Policy Research Working Paper 10609 Trade and Infrastructure Integration in Africa Lionel Fontagné Mathilde Lebrand Siobhan Murray Michele Ruta Gianluca Santoni Africa Region & Infrastructure Chief Economist Office November 2023 Policy Research Working Paper 10609 Abstract Economic integration of the African continent rests on two the African Continental Free Trade Area, would increase pillars: the ratification of an ambitious trade agreement African countries’ exports by an average of 3.4 percent and and massive investment in transportation infrastructure. increase gross domestic product by 0.6 percent. Comple- Leveraging a newly created city-level database on African menting trade integration by reducing transportation time exporters’ transport times, transport route optimization and on roads, ports and border posts would increase exports general equilibrium modeling of international trade, the by 11.5 percent and increase gross domestic product by 2 paper quantifies the impact of greater trade and transport percent. Major transport investments are necessary to reap integration in Africa. A pan-African agreement, such as the full benefits of the African Continental Free Trade Area. This paper is a product of the Office of the Chief Economist, Africa Region and the Infrastructure Chief Economist Office. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// www.worldbank.org/prwp. The authors may be contacted at mlebrand@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Trade and Infrastructure Integration in Africa∗ e (Bank of France, CEPII and PSE)† Lionel Fontagn´ Mathilde Lebrand (World Bank)‡ Siobhan Murray (World Bank)§ Michele Ruta (IMF)¶ Gianluca Santoni (CEPII)‖ Key Words: infrastructure; preferential trade agreements; structural gravity; general equilibrium. JEL Codes: F14, F15. ∗ The views expressed in this paper are the authors’ and do not necessarily reflect those of the institutions they belong to. We thank Antoine Bou¨ eline Carr` et, C´ ere, Thierry Mayer, Tristan Reed, Ariell Reshef, Hakobyan Shushanik, Farid Toubal and participants to various seminars and conferences who provided very useful comments on previous versions of this paper. Lionel Fontagn´ e thanks the support of the EUR grant ANR-17-EURE-0001. † lionel.fontagne@banque-france.fr. ‡ Email: mlebrand@worldbank.org. § Email: smurray@worldbank.org. ¶ Email: mruta@imf.org. ‖ Email: gianluca.santoni@cepii.fr Introduction Economic integration of the African continent rests on two pillars: large investments in transportation infras- tructure and the ratification of an ambitious pan-African trade agreement. Both aim at reducing barriers to trade, but in different ways. Better transport infrastructure reduce transport costs between African countries and between them and the rest of the world, and trade facilitation programs reduce queuing times at borders. Preferential trade agreements (PTA) reduce frictions directly between trading partners by reducing tariffs and harmonizing policies on e-commerce, investment, and intellectual property. Infrastructure improvements and trade reforms are therefore potentially complementary: reducing trade frictions alone may not be sufficient if the contracting parties are poorly connected; similarly, improving infrastructure connectivity alone may not be sufficient in the presence of trade barriers (tariff and non-tariff). The contribution of this paper is to address the question of how the effects of both pillars of integration operate in combination using granular data on transport infrastructure, information on the depth of regional trade agreements and general equilibrium trade modeling. To quantify the joint impact of trade reforms and infrastructure investment in Africa, we first build a new city- level database of travel times using a geographic information system (GIS). We map the current infrastructure and border delays for the entire continent (both within Africa and between Africa and the rest of the world). We then add the infrastructure projects as listed in the Program for Infrastructure Development in Africa (PIDA) and assess their impacts on reducing transportation times. For trade policy, we rely on data from the World Bank’s Deep Trade Agreements database (Mattoo, Rocha & Ruta 2020); which contains detailed information on the content of trade agreements on a range of topics, including competition policy, investment, capital movements, and intellectual property rights protection. We follow the cluster analysis on the distribution of e, Rocha, Ruta & Santoni (2023) to sort the agreements according to their ambition provisions done by Fontagn´ into Shallow, Medium, and Deep treaties. We use both datasets to quantify the combined trade and welfare impacts of the infrastructure investments as planned in the PIDA and an ambitious (i.e., Deep) trade agreement for the African continent. We quantify the impacts in four steps. First, we estimate a structural gravity model (Anderson & van Wincoop 2003) in which we include the level of ambition in trade agreements to explain bilateral trade flows. We include both domestic and international trade flows, as well as the trade deflection effect due to border crossings (an inverse measure of deepening globalization). Second, we give a functional form to the unobserved bilateral component that depends on transportation time, contiguity, language proximity, past colonial relations, and other unobservable country characteristics for the year 2018. This strategy allows us to determine the elasticity of domestic and international sales to transportation time. Third, after verifying that the elasticity resulting from our data is consistent with estimates widely used in the literature (Hummels 2007, Hummels & Schaur 2013), we quantify the effect of planned infrastructure investments on transportation. As a fourth and final step, we construct counterfactuals for trade policy and the transportation network. For infrastructure, we construct two alternative transportation networks based on the projected investments for 2 roads alone and for the combination of roads, ports and land borders. For each of these networks we obtain a new matrix of transport times between African countries and with the rest of the world. On trade policy, we consolidate the various existing trade agreements in Africa into a new “ambitious” trade agreement, such as the African Continental Free Trade Area (AfCFTA). Then we simulate the single and combined impact of these reforms on trade and gross domestic product of African economies. We show that an ambitious trade agreement would increase African countries’ exports by an average of 3.6 percent and their GDP by 0.6 percent. Complementing trade integration by reducing transportation time on roads, ports and border posts would increase exports by 11.5 percent and GDP by 2 percent. These results suggest that infrastructure improvements are essential to reap the full benefits of the AfCFTA. Our paper draws on several strands of research. First, there is a growing body of literature that examines the impact of transportation infrastructure projects on transportation costs, including Donaldson & Hornbeck (2016), Donaldson (2018), Fajgelbaum & Redding (2022), Allen & Arkolakis (2022), and Fan, Lu & Luo (2023). While most of this research focuses on one country and one transportation network, our work covers most African countries (and the rest of the world) and considers roads, ports and land borders simultaneously. A related stream of research examines the impact of transportation costs on exports as mediated by infrastructure, (Limao & Venables 2001, Martincus, Carballo & Cusolito 2017).1 A second strand of literature aims to quantify the income gains from new transportation infrastructure sar & Fajgelbaum 2016, Fajgelbaum & Schaal 2020). We use on using general equilibrium models of trade (Co¸ a general equilibrium model of an endowment economy to uncover the trade and income effects of a fall in the cost of trade that induces a reshaping of the geography of trade flows. This fall in the trade cost is the result of a combination of regulatory change (a regional trade agreement) and infrastructure investment. ao Third, trade models have been used extensively to quantify the effects of trade agreements (see Lim¯ 2016 for a survey), including recent Structural Gravity General Equilibrium models (Anderson & Yotov 2016, e & Santoni 2021, Fontagn´ Fontagn´ ıguez-Clare e et al. 2023), New Quantitative Trade Models (Costinot & Rodr´ 2014, Caliendo & Parro 2015) and Computable General Equilibrium models.2 While most papers examine the impact of infrastructure investment and trade reforms in isolation, we quantify their combined impact for Africa. Finally, by combining GIS analysis and trade modeling, our work relates to a final stream of research that uses granular geographic data to shed light on the consequences of major infrastructure projects. The analysis of China’s Belt and Road Initiative is a recent and striking example of this (de Soyres, Mulabdic & Ruta 2020, Baniya, Rocha & Ruta 2020, Bird, Lebrand & Venables 2020, Lall & Lebrand 2020). We depart from the latter works in three ways: (i) we extend the transportation network developed by de Soyres, Mulabdic, Murray, Rocha & Ruta (2019) to the entire African continent for both regional and international transportation; this provides us with an ideal setup for quantifying the effect of a change in relative trade costs. (ii) We expand the types of transportation projects analyzed. (iii) We use a structural gravity general equilibrium approach 1 A related but distinct line of research studies the impact of infrastructure development on allocative efficiency in developing economies. Asturias, Garc´ ıa-Santana & Ramos (2019) use plant-level data and focus on domestic sales in India, while we use multi-country aggregate data and consider domestic and international sales. 2 Examples of recent applications of New Quantitative Trade Model (NQTM) and Computable General Equilibrium models (CGE) to the Africa region are presented in the following studies Abrego, Amado, Gursoy, Nicholls & Perez-Saiz (2019) (for NQTM) and Bengoa, Mathur, Narayanan & Norberg (2021), World Bank (2020), African Development Bank (2019) (for CGE). 3 to study the effects of “deep” integration through preferential trade agreements combined with infrastructure investment. More generally, our work is also related to the findings of Goldberg & Reed (2022) on the role of demand constraints in national development and, in particular, on how these constraints can be alleviated through international market integration. The remainder of the paper is organized as follows. Section 1 describes the data used in the analysis. Section 2 presents the ex-post quantification of the trade effects of trade and transport reforms in Africa. Section 3 quantifies the impact of planned infrastructure investments on transport time. Section 4 presents a general equilibrium counterfactual analysis. Concluding remarks follow. 1 Data overview Analyzing trade and infrastructure in a general equilibrium structural gravity framework requires specific data. The focus of this section is the development and validation of an original dataset on transport times of African exporters. Built using Geographic Information System tools and network analysis, it provides a detailed, fine- scale perspective on transport times. Given the critical role that transport times play in shaping trade patterns, the production of such data is a contribution in its own right. We also discuss the sources of trade and production data and how we combined them to get an accurate picture of domestic and international trade flows. 1.1 Transportation Time Transportation times become prohibitive not only when roads are in poor condition, but also when there are delays at borders or operations at ports are inefficient. To get a complete picture of transportation costs for African exporters, we take all these aspects into account and construct a new database of transportation times at the city level. We first map the current infrastructure network to obtain our baseline transport time matrix. We then use the list of investments from the Infrastructure Development Program for Africa (PIDA) in its long-term vision (2012-2040) to create our counterfactual. We start by mapping the current transportation network to calculate transportation times between large cities in African countries and the rest of the world. The nodes in the global network, which serve as export destinations in the analysis, are all cities (worldwide) with a population over 500,000 and the two most populous cities, data permitting, in each African country. This gives a total of 1,037 cities connected to the transportation network and 32 cities with a population of less than 50,000. The subset for Africa includes 165 connected cities that serve not only as destinations but also as export origins. The final network is solved for each origin- destination pair, and routing is determined by the shortest path, adjusting for a maritime preference parameter. The construction of the transportation network at the city level proceeds as follows.3 We first build a transportation network for Africa and the rest of the world that includes land and maritime connections. For Africa, we rely primarily on major roads from Open-Street Map, including different classes of roads for which we assume different speeds (see Table A1 in the Appendix). For the rest of the world, we use the land and maritime 3 To be compatible with trade data, the city-level network is then aggregated to the country level using a theoretically consistent approach. The aggregation procedure, following Hinz (2017), is described in detail below. 