal43.141.742.3
Madagascar28.446.538.0
Eritrea28.040.536.3
Zimbabwe31.433.532.5
Togo12.853.131.4
Kenya25.430.628.1
Gambia, The19.021.920.7
Guinea22.818.420.4
Ghana14.722.318.9
Tanzania16.619.518.6
Cote d'Ivoire14.117.216.0
Comoros23.66.314.3
Burkina Faso12.711.512.1
Uganda6.914.511.0
Malawi8.411.910.4
Zambia10.710.110.4
Mali4.113.69.9
Mozambique11.66.98.8
Ethiopia7.59.48.6
Sierra Leone9.77.58.6
Benin5.88.77.0
Niger2.89.36.6
Burundi2.210.26.4
Rwanda5.85.95.9
Cameroon6.84.05.1
Gabon2.77.85.0
Sao Tome and Principe2.04.33.9
Congo, Rep.2.4−2.4
Seychelles1.13.72.4
Nigeria1.62.92.3
Sudan3.70.62.1
Guinea-Bissau0.21.21.0
Mauritania0.20.00.1

Notes: All numbers denote percentages. Angola, Chad, the Democratic Republic of Congo, Equatorial Guinea, Liberia and Somalia are not shown in the Table due to lack of information.

Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Table B2.
Open in new tab

Average Share of Manufacturing in Merchandise Exports

Avg. % Manufacturing in Total Merchandise Exports
Country1993–20002001–20091993–2009
Sub-Saharan Africa29.432.431.1
Djibouti−90.790.7
Lesotho94.988.890.0
Botswana89.682.983.6
Cape Verde81.368.273.8
Mauritius72.266.068.9
Swaziland54.465.263.9
South Africa49.654.752.3
Namibia55.847.148.0
Central African Republic49.037.444.5
Senegal43.141.742.3
Madagascar28.446.538.0
Eritrea28.040.536.3
Zimbabwe31.433.532.5
Togo12.853.131.4
Kenya25.430.628.1
Gambia, The19.021.920.7
Guinea22.818.420.4
Ghana14.722.318.9
Tanzania16.619.518.6
Cote d'Ivoire14.117.216.0
Comoros23.66.314.3
Burkina Faso12.711.512.1
Uganda6.914.511.0
Malawi8.411.910.4
Zambia10.710.110.4
Mali4.113.69.9
Mozambique11.66.98.8
Ethiopia7.59.48.6
Sierra Leone9.77.58.6
Benin5.88.77.0
Niger2.89.36.6
Burundi2.210.26.4
Rwanda5.85.95.9
Cameroon6.84.05.1
Gabon2.77.85.0
Sao Tome and Principe2.04.33.9
Congo, Rep.2.4−2.4
Seychelles1.13.72.4
Nigeria1.62.92.3
Sudan3.70.62.1
Guinea-Bissau0.21.21.0
Mauritania0.20.00.1
Avg. % Manufacturing in Total Merchandise Exports
Country1993–20002001–20091993–2009
Sub-Saharan Africa29.432.431.1
Djibouti−90.790.7
Lesotho94.988.890.0
Botswana89.682.983.6
Cape Verde81.368.273.8
Mauritius72.266.068.9
Swaziland54.465.263.9
South Africa49.654.752.3
Namibia55.847.148.0
Central African Republic49.037.444.5
Senegal43.141.742.3
Madagascar28.446.538.0
Eritrea28.040.536.3
Zimbabwe31.433.532.5
Togo12.853.131.4
Kenya25.430.628.1
Gambia, The19.021.920.7
Guinea22.818.420.4
Ghana14.722.318.9
Tanzania16.619.518.6
Cote d'Ivoire14.117.216.0
Comoros23.66.314.3
Burkina Faso12.711.512.1
Uganda6.914.511.0
Malawi8.411.910.4
Zambia10.710.110.4
Mali4.113.69.9
Mozambique11.66.98.8
Ethiopia7.59.48.6
Sierra Leone9.77.58.6
Benin5.88.77.0
Niger2.89.36.6
Burundi2.210.26.4
Rwanda5.85.95.9
Cameroon6.84.05.1
Gabon2.77.85.0
Sao Tome and Principe2.04.33.9
Congo, Rep.2.4−2.4
Seychelles1.13.72.4
Nigeria1.62.92.3
Sudan3.70.62.1
Guinea-Bissau0.21.21.0
Mauritania0.20.00.1

Notes: All numbers denote percentages. Angola, Chad, the Democratic Republic of Congo, Equatorial Guinea, Liberia and Somalia are not shown in the Table due to lack of information.

Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

First of all, the final rows of Table 1 show that also in case of our trade sample restricted to SSA bilateral exports, the usefulness of the Helpman and others (2008) approach can not be rejected:17 religious similarity has a significantly positive effect on the probability to trade18 and, moreover, the inverse Mills' ratio is significant in the second stage (hereby not rejecting the need to take account of endogenous sample selection).

Turning to the results on the importance of our included trade costs variables,19 we confirm the standard result that distance negatively affects the amount of trade between countries. We do not find convincing evidence that the penalty on distance is significantly higher for intra-SSA trade (see also Foroutan and Pritchett, 1993).20 We only find a significantly higher distance penalty for intra-SSA trade in 50 percent of our sample years. Interestingly, we find this significantly higher distance penalty during the later years in our sample in particular, suggesting that the improvements in SSA trade costs that have been made in recent years have been biased towards improving trade costs with the ROW instead of better connecting the sub-continent.

Second, we do not find evidence of a border effect in SSA trade. For SSA trade with the ROW this may not be that surprising. The only SSA countries that border non-SSA countries are those bordering North African countries, and SSA countries trade less with these countries than with other non-African countries (see e.g. IMF, 2007). The lack of a “border-effect” is arguably more surprising for intra-SSA trade given that studies looking at other parts of the world (e.g. Europe or the United States) usually find strong evidence that neighbors trade disproportionately more with each other.

By contrast, we do find strong effects of language and colonial history on trade volumes of SSA countries. Sharing a colonial history has a strong positive effect on the amount of trade. Especially SSA trade with its former colonizer(s) is much higher than trade with other countries in the world. Having a common colonizer also boosts bilateral trade, and this effect is not significantly different for intra-SSA trade compared to trade with other former colonies in the ROW. Sharing a common language also stimulates both intra-SSA and SSA-ROW trade in largely the same way (see also e.g. Foroutan and Pritchett, 1993; or Coe and Hoffmaister, 1999).

Finally, we do not find very convincing evidence that SSA trade in manufactures benefits significantly from the many RFTAs that are in existence on the sub-continent. We only find a significant positive effect of having an RFTA on export volumes in 50 percent of the years in our sample period. It is interesting to note however, that we find these significant positive effects of RFTAs on export volumes for the latest years in the sample (2006–2009) in particular. A tentative indication that African RFTAs, many of whom often only exist(ed) on paper, could be becoming more effective in coordinating policies favorable to trade (see also UNCTAD, 2009).

Constructing Market Access, Distinguishing Between Access to SSA and to the ROW

Using the yearly-estimated coefficients of (7) and the relationship between the trade equation in (2) and market access in (1), the next step is to construct market access using (4). Besides calculating overall market access, we also look at three different subcomponents of market access. Following among others Redding and Venables (2004), Breinlich (2006) or Head and Mayer (2010), we distinguish explicitly between the respective contribution of domestic market access (DMA) and foreign market access (FMA). Furthermore, in order to be able to distinguish between the relevance of access to other SSA markets and to markets in the rest of the world (ROW) respectively, we in turn split foreign market access (FMA) into access to other SSA markets and access to ROW markets:
(8)
where formula⁠, formula and formula⁠. We construct the different components of a country's total market access for each year separately according to (4), adapted to take into account the estimated parameters of (7):
(9)
where coefficients with superscript “SSA” denote the estimated effect of a variable on intra-SSA trade (i.e. the coefficient on a variable plus the coefficient on that variable interacted with the intra-SSA dummy). We construct domestic market access (DMA) in Redding and Venables' (2004) preferred way. That is, we use a country's internal distance (see Appendix A for its exact definition) as input in the trade cost function, assuming that the speed at which internal trade decays with distance is half as strong as for trade with other countries (i.e. we divide formula by two). Moreover, we take each country as sharing a common border, a common language, a common colonial history, and having an RFTA with itself.

