sign in sign up
<<
Home , French Community, French Equatorial Africa, Imperial Preference, Metropole

Independence and trade: the specic eects of French

Emmanuelle Lavallée∗ Julie Lochard†

June 2012‡

Very preliminary. Please do note cite.

Abstract Empirical evidence suggests that colonial rule and subsequent indepen- dence inuence past and current trade of former . Independence ef- fects could dier substantially across former empires if they are related to the end of dierent preferential trade arrangements. Thanks to an original dataset including new data on pre-independence bilateral trade, this paper explores the impact of independence on former colonies' trade (imports and exports) for dierent empires on the period 1948-2007. We show that independence reduces trade (imports and exports) with the former metropole and that this eect is mainly driven by former French colonies. We also nd that, after independence, trade of all former colonies increase with third countries. A close inspection of the eects over time highlights that independence eects are gradual but tend to be more rapid and more intense in the case of exports. These results oer indirect evidence for the long-lasting inuence of colonial trade policies.

Author Keywords: Trade; ; French Empire. JEL classication codes: F10; F54.

∗Université -Dauphine, LEDa, UMR DIAL. Email: [email protected]. †Erudite, University of Paris-Est Créteil. Email: [email protected]. ‡We would like to thank participants at the IRD-DIAL Seminar and participants at the Con- ference on International Economics (CIE) 2012 for useful comments and suggestions.

1 1 Introduction

Several studies highlight the consequences of colonial rule on bilateral trade. Mitchener and Wei- denmier (2008) assess the contemporaneous eects of empire on trade over the period 1870-1913. They show that belonging to an empire doubled trade relative to those countries that were not part of an empire. A country's prior colonial status is also shown to exert a still large and statis- tically signicant positive eect on current bilateral trade (see e.g. Rose, 2000; Glick and Taylor, 2006).1 This raises the question of the impact of decolonization on post-colonial trade. Head et al. (2010) investigate this issue using a worldwide data set over the period 1948-2006. They show that post-independence trade with the colonizer does not exhibit immediate signicant changes, but that after several decades, the accumulated trade erosion is large and statistically signicant. On average, trade between a and its colonizer is reduced by 65% after four decades. They obtain two other sobering results. Decolonization reduces trade between siblings, i.e. former parts of the same , in a comparable extent. Independence also decreases trade of former colonies with the rest of the world. They interpret their ndings as the result of the deterioration of business networks. However, another interpretation lies in the end of preferential trade arrangements within empires. We focus on this second interpretation and assess two additional questions related to the inuence of de- colonization. Do the eects of independence on post-colonial trade patterns change according to the colonial power? Are exports and imports impacted in the same way? Indeed, the impact of independence on bilateral trade may change drastically across empires since they implemented very dierent colonial trade policies. In this respect, the liberal is generally opposed to the protectionist (see Mitchener and Weidenmier, 2009). Moreover, colonial trade policies also diered systematically between exports to and imports from the metropole. To gain some perspective on these questions, this article investigates the eect of indepen- dence on former colonies' exports and imports and compares the consequences of independence for dierent empires. While some studies explore this issue in a comparative fashion, they only rely on descriptive statistics (Kleiman, 1976, 1977). To examine the consequences of independence on trade, we construct a new bilateral database of trade between 71 developing countries and 190 partner countries over the period 1948-2007. Most of our data are extracted from the International Monetary Fund's Direction of Trade Statistics (DOTS). But since the DOTS scarcely report data prior to colonies' independence, in particular for French colonies, we increase our database by gath- ering data from various ocial French sources on bilateral trade of former French colonies from

1Rose (2000) shows in his benchmark results for 1990 that the colonial relationship raises bilat- eral trade by a factor of 5.75, everything else equal, while having had a common colonizer makes countries bilateral trade 80% larger. The positive and persistent eect of a common colonial history is related to similar institutions or surviving business networks.

2 1948 to their independence. We are thus able to compare properly the impact of independence for former French colonies and for other former colonies, in particular the British ones. Using a gravity model of trade, we rst investigate the impact of independence on bilateral trade (exports and imports) of former colonies worldwide and examine the specic impact of independence for former French colonies as compared to other former colonies. We show that independence reduces trade (imports and exports) with the former metropole and that this eect is mainly driven by former French colonies. We also nd that, after independence, trade (imports and exports) of all former colonies increase with third countries. Second, we explore the impact of independence over time for each former colonial empire and highlight that independence eects are gradual but tend to be more rapid and more intense in the case of exports. These results oer indirect evidence for the long-lasting inuence of colonial trade policies. Finally, we investigate the impact of independence on the geographical diversication (extensive margin) of former colonies' trade. The paper is structured as follows. In section 2, we expose the various preference systems implemented by the two major European colonial powers (Britain and France) and their potential consequences on the independence eects for former colonies' trade. In section 3, we describe our data, introduce the empirical model and discuss some estimation issues. In section 4, we estimate the average impact of independence on former colonies' trade, compare the impact of indepen- dence for dierent former colonial empires and compute some robustness checks. In section 5, we investigate the impact of independence over time. In section 6, we explore the consequences of independence on the geographical extensive margin of trade. Finally, we summarize our ndings and add concluding remarks in section 7.

2 Preferential arrangements within empires and in- dependence eects

Independence may put an end to a series of special relationships - some formal, some informal - built up within empires so as to promote imperial trade at the expense of trade with the rest of the world. Informal arrangements cover business networks and all the products of historical and political connections between metropoles and their dependencies. These connections are likely to weaken following independence. As argued by Head et al. (2010), the gradual retirement of business people who facilitated trade within the empire could have induced a gradual decrease in bilateral trade relationships between the former colonizer and its former colonies (p.9). In this case, there is no obvious reason to suppose that the erosion of trade linkages could change drastically from an empire to another. On the contrary, formal relationships such as preferential trade agreements

3 and imports licensing policies implemented by colonial powers diered substantially across empires, in particular as regards the advantages they granted to their members. In this section, we briey describe the preferential arrangements implemented by the two major European colonial powers, namely Britain and France, at the time of colonization and their potential consequences on the post-colonial trade patterns of their colonies. In particular, we will argue that, given the nature of preferential schemes, independence is very likely to inuence more rapidly and more extensively the post-colonial trade patterns of former French colonies as compared to British ones. The liberal Great Britain is usually opposed the protectionist France.2 The generally favored policies and trade of British colonies was open to all foreign countries at least until 1932. Following World War I and the , the British empire departed from free trade. Since the Ottawa Imperial Economic Conference in 1932, Britain, its colonies and its dominions (, Australia, New Zealand, Union of South Africa, Irish Free State, Canada and Southern Rhodesia) have accorded to each other certain special privileges. However, although Ottawa agreements are the most important preferential agreements at the time of colonization, their importance should not be overestimated.3 The rst breach in the system existed from its creation. Indeed, a lot of British colonies (including nearly all African colonies) were excluded from imperial preferences since they were committed by international agreements preserving free trade and signed before 1932 (see de Sousa and Lochard, 2012). It is dicult to assess the eect of Ottawa preferences on members' trade notably because of the multiplicity of taris involved. In the mid-fties, three sets of colonies can be distinguished according to the preferences they gave to and received from Britain. The rst one covering Ghana, Kenya, Uganda, Tanganyika (now Tanzania), and Nigeria gave no preference to Britain and received some form of preference on over 70 per cent of their exports there. India, Malaya (now Malaysia) and Singapore, Pakistan, South Africa and Borneo territories gave extremely limited preferences but received signicant ones. At the other hand of the spectrum, Rhodesia (now Zambia and Zimbabwe) and Nyasaland (now Malawi) gave preferences on a wide range of British goods and get preference on 28 per cent of their sales to the UK (The Economist Intelligence Unit, 1960). Thus, trade preferences within the British Empire were generally quite loose and contrasted sharply with the preferential arrangements implemented by France. France and most of its colonies adopted a customs union in 1892 (Mitchener and Weidenmier, 2008). Under this regime, colonies enjoyed free trade with France for most products while non-colonies were subject to taris. Gen-

2The other empires applied a wide array of trade policies towards their colonies. For instance, Belgian colonies had low to moderate taris, whereas adopted preferential taris systems with most of its colonies. 3In the post war era, other preferential trade agreements covering UK-colonies trade are few and far between. The most important is the Commonwealth Sugar Agreement under which Britain bought xed quantities of Commonwealth sugar at higher than world prices.

