French Colonial Trade Patterns and European Settlements

Transcription

1 Comp Econ Stud (2018) 60: French Colonial Trade Patterns and European Settlements Tania El Kallab 1,2 Cristina Terra 2,3 Published online: 21 December 2017 Ó Association for Comparative Economic Studies 2017 Abstract We construct a new database relying on various primary historical sources containing information on the value of French sectoral trade between 1880 and 1913 in order to assess the contemporaneous effects of colonial European settlements on French trade patterns. Our empirical results show that French colonies with more European settlements traded more with France. The impact is stronger with respect to the of raw materials and of manufactured goods from France to their colonies, suggesting that those territories were a source of resources for France and a market for its products. European settlements in colonies other than the French ones did not impact the trade of those colonies with France. We also explore to what extent the impact of European settlements on trade was exerted through the channel of the institutions brought by the settlers. Separating the part of European settlements associated with institutions, we find that the settlements associated with stronger institutions in French colonies had a positive impact on trade with those colonies. No significant impact was found for non- French colonies. Keywords International trade Colonization Institutions JEL Classification F14 N53 N74 O5 & Tania El Kallab tania.kallab@essec.edu Cristina Terra cristina.terra@essec.edu Lebanese American University, Beirut, Lebanon ESSEC Business School, Cergy-Pontoise, France THEMA, Cergy-Pontoise, France

2 292 T. El Kallab, C. Terra Introduction Economic interests were the main driver of colonization, and trade was one of the main tools of colonial extraction. Mitchener and Weidenmier (2008) assess the impact of colonization on colonial trade flows by adding colonial status dummies to a standard gravity model. They find that trade is twice as large among countries that are part of an empire. We propose to deepen the understanding of the impact of colonialism on trade in three dimensions. First, we investigate whether the share of European settlers in the colonies affected that relation. The colonial dummies used by Mitchener and Weidenmier (2008) indicate whether a country is a colony, but it does not capture the extent and depth to which a country is colonized. The size of European settlements in the colonies can be taken as one measure of the intensity of the colonization, since settling in the colonies was a way to concretize it. Second, we examine the impact of colonization on trade of agricultural and of manufactured goods separately. Looking at sectoral trade, we inspect the pattern of trade between the colonizer and different groups of colonies. Third, we explore differences on the impact of colonialism on trade with the colonizer s own colonies and with colonies from other empires. More precisely, we examine the trade pattern of France with its own colonies and with colonies of other nations, using a new data set we constructed containing trade data of France with all its trading partners from 1880 to European powers were able to establish their colonial control by settling in the colonies and exploiting their resources through trade. This assertion is not new to the literature and can be traced back to Lenin and farther back to Hobson and Marxist thinkers who regarded trade as the primary cause of imperialist expansion (Kleiman 1976). French colonization offers a quite representative example: French colonizers settled in the form of military troops, imposed territorial control, and set up trade policies and preferential trade agreements in order to transfer resources from the colony to themselves and to secure favorable markets for their products (Crowder 1968). The presence or absence of European settlers affected the insertion of European trading companies and the type of institutions set up by the colonial administration. In some French colonies in Africa, for instance, where production was in the hands of domestic farmers, European settlers helped create large companies to lobby the colonial government and establish a controlled system of oligopolistic firms. In turn, the colonial administration generally supported the activity of trading companies by implementing coercive institutions (Hopkins 2014). In British colonies, production was often controlled by the European settlers who had a political influence over the colonial government. In that case, the cost of imposing extractive institutions was higher (Tadei 2017). Whether production was organized through small domestic farmers or plantation companies, trade revolved around the activity of European trading firms, whose relation with settlers was crucial. Notice that there are other features of colonization, such as the establishment of currency unions and trade agreements, that also impacted trade and are not necessarily related to the extension of colonial settlements. Trade agreements are

3 French Colonial Trade Patterns and European Settlements 293 likely to not only increase trade among partners to the agreement, but also to affect trade with other partners (Glick et al. 2017). For instance, France applied policies of tariff assimilation/ custom union 1 with its colonies to encourage intra-empire trade, while Great Britain applied an open-door policy. Such policies surely shaped trade patterns of France with British colonies and ought to be accounted for. We try to control for those variables in our empirical investigation. To carry out our empirical investigation, we constructed a new data set on French trade between 1880 and The data were collected from Tableau General du commerce which is the most complete and reliable database, for it is the official French customs record. 2 Our data set contains information on bilateral French and with all trade partners, disaggregated into three sectors: agricultural raw materials, raw materials necessary for industry, and manufactured goods. With data on trade by sector, we are able to examine the pattern of trade between France and the colonies. More specifically, we are able to identify whether the trade was directed toward primary or manufactured products. Using this new data set, we investigate the impact of European settlements on the trade pattern between France and its own colonies, and between France and non-french colonies. Once we establish this relation, we explore whether that impact is related to the institutions brought by the European settlers. Typically, colonies with more European settlers developed more European-style institutions with property rights, checks and balances and accountability (Acemoglu et al. 2001). Higher quality of institutions, in turn, may have had a positive impact on bilateral trade flows due to the lower transaction costs and higher level of trust they produce (Briant et al. 2009; Sousa and Lochard 2012; Groot et al. 2004; Linders 2004). A number of studies has investigated the impact of colonial status on both historical and current trade (Estevadeordal et al. 2002; Mitchener and Weidenmier 2008; Sousa and Lochard 2012), while others have examined the impact of independence on post-colonial trade (Lavallée and Lochard 2015; Head et al. 2010). Some of these studies look at the effect of colonization on bilateral trade (Mitchener and Weidenmier 2008; Rose 1999; Sousa and Lochard 2012) using colonial dummies as an explanatory variable of total bilateral trade, and they find that being colonized exerts a positive impact on total bilateral trade. In our empirical investigation, we add new insights by using European settlements in addition to colonial dummies. Furthermore, we investigate sectoral bilateral trade, instead of only total bilateral trade. To our knowledge, this is the first study to investigate the impact of colonial settlements on trade. We find that greater French settlements increased the overall French and with French colonies, and the impact is strongest with respect to the of raw materials. The British or other European settlements in their respective 1 Tariff assimilation is a regime where the tariff rates on goods are the same in the metropolis and the colony. 2 Two recent projects have collected bilateral trade data at the product level for the colonial period: the Montesquieu Dataset (Becuwe et al. 2015), from University of Bordeaux that is not as yet publicly available, and the RICardo project (Dedinger and Girard 2017).

