Trade and Empire

The Economic Journal, 118 (November), 18051834. The Author(s). Journal compilation Royal Economic Society 2008.
8. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.
TRADE AND EMPIRE*

Kris James Mitchener and Marc Weidenmier
We employ a new database of over 21,000 bilateral trade observations from 18701913 to assess the contemporaneous effects of empire on trade. Our analysis shows that belonging to an empire roughly doubled trade relative to those countries that were not part of an empire. The use of a common language, the establishment of currency unions, the monetisation of recently acquired colonies, and the establishment of preferential trade agreements and customs unions help to account for the observed increase in trade associated with empire.
What factors determine the size of trade ows between countries? To assess the determinants of bilateral trade ows, many empirical studies employ a standard augmented gravity model that includes mass, distance and a host of economic and political variables often including currency unions, tariffs, wars and exchange-rate regimes. One empirical nding of many studies is that a countrys prior colonial status exerts a large and statistically signicant positive effect on current bilateral trade.1 This result raises several interesting additional questions that, to date, have received comparatively little attention by economists. First, to what extent was trade inuenced by colonial status when these former colonies were still part of formal empires? Second, if we could examine the contemporaneous impact, would membership in an empire increase or reduce trade? Third, what are the channels through which colonial status impacts trade? To gain some perspective on these questions, this article provides a thorough examination of the contemporaneous effect of empire on trade during the Age of High Imperialism, 18701913.2 Although a few previous studies have attempted to control for the contemporaneous effects of empire on trade using historical data sets (Estevadeordal et al., 2003; Lopez-Cordova and Meissner, 2003), they did not focus on understanding how empire impacted trade, in part because their samples lacked adequate bilateral trade data on colonies and global trade.3 To examine empires effect on trade, we rst construct a new bilateral trade database of over 21,000 observations from 18701913 that is nearly 20 times larger than existing databases for this sample period.
* We thank Tai-Yu Chen, Mitch Frass, Justin Jones, Julie Van Tighem and Maria Revutsky for research assistance and the Dean Witter Foundation, Santa Clara University (Presidential Research Grant) and the National Science Foundation (NSF Grant 0518661) for nancial support. We also thank Carolyn Evans, Jeffry Frieden, Alan Taylor and seminar participants at the Harvard Business School, University of Michigan, UCLA, Universitat Pompeu Fabra and the St. Louis Federal Reserve Bank for comments and suggestions. 1 For example, see Baldwin (2005), Glick and Taylor (2006) and Rose (2000, 2002). 2 This article also relates to a growing body of scholarship that examines how empires and colonial relationships affect economic outcomes (Ferguson, 2004; Lal, 2001, 2004; Acemoglu et al., 2001). Whereas some of these studies re-examine the institutional legacies of empire and their impact on economic development while others provide an overall accounting of the economic effects of colonial relationships, the main objective of this article is a narrower one: to provide an empirical assessment of how they affect trade. We acknowledge that our empirical approach focuses on total trade ows and may not account for the extent to which some of the trade between metropole and colony was based, not on mutual exchange but on coercion. 3 Accominotti and Flandreau (2005) and Flandreau (2000) have employed gravity models to examine the effects of bilateral trade agreements and currency unions on trade during the nineteenth century. [ 1805 ]
1806
THE ECONOMIC JOURNAL
[NOVEMBER
We then estimate the impact of empire on trade using an augmented gravity model of trade and examine some channels through which this effect may have operated. Our results suggest that being in an empire roughly doubled trade relative to those countries that were not part of an empire. Moreover, the positive effect of empire on trade does not appear sensitive to whether the metropole was Britain, France, Germany, Spain or the United States, nor does it appear to be sensitive to a variety of different econometric specications or robustness checks (including endogeneity, multilateral resistance, propensity score matching and selection models). We further examined whether tariff policies and transactions costs help to account for the observed boost in bilateral trade.4 We nd that preferential trading agreements, customs unions between colonies and metropoles, empire-based currency unions and sharing a common language increased bilateral trade. Consistent with earlier studies, our analysis also conrms that, in most specications, the gold standard had a positive effect on bilateral trade ows during the period 18701913; however, the effect of joining the gold club is substantially smaller than the empire effect.5 Although our article does not claim to capture the overall welfare costs or benets of belonging to an empire, it provides new estimates of the contemporaneous effect of empire on trade and examines the mechanisms through which this effect may have operated. In addition, our results shed light on the historical origins of the large legacy effect of empire often reported in studies examining recent bilateral trade ows. In the next Section, we review the empirical literature on bilateral trade, present an augmented gravity model of trade and describe the data we use to test it. Section 2 provides empirical estimates of the effects of empire, non-empire currency unions and the gold standard on bilateral trade. Section 3 identies the channels of empire and the effects of these mechanisms on bilateral trade. The last Section offers some concluding comments and discusses how empire may have imparted a positive effect on trade.
1. Explaining Bilateral Trade Flows During the Classical Gold Standard Era
1.1. Empirical Research on Bilateral Trade Empirical research has drawn attention to the effects of policies, institutions and geography on trade. Using augmented gravity models of trade, economists have examined the importance of tariffs, transport costs, exchange rate volatility and transactions costs in explaining the cross-country variation in bilateral trade ows. Numerous studies utilising the gravity model framework (with data from different time periods and for different country samples) have reported on the costs of trade-policy frictions, the tyranny of distance and the benets of being part of a common currency area (Estevadeordal et al., 2003; Glick and Rose, 2002; Rose, 2000). Reviewing the empirical evidence and
4 Relatedly, Ferguson (2002, 2003) has argued that England enforced free trade during the gold standard period while other countries moved towards more protectionist policies near the end of the century. 5 In a similar vein, Ferguson and Schularick (2004) nd a large empire effect for British colonial bonds during the classical gold standard period. Membership in the British Empire signicantly reduced the cost of capital for colonial borrowers since the mother country guaranteed the bonds of its possessions. Flandreau and Zumer (2004) nd that adherence to the gold standard did not lower the cost of capital for sovereign borrowers during the gold standard period.