4 erally suspended during World War II, taris had been progressively reintroduced in most French colonies. Broadly speaking, trade within the French customs territory (France, , and over- seas departments) and trade between the French customs territory and the French empire was exempted from customs duties for most products or beneted from preferential duties.4 As regards trade with the rest of the world, colonies belonging to the French customs territory applied the metropolitan taris and other colonies generally applied their own taris (that could be nil as in the case of ). For instance, in , a scal import duty was levied in all goods and additional imports duties were charged on goods originating outside the French empire. Moreover, in the , unlike the British empire, protection for agricultural prod- ucts was particularly draconian. An elaborate system of commodity agreements and of purchasing arrangements for agricultural products - the major production of French colonies - guaranteed that French Community products always had a preferred market in France (i.e. even if France imported duty free from the rest of the world) where they were sold at higher than world price. For instance, bananas, cocoa, coee and citrus fruit enjoyed taris and quotas preferences in France; cereals and oilseeds a guaranteed market with prices above world level; cotton and other bres were protected by licensing and by a purchasing cartel and were heavily subsided. Lastly, as regards import licensing, and given the weakness of the Franc, restrictions on non- franc imports were more severe than restrictions on non sterling imports.5 As a result, in 1956, France provided 71% of the French Community imports and took 69% of their exports (The Economist Intelligence Unit, 1960). A clear conclusion emerges from this comparison: whereas the Commonwealth system was quite loose, the French one seemed strong enough to have distorted what might be considered as the normal pattern of trade within the French empire. Thus, independence is very likely to have a stronger inuence on trade of former French colonies.6 One can think that the Yaounde Convention (1963), which aimed at enlarging the trade preferences prevailing in the French Community to all European Economic Community (EEC) members, may have softened the eects of independence for French colonies attaining independence in the 1960s.7 However, the Community trade preferences

4Except Equatorial French Africa, , and obliged by international con- ventions to not introduce preferential regime and Morocco. 5If quota restrictions tended to promote trade between Britain and the Commonwealth in the early postwar years, the preferential element progressively disappeared. Indeed, in the 1950s, many British colonies stopped doing distinction, in their import licensing policy, between sterling and other soft currencies, dividing the world between the Dollar area and the rest. 6Furthermore, note that, contrary to Britain, France did not generally adopted a policy of peaceful disengagement from its colonies. 7The Yaounde convention was signed between the EEC and 18 African and Malagasy States, named Associated States, including Burundi, Cameron, Congo (), Congo (Kinshasa), Côte-d'Ivoire, , , Burkina-Faso, Madagascar, , , , Rwanda, Sene- gal, Somalia, and Togo.

5 which should have favored exports of tropical foodstus of the Associated States to the EEC remained very low. The major exports products of the Associated States, such as banana or rice, were either excluded from preferential treatment or subject to quantitative restrictions (IEDES, 1969). A further implication of these colonial trade arrangements is that independence may have a larger impact on exports of former British and French colonies, as compared to imports. Indeed, before independence, Britain gave more preferences to its colonies than it received in return. In- deed, Glickman (1947) argues that the Ottawa preferential scheme mainly resulted in the increase of the exports of the British empire to Britain. In the case of former French colonies, independence, concomitantly with the creation of the EEC, tended to limit exports to France and to the EEC, whereas imports from France remained exempt from customs duties. In addition, the composition of imports and exports of former colonies may have played an important role. Former colonies exported mainly homogeneous primary products to their metropole, whereas they imported manu- factured goods from their colonial power. These imports may ultimately reect the slowly changing preference patterns of the whole society (Kleiman, 1976). Consequently, one can think that, after independence, imports may have been reoriented less easily than exports. The following section explains in details our data and empirical strategy to assess the eects of independence on exports and imports of former colonies.

3 Data and empirical model

3.1 Trade data

In our sample we have data for bilateral trade of 71 reporting countries or of countries with 190 partner countries. These reporting countries or federations of countries include 13 for- mer French colonies, 34 former British colonies, 11 former colonies from other colonizers8 and 13 countries that became independent before 1945 or that have never been colonized (see Table 4 in appendix). The sample ranges from 1948 to 2007. As mentioned in the introduction, the main database recording bilateral trade for a long period of time is the International Monetary Fund's Direction of Trade Statistics (DOTS). Even if it provides data on bilateral trade for a very large number of countries or entities since 1948, it has two major drawbacks. First, trade of some former colonies is not recorded in the DOTS before their independence. This is especially the case for former French and Belgian colonies. For instance,

8Our sample includes former colonies from Belgium (Burundi, Congo and Rwanda), from the Netherlands (Indonesia, Surinam, the Netherlands Antilles) and from Portugal (Angola, Cape Verde, Bissau, Mozambique, Sao Tome and Principe).

6 only a few former French colonies have their trade recorded in the DOTS since 1948 (Cameroon, Madagascar, Algeria, Morocco, Tunisia, Vietnam, and Cambodia). Most trade ows of the French Central and West African colonies are simply not included in the DOTS database even as part of the French imports or exports (Lavallée and Vicard, 2010). In contrast, trade of other former colonies (in particular the British ones) are much more precisely recorded. Indeed, the DOTS dataset reports pre-independence trade data for all former British colonies, except for a few ones (Botswana, Kiribati, , Lesotho, Malawi, Swaziland and Tonga).9 Second, when trade of former colonies is recorded, it is often recorded as zero. As noticed by Head et al. (2010), trade between France and Vietnam, Laos and Cambodia is recorded as zero in the DOTS between 1948 and 1953 or 1954 (the year of independence) and then becomes positive. These missing and `fake' zeros may create serious estimation bias tending to overestimate independence eects. In order to improve the coverage of these data, we use mirror data. The DOTS database reports two values for the same ow (imports of country A from country B and exports of B to A). When imports data are missing or recorded as zero, we substitute imports data with the reverse ow (exports), whenever available. As in Head et al. (2010), we add 10% to the export ow to adjust for the fact that exports are reported FOB and imports are recorded CIF. Secondly, we complement the DOTS data with bilateral trade data during the colonial era of former French colonies coming from three main sources:

• Commerce extérieur des Etats d'Afrique et de Madagascar de 1949 à 1960, Institut National de la Statistique et des Etudes Economiques (INSEE);

• Annuaire statistique de l'Union Française d'Outre Mer, Ministère de la France d'Outre-Mer (1938-1949);

• Annuaire statistique des Territoires d'Outre Mer, INSEE (1959, 1960, 1961);

During the colonial era, trade was generally reported for federations of colonial possessions, like the (Afrique Equatoriale Française) which contained four territories: Gabon, Middle Congo (now the ), Oubangui-Chari (now the ) and Chad (see Table 4 in appendix). Thus, for these federations of colonies, we aggre- gated data for post-colonial trade (coming from the DOTS) and for every determinant of trade on the entire period of estimation.10 For former French colonies, there are 2,088 pre-independence (non-zero) imports ows in the DOTS database. As mentioned above, these DOTS observations concern only a few colonies. With

9The DOTS dataset also reports some pre-independence trade data for only one Belgian former colony (the Democratic Republic of Congo), but for all former Dutch and Portuguese colonies included in our sample. 10For monadic determinants (such as population or GDP per capita), we simply computed the sum on individual countries belonging to the .