4 294 T. El Kallab, C. Terra colonies had either a negative or a nonsignificant impact on the trade of those colonies with France. We then examine the impact of European settlements through colonial institutions. Using principal component analysis, we constructed a new variable capturing the common variability between the three variables capturing institutions, namely democracy, constraints on the executive, and the duration of colonization. To try to disentangle the part of European settlements corresponding to colonial institutions from the residual part corresponding to other channels, we then estimate the predicted value of European settlements explained by this variable that captures institutions. We find that the part of European settlements associated with better formal institutions has a positive impact on French of raw material and of manufactured goods to their own colonies. These results indicate that strong institutions facilitated the trade of France with their own colonies, and this pattern of trade is likely to be related to the comparative advantages of the trade partners. In non-french colonies, on the other hand, European settlements with better institutional quality do not impact trade between those colonies and France. This paper is organized as follows. The second section presents an overview of the literature examining the link between colonialism and trade. The third section presents the empirical model and the data, and the fourth section presents the baseline results along with the various specifications and robustness checks. The fifth section explores the channels through which settlement might have affected trade. The last section concludes. Colonization, Settlement, and Trade The economics and history literature suggest a variety of explanations for the impact of colonization on the patterns of trade during the period of imperialism. Betts (1961) and Kwon (2011) relate the expansion of imperial control to mercantile economic policies, which led to a demand for formal political control. Mitchener and Weidenmier (2008) argue that, prior to the Industrial Revolution, colonial acquisitions were continuously sought by imperial powers to complement their growing economies, which ultimately affected colonial trade. The authors find that belonging to an empire doubled trade relative to those countries that were not part of an empire. 3 These findings are consistent with previous literature suggesting that colonial domination increased the colonies trade with their metropolitan countries (Kleiman 1976; Bairoch 1999). In its extreme form, colonization reflects some form of forced trade, which implies some monopolization of colonial trade, forcing the colony s population to buy their for more and to sell their for less than going world prices (Kleiman 1976, p. 1). In a more moderate view, colonies are not entirely subjugated 3 Agricultural boomed in the Gold coast, including Senegal, Ivory Coast and other French colonies in West Africa; in return, these colonies began to import Europeans manufactured goods. In Indochina, also a French colony, the land under cultivation increased dramatically, allowing Indochina to become the third largest producer of rice in the world (Mitchener and Weidenmier 2008).

5 French Colonial Trade Patterns and European Settlements 295 to the colonizer and have some level of internal sovereignty, hampering the exploitation of the colony. In that case, a trade structure biased toward the metropolitan country is needed for the economic exploitation of colonial territories through trade (Kleiman 1976, p. 1). Colonization facilitates trade by using power to impose preferential trade policies, currency and custom unions (Crowder 1968; Estevadeordal et al. 2002; Ferguson and Schularick 2006; Lal 2004; Mitchener and Weidenmier 2008), as is the case of the French empire (Amin 1972; Lavallée and Lochard 2015). Egger et al. (2012) argue that migrants acquire economic, cultural, and institutional knowledge about both the home and the host markets, enabling them to mediate economic exchanges between those markets. In the case of colonies, European settlements deepened the colonizers insertion, enabling them to exploit resources more easily. Moreover, the longer the settlements lasted, the deeper were the roots of colonial institutions. Acemoglu et al. (2001) argue that, in viable areas, Europeans tended to settle, to build infrastructure, and to promote European-style institutions that have persisted until today; wherever they faced tropical soils and diseases, they confronted the high costs of cultivation, building, and trade, which demotivated the settlers, leaving them with only extractive institutions. In order to empirically investigate how European settlements affected the French trade patterns with the various colonial groups and to assess their importance on trade, we use an augmented gravity model, as explained in the next section. Empirical Model and Data Baseline Model The gravity model, the workhorse model for examining bilateral trade flows, is extensively used in the literature due to its good fit to the data. In its basic form, the model asserts that trade is greater between larger trade partners, whereas trade is smaller the greater the distance between them. In our empirical analysis, we use an augmented version of the gravity model, adding the share of European settlers in 1900 in the colonies or former colonies as an additional explanatory variable, as well as a number of additional variables likely to affect trade. More specifically, our baseline regression is based on the following gravity equation: LnðTradeÞ i ¼ b 0 þ X b j D ij þ c 0 ES i þ X c j D ij ES i þ ax i þ i ; j2j j2j ð1þ where subscript i represents the country France is trading with and j 2 J refers to the colonial group the country belongs to, within the set J = (Great Britain, other empires, former colonies). Since we take French colonies as the reference group, France is not included in the set J. We ran six different cross-section regressions, one for each of the following trade variables: total French, total French