4 networks and speed assumptions from de Soyres et al. (2019). In doing so, we limit the analysis to road and maritime routes for shipments within Africa, while considering rail and maritime connections for shipments to the rest of the world.4 Two main reasons justify this approach. First, road is by far the main transport mode for trade within Africa, while most international trade in the rest of the world travels by sea and by rail.5 Maritime shipping accounts for 80% in volume and 70% in value of global trade and rail is the second transportation mode used (International Transport Forum 2017). Second, most transport reforms in Africa included in the PIDA, as discussed in detail in section 3, consist of road and maritime infrastructure investments. This approach allows us to incorporate most of PIDA’s proposed interventions. We then integrate the land transport network by including delays at (land and sea) borders and cargo loading and unloading times at seaports. For this purpose, we first use the World Bank’s Doing Business indicators, which provide country-level information on the ease of crossing land or sea borders by measuring the number of documents required to trade, days to cross the border, and costs associated with exporting and importing.6 We limit our analysis to waiting times to import and export for the latest year available in the dataset. If a country has a maritime gateway, we consider both delays at the land border and delays at the maritime border. For each country, we define both land and sea border delays if the country is coastal and only land border delays if the country is landlocked. Because data are reported only for the most used gateway per country, which for coastal countries is generally a port, we assume that unobserved land-border times are a fraction of reported sea-border times.7 In addition to delays at the maritime border, ships may experience additional delays for loading and unloading, which we calibrate using the median number of days container ships stay in ports, derived from UNCTAD data.8 Finally, to combine data on land borders and port delays, we rely on the following assumptions. (i) Trucks crossing a border must pass through customs in the country of departure and the country of arrival (total crossing times are equal to the sum of times at the border of departure and the border of arrival). (ii) Export and import procedures at sea borders add up to logistics delays at ports (total times at sea borders are the sum of the country’s sea clearance time plus delays at ports). In order to reduce border delays, several countries have implemented One-Stop-Border Posts (OSBP) to enable goods, people, and vehicles to stop in a single facility in which they undergo necessary controls to exit one state and enter the adjoining state. We include such facilities using maps and status information from the PIDA dashboard which includes up-to-date implementation status and various sources (World Bank reports 4 Air transport is likely to be the preferred option for perishable goods, especially flowers (HS06) or vegetables (HS07). These goods represent less than 0.04 percent of an African country’s total exports in 2018. If all chapters from HS01 to HS08 are considered as perishable goods, the share is still less than 0.09 per cent. 5 The prevalence of roads as a mode of transportation is also reflected in the PIDA action plan, where most of the priority transportation projects concern the road network (or a combination of roads and borders). Railways, on the other hand, are typically associated with port hub development, except for the Djibouti-Addis corridor and the Kinshasa-Brazzaville road-rail bridge project. We also prioritize road transport because the rail network in Africa is composed of several disconnected components with heterogeneous market sizes, making them difficult to compare. 6 Note that we use border compliance time and do not include export documentation compliance (which is a one-time cost). Whether the country of entry is the final destination of the shipment or just a transit point, the border compliance time is the same. There may be a risk of overestimating border compliance time for transit shipments. However, in a non-free trade zone, it is not obvious that transit trucks would have less wait time than trucks arriving at their final destination. In the context of a broad study of Africa, using a single border wait time seems a safe simplifying assumption. 7 Details can be found in Table A2 in the Appendix A. 8 UNCATD reports ship dwell time at ports, this information is constructed from Maritime Traffic data and can be accessed here: https://unctadstat.unctad.org/wds/TableViewer/tableView.aspx 5 and the 2016 One-Stop Border Post Sourcebook from NEPAD/ADB/ICA/JICA) on current, in construction, and planned One-Stop-Border Posts for Africa. We assume that land border delays are reduced by half when a functioning One-Stop-Border Post allows trucks to stop at only one facility instead of two. Figure 1 shows the different road types included in the major road network in Africa, the location of ports and currently active One-Stop Border Posts.9 Figure 1: Existing transportation with roads, ports and border posts. Note : Authors’ network based on OSM road networks and locations of ports and OSBP (i.e. One-Stop-Border Posts) according to available information detailed in Appendix. The choice of transport mode is difficult to model and is subject to different trade-offs specific to each sector. As expected, given the very low speed of maritime connections, most routes with a land-based alternative do not use maritime segments. Similar to de Soyres et al. (2019), to account for the very high share of maritime transport in international trade, we run the baseline twice: without any constraints and then with a preference for maritime routes. Within the optimal path algorithm, this preference is implemented as follows: for each city pair with access to both a maritime and a rail link, the algorithm selects the maritime option whenever the shipping time is less than four times the shipping time using the rail link.10 The final network is then processed through an optimal routing algorithm, which isolates the shortest path for each city pair.11 9 The rest of the transport network for locations outside of Africa is detailed in de Soyres et al. (2019) 10 Although this criterion can be seen as a simplification of modal trade-offs, imposing this preference for maritime connections allows us to more accurately represent the observed trade-offs. 11 As an example, for a transport from Sousse (Tunisia) to Pretoria (South Africa), there are two optimal paths depending on the degree of maritime preference. In the absence of maritime preference, the optimal path is entirely by land (12,400 km of roads, i.e. 19.5 days including delays at the land borders); in the presence of maritime preference, the optimal path is by sea for a total of 30 days (including delays in ports). 6 From the previous analysis we derived the (shortest) transport time between each pair of cities. As our trade and production data are at the country level, a consistent aggregation approach is required. We address this by considering three essential parameters: location-specific friction, trade elasticity, and trade elasticity with respect to transport time. We aggregate transport time between cities at the country pair level using a theoretically robust iterative method that integrates population weights and a calibrated friction elasticity, as illustrated in Hinz (2017). The theoretically consistent aggregation for the frictions at the level of the city pairs is as follows:  1/δθ sk ml #daysij =  #daysδθ kl  si mj k∈i l∈j When aggregating at the level of a pair of countries ij , with exporter country i and destination country j , we take into account both the relative population of cities k in i and l in j , as well as the transport time elasticity, denoted as δθ.12 The transport time elasticity is calibrated using a simple iterative procedure as follows. We start with an arbitrary initial value of the elasticity (i.e. δθ = −1). Then, we estimate the elasticity in a cross-section gravity equation for the year 2018 with origin and destination fixed effects and controlling for PTAs, language proximity, and shared colonial history (border dummies are not included because border crossing time is already included in transport time).13 We retrieve the transportation time coefficient and then use it as a parameter in the aggregation equation (δθ = β transportation ). In the next iteration, this new aggregation parameter is used.14 Using a network analysis to quantify shipping times and the impact of new and improved transportation infrastructure has advantages and disadvantages. On the plus side, a network analysis can be a useful tool to support modeling of international trade flows. It can enhance the understanding of the impact of changes in infrastructure or other factors on an interconnected system allowing to extract information from spatial data. Specifically, a network analysis enables the measurement of the time it takes to go from one location to another taking into account the overall quality and quantity of infrastructure, physical obstacles (e.g. rough terrain) or other barriers (e.g. border delays), and relative performance measures (e.g. processing time at ports or when crossing a border). As with any modeling exercise, it should be acknowledged that our network is a simplified version of reality and results are affected by the quality of inputs. For example, due to limited data availability, we cannot include some elements that might influence the definition of optimal routes, such as freight service quality, service frequency, conflicts or adverse weather events. 1.1.1 Transportation Time Validation Before using it as an input in the general equilibrium counterfactual analysis of trade and infrastructure reforms, we validate the transportation time in our dataset using evidence from Hummels & Schaur (2013) and Hummels 12 The elasticity is a combination of trade elasticity θ and location specific friction elasticity δ . 13 To retrieve the aggregation parameter we rely on a PPML estimator. 14 We repeat this process until the estimated coefficient remains unchanged in its 5th digit. We do this separately for transport times with and without the sea route preference assumption. The procedure converges in 7 iterations in both cases. Without sea preference, the estimated δθ is -0.852, with sea preference, the estimated parameter is -0.780. 7 (2007). Table 1 reports the results of the two-step validation exercise. First we aim at identifying the tariff equivalent of one additional day of shipment for African exporters with the rest of the world by regressing the log of trade in 2018 on the number of days (days# ij ) needed to reach any of the destination markets in the sample.15 The estimated equation reads: ln Xij = βdays days# ij + γ Dij + ϵij (1) where i is the exporter country, j the importer country, Dij a vector of bilateral trade frictions (excluding distance), with i being restricted to African exporter countries while j being any destination in the sample. Results are shown in column 1 and 2 of Table 1. In column 1 we focus on the shortest path in the trans- portation matrix without restrictions while in column 2 we add, whenever possible, a maritime preference to the algorithm searching for the shortest path. The estimated coefficients are then translated into tariff equivalent using an elasticity of substitution of σ = 6, according to the following formula: Tariff Equivalent = (exp[(β days / − σ ) − 1]) ∗ 100.16 In our preferred specification, column 1, an additional day of transport is equivalent to a 2.4 percent tariffs for African exporters, a plausible value considering that Hummels & Schaur (2013) estimate for the United States a tariff equivalent in the range of 0.6 to 2.1 percent. With a preference for maritime routes in the calculation of the optimal transport time, we obtain a tariff equivalent of 1.8 percent (in column 2). In the second step, that is, in column 4 of Table 1, we regress the number of days against distance (without maritime preferences) and then use the equivalent tariff estimates from column 1 to infer the elasticity of trade costs (which we do not observe) to distance as in Hummels (2007), using exp[(β days / − σ ) − 1]) ∗ βdist . The results in column 6 replicate the analysis on data with maritime preferences, using the equivalent tariffs derived from column 2. The estimated equation reads: days# ij = βdist distij + ω Dij + ϵij (2) Column 3 and 5 show the estimation using the log number of days as a dependent variable by sake of comparison. In both columns 3 and 5, the estimated elasticity is around one: the number of days is proportional to the distance, as expected. The regression results on the number of days indicate an elasticity of trade costs to distance of 0.180 when transportation time is calculated without the maritime preference (column 4) and 0.178 when more weight is given to the maritime route (column 6). The implied elasticity for African exporters is therefore slightly higher than 0.151, the one estimated for the United States in Hummels (2007). The evidence presented here suggests that transport time data can be used with some confidence to quantify the economic impact of trade reforms and infrastructure investments in Africa.17 15 Note that the trade data used for the transport time validation also include domestic sales. The construction of the trade data is detailed in Section 1.2. 16 The average of elasticities of substitution significant at the 1 percent level is 6.3 in Fontagn´ e, Guimbard & Orefice (2022). 17 Note that the estimates presented to date are for 2018, and repeating the analysis for 2014 produces virtually unchanged results. The corresponding tables are available on request. 