Having constructed the three different components of a country's total market access using (9), it is straightforward to obtain both total market access (MA) and access to foreign markets (FMA) according to equation (8).

III. Evidence on the Role of Economic Geography in SSA Economic Development

Having constructed yearly market access measures for all 48 SSA countries in our sample, we are finally in a position to assess the effect of market access on economic development.

Market Access and Economic Development

We start by visualizing the relationship between market access and income levels. Figure 1a plots market access (MA) against GDP per worker (our preferred proxy of wages, see below for more on this choice) over our entire sample period. It shows an overall positive relationship between GDP per worker over our sample period.

Market Access and GDP Per Worker in SSA
Figure 1a.

Market Access and GDP Per Worker in SSA

Open in new tabDownload slide
Notes: The raw correlation between ln MA and ln GDP per worker is 0.22 [p-value: 0.00]. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Furthermore, when looking for possible differences over our 17 year sample period, Figure 1b suggests that the relationship between income levels and market access has strengthened in the most recent years of our sample.

Market Access and GDP Per Worker in SSA, Changes Over Sample Period
Figure 1b.

Market Access and GDP Per Worker in SSA, Changes Over Sample Period

Open in new tabDownload slide
Notes: The raw correlation between ln MA and ln GDP per worker is 0.05 [p-value: 0.35] for the 1993–2000 period and 0.26 [p-value: 0.00] for the 2001–2009 period. Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

To look at this in a more rigorous way, we turn to the estimation of the NEG wage equation, (5). In the absence of reliable wage data for all SSA countries in all years of our sample, we need to proxy wages. Since many SSA face high unemployment rates, we decided not to use GDP per capita (as e.g. Redding and Venables, 2004; Breinlich, 2006; or Head and Mayer, 2010 do), but take GDP per worker as a more appropriate measure.21 The error term ηit in (5) captures cit, a country's level of production efficiency. Again following Redding and Venables (2004), we start by assuming that these cross-country differences in technology are captured by an idiosyncratic error term and estimate (5) using pooled OLS (implicitly only allowing for other variables determining technological efficiency that are uncorrelated with our market access measure). The result is shown in the first column of Table 2 below.22 We find that the estimated market access coefficient is positive and significant. A 1 percent increase in a country's market access increases GDP per worker by 0.07 percent.

This conclusion is, however, somewhat premature. It is only valid under the earlier-mentioned assumption of idiosyncratic differences in countries' production efficiency, cit, that are uncorrelated with market access. As this assumption is likely to be violated, we subsequently make use of the panel data nature of our data set. We include country fixed effects to capture all time-invariant country-specific variables that affect a country's production efficiency. Most notably, we hereby control for differences in physical geography that is often blamed for Africa's poor development (climate, primary resource endowments, soil quality, etc). By including time (year) fixed effects as well, we also take into account any shocks that are affecting all countries similarly. Examples are the introduction of new technological innovations made in developed countries (a prime example here are mobile phones, which have rapidly spread all over SSA) or worldwide economic shocks such as changes in the world price of agricultural products or natural resources. The second column of Table 2 shows that the inclusion of fixed effects is quite important (corroborating findings by Head and Mayer, 2010): the effect of total market access on GDP per worker is still positive and significant, but it is about half that found in column 1: a 1 percent increase in a country's total market access increases GDP per worker by 0.03 percent.

However, the inclusion of country- and year-fixed effects may still not provide us with accurate estimates of the effect of market access. They only control for time-invariant country-specific or country-invariant time-specific variables. It is not unlikely that a country's production efficiency is also determined by time- ànd country-varying variables that are correlated with market access. If this is the case, we would still obtain biased estimates of the coefficient on market access, even when allowing for country- and year fixed effects. We therefore include several additional control variables to our regressions (see also Breinlich, 2006; Redding and Venables, 2004; Hering and Poncet, 2010; Fally and others, 2010). They are related to a country's quality of institutions (polity IV), its human capital (gross primary enrolment), its scope for economics of density (working population per km2 urbanization rate), and whether or not it is in a state of civil war or conflict.

Moreover, given that our market access measures all focus on the importance of manufactures, whereas SSA countries vary widely in the importance of this sector in their overall exports (see Figure B2 and Table B2 in Appendix B), we control for a country's economy's dependence on natural resources by controlling for the importance of oil, and of agriculture respectively in its overall economy (note that, given its time-invariant nature, the presence of oil, or any other natural resource, is already controlled for by the included country fixed effects). Column 3 of Table 2 shows the corresponding estimation results.

Adding these additional controls further lowers the effect of market access23 but it remains significantly related to GDP per worker: a 1 percent increase in a country's market access raises its income level by 0.02 percent. As to the control variables, we find that three of them are significant.24 SSA countries that are more dependent on agriculture, and those plagued by civil war tend to have lower levels of GDP per worker. Also, we find that the more oil-dependent SSA countries have higher levels of income per worker.

Finally, column 4 and 5 show the results of estimating the relationship between market access and GDP per worker for the first and second half of the sample respectively (always including fixed effects and the eight above-mentioned control variables). This confirms the preliminary evidence shown in Figure 1b: the positive relationship between market access and economic development is strongest in the second half of our sample period (2001–2009). Although positive, we do not find a significant effect of market access on income levels for the early years in our sample.25

Columns 3–5 constitute our baseline results. They show that market access is a significantly positive determinant of a SSA country's economic development. Moreover, the importance of market access has increased over the last two decades. Based on the results shown in column 5, a 1 percent increase in total market access increases GDP per worker by 0.031 percent. When comparing this result to similar studies using samples encompassing both developed and developing countries (e.g. Redding and Venables, 2004; or Head and Mayer, 2010), but also to other studies looking at developing economies like Brazil (Fally and others, 2010), China (Hering and Poncet, 2010) or Indonesia (Amiti and Cameron, 2007), we find a substantially lower effect of market access on economic development.26 Given the fact that the manufacturing sector is still relatively undeveloped in SSA compared to many countries considered in these other studies, this finding may not be that surprising. Nevertheless, our results show that economic geography matters, also in SSA. Moreover, its importance has increased in recent years.

Decomposing the Importance of Access to SSA Markets and to Markets in the ROW

Having established the importance of market access for SSA economic development, we further decompose this finding in this section. In particular, we look at the relative importance of access to other SSA markets and markets in the ROW respectively. To do so, we use the decomposition of a country's market access in domestic market access (DMA) and foreign market access (FMA) [see (9)], further decomposing the latter into access to other SSA markets (MAitSSA) and access to markets in the ROW (MAitROW).

SSA Market Access and ROW Market Access Compared

In Figures 2a and 2b, we start by showing some stylized facts about the two most interesting components of market access, ROW and SSA market access. The left-hand panel of Figure 2a shows that having good access to SSA markets is virtually uncorrelated with good access to markets in the ROW. The right-hand panel adds to this by plotting the mean relative importance of SSA in a country's FMA against mean GDP per worker over our sample period. This shows that ROW market access dominates FMA for SSA's island nations, countries located in SSA's north east (e.g. Sudan Djibouti, or Eritrea), and South Africa (SSA's economic powerhouse27). By contrast, for countries close to South Africa or Nigeria (e.g. Botswana, Swaziland; Togo or Benin), SSA market access dominates their overall FMA. But, despite the substantial difference between countries in the degree to which the ROW dominates their FMA, Figure 2a shows no clear relationship between the relative importance of SSA or the ROW in FMA, and income levels.

Decomposing Market Access – variation in access to SSA and access to ROW
Figure 2a.

Decomposing Market Access – variation in access to SSA and access to ROW

Open in new tabDownload slide
Notes: The overall mean share of SSA in FMA is 56% (s.d. 0.31). Source: Authors' analysis based on data sources discussed in the main text or Appendix A.
Market Access and Distance to Major Markets
Figure 2b.

Market Access and Distance to Major Markets

Open in new tabDownload slide
Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Figure 2b shows part of the reason why SSA and ROW market access are little correlated. Given the strong distance decay effects we found when estimating the trade equation (see Table 1), countries located closest to Europe (Africa's largest export market) have the best ROW market access.28 On the contrary, countries closest to South Africa, but also those in West Africa (see Figure B1 in Appendix B), tend to have the best SSA market access.29

SSA market access, ROW market access and SSA economic development

Next, Figure 3 depicts the relationship between the various (sub)components of total market access and economic development. We separately plot DMA, ROW market access and SSA market access against GDP per worker. Given our finding of an increased importance of market access over the years (see Table 2), Figure 3 distinguishes between the first and second half of our sample (1993–2000 and 2001–2009). These simple scatterplots show that the relationship of all three different (sub)components of a country's market access with GDP per worker appears strongest during the 2nd half of our sample period.30

Market Access' Different Components and Economic Development
Figure 3.