The Author(s). Journal compilation Royal Economic Society 2008
2008 ]
TRADE AND EMPIRE
1807
performing a meta-analysis of earlier studies, Rose (2004) nds that belonging to a currency union has a positive and statistically signicant effect on trade and that this result appears robust to various econometric specications, denitions of currency union, measures of distance, exchange rate volatility and country samples.6 Estevadeordal et al. (2003) has argued that historical episodes, such as the gold standard era and the interwar period, may be even better suited for testing the effects of currency unions on trade since there was considerable variation in trade ows that existed during these periods and since the gold standard represented the formation of the largest currency arrangement in history. Using quinquennial panel data for the period 18701910, Lopez-Cordova and Meissner (2003) test for the effects of currency unions and gold-standard membership on bilateral trade. After controlling for other inuences such as distance, language and a common border, they nd a large, positive effect for historical currency unions, similar to what Rose and others nd using more recent samples, as well as a large effect from gold-standard membership. Estevadeordal et al. (2003) and Flandreau and Maurel (2001) report similar results for historical currency unions. One puzzle that arises from this empirical literature is that exchange rate volatility appears to be unimportant in explaining bilateral trade ows; some studies have found that exchange rate volatility impacts trade negatively but no consensus has emerged. Even in studies where the coefcient is negative, the size of the effect is generally small and the statistical signicance varies widely.7 Many of the gravity model studies (including those examining historical periods) include a currency union indicator variable as well as a measure for exchange rate volatility as independent variables. If we assume that multicollinearity is not a severe problem, then the high degree of statistical signicance on the currency union indicator variable must reect a benet other than exchange rate stability. Rose (2000) suggests that currency unions could also reduce transaction costs that arise from operating with various currencies and that they provide a more serious and durable commitment than simply having a xed rate. Nevertheless, he professes ignorance as to why the estimated effects of currency unions on trade are so large: It is wisest to conclude that we simply dont know why a common currency seems to facilitate trade. (Rose, 2000, p. 24). This leaves open the possibility that currency union dummy variables are proxying for omitted inuences.8 As we suggested in the Introduction, one institutional factor that may have affected bilateral trade ows in the late nineteenth and early twentieth centuries and that has received little attention in previous empirical studies is empire. The notion that trade and empire are linked is certainly not new. Scholarly debate reaches back at least a century to the era of High Imperialism, when France, Germany and Great Britain (and to a lesser extent Russia, Portugal, Belgium and the US) renewed their quest for territorial acquisition. Even though the British Empire, which spanned ve continents, was still unrivalled during the nineteenth century, continental European countries began to challenge Britains role on the world stage more actively in the latter half of
See Baldwin (2005) for a more recent survey. Edison and Melvin (1990) review the empirical studies. 8 Although studies examining the period 18701913 conrm the salutary effects of currency unions on trade, they provide little direct evidence that the gold standard reduced transaction costs or payments frictions.
7 6
1808
[NOVEMBER
the century. New imperial powers sought overseas territories to complement their growing economies, which had been stimulated by the industrial revolution. Colonial acquisitions during this phase of expansion included Britain extending its holdings in Burma, Malaysia and Africa, France consolidating its Indo-Chinese Empire and its foothold in Madagascar, and Germany carving out an empire in Africa. The Age of High Imperialism also included the US, which had acquired the Philippines and Hawaii after its war with Spain. A careful reading of the economics and history literatures suggests a variety of reasons why membership in an empire could have affected trade ows during the gold standard era (Bairoch, 1989; Ferguson, 2002; Frieden, 2006; Lal, 2004; Porter, 1999). These include preferential trade policies and other transaction costs, which arise from monetary arrangements, developing marketing or distribution networks, or sharing a common language. In order to determine whether trade among empire members differed from non-empire countries and to assess the relative importance of empire versus other institutional factors, such as currency unions and monetary regimes, we now turn to estimating an augmented gravity model of trade for the Age of High Imperialism. 1.2. Gravity Model of Bilateral Trade The gravity model is the workhorse empirical model of studies examining trade ows and continues to be used widely by economists due in part to its straightforward implementation and theoretical underpinnings.9 In its simplest form, the gravity model captures two main forces affecting trade: mass (a force of attraction) and distance (a force of resistance). Mass (measured here by the size of countries) is proportional to trade whereas distance varies inversely. The model thus predicts that, all else equal, larger economies ought to trade more than smaller economies and those that are located closer to each other will also experience greater trade. The second prediction seems particularly relevant to our sample period, since transportation costs declined dramatically during the nineteenth century and appear to have been an important driver of trade during this period (Estevadeordal et al., 2003; ORourke and Williamson, 1999). We augment this basic model with an additional set of covariates to capture the effects of differences in geography, institutions, tastes and preferences on bilateral trade. In particular, previous studies have argued that, holding other factors constant, having a common border ought to boost trade while being landlocked will reduce trade with other countries. Following earlier work, we include a measure of exchange-rate volatility since more volatility ought to reduce trade. Finally, we include historicalinstitutional variables: whether countries were on the gold standard, whether they were part of a formal currency union, and whether they were part of an empire, all of which may have boosted trade between countries. Given our greatly expanded trade database, our model will also allow us to comment on the reliability of previous empirical studies focusing on currency unions and gold standard membership during the late9 There are a variety of theoretical models used to justify the implementation of a gravity model of trade. See Deardorff (1998), Evenett and Keller (1998) and Feenstra et al. (1998) for discussion and additional references.
2008 ]
TRADE AND EMPIRE
1809
nineteenth and early-twentieth centuries. Our basic estimation equation takes the following form: ln BITRADEijt b0 b1 ln RRi RRj t b2 ln Distanceij b3 Lndlckij b4 Borderij b5 ExVolatilityijt b6 ln Popi Popj t b7 Warijt b8 Goldijt b9 NONEMPCUijt b10 Empireijt eijt ; where i and j denote countries, t denotes time, and other variables are dened as follows:
BITRADEijt denotes the average value of real bilateral trade between i and j at time t;10 RR is railway track miles; Distance in miles between i and j; Lndlck is the number of landlocked countries in the country-pair dyad (0,1, or 2); Border is a binary variable which is unity if i andj share a border; ExVolatility is exchange rate volatility; Population is a nations population; NONEMPCU is a binary variable which is unity if both countries are part of either the Latin or Scandinavian currency unions; Gold is a binary variable which is unity if i and j both are on the gold standard; War is a binary variable which is unity if countries i and j are at war; Empire is a binary which is unity if both countries are part of the same political empire; b are estimated coefcients; and e is a white noise error term capturing other inuences on bilateral trade. For this study, the key coefcients of interest are b8 b10, which show the effects of the gold standard, non-colonial currency unions and empire on trade. We estimate (1) using a variety of econometric specications: pooled ordinary least squares and xed-effects models. The xed effects model, or within estimator, is equivalent to adding a complete set of dyad-pair or country-specic intercepts to the estimating equation (Anderson and van Wincoop, 2003). The dyad or country-pair xed effects are designed to capture bilateral trade resistance. Time dummies can also be added to the dyad xed effects models to control for annual shocks that impact bilateral trade ows. Country-xed effects are included to capture the idea that each country may have a different general resistance to trade. The country-xed effects can also be interacted with year dummies to allow for time-varying multilateral resistance to trade such as distance. 1.3. Data To estimate (1), we created a new, large database of annual bilateral trade that draws most extensively on a consistent set of British statistical sources published by the Board
10 The average value of real bilateral trade is either the average value of one or two dyadic trade pairs, depending on data availability.