8 Table 1: Transportation time data validation Dep Var: log (Xij ) Daysij Daysij DaysMij AR DaysM ij AR log(#) # log(#) # (1) (2) (3) (4) (5) (6) # Daysij -0.143*** (0.012) # DaysM ij AR -0.105*** (0.009) log(Distance)ij 1.116*** 7.472*** 1.203*** 10.090*** (0.062) (0.408) (0.066) (0.535) Contiguityij 2.309*** 1.983*** 0.403** -0.950 0.174 -4.492*** (0.415) (0.453) (0.165) (1.037) (0.182) (1.624) Language Proximityij 0.441*** 0.360*** 0.016 0.273** 0.001 -0.417** (0.062) (0.062) (0.013) (0.126) (0.013) (0.175) Colonial Tiesij 0.718** 0.768** -0.007 0.192 0.004 0.741 (0.316) (0.323) (0.079) (0.608) (0.074) (0.689) Observations 2038 2,038 2,038 2,038 2,038 2,038 # Origins 20 20 20 20 20 20 # Destinations 110 110 110 110 110 110 Period 2018 2018 2018 2018 2018 2018 Transportation no MAR MAR no MAR no MAR MAR MAR Tariff Equivalent 2.4 1.8 Distance Elasticity 0.180 0.178 Note: MAR indicates that the transportation times are estimated considering maritime preferences. Exporter(i) and Importer (j ) fixed effects are included in all regressions. Standard errors in parenthe- ses are clustered by country-pair, *** p < 0.01, ** p < 0.05, * p < 0.1. To compute tariff equivalents we assume σ = 6. log(Distance)ij , Contiguityij and Colonial Tiesij variables are from Mayer & Zignago (2011); while for Language Proximityij we the use “lp” variable from Gurevich, Herman, Toubal & Yotov (2021) (to be more precise we use log (lp + 0.01)). 9 1.2 Trade, Production and PTAs A fully-fledged general equilibrium structural gravity approach relies on a trade matrix that includes both intra- national and inter-national manufacturing and agricultural trade flows. The inclusion of intra-national sales is critical, as the correct benchmark for trade integration is the domestic economy (Head & Mayer 2014, Yotov 2021). The trade database used for the empirical analysis includes both domestic and international sales of agricultural and manufacturing products. Domestic sales are calculated as the difference between the value of production and total exports (both measured in gross terms).18 Manufacturing output and trade come from UNIDO INDSTAT2 (2022) and UN-ComTrade, respectively.19 We follow standard practice for filling gaps in intra-national flows if output is missing or implies negative domestic sales. Specifically, we linearly interpolate production values between non-missing data points and extrapolate the remaining missing values using the ratio of gross output to value added as in Head & Mayer (2021). Agricultural product flows come from the Food and Agriculture Organization of the United Nations Statistics Division (FAOSTAT).20 The FAOSTAT reports a Detailed Trade Matrix in both quantities (in tons) and values (in thousands of US dollars) from 1986 to 2018 and a disaggregated agricultural production data. As done for the manufacturing sector, after filling in some missing values following the same procedure mentioned earlier, we combined production and trade data into a dataset containing both domestic and international sales. Finally, we combine manufacturing and agricultural data and aggregate, for each year, the underlying flows by country of origin and destination. The final dataset includes, for each year, only countries with non-missing (or nonnegative) domestic sales. Overall, the estimation dataset covers 137 countries (34 of which in the African continent) on 1-year intervals from 1986 to 2018. The counterfactual exercises are computed for year 2018 and 2014.21 Trade and production data are then combined with the list of in force Preferential Trade Agreement (PTAs) e et al. (2023). Using various clustering techniques, Fontagn´ as provided in Fontagn´ e et al. 2023 show that existing PTAs can be classified into three optimal groups based on the distribution of provisions included in each of the 18 policy areas covered by the World Bank Deep Agreement database. In what follows, we use this classification and estimate the elasticity of bilateral trade flows to different types of PTAs. 18 If the resulting value of domestic sales is negative, it is considered missing. Head & Mayer (2014) discuss an alternative options based on ratios and proposed by Romalis (2007), and Caliendo & Parro (2015) . 19 Manufacturing data are available at https://stat.unido.org/ and https://comtrade.un.org/ 20 Agricultural data are available at http://faostat3.fao.org/home/E. 21 In year 2018, the dataset covers 112 and (20 of which in the African continent); while in 2014 the total number of countries is 132 (30 of which in the African continent). Importantly, these data do not include informal trade, which can be a significant part of total agricultural trade (Bouet, Pace & Glauber 2018). The likely underestimation of trade flows between African countries may also have an impact on the estimated gains from the reduction of trade barriers between them. 10 2 Assessing Trade Impacts of PTAs and Infrastructure Investments In this paper, we employ a four-step strategy to investigate the trade implications of the AfCFTA and of the improvements in transportation infrastructure. The results of the first two stages are outlined in this section: i) estimating a structural gravity model where the ambition of signed trade agreements explains bilateral trade intensity; ii) giving a functional form to the unobserved bilateral frictions in relations between African countries and trading partners by making them dependent on transport time as measured in our data set. The third stage, which quantifies the impact of planned infrastructure investments on transport time, is presented in the section 3. The final stage, simulating the general equilibrium effects of policy and infrastructure integration in Africa, is discussed in section 4. 2.1 The Effects of Different Types of PTAs on Trade e et al. (2023) in In this section, we estimate the trade impact of the three clusters of PTAs identified in Fontagn´ a gravity framework in order to recover the partial effect on trade of existing PTAs associated with the different clusters.This estimation is performed on the worldwide sample of countries and the obtained elasticity is used in the trade policy counterfactuals addressing the economic impact of integration among African countries.22 To this end we estimate the following structural gravity equation with PPML on panel data: 3 2010 Xij,t = exp βz P T Az ij,t + β Opted Out P T Aij,t + βT IN T L BRDRij ∗ T + πi,t + χj,t + µij + ϵij,t z =1 T =1990 (3) Where Xij,t includes both intra-national and international manufacturing plus agricultural trade flows on 1-year intervals from 1986 to 2018.23 Bilateral flows are normalized with total absorption at destination (X.j,t ), as to mitigate the differences in the penalization of large and small trade flows of the PPML estimator (Eaton, Kortum & Sotelo 2013, Head & Mayer 2014). Our main variable of interest, PTAij,t , is sliced across clusters: PTAz ij,t . The dummy IN T L BRDRij takes the value of one for international trade flows. We interact this dummy with decades indexed by T (leaving the first period as reference). Exporter-time and importer-time fixed effects control for time-varying Multilateral Resistance Terms, whereas bilateral fixed effects control for time invariant unobserved characteristics of the country pairs, that can potentially lead to self-selection into PTAs (Baier & Bergstrand 2007). As an additional control, we also include the variable Opted Out P T Aij,t , to control for agreements that are no longer in force. The results presented in Table 2 present the elasticity of bilateral trade to the different types of PTAs (as classified by the clustering procedure) in column 1 to 3 for manufacturing and agriculture, manufacturing only, and agriculture only respectively. Two results emerge. First, PTAs of different ambition have a different impact on trade, with PTAs in cluster # 1 having the largest impact and the PTAs in cluster # 3 the lowest respectively. 22 The estimated elasticity captures the world average effect of trade integration. African countries register only two agreements within the “deep” class: Common Market for Eastern and Southern Africa (COMESA) and Economic Community of West African States (ECOWAS). Estimating a regional specific elasticity would likely be affected by such scanty sample. 23 Although results are robust to the use of longer intervals, e.g. 4-year intervals as in Anderson & Yotov (2016), we privilege a specification on sequential data as suggested by Egger, Larch & Yotov (2022). 11 Second, the trade impact of the most ambitious PTAs in cluster # 1 is magnified for agricultural products, a sector where ensuring compatibility of norms and controls is an important challenge. Table 2: Partial equilibrium impacts of different types of PTAs, PPML estimates Dep Var: Manufacturing Manufacturing Agriculture Xijt /Xjt and Agriculture only only (1) (2) (3) PTAz #1 ij,t 0.519*** 0.447*** 0.981*** (0.065) (0.066) (0.112) PTAz #2 ij,t 0.268*** 0.262*** 0.174*** (0.032) (0.034) (0.060) PTAz #3 ij,t 0.107*** 0.011 0.121* (0.035) (0.036) (0.072) Opted Out PTAij,t 0.157*** 0.108** -0.086 (0.043) (0.045) (0.107) INTL BRDR∗1990 0.274*** 0.233*** 0.155** (0.032) (0.034) (0.063) INTL BRDR∗2000 0.761*** 0.616*** 0.825*** (0.043) (0.048) (0.079) INTL BRDR∗2010 1.130*** 0.963*** 1.269*** (0.050) (0.056) (0.097) Observations 486,623 413,265 388,594 R-squared 0.984 0.974 0.997 Period 1986-2018 1986-2018 1986-2018 Interval 1-year 1-year 1-year FEs it jt ij it jt ij it jt ij Notes : Exporter-time (it), Importer-time (jt) and Exporter-Importer (ij ) fixed effects. Standard errors in parentheses are clustered by country-pair, *** p < 0.01, ** p < 0.05, * p < 0.1. The excluded category for the international border dummy is the period 1986-1989. e & Santoni (2021), we derive a matrix of bilateral frictions on the basis of the estimates from As in Fontagn´ column 1 of Table 2: 1−σ Xij,t τij,t = exp(µij + βT IN T L BRDRij ∗ T + β0 Opted Out P T Aij,t + βz RT Az ij,t ) × ( ) if Xij,t > 0 z Xij,t 1−σ τij,t = exp(µij + βT IN T L BRDRij ∗ T + β0 Opted Out P T Aij,t + βz RT Az ij,t ) if Xij,t = 0 z 1−σ where (Xij,t /Xij,t ) is the ratio of observed to predicted trade from Equation 3. The “estibrated” τij,t as well as the estimated inward and outward multilateral resistance terms represent the key elements for the baseline and 1−σ the counterfactual exercise that follows. Note that given σ = 6, the estimated term τij,t should be interpreted as the reciprocal of the bilateral trade friction, denoted by τij,t . Consequently, this term serves as a measure of 12 24 the ease with which trade occurs between countries i and j at time t. 2.2 The Role of Transportation in Shaping Bilateral Trade Frictions In this section we want to quantify to what extent transportation – as revealed by the network analysis above – is a source of bilateral frictions in trade relations between African countries and their trading partners. To 1−σ proceed, we start from τij,t , the estimated trade friction matrix from the previous step and regress them on transportation time. In the regression, we include the existing PTAs (i.e. the baseline) and additional controls that capture geography and historical bilateral determinants of trade (i.e. language proximity, contiguity and colonial ties) and control for unobserved exporter and importer characteristics using fixed effects. The estimation is performed in cross section for the year 2018 because we do not have time variation in our analysis of transport networks.25 Having identified the functional form for bilateral frictions, we will be able in the next section to build coun- 1−σ terfactuals using the estimated elasticity of τij,t to changes in transportation time to model our counterfactual scenarios. Table 3 presents the results of these estimates across specifications. Column 1 shows the average im- pact of transportation time. Column 2, on the other hand, distinguishes between domestic trade (namely i = j ) and cross-border trade, and shows that the elasticity of cross-border trade frictions with respect to transport time is significantly larger than that of domestic trade. In column 3 we introduce a maritime preference in the choice of the optimal route, which has a limited impact on the measured elasticity, as shown by the standard errors. Our preferred estimates are reported in columns 2 and 3. These coefficients are used in the next section to translate changes in transport time into changes in bilateral trade frictions. In our preferred specifications, we include the PTA variable as a control variable. This ensures that the elasticity of trade frictions with respect to transport time is orthogonal to the trade agreement in question. Omitting PTAs from the equation would have a limited effect on the estimate of the cross-border elasticity, 26 while the effect on the estimate of the domestic elasticity would be slightly reduced. However, it could be argued that there is an interdependence between PTAs and transportation time. Columns 4 and 5 examine this relationship by introducing an interaction between these two variables. The resulting interaction term is negative and statistically significant. For example, in column 4, the elasticity of cross-border trade frictions increases slightly when the interaction with PTAs is taken into account (−1.836 − 0.581 = −2.417). Interestingly, the elasticity of domestic trade frictions also increases when the interaction term is included (−1.024). These results suggest that the expected benefits associated with improved connectivity and synchronized trade policies 24 All other things being equal, an increase in one of the determinants of trade frictions, such as distance or transportation time, implies a decrease in the ease of trade. The underlying values of β and µij are calculated over the entire period and are therefore time-invariant, although both the baseline and the counterfactual are calculated in a cross-section (either 2018 or 2014, depending on data availability). Since 0.6% of the estimated µij are unidentified, we replace missing trade costs with standard gravity regression predictions using log distance, common border, colonial ties, and language proximity as explanatory variables. 