Market Access' Different Components and Economic Development

Open in new tabDownload slide
Notes: All figures plot demeaned ln GDP per worker against a demeaned subcomponent of market access. Demeaned meaning that we substract a country's mean ln GDP per worker or mean subcomponent of market access from each observation. Doing this already removes the influence of any unobserved time-invariant from the data, making our plots more directly related to our findings in Table 2 that always include a full set of country fixed effects. It is also done to show that there exists substantial within country-variation in both each respective subcomponent of market access as well as in ln GDP per worker. Source: Authors' analysis based on data sources discussed in the text.

To more formally assess the importance of market access's different components, we re-estimate equation (5) replacing total market access by DMA, access to SSA markets (denoted as FMA-SSA in Table 3) and ROW market access (denoted as FMA-ROW in Table 3). The estimation results are shown in Table 3 below, always distinguishing between the first and second half of our sample period. All regressions include the same control variables as in columns 3–5 of Table 2 and a full set of country- and year fixed effects.31

First of all, the results confirm our earlier finding that market access has become increasingly important for SSA countries. In the first half of our sample period we only find some evidence that, of the various market access categories, domestic market access is weakly significant (at 10 percent). FMA, also when further split between SSA and ROW market access, is not significantly related to income levels in these years (see column 1 and 2). This changes markedly during the latest years in our sample (2001–2009). In particular, we find that access to foreign markets has become much more important in explaining the differences in economic development between SSA countries (see column 3). Even more interesting, when further decomposing this effect of FMA in column 4, we find that it is access to SSA markets in particular that drives these findings: A 1% increase in a country's access to other SSA markets is associated with a 0.05% increase in GDP per worker. The effect of ROW market access instead is insignificant.

Some care has to be taken in interpreting these results. Our findings do go against those proclaiming that intra-SSA economic linkages are too weak and under-developed to be of importance to SSA countries. They support the view that stimulating intra-SSA manufacturing export markets is very important for the future viability of SSA countries that want to become less dependent on natural resource revenue (Collier and Venables, 2007). Indeed, manufacturing already dominates intra-SSA exports (UNCTAD, 2009). Moreover, intra SSA-exports have increased faster than SSA-exports to the rest of the world in recent years (Easterly and Reshef, 2010; ARIA IV). However, given that we focus on market access for manufacturing goods (see also our discussion at the beginning of section II), our results should not be taken as saying that access to markets in the ROW does not matter for SSA. SSA exports to the ROW are dominated by natural resources (for more than 75 percent). The fact that market access to the ROW for SSA manufacturing products is not significant, does not say much about the importance of, for example, lowering tariffs on SSA agricultural products for SSA countries' economic development. It can also be taken as an indication that, to date, most SSA manufactures are not yet finding their way to markets outside the (sub)continent.32

Finally, also note that DMA loses its (weak) significance in the second half of our sample period. It confirms the idea that for most SSA countries their own domestic market size is too small to be of significant importance.33; possibly even posing constraints on firms' prospects (see Collier and Venables, 2007). However, endogeneity problems are inherently present when including DMA. One basically regresses a measure dominated by a country's own GDP (DMA) on its GDP per worker (see Head and Mayer (2010) for a (critical) discussion on this issue). Therefore, columns 5–8 in Table 3 show that our results on FMA, and its two subcomponents (FMA-SSA and FMA-ROW), also come through when totally abstracting from DMA. To summarize our main findings: access to foreign markets is increasingly important for SSA countries. Moreover, it is access to other SSA countries in particular that is positively associated with income levels.

Additional Robustness Checks

Several issues could still invalidate our main findings. First, even when abstracting from domestic market access (DMA), there is still the issue of endogeneity. The assumption under which our baseline results are valid is that, after controlling for fixed effects and the included control variables, the remaining error term is uncorrelated with our measures of foreign market access. One way in which this may be violated is when the error term still contains other variables influencing a country's GDP per worker that are correlated with market access. Another way is reverse causality: if market access not only influences GDP per worker, but GDP per worker in turn also influences market access, the error term is by construction correlated with market access.

To control for both possible sources of endogeneity, we employ an instrumental variable approach,34 using the distance to SSA's most important export markets in SSA and in the ROW as instruments for our measures of foreign market access (i.e. the EU, South Africa and Nigeria; see Figures 2b and B1). The relevance of this approach relies on the arguments put forward to justify the usefulness of these ‘distance instruments’ (see among others Redding and Venables, 2004 and Hanson, 2006 for more on this). Table 4a shows the results.

First of all, Table 4a shows that we cannot reject the validity of our instruments (see the statistics at the bottom of the Table). The F-statistic for their joint significance in the first stage is larger than 10 (Staiger and Stock, 1997), except in column 3 where it is 7.44. Moreover, they always pass the Hansen J test for overidentification. The results confirm our baseline findings. Foreign market access significantly positively affects income per worker in the latest years of our sample. When further subdividing this into ROW- and SSA market access, we again find that this positive effect holds for access to SSA markets only.

A drawback of our IV results is that the distance instruments used are time-invariant. This precludes the use of country-fixed effects. Columns 1–4 of Table 4b therefore also show results when including each market access measure lagged one period (all control variables are also lagged one period). This to some extent controls for reverse causality, while still allowing for the inclusion of country-fixed effects.35 Reassuringly, all our baseline results again come through.

Finally, our last robustness check extends our empirical model.36 Footnote 5 already hinted at the possibility of extending the NEG model that we use (see Appendix C) to also include an intermediate goods sector. This would (see Redding and Venables 2004 for the details) add an additional term to equation (5). Besides a country's market access (i.e. its ease of access of final goods markets), a country's supplier access (i.e. its ease of access to markets for intermediate goods needed in final goods production) would be an important determinant of its income level. A yearly-measure of each SSA country's supplier access can be constructed in a very similar way as its market access (basically replacing formula with formula in (4), but again see Redding and Venables 2004 for more details). Table 4c shows the results when also taking account of supplier access (SA). Similar to market access, it is very straightforward to decompose overall supplier access into domestic supplier access (DSA), and access to suppliers in SSA and in the ROW respectively.

For supplier access too, we find that it increased in importance over the last two decades, and that access to SSA suppliers in particular is positively associated with higher GDP per worker. However, when focusing on foreign supplier access, or supplier access as a whole, results are much weaker than those for market access. Most importantly for our purposes, all our baseline findings regarding the importance of market access hold up to also considering countries' supplier access.

IV. The Effect of Different Policies Aimed at Improving Market Access

Our main findings show that improving a country's market access, and in particular access to other SSA countries, will have significant positive effects on its economic development. In this section we use this positive relationship between market access and income levels to gain insight into the relative effect of different policies aimed at improving a country's market access.37 For example, we look at the effect of improving a country's infrastructure, increasing SSA regional integration, or alleviating the burden of landlocked countries. To be able to do so, we change our baseline estimation strategy in one important way (see also Elbadawi and others, 2004 or Redding and Venables 2004).

Our baseline strategy includes importer-year and exporter-year dummies when estimating the trade equation (3). This does, however, not allow one “to quantify the effects … of particular country characteristics (for example, landlocked or infrastructure), since all such effects are contained in the dummies” (Redding and Venables 2004, p. 75). As such, it becomes impossible to look at the effect of country-specific policies aimed at lowering trade costs. To overcome this problem we follow Redding and Venables (2004, section 7), and estimate the following trade equation instead of equation (3), proxying each country's market and supplier access by its GDP (⁠formula⁠):
(10)
In order to conduct our various ‘policy experiments’, we augment our specification of trade costs (6) by three different exporter- and importer-specific determinants of trade costs: being landlocked, being an island, and the state of a country's infrastructure, that is,
(11)

Furthermore we take note of Martin and others (2008) and control for whether or not a country is experiencing civil war or civil conflict (see Appendix A for the full details on all these additional variables included to the trade equation).