1810
[NOVEMBER
of Trade. In particular, we relied on numerous volumes of 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 for the period 1870 to 1913. Some additional data for French colonies is from the Tableau General du Commerce Exterieur. Overall, the data consist of more than 21,000 bilateral trade observations and 880 distinct country pairs or dyads.11 The trade data we collected from British Board of Trade publications are converted into current pounds using annual exchange rates from the Global Financial Database and Ferguson and Schularick (2004). We deated the data using the UK PPI and expressed the gures in 2000.12 Although we would like to have included GDP to measure mass in our gravity model, reliable annual estimates for a wide range of non-OECD countries prior to 1914 (including smaller colonies) are scarce. We therefore used population to capture mass. Total railway miles are employed to measure a countrys transportation network that proxies for internal transport costs that might affect bilateral trade ows. These data series (as well as population) are from Banks (1976) and the aforementioned Board of Trade publications. Data on (log) distance in miles are from Rose (2002) and an online distance calculator that employs US Geographical Survey information.13 Data on when countries joined the gold standard and joined the Latin and Scandinavian Monetary Unions are from Flandreau and Maurel (2001), Bae and Bailey (2003), Ferguson and Schularick (2004), Meissner (2005) and Ofcer (2004). We computed exchange-rate volatility following the methodology of Rose (2000) but using the exchange rate sources listed above.14 We limit our denition of empire to include only formal empires and only those with more than one dependency, which rules out Sweden-Norway, but otherwise initially code data for all empires that existed during this period and for which trade data existed.15 We use information on empire afliation from the Correlates of War Database (COW) described in Sarkees (2000), Olson (1991), OBrien (1991) and the online historical encyclopedia available at http:// regiments.org/nations/index.htm. Following Glick and Taylor (2006), the COW database is also used to code interstate conicts between bilateral trading partners during the gold standard period. Our database signicantly improves upon the trade data used in earlier studies of the classical gold standard period in that it is better suited for sorting out the relative
11 The colonies included in the sample are Aden, Algeria, Australia (New South Wales, Western Australia, Queensland, South Australia, Tasmania, Victoria), Bahamas, Barbados, Belgium Congo, Bermuda, British Guiana, British Honduras, Brunei, Canada, Ceylon, Cuba, Cyprus, Djibouti, Dutch Guiana, Egypt, Falkland Islands, Fiji, French Guiana, French Indochina, Gambia, German East Africa, German SW Africa, German West Africa, Gibraltar, Gold Coast, Guadeloupe, Hawaii, Hong Kong, India, Jamaica, Labuan, Lagos, Madagascar, Maldives, Malta, Martinique, Mauritius, Morocco, Netherlands East Indies, New Caledonia, New Hebrides, New Zealand, Newfoundland, Nyasa, Philippines, Portonovo, Portuguese West Africa, Puerto Rico, Reunion, Sarawak, Senegal, Seychelles, Sierra Leone, Somalia, South Africa (Natal Province, Cape Province, and Transvaal), Southern Nigeria, St. Helena, St. Pierre/Miquelon, Straits Settlement, Togo, Trinidad and Tobago, Tunis, Uganda, UK East Africa, and Zanzibar. 12 We thank Moritz Schularick for generously sharing his data with us. 13 We use information from http://www.wcrl.ars.usda.gov/cec/java to calculate great circle distance. 14 Rose (2000) computes exchange-rate volatility as the standard deviation. 15 Belgium, Italy, Japan, Portugal and Russia also had colonial empires during this period. We have very limited bilateral trade data for the Belgian, Italian, Japanese and Portuguese colonial empires. We do not have any bilateral trade data for members of the Russian Empire. As a result, we could not consider these empires in the empirical analysis.
2008 ]
TRADE AND EMPIRE
1811
impact of belonging to a monetary standard or an empire. The rst reason is its sheer size. To date, the most comprehensive bilateral trade database for the gold standard period, at least in terms of country coverage, is Lopez-Cordova and Meissner (2003), which augmented Barbieris (1996) trade data with information from general statistical compendiums.16 It is roughly 20 times smaller than the one we have constructed. Flandreau and Maurel (2001) use annual data, but limit it to a sample of sixteen European countries, and Estevadeordal et al. (2003) uses only one year of data from the pre-World War I period. Our data are superior to Lopez-Cordova and Meissner (2003) in both dimensions of the panel number of years and number of country pairs. They constructed trade for ve-year intervals from 18701910 whereas our trade data are annual. Moreover, nearly 70% of the observations in their database come from just four years. Perhaps even more signicant is that the early years in the sample are drawn overwhelmingly from intra-European trade. For example, in 1875, 70% of the observations are European trade pairs. Even in later periods, only a small portion of the sample involves trade pairs that are both non-European and non-US. This is an important omission considering that non-European/non-US and colonial trade constituted more than 21% of world trade in 1903 and nearly 23% of world trade in 1913 (US Tariff Commission, 1922). Hence, existing databases are insufcient in country and colony coverage to permit an analysis of empire on trade. In contrast, our database contains the universe of readily available bilateral trade data reported in the British Board of Trade Statistics, which contains a signicant portion of non-European, non-US and colonial bilateral trade ows. This distinction is non-trivial if one is attempting to estimate the impact of belonging to an empire and joining the gold standard. In this case, it is important to have identifying variation in the crosssection and time-series coming from two different sources: (1) trade pairs that consist of colonies and non-colonisers and (2) colonies that are both on and not on the gold standard. Since our database provides both types of identifying variation we should be able to provide new insight into the importance of empire and monetary arrangements for trade.
2. Analysis of Bilateral Trade Flows, 18701913

2.1. Gravity Model Estimates Table 1 displays pooled, ordinary least squares regressions with clustered standard errors (since we cannot assume that eijt is independent over the country pair or dyad). The most basic specication explains approximately 40% of the variation in bilateral trade ows. Most variables enter with the correct expected signs and are, for the most part, statistically signicant at conventional levels. In terms of geographical inuences on trade, all else constant, sharing a border boosts bilateral trade whereas being
16 Barbieris (1996) data set was merged with data from at least a dozen other sources to create LopezCordova and Meissner (2003). The authors do not describe how they reconciled the differences in statistical presentation of data across sources and countries.