6 296 T. El Kallab, C. Terra, and of agricultural goods, of raw materials for industry, and of manufactured goods. D ij are dummies that equal one when country i belongs to the colonial group j, and zero otherwise. These dummies indicate whether trade with France was, on average, different across groups of colonies, compared to French colonies. More specifically, the coefficient b 0 is the constant term associated with France s trade with French colonies, while the coefficients b j capture the average difference in France s trade with the group of colonies j compared to trade with their own colonies. ES i represents European settlements in 1900 in colony i. The interaction term between European settlements (ES) and colonial dummies allows us to examine how the impact of ES on trade differs across groups of colonies. c 0 and c j are, then, our main coefficients of interest. X i is a vector of control variables. We follow Mitchener and Weidenmier (2008) and use population (in logs) to stand for the mass in the gravity model, since we do not have GDP measures for 1900 for the colonies and former colonies. The other standard gravity variables we use are the distance between France and the trading partner (also in logs), and a dummy variable for being landlocked. Following earlier studies, we also incorporate historical institutional dummies: Four distinct dummies capture whether the countries were part of a formal or informal trade preference agreement with France or with non-french colonizers, or whether the countries were in a currency union with France or with other colonizers. We control for climate indicators (temperature and humidity). Those variables are particularly relevant because they affect agricultural productivity, which is a major sector in colonial trade. We also include a dummy capturing whether the partner country was at war, following Martin et al. (2008) who show this variable affects the volume of trade. Finally, we add a variable capturing the extent to which European languages are spoken in the country, since sharing a common language reduces informational, operational, and transaction costs, thus facilitating trade. Since most of our explanatory variables are time invariant, including our main variable of interest, European settlements, we estimate Eq. (1) using cross-section ordinary least-squares for the average of the trade values between 1897 and We incorporate a large number of control variables in order to capture cross-country differences and to mitigate the omitted variable problem. Data Trade Data To conduct our empirical investigation, we have constructed a new data set of bilateral and of France with all their trade partners from 1880 to We collected the data from several volumes of the Tableau General du commerce de la France and the Tableau decennal du commerce de la France. In our cross-section regression, we take the average of the trade values in Francs from 1897 to We chose that time frame in order to keep the period close to that of European settlements, which is measured around 1900.

7 French Colonial Trade Patterns and European Settlements 297 The data are disaggregated into the following sectors: agricultural raw materials, raw materials for industry and manufactured products. France s of agricultural goods consisted mostly of food. Raw materials for industry, included mining products such as gold, cobalt, phosphate, iron, wood and wool. Manufactured goods comprised goods such as machinery, tools, fabric and weapons. Notice that our database significantly improves upon historical trade data used in previous studies, for it includes data on and separately, as well as trade data disaggregated into sectors, as described in the previous paragraph. We have yearly data covering all French trading partners. This detailed and comprehensive data allows us to compare French trade with its own colonies to that with colonies from other empires as well as with former colonies. The trade data were originally collected at the aggregate level by colonial groups and geographic regions, whereas our control variables are assigned to countries according to their current geographic division. We have disaggregated the regions into countries to make the trade data compatible with the geographical units of the other variables. Taking into account, on a yearly basis, the political and historical events (i.e., colonization, independence, creation of a nation, new groups entering the sample), we were able to assign the appropriate countries to each colonial group, and we used the population size collected from Mitchell (2007a, b) and Maddison (2006) as weights to assign trade values for each country. Over the time frame we use in this paper, from 1897 to 1903, there was no change in the colonial status of the countries in our data set. Table 1 shows the list of the 91 countries in our data set, divided into 24 French colonies, 36 British colonies, 14 colonies of other empires and 17 former colonies, while Fig. 1 shows the world map containing all the countries included in our sample and how they were divided among the colonial powers. We are aware that the disaggregation of the geographical regions from the original data set into the current division of countries is likely to produce measurement errors. Were we to keep the trade data in their original regional division, we would have to aggregate the independent variables instead, given that the control variables we use in the gravity equation are assigned by country. To that end, we would have to make a number of different assumptions, depending on the variable, that would most probably lead to even more measurement errors. This is why we chose to disaggregate the trade data to make them compatible with the rest of our data. To verify the extent of the possible distortions produced by the disaggregation procedure, we compare our disaggregated data to the bilateral trade data from Mitchener and Weidenmier (2008), for the set of countries available in both datasets. The sources for their data are the Statistical Abstract for the United Kingdom, the Statistical Abstract for the Several British, Colonies, Possessions and Protectorates, and the Statistical Abstract for the Principal and other Foreign Countries. They also use the Tableau General du Commerce Exterieur for some additional data for French colonies. Since the two datasets use different currencies, instead of comparing the trade values, we compare France s trade with each partner country as a share of total trade. As can be seen in Tables 12 and 13 in Appendix