7 new colonial data, we add 1,603 non-zero observations (some of which are already recorded in the DOTS database) and 998 pre-independence new observations (which are never recorded in the DOTS database), counting for 31% of the total of pre-independence imports data (3,211) of former French colonies. For exports, we add 1,274 pre-independence (non-zero) observations and 422 pre- independence new observations (which are never recorded in the DOTS database), representing about 16% of the total of pre-independence exports data (2,712) of former French colonies. Thus, for some countries or federation of countries not recorded in the DOTS data such as the or French West Africa, adding colonial data provides a clear improvement. This allows to estimate more accurately the impact of independence on trade for former French colonies and to compare independence eects across former colonial empires.

3.2 Baseline model and estimated equation

To investigate the role of independence on bilateral trade of former colonies, we use a gravity model.

The gravity model relates bilateral trade, Mijt, (e.g. imports) between country i and country j at time t, to their economic sizes (Yit and Yjt), bilateral trade costs (τijt) and multilateral trade resistances (Pit and Pjt). The gravity equation can be written as:

µ ¶1−σ YitYjt τijt Mijt = , (1) Ywt PitPjt where Ywt is the nominal world income and σ > 1 the elasticity of substitution between all goods.

Trade costs (τijt) are generally modeled as a function of some observable factors, including bilateral distance between trade partners, the existence of a common border or a common language, or regional trade agreements (RTA). We also introduce a set of indicators (Indep) related to past colonial ties.

γ1 γ2 γ3 γ4 γ5 (2) τijt = distij × exp (borderij ) × exp (langij ) × exp (RTAijt) × exp (Indepijt) ,

In our basic specication, we break up the set of indicators related to past colonial ties into three dummy variables:

Indepijt = {Indep_COLijt, Indep_SIBijt, Indep_ROWijt}

The rst dummy variable is equal to 1 for relationships between former colonies and their colonizer for each year since independence and 0 otherwise. The second one (Indep_SIBijt) is equal to 1 for relationships between former colonies (siblings) for each year since independence and

8 11 0 otherwise and the third one (Indep_ROWijt) is equal to 1 for relationships between former colonies and the rest of the world for each year since independence and 0 otherwise. As argued in section 2, independence may increase bilateral trade costs with the former col- onizer and other countries in the colonial empire (siblings). As a consequence, we expect inde- pendence to decrease trade with the former colonizer and other former colonies from the same empire. We also expect independence to increase trade of former colonies with the rest of the world since increasing trade costs with the former colonial power and the other former colonies may raise multilateral resistance indices (see Anderson and van Wincoop, 2003). However, we also expect dierent independence eects depending on the identity of the colo- nial power. Indeed, colonial powers implemented dierent colonial trade policies in their colonies (see section 2). To investigate the impact of independence on trade according to the colonial power, we break up our set of indicators related to past colonial ties (Indep_COL, Indep_SIB, Indep_ROW ) according to the metropole (France, Great Britain and other colonial empires, i.e. Portugal, Belgium and the Netherlands). For instance, the dummy Indep_FRA takes the value 1 for ows between France and its former colonies for each year since independence. The

Indep_SIB_FRA and Indep_ROW _FRA dummies take the value 1 for trade between a former French colony and, respectively, another former French colony (sibling) after independence and the rest of the world after independence. Replacing the trade cost factor in equation (1), we obtain the estimated equation in its multi- plicative form :

β1 β2 β3 β4 Mijt = GDP Capit × P opit × GDP Capjt × P opjt

× exp (β5RTAijt + β6Indepijt + αij + λt) × ²ijt, (3)

We proxy countries' economic size (Y ) by GDP per capita and population to account for size and development eects. Bilateral time-invariant factors aecting trade, such as bilateral distance, common language or common border, are accounted by bilateral xed eects (αij ). In our basic regression model, bilateral xed eects also capture multilateral resistance terms (see below). All variables and data sources are described in appendix (Table 5). Our empirical model also includes time dummies (λt) which control for the general evolution of trade.

11 We adopt a `restricted' denition by granting that the Indep_SIBijt is equal to one when the two trade partners obtain independence. For instance, for the trade relationship between two former French colonies, Cameroon (independent in 1960) and Djibouti (independent in 1977), the

Indep_SIBijt dummy will be equal to one from 1977 onwards.

9 3.3 Estimation issues

Anderson and van Wincoop (2003) argue that multilateral resistance terms are crucial determinants of trade. In panel empirical analysis, these multilateral resistance indices are generally taken into account by country-year xed eects. However, this method is computationally burdensome and even impossible to apply in the case of large datasets including many countries and years. In our case, this method would imply adding more than 15,000 country-year dummies, which is practically unfeasible.12 We therefore adopt another solution which consists in using the xed eects (within) method of estimation. While this solution is not fully satisfactory notably because it amounts to consider that multilateral resistance terms are time-invariant, it has several advantages. It exploits the time series properties of the data. The eect of independence is thus estimated by comparing, within each pair of countries, the evolution of trade before and after independence. Moreover, it allows to control for every unobservable time-invariant country and country-pair characteristics that aect trade and that are potentially correlated with other determinants of trade. A second estimation issue relates to potential simultaneity in the relationship between indepen- dence and trade. Indeed, it may be the case that a colonizer chooses to give independence because it does not expect any further gain from trade with its former colonies. In this case, the traditional estimators, such as the OLS estimator or even the xed eects estimator, are biased. However, we argue that simultaneity is quite unlikely. The decolonization process is more related to politi- cal issues than to strictly economic and international trade issues. Indeed, the importance of the colonies in the trade of their metropole has been increasing rather than decreasing in the three or four decades before independence. Kleiman (1976) shows that the share of colonies in metropolitan countries' trade has doubled or even trebled between the late 1920's and the mid-1950's. Moreover, for the colonial powers, colonies, with a few exceptions (notably Algeria, Angola and Mozambique), accounted for a small part of their total trade (about 10%). Thus, any potential gains of colonial trade could have been of only limited signicance to them (Kleiman, 1976, p. 478). A third issue relates to the presence of zero trade in the dataset. Indeed, the most standard approach consists in estimating the gravity model in a logarithmic form, which amounts to drop zero values of the dependent variable (i.e. trade). In our dataset, we have 802,852 observations in total, of which 207,062 correspond to zero trade for imports (26%) and 223,932 for exports (28%). There are several ways to handle this problem. The rst one is to simply drop the zero trade observations. However, this method will yield biased estimates if the zeros are not randomly distributed, which is quite likely.13 The second one is to use a Tobit estimator. However, this

12Another possible method to account for multilateral resistance indices is to use the `tetrad' approach of Head et al. (2010), but the results are sensitive to the choice of the reference countries. 13Zero trade is more likely to occur for instance for small and distant countries (see Santos Silva and Tenreyro, 2006).