1812
[NOVEMBER
Table 1 The Effects of Empire on Trade, 18701913

Independent Variables Empire Membership Gold Standard Distance Border Number Landlocked Railway Track Population Non-Empire Currency Union Exchange Rate Volatility War War(1) War(2) War(3) War(4) War(5) War(6) Year Dummies Observations R-squared 0.151 (0.20) 0.406* (0.22) 0.238 (0.16) 0.166 (0.16) 0.003 (0.19) 0.069 (0.26) 0.012 (0.23) No 21,630 0.41 0.202 (0.20) 0.468** (0.21) 0.189 (0.16) 0.24 (0.17) 0 (0.20) 0.005 (0.27) 0.082 (0.24) Yes 21,630 0.42 Pooled OLS 1.318*** (0.19) 0.094 (0.10) 0.560*** (0.07) 0.676*** (0.20) 0.3 (0.23) 0.188*** (0.02) 0.368*** (0.03) 0.679** (0.34) Pooled OLS 1.292*** (0.19) 0.189* (0.10) 0.564*** (0.07) 0.624*** (0.20) 0.225 (0.23) 0.200*** (0.03) 0.365*** (0.03) 0.646** (0.32) Pooled OLS 1.462*** (0.24) 0.379*** (0.12) 0.523*** (0.09) 0.709*** (0.21) 0.028 (0.25) 0.164*** (0.04) 0.444*** (0.04) 0.317 (0.38) 2.212 (2.18) 0.343 (0.25) 0.559 (0.51) 0.088 (0.34) 0.106 (0.32) 0.347 (0.34) 0.639* (0.39) 0.442 (0.41) Yes 12,178 0.48
Robust standard errors in parentheses. * Signicant at 10%; ** Signicant at 5%; *** Signicant at 1%.
landlocked or far away from your trading partner reduces trade ows. Most of the specications also show that larger countries (as measured here by population and miles of railway) trade more. Finally, with respect to our institutional variables, being part of an empire, being on the gold standard, or belonging to a non-empire currency union all seem to have signicant positive effects on bilateral trade ows. For example, being part of an empire resulted in more than 2.5 times as much trade (273%) compared to those areas that were not part of an empire. In the initial specication, those countries on the gold standard realised a boost in their bilateral trade, although the effect is relatively small (less than 10% and not statistically signicant), when compared to the effect on empire. On the other hand, being a member in a currency union increased trade by more than 97%. We also nd that the interstate conict variables are generally not signicant during the gold standard period. This may reect the fact that
2008 ]
TRADE AND EMPIRE
1813
the gold standard period is generally considered a period of economic and political stability characterised by the absence of a global conict.17 Indeed, there are only 29 dyads exhibiting conict between bilateral trading partners in our sample. Interstate conict probably did not have an effect on bilateral trade until the outbreak of World War I, as shown by Glick and Taylor (2006). Column 2 adds year dummies to the initial specication while Column 3 additionally includes exchange rate volatility. The basic tenor of the results remains unchanged. Countries that share a border have higher trade and countries that are landlocked or far away from its trading partner have lower bilateral trade ows. Countries with greater population and more railway miles also have larger trade ows. Membership in an empire more than doubles trade and belonging to the gold standard or a non-empire currency union boosts trade by 21% and 91%, respectively. Exchange rate volatility does not have a statistically signicant effect on bilateral trade ows. Including exchange rate volatility, however, reduces the sample size by more than 50% since we do not have data on bilateral exchange rates for all trade pairs in our sample.18 Interstate conict generally does not have a statistically signicant effect on bilateral trade. Table 2 examines alternative specications of the regression model and performs some robustness checks to provide additional information on the size and signicance of the empire effect. Column 1 considers whether the empire effect is signicant for all the major empires that existed in the late-nineteenth and earlytwentieth centuries. We coded separate indicator variables for whether dyads were part of the British, Spanish, French, German, or American empires. The statistically signicant and positively signed coefcients on all the empire indicator variables suggest that empire consistently boosted trade but, as the point estimate indicates, the effects were not uniform across empires. As we discuss later in the article, differences across empires in terms of their effects on empire may relate to metropoles requiring their colonies to establish common tariff policies, other trade policies that biased trade in favour of within-empire trade, or the extent to which an empire succeeded in reducing transactions costs. Columns 2 and 3 exploit the panel nature of the data by estimating country-xed effects and country-year xed effects. Columns 4 and 5 report the results from the dyadpair and dyad-pair-year, xed-effects models. The results are robust to these alternative specications in that all the models show a strong positive association between empire membership and bilateral trade ows. Depending on the specication, the point estimates suggest that empire boosted bilateral trade between 43% and 491%.19
17 The Age of High Imperialism (18701913) is also part of the period that is commonly referred to as Pax Britannica (18251913), the era of British peace. 18 Foreign countries often report their bilateral trade in pounds sterling during the classical gold standard period. This explains why the sample shrinks nearly 50% when we include exchange rate volatility. We were forced to rely on other primary and secondary sources for bilateral exchange rate data. 19 As a robustness check, we also estimated a difference-in-differences model. By differencing the dependent variable and including year dummies we are able to control for country-specic and global trends in trade that might be driving the large empire effect found in the OLS and panel regressions. The difference-in-differences estimator shows that membership in an empire (and its 5 and 10-year lagged effects) increased trade approximately 1% per year (not reported). The empire effect is statistically signicant at the 1% level of signicance in all regressions.