8 298 T. El Kallab, C. Terra Table 1 France s trade partners: colonies and former colonies French colonies British colonies Other colonies Former colonies Algeria Antigua et Barbuda Aruba Argentina Benin Australia Angola Bolivia Burkina Fasso Bangladesh Cameroon Brazil Cambodia Bahamas Cuba Canada Central African republic Barbados Western Sahara Chile Chad Botswana Guinea Bissau Colombia Congo Cyprus Equatorial Guinea Costa Rica French Guiana Egypt Mozambique Dominican Republic French Polynesia Fiji Namibia Ecuador Gabon Ghana Philippines Guatemala Guadeloupe Gibraltar Puerto Rico Haiti Guinea Gambia Sao Tome and Principe Honduras Ivory Coast Grenada Suriname Mexico Laos Guyana Togo Peru Madagascar India USA Mali Jamaica Uruguay Martinique Kenya Venezuela Mauritania Saint Lucia Morocco Malta New Hebrides (Vanuatu) Myanmar Niger Mauritius Reunion Malawi Tunisia Nigeria Vietnam New Zealand Pakistan Sudan Solomon islands Sierra Leone Somalia Trinidad and Tobago Tanzania Uganda British virgin Islands South Africa Zambia Zimbabwe 2, the import and export shares using the data from Mitchener and Weidenmier (2008) are very similar to the ones computed using our data set. 4 4 Please, refer to Appendix for a more detailed discussion on the comparison of the two data sets.

9 French Colonial Trade Patterns and European Settlements 299 Fig. 1 The trading partners of France per colonial group To mitigate distortions from the disaggregation of groups into countries, we cluster the standard errors by the original country groups. Also, as a robustness check, we used an alternative weight measure based on arable land area for the disaggregation. In the Robustness Checks section, we show that the results are robust to this alternative specification. Table 2 presents some summary statistics for the total value of trade per sector with each group of colonies. The table shows a fair variation in the value of trade, as indicated by the means and the standard deviations. As we can see in column (1), 40% of French trade with its own colonies corresponds to of manufactured goods, while 35% corresponds to of agricultural goods. Regarding the other groups of countries, the highest share of French trade is in of raw material for industry: 74% for British colonies, 65% for other colonies, and 52% for former colonies. Column (2) depicts the same pattern, but presenting trade shares in and in separately. Notice that the shares of French of raw materials for industry and that of of manufactured goods are very small in all the country groups between 1 and 5%. We do not include those two sectors in our regressions since they are not economically significant. The trade between France and its own colonies accounts for 30% of the total value of French trade with all the colonies, compared to 15% with British colonies, 1% with other colonies, and 54% with former colonies. If we exclude Canada and the USA, the share of French colonies increases to 42% and that of former colonies drops to 36% (these numbers are in the line of columns (3) and (4) of Table 2).

10 Table 2 French trade with colonies Summary statistics Mean SE (1) (%) (2) (%) (3) (%) (4) (%) P Tsj P Tsj ðxsjþmsj Þ Tsj Tsj, excluding USA and Canada s Pj Pj Tsj Tsj Tsj French colonies Exports Manufactured goods 316,000,000 14,400,000 8,821, Agricultural goods 76,400,000 3,472,435 1,724, Raw material for industry 36,400,000 1,655,525 1,018, Imports Manufactured goods 6,637, , , Agricultural goods 280,000,000 12,700,000 8,531, Raw material for industry 76,800,000 3,492,524 1,671, ,000,000 36,000,000 21,900, British colonies Exports Manufactured goods 46,200,000 1,320, , Agricultural goods 8,245, ,599 98, Raw material for industry 6,094, ,137 95, Imports Manufactured goods 2,785,935 79,598 58, Agricultural goods 36,700,000 1,048, , Raw material for industry 290,000,000 8,296,335 4,464, ,000,000 11,200,000 5,615, Other colonies Exports Manufactured goods 2,324, ,041 72, s 300 T. El Kallab, C. Terra

11 Table 2 continued Summary statistics Mean SE (1) (%) (2) (%) (3) (%) (4) (%) P Tsj P Tsj ðxsjþmsj Þ Tsj Tsj, excluding USA and Canada s Pj Pj Tsj Tsj Tsj Agricultural goods 1,157,049 82,646 35, Raw material for industry 298,050 21,289 7, Imports Manufactured goods 1,870, ,628 89, Agricultural goods 4,314, , , Raw material for industry 15,900,000 1,132, , ,300,000 1,733, , Former colonies Exports Manufactured goods 302,000,000 17,800,000 9,696, Agricultural goods 53,000,000 3,118,234 1,331, Raw material for industry 65,400,000 3,846,529 3,442, Imports Manufactured goods 39,600,000 2,327,361 2,282, Agricultural goods 208,000,000 12,200,000 5,414, Raw material for industry 735,000,000 43,300,000 23,700, ,410,000,000 82,700,000 43,800, T stands, alternatively, for or, s refers to the sector and j to the colonial group s French Colonial Trade Patterns and European Settlements 301