10 method is highly sensitive to the trade value used as the left censor value (see Head et al, 2010). A third solution consists in using a Poisson quasi-maximum likelihood (PQML) estimator. The PQML estimator incorporates the zeros and is robust to dierent patterns of heteroskedasticity (see Santos Silva and Tenreyro, 2006).14 These arguments extend to panel data. Among panel data models, the xed eects Poisson (FEP) estimator has strong robustness properties. In particular, it allows for arbitrary dependence between the xed eects and the explanatory variables, as in the linear model. The only assumption required for the estimator to be consistent concerns the conditional mean of the dependent variable (see Wooldridge, 1999 and Wooldridge, 2002, ch. 19). Numerous recent papers provide supporting evidence for the PQML estimator or the xed eects Poisson estimator (e.g. Siliverstovs and Schu- macher, 2009; Westerlund and Wilhelmsson, 2011). Therefore, in our main empirical estimations, we use the xed eects Poisson estimator. Note that this estimator may also produce biased esti- mates if the zeros are not `true' zeros (i.e. if trade ows are incorrectly recorded as zero trade ows). Head et al. (2010) report several cases in the DOTS database where there are zeros which should be indeed coded as missing. Thanks to the original data we gathered, we are able to solve partly this issue since we can distinguish between `true' zeros and missing trade for former French colonies.

In the next sections, we present our empirical results. First, we estimate the average eect of independence on former colonies' trade (imports and exports) over the entire post-colonial period and compare the results for dierent empires (section 4). Second, we investigate the impact of independence over time (section 5). Finally, we explore the impact of independence on the extensive margin of trade (section 6).

4 The overall eect of independence on former colonies' trade

4.1 The average impact of independence

We rst estimate the average impact of independence on the whole post-colonial period without dierentiating among former colonial powers. Estimation results of equation (3) are reported in Table 1 for former colonies' imports (columns 1 to 3) and for former colonies' exports (columns 4 to 6). We report results using the simple xed eects (within) estimator which drops the zero

14Santos Silva and Tenreyro (2006) show that heteroskedasticity is quantitatively and qualita- tively important in the gravity equation, even when controlling for xed eects (p. 643) and that the log-linear gravity specication leads to inconsistent estimates in this case.

11 trade observations (columns 1 for imports and 4 for exports). We next turn to the xed eects Poisson estimator without and with the zero trade values (resp. columns 2 and 5 and columns 3 and 6). There are some notable dierences between the within estimates and the xed eects Poisson estimates. For instance, using the within estimator, we nd that independence reduces trade with other former colonies (siblings) on average over the whole post-colonial period. When we use the Poisson estimator (with and without the zeros), the negative overall impact of independence on trade with siblings disappears. More generally, estimated coecients are quite similar using the positive trade sample and using the whole sample. This suggests that heteroskedasticity is potentially more problematic than truncation. This is in line with the results obtained in other papers using the Poisson estimator (see Santos Silva and Tenreyro, 2006). Therefore, we will concentrate on the interpretation of the coecients using the Poisson estimator on the sample including zero trade (columns 3 and 6). On the whole, our empirical model seems to work reasonably well. All control variables are signicant and have the expected sign. Results depicted in column 3 of Table 1 show that indepen- dence aects former colonies' imports patterns. We nd that independence implies a substitution of imports from the rest of the world for imports from the metropole. More precisely, independence reduces imports from the metropole by 36%[= (exp(−0.46)−1)∗100]15 and increases imports from the rest of the world by 118%[= (exp(0.78) − 1) ∗ 100] on average over the whole post-colonial period.16 Moreover, we do not nd any global impact of independence on imports from other former colonies (siblings). In order to check whether the inclusion of new pre-independence data aect our results, we estimate the same set of equations with the DOTS data only. Estimation results reported in Table 6 in appendix reveal that the eects of independence is generally lower when including new colonial data rather than the DOTS dataset only.17 This suggests that inde- pendence has a lower impact on former colonies' imports when accounting more exhaustively for pre-independence trade. Columns 4 to 6 of Table 1 report our estimation results for exports. Here again our empirical model seems to work reasonably well. All control variables are signicant and intuitively signed. It is worth noting that in all cases with the Poisson estimator, there is no signicant impact of the RTA variable on former colonies' exports. This may be due to the specialization of several former

15The average impact of independence on imports from the metropole is marginally signicant (p-value=5.3). 16Note that in the Poisson regression model, the interpretation of the estimated coecients is similar to the one in the standard log-linear model. Coecients on variables in logarithmic form (such as the GDP per capita or the population) can be interpreted as elasticities and the impact of dummy variables is measured as the exponential of the coecient minus one (see Winkelmann, 2003). 17For instance, the corresponding coecient for the Indep_COL variable is -0.55 (signicant at the 5% level) when estimated on the restricted DOTS dataset. Comprehensive results are reported in appendix.

12 Table 1: The eect of independence on former colonies' trade

Trade ows Imports Exports Method Within FEP T>0 FEP Within FEP T>0 FEP (1) (2) (3) (4) (5) (6) a a a a a a ln(GDP per capit) 1.01 0.99 1.03 1.67 1.05 1.09 (0.04) (0.08) (0.08) (0.05) (0.17) (0.16) a a a a a a ln(GDP per capjt) 1.37 1.25 1.25 1.02 0.95 1.05 (0.05) (0.11) (0.11) (0.06) (0.15) (0.15) a a a a a a ln(Populationit) 1.11 0.88 0.90 0.88 1.02 1.17 (0.11) (0.18) (0.18) (0.14) (0.26) (0.24) a a a a a a ln(Populationjt) 0.78 1.64 1.74 1.18 1.50 1.65 (0.08) (0.17) (0.17) (0.09) (0.27) (0.27) a b b a RTAijt 0.50 0.16 0.16 0.54 0.08 (0.06) (0.06) (0.06) (0.06) (0.08)

ACP_EUijt 0.09 -0.01 0.01 (0.07) (0.13) (0.12) a b c a a Indep_COLijt -0.36 -0.45 -0.46 -0.29 -0.72 -0.61 (0.14) (0.22) (0.24) (0.21) (0.17) (0.18) a a Indep_SIBijt -0.87 -0.01 0.06 -0.69 -0.07 -0.07 (0.12) (0.22) (0.20) (0.12) (0.14) (0.12) a a a a a a Indep_ROWijt 0.66 0.68 0.78 0.61 1.06 1.26 (0.06) (0.13) (0.13) (0.07) (0.25) (0.20) # of observations 174,800 174,536 291,176 166,093 165,852 293,086 Country-pair dummies yes yes yes yes yes yes Year dummies yes yes yes yes yes yes Notes: Within and Fixed eects Poisson estimations. Standard errors clustered at the country-pair level in parentheses. a, b and c denote signicance at the 1%, 5% and 10% level respectively. Year dummies estimates are not reported.