1814
[NOVEMBER
Table 2 The Effects of Empire on Trade, 18701913: Sensitivity Tests and IV Estimates
Independent Variables Empire Membership British Empire Membership French Empire Membership German Empire Membership US Empire Membership Spanish Empire Membership Distance Gold Standard Border Number Landlocked Railway Track Population Non-Empire Currency Union Exchange Rate Volatility War War(1) War(2) War(3) War(4) War(5) War(6) Year Dummies Observations R-squared Pooled OLS Country Fixed Country Yr. Effects Fixed Effects 1.771*** (0.04) 1.178*** (0.20) 1.135* (0.62) 1.027*** (0.32) 2.158*** (0.53) 2.436*** (0.19) 0.557*** (0.07) 0.225** (0.10) 0.633*** (0.20) 0.234 (0.23) 0.198*** (0.03) 0.360*** (0.03) 0.618* (0.32) 1.776*** (0.038) Dyad Fixed Effects 0.394*** (0.11) Dyad Yr. Instrumental Fixed Effects Variables 0.359*** (0.11) 1.345*** (0.04)
0.724*** (0.01) 0.327*** (0.03) 0.595*** (0.04) 0.745*** (0.12) 0.098*** (0.01) 0.230*** (0.02) 0.809*** (0.08)
0.752*** (0.015) 0.396*** (0.050) 0.706*** (0.043) 0.809*** (0.136) 0.616 (1.101) 0.015 (0.202) 0.740*** (0.092) 0.251 (0.367) 0.171 (0.357) 0.127 (0.359) 0.171 (0.354) 0.171 (0.362) 0.267 (0.353) 0.108 (0.332) Yes 21,630 0.75
0.345*** (0.02)
0.155*** (0.02)
0.214 0.028 (0.21) (0.23) 0.476** 0.37 (0.21) (0.25) 0.199 0.217 (0.17) (0.22) 0.25 0.171 (0.17) (0.22) 0.009 0.002 (0.21) (0.25) 0.013 0.016 (0.27) (0.26) 0.095 0.067 (0.24) (0.23) Yes No 21,630 21,630 0.41 0.69
0.191*** 0.058*** (0.01) (0.01) 0.370*** 0.129*** (0.02) (0.02) 0.486*** 0.091 (0.11) (0.11) 1.767*** 0.975** (0.45) (0.45) 0.363 0.438* 0.203 (0.26) (0.24) (0.37) 0.618*** 0.553** 0.468 (0.24) (0.23) (0.38) 0.015 0.01 0.189 (0.24) (0.23) (0.38) 0.082 0.07 0.237 (0.24) (0.23) (0.38) 0.262 0.153 0 (0.23) (0.22) (0.38) 0.194 0.152 0.006 (0.21) (0.20) (0.38) 0.012 0.101 0.081 (0.20) (0.19) (0.37) No Yes Yes 12,178 12,178 21,630 0.28 0.35 0.42
0.568*** (0.02) 0.186*** (0.03) 0.621*** (0.05) 0.220*** (0.06) 0.200*** (0.01) 0.368*** (0.01) 0.656*** (0.11)
Including a term for time-varying multilateral trade resistance also does not reduce the size of the empire effect (Anderson and van Wincoop, 2003).20 In terms of other determinants of trade, being on the gold standard increases bilateral trade by between 17% and 49%. The currency union variable is signicant and
20 It may be the case that some bilateral trade is zero or close to zero in our sample. To deal with this issue, we also estimated a series of Tobit and median regressions. For both of these specications, the empire effect on trade remains large, positive and statistically signicant.
2008 ]
TRADE AND EMPIRE
1815
positive in ve out of the six specications. Exchange rate volatility has a negative and statistically signicant effect on bilateral trade ows in the two country-pair xed effects models. Interstate conict generally does not have a statistically signicant effect on trade in the different empirical specications even when the joint signicance of the lagged variables is tested. The possibility that the ag may have followed trade could produce biased results in the OLS and panel regressions. That is, empires may have colonised areas where there were already well-established trade ties or where there was potential for strong trade linkages between the region and the metropole.21 To control for this source of endogeneity, we use an instrumental variables model. Our instrument for empire is the ve-year lagged value of the size of other empires (measured by area). Our constructed instrument suggests that an empire, such as Britain, may have increased its size, in part, because it felt threatened economically, politically, or militarily by France or Germanys territorial acquisitions. Historians have noted that this was a primary reason why the Great Powers sought out new colonies during the second half of the nineteenth century. For example, historians have described the dramatic expansion by European powers in Africa as a scramble that was unsuccessfully held in check by the failed partition arrangement of the Berlin Conference in 18845. Germanys expansion beyond Africa and into the Middle East and Far East led British policy makers to worry about this new European colonial rival; Bismarck, in turn, seemed to have decided to engage Germany in territorial expansion as a response to what he saw as aggressive actions of European rivals (Townsend, 1966 pp. 71, 87). Americas growing naval power and acquisitions in the Caribbean and the Philippines in 1898 signalled the presence of a new rival to the West and led to territorial disputes in South America. And Britain and France competed in Asia over areas once occupied by China (Porter, 1999). Since our instrumental variable tracks the growth of other empires, it is likely to be highly correlated with the empire dummy variable but uncorrelated with bilateral trade of a given empire. The rst stage regression suggests that we probably do not suffer from a weak instrument problem.22 The results from the instrumental variables estimation are very similar to the baseline regressions as well as the xed effect specications. Empire membership signicantly increases trade. To provide some additional insight into the effects of colonisation on trade, consider the trade of West Africa an area of new colonisation in the late nineteenth century. Exports, such as groundnut oil, which was used as a substitute for olive oil and palm oil to lubricate machinery, more than quadrupled between 1897 and 1913. The export boom was especially pronounced in the British colonies of Nigeria and Gold Coast and the French colonies of Senegal and Ivory Coast. In Nigeria, groundnut exports went from a few million pounds to over 130 million pounds. Cocoa exports boomed in the Gold Coast and timber exports from the Ivory Coast increased by a factor of six in twenty years (Frieden, 2006). In return, these colonies began to import European manufactures. In Indochina (under the French colonial regime), the land under cultivation dramatically increased, allowing it to become the third largest producer of rice in the world. In British
21 Frieden (2006, p. 74) suggests that it is a matter of continuing controversy how important foreign economic interests were in colonial expansion. 22 The ve-year lagged value for the land area of other major empires is signicant at the 1% level as an instrument in the rst stage regression. The R-squared in the rst stage regression is approximately 90%.