12 302 T. El Kallab, C. Terra Table 3 European settlement: summary statistics Variable # Obs. Mean (%) SD (%) Min Max European settlement ES in French colonies ES in British colonies ES in former colonies ES in other colonies European Settlements European settlements in 1900 is the percentage of the population that was European or of European descent in the year We use the data from Acemoglu et al. (2001), completing the missing values for some French colonies with data from Huillery (2011), as shown in Table 11 in Appendix 1. Table 3 presents some summary statistics for the European settlements (ES) variable. The variable presents appreciable variability across and within colonies. For instance, among French colonies, ES ranged from 0 to 30% with a mean of 1.21% and a standard deviation of 4.11%, whereas among British colonies, it ranged from 0 to 93% with a mean of 3.1% and a standard deviation of 12.4%. European Language The European language variable is an index we create to capture the extent to which European languages are spoken in a country. The languages are French, English, Spanish, Dutch, Italian and Portuguese. The index, ranging from 0 to 1, was constructed based on the CEPII dataset, which reports two variables: the main four languages spoken by at least 9% and at most 20% of the population, and the main four languages spoken by more than 20% of the population. The data also contains information on whether two countries share a common ethnological language, including the creole language and a lingua franca. Based on these data, we constructed our index for language according the following formula: EL ¼ n EL þ m EL 0:2 N TL M TL where EL is the European language index, n EL is the number of European languages spoken by more than 20% of the population, N TL is the total number of languages spoken by more than 20% of the population. If, for instance, more than 20% of the population in Benin speaks French while the spoken languages are French and Fon (the local language), the share of EL would be equal to 0.5. m EL is the number of European languages spoken by 9 to 20% of the population, and M TL is the total number of languages spoken by 9 20% of the population. The second ratio is multiplied by 0.2 to take into account that this ratio concerns at most 20% of the population. Let us take Ghana as an example. There are two languages spoken in Ghana, English and Ewe, the local language, and English is spoken by at most 20% of the population. The EL would be equal to 0 2 þ 1 2 0:2 ¼ 0:1. The value of 1 is

13 French Colonial Trade Patterns and European Settlements 303 attributed to those countries whose second official language is a European one. This additional information is gathered from various historical backgrounds. 5 Notice that the European language index is different from the colonial dummy. The fact that France colonized Benin, for instance, does not mean that the Beninese speak French. We chose to create the index including all European languages, and not only French, because speaking any European language should facilitate trade with France, compared to speaking only local languages. Finally, note that the information on language is based on current data: we use the language spoken today in a country as a proxy of the language spoken during the colonial era. The reasoning is that, if a former colony speaks a European language today, the use of that language must have started during colonial times due to the presence of missionaries and colonizers. Once the colony acquired its independence, it had less incentive to learn the colonizer s language. Mass, Distance, Landlocked and Other Controls We would have liked to include GDP as a measure of size in our gravity model, but reliable estimates of GDP for most of the colonies are not available for our period of study. We follow Mitchener and Weidenmier (2008) and use population as a proxy for GDP instead. Our data are from Maddison (2006) and Mitchell (2007a). The geographical distance between France and its trade partners is from Geo CEPII, constructed by Mayer and Zignago (2011). Landlocked data and colonial status dummies are from the Geo CEPII data and from Nunn (2008). The preferential trade agreements and the currency union data are from Mitchener and Weidenmier (2008). 6 The data on whether each country was at war in a particular year were constructed based on the data from the Correlates of War Web site. The climate variables, average temperature and average humidity are from Acemoglu et al. (2001), completed with those from Parker (1997). We took the average temperature in Centigrade and the average humidity in percentage. The Impact of Settlements on Trade OLS Results We start by looking at the impact of colonial status on trade, without considering European settlements. To that end, we estimate Eq. (1) without the terms with ES, using ordinary least-squares (OLS) with standard errors clustered at the original geographical division of countries. From the results in Table 4, we see that the trade of France with it own colonies is on average greater than the trade with the other groups of colonies, given that the estimated constant is positive and significant, 5 Notice that, theoretically, the index could take a value larger than one if, for example, there is only one language spoken by at least 20% of the population and only one language spoken by 9 20%, and these two languages are European. In this case, the number would be 1.2. First, such case does not occur in our data, second, we would take the minimum value between the EL index and 1. 6 We thank Marc Weidenmier and Kris Mitchener for generously sharing their data.