13 colonies in primary goods (which are mostly non-traded intra-regionally). We also include a specic dummy variable for the ACP (AfricaCaribbeanPacic) agreements. Indeed, since its creation, the European Union has oered trade preferences to a large number of African Caribbean and Pacic countries through the Cotonou agreement (previously Yaoundé and Lomé conventions). The ACP agreements are found to have no impact on former colonies' exports. This is not surprising since many papers argue that these agreements do not have any signicant impact on developing countries (Panagariya, 2002). Our results also indicate that independence aects former colonies' exports in the same direction as imports. Independence is found to reduce exports to the former colonizer, to increase exports to the rest of the world and to have no eect on exports to siblings on average over the whole post-colonial period. However, the coecient estimates are larger than those for imports.

More precisely, independence reduces exports to the metropole by 46%[= (exp(−0.61) − 1) ∗ 100] and increases exports to the rest of the world by 252%[= (exp(1.26) − 1) ∗ 100] (column 6). Even if such dierences are probably not statistically signicant, they suggest that for former colonies, exports are easier to redirect than imports.

4.2 Independence eects across former colonial powers

Are these independence eects common to all former empires? To answer this question, we break up our set of indicators related to past colonial ties according to the former colonial power as described in section 3.2. To conserve on space, we only report estimates for independence dummies in Table 2.18 Estimation results highlight that the eects of independence on bilateral imports and exports of former colonies are very sensitive to the identity of the former colonial power. For imports, we nd very dierent impacts of independence on post-colonial trade patterns across empires. As regards the French empire, independence reduces imports from France and from other former French colonies (siblings) by an amount that is statistically and economically signicant. Our results indicate that, following independence, imports from France and imports from other former French colonies have decreased respectively by 52% [= (exp(−0.74) − 1) ∗ 100] and 66% [= (exp(−1.09)−1)∗100] on average over the entire post-independence period (column 1). Independence is also found to increase imports of former French colonies from the rest of the world by 85% [= (exp(0.62) − 1) ∗ 100]. This suggests that, for these countries, independence implies a redirection of trade from France and its empire to the rest of the world. As regards the British empire, the consequences of independence on former colonies' imports are drastically dierent. First, we nd that independence has no eect on imports from Great Britain and from siblings on average over the post-colonial period. Indeed, the coecients of

18Coecient estimates for control variables are very similar to that of Table 1 and are available upon request.

14 Indep_GBR and Indep_SIB_GBR are not signicant at the 10% level. This result could be related to the mixed eects of the Ottawa agreement introducing imperial preference but only for a small subset of British colonies (see section 2). At last, as in the case of former French colonies, we nd that former British colonies' imports from the rest of the world increase dramatically after independence. The average estimated eect for these countries is 110% [= (exp(1.02) − 1) ∗ 100]. For the other empires (Belgian, Portuguese and Dutch empires), we nd no impact of inde- pendence on imports of former colonies from their colonizer, an increase in imports from the rest of the world and a (large) decrease in imports from siblings. However, we have to interpret these results with caution because the other empires' group gather very dierent situations and represent relatively few observations.19 Estimations results for exports also exhibit large dierences across empires. For former French colonies, independence eects for exports are broadly similar to those found for imports. Indeed, our results indicate that, following independence, exports to France and exports to other former French colonies have decreased respectively by 45% [= (exp(−0.60) − 1) ∗ 100] and 67% [= (exp(−1.12) − 1) ∗ 100] on average over the entire post-independence period (column 2). Furthermore, following independence, exports of former French colonies to the rest of the world have been multiplied by more than 4 [= exp(1.52)]. Thus, for former French colonies, independence has also generated a redirection of exports from the former colonizer and from siblings to the rest of the world. The fact that the independence eect on exports to the rest of the world is greater than the corresponding eect on imports tends to conrm the hypothesis according to which exports are easier to reorient than imports. For former British colonies, we nd some evidence that independence leads to a decrease in exports to the metropole and an increase in exports to other (ROW) countries in average over the whole post-colonial period. The estimated eects are respectively -42%[= (exp(−0.55) − 1) ∗ 100] and +177% [= (exp(1.02)−1)∗100]. Thus, it seems that former British colonies have not imported less but have exported less to the UK after independence. This could be related to the fact that, in the imperial scheme, Britain gave more preferences to its colonies than it received in return (see section 2). As for imports, we also nd that independence does not aect exports of British colonies with their siblings. As regards other empires, independence seems to reduce exports to their metropole and to their former siblings and increase drastically exports to the rest of the world. But, here again these results should be interpreted cautiously. On the whole, our results indicate that, following independence, trade (imports and exports)

19In particular, the Indep_SIB_OTH is equal to one for only 974 observations (on a total of 802,852). The estimation results for this variable depend mainly on (few) former Portuguese colonies.

15 with the metropole and with siblings has decreased mainly for former French colonies (by about respectively 50% and 67% on average over the whole post-colonial period). For other empires, the evidence is less clear-cut. The specicity of independence eects on bilateral trade patterns in the case of the French empire is consistent with our previous discussion on the imperial preferences systems implemented by the main colonizers. Finally, for every empire, we also show that trade (imports and exports) with other (ROW) countries has strongly increased following independence. This result contrasts sharply with the ndings of Head et al. (2010) who show that independence has reduced trade of former colonies with the rest of the world. The increase in trade with ROW countries could be related to the intensive margin (value of trade per product and/or partner) and the extensive margin of trade (number of products and/or partners). In particular, the end of preferential trade policies related to the colonial rule could have induced newly independent countries to diversify their sources and export markets. We further investigate this issue in section 6.

4.3 Robustness analysis

In our regression analysis, we control for every time-independent factors using country-pair xed eects. However, our independence estimates may still be biased due to omitted variable bias. The reduction in colonial trade after independence may be due to other (time-dependent) factors not related to the end of the colonial empire. We undertake several strategies to check the robustness of our results in Table 3. First, we drop the countries that became independent before 1945 or that have never been colonized (column 1 for imports and 5 for exports). In this case, our control group only includes countries before their independence. Thus, our estimates strictly compare trade after independence to trade before independence for each pair of countries. Second, we try to capture the specicities of the French decolonization process through extra control variables. Independence from France is characterized both by the continuation of some form of integration in West and through economic and monetary arrangements and by a more clear-cut break-up in other former French colonies through bloody independence wars.20 With our xed eects method of estimation, the specic currency arrangements among CFA countries (monetary unions with a xed exchange rate against the , now the euro) are captured through bilateral xed eects because they exist from the beginning of our period of estimation.21 In columns (2) and (6), we distinguish between peaceful and hostile independence events (see also Head et al., 2010). More precisely, we compute a dummy variable for hostile independence events

20In contrast, decolonization in the former British empire generally proceeded more smoothly. 21In unreported robustness checks, we also tried to control for the specic impact of the West African Economic and Monetary Union (WAEMU) and the Economic and Monetary Community of Central Africa (CEMAC), that entered into force in 2000 and 1999 respectively, but these two additional dummy variables (WAEMU and CEMAC) were not statistically signicant.