1816
[NOVEMBER
Malaya, tin became the most important commodity export, supplying half of the worlds demand and, after 1900, Malaya and British Ceylon (already exporting large quantities of tea and coconuts) became major exporters of rubber (Frieden, 2006). 2.2. Robustness Checks To test whether our results are sensitive to the specication of the econometric model, we also conducted a series of robustness checks. We rst test whether the inclusion of gross domestic product (GDP) a measure of country mass that is widely used in gravity models that analyse the modern period (when estimates of GDP are widely available for most countries) changes our results. We include GDP for all countries and colonies for which reliable historical estimates have been assembled by employing the data set of Clemens and Williamson (2004).23 Their data set includes estimates for 35 countries and colonies and is assembled using a variety of sources including Maddison (1995). The inclusion of GDP reduces the sample size to approximately 6,700 observations; however, as shown in Appendix Table A1, membership in an empire still has an economically large and statistically signicant effect on bilateral trade ows. Using this more limited sample of countries and colonies, membership in a colonial empire raises trade between 159% and 780% in the dyad and country xedeffects specications.24 The other variables in the gravity model generally have the correctly predicted signs and are statistically signicant at conventional levels. GDP is positive and statistically signicant. Bilateral trade ows are lower for countries/ colonies that are located farther away from each other and in cases where they are landlocked. Countries and colonies that border one another have greater trade, and membership in a non-empire currency union raises trade in three out of the four specications. Although adherence to the gold standard raises bilateral trade ows, the effect is only statistically signicant in the dyad xed-effects model. The most notable change in the empirical results from the baseline regressions is the statistical insignicance of the population variable, which has a correlation coefcient of nearly 90% with the GDP variable. The fact that these two variables are so highly correlated suggests that we are losing very little in our analysis when we include population to analyse the broader sample of countries and colonies contained in our bilateral trade database. Some of the literature on gravity models has suggested that treatment effects of policy variables (such as currency unions or xed exchange rate regimes) can be distorted if the characteristics shaping trading costs are very different between the treated and untreated populations (Persson, 2001; Baier and Bergstrand, 2007; Ritschl and Wolf, 2003). In our case, it may be that the determinants of trading costs are quite different between those belonging to empires and those that do not. To address this potential shortcoming, we employ propensity score matching techniques to estimate the treatment effect of becoming part of an empire. The basic idea of propensity scoring is to estimate the probability of policy adoption (becoming part of an empire) conditional on observable characteristics.25 To estimate the propensity score, we rst
We thank Michael Clemens for generously providing these data. We estimated a country xed effects model with year dummies. 25 Of course, the underlying assumption in propensity score matching is that unobserved heterogeneity is not a signicant problem.
24 23
2008 ]
TRADE AND EMPIRE
1817
employ a probit model which is a function of the variables employed in the baseline gravity models in Table A2 (membership in the gold club, number of countries landlocked in the dyadic trade pair, distance, border, population and common language). As shown in Appendix Table A2, all of the independent variables are statistically signicant at the 1% level in the probit regression. Membership in the gold club, distance and a common language increase the probability of empire membership. The positive coefcient on distance probably reects the fact that metropoles often colonised far away areas. The negative coefcient on population suggests that it was easier for metropoles to colonise areas that had fewer people. We then rank the propensity scores and match treated (empire) observations with observations (non-empire) that possess similar characteristics based on the probit model. Appendix Table A3 shows that matching based on propensity scores greatly reduces the selection bias when the sample means of the matched and unmatched samples (what would be used in a pooled OLS regression). By comparing empire and non-empire dyads with similar observable characteristics, we can reduce selection bias by at least 77% for each independent variable. In the case of population, matching techniques reduce the bias by almost 100%. We then employed a number of commonly used propensity score matching methods to estimate the average treatment effect (ATE) (the effect of empire on bilateral trade ows).26 The rst method is the nearest neighbour method with replacement. This technique matches each treated (empire) observation with the n control units (nonempire) that have the closest propensity score. We selected n 1 and n 3 for the empirical analysis. Stratication, the second method, compares treated (empire) and non-treated (non-empire) observations within a number of subsamples or blocks based on propensity scores. The third method is radius matching which matches a treated (empire) observation to a control group (non-empire) within a specied radius (r). We use a wide radius (r 0.03), a medium radius (r 0.01), and a tight radius (r 0.005). Appendix Table A4 shows that the average treatment effect is between 228% and 268%, which is similar in size to the empire effect estimated in the baseline gravity models. The (empire) treatment effect is statistically signicant at the 1% level in the three different matching models.27 If the decision to join an empire is correlated with a set of variables that also affect economic outcomes, then there may be a selection on observables problem. In addition to the IV estimates we reported in Table 2 we use to address endogeneity, Appendix Table A5 estimates a Heckman selection model as a robustness check. In the rst stage regression, membership in a colonial empire is estimated as a function of common language and the exogenous variables in our gravity model (distance, membership in the gold club, distance, border, number of countries landlocked in the dyadic trade pair, miles of railway track, population, membership in a non-empire currency union, war and its lagged effects). The second stage uses the same variables used in the baseline gravity equations shown in Table 1 and 2 and the predicted values of empire membership from the rst stage. Appendix Table A5 shows that membership
26 Persson (2001), Baier and Bergstrand (2007) and Ritschl and Wolf (2003) are some recent papers that have used the propensity score methodology to estimate the effect of currency unions on trade. 27 The average treatment effect is similar if we use group data (where the groups are dyadic averages) for propensity score matching.
1818
[NOVEMBER
in an empire has a large positive and statistically signicant effect on bilateral trade ows in both the two-step and maximum likelihood models; other effects on trade are quite similar to those shown in earlier tables. The statistical precision of the estimated effects of trade on empire in the xed effects models also depends on the extent to which our treatment variable (empire) has sufcient identifying variation; if there are too few countries/colonies that change their status during our sample period, we may not be able to identify a statistically signicant effect of empire on trade (i.e., the standard errors will be too large). Although the variation in our treatment variable seems sufcient to identify a statistically signicant empire effect, we nevertheless run a more stringent test to check the robustness of our xed effects estimates. We re-estimated our model including only those colonies that changed their empire status during our sample period (either by exiting, switching or becoming part of an empire). Many (but not all) of the dyadic trade pairs belonged to an empire for the whole sample period; our new sample thus contains 552 observations. Appendix Table A6 shows that membership in an empire has a large positive and statistically signicant effect on bilateral trade ows even when the sample is limited to only those countries that changed their empire status. Membership in an empire increases bilateral trade ows by approximately 115%. Overall, the empire effect appears to be robust to a wide range of sensitivity checks.
3. What Explains the Empire Effect?

Our empirical results suggest that being part of an empire in the late-nineteenth and early-twentieth centuries had a large positive and statistically signicant effect on trade. In this Section, we hope to shed further light on this nding by exploring some of the channels through which empire may have boosted bilateral trade ows. Although there are many possible ways in which a particular colonys trade was directly impacted by membership in an empire, our aim is to examine general effects that can be discerned across empires and over time. We focus on two channels that both economists and historians have highlighted as important determinants of trade: (1) trade policy and (2) transactions costs. 3.1. Trade Policy Trade policies of metropoles may have been shaped by colonial ties. Economists have recently suggested that European powers viewed colonies as a way of setting up preferential trading relationships (Estevadeordal et al., 2002) and of ensuring markets (Alesina, 2002).28 This system of trade preferences and agreements between an empire and its colonies often took three different forms:
28 Alesina (2002, p. 20) quotes the British premiers response to the French ambassador in 1897 to make his case: If you were not such persistent protectionists you would not nd us so keen to annex territories. One could perhaps also view the motivation for preferential trade arrangements in light of the earlier literature on economic imperialism. Hobson (1902) and Lenin (1916) suggested that capitalist economies needed to acquire colonies in order to sustain themselves; securing favourable terms for trade with colonies may have been another way to sustain their economies. Of course, these writers emphasised ows of nancial capital rather than trade ows as the motivation for acquiring colonies.