14 304 T. El Kallab, C. Terra Table 4 OLS: the impact of colonial status on trade (1) (2) (3) (4) (5) (6) Exports of Imports of Agricultural goods Manufactured goods Agricultural goods Raw material for industry Constant 15.00*** 12.22*** 13.96** 5.62*** *** (4.04) (2.40) (5.46) (2.04) (6.40) (2.18) Dummy British *** *** *** *** *** colonies (0.51) (0.40) (0.70) (0.37) (0.72) (0.43) Dummy other *** *** *** *** *** colonies (0.70) (0.56) (0.90) (0.66) (1.43) (0.57) Dummy former * ** * colonies (0.77) (0.59) (0.95) (0.71) (1.05) (0.72) Log population 1.02*** 0.90*** 1.08*** 0.95*** 0.92*** 1.02*** in 1900 (0.09) (0.08) (0.10) (0.07) (0.16) (0.08) Log distance * km 2 (0.46) (0.31) (0.58) (0.25) (0.74) (0.27) Landlocked ** ** * *** ** *** (0.56) (0.46) (0.66) (0.35) (0.67) (0.36) War 4.88* (2.76) (2.38) (3.31) (2.63) (6.32) (2.45) Temperature * (0.04) (0.02) (0.05) (0.03) (0.07) (0.03) Average humidity (0.01) (0.01) (0.01) (0.01) (0.02) (0.01) Trade ** preference with non- French empire (0.46) (0.40) (0.54) (0.32) (0.83) (0.34) Trade preference with French empire Currency union with French empire Currency union with non- French empire 2.01** 1.79** 2.34** (0.93) (0.86) (1.06) (0.57) (1.21) (0.50) *** * (0.85) (0.81) (0.95) (0.66) (1.30) (0.62) (0.34) (0.23) (0.41) (0.23) (0.48) (0.25) European 2.05*** 2.18*** 1.93*** 2.10*** 3.70*** 1.42*** language index (0.51) (0.48) (0.63) (0.43) (0.92) (0.51) Observations

15 French Colonial Trade Patterns and European Settlements 305 Table 4 continued (1) (2) (3) (4) (5) (6) Exports of Imports of Agricultural goods Manufactured goods Agricultural goods Raw material for industry R Robust standard errors in parentheses The data on and are in log. ES refer to European settlement in The standard errors are clustered at the original geographical division of countries *** p \ 0.01; ** p \ 0.05; * p \ 0.1 while the coefficients for the other groups of countries are negative and significant. This is true for (coefficients in columns (1) (3)) and (columns (4) (6)), except for France s of agricultural goods, which do not seem to depend on colonial status. More populous countries trade more with France, while landlocked countries trade less, as expected. Interestingly, distance is not a significant explanatory variable of trade between France and the colonies and former colonies. Indeed, when countries are linked by colonial ties, traditional gravity variables such as distance or multilateral resistance may become relatively less important in governing trade compared to the case of two independent countries. Institutional and historical factors play a more prominent role. 7 Trade preference agreements with France increase France s, but they have no significant impact on. Finally, colonies and former colonies where European languages are more spread tend to trade more with France. Table 5 reports the OLS regressions of Eq. (1), this time including the terms with European settlements. Standard errors are clustered as before. The results indicate that ES brings additional information to explain colonial trade, compared to colonial status alone. The estimated coefficients for the constant and the dummies for colonial groups are very similar to the ones without the ES terms, presented in Table 4, while the estimated coefficients for ES and its interaction with colonial groups dummies are significantly different from zero. More specifically, France traded more with French colonies with a greater share of European settlers: One more percentage point in ES increases France s and with French colonies by 7 and 6%, respectively. The amount of ES, however, did not impact French trade with the other group of colonies: the coefficients for ES interacted with colonial dummies are negative, significant, and equal in magnitude to the coefficient of non-interacted ES. This result is in line with the assumption that settlers in other colonies were mainly their respective colonizers, so that their settling had no significant impact on the trade of those colonies with France. 7 In a regression not reported here, we have tried to control for a multilateral resistance variable as suggested by Baier and Bergstrand (2009), but the variable was highly collinear with distance since we only have the bilateral trade with France.

16 306 T. El Kallab, C. Terra Table 5 OLS: the impact of European settlement on trade European settlement in 1900 ES in British colonies ES in other colonies (1) (2) (3) (4) (5) (6) Exports of Imports of Agricultural goods Manufactured goods Agricultural goods Raw material for industry 0.07*** 0.04*** 0.08*** 0.06*** *** (0.02) (0.01) (0.02) (0.01) (0.03) (0.01) *** *** *** *** *** (0.02) (0.01) (0.02) (0.01) (0.04) (0.02) *** *** *** *** *** (0.02) (0.01) (0.02) (0.01) (0.03) (0.01) ES in former * ** * colonies (0.03) (0.02) (0.04) (0.02) (0.04) (0.03) Constant 11.70*** 9.65*** 10.18* 4.27** ** (3.94) (2.34) (5.28) (2.08) (6.38) (2.36) Dummy British *** *** *** * colonies (0.44) (0.40) (0.59) (0.40) (0.75) (0.42) Dummy other *** *** *** * * colonies (0.62) (0.56) (0.76) (0.80) (1.52) (0.69) Dummy former ** colonies (1.00) (0.79) (1.21) (1.02) (1.36) (1.00) Log population 0.96*** 0.85*** 1.01*** 0.89*** 0.88*** 0.97*** in 1900 (0.08) (0.08) (0.09) (0.08) (0.17) (0.08) Log distance km 2 (0.39) (0.25) (0.50) (0.26) (0.67) (0.26) Landlocked ** *** ** *** ** *** (0.47) (0.42) (0.54) (0.38) (0.66) (0.36) War (5.44) (3.93) (6.91) (5.00) (7.61) (5.99) Temperature * (0.03) (0.02) (0.04) (0.04) (0.08) (0.04) Average humidity (0.01) (0.01) (0.01) (0.01) (0.02) (0.01) Trade preference with non- French empire (0.49) (0.45) (0.57) (0.34) (0.89) (0.37) Trade preference with French empire 2.10** 1.88** 2.39*** 0.91* (0.80) (0.82) (0.87) (0.54) (1.20) (0.49)