16 Table 2: The eect of independence on former colonies' trade by empire

Imports Exports (1) (3) a a Indep_FRAijt -0.74 -0.60 (0.27) (0.19) a a Indep_SIB_FRAijt -1.09 -1.12 (0.27) (0.31) b a Indep_ROW_FRAijt 0.62 1.52 (0.26) (0.25) b Indep_GBRijt -0.07 -0.55 (0.20) (0.28)

Indep_SIB_GBRijt 0.09 -0.05 (0.21) (0.12) a a Indep_ROW_GBRijt 1.02 1.02 (0.15) (0.21) c Indep_OTHijt -0.33 -0.80 (0.43) (0.44) a a Indep_SIB_OTHijt -2.77 -2.80 (0.43) (0.44) a a Indep_ROW_OTHijt 0.62 1.38 (0.17) (0.37) # of observations 291 176 293 086 Country-pair dummies yes yes Year dummies yes yes Notes: Fixed eects Poisson estimations. Standard errors clustered at the country-pair level in parentheses. Other coecients (for the control variables) are not reported. a, b and c denote signicance at the 1%, 5% and 10% level respectively. Year dummies estimates are not reported.

17 (Algeria, Indochina and Syria) and interact the Indep_FRA variable with this dummy. Finally, we introduce two additional variables inspired from Berger et al. (2011) to account for a potential omitted variable bias. We introduce a ve year trend in the dependent variable (columns 3 and 7) and we control for a dummy variable that is equal to one if the observation is between 4 years before independence and the year of independence (columns 4 and 8). In all cases, we obtain independence estimates that are very similar to our baseline results. For instance, as in Head et al. (2010), we nd that both hostile and more peaceful independence events lead to a decline in trade (imports and exports) with France and that this decline is larger in the rst case (columns 2 and 6). The extra variable that accounts for the trend in the dependent variable is positive and signicant (columns 3 and 7), indicating that countries that traded more with one another continue doing so. However, this does not modify the independence eects which are even larger in some cases as compared to the baseline estimates of Table 2. Thus, even if we control for the trend in trade, we still nd that independence decreases trade of former French colonies with the metropole and with other former colonies and increases trade with the rest of the world. In the following section, we investigate in more details the eects of independence over time.

18 Table 3: Robustness Analysis Trade ows Imports Exports Ex-col War 5-y trend Pre-indep Ex-col War 5-y trend Pre-indep (1) (2) (3) (4) (5) (6) (7) (8) b b b b c b a b Indep_FRAijt -0.60 -0.35 -0.77 -0.68 -0.46 -0.42 -0.69 -0.50 (0.29) (0.18) (0.30) (0.26) (0.25) (0.20) (0.17) (0.22) a a a a a a a a Indep_SIB_FRAijt -0.83 -0.71 -1.42 -1.08 -1.00 -1.15 -1.55 -1.11 (0.27) (0.22) (0.39) (0.27) (0.34) (0.42) (0.43) (0.31) a b b b a a a a Indep_ROW_FRAijt 0.77 0.59 1.32 0.63 1.61 1.07 1.55 1.52 (0.25) (0.28) (0.17) (0.26) (0.26) (0.18) (0.26) (0.25) c c b Indep_GBRijt 0.01 -0.07 -0.18 -0.01 -0.51 -0.55 -0.84 -0.46 (0.20) (0.20) (0.23) (0.22) (0.27) (0.28) (0.33) (0.30)

Indep_SIB_GBRijt 0.16 0.09 0.12 0.09 -0.08 -0.05 0.02 -0.05 (0.20) (0.21) (0.22) (0.21) (0.12) (0.12) (0.12) (0.12) a a a a a a b a 19 Indep_ROW_GBRijt 1.16 1.05 0.72 1.05 1.05 1.02 0.73 1.02 (0.14) (0.15) (0.17) (0.15) (0.23) (0.21) (0.33) (0.21) c c Indep_OTHijt -0.09 -0.32 -0.19 -0.27 -0.43 -0.79 -0.91 -0.70 (0.45) (0.43) (0.43) (0.43) (0.46) (0.44) (0.47) (0.44) a a a a a a a a Indep_SIB_OTHijt -2.52 -2.77 -3.01 -2.77 -2.57 -2.80 -3.31 -2.80 (0.45) (0.43) (0.25) (0.43) (0.43) (0.44) (0.28) (0.44) a a a a a a a a Indep_ROW_OTHijt 0.89 0.63 0.75 0.62 1.73 1.38 1.30 1.38 (0.18) (0.17) (0.17) (0.17) (0.34) (0.37) (0.40) (0.37) a a WIndep_FRAijt -0.95 -0.45 (0.15) (0.17) a WIndep_SIB_FRAijt -0.79 0.11 (0.39) (0.43) b WIndep_ROW_FRAijt 0.09 1.07 (0.55) (0.52) a a 5-year trendijt 0.18 0.13 (0.01) (0.01)

Pre-indep dummyijt 0.14 0.20 (0.09) (0.15) # of observations 232 952 291 176 126 801 291 176 227 627 293 086 117 954 293 086 Country-pair dummies yes yes yes yes yes yes yes yes Year dummies yes yes yes yes yes yes yes yes Notes: Fixed eects Poisson estimations. Standard errors clustered at the country-pair level in parentheses. Other coecients (for the control variables) are not reported. a, b and c denote signicance at the 1%, 5% and 10% level respectively. Year dummies estimates are not reported. 5 The eect of independence over time

5.1 The timing of overall independence eects

So as to investigate the impact of independence over time, we break up our set of indicators related to past colonial ties (Indep_COL, Indep_SIB, Indep_ROW ) into dummy variables denoting the number of years since at least one trade partner is independent (or the two partners in the case of siblings), up to a maximum we set at 45. For instance, the dummies Indep_COL1 to

Indep_COL45 take the value 1 for ows between the colonial power and its former colonies for each year (1 to 45) since independence. Figure 1 portrays the average eects of independence on imports and exports over time. The rst conclusion that emerges from this gure is that the eects of independence are progressive. This is particularly true for trade with the metropole. Whereas independence has no signicant impact on imports from the former metropole in the rst two decades following independence; after 30 years of independence the accumulated trade erosion is large and signicant: a former colony imports 53% less from its former metropole. These results are in line with Head et al. (2010). Nonetheless, it is worth noting that the progressive impact of independence is also found for trade with ROW and exports to siblings. Although signicant in the rst years following independence, the independence eects on imports from and exports to third (ROW) countries keep on increasing in the rst two decades after independence. After 20 years of independence, a former colony imports almost 3 times more from the rest of the world and exports 4 times more to these countries. This gure also yields a transversal nding : independence has more intense and more rapid eects on exports than on imports whatever the status of the trade partner, i.e. former metropole, sibling or ROW. Independence has signicant eects on trade with the former metropole 25 years after independence for imports and only 12 years after independence for exports. In the same vein, whereas independence reduces lately imports from siblings, it decreases exports to siblings almost immediately. The fact that independence impacts more heavily and more rapidly exports than imports conrms our previous results (see section 4). This close inspection of the timing of independence eects shows that independence inuences post-colonial exports and imports dierently. In the following section we will investigate whether the timing of independence eects dier for former British and French colonies.

20 Figure 1: Eects of independence over time

On imports On exports 2.00 2.00 1.00 1.00 0.00 0.00 −1.00 −1.00 −2.00 −2.00 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 # of years since independence # of years since independence

Note: Solid lines represent the effects of independence on trade with metropole, dash lines on trade with siblings, dash−dot lines on trade with RoW.