2008 ]
TRADE AND EMPIRE
1819
(1) policies of tariff assimilation/customs union, (2) preferential tariff policies and/or (3) open door policies. The policy of tariff assimilation is a policy regime where the tariff rates on goods are the same in the metropole and the colony. Under this arrangement, the metropole and colony form a customs union. A preferential tariff system describes a trade policy where colonies and the mother country have differential tariffs but non-empire goods are generally taxed at a higher rate. An open door trade policy refers to a tariff regime where there is no distinction made between the products of the mother country and non-empire trading partners. In other words, a colony or metropole with an open door trade policy does have not a preferential tariff policy or trade agreement (i.e., customs union) with some of its trading partners. The open door trade policy should not be confused with a free trade policy, however. Many countries with open door policies levied duties to protect local industries or to raise revenue for the scal authority (US Tariff Commission, 1922). From the colonisers perspective, Britain was the least protectionist of the imperial powers as of 1913.29 According to Bairoch (1989, p. 139), average tariff rates on imported manufacturers were around 13% in Germany, over 20% in France, over 40% in the US, and more than 80% in Russia. Table 3 breaks down the trade policies of the major empires during the era of high imperialism. Great Britain generally maintained an open door policy during the gold standard period, while many of the British colonies in the West Indies (such as Jamaica, British Guiana and the Bahamas) adopted differential import duties to promote domestic industries. The British Dominions Canada, Australia (in particular, Victoria), New Zealand and South Africa also implemented preferential trading agreements in favour of Great Britain between 1898 and 1907 to protect domestic producers and manufacturers (US Tariff Commission, 1922). On the other hand, France and most of its colonies adopted tariff assimilation or a customs union as its predominant trade policy in 1892. Under this regime, colonies enjoyed free trade with France for most products while non-colonies were charged tariffs to promote trade within the empire. As shown in Table 3, Algeria, Indochina and Tunis, three of Frances most important colonies, formed a customs union with the metropole. Many of its remaining colonies, including French West Africa as well as its island dependencies adopted open door or preferential trading policies. The smaller colonial empires tended to have a more uniform colonial trade policy. For example, colonies of the Belgian, Dutch and German empires had open-door trading policies and low to moderate tariffs that were levied strictly for revenue purposes. Spain, Portugal and the US generally adopted preferential tariff systems with its colonies. There were a few exceptions, however. Macao, Portuguese Congo, the Canal Zone and American Samoa had open door trade policies and the US maintained a customs union with Puerto Rico after acquiring the colony in the Spanish-American War in 1898.
29
The average level of import duties on manufactures was approximately zero for the UK.
1820
[NOVEMBER
Table 3 Colonial Tariff Systems

Countries Belgium France Assimilated Algeria French Indo-China Tunis Madagascar Reunion Martinique Guadeloupe New Caledonia French Guiana Gabon German East Africa German Southwest Africa Kamerun Togo German Samoa New Guinea Kiaochow British India Newfoundland Papua Norfolk Island Colonies in Asia: Aden Ceylon Straits Settlements Federated Malay States Protected Malay States Hong Kong Weihaiwei Colonies in Africa: Nigeria Gold Coast Sierra Leone Gambia British Somaliland Kenya and Uganda Zanzibar and Pemba Nyasaland Egypt Anglo-Egyptian Sudan Other colonies: Gibraltar Malta British North Borneo Brunei Sarawak Tonga Solomons Gilbert and Ellice Islands Mauritius Seychelles Falkland Islands Bermuda St. Helena Preferential Open Door Belgium Congo
Germany
Great Britain
Dominions: Canada Australia New Zealand Cook Islands Union of South Africa Rhodesia Colonies: Trinidad British Guiana Jamaica and Caymans Turks and Caicos Barbados Leeward Islands: Dominica Montserrat St. Christopher-Nevis Virgin Islands Antigua Windward Islands Grenada St. Lucia St. Vincent British Honduras Bahamas Fiji
2008 ]
TRADE AND EMPIRE
1821
Table 3 Continued
Countries Italy Japan Netherlands Portugal Mozambique Angola Cape Verde Islands Portuguese India Timor Sao Thome and Principe Portuguese Guinea Fernandi Po Spanish Guinea Rio de Oro Philippines Virgin Islands Guam Formosa Saghalin Korea Assimilated Preferential Eritrea Somalia Libya Open Door Italian Northern Somaliland Rhodes Kwangtung Kiacochow (leased territory) Dutch East Indies Curacao Dutch Guiana Macao Portuguese Congo
Spain United States Puerto Rico
Canary Islands Spanish Morocco American Samoa Canal Zone
Source: US Tariff Commission (1922).
3.2. Transactions Costs A second channel through which membership in an empire may have beneted international trade is by lowering transactions costs and payments frictions. One way empires reduced payments frictions was by promoting a common language among merchants. Even in cases where the dominant language of the population differed from that of the imperial power, a lingua franca often developed around commercial centres (Ferguson, 2002). Since trade in the eighteenth century and early nineteenth century had been initially organised around principles of mercantilism and imperial preference, all else equal, merchants had a nancial incentive to learn the language of colonial masters in order to sell more goods. Even as the incentive to learn colonial languages receded, as trade relationships changed over time, the process was path dependent: the foreign language of the imperial power continued to be used by merchants. Merchants who had been trading within an empire were already acclimatised to local customs and habits. They had well-established contacts and may have developed social networks as well as distribution and marketing channels for buying and selling goods; this would tend to lower the transactions costs associated with trade. Greif (1997), McMillan (1997) and McMillan and Woodruff (1999) have noted the importance of informal relationships in fostering deals where laws of contract are weak; social networks can, in turn, support contracting and foster trade (Rauch and Trindade, 2002). The historical record suggests that colonial ofcials were often urged to foster ties between locals and merchants.30
30
See Frieden (2006, pp. 901) for discussion of this point.