25 When we use year 2014, although the regional coverage increases to 30 African countries, the estimated elasticity of τ 1−σ to ij,t transportation time hardly changes. In 2014, without maritime preferences (corresponding to column 4 of Table 3) the coefficient for log(Transportation Time)i̸=j is -2.369 (se 0.119); whereas the estimated coefficient for log(Transportation Time)i=j is -0.606 (se 0.157). With maritime preferences (corresponding to column 5 of Table 3) the estimated coefficient for log(Transportation Time)i̸=j is -2.029 (se 0.101); whereas the estimated coefficient for log(Transportation Time)i=j is -0.651 (se 0.157). 26 Our preferred specification in column 2 of Table 3 without the PTA variable reports an estimated coefficient of -2.454 (se 0.130) for cross-border trade and -0.408 (se 0.169) for domestic trade. The corresponding general equilibrium tables are available upon request. 13 may be larger when this relationship is incorporated into the general equilibrium analysis. As a result, using the estimates in columns 2 and 3 leads to more conservative effects. Table 3: Elasticity of bilateral frictions to transportation time 1−σ Dep Var: τij,t (1) (2) (3) (4) (5) log(Transportation Time)ij -1.572*** (0.105) log(Transportation Time)i̸=j -2.252*** -1.867*** -1.836*** -1.515*** (0.135) (0.115) (0.236) (0.228) log(Transportation Time)i=j -0.678*** -0.734*** -1.024*** -1.043*** (0.181) (0.189) (0.252) (0.266) Language Proximityij 0.206*** 0.171*** 0.151** 0.164*** 0.151** (0.059) (0.058) (0.060) (0.058) (0.060) Colonial Tiesij 0.795** 0.735** 0.737** 0.690** 0.692** (0.329) (0.334) (0.343) (0.339) (0.349) PTAij 1.085*** 0.777*** 0.870*** 2.500*** 2.447*** (0.163) (0.168) (0.170) (0.723) (0.824) PTAij × log(Transportation Time)i̸=j -0.581** -0.495* (0.242) (0.256) Observations 2,220 2,220 2,220 2,220 2,220 # Origins 20 20 20 20 20 # Destinations 110 110 110 110 110 Period 2018 2018 2018 2018 2018 Transportation no MAR no MAR MAR no MAR MAR Estimator OLS OLS OLS OLS OLS Note: MAR indicates that transportation time is estimated considering maritime preferences. Exporter(i) and Importer (j ) fixed effects are included in all regressions. Standard errors in parentheses are clustered by country-pair, *** p < 0.01, ** p < 0.05, * p < 0.1. Colonial Tiesij variable is from Mayer & Zignago (2011); while for Language Proximityij we the use “lp” variable from Gurevich et al. (2021) (to be more precise 1−σ we use log (lp + 0.01)). The dependent variable τij,t is the unexponentiated reciprocal of bilateral trade friction τij,t and indicates trade ease between countries i and j : increasing transport time implies decreasing bilateral trade easiness. 14 3 The Effects of New Infrastructures on Transportation Times Before moving to the general equilibrium exercise, we need to assess how new transport infrastructure would impact transportation times in Africa. We draw upon the list of infrastructure investments from the Program for Infrastructure Development in Africa (PIDA) for our simulations.27 PIDA has compiled a list of 232 priority projects for roads, ports, borders, airports, and rails, aiming to enhance regional integration and stimulate economic development by 2040. The list of active and future One-Stop-Border Posts is updated using the OSBP Sourcebook 2nd Edition May 2016.28 From this list, we retain projects related to roads, ports, and border posts that can be geo-referenced and supplement it with 48 additional border posts from the OSBP Sourcebook for Africa. This results in 208 geo-referenced projects (roads, ports, and border posts).29 Figure 2 displays the newly selected investments in roads, ports, and border posts. By comparing pre- and post- PIDA scenarios, we can quantify the changes in shipping times induced by the new and improved transport infrastructure projects. The implementation of PIDA projects would significantly increase the number of road and port connections and enhance the speed and processing times along numerous road segments, borders, and ports, thereby reducing shipping times for many African cities. We begin by constructing a new counterfactual transport network that includes the new road classes and investments along the land borders and ports.30 First, we incorporate road improvements into the new scenario by increasing the speed from the current road type to the average speed of the region’s freeways, based on regional road speeds that are an average of country-level speeds used in Weiss et al. 2020. Second, we reduce queuing time for border points in the new list of one-stop border posts by half, resulting in a reduction of average land border delays between African countries from 3.5 days to 2.9 days. Figure 3 reveals the large heterogeneity of delays across countries, displaying the average border delay per country. Eritrea, Saudi Arabia, Democratic Republic of the Congo, and Central African Republic have the longest current border delays, while Morocco has the shortest border delay. The countries that will benefit from the largest reductions in land border times are Democratic Republic of Congo, Congo-Brazzaville, South Sudan, Burundi, Cameroon, Malawi, and South Africa.31 Third, when a project involves building a new port or upgrading an old port, we assume that the associated processing time decreases according to the following rule: The improved port delay equals the second-best performance in the region or the second-best performance in the continent if the port is already among the best in its region. Overall, new border and port improvements at maritime gateways will reduce the average sea border delays from and to Sub-Saharan African coastal countries by 16 percent on average, with larger gains in ports with currently the highest delays. The countries that will benefit from significant reductions in sea ote d’Ivoire, Congo-Brazzaville, Democratic Republic of Congo, Kenya, Togo, border times are Cameroon, Cˆ 27 Access the PIDA projects dashboard online at https://www.au-pida.org/pida-projects/. 28 Further details available at https://www.tralac.org/documents/resources/african-union/1682-osbp-sourcebook-2nd-edition-may-2016/file.html 29 All projects included in the analysis are listed in the Appendix in Table B2 for the road projects, in Table B3 for the border improvements, and in Table B4 for the port projects. 30 A detailed description of the assumptions used to create our counterfactuals can be found in Appendix B. 31 Figure 3 is an average at the country level, with some border points that benefit from a reduction in queuing time, others not. 15 and Tanzania (Figure 4). Table 4 presents the average transportation time (Baseline ) and the expected change with the implementa- tion of all major transportation infrastructure projects, including road, port, and border improvements (with and without the preference for sea routes) for the countries in the estimation sample.32 Changes in the aver- age shipping time before and after the implementation of the PIDA projects vary across destination countries. Considering all possible investments, countries in the Middle-East and North African region benefit the most from the investments reducing border and port delays, mostly when shipping to other countries in the regions. Without imposing a preference for maritime routes, the proportional reduction in shipping time in the Africa region ranges from 2.57 percent with countries in East Asia to 19 percent within the Africa region. In general, when assuming a preference for maritime routes, changes in transportation times are more pronounced. The country-level results (shown in Appendix Table B6) exhibit significant heterogeneity, demonstrating why it is crucial to use a network approach to include interconnections between different projects and countries. For each country, transport time varies as much with the infrastructure projects that affect it as with those of neighboring countries traversed by its optimal routes. The optimal mix of sea and land routes also changes between scenarios. In the baseline, on average, 86% of the total trip length is at sea if maritime routes are preferred, and 51% if no preference is assumed. After all infrastructure investments are made, the shares drop to 84% and 45%, respectively. For certain pairs of countries, there are significant mode shifts. For example, simulated trips between Gabon and Senegal are mostly by sea in the baseline, but switch to land corridors when transport and border investments occur (in the scenario without maritime preference). In fact, the reduction in border waiting times favors land routes, and trips that took place by sea in the baseline then switch to land. On the other hand, trips between Burkina Faso and Algeria are mostly made by land in the baseline, but switch to maritime routes when border times are reduced (in the scenario with maritime preference). Indeed, the reduction of border times between Burkina Faso and its coastal neighbors makes the maritime route more accessible. In the next section, we present the results of three counterfactual scenarios that examine the combined effects of policy and infrastructure reforms in Africa. These scenarios explore various combinations of the implementation of the African Continental Free Trade Area (AfCFTA) and the Program for Infrastructure Development in Africa (PIDA), and their impact on the continent’s economic development. 32 Appendix Table B7 details the incoming and outgoing transport shocks for each country in the estimation sample. Appendix Table B5 provides regional transport statistics for all African economies. 16 Figure 2: Existing transportation network and new investments in roads, ports and border posts Note : Authors’ transport network based on the PIDA list of projects available online as documented in the Appendix. 17 Figure 3: Average land border time (baseline and counterfactual) per country 18 Note : Authors’ construction based on World Bank Doing Business Indicators. The graph shows the average land border delays per country before, in the baseline scenario, and after, in the counterfactual scenario, for each African country. Border improvements come from the reduction in border delays due to new OSBPs as shown on Figure 2. Border delays are calculated using border delays from the World Bank’s Trade Across Border database using border delays for each country. More details can be found in the Appendix. Figure 4: Total sea border time (baseline and counterfactual) per country Note : Authors’ construction based on World Bank Doing Business Indicators and UNCTAD data on port calls. The graph shows the average sea border delays per country before, in the baseline scenario, and after, in the counterfactual scenario, for each coastal African country. Border improvements come from the reduction in border delays assumed from new investments in port infrastructure as shown on Figure 2. Border delays are calculated using border delays at maritime gateways from the World Bank’s Trade Across Border database and UNCTAD data for port calls. More details can be found in the Appendix. 19 Table 4: Proportional decrease in transportation time, aggregated by regions Baseline Improvements in Origin Destination Roads land borders and ports # Days % Change Without Maritime Pref. Africa 17 -19.06 East Asia & Pacific 33 -2.57 Europe & Central Asia 23 -5.89 Africa Latin America & Caribbean 30 -4.77 Middle East 24 -6.30 North America 29 -6.69 South Asia 29 -4.35 With Maritime Pref. Africa 27 -22.77 East Asia & Pacific 38 -7.39 Europe & Central Asia 28 -10.74 Africa Latin America & Caribbean 34 -9.44 Middle East 30 -11.65 North America 33 -9.34 South Asia 33 -8.81 Note : city level transport time are aggregate at the country-pair level using a theory consistent population- weighted average. 20 4 Quantification Strategy and Counterfactual Results To assess the impact of policy and infrastructure reforms, we use a general equilibrium structural gravity model of an endowment economy. In this model, bilateral trade responds to simulated reforms through: (i) a direct effect driven by the estimated parameters β z , with z = (1, . . . , 3) in Equation 3; (ii) an indirect effect induced by third-country adjustments, such as the effect of simulated East African Community deepening on EU- Kenya trade. The usual trade diversion effect appears, corresponding to the “general equilibrium effects” of an a la Anderson & Van endowment economy (Head & Mayer 2014). The Multilateral Resistance Terms (MRT) “` Wincoop” 2003 serve as general equilibrium trade cost indices. They transmit local shocks to the overall trade friction matrix. The inward MRTs transmit the impact of the policy shock to consumers, while the outward MRTs transmit it to producers in the exporting country. We analyze three different scenarios to explore the impact of different policy and infrastructure reforms: Scenario #1: Deepening trade integration in the African region This scenario examines the AfCFTA’s enactment without additional resources for enhancing Africa’s transport infrastructure. We assume that the existing transport network remains unchanged, focusing on the trade integration effects facilitated by the AfCFTA. The results provide insights into trade liberalization’s potential benefits, excluding improvements to transportation infrastructure.33 Scenario #2: Improving connectivity in the African region The second scenario evaluates the isolated impacts of PIDA’s anticipated infrastructure investments, including enhancements to roads, ports, and borders, without the successful completion of the AfCFTA. The primary objective is to assess the potential economic benefits arising from upgraded transport infrastructure across the continent, disregarding the AfCFTA’s trade integration effects.34 Scenario #3: Combining trade and infrastructure integration The final scenario combines both trade integration through the AfCFTA and transport infrastructure invest- ment under PIDA. We differentiate between two alternative assumptions regarding the transport network: with and without a preference for maritime transport. Comparing these sub-scenarios’ outcomes, we aim to under- stand better how transport mode selection influences the overall consequences of combined reforms on Africa’s economic development. 