5.2 The timing of independence eects for former French and British colonies

Figures 2 and 3 display the eects of independence over time according to the former colonial power. For the reasons exposed in section 4.2, our comments will focus mainly on former British and French colonies.22 The results depicted in these gures are consistent with the ndings presented in section 4: independence eects are mainly driven by former French colonies, except as regards trade with ROW. Indeed, independence has no signicant eect on former British colonies' imports from their former metropole and siblings. On the contrary, independence reduces progressively former French colonies' imports from France and their siblings. Although the impact of independence on exports are a bit trickier, Figure 3 shows that, compared to former British colonies, independence reduces more intensely and more rapidly former French colonies' exports to France and siblings.

22Regressions also include independence indicators for the other former empires but for space and clarity considerations we do not present their coecient estimates. They are available upon request.

21 Figure 2: Eects of independence on imports for former French and British colonies

From Metropole From siblings 1.00 0.50 0.00 0.00 −0.50 −1.00 −1.00 −1.50 −2.00 −2.00 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 # of years since independence # of years since independence

From RoW 2.00 1.00 0.00 −1.00 0 5 10 15 20 25 30 35 40 45 # of years since independence

Note: Solid lines represent the effects of independence for former French colonies, the dash lines for Bristish colonies

Another conclusion that emerges from Figures 2 and 3 but also from Figure 1 is that inde- pendence increases instantaneously trade with ROW. This is particularly pronounced for former British colonies' exports. There may be two reasons for this nding. First, as exposed by Head et al. (2010), DOTS data have a fake zero problem. They report some trade ows as zero whereas they should be coded as missing. This is particularly the case for pre-independence trade data. In our data set, we partly solve this problem in the case of former French colonies (see section 3.1). Nevertheless, the remaining incorrect zero (British) information can lead to overestimate the eect of independence. Indeed, for recording reasons, trade ows may jump from zero to a positive value the years following independence. This is likely the case when estimating the eect of independence on trade with ROW due to both trade partners' poor statistical capacities. In order to test whether our results are due to fake zeros, we drop observations for which trade ows are recorded as zeros

22 Figure 3: Eects of independence on exports for former French and British colonies

To metropole To siblings 1.00 1.00 0.00 0.00 −1.00 −1.00 −2.00 −2.00 −3.00 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 # of years since independence # of years since independence

To RoW 3.00 2.00 1.00 0.00 0 5 10 15 20 25 30 35 40 45 # of years since independence

Note: solid lines represent the effects of independence for former French colonies, dash lines for former British colonies

before independence. Estimations on this restricted sample lead to the same result: independence increases immediately exports to ROW and this result is driven by former British colonies.23 The second possible reason lies in the anticipation of independence. Exporters in former colonies may have anticipated independence and progressively diversied their exports. This scenario is more likely in the case of the British empire since Britain, contrary to France, adopted a policy of peace- ful disengagement from its colonies. In order to test this hypothesis, we introduce dummy variables denoting the number of years before and after independence. Our results reported in Figure 6 (in appendix) suggest an anticipation of independence. Indeed, the dummy variables indicating four to one year prior to independence are economically and statistically signicant. The decomposition of this eect by empire shows that this eect is mainly due to the British empire.

23Estimation results are not reported in the paper but are available upon request.

23 6 The consequences of independence on the exten- sive margin of trade

In the previous sections we show that independence increases former colonies' trade with third countries. This suggests that independence may increase trade in its geographical extensive margin. In other words, the end of the colonial rule may incite newly independent countries to initiate trading relationships with new partners. Indeed, empirical evidence suggests that geographical trade diversication is important, in particular for developing countries (see e.g. Evenett and Venables, 2002) and that reducing taris and export costs increase the geographical extensive margin of trade (Shepherd, 2010). So as to estimate the eects of independence on the geographical extensive margin of trade, we compute a variable that counts, for each year, the number of countries from which a reporting country imports (respectively exports) as in Shepherd (2010). Then, we estimate the impact of independence on the extensive margin of trade using a Fixed eects Poisson estimator. We control for countries' size and level of development using their GDP per capita and population, all time invariant trade costs (e.g. distance) being captured in country xed eects. We also include year xed eects to control for the global evolution of geographical diversication in our sample of countries.24 Figures 4 and 5 portray the eects of independence over time on the extensive margins of imports and exports respectively.25 For former British colonies, estimation reveals that independence has a large impact on the extensive margin of imports and exports. The geographical diversication of trade takes place mainly in the rst fteen years after independence. For former French colonies, the evidence supporting the increase in geographical diversication is less conclusive. We nd that independence has a smaller (and marginally signicant) eect on the number of trade partners for former colonies' imports and a non signicant eect for exports.

7 Conclusion

This paper investigates the consequences of independence on bilateral trade of former colonies. We obtain more accurate results as compared to the one obtained in the literature thanks to a new dataset on colonial trade of former French colonies from 1948 to their independence and by estimating independence eects for each direction of trade ows (imports and exports) and for each

24The main determinants of the geographical extensive margin of trade include market size, development level, trade costs (distance, taris) and export costs (such as border formalities) in the exporting country (Shepherd, 2010). Here, we cannot introduce taris or export costs measures since data for these variables are typically not available for a long time span. 25Estimation tables corresponding to this gure are available upon request.

24 Figure 4: Eects of independence on the extensive margin of imports

Whole effects Effects by Empire 1.00 1.50 1.00 Former Bristish colonies 0.50 0.50

Former French colonies 0.00 0.00 −0.50 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 # of years since independence # of years since independence

Figure 5: Eects of independence on the extensive margin of exports

Whole effects Effects by Empire 1.00 1.50 1.00

0.50 Former Bristish colonies 0.50

Former French colonies 0.00 0.00 −0.50 −0.50 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 # of years since independence # of years since independence

former empire. Our study yields several sobering ndings. First, we nd that independence reduces trade (imports and exports) with the former metropole. However, this eect depends on the identity of the colonizer and is mainly driven by former French colonies. For these countries, trade (exports and imports) with the former metropole is reduced by about 50% on average over the whole post-

25 colonial period as compared to pre-independence levels. We also nd that former French colonies trade less with their siblings (other former French colonies) after independence (by about 67%). These eects are long-lasting but appear quite quickly after independence, in particular for exports. Trade erosion with the former metropole reaches its long-term value fteen years after independence for former colonies' exports and twenty ve years after independence for imports. This suggests that, after independence, imports may have been reoriented less easily than exports. We do not nd similar trade erosion with the former metropole or with siblings for former British colonies. This nding is consistent with historical evidence emphasizing that Britain was the least protectionist of the imperial powers. Thus, we interpret these results as oering an indirect evidence for the long-lasting inuence of colonial trade policies. Finally, we show that, after independence, trade (exports and imports) of all former colonies increase with the rest of the world (other countries not belonging to the same empire). Part of this increase in trade comes from the increase in the number of trade partners (geographical diversication of trade). Some of our results contrast with the ndings of Head et al. (2010) who use the DOTS database on a larger sample of countries. Further research could complement the DOTS dataset for former British colonies with data on pre-independence trade coming from colonial sources.