We estimate (10) with (11) substituted for trade costs Tijt for the latest possible year in our sample (which is 2008 instead of 2009 because of missing infrastructure and conflict data for 2009). We take the latest possible year in order to make our prediction of the effect of each of our ‘policy experiments’ as up-to-date as possible. Table B1 in Appendix B shows the resulting estimated coefficients. Again, we allow for a different effect of each trade-cost related variable on intra-SSA trade. Confining our discussion to the newly added country-specific trade cost variables,38 we find a large burden of being landlocked, and to a lesser extent of being an island (landlocked countries export and import significantly less (84 percent and 125 percent respectively) than coastal nations; and these numbers are 80 percent and 34 percent for island countries). Moreover, we find that bad infrastructure is an important deterrent to trade, and exports in particular.

Based on the estimates shown in Table B1, we calculate each SSA country's market access and its components [in a similar way as in (9)]. Next, we recalculate these measures taking into account one of eight different policy experiments that we set out in more detail below, and calculate the resulting change in foreign market access (and its two subcomponents). We do not look at domestic market access as most of our policy experiments are not that interesting to look at for domestic market access (e.g. no longer being landlocked only affects a country's trade costs with other countries, and it is hard to think about a country establishing an RFTA with itself).

Finally, the effect of the resulting improvement in market access on GDP per worker easily follows from the estimated coefficient(s) on foreign market access in Table 3. In particular, we take our finding on the relative importance of access to other SSA markets seriously and use the coefficient on SSA market access, reported in column 8, in combination with the change in SSA market access to get at the overall impact of each policy experiment on economic development.39 Table 5 shows the results of doing this for the first six of our eight different “policy experiments,” focusing on five different countries.

Given the way we modelled trade costs, see (11), the first four policy experiments affect all countries' SSA market access and ROW market access similarly. This is not true for overall FMA. Since we allowed all trade costs variables to have a different effect on intra-SSA trade and SSA trade with the ROW respectively, changes in FMA depend on the relative importance of SSA market access and ROW market access in overall FMA (explaining why the overall change in FMA resembles that in ROW market access for Cape Verde and Sudan, and that in SSA market access for the other three countries).

Halving distances to all trade partners (a rough proxy for improving SSA countries' connectivity through, e.g., cross-border infrastructure projects, or more effective border procedures) results in the largest improvement in GDP per worker, raising it by about 6 percent. Next comes alleviating a landlocked country's burden of having no direct access to the coast (raising incomes by almost 5 percent), followed by a 4 percent increase in GDP per worker as a result of a one standard deviation improvement in a country's infrastructure (e.g. corresponding to upgrading Ethiopia's infrastructure to resemble that in Botswana). With a resulting increase of 2.8 percent, alleviating the remoteness of an island country has the smallest effect on GDP per worker.

Finally, columns 5 and 6 show the effects of a newly established RFTA. These are also positive but much smaller compared to the other policy experiments. Not surprisingly, the effect on GDP per worker is larger, the larger the number of new partner countries in the new RFTA.40 (compare columns 5 to 6, or the impact of the SSA-wide free trade zone on Cape Verde to that on Botswana [a SSA-free trade zone would more than triple the number of RFTA partners for Cape Verde whereas ‘only’ doubling it for Botswana]). This much smaller effect of the establishment of an SSA-wide free trade zone compared to those of our other ‘policy experiments’ should in our view be taken with a pinch of salt. Due to the plethora of RFTAs officially in existence in SSA in 2008, the average SSA country already shared official RFTA membership with 46 percent of its SSA trade partners. However, the effectiveness of SSA RFTAs in actually implementing policies favourable to intra-SSA trade varies widely (compare, e.g., SADC to CENSAD). Our simple dummy variable for the existence of an RFTA is unable to take the varying degrees of effectiveness of each RFTA into account, so that our findings in Table 5 are most likely an underestimate of the effect on economic development were SSA countries able to establish an SSA-wide trade zone operating at the same level effectiveness as e.g. ASEAN or MERCOSUR, let alone NAFTA or the EU (see also UNCTAD, 2009).

Our last two experiments do not so much concern policy. They are aimed at giving an idea of the magnitude of spatial spillovers across SSA countries. How large are the benefits of growth in one particular country for its neighbors as a result of the increased market access that these neighboring countries enjoy? Figure 4 gives some idea of this. It plots the increase in GDP per worker in all SSA countries resulting from a 10 percent increase in the GDP of one of SSA's economic powerhouses, South Africa and Nigeria, against distance to these countries.

Spillovers to Neighbors of Positive 10 Percent GDP Shock in Nigeria or South Africa
Figure 4.

Spillovers to Neighbors of Positive 10 Percent GDP Shock in Nigeria or South Africa

Open in new tabDownload slide
Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Given the magnitude of the estimated distance penalty in SSA (see Table 1), we find that the effects of such shocks quickly peters out with distance. Countries located closest to South Africa and Nigeria respectively experience the largest spillovers. The overall spillover effect is small compared to some of our earlier “trade cost experiments” (the nearest neighbors experiencing “spillover-growth” of about 0.2 percent).41 This is due to the fact that for each country, South Africa and Nigeria constitute only one of many trading partners, that determine a country's market access.

V. Conclusions

The role of geography in explaining sub-Saharan Africa's poor economic performance is often confined to its physical geography, focusing on, for example, its hostile disease environment or poor climate. This paper focuses on a different role of geography and establishes the importance of relative or economic geography for economic development in sub-Saharan Africa (SSA). Using an empirical strategy that is firmly based upon a new economic geography model, our paper is among the first to test for the importance of market access and thereby of economic geography in explaining the observed differences in economic development between SSA countries.

Building on the framework introduced by Redding and Venables (2004), we first construct theory-based measures of market access for manufactures for each SSA country, relying on bilateral manufacturing trade data to reveal the relative importance of trade costs and market size in determining each country's market access. In doing so, we explicitly allow for a different impact of trade costs on intra-SSA trade and SSA trade with the rest of the world (ROW), and subsequently decompose each country's total market access into market access to other SSA countries and into market access to the ROW respectively.

Using these constructed measures of market access, we estimate the impact of market access for manufactures on GDP per worker. We find that market access positively affects income levels. Economic geography matters for economic development, also in SSA. Moreover, it has increased in importance over the last two decades. The relationship between market access and economic development is strongest and most robust during the 2001–2009 period: a 1 percent increase in a country's market access is associated with a 0.03 percent increase in its GDP per worker. This finding is robust to controlling for other variables affecting economic development (most notably human capital, institutions and natural resource dependence), to controlling for unobserved heterogeneity by allowing for country (and year) specific fixed effects, and to instrumenting market access by distance to major markets.

Arguably even more interesting is our finding that, when decomposing our overall market access effect into the respective effects of domestic, SSA-, and ROW-market access, access to other SSA markets has the most significant (and also the most robust) impact on a country's economic development. This finding becomes less surprising when considering the fact that most SSA countries sell the bulk of their manufacturing exports to other SSA countries (UNCTAD 2009; 2010). Moreover, SSA export growth has been regional in recent years. Intra SSA-exports have increased faster than SSA-exports to the ROW (Easterly and Reshef 2010; UNCTAD 2010). By contrast, SSA exports to the ROW are to date still dominated by natural resources and agricultural products, with only little SSA manufactured goods finding their way to European, US, or Asian markets. Our findings stress the importance of the rest of SSA for most SSA countries' prospects on developing a more diversified economy with a profitable, exporting, manufacturing sector (one of the backbones of Asia's sustained growth over the last decades). Improving market access among SSA countries alleviates the constraint of small domestic market size faced by most SSA countries (Collier and Venables 2007), carrying positive effects for economic development.

Based on our findings, we also show tentative evidence on the impact of several policies specifically aimed at improving SSA countries' market access. Overall, this lends support to the view that current efforts to improve SSA market access by, for example, investing in infrastructure (Sub-Saharan African Transport Policy Program or The Infrastructure Consortium for Africa), alleviating the burden of landlocked countries (the Almaty Programme), or by aiming to increase effective intra-SSA integration (African Union), are indeed important, although in varying degrees, in further stimulating SSA economic development.

Above all, see also Henderson, Shalizi and Venables (2001), our results are a reminder that distance or relative geography matters for economic development. Despite room for (policy-induced) improvements in market access, the (economic) remoteness of many SSA countries remains an important burden on their economic development prospects.