4.1 Scenario # 1: Deepening Trade Integration in the Africa Region We now proceed to the general equilibrium analysis of the trade impact of deepening trade integration in Africa with the AfCFTA consolidating and extending all existing trade agreements between African countries, absent the improvement of transport infrastructure. Results for the first scenario, deepening regional trade integration in Africa at the extensive and intensive 33 We consolidate the existing trade agreements in Africa – and add the missing ones – into a high ambition pan-African trade agreement, the AfCFTA. We impose that non-positive entries in PTAz are switched to 1, with depth class z = 1 (the most ambitious type). We also deepen the existing agreements signed between African countries. Under this simulation, positive entries in either PTAz=3 or PTAz=2 are switched to zero while the corresponding entries in PTAz=1 are set to 1 at the intensive margin, conditional on i and j being African countries. 34 Operationally, to derive a counterfactual trade cost matrix, we combine the percentage change in transport time with the estimated elasticity from column 4 of table 3. 21 margins described above, are shown in Table 5. In column 1, we report the baseline share of intra-African export for each African country, followed by the changes in intra-African exports in column 2 and African exports with the rest of the world in column 3. Column 4 shows the change in total trade, and column 5 the change in GDP. All changes are expressed as percentage deviation from the baseline. On average, our trade integration scenario increases intra-African trade by 34.7 percent, while reducing trade with the rest of the world by 1.3 percent (diversion effects). This translates into an overall increase in GDP by 0.6 percent. These values may be viewed as modest, but one must recall that intra-African trade is hampered by trade frictions (e.g. roads). In the next sections, we gradually add the simulated impacts of the infrastructure investments. 22 Table 5: General-Equilibrium effects of deeper trade integration in the African region, no improvement of transport infrastructure Regional Trade ∆Xi ∆Xi ∆GDPi Country Baseline (%) Regional RoW (1) (2) (3) (4) (5) Algeria 2018 5.32 47.94 -0.82 1.77 0.28 Botswana 2018 15.70 33.98 -4.76 1.33 7.34 Burkina Faso 2014 27.44 35.02 3.29 11.99 0.67 Burundi 2018 16.75 40.77 8.94 14.27 0.28 Cabo Verde 2014 8.85 61.55 -1.28 4.26 0.26 Cameroon 2014 3.62 32.31 0.77 1.91 0.23 Congo, Rep. 2014 17.02 50.76 -5.98 3.67 2.70 ote d’Ivoire Cˆ 2018 7.25 26.33 -0.05 1.87 0.32 Egypt, Arab Rep. 2018 11.84 34.34 -1.65 2.61 0.32 Ethiopia 2014 3.79 19.70 0.51 1.24 0.06 Ghana 2018 8.89 44.65 -1.68 2.44 0.38 Kenya 2018 27.16 33.97 0.18 9.36 0.44 Lesotho 2018 39.69 32.70 -9.94 6.99 9.45 Madagascar 2014 7.69 43.69 0.28 3.62 0.46 Malawi 2014 31.40 28.68 2.21 10.53 0.88 Mauritius 2018 15.70 33.79 -1.90 3.70 1.13 Morocco 2018 4.08 42.72 -0.57 1.19 0.34 Mozambique 2018 14.18 44.02 -0.43 5.88 2.54 Namibia 2018 28.60 25.72 -6.32 2.84 7.42 Niger 2018 14.74 8.59 2.59 3.47 0.22 Nigeria 2018 11.27 34.59 -2.02 2.10 0.18 Rwanda 2018 14.23 59.26 11.56 18.34 0.85 Senegal 2018 17.50 38.32 -0.29 6.46 0.56 South Africa 2018 21.76 24.90 -1.33 4.38 0.68 Eswatini 2018 86.81 17.09 -18.96 12.33 9.96 Tunisia 2018 6.79 47.97 -1.58 1.78 0.78 Uganda 2018 44.72 38.64 -3.40 15.40 1.39 Tanzania 2018 40.24 30.45 -3.41 10.21 0.85 Zambia 2014 17.07 35.83 0.98 6.93 2.34 Zimbabwe 2014 67.28 17.23 -17.22 5.96 9.92 Overall 34.68 -1.34 3.41 0.64 Note : “Regional” stands for intra African exchanges. In column 1 the Regional trade percentages are computed at the baseline levels. In column 2 to column 5 percent changes with respect to the baseline in total exports and GDP of participating countries. We assume σ = 6. The reference country for the normalization is New Zealand. 23 4.2 Scenario # 2: Improving Connectivity in the Africa Region In this section, we investigate the outcomes of a counterfactual scenario that focuses on improving connectivity in Africa without creating an extensive and ambitious free trade area. Table 6 presents the results for this scenario, showcasing the impacts of improved roads, ports, and border delays by country (without mode preference) in columns (1) and (2), and the results assuming a preference for sea routes in columns (3) and (4). Overall, improving infrastructure connectivity in Africa would increase trade by 7.56 percent and GDP by 1.15 percent, relative to the baseline scenario. This could reach up to 1.46 percent if maritime transport is pri- oritized. The gains are not evenly distributed across the three levers considered: roads, ports, and land borders. Although the implementation of one-stop border posts accounts for a substantial share of the estimated gains in the simulated scenario,35 the impact of other levers is not trivial. Improvements in road infrastructure alone increase GDP by 0.11 percent on average. This increase in GDP rises to 0.33 percent when road infrastructure improvements are combined with reductions in port delays. Related infrastructure projects are particularly at- tractive from a cost-benefit perspective. According to the PIDA Priority Action Plan, the estimated investment in transportation infrastructure $25.4 billion.36 Based on the GDP impacts estimated above, road infrastructure development could yield a cumulative rate of return of 4.1 percent for the African continent, calculated as the sum of the GDP gains for the countries in our sample. The rate of return for the continent rises to 12.6 percent if road development is complemented by port improvement projects.37 At the country level, the results exhibit significant heterogeneity. When border delays are high, road network development primarily boosts internal connectivity. However, when combined with additional reductions in border and port delays, intra-continental and intercontinental connectivity significantly improves, generating ote substantial benefits. Major increases in trade flows are observed in Burkina Faso, Burundi, Cameroon, Cˆ d’Ivoire, the Arab Republic of Egypt, Kenya, Malawi, Senegal, Uganda, Zambia, and Zimbabwe. Uganda, Zambia, and Zimbabwe, being landlocked countries, indirectly benefit from port investments in transit countries such as Tanzania or Mozambique. Differences in outcomes without and with maritime preferences stem from variations in the assumptions used to construct the transport network and its counterfactuals. For instance, Egypt experiences a significant decrease in transport costs when a maritime preference is assumed. In this case, flows transition from longer maritime routes in the baseline scenario to shorter land routes when land border delays are substantially reduced between Egypt and its neighbors. This explains why the average impacts are considerably larger with the maritime preference than without. GDP effects arise from the interplay of complex general equilibrium effects, relying on a country’s character- istics and position in the global trade network. The countries that gain the most include Botswana, Cameroon, Namibia, Niger, and Zambia. Egypt and Tunisia also observe substantial GDP gains when assuming a mar- 35 In line with previous studies, customs delays have been shown to be a significant barrier to trade. For example, Volpe Martincus, Carballo & Graziano (2015) find that if all exports in Uruguay had been subject to physical inspection, and if such inspections had lasted two days, then total exports in 2011 would have been 16.4 percent lower than they actually were. Using transaction-level data from El Salvador, Carballo, Graziano, Schaur & Martincus (2022) show that upgrading transit systems to streamline border processing in developing countries can be a high-return investment (with a back-of-the-envelope estimated return of US$3 to 1). 36 Link to PIDA Priority Action Plan 37 Considering that the general equilibrium analysis is done in comparative statics we do not apply a discount factor to compute the return. 24 itime preference. Some of these countries directly benefit from infrastructure investments in their territories, while others gain from spillovers originating from investments in neighboring countries. Although some of these countries experience significant increases in their trade flows, others do not, highlighting the importance of employing a general-equilibrium model that considers the complex trade interactions among African countries and the rest of the world. 25 Table 6: General Equilibrium Effects, improving connectivity of the African Region Roads, land borders and ports Without With Maritime Pref. Maritime Pref. Country Baseline ∆Xi ∆GDPi ∆Xi ∆GDPi (1) (2) (3) (4) Algeria 2018 6.63 0.93 11.10 1.76 Botswana 2018 2.98 11.81 2.86 10.36 Burkina Faso 2014 38.37 2.01 38.38 1.99 Burundi 2018 37.07 1.35 43.22 1.45 Cabo Verde 2014 0.12 0.01 2.55 0.17 Cameroon 2014 17.46 2.70 9.47 1.89 Congo, Rep 2014 5.24 4.60 8.43 6.24 ote d’Ivoire Cˆ 2018 9.39 1.41 15.01 2.17 Egypt, Arab Rep. 2018 8.43 0.99 7.36 1.01 Ethiopia 2014 14.39 0.59 8.92 0.36 Ghana 2018 5.88 0.99 5.12 0.90 Kenya 2018 12.74 0.94 24.83 1.70 Lesotho 2018 0.60 -0.25 -0.16 -0.51 Madagascar 2014 0.12 0.01 0.74 0.09 Malawi 2014 25.22 1.99 17.07 1.36 Mauritius 2018 0.00 -0.03 -0.03 -0.02 Morocco 2018 2.20 0.59 2.09 0.61 Mozambique 2018 2.90 0.67 2.20 0.55 Namibia 2018 3.22 7.62 3.43 7.10 Niger 2018 4.28 4.15 1.59 4.51 Nigeria 2018 7.50 0.58 13.59 1.38 Rwanda 2018 34.36 1.60 36.46 1.50 Senegal 2018 9.84 1.05 13.98 1.48 South Africa 2018 4.77 0.58 3.89 0.57 Eswatini 2018 1.78 0.37 2.94 1.24 Tunisia 2018 2.90 1.19 3.05 1.36 Uganda 2018 8.97 1.06 14.49 1.72 Tanzania 2018 12.79 1.68 20.83 2.78 Zambia 2014 19.74 6.63 14.14 4.86 Zimbabwe 2014 11.81 18.85 10.02 15.76 Overall 7.56 1.15 9.61 1.46 Note : percent change compared to the baseline in total exports and GDP of participating countries. We assume σ = 6. The reference country for the normalization is New Zealand. 26 4.3 Scenario # 3: Combining Trade and Infrastructure Integration In the third scenario, we explore the potential outcomes of concurrently implementing both trade integration through the AfCFTA and transport infrastructure investment as per PIDA’s projections. By considering the combined effects of these two initiatives, we aim to provide a comprehensive understanding of the potential synergies that may arise from simultaneous policy and infrastructure development. Two sub-scenarios are considered: the first without a maritime preference (columns 1 and 2), and the second with a maritime preference (columns 3 and 4) in Table 7. The results indicate that reducing both types of trade frictions simultaneously offers cumulative benefits that exceed the sum of the individual policy effects. On average, when no maritime preference is considered, exports increase by 11.49 percent and GDP by 2 percent. When a maritime preference is assumed, the expected gains are slightly higher, with total exports and GDP increasing by 14.80 percent and 2.27 percent, respectively. 27 Table 7: General Equilibrium Effects, improving connectivity and deepen integration of the African Region Deep PTAs plus roads, borders and ports improvements Without With Maritime Pref. Maritime Pref. Country Baseline ∆Xi ∆GDPi ∆Xi ∆GDPi (1) (2) (3) (4) Algeria 2018 9.12 1.40 14.42 2.22 Botswana 2018 4.79 21.45 4.70 19.56 Burkina Faso 2014 56.63 3.09 56.00 3.02 Burundi 2018 59.20 1.83 66.42 1.89 Cabo Verde 2014 4.23 0.28 6.86 0.46 Cameroon 2014 19.96 3.01 11.63 2.14 Congo, Rep. 2014 9.57 8.28 12.91 10.12 ote d’Ivoire Cˆ 2018 11.57 1.85 18.74 2.58 Egypt, Arab Rep. 2018 12.00 1.50 13.30 1.67 Ethiopia 2014 16.51 0.67 10.64 0.43 Ghana 2018 8.15 1.39 7.90 1.27 Kenya 2018 22.93 1.43 38.71 2.21 Lesotho 2018 7.50 9.10 7.09 8.75 Madagascar 2014 3.75 0.47 4.60 0.57 Malawi 2014 41.00 3.43 31.95 2.68 Mauritius 2018 3.67 1.07 3.70 1.07 Morocco 2018 3.97 1.13 3.97 1.15 Mozambique 2018 8.47 3.20 8.01 3.03 Namibia 2018 6.24 16.86 6.55 15.96 Niger 2018 9.15 4.45 9.51 4.73 Nigeria 2018 9.55 0.81 17.65 1.49 Rwanda 2018 60.82 2.88 61.22 2.73 Senegal 2018 18.08 1.78 24.73 2.26 South Africa 2018 9.34 1.36 9.37 1.35 Eswatini 2018 13.77 10.30 15.15 11.01 Tunisia 2018 5.59 2.47 5.91 2.66 Uganda 2018 25.47 2.65 32.16 3.29 Tanzania 2018 23.83 2.77 33.30 3.65 Zambia 2014 27.90 10.53 22.33 8.43 Zimbabwe 2014 17.83 32.58 16.08 28.97 Overall 11.49 2.00 14.80 2.27 Note : percent change compared to the baseline in total exports and GDP of participating countries. We assume σ = 6. The reference country for the normalization is New Zealand. 