17 French Colonial Trade Patterns and European Settlements 307 Table 5 continued Currency union with French empire Currency union with non- French empire (1) (2) (3) (4) (5) (6) Exports of Imports of Agricultural goods Manufactured goods Agricultural goods Raw material for industry ** (0.80) (0.79) (0.88) (0.64) (1.36) (0.57) (0.39) (0.27) (0.49) (0.22) (0.50) (0.24) European 1.90*** 2.12*** 1.76*** 1.84*** 3.61*** 1.20** language index (0.52) (0.53) (0.62) (0.43) (0.97) (0.46) Observations R Robust standard errors in parentheses The data on and are in log. ES refer to European settlement in The standard errors are clustered at the original geographical division of countries *** p \ 0.01; ** p \ 0.05; * p \ 0.1 Our results are compatible with the idea that settlements provided a way for the French to establish control in order to exploit the colonial markets raw materials and to export to them their manufactured products. Marseille (1984) observed that France, isolated within its empire and imposing the prices of products in the colonial markets through custom unions and trade preferences, exported its manufactured products to ensure the existence of solvent markets for the empire during times of crisis. Interestingly, like the result for colonial status, France s of agricultural goods are insensitive to the size of ES in the colonies and former colonies. One possible explanation is that the French were exploiting the natural resources even before formal colonization took place, but further evidence would be necessary to confirm this hypothesis. The coefficients of currency union with French empire are positive, significant and high in magnitude for France s of agricultural goods (column (5) of Tables 4, 5), whereas they are not significant for the regressions using the other trade variables. This indicates that being part of a currency union with the French empire increases France s of agricultural goods by 360%. This further highlights the importance of preferential trade policies set by the French in order to increase their exploitation of their colonies raw materials.

18 308 T. El Kallab, C. Terra The estimated coefficients for the other control variables for the regressions with ES in Table 5 are very similar to ones from the regressions without ES in Table 4.In the next subsection, we show that our results are robust for various alternative specifications. For some anecdotal evidence on the impact of ES on sectoral French trade, it is interesting to look at the export boom in French West Africa, namely in Senegal and the Ivory Coast, between 1897 and Timber from the Ivory Coast increased by a factor of six in 20 years (Frieden 2006), as colonial of European manufactures grew. In Indochina, which was under a French colonial regime, the area of cultivated land dramatically increased, allowing it to become the third largest producer of rice in the world (Mitchener and Weidenmier 2008). French settlements strengthened ties between the empire and its colonial markets, achieving the ultimate goal of French colonization, as stated by Jules Ferry (1911) Colonial policy was the daughter of industrial policy. France used the colonial markets as both a buffer in times of crisis and a profitable alternative in times of expansion to dispense its products (Marseille 1984). Robustness Checks To check the sensitivity of our results to the specification of the econometric model, we conducted a series of robustness checks. Instrumental Variable First of all, we employ an instrumental variable approach to control for possible endogeneity of European settlements. Trade may have preceded colonial expansions, or even have led to it, so that colonization and the choice of settling in the colonies may be endogenous to trade. In order to correct for this, we employ a twostage least-square estimation method where we instrument ES by pre-colonial population density in the 1500s from Acemoglu et al. (2002). It is reasonable to think that pre-colonial population density is related with ES in According to Engerman and Sokoloff (1997), Europeans were more likely to settle in previously poor and less populated regions because indigenous population density raised costs for Europeans to obtain and secure land for new settlers. Acemoglu et al. (2002) argue that the density of pre-colonial indigenous population affected the returns from setting up extractive institutions. Also, population density is positively correlated with the formation of pre-colonial states, and Fenske (2014) claims that strong pre-colonial states might have discouraged new settlements because of the costs incurred from creating extractive trade institutions. Following the same line of arguments, Huillery s (2011) recent empirical results show that high population density in the pre-colonial period increased the risk of indigenous hostility and hence discouraged new settlements. In sum, population density in 1500

19 French Colonial Trade Patterns and European Settlements 309 is likely to have had a negative impact on settlement decisions for Europeans over the colonial period. 8 In the first-stage regressions reported in the lower panel of Table 6, the coefficient for population density is negative and significant, with an R-squared of The second-stage regressions, in columns (1) and (2) of the table, confirm that France traded more with its own colonies with larger ES. ES did not impact France s trade with British colonies. The coefficient of the interaction term between ES and British colonies dummy is negative and of similar magnitude as the coefficient of the non-interacted ES. The validity of our instrument is threatened if other factors correlated with population density in 1500 affect trade in It is reasonable to assume that the potential impact of population density in 1500 on trade through the remainder of the factors not controlled for, if such impact exists, should be a positive one, for they are likely to have a positive impact on economic activity, hence on trade expansion later on. Since we found a negative correlation between population density in 1500 and ES, and a positive correlation between ES and trade, we have that the impact of population density in 1500 on trade through ES is a negative one. Therefore, the impact we are capturing cannot be driven by these other factors. In any case, to check for that possibility, we regress ES, instrumented by population density in 1500, on trade in 1900 while controlling for population density in 1900, which is likely to be correlated with population density in 1500 (those results are not reported here, but available from authors upon request). We still find a negative and significant correlation between our instrument population density in 1500 and ES in the first-stage least-square regression, and a positive impact of ES on trade in the second-stage least-squares regression. Country Groups Our choice of country groups may raise questions. For instance, German, Portuguese, and Spanish colonies are grouped together under Other colonies. The number of countries associated with each of those colonizers is too small, preventing us from having separate groups for each of them, while British colonies are numerous enough to be in a group of their own. As a robustness check, we estimated our regression with only two large groups of colonies: French versus non- French colonies. The results, presented in columns (3) and (4) of Table 6, are similar to our previous results from Table 5. Splitting the countries into only these two groups, we get that one additional percentage point in ES in French colonies increases by 4% and by 6%, while ES in non-french colonies has no significant impact on French from those colonies. One may also wonder whether the results from the former colonies groups could be biased by the former British (USA and Canada) and French (Haiti) colonies in 8 Two other variables, latitude and settler mortality, are used as instruments of ES in previous studies. However, they are not valid instruments in our case, since both of them have a direct impact on trade. Latitude is correlated with climate and geographical conditions that affect agricultural productivity, hence trade. Settler mortality might be correlated to low agricultural productivity due to the high burden of diseases.