Appendix A. Data Definitions and Sources

GDP (also per capita and per worker): Gross Domestic Product (also per capita and per worker) in current US dollars. From World Bank Development Indicators, 2011, or World Bank Africa Database, 2010.

Distance: Great circle distance between main cities, from CEPII.

Internal distance: This often-used specification of Dii reflects the average distance from the centre of a circular disk with areai to any point on the disk (assuming these points are uniformly distributed on the disk). It is calculated on the basis of a country's area: formula⁠.

Contiguity: Dummy variable indicating if two countries share a common border, from CEPII.

Common official language: Dummy variable indicating if two countries share a common official language, from CEPII

Common colonizer: Dummy variable indicating if two countries have been colonized by the same colonizer, from CEPII.

Colony – Colonizer relationship: Dummy variable indicating if two countries have ever had a colony-colonizer relationship, from CEPII.

Landlocked: Dummy variable indicating if a country has no direct access to the sea.

Island: Dummy variable indicating if a country is an island.

Infrastructure index: Following Limao and Venables (2001), the index is constructed as the unweighted average of four variables (each normalized to have a mean of 0 and standard deviation 1 over the whole sample period as well as in each year). As Limao and Venables (2001), we ignore missing values, making the implicit assumption that the four variables are perfect substitutes to a transport services production function. The four components are:

 - Roads: Km road per km2.

 - Paved roads: Km paved road per km2.

 - Railways: Km railways per km2.

 - Telephone main lines: Telephone main lines per 1000 inhabitants.

All four are taken from the World Bank Development Indicators 2011, or the World Bank Africa Database, 2010.

African regional or free trade agreement: Dummy variable indicating if two countries in a particular year are both a member of one of the following African regional or free trade agreements: ECOWAS, ECCAS, COMESA, SADC, UEMOA, CEMAC (or UDEAC), EAC, IGAD, CENSAD, or the recently established AFTZ.

Civil conflict: Dummy variables indicating if a country experienced the use of armed force between two parties, of which at least one is the government of a state that resulted in at least 25 and at most 999 battle-related deaths, from the International Peace Research Institute, Oslo. Source: World Development Indicators, 2011.

Civil war: Dummy variables indicating if a country experienced the use of armed force between two parties, of which at least one is the government of a state that resulted in at least 1000 battle-related deaths, from the International Peace Research Institute, Oslo. Source: World Development Indicators, 2011.

Urbanization rate: Share of the population living in urban areas, from the World Bank Development Indicators, 2011, or the World Bank Africa Database, 2010.

Gross primary enrollment: Gross enrollment ratio is the ratio of total enrollment, regardless of age, to the population of the age group that officially corresponds to the level of education shown. Primary education provides children with basic reading, writing and mathematics skills along with an elementary understanding of such subjects as history, geography, natural science, social science, art and music. From the World Bank Development Indicators, 2011, or the World Bank Africa Database, 2010.

Percentage oil rents in total GDP: Oil rents are the difference between the value of crude oil production at world prices and total costs of production. From the World Bank Development Indicators, 2011, or the World Bank Africa Database, 2010.

Percentage agriculture in total GDP: Agriculture includes forestry, hunting and fishing as well as cultivation of crops and livestock production. From the World Bank Development Indicators, 2011, or the World Bank Africa Database, 2010.

Working population per km2: Data on the working population and a country's overall area are separately taken from the World Bank Development Indicators, 2011, or the World Bank Africa Database, 2010.

Polity IV: The "Polity Score" captures a regime's authority spectrum on a 21-point scale ranging from -10 (hereditary monarchy) to +10 (consolidated democracy). From the Polity Project.

Religious similarity: Fraction measuring the probability of two people from different countries adhering to the same religion. In particular we follow Helpman and others (2008) and construct this variable as: (% Protestants in country i · % Protestants in country j) + (% Catholics in country i · % Catholics in country j) + (% Muslims in country i · % Muslims in country j). Information on each country's religious composition is taken from Robert Barro's website at Harvard.

Percentage manufacturing exports in merchandise exports: Manufactures comprise commodities in SITC sections 5 (chemicals), 6 (basic manufactures), 7 (machinery and transport equipment), and 8 (miscellaneous manufactured goods), excluding division 68 (non-ferrous metals). From the World Bank Development Indicators, 2011, or the World Bank Africa Database, 2010.

Appendix B

ROW and SSA Market Access, and Distance to the USA and Nigeria Resp.
Figure B1.

ROW and SSA Market Access, and Distance to the USA and Nigeria Resp.

Open in new tabDownload slide
Source: Authors' analysis based on data sources discussed in the main text or Appendix A.
Share of Manufacturing in Total Merchandise Exports 1993–2009
Figure B2.

Share of Manufacturing in Total Merchandise Exports 1993–2009

Open in new tabDownload slide
Notes: The figure shows box-plots for each year in our sample. Each box ranges from the 25th to the 75th percentile of the distribution of SSA countries' percentage of manufacturing exports in total merchandise exports. The horizontal line within each box denotes the median percentage of manufacturing exports in total merchandise exports. This median increases from less than 10 % to more than 20 % over our sample period. The lines extending from each box denote the upper and lower adjacent values respectively. These are calculated as 1 · 5 times the interquartile range (IQR), the difference between the 25th and 75th percentile of the overall SSA distribution of manufacturing shares in total merchandise trade in each particular year. The countries explicitly shown in the figure are those for which manufacturing exports constitute a share in overall merchandise exports that is even higher that the upper adjacent value of the overall SSA distribution in a particular year.
Source: Authors' analysis based on data sources discussed in the main text or Appendix A.

Appendix C

In this Appendix, we briefly set out the new economic geography (NEG) model that underlies our empirical framework.42 Assume the world consists of i = 1, … ,R countries, each being home to an agricultural43 and a manufacturing sector. As in virtually all NEG models, we focus on the manufacturing sector. Moreover, and in line with e.g. Redding and Venables (2004), Breinlich (2006), Knaap (2006) and Head and Mayer (2006), we restrict our attention to the ‘short-run’ version of the model. This amounts to, as Redding and Venables 2004, p.59) put it, “taking the location of expenditure and production as given and asking the question what wages can manufacturing firms in each location afford to pay its workers.”

In the manufacturing sector, firms operate under internal increasing returns to scale, represented by a fixed input requirement ciF and a marginal input requirement ci. Each firm produces a different variety of the same good under monopolistic competition using the same Cobb-Douglas technology combining two different inputs. The first is an internationally immobile factor (e.g. labor), with price wi and input share β, the second is an internationally mobile factor with price vi and input share γ, where γ + β= 1.44

Manufacturing firms sell their products to all countries. This involves shipping them to foreign markets incurring trade costs in the process. These trade costs are assumed to be of the iceberg-kind and the same for each variety produced. In order to deliver a quantity xij(z) of variety z produced in country i to country j, xij(z)Tij has to be shipped from country i. A proportion (Tij-1) of output ‘is paid’ as trade costs (Tij= 1 if trade is costless). Taking these trade costs into account gives the following profit function for each firm in country i,
(C1)
where pij(z) is the price of a variety produced in country i.
Turning to the demand side, consumers combine each firm's manufacturing variety in a CES-type utility function, with σ being the elasticity of substitution between each pair of product varieties. Given this CES-assumption, it follows directly that in equilibrium all manufacturing varieties produced in country i are demanded by country j in the same quantity (for this reason varieties are no longer explicitly indexed by (z)). Denoting country j's expenditure on manufacturing goods as Ej, country j's demand for each product variety produced in country i can be shown to be
(C2)
where Gj is the price index for manufacturing varieties that follows from the assumed CES-structure of consumer demand for manufacturing varieties. It is defined over the prices, pij, of all goods produced in country i and sold in country j,
(C3)
Maximization of profits (C1) combined with demand as specified in (C2) gives the well-known result in the NEG literature that firms in a particular country set the same f.o.b. price, pi, depending only on the cost of production in location i, i.e. pi is a constant markup over marginal costs:
(C4)

As a result, price differences between countries in a good produced in country i can only arise from differences in trade costs, i.e. pij= piTij.