28 5 Conclusions Major trade reforms and infrastructure investments are essential levers of economic integration for Africa. Quantifying their contribution can be challenging because of the lack of data and the presence of general equilibrium effects. In this paper, we start by constructing a new granular dataset of transport times faced by African exporters to quantify the reduction in times from the implementation of the Program for Infrastructure Development in Africa, an ambitious list of transport investments for the next decade. We then use data from the World Bank’s Deep Trade Agreements project to assess the response of trade to different degrees of agreement ambition using a structural gravity model. We recover the elasticity of domestic and international sales to transportation time for African countries, with a functional form for unobserved bilateral frictions, making them depend on transport time and other unobservable characteristics of the origin and destination countries. Finally, we build on our calibrated model to simulate the trade and economic impacts of our investment scenarios in roads, ports, and land borders, and the consolidation of various trade agreements signed in Africa into the African Continental Free Trade Area. Our results show that trade reforms and infrastructure improvements complement each other. The combi- nation of tariff reduction and trade facilitation, which improves border delays and behind-the-border frictions, can substantially reduce trade costs and increase exports by 11.5 percent and GDP by 2 percent for countries in the continent. 29 Bibliography Abrego, M. L., Amado, M. A., Gursoy, T., Nicholls, G. P. & Perez-Saiz, H. (2019), The African Continental Free Trade Agreement: welfare gains estimates from a general equilibrium model, International Monetary Fund working paper 19/124. African Development Bank (2019), Integration for Africa’s Economic Prosperity, African Economic Outlook, Ch.3, African Development Bank Group. Allen, T. & Arkolakis, C. (2022), ‘The Welfare Effects of Transportation Infrastructure Improvements’, The Review of Economic Studies 89(6), 2911–2957. Anderson, J. E. & van Wincoop, E. (2003), ‘Gravity with gravitas: A solution to the border puzzle’, American Economic Review 93(1), 170–192. Anderson, J. E. & Yotov, Y. V. (2016), ‘Terms of trade and global efficiency effects of free trade agreements, 1990–2002’, Journal of International Economics 99, 279–298. ıa-Santana, M. & Ramos, R. (2019), ‘Competition and the welfare gains from transporta- Asturias, J., Garc´ tion infrastructure: Evidence from the golden quadrilateral of india’, Journal of the European Economic Association 17(6), 1881–1940. Baier, S. L. & Bergstrand, J. H. (2007), ‘Do free trade agreements actually increase members’ international trade?’, Journal of international Economics 71(1), 72–95. Baniya, S., Rocha, N. & Ruta, M. (2020), ‘Trade effects of the new silk road: A gravity analysis’, Journal of Development Economics 146, 102467. Bengoa, M., Mathur, S., Narayanan, B. & Norberg, H. C. (2021), ‘Environmental effects of the african conti- nental free trade agreement: a computable general equilibrium model approach’, Journal of African Trade 8(2), 36–48. Bird, J., Lebrand, M. & Venables, A. J. (2020), ‘The belt and road initiative: Reshaping economic geography in central asia’, Journal of Development Economics 144, 102441. Bouet, A., Pace, K. & Glauber, J. W. (2018), Informal cross-border trade in Africa: How much? Why? And what impact?, IFPRI discussion papers 1783, International Food Policy Research Institute (IFPRI). Caliendo, L. & Parro, F. (2015), ‘Estimates of the trade and welfare effects of NAFTA’, The Review of Economic Studies 82(1), 1–44. Carballo, J., Graziano, A. G., Schaur, G. & Martincus, C. V. (2022), ‘The Effects of Transit Systems on International Trade’, The Review of Economics and Statistics pp. 1–41. Costinot, A. & Rodr´ ıguez-Clare, A. (2014), Chapter 4 - trade theory with numbers: Quantifying the conse- quences of globalization, in G. Gopinath, E. Helpman & K. Rogoff, eds, ‘Handbook of International Eco- nomics’, Vol. 4 of Handbook of International Economics, Elsevier, pp. 197–261. sar, A. K. & Fajgelbaum, P. D. (2016), ‘Internal geography, international trade, and regional specialization’, Co¸ American Economic Journal: Microeconomics 8(1), 24–56. de Soyres, F., Mulabdic, A., Murray, S., Rocha, N. & Ruta, M. (2019), ‘How much will the belt and road initiative reduce trade costs?’, International Economics 159, 151–164. 30 de Soyres, F., Mulabdic, A. & Ruta, M. (2020), ‘Common transport infrastructure: A quantitative model and estimates from the belt and road initiative’, Journal of Development Economics 143, 102415. Donaldson, D. (2018), ‘Railroads of the raj: Estimating the impact of transportation infrastructure’, American Economic Review 108(4-5), 899–934. Donaldson, D. & Hornbeck, R. (2016), ‘ Railroads and American Economic Growth: A “Market Access” Ap- proach’, The Quarterly Journal of Economics 131(2), 799–858. Eaton, J., Kortum, S. & Sotelo, S. (2013), International Trade: Linking Micro and Macro, Vol. 2 of Econometric Society Monographs, Cambridge University Press, p. 329–370. Egger, P. H., Larch, M. & Yotov, Y. (2022), ‘Gravity estimations with interval data: Revisiting the impact of free trade agreements’, Economica 89(353), 44–61. Fajgelbaum, P. D. & Schaal, E. (2020), ‘Optimal transport networks in spatial equilibrium’, Econometrica 88(4), 1411–1452. Fajgelbaum, P. & Redding, S. J. (2022), ‘Trade, structural transformation, and development: Evidence from Argentina 1869–1914’, Journal of Political Economy 130(5), 1249–1318. Fan, J., Lu, Y. & Luo, W. (2023), ‘Valuing Domestic Transport Infrastructure: A View from the Route Choice of Exporters’, The Review of Economics and Statistics pp. 1–18. Fontagn´e, L., Guimbard, H. & Orefice, G. (2022), ‘Tariff-based product-level trade elasticities’, Journal of International Economics 137, 103593. e, L. & Santoni, G. (2021), ‘GVCs and the endogenous geography of RTAs’, European Economic Review Fontagn´ 132, 103656. e, L., Rocha, N., Ruta, M. & Santoni, G. (2023), ‘The Economic Impact of Deepening Trade Agree- Fontagn´ ments’, The World Bank Economic Review . lhad005. Goldberg, P. K. & Reed, T. (2022), Demand-Side Constraints in Development: The Role of Market Size, Trade, and (In)Equality, Technical report, Working Paper. Gurevich, T., Herman, P. R., Toubal, F. & Yotov, Y. V. (2021), One Nation, One Language? Domestic Language Diversity, Trade and Welfare, CESifo Working Paper Series 8860, CESifo. Head, K. & Mayer, T. (2014), Gravity Equations: Workhorse, Toolkit, and Cookbook, Vol. 4 of Handbook of International Economics, Elsevier, chapter 3, pp. 131–195. Head, K. & Mayer, T. (2021), ‘The United States of Europe: A gravity model evaluation of the four freedoms’, Journal of Economic Perspectives 35(2), 23–48. Hinz, J. (2017), The view from space: Theory-based time-varying distances in the gravity model, Annual ur Socialpolitik / German conference 2017 (Vienna): Alternative structures for money and banking, Verein f¨ Economic Association. Hummels, D. (2007), ‘Transportation costs and international trade in the second era of globalization’, Journal of Economic Perspectives 21(3), 131–154. Hummels, D. L. & Schaur, G. (2013), ‘Time as a trade barrier’, American Economic Review 103(7), 2935–59. International Transport Forum (2017), ITF Transport Outlook 2017. 31 Lall, S. V. & Lebrand, M. (2020), ‘Who wins, who loses? understanding the spatially differentiated effects of the belt and road initiative’, Journal of Development Economics 146, 102496. Limao, N. & Venables, A. J. (2001), ‘Infrastructure, geographical disadvantage, transport costs, and trade’, The world bank economic review 15(3), 451–479. ao, N. (2016), Chapter 6 - preferential trade agreements, Vol. 1 of Handbook of Commercial Policy, North- Lim˜ Holland, pp. 279–367. Martincus, C. V., Carballo, J. & Cusolito, A. (2017), ‘Roads, exports and employment: Evidence from a developing country’, Journal of Development Economics 125, 21–39. Mattoo, A., Rocha, N. & Ruta, M. (2020), Handbook of Deep Trade Agreements, number 34055 in ‘World Bank Publications’, The World Bank. Mayer, T. & Zignago, S. (2011), Notes on CEPII’s distances measures: The geodist database, Working Papers 2011-25, CEPII. Romalis, J. (2007), ‘NAFTA’s and CUSFTA’s impact on international trade’, The review of Economics and Statistics 89(3), 416–435. Volpe Martincus, C., Carballo, J. & Graziano, A. (2015), ‘Customs’, Journal of International Economics 96(1), 119–137. Weiss, D., Nelson, A., Vargas-Ruiz, C., Gligoric, K., Bavadekar, S., Gabrilovich, E., Bertozzi-Villa, A., Rozier, J., Gibson, H., Shekel, T., Kamath, C., Lieber, A., Schulman, K., Shao, Y., Qarkaxhija, V., Nandi, A., Keddie, S., Rumisha, S., Amratia, P. & Gething, P. (2020), ‘Global maps of travel time to healthcare facilities’, Nature Medicine 26, 1–4. World Bank (2020), The African Continental Free Trade Area : Economic and Distributional Effects. Yotov, Y. (2021), The Variation of Gravity within Countries, School of Economics Working Paper Series 2021-12, LeBow College of Business, Drexel University. 32 A Network analysis - Method and assumptions As a starting point, we build a network model which accounts for the current transportation network and we use it to compute transportation times between all city-pairs using a shortest-path algorithm. From this reference point, we run multiple improved scenarios that account for the planned infrastructure projects linked to the PIDA and we assess the reduction in shipping times resulting from these projects. We build a transport network for Africa and the rest of the world which includes land and maritime transport links. To simplify the network, we restrict the analysis to road and maritime links for Africa abstracting from rail and air connectivity, and to rail and maritime links in the rest of the world. The multimodal network used in this analysis was constructed by merging two types of features: maritime and land (roads and rail). These features are largely non-overlapping and separated in space. For Africa, we rely on the road network from Open-Street Map, including most major road types for which we assume different speeds, as well some minor roads and ferries that provide essential connectivity. For the rest of the world, we use the rail and maritime networks from de Soyres et al. (2019) assuming different speeds on maritime and rail links. The rail network comes from Delorme Atlas of the Earth (DAE), 2015 release. Maritime features were generated separately for the Pacific, Atlantic and Indian Oceans, and for the Caribbean Basin and Seas of Indonesia. To simulate plausible transoceanic shipping routes, great circle arcs were generated from near-port locations along the boundary of Exclusive Economic Zones. More details on the maritime and rail networks can be found in de Soyres et al. (2019). Each type of network element is assigned a cost calculated as the segment length divided by speed, where the speed of distinctive features is presented in Table A1. Segments that cross land borders and port links incur additional costs, as described in sections below. Table A1: Speed assumptions (km/h) for different types of roads Eastern Africa Middle Africa Northern Africa Southern Africa Western Africa motorway 90 80 106 110 107 motorway link 40 40 68 60 43 trunk road 81 64 79 95 73 trunk link 71 54 69 85 63 primary road 62 51 62 76 71 primary link 47 30 38 45 50 secondary road 54 58 62 66 51 secondary link 44 48 52 56 41 minor road 30 rail 50 ferry 25 + 2 hours maritime 25 Note : Regional road speeds are averaged over country level speeds used in Weiss et al. (2020). Link speeds not reported (trunk link, secondary link) in the Weiss analysis are populated by subtracting 10 from the main type. Once the network features are assembled, a shortest path algorithm is used to find the least-cost route (in 33 terms of shipping time) between every city-pair in our network, considering all possible ways to link the origin city to the destination city. The output origin-destination (O-D) matrix includes 165 cities in Africa and over 1000 cities globally (inclusive of the African cities). While the network solution minimizes cost in time, we also calculate the total route length in kilometers and length of maritime portion. The construction of a transport dataset for network analysis requires many assumptions and simplifications. First, the lack of data on traffic and road condition surely limits our ability to properly capture some routing. These limitations may also lead to an underestimation of gains resulting from some road projects, due to the utilization of roads in the baseline that are unsuitable for trucking, may require much lower speeds or even be impassable in rainy season. Nevertheless, comparison of baseline results with other sources (google.com, openstreetmap.org, searoutes.com) demonstrates general agreement with a small number of route compar- isons. We account for the very high share of maritime shipping in international trade by using a network preference to produce a maritime alternative where land connectivity exists (same as in de Soyres et al. (2019)). Within the optimal path algorithm, this preference is implemented in the following way: for any city pair that has access to both a maritime and a rail link, the algorithm selects the maritime option whenever the shipping time is lower than four times the shipping time incurred using the rail link. While such a criterion can be seen as a strong simplification of the actual trade-offs, the imposition of such a preference for maritime links allows us to have a representation of the world that is very much in line with other data sources in terms of both modal shares and total shipping times. Assumptions for port delays On top of the time cost computed along the transport network elements, we account for processing times when a shipment is reaching / leaving ports due to delays in loading and unloading and other inefficiencies creating congestion at ports. We calibrate port congestion using the median number of days for container ships at ports from UNCTAD data on time vessels spend in port. The aggregated figures are derived from the fusion of automatic identification system (AIS) information with port mapping intelligence by MarineTraffic (at http://marinetraffic.com), covering ships of 1000 GT. The data used is the median time vessels spent within port limits (in days). Figure A1 shows the average port time based on UNCTAD delays at port for 2019 per maritime area. Ports in Sub-Saharan African countries have on average the longest delays at port, with ships waiting on average 1.8 days at port. Processing time when crossing an international border is also an important part of overall transportation time for international trade. To account for border delays in our computations, we include importing and exporting time from the “trading across borders” section in the World Bank’s Doing Business Database. The data is available for each country considered in our analysis. World Bank Doing Business indicators provide country-level information on the ease of crossing land or sea borders by measuring the number of documents required to trade, the days to cross the border, and the related costs to export and import. We restrict our analysis to time delays for importing and exporting for the latest year available. For each country, we define both land and sea border delays if the country is coastal and only land border delays if the country is landlocked. 