20 310 T. El Kallab, C. Terra Table 6 Robustness checks (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Instrumental With only two Excluding USA, Disaggregation with Excluding countries Averaging the full variables colonial groups Canada and Haiti land area with data different sample from M&W European settlement in 0.06*** 0.07*** 0.04* 0.06*** 0.07*** 0.06*** 0.07*** 0.07*** ** 0.05*** 0.06*** 1900 (0.022) (0.02) (0.02) (0.02) (0.02) (0.01) (0.017) (0.016) (0.03) (0.02) (0.01) (0.01) ES in British colonies ** ** *** *** *** ** *** ** *** *** (0.035) (0.023) (0.02) (0.01) (0.016) (0.024) (0.03) (0.03) (0.02) (0.02) ES in other colonies *** *** *** *** * ** *** *** (0.048) (0.065) (0.02) (0.01) (0.017) (0.016) (0.03) (0.03) (0.02) (0.01) ES in former colonies * * ** * * (0.038) (0.031) (0.03) (0.02) (0.023) (0.026) (0.04) (0.04) (0.02) (0.02) ES in non-fr colonies ** (0.03) (0.02) Constant 13.18** 4.69* 11.90** *** 4.27** 14.95*** *** *** 3.78* (5.419) (2.515) (4.42) (2.24) (3.94) (2.08) (3.486) (4.319) (3.88) (2.54) (3.45) (2.05) Dummy British colonies *** *** ** *** * *** ** (0.449) (0.405) (0.44) (0.40) (0.585) (0.756) (0.55) (0.46) (0.43) (0.39) Dummy other colonies *** *** *** * *** *** ** *** ** (0.502) (0.315) (0.62) (0.80) (0.523) (0.778) (0.58) (0.72) (0.62) (0.58) Dummy former colonies ** * (1.385) (1.117) (1.00) (1.02) (1.116) (1.217) (1.17) (1.20) (0.87) (0.90) Dummy non-french colonies *** ** (0.46) (0.47)

21 French Colonial Trade Patterns and European Settlements 311 Table 6 continued (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Instrumental With only two Excluding USA, Disaggregation with Excluding countries Averaging the full variables colonial groups Canada and Haiti land area with data different sample from M&W Log population 0.94*** 0.87*** 1.07*** 1.02*** 0.96*** 0.89*** 0.95*** 0.90*** 0.99*** 0.90*** 0.93*** 0.84*** (0.089) (0.074) (0.08) (0.09) (0.08) (0.08) (0.098) (0.104) (0.10) (0.11) (0.09) (0.07) Log distance km (0.529) (0.239) (0.43) (0.26) (0.39) (0.26) (0.306) (0.419) (0.37) (0.29) (0.34) (0.24) Landlocked ** *** ** *** ** *** *** *** ** *** (0.521) (0.346) (0.46) (0.33) (0.47) (0.38) (0.540) (0.578) (0.39) (0.37) (0.41) (0.20) War *** (4.111) (3.105) (2.94) (3.13) (5.44) (5.00) (6.166) (5.403) (6.45) (5.76) (4.63) (4.21) Temperature *** ** * ** ** (0.045) (0.039) (0.04) (0.03) (0.03) (0.04) (0.071) (0.059) (0.04) (0.04) (0.03) (0.03) Average humidity * (0.009) (0.007) (0.01) (0.01) (0.01) (0.01) (0.012) (0.013) (0.01) (0.01) (0.01) (0.01) Trade preference with non- French empire Trade preference with French empire Currency union with French empire * * (0.571) (0.351) (0.51) (0.34) (0.49) (0.34) (0.814) (0.628) (0.55) (0.54) (0.42) (0.22) 2.43** 1.35** 2.22** 0.96* 2.10** 0.91* 2.32** 1.47* 2.40*** * (0.942) (0.577) (0.80) (0.54) (1.089) (0.765) (0.86) (0.65) (0.79) (0.44) (0.898) (0.606) (0.80) (0.63) (0.80) (0.64) (0.952) (0.801) (0.83) (0.71) (0.81) (0.63)