Next, free entry and exit drive (maximized) profits to zero, pinpointing equilibrium output per firm at formula⁠. Combining equilibrium output with equilibrium price (C4) and equilibrium demand (C2), and noting that in equilibrium the price of the internationally (perfectly) mobile primary factor of production will be the same across countries (vi = v for all i), gives the equilibrium manufacturing wage:
(C5)
where A is a constant that contains inter alia the substitution elasticity, σ, and the fixed costs of production, F), and mj denotes country j's market capacity that is a combination of its expenditure on manufacturing goods (Ej) and the price index for manufacturing varieties that it faces (Gj), i.e. mj = Ej Gj(σ−1). In log-linear form (C5) equates to equation (1) that underlies all our estimation results in the main text of the paper.
Finally, aggregating demand from consumers in country j for a good produced in country i (see (C2)) over all firms, ni, producing in country i, gives the following aggregate export equation describing the total amount country i exports to county j.
(C6)
where we make use of the fact that pij = piTij and redefine si = nipi1−σ (what Redding and Venables 2004 refer to as a country's supplier capacity). Equation (C6) forms the basis of the two-step estimation procedure that we use in our paper.

Notes

1

Moreover, Amiti and Javorcik (2008) find that market access positively affects the amount of FDI in Chinese provinces and Lall, Shalizi and Deichmann (2004) show that market access is an important determinant of firm level productivity in India.

2

The only paper we know of focusing on the role of market access in SSA is Elbadawi, Mengistae and Zeufack (2006) that shows that differences in terms of export performance between firms in 10 SSA countries and firms in other developing countries (e.g. India, China, Malaysia or Peru) can partly be explained by SSA's poor market access. Their paper does not link export performance — or market access — to income per capita. Another paper that is similar in spirit to ours is that by Arora and Vamvakidis (2005), which looks at how South Africa's economy influences development in the rest of SSA.

3

Throughout the paper, unless explicitly noted otherwise, market access refers to a country's market access for manufactures.

4

See Fujita, Krugman, and Venables (1999), Puga (1999), Head and Mayer (2004) for more detailed expositions of various basic NEG models. See also Head and Mayer (2010), who show that the relationship between market access and economic development not only follows from NEG models but can be derived from a more general class of models.

5

In Redding and Venables (2004) and Knaap (2006), each firm also uses a composite intermediate input (made up of all manufacturing varieties) in production, allowing them to also look at the relevance of so-called supplier access for income levels. Since our goal is to establish the relevance of market access we, in line with Breinlich (2006), skip intermediate inputs and thereby ignore supplier access [this also has the advantage that we avoid the multicollinearity problems when including both market and supplier access in the estimations, see Redding and Venables (2004) and Knaap (2006)]. In this respect our derivation and application of the wage equation is similar to Hanson (2005), see also Head and Mayer (2004, pp. 2622–2624), or Head and Mayer (2010). As a robustness check, Table 4c shows results when also including constructed measures of supplier access to (1).

6

Moreover, this direct estimation strategy jointly identifies the relative importance of the different components making up a country's overall market access and the overall effect of market access on income levels. It does so solely from the spatial distribution of GDP (per capita) across countries. The nonlinear nature of (1) makes this an impossible task without putting a priori restrictions on (some) of the parameters (see e.g. Amiti and Cameron, 2007). Econometrically, the parameter on market access, χ2, and the parameters within the market access term (e.g. σ) are not separately identified when directly estimating (1).

7

Other papers using this strategy include inter alia Knaap (2006), Breinlich (2006), Mayer (2008), Head and Mayer (2006, 2010), Hering and Poncet (2010) or Bosker and Garretsen (2010).

8

See Appendix A for a full list of variables (including data sources) that we use in our analysis.

9

This problem is much less present when looking at different samples of countries (e.g. Europe, North America and even South-East Asia and parts of Latin America) where trade is dominated by manufacturing goods. We have also done our analysis using total bilateral SSA exports as the basis for constructing our market access measures. When using these measures we do not find a significant effect of market access on economic development (results are available upon request). This could be an indication that relative location to markets for a country's natural resources (which dominate SSA exports to the rest of the world) does not matter. However, given that (as stressed in the main text) a theoretical underpinning of a relationship between market access for primary products and economic development is lacking, we decided not to particularly stress this finding. We do add controls for a country's economy's dependence on natural resources when estimating (5).

10

This is the usual choice in the gravity literature (see e.g. Limao and Venables, 2001; Subramanian and Tamarisa, 2003). See Hummels (2001) for a critique on this, arguing in favor of an additive specification instead.

11

Tariffs are also an important component of trade costs. However, using the available SSA tariff data available in UN TRAINS, reduces our sample from 16560 to 546, 1251 or 4859 for 1993, 2000 and 2009 respectively. For this reason we excluded tariffs from our trade cost specification.

12

Note that, due to the assumed CES utility function, the NEG model set out in Appendix C in principle implies that each country trades at least something with every other country. This implies that using the NEG trade equation in explaining both the zero and the non-zero trade flows ascribes the zero observations to the error term only (relying on arguments of measurement error or reporting errors, see also Santos Silva and Tenreyro, 2006, p. 643). Although maybe defendable when looking at samples with a limited amount of ‘zeroes’, we think this is very unlikely in our SSA case, where about 80 percent of the observations are zeroes.

13

Another disadvantage of the Heckman two-step method is that it does not adequately take account of the heteroscedasticity inherently present in bilateral trade data (see Santos Silva and Tenreyro, 2006). However, we think that the disadvantages of the current methods available that do do this (see the discussion in the text and also footnote 12), that is, either assuming exogenous sample selection (OLS on the non-zeroes or zero-inflated Poisson) or imposing that the zero trade flows are the result of measurement or reporting errors (PPML), do not outweigh the ability of the Heckman two-step procedure to take account of endogenous sample selection.

14

We use religious similarity in all main results presented in this paper. We have also looked into the possibility of using the other two ‘instruments’ proposed by Helpman and others (2008): “the number of days and procedures needed to start a business” and “the costs incurred when starting a business.” We constructed the same two variables using the available data in the World Bank's “Doing Business Survey”. However, using these two alternative variables in the first stage reduces our sample size significantly (data are missing for the period up to 2003, and for the 2003–2009 period using them reduces the average yearly sample size from 16560 to 12781). Moreover, we find that, in case of our sample, these two variables are poor predictors of the probability to trade in the first stage of our Heckman-estimation strategy. Results are available upon request.

15

Helpman and others (2008) also show the importance of taking explicit account of firm heterogeneity when estimating the trade equation. We decided not to do this in this paper because it is not clear what the consequences are of introducing firm heterogeneity in the NEG model that we use, and for the wage equation (5) in particular. It lies beyond the current scope of the paper to develop a fully-fledged NEG model incorporating firm heterogeneity. As such, we decided to stick to the more standard NEG model used in all previous empirical work looking at the relationship between market access and economic development, and refrain from explicitly incorporating firm heterogeneity into our analysis. This is certainly not to deny that it would be a very interesting avenue for future research.

16

Although religious similarity itself does not change over time, we hereby do allow its effect on the probability to trade to differ between each year in our sample.

17

Note that, as in case of two stage least squares, one can never fully test the validity of religious similarity as our ‘instrument’. This ultimately hinges upon believing the arguments put forward by Helpman and others (2008) in favour of using this variable to satisfy the needed exclusion restriction.

18

Religious similarity is significant at the 1 percent level in all years, except for 1993 (p-value = 0.43), 1994 and 1995 (in both years it is significant at the 10 percent level).

19

Given our main interest in the estimated coefficient of the trade equation in the second stage for our main purpose to construct various market access measure(s), the results of the first stage probit estimation are not explicitly discussed.

20

Note that comparing our findings to other studies looking at SSA-trade (e.g. Foroutan and Pritchett, 1993; Coe and Hoffmaister, 1999; Subramanian and Tamirisa, 2003 or Limao and Venables, 2001) is difficult due to the difference in estimation strategy used. These other studies use, for example, Tobit or NLS techniques to estimate the trade equation. Moreover, they usually do not include importer-year and exporter-year dummies in their regressions.

21

All results in this paper also hold when using GDP per capita instead. They are available upon request.

22

We only show bootstrapped standard errors for all our estimation results (they are based on 200 replications). The bootstrapped standard errors take explicit account of the fact that our measures of market access are all generated regressors. See Redding and Venables (2004, p. 64) for more details. Results only become stronger when using robust standard errors instead.