34 Figure A1: Average port time based on UNCTAD delays at port for 2019 Note : Authors’ calculations based on UNCTAD data on port calls for the median number of days for container ships at ports. Port data are averaged at the maritime area level. Because data are reported only for the most used gateway per country, which for costal countries is generally a port, we assume that unobserved land-border times are a fraction of reported sea-border times. When only sea-border delays are available in the data, we assume that land-border delays are equal to a fourth of sea-border delays. We check that this assumption is in line with other cases when both land and sea border delays are available. In addition, we compare the obtained delays with data from external documents to ensure that this coefficient is acceptable. In order to reduce border delays, several countries have implemented One-Stop-Border Posts (OSBP) to enable goods, people, and vehicles to stop in a single facility in which they undergo necessary controls to exit one state and enter the adjoining state. For border posts in Africa, we include such facilities using maps and status information from the Program for Infrastructure Development in Africa (PIDA) and various sources (World Bank reports and the 2016 One-Stop Border Post Sourcebook from NEPAD/ADB/ICA/JICA) on current, in construction, and planned One-Stop-Border Posts for Africa. We assume that land border delays are reduced by half when a functioning One-Stop-Border Post allows trucks to stop at only one facility. Table A2 summarizes the construction of the land border delays depending on the information available from the World Bank Doing Business indicators and on whether a country-pair border has an OSBP or not. Finally, we combine these data to calculate delays for country-pair land borders and for each port. To get bilateral values for land border delays for each country-pair, we compute the total border delay as the sum of time for each border. Whenever a border is crossed along the optimal route, the “border penalty” is incurred. Figure B2 shows the average border time per type of borders. The average land border delays in Africa are 90 hours for the borders without OSBPs and 50 hours for the borders with OSBPs. Figure B3 shows the distribution of these delays for all land borders with delays below 10 hours to delays above 250 hours. The fastest land-border delays are mostly in Northern Africa and between certain countries such as Botswana and 35 Table A2: Construction of final land border times without and with OSBP Final land border time Country A without OSBP Land: timeA Land Sea: timeA Sea Country B Land: timeB Land timeA B Land + timeLand 0.25 × timeA B Sea + timeLand Sea: timeB Sea timeA B Land + 0.25 ×timeLand 0.25 ×(timeA B Sea timeLand ) Final land border time Country A with OSBP Land: timeA Land Sea: timeA Sea Country B Land: timeB Land 0.5 ×(timeA B Land + timeLand ) 0.5 ×(0.25 × timeA B Sea + timeLand ) Sea: timeB Sea 0.5 ×(timeA B A B Land + 0.25 × timeLand ) 0.5 ×(0.25 × timeSea + 0.25 × timeLand ) Namibia, Mozambique and South Africa, Kenya and Tanzania (Table A3). The longest delays are mostly in Central Africa around DRC, Central African Republic, and in Eastern Africa around North Sudan, South Sudan and Eritrea (Table A3). We finally cross-check our assumed border delays with observed delays from diverse external sources to validate our assumptions. To get delays at port, we compute the total delay as the sum of the processing time at the port to load or unload goods as defined previously and of the processing time at the maritime border. 36 Table A3: Land-border delays in hours (lowest 10 versus highest 10) Top 10: hours BWA NAM Botswana Namibia 10 MOZ ZAF Mozambique South Africa 15 KEN TZA Kenya Tanzania 20 LBY TUN Libya Tunisia 21 DZA TUN Algeria Tunisia 23 AGO NAM Angola Namibia 24 CIV GHA ote d’Ivoire Cˆ Ghana 26 BEN NGA Benin Nigeria 26 BWA ZAF Botswana South Africa 26 DZA MAR Algeria Morocco 26 Bottom 10: CAF COG Central African Republic Congo 191 COD ZMB Congo, Dem. Republic of Zambia 194 CAF COD Central African Republic Congo, Dem. Republic of 196 ETH SSD Ethiopia South Sudan 197 DJI ERI Djibouti Eritrea 212 COD SSD Congo, Dem. Republic of South Sudan 220 ERI ETH Eritrea Ethiopia 227 CAF TCD Central African Republic Chad 228 CAF SSD Central African Republic South Sudan 268 ERI SDN Eritrea North Sudan 271 Note : Authors’ calculations based on World Bank Trade Across Border Indicators. Land-border delays are assumed to be non-directional. 37 B Impacts of new PIDA projects on transportation times Description of assessed projects To build our estimates of shipping time reductions to be associated with the new continental free trade agree- ment, we need to enrich the current transportation network with planned projects. We use the set of projects described in the Program for Infrastructure Development in Africa to define the improvements. PIDA’s long- term strategic planning for Africa’s regional infrastructure (2012-40) has been conducted under the coordination of the African Union Commission, United Nations Economic Commission for Africa, African Development Bank and NEPAD Planning and Coordinating Agency, in cooperation with all African stakeholder. It aims at prior- itizing the infrastructure necessary for more integrated transport, energy, ICT and transborder water networks to boost trade, spark growth and create jobs. Among the 409 projects, 232 projects focus on transport invest- ments in ports, railways, roads, airports, and borders. As of 2018, among the 232 transport projects, only 37 are in operation, others are either in construction or at conception according to the PIDA dashboard webpage. In addition, we only keep the road, ports, and border projects with sufficient information for mapping. The result is 208 geo-referenced projects (roads, ports and border posts), summarized by country in Table B1. Projects are listed individually for roads, border posts, and for port investments. Figure B1 shows the location of all PIDA projects included in our counterfactual. Port improvements will happen in the following places: Santa Cruz de Tenerife, Praia, Matadi, Beira, Lome, Maputo, Dar Es Salaam, Douala, Moanda, Mombasa, Nacala, Walvis Bay, Luanda, Kribi, Tema, Cotonou, Abidjan, Durban, Dakar, and Pointe Noir. Road upgrades range from less than 20 km to greater than 2000 km in length for sections of some Trans-African Highways. New border posts investments will reduce delays in all regions of Africa. As shown in the right map in Figure B1, projects can have a synergistic effect, resulting in dramatic reduction in transport time when road improvements, port upgrades and OSBP are implemented along the same route. 38 Table B1: Infrastructure Improvements by Country Region Country Road (km) OSBP Port Eastern Africa: Burundi 417 5 (2) 0 Djibouti 79 1 0 Eritrea 0 0 0 Ethiopia 25 3 0 Kenya 2756 8 (5) 1 Madagascar 0 0 0 Malawi 238 5 0 Mozambique 0 7 (1) 3 Rwanda 157 9 (3) 0 Somalia 246 0 0 South Sudan 1292 3 0 Tanzania 2512 10 (4) 1 Uganda 546 7 (2) 0 Zambia 670 8 (2) 0 Zimbabwe 1480 6 (2) 0 Middle Africa: Central African Republic 445 1 0 Angola 1418 2 1 Cameroon 2554 5 2 Chad 2576 2 0 Congo, Dem. Rep. 3063 7 2 Equatorial Guinea 256 1 0 Gabon 929 1 0 Congo, Rep. 1718 2 1 Northern Africa: Algeria 2332 2 0 Egypt, Arab Rep. 0 1 0 Libya 1240 2 0 Morocco 0 2 0 Sudan 1596 2 0 Tunisia 0 2 0 Southern Africa: Botswana 0 5 0 Lesotho 0 0 0 Namibia 0 3 1 South Africa 0 6 (1) 1 Eswatini 0 2 0 Western Africa: Benin 137 3 (1) 1 Burkina Faso 0 6 (3) 0 Cabo Verde 0 0 1 Cˆote d’Ivoire 167 6 2 Gambia, The 0 1 0 Ghana 544 3 1 Guinea 0 2 0 Guinea-Bissau 0 1 0 Liberia 0 2 0 Mali 1363 7 0 Mauritania 481 1 0 Niger 2291 4 (3) 0 Nigeria 115 2 0 Senegal 234 5 1 Sierra Leone 0 0 0 Togo 52 3 (1) 1 39 Figure B1: Existing transportation network and new investments in roads, ports and border posts Note : Authors’ construction based on the PIDA dashboard. 40 Table B2: List of geo-referenced road projects from PIDA Program Project Location Abidjan Lagos Coastal Abidjan - Grand Bassam Missing Road Links ote d’Ivoire Cˆ Transport Corridor Abidjan - Lagos Corridor Highway Benin, Cˆote d’Ivoire, Ghana, Nigeria, Togo Agona Junction - Alubo Road Ghana Godomey - Pahou Road Benin Central African Huambo-Kuito Road Angola Inter-Capital Ketta-Djoum Road Cameroon, Congo, Rep. Connectivity Kinshasa - Luanda Road (Angola section) Angola Kinshasa - Luanda Road (DRC section) Congo, Dem. Rep. Kribi-Campo Bata Road Cameroon, Equatorial Guinea Kuito-Luena Road Angola Libreville-Bata Road Equatorial Guinea, Gabon Libreville-Brazzaville Road (Congo section) Congo, Rep. Libreville-Brazzaville Road (Gabon section) Gabon Luena-Luau-Dilolo Road Angola Ndende - Doussala - Dolisie Gabon, Congo, Rep. Yaounde-Brazzaville Road (Cameroon section) Cameroon Yaounde-Brazzaville Road (Congo section) Congo, Rep. Central Multimodal Bujumbura - Bugarama Road Burundi Transport Corridor Bujumbura - Rumonge (RN3) Road Section Burundi Chaya-Manyoni Road Tanzania Dar es Salaam - Chalinze - Morogoro Road Capacity Upgrade Tanzania DSM-Manyoni-Isaka Road Tanzania Gitega - Karuzi - Muyinga (RN12) Road Burundi Isaka-Lusahunga Road Tanzania Kasulu-Kibondo Nyakanazi Road Tanzania Kayanza - Bugarama Road Burundi Kidahwe-Kanazi-Kasulu Road Tanzania Kidahwe-Uvinza Road Tanzania Kigali - Ngoma Road Rwanda Kigoma-Kidahwe Road Tanzania Kobero - Muyinga Road Burundi Lusahunga-Rusumo Road Tanzania Mpanda - Kanazi Road Tanzania Mpanda-Kasulu-Nyakanazi and Kasulu-Kigoma Roads Tanzania Mpanda-Kasulu-Nyakanazi and Kasulu-Kigoma Roads Tanzania Nyahua-Chaya, Urambo-Kaliua Tanzania Kazirambwa-Chagu and Malagarasi-Uvinza Roads Tanzania Nyakarara - Mwaro - Mweya -Gitega (RN18) Road Upgrade Burundi Nyanza Lac-Mugina (RN3) Road Burundi Nzega-Tabora Road Tanzania Rumonge-Nyanza Lac Road Burundi Tabora-Nyahua Road Tanzania Urambo-Tabora Road Tanzania Uvira - Bukavu Road Dem. Rep. of Congo Dialocoto to Tambacounda on Dakar-Bamako Road Senegal Kaolack to Fatick on Dakar-Bamako Road Senegal Mako - Dialocoto Road Mali 41 Table B2: List of geo-referenced road projects from PIDA (continued) Djibouti-Addis Berbera-Hargeisa-Kalabayd-Togowuchale Road Somalia Transport Corridor Dobi - Galafi Road Ethiopia Galafi - Yakobi Road Djibouti Douala-Bangui Douala-N’Gaound´ e-N’Djamena Road er´ Chad Douala-N Djamena Douala-N’Gaound´ e-N’Djamena Road Links in Cameroon er´ Cameroon Multimodal Corridor Garoua - Boulai - N’Gaound´ e Road er´ Cameroon Lamu Gateway Eldoret - Nadapal Road Kenya Development Garissa - Isiolo Highway Kenya Isiolo - Lokichar - Lodwar - Nadapal Highway Kenya Juba-Torit-Kapoeta-Nadapal Road South Sudan Lamu - Garissa Highway Kenya North-South Bulawayo - Gwanda Road Zimbabwe Multimodal Chirundu - Harare Road Link 1 (Chirundu Border Post) Zimbabwe Transport Corridor Chirundu - Harare Road Link 2 (Makuti to Karoi) Zimbabwe Chirundu - Harare Road Link 3 (Karoi to Chinhoyi) Zimbabwe Chirundu - Harare Road Link 4 (Chinhoyi to Harare) Zimbabwe Gwanda - Beitbridge Road Zimbabwe Harare - Beitbridge Road Link 1 (Harare to Chivhu) Zimbabwe Harare - Beitbridge Road Link 2 (Chivhu to Masvingo) Zimbabwe Harare - Beitbridge Road Link 3 (Masvingo to Turn Off) Zimbabwe Harare - Beitbridge Road Link 4 (Turn Off to Beitbridge) Zimbabwe Harare - Nyamapanda Road Project Zimbabwe Kamuzu International Airport (KIA) Malawi Kitwe - Chingola Road Zambia Serenje - Nakonde Road Link 1 (Serenje - Mpika) Zambia Serenje - Nakonde Road Link 2 (Mpika - Chinsali) Zambia Serenje - Nakonde Road Link 3 (Chinsali - Nakonde) Zambia Northern Bachuma Gate - Maji ya Chumvi Road Kenya Multimodal Bungoma - Eldoret Road Kenya Transport Corridor Eldoret - Kitale Road Kenya Goma-Kisangani Road (DRC National Road No.2) Congo, Dem. Rep. Juba-Bor-Malakal-Renki-Sudan border Road South Sudan, Sudan Kabale - Kisoro Road (100km) Uganda Kampala - Eldoret Road Kenya Kampala Bombo Expressway Road Uganda Kampala-Jinja Road Uganda Kampala-Kibuye Busega-Mpigi Expressway Road Uganda Katuna - Biumba Road Uganda Masaka - Malaba Road Uganda Mbarara-Ntugamo Road Uganda Molo-Eldoret Road Kenya Mombasa - Voi Road Kenya Mombasa Southern Bypass Road Kenya Mombasa-Nairobi Toll Road Kenya Voi - Athi Road Kenya Pointe Noire - N Djamena Multimodal Dollisie - Brazzaville Road Congo, Rep. Transport Corridor Mambili - Ouesso Road Congo, Rep. Trans-African TAH2: Algiers to Lagos - Missing Road Links in Algeria Algeria Highway Programme TAH2: Algiers to Lagos - Missing Road Links in Niger Niger TAH3: Tripoli to Cape Town - Missing Road Links Libya, Niger, Nigeria TAH6: Ndjamena to Djibouti - Missing Road Links in Chad Chad TAH6: Ndjamena to Djibouti - Missing Road Links in Sudan Sudan TAH8: Lagos to Mombasa - Missing Road Links in Cameroon Cameroon TAH8: Lagos to Mombasa - Missing Road Links in CAF CAF TAH8: Lagos to Mombasa - Missing Road Links in DRC Congo, Dem. Rep. Trans-Maghreb Highway Nouakchott - Nouadhibou Road Mauritania Notes : CAF stands for the Central African Republic ; DRC stands for the Democratic Republic of the Congo. 