26 Appendix

Table 4: List of reporting countries

Former French colonies Former British Colonies Other former colonies Not colonized (in 1945) French Equatorial Africa Bahamas; Belize; Barba- Burundi; the Democratic Afghanistan; Cuba; Do- (Congo, Gabon, Chad, dos; Botswana; ; Republic of the Congo; minican Republic; Egypt; Central African Republic); Fiji; Ghana; Gambia; Rwanda; Netherlands An- Ethiopia; Haiti; Iran; French West Africa 1 Grenade; Guyana; In- tilles; Indonesia; Suri- Iraq; Lebanon; Liberia; (, Mali, Mau- dia; Jamaica; Jordan; nam; Angola; Cape Verde; Nepal; Oman; South ritania); French West Kenya; Kiribati; Kuwait; Guinea Bissau; Mozam- Africa Africa 2 (Niger, Benin, Saint Lucia; Sri Lanka; bique; Sao Tome and , Guinea); Lesotho; Myanmar; Principe Ivory Coast; Cameroon; Malawi; Malaysia; Nige- Côte française des Soma- ria; Pakistan; ; lis (Djibouti); Algeria; Solomon Islands; Sierra Indochina (Vietnam, Cam- Leone; Swaziland; Tonga; bodia, Laos); Morocco; Tanzania; Ugandan; Madagascar; Syria; Togo; and the Tunisia Grenadines; Zambia; Zimbabwe

Table 5: Data and variable denitions

Tijt Import and export data come from the IMF (DOTS database) and from other ocial French sources (see section 3.1). We converted colonial trade data recorded in French Franc in dollars using the exchange rate coming from the IMF (IFS database).

GDPCapit/jt; Current GDP per capita and population data come from the Historical Statistics for Popit/jt the World Economy provided by Angus Maddison (http://www.ggdc.net/maddison).

RTAijt The Regional Trade Agreement dummy is computed using informations from the WTO. It covers several free trade agreements or customs unions relevant for our countries of interest. For instance, the COMESA (Common Market for Eastern and Southern Africa) (in force since 1994), the ECOWAS (Economic Community of West African States) (in force since 1993), the CAEMC (Central African Economic and Monetary Community) (in force since 1999) and the PAFTA (Pan-Arab Free Trade Area) (in force since 1998).

27 Table 6: The eect of independence on former colonies' trade with the DOTS dataset

Trade ows Imports Exports (1) (2) (3) (4) a a a a ln(GDP per capit) 1.03 1.03 1.09 1.09 (0.08) (0.08) (0.16) (0.16) a a a a ln(GDP per capjt) 1.25 1.26 1.03 1.03 (0.11) (0.11) (0.15) (0.15) a a a a ln(Populationit) 0.90 0.87 1.18 1.20 (0.18) (0.18) (0.24) (0.24) a a a a ln(Populationjt) 1.74 1.75 1.62 1.63 (0.17) (0.17) (0.27) (0.27) b b RTAijt 0.16 0.16 0.08 0.08 (0.06) (0.06) (0.08) (0.08)

ACP_EUijt 0.01 0.01 (0.13) (0.13) b a Indep_COLijt -0.55 -0.57 (0.26) (0.19)

Indep_SIBijt 0.05 -0.07 (0.20) (0.12) a a Indep_ROWijt 0.90 1.27 (0.12) (0.20) a b Indep_FRAijt -1.04 -0.50 (0.20) (0.25) a a Indep_SIB_FRAijt -1.17 -1.17 (0.27) (0.31) a a Indep_ROW_FRAijt 1.17 1.59 (0.18) (0.26) b Indep_GBRijt -0.09 -0.55 (0.20) (0.28)

Indep_SIB_GBRijt 0.09 -0.05 (0.21) (0.12) a a Indep_ROW_GBRijt 1.02 1.01 (0.15) (0.21) c Indep_OTHijt -0.33 -0.80 (0.43) (0.43) a a Indep_SIB_OTHijt -2.77 -2.80 (0.43) (0.44) a a Indep_ROW_OTHijt 0.61 1.38 (0.17) (0.37) # of observations 289,634 289,634 292,319 292,319 Country-pair dummies yes yes yes yes Year dummies yes yes yes yes Notes: Fixed eects Poisson estimations. Standard errors clustered at the country-pair level in parentheses. a, b and c denote signif- icance at the 1%, 5% and 10% level respectively. Year dummies estimates are not reported.

28 Figure 6: Has independence been anticipated? Evidence from exports

Whole effect According to colonial power 4.00 3.00 3.00

Former French colonies 2.00 2.00

Former British colonies 1.00 1.00 0.00 0.00 −1.00 −10 0 10 20 30 40 −10 0 10 20 30 40 # of years before and after independence # of years before and after independence

References

Anderson, J.E. and E. van Wincoop, 2003. Gravity With Gravitas: A Solution to the Border Puzzle. American Economic Review 93(1), pp. 170-192. de Sousa, J. and Lochard, J. (2012). Trade and Colonial Status, Journal of African Economies, forthcoming. The Economist Intelligence Unit (1960). The Commonwealth and Europe. Published by The Economist Intelligence Unit, London, New York. The Economist Intelligence Unit (1962). Evenett, S, and A. Venables (2002). Export growth in developing countries: Market entry and bilateral trade ows. Working Paper. Fram, N. (2006). Decolonization, the Commonwealth, and British Trade, 1945-2004. Working Paper. Glick, R. and A.M. Taylor (2006). Collateral Damage: Trade Disruption and the Economic Impact of War. NBER Working Paper 11565. Glickman, D. (1947). The British Imperial Preference System. The Quarterly Journal of Eco- nomics, 61(3), pp. 439-470. Head, K., Mayer, T. and J. Ries (2010). The Erosion of Colonial Trade Linkages after Indepen- dence. Journal of International Economics 81(1), pp. 1-14.

29 IEDES (1969). Renouvellement de la Convention de Yaoundé. Tiers-Monde 10(37). pp. 213- 236.

Kleiman, E. (1976). Trade and the Decline of Colonialism. The Economic Journal 86, pp. 459-480.

Kleiman, E. (1977). Heirs to Colonial Trade. Journal of Development Economics 4, pp. 93-103.

Lavallée, E., Vicard, V. (2010). National Borders Matter...where one draws the line too. Banque de France Working Paper.

Mitchener, K.J. and M. Weidenmier (2008). Trade and Empire. The Economic Journal 118, pp. 18051834.

Panagariya, A. (2002). EU Preferential Trade Arrangements and Developing Countries. The World Economy, 25(10), pp. 1415-1432.

Rose, A. K. (2000). One money, one market: the eect of common currencies on trade. Economic Policy, 15(30), pp. 7-45.

Santos Silva, J.M.C. and S. Tenreyro (2006). The log of gravity. The Review of Economics and Statistics 88 (4), pp. 641-658.

Siliverstovs, B. and D. Schumacher (2009). Estimating gravity equations: to log or not to log?. Empirical Economics 36(3), pp. 645-669.

Shepherd, B. (2010). Geographical Diversication of Developing Country Exports. World De- velopment 38(9), pp. 12171228.

Westerlund J. and F. Wilhelmsson (2011). Estimating the gravity model without gravity using panel data. Applied Economics 43, pp. 641-649.

Winkelmann, R. (2003). Econometric analysis of count data, Springer.

Wooldridge J. (1999), Distribution-free estimation of some nonlinear panel data models. Journal of Econometrics 90, 77-97.

Wooldridge J. (2002), Econometric Analysis of Cross Section and Panel Data, Cambridge: MIT Press.

30