23

This is also partly driven by the significant reduction in sample size resulting from the fact that not all controls are available for all countries in all years (using the reduced sample in column three without including any of the controls but only country- and year-FE gives an estimated coefficient of 0.23 [p-value: 0.036]).

24

Note that the non-significance of some of our included controls may also be the result of them having very little within-variation, leaving us with a danger of making type II errors on these variables.

25

But we note that it is also not significantly different from the effect we find during the later years.

26

The closest estimate we found in these studies is the one reported on China by Hering and Poncet (2010). In their specification that includes most possible other controls related to wages, they find a positive effect of 0.05 percent in response to an increase in market access of 1 percent.

27

We note that all results presented in the paper are robust to the exclusion of South Africa from the sample.

28

We find a very similar picture when plotting ROW market access against distance to the USA. See Figure B1 in Appendix B.

29

Congo (COG) and the Democratic Republic of Congo (ZAR), are two exceptions here. These countries' SSA market access is the best in our sample, an artifact of the fact that the two main cities in these two countries (used to calculate the distance between them) are located only 10.5 km apart (the next smallest distance is that between Nigeria and Benin: 105km). Leaving these two countries out of our sample does not change any of the results presented in our paper.

30

Note that Equatorial Guinea (GNQ) shows as somewhat of an outlier in these scatterplots. This is due to its rapid economic growth over our sample period following the discovery of large oil and gas reserves. All results in our paper our fully robust to leaving this country out of the sample.

31

The results on these control variables are very similar to those in Table 2. They are available upon request, as are the results when doing the estimations using our entire sample period.

32

One could argue that our ROW market access measure suffer from too little cross-sectional variance to find any effect when controlling for country- and year-specific fixed effects. However, the scatterplots in Figure 3 suggest otherwise. Also, when including SSA and ROW market access separately we find the same results.

33

Note that this was also already borne out by the estimated effect of DMA in the early years of our sample. Although significantly positive at the 10 percent level, this effect is very small: a 1 percent increase in DMA increasing GDP per worker by only 0.004 percent.

34

This also controls for the third way by which endogeneity issues may be raised, that is, measurement error.

35

Note that this argument breaks down in case of autocorrelation in the residuals. Also including lagged variables does not solve possible endogeneity resulting from omitted variables or measurement error.

36

We also estimated (5) in first differences as an alternative way to deal with unobserved country-specific variables that are correlated with market access. Again, we find that SSA market access is the only component of market access for which we systematically find a significant positive effect on GDP per worker (results available upon request).

37

We note at this point that our policy experiments do not quantify the full general equilibrium effects on income levels. We are confining ourselves to the “short-run” effects of improving market access on income levels. We abstract from any subsequent changes in economic geography induced by e.g. firms or consumers changing their location decistion as a results of the changes in income levels induced by the change in market access resulting from one of our policy experiments.

38

Given that we no longer include importer- and exporter dummies the results on the magnitude of the effect of bilateral trade costs on bilateral exports differs from those reported in Table 1. However, the direction of their effect is never different.

39

It would be straightforward to redo these calculations using any other reported coefficient in Table 3 or Table 2 for that matter. This would change the absolute effect of each of the different policy measures on GDP per worker, but it leaves the relative magnitude of each of the different policy experiments unchanged.

40

Moreover, the impact also depends on the relative importance of a country's newly added RFTA partners for its market access compared to that of the countries with which it already shares an RFTA.

41

This effect is much smaller that the spillover-effects of South African growth on its neighbors found by Arora and Vamvakidis (2005). Their analysis is a reduced form exercise, making it hard to compare to our theory-based approach (although our findings could be reconciled with theirs by arguing that we only capture the trade-induced spillover effect, whereas they capture a composite spillover effect of South African growth including also other non-trade related spillovers). Moreover, given their chosen empirical strategy they are unable to include year fixed effects in their panel estimation so that it is impossible to exclude the possibility that (part of) their findings are driven by an omitted variable affecting both South African economic growth as well as that of other SSA countries.

42

See Fujita, Krugman, and Venables (1999), Puga (1999), Head and Mayer (2004) for more detailed expositions of various NEG models and the derivation of market access and the equilibrium wage equation in particular.

43

The agricultural sector uses labor and land to produce a freely tradable good under perfect competition that acts as the numéraire good.

44

Since our main aim is to establish the relevance of market access we, in line with for instance Breinlich (2006), skip intermediate inputs and thereby ignore supplier access for most of our analysis, do however see Table 4c for estimation results for supplier access.

References

Amiti
M.
Javorcik
B.S.
,
Trade Costs and Location of Foreign Firms in China
, 
Journal of Development Economics
,
2008
, vol.
85
1–2
(pg.
129
-
149
)

Google Scholar

Crossref
Search ADS

 
Amiti
M.
Cameron
L.
,
Economic Geography and Wages
, 
Review of Economics and Statistics
,
2007
, vol.
89
1
(pg.
15
-
29
)

Google Scholar

Crossref
Search ADS

 
Amjadi
A.
Yeats
A.J.
,
Have Transport Costs Contributed to the Relative Decline of Sub-Saharan African Exports? Some Preliminary Evidence
,
1995
 
World Bank Policy Research Paper no. 1559
OpenURL Placeholder Text

Anderson
J.E.
van Wincoop
E.
,
Trade Costs
, 
Journal of Economics Literature
,
2004
(pg.
691
-
751
) 
vol. XLII:

Google Scholar

OpenURL Placeholder Text

 
Arora
V.
Vamvakidis
A.
,
The implications of South African economic growth for the rest of Africa
, 
IMF Working Paper
,
2005
 
WP/05/58

Google Scholar

OpenURL Placeholder Text

 
Bigsten
A.
Söderbom
M.
,
“What have we learned from a decade of manufacturing enterprise surveys in Africa?”
, 
The World Bank Research Observer
,
2006
, vol.
21
(pg.
241
-
265
)

Google Scholar

Crossref
Search ADS

 
Bosker
E.M.
Garretsen
H.
,
Trade Costs, Market Access and Economic Geography
, 
Papers in Regional Science
,
2010
, vol.
89
1
(pg.
485
-
513
)

Google Scholar

Crossref
Search ADS

 
Bosker
E.M.
Brakman
S.
Garretsen
H.
Schramm
M.
,
Relaxing Hukou-Increased Labor Mobility and China's Economic Geography
,
2010
 
CEPR discussion paper 8106
OpenURL Placeholder Text

Breinlich
H.
,
The spatial income structure in the European Union – what role for Economic Geography?
, 
Journal of Economic Geography
,
2006
, vol.
6
(pg.
593
-
617
)

Google Scholar

Crossref
Search ADS

 
Buys
P.
Deichmann
U.
Wheeler
D.
,
Road Network Upgrading and Overland Trade Expansion in Sub-Saharan Africa
, 
Journal of African Economies
,
2010
, vol.
19
(pg.
399
-
402
)

Google Scholar

Crossref
Search ADS

 
Coe
D.T.
Hoffmaister
A.W.
,
North-South Trade: Is Africa Unusual?
, 
Journal of African Economics
,
1999
, vol.
8
(pg.
229
-
256
)

Google Scholar

Crossref
Search ADS

 
Collier
P.
,
Primary Commodity Dependence and Africa's Future
, 
Keynote Speech, World Bank's Annual Conference on Development Economics.
,
2002

Google Scholar

OpenURL Placeholder Text

 
Collier
P.
Gunning
J.W.
,
Explaining African Economic Performance
, 
Journal of Economic Literature
,
1999
, vol.
37
(pg.
64
-
111
)

Google Scholar

Crossref
Search ADS

 
Collier
P.
Venables
A.J.
,
Rethinking trade preferences: how Africa can diversify its exports
, 
The World Economy
,
2007
, vol.
30
8
(pg.
1326
-
1345
)

Google Scholar

Crossref
Search ADS

 
Deichmann
U.
Lall
S.V.
Redding
S.J.
Venables
A.J.
,
“Industrial location in developing countries
, 
World Bank Research Observer
,
2008
, vol.
23
2
(pg.
219
-
246
)

Google Scholar

Crossref
Search ADS

 
Easterly
W.
Reshef
A.
,
African export successes: surprises, stylized facts, and explanations
, 
NBER Working Paper
,
2010
 
no. 16597

Google Scholar

OpenURL Placeholder Text

 
Elbadawi
I.
Menigstae
T.
Zeufack
A.
,
Market access, supplier access, and Africa's manufactured exports: a firm level analysis
, 
Journal of International Trade & Economic Development
,
2006
, vol.
15
4
(pg.
493
-
523
)