42 Table B3: List of border post improvements from PIDA and other sources Program Project Location Abidjan-Lagos Elubo/ Noe One-Stop Border Post (OSBP) ote d’Ivoire, Ghana Cˆ Coastal Hillacondji-Sanveekondji OSBP Benin, Togo Transport Corridor Krak´e/ S´ e Badagry OSBP* em´ Benin, Nigeria No´ e OSBP* ep´ Ghana, Togo Ouidah/Hillacondji/Sanveekondji OSBP Togo Abidjan-Ouagadougou-Bamako Kaouara-Niangoloko OSBP Burkina Faso, Cˆote d’Ivoire Multimodal Corridor Pogo-Zegoua OSBP* ote d’Ivoire, Mali Cˆ Beira-Nacala Colomue/Dedza OSBP Malawi, Mozambique Multimodal Transport Forbes/Machipanda OSBP Mozambique, Zimbabwe Corridors Nyamapanda/ Cuchimano OSBP Mozambique, Zimbabwe Central Multimodal Gatumba/Kavimvira OSBP Burundi, Congo, Dem. Rep. Transport Corridor Kabanga/Kobero OSBP* Burundi, Tanzania Mutukula OSBP* Tanzania, Uganda Rusumo OSBP* Rwanda, Tanzania Dakar-Bamako-Niamey Kidira/ Diboli OSBP Mali, Senegal Multimodal Corridor Koloko/Heremakono OSBP Burkina Faso, Mali Djibouti-Addis Corridor Galafi OSBP Djibouti, Ethiopia Douala-Bangui Douala-N Djamena Campo OSBP Cameroon, Equatorial Guinea Multimodal Transport Garoua Boulai OSBP Cameroon, CAF Corridor Kouss´ e OSBP er´ Cameroon, Chad Kout´ e OSBP er´ Cameroon, Chad Kinshasa-Brazzaville Bridge Road and Rail Project & Rail to Ilebo Brazzaville-Kinshasa OSBP Congo, Dem. Rep. Congo, Rep. North-South Beitbridge OSBP South Africa, Zimbabwe Multimodal Corridor Martin’s Drift OSBP Botswana, South Africa Northern Multimodal Transport Corridor Katuna/Gatuna OSBP* Rwanda, Uganda Malaba OSBP* Kenya, Uganda Mpondwe OSBP DRC, Uganda Nadapal OSBP Kenya, South Sudan Renk (South Sudan/Sudan) OSBP South Sudan, Sudan Rusizi/Bukavu OSBP Congo, Dem. Rep. , Rwanda Trans-African Highway Zobue/Mwanza OSBP Malawi, Mozambique Trans-Maghreb Highway Dakla/Nouadhibou OSBP Mauritania, Morocco Ghardimaou OSBP Algeria, Tunisia Musaid-Soloum OSBP Egypt, Arab Rep. , Libya Oujda Tlemcen OSBP Algeria, Morocco Ras Adjir OSBP Libya, Tunisia Notes : CAF stands for the Central African Republic ; DRC stands for the Democratic Republic of the Congo. * Projects that are considered already open in the baseline. 43 Table B3: List of border post improvements from PIDA and other sources (continued) Program Project Location added from OSBP Source book Trans-Gambia Bridge Senegal, Gambia, The 2nd Edition May 2016 Lebombo/Ressano Garcia* South Africa, Mozambique Mandimba/Chiponde Mozambique, Malawi Victoria Falls* Zimbabwe, Zambia Taveta/Holili* Kenya, Tanzania Moussala Senegal, Mali Lunga Lunga/Horo Horo* Kenya, Tanzania Tabou ote d’Ivoire, Liberia Cˆ Wenela/Katima Mulilo Namibia, Zambia Unity Bridge(Mtambaswala/Namoto) Tanzania, Mozambique Lavumisa Eswatini, South Africa Oshoek/Ngwenya Eswatini, South Africa Gaya/Malanville* Niger, Benin Paga/Dakola Ghana, Burkina Faso Danane ote d’Ivoire, Liberia Cˆ Nemba/Gasenyi I* Rwanda, Burundi Mamuno/Trans Kalahari Botswana, Namibia Pioneer’s Gate/Skilpadeshek South Africa, Botswana Kagitumba/Mirama Hills* Uganda, Rwanda Lab´ezanga Mali, Niger Chirundu* Zimbabwe, Zambia Akanyaru/Kanyaru Rwanda, Burundi Kasumbalesa Zambia, DRC P´ e* etel Kol´ Niger, Burkina Faso Doussala Gabon, Congo, Rep. Rubavu/Goma Rwanda, DRC Busia/Busia* Uganda, Kenya Mpack Guinea Bissau, Senegal Kantchari/Makalondi* Niger, Burkina Faso Namanga/Namanga* Kenya, Tanzania Isibania/Sirari* Kenya, Tanzania Gisenyi/Goma Rwanda, DRC Kasumulu/Songwe Tanzania, Malawi Nigouni ote d’Ivoire, Mali Cˆ Oshikango/Santa Clara Namibia, Angola Ruhwa/Ruhwa* Rwanda, Burundi Tunduma/Nakonde Tanzania, Zambia Moyale Kenya, Ethiopia Bibia/Elegu-Nimule Uganda, South Sudan Gallaba/Metema Sudan, Ethiopia Plumtree/Ramokgwebane Zimbabwe, Botswana Mwami/Mchinji Zambia, Malawi Kazungula Botswana, Zambia Boundou/Fourdou Guinea, Senegal Kouremale Mali, Guinea Mfum Cameroon, Nigeria Cinkans´e* Togo, Burkina Faso Chavuma/Caripande Angola, Zambia Notes : CAF stands for the Central African Republic ; DRC stands for the Democratic Republic of the Congo. * Projects that are considered already open in the baseline. 44 Table B4: List of port projects from PIDA Program Project Location Central Africa Hub Port Banana Port Upgrade DRC and Rail Programme Douala Port Upgrade Cameroon Kribi Port Upgrade Cameroon Matadi Port Upgrade DRC Pointe Noire Port Upgrading Congo, Rep. Central Multimodal Dar es Salaam New Berths Vijibweni, Transport Corridor Mbwamaji and Kunduchi Tanzania Dar es Salaam New Container Terminal Tanzania Dar es Salaam New SPM Oil Terminal Tanzania Dar es Salaam Port Modernisation Tanzania Northern Multimodal Transport Corridor Mombasa Port New Container Terminal Kenya Mombasa Port New Petroleum Facility Kenya Praia-Dakar-Abidjan Marine transport system for operations Multimodal Transport Corridor from Praia to West African ports (with PPP) Cabo Verde Southern Africa Hub Port Beira New Coal Terminal Development Mozambique and Rail Programme Beira Port Dredging Mozambique Durban Port Expansion South Africa Luanda Port Expansion Angola Maputo Port Expansion Mozambique Nacala Port Container Terminal Expansion Mozambique Nacala Port New Coal Terminal Mozambique Walvis Bay Port New Container Terminal Namibia West Africa Hub Port Abidjan Port Upgrading ote d’Ivoire Cˆ and Rail Programme Cotonou Port Upgrading Benin Dakar Port Upgrading Senegal Ile Boulay New Port / Container Terminal ote d’Ivoire Cˆ Lome Port Upgrading Togo Tema Port Upgrading Ghana Notes : CAF stands for the Central African Republic ; DRC stands for the Democratic Republic of the Congo. Effects of new projects on transportation times In this section, we present the results on transportation times from the network analysis before and after the PIDA projects are implemented. We start with an initial network of links connecting all cities considered in our analysis, and we then add additional links and improved speed corresponding to transport infrastructure projects that are linked to the PIDA list of projects. In the improved scenario, road improvements are introduced as an increase in speed from the current road type to average speed of motorway links in the region. We do not add any new road links to the network, which limits the potential gains deriving from road improvements only. Similarly, when a project involves building a new port or upgrading an old port, we assume that the associated “processing time” decreases according to the following rule. The improved port delay equals the second-best performance in the region or the second-best performance in the continent if the port is already among the best in its region. Dividing Africa into 5 regions (North, West, Central, East, and South), we assume that a port investment will reduce port and border delays to the level of the second-best performer in the same region according to the UNCTAD dataset and World Bank Doing Business Indicators. 45 Figure A1 in the previous section reports the baseline and new total sea-border time per country which is the sum of the port delays for ships as reported by UNCTAD and the sea-border delays as reported by World Bank Doing Business Indicators. When a project involves a border improvement to create a One Stop Border Post, the border delay is divided by two following the rule described in Table A2. Figure B2 shows the average land border time in the baseline and counterfactual scenarios. For all borders, the average time will decrease form 3.5 days to 2.9 days. For the borders with a planned investment, the average border delays will drop from 3.8 to 1.9 days. Figure B3 shows the new distribution of land border delays when considering new OSBPs. The new distribution clearly moves to the left with fewer borders with delays from 5 to 9 days. The list of PIDA projects does not include investments for the borders with the longest delays, above 9 days. Figure B2: Average border time per type of border Note : Authors’ construction based on World Bank Doing Business Indicators 46 Figure B3: Distribution of land border delays in Africa, before and after the new investments Note : Authors’ construction based on World Bank Doing Business Indicators. 47 Total impacts on transportation times Based on the method detailed above, we estimate reductions in transportation times between all the pairs of cities considered in our analysis. To aggregate all values at the country-pair level, we take a theory consistent approach based on Hinz (2017) and detailed in the main text. At a more disaggregated level, the results assuming maritime preference feature significant heterogeneity when looking across country-pairs (Table B6 for the results from each country towards the Middle East and North Africa region, and towards the Sub-Saharan Africa region.). The countries experiencing the largest reduction in shipping times assuming maritime preference to Sub-Saharan African countries when investing in roads, borders and ports are South Sudan, Burundi, Rwanda, the Republic of Congo, Togo, Uganda, Zambia, Kenya, Tanzania and Tunisia. The countries benefiting the least are the Arab Republic of Egypt, Eritrea, Madagascar, Guinea-Bissau, Liberia and Sao-Tome and Principe. These results show why it is important to use a network approach in order to understand the interconnections of all planned infrastructure: the gains in landlocked countries are the results of the infrastructure in the country of interest as well as transit countries to reach a port. Finally, Table B7 details for each country the average percentage change in transportation time for incoming (In) and outcoming (Out) shipments as projected in the different simulation scenarios. Those shocks are used in the general equilibrium assessment as the counterfactual changes in connectivity for African economies. 48 Table B5: Proportional decrease in transportation time, aggregated by regions (all African countries) Baseline Improvements in Origin Destination Roads land borders and ports # Days % Change Without Maritime Pref. Africa 17 -18.63 East Asia & Pacific 37 -2.26 Europe & Central Asia 22 -5.47 Africa Latin America & Caribbean 29 -4.82 Middle East 25 -6.75 North America 28 -6.00 South Asia 31 -4.46 With Maritime Pref. Africa 26 -19.69 East Asia & Pacific 41 -5.77 Europe & Central Asia 27 -8.95 Africa Latin America & Caribbean 32 -8.40 Middle East 31 -9.17 North America 32 -7.79 South Asia 35 -7.14 Note : city level transport time are aggregate at the country-pair level using a theory consistent population- weighted average. 49 Table B6: Decrease in transportation time, aggregated by country Without Maritime pref With Maritime pref Baseline Improvements in: Baseline Improvements in: Roads, land borders Roads, land borders ports land borders # Days % Change # Days % Change AGO 25 -3.42 30 -9.99 BDI 28 -12.68 38 -25.63 BEN 23 -4.60 27 -5.50 BFA 25 -7.95 34 -16.00 BWA 25 -4.60 28 -6.89 CAF 29 -11.19 39 -16.10 CIV 24 -7.39 31 -18.64 CMR 31 -12.24 40 -10.82 COD 28 -9.73 37 -15.69 COG 29 -12.04 39 -23.26 COM 25 -1.34 26 -2.74 CPV 24 -1.98 26 -4.02 DJI 22 -7.95 26 -3.24 DZA 20 -6.54 25 -13.48 EGY 20 -6.77 27 -3.69 ERI 23 -2.45 26 -2.69 ETH 23 -10.23 28 -8.30 GAB 26 -5.56 28 -2.61 GHA 25 -6.45 28 -7.97 GIN 23 -6.55 27 -3.31 GMB 23 -7.38 31 -17.89 GNB 23 -4.43 27 -3.09 GNQ 28 -9.26 32 -2.87 KEN 23 -8.69 32 -21.72 LBR 26 -7.14 32 -2.96 LBY 20 -6.22 25 -11.48 LSO 30 -3.16 33 -4.43 MAR 20 -6.24 25 -6.51 MDG 26 -1.21 28 -2.53 MLI 23 -7.05 28 -7.57 MOZ 24 -4.59 27 -5.63 MRT 21 -7.28 27 -4.58 MWI 26 -9.91 29 -6.81 NAM 25 -3.57 29 -6.44 NER 23 -6.53 28 -9.27 NGA 25 -7.35 34 -18.88 RWA 28 -12.25 37 -22.44 SDN 22 -8.74 36 -7.41 SEN 22 -4.97 25 -6.75 SLE 24 -4.74 27 -3.06 SOM 23 -6.13 26 -2.72 SSD 28 -18.15 39 -21.30 STP 28 -0.98 30 -2.54 SWZ 26 -4.09 29 -5.66 TCD 26 -9.91 34 -18.37 TGO 27 -10.28 31 -17.35 TUN 19 -5.86 25 -14.28 TZA 25 -6.97 34 -24.29 UGA 25 -8.95 34 -21.31 ZAF 24 -3.98 28 -5.43 ZMB 28 -12.82 31 -9.73 ZWE 28 -10.71 30 -8.64 50 Table B7: Average transportation time shocks by country and scenario Without Maritime pref With Maritime pref Improvements in: Improvements in: Roads Roads land borders land borders ports ports Country: In Out In Out Burundi -7.37 -13.96 -7.35 -25.62 Botswana -7.47 -14.06 -7.87 -20.62 ote d’Ivoire Cˆ -27.64 -7.49 -27.62 -9.04 Algeria -2.76 -6.53 -2.26 -21.38 Egypt, Arab Rep. -7.73 -12.88 -7.53 -9.53 Ghana -6.44 -11.10 -7.45 -21.40 Kenya -0.01 -0.22 0.00 -2.45 Lesotho -0.65 -5.94 -1.21 -13.46 Morocco -0.31 -6.59 -0.51 -7.17 Mozambique -0.36 -9.80 -0.21 -7.45 Mauritius -1.67 -4.36 -1.41 -5.87 Namibia -2.74 -8.73 -3.90 -20.86 Niger 0.00 -1.28 -0.01 -0.24 Nigeria -0.85 -1.83 -1.26 -1.86 Rwanda -0.06 -0.18 -0.24 -0.60 Senegal -0.34 -3.86 -0.29 -3.87 Eswatini 0.00 0.00 0.00 0.00 Tunisia -3.83 -18.64 -3.81 -12.62 Tanzania -17.49 -7.36 -17.49 -9.93 Uganda -28.73 -5.23 -29.20 -6.29 South Africa -0.20 -7.43 -0.23 -20.53 Burkina Faso -7.67 -10.53 -7.66 -18.42 Cameroon -2.73 -5.01 -2.62 -10.16 Congo, Rep. -0.36 -2.37 -0.37 -6.13 Cabo Verde -1.72 -2.81 -2.66 -3.39 Ethiopia -3.44 -13.02 -3.75 -25.42 Madagascar -3.10 -6.88 -4.23 -13.96 Malawi -0.54 -6.28 -0.76 -6.51 Zambia -14.96 -16.38 -14.85 -12.15 Zimbabwe -28.66 -31.33 -28.59 -30.42 Notes : Transportation time changes induced by the different infrastructure improvement detailed in the previous sections. “In” stands for reduction of transportation time for incoming shipments (this includes domestic sales); whereas “Out” stands for transportation time for outcoming shipments (this only includes international shipments). 51