Google Scholar

Crossref
Search ADS

 
Fally
T.
Paillacar
R.
Terra
C.
,
Economic geography and wages in Brazil: evidence from micro-data
, 
Journal of Development Economics
,
2010
, vol.
91
(pg.
155
-
168
)

Google Scholar

Crossref
Search ADS

 
Foroutan
F.
Pritchett
L.
,
Intra-Sub-Saharan African Trade: Is It Too Little?
, 
Journal of African Economics
,
1993
, vol.
2
(pg.
74
-
105
)

Google Scholar

Crossref
Search ADS

 
Frazer
G.
van Biesebroeck
J.
,
Trade growth under the African growth and opportunity pact
, 
Review of Economics and Statistics
,
2010
, vol.
92
1
(pg.
128
-
147
)

Google Scholar

Crossref
Search ADS

 
Freund
C.
Rocha
N.
,
What constrains Africa's exports?
, 
World Bank Economic Review
,
2011
, vol.
25
3
(pg.
361
-
386
)

Google Scholar

Crossref
Search ADS

 
Fujita
M.
Krugman
P.
Venables
A.J.
, 
The Spatial Economy
,
1999
MIT Press

Google Scholar

OpenURL Placeholder Text

 
Gallup
J.L.
Sachs
J.D.
Mellinger
A.D.
,
Geography and Economic Development, International Regional Science Review
,
1999
, vol.
22
(pg.
179
-
232
)
OpenURL Placeholder Text

Hanson
G.
,
Market Potential, Increasing Returns and Geographic Concentration
, 
Journal of International Economics
,
2005
, vol.
67
1
(pg.
1
-
24
)

Google Scholar

Crossref
Search ADS

 
Head
K.
Mayer
Th.
Henderson
V.
Thisse
J-F.
,
The Empirics of Agglomeration and Trade
, 
The Handbook of Regional and Urban Economics, volume IV
,
2004
North Holland
(pg.
2609
-
2665
)

Google Scholar

OpenURL Placeholder Text

 
Head
K.
Mayer
Th.
,
Regional wage and employment responses to market potential in the EU
, 
Regional Science and Urban Economics
,
2006
, vol.
36
5
(pg.
573
-
594
)

Google Scholar

Crossref
Search ADS

 
Head
K.
Mayer
Th.
,
Gravity, Market Potential and Economic Development
,
2010
U British Columbia/Sciences-Po. Processed

Google Scholar

Crossref
Search ADS
 
Helpman
E.
Melitz
M.
Rubinstein
Y
,
Estimating trade flows: trading partners and trading volumes
, 
Quarterly Journal of Economics
,
2008
, vol.
123
(pg.
441
-
487
)

Google Scholar

Crossref
Search ADS

 
Henderson
J.V.
Shalizi
Z.
Venables
A.J.
,
Geography and Development
, 
Journal of Economic Geography
,
2001
, vol.
1
(pg.
81
-
105
)

Google Scholar

Crossref
Search ADS

 
Hering
L.
Poncet
S.
,
Market Access and individual wages: evidence from China
, 
Review of Economics and Statistics
,
2010
, vol.
92
(pg.
145
-
159
)

Google Scholar

Crossref
Search ADS

 
Hummels
D.
,
Toward a Geography of Trade Costs
, 
mimeo, Purdue University
,
2001

Google Scholar

OpenURL Placeholder Text

 
IMF
, 
Regional Economic Outlook: Sub-Saharan Africa
,
2007
Washington, D.C

Google Scholar

OpenURL Placeholder Text

Johnson
S.
Ostry
J.
Subramanian
A.
,
Levers for growth
, 
Finance and Development
,
2006
, vol.
43
1
(pg.
28
-
31
)

Google Scholar

OpenURL Placeholder Text

 
Jones
B.F
Olken
B.A.
,
The anatomy of start-stop growth
, 
Review of Economics and Statistics
,
2008
, vol.
90
3
(pg.
582
-
587
)

Google Scholar

Crossref
Search ADS

 
Knaap
T.
,
Trade, location, and wages in the United States
, 
Regional Science and Urban Economics
,
2006
, vol.
36
5
(pg.
595
-
612
)

Google Scholar

Crossref
Search ADS

 
Krugman
P.
,
Increasing Returns and Economic Geography
, 
Journal of Political Economy
,
1991
(pg.
483
-
499
) 
nr. 3

Google Scholar

OpenURL Placeholder Text

 
Lall
S.V.
Shalizi
Z.
Deichmann
U.
,
Agglomeration economies and productivity in Indian industry
, 
Journal of Development Economics
,
2004
, vol.
73
(pg.
643
-
673
)

Google Scholar

Crossref
Search ADS

 
Limao
N.
Venables
,
Infrastructure, Geographical Disadvantage, Transport Costs, and Trade
, 
The World Bank Economic Review
,
2001
, vol.
15
(pg.
451
-
479
)

Google Scholar

Crossref
Search ADS

 
Mayer
Th
,
Market Potential and Development
, 
CEPR Discussion Paper
,
2008
London
CEPR
 
no. 6798

Google Scholar

OpenURL Placeholder Text

 
Ndulu
B.
, 
Challenges to African Growth
,
2007
World Bank

Google Scholar

Crossref
Search ADS

 
Nunn
N.
Puga
D.
,
Ruggedness: the blessing of bad geography in Africa
, 
Review of Economics and Statistics
,
2011
 
forthcoming

Google Scholar

OpenURL Placeholder Text

 
Portugal-Perez
A.
Wilson
J.S.
,
Trade costs in Africa: barriers and opportunities for reform. Policy research working paper, no.4619
,
2008
 
World Bank

Google Scholar

Puga
D.
,
The rise and fall of regional inequalities
, 
European Economic Review
,
1999
, vol.
43
(pg.
303
-
334
)

Google Scholar

Crossref
Search ADS

 
Redding
S.
Venables
A.J.
,
Economic Geography and International Inequality
, 
Journal of International Economics
,
2004
, vol.
62
1
(pg.
53
-
82
)

Google Scholar

Crossref
Search ADS

 
Subramanian
A.
Tamirisa
N.T.
,
Is Africa Integrated in the Global Economy?
, 
IMF Staff Papers
,
2003
, vol.
50
3
 
Washington, D.C

Google Scholar

OpenURL Placeholder Text

 
Staiger
D.
Stock
J.H.
,
Instrumental variables regression with weak instruments
, 
Econometrica
,
1997
, vol.
65
(pg.
557
-
586
)

Google Scholar

Crossref
Search ADS

 
Temple
J.
Wössmann
L.
,
Dualism and Cross-Country Growth Regressions
, 
Journal of Economic Growth
,
2006
, vol.
11
(pg.
187
-
228
)

Google Scholar

Crossref
Search ADS

 
United Nations Economic Commission for Africa
,
2010
Addis Ababa
Ethiopia
 
Assessing Regional Integration in Africa IV (ARIA IV) – Enhancing Intra-African Trade
United Nations Conference on Trade and Development
, 
Economic Development in Africa Report 2009 – Strengthening Regional Economic Integration for Africa's Development
,
2009
New York and Geneva

Google Scholar

OpenURL Placeholder Text

Van Biesebroeck
J.
,
Exporting raises productivity in Sub-Saharan African manufacturing firms
, 
Journal of International Economics
,
2005
, vol.
67
(pg.
373
-
391
)

Google Scholar

Crossref
Search ADS

 
Wooldridge
J.M.
, 
Introductory Econometrics-A Modern Approach
,
2003
Thomson
USA

Google Scholar

OpenURL Placeholder Text

 
World Bank
, 
Accelerating Development Outcomes in Africa-Progress and Change in the Africa Action Plan
,
2007
Washington, D.C
The World Bank
 
Africa Region

Google Scholar

OpenURL Placeholder Text

World Bank
, 
World Development Report 2009
,
2008
Washington, D.C
The World Bank

Google Scholar

OpenURL Placeholder Text

© The Author 2012. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / the world bank. All rights reserved. For permissions, please e-mail: [email protected]
JEL
F1 - Trade O10 - General O19 - International Linkages to Development; Role of International Organizations O55 - Africa
Issue Section:
Articles
Download all slides