1822
[NOVEMBER
Finally, empires may have also reduced transaction costs associated with trade via the widespread propagation and use of colonial currencies. As discussed above, previous scholarship on currency unions has highlighted the role of the gold standard in reducing payments frictions, but during this period other currency unions also existed. The Latin Monetary Union (LMU) and Scandinavian Monetary Union (SMU) were included in the gravity models estimated in Section 2 but during our sample period, some colonies also participated in less formal monetary arrangements that functioned like currency unions (The Banker, 1950). Table 4 displays the considerable cross-sectional and times-series variation in the establishment and formation of currency unions within colonial empires during the classical gold standard period. Belgian, Dutch and French colonies generally linked up to their respective metropoles currencies to form a xed-exchange-rate area. The US introduced the dollar in its dependencies after acquiring many of its colonies in the Spanish-American War of 1898. British and German colonies, on the other hand, either joined the sterling or mark block or formed currency unions with other colonies in the region. British colonies in East Africa, for example, formed a silver rupee union with India that also included some areas in East Africa that were members of the German Empire. Brunei, Johore, Labuan and the Straits Settlements also formed a currency union while British colonies in West Africa left the sterling union to form a West African (Silver) Currency Union in 1913. Although many different currency unions were formed within and even across the colonial empires, the pound sterling remained the largest and most important currency during the classical gold standard period. As the pound sterling became the preferred means for settling accounts, countries and colonies began to hold sterling balances with foreign banks, which set up ofces in London, and sterling functioned as the reserve currency of the world. Thus, while previous research points to the gold standard as playing an important role in reducing payments frictions and currencies like the pound sterling were as good as gold, it may have been the case that sterling was in many ways better than gold. It was more convenient in that British exporters and importers preferred to draw and be drawn on in pounds sterling. Investor and trader preference for carrying out transactions in sterling or sterling-denominated bills of exchange meant that it was advantageous for dominions and colonies to also carry out their transactions in sterling.31 3.3. The Effects of Transactions Costs and Trade Policies on Bilateral Trade Before assessing the direct effects of these channels on bilateral trade, we rst assess whether trade policies and transactions costs are correlated with empire in order to provide better insight into this cross-sectional indicator variable. This exercise should help to unpack the empire variable that may bundle several economic policies and political relationships between a metropole and its colony. We do this by estimating a series of simple OLS regressions where we model the determinants of empire. Column 1 of Table 5(a) shows the regression of empire on a constant and the common language variable. Columns 2 to 5 of Table 5(a) show the contribution of including
31 For example, Butlin (1986) has noted that the use of the pound sterling was widespread in Australia and New Zealand.
2008 ]
TRADE AND EMPIRE
1823
Table 4 Empire Currency Unions, 18701913

Dates Belgium Franc Union Belgium Congo Great Britain Gold Rupee Union India Silver Rupee Union Aden Ceylon India Maldives Mauritius Seychelles Somaliland Uganda Sterling Union Australia Bahamas Barbados Bermuda Botswana Canada Cape of Good Hope/South Africa Ceylon Cyprus Falkland Islands Fiji Gambia Gibraltar Gold Coast Ireland Jamaica Malta Natal New Hebrides New Zealand Newfoundland Nigeria Orange Free State Rhodesia St. Helena Sarawak Seychelles Sierra Leone Transvaal Trinidad Straits Union Brunei Johore Labuan Straits West African Currency Union Gambia Gold Coast Nigeria Sierra Leone
18791913
18981913 18701913 18721900 18701897 18701913 18701913 18771913 18891913 18701913 18701913 18701913 18701913 18701913 18701913 18701913 18701913 18701871, 19011913 18781913 18701913 18741913 18701912 18991913 18861912 18701913 18701913 18701913 18821913 18781913 18701913 18701913 18801912 18701913 18911913 18701913 19061913 18701876 18801912 18701913 18701913 19031913 18701913 18701913 18701913 1913 1913 1913 1913
1824
[NOVEMBER
Table 4 Continued
Dates France Franc Union Algeria Guadeloupe Madagascar Martinique New Caledonia New Hebrides Senegal St. Pierre and Miquelon Tunis Germany Mark Union German West Africa German SW Africa Togoland Silver Rupee Union Burundi Rwanda Tanzania Netherlands Guilder Union Indonesia Surinam (Dutch Guiana) United States Dollar Union British Honduras Hawaii Puerto Rico
18701913 18701913 19001913 18701913 18701913 18781913 18701913 18701913 18911913
18861913 19011913 18841913 19041913 19041913 18881913
18701913 18701913
18941913 18981913 18981913
Sources: Pick and Sedillot, 1971; http://www.dollarization.org; The Banker, 1950.
one additional variable to the variable(s) listed in the previous column. (The correlation coefcients are shown in Table 5(b)). Column 1 shows that common language (one measure of transactions costs) is positively correlated with empire and is statistically signicant, although the R-squared in the regression is only 4%.32 Column 2 augments the simple model with the number of years a colony has been a member of an Empire since the conclusion of the Napoleonic Wars. We include this variable to capture a reduction in transactions costs that may have arisen from the long-run relationship between a colony and its metropole, such as familiarity with local customs and culture, the development of distribution and marketing channels, or the formation of social networks all of which may benet trade or its participants gradually, over many decades. This measure of transactions costs has a positive and statistically signicant effect on empire; the t of the model
32 With respect to the colonial empires, settler colonies were coded as having a common language with the metropole. For example, British settler colonies such as Australia, Canada, New Zealand and South Africa were coded as having the same language as Britain. As a result of the coding scheme, the language variable may also capture institutional differences.
2008 ]
TRADE AND EMPIRE
1825
Table 5 The Channels of Empire, 18701913

Independent Variables (a) Regressions Common Language Years in Empire Customs Union Trade Preferences Empire Currency Union Observations R-squared 22,580 0.04 Common Language (b) Correlation Matrix Common Language Years in Empire Customs Union Trade Preferences Empire Currency Union 1 0.16 0.02 0.11 0.32 22,580 0.23 Years in Empire 22,580 0.25 Customs Union 22,580 0.25 Pooled OLS 0.270*** (0.06) Pooled OLS 0.187*** (0.05) 0.011*** 0.00 Pooled OLS 0.195*** (0.05) 0.010*** 0.00 0.519*** (0.10) Pooled OLS 0.189*** (0.05) 0.010*** 0.00 0.533*** (0.10) 0.286*** (0.10) Pooled OLS 0.028 (0.02) 0.003*** (0.00) 0.182* (0.10) 0.281** (0.13) 0.782*** (0.03) 22,580 0.53
Trade Preferences
Empire Currency Union
1 0.16 0.18 0.46
1 0.02 0.21
1 0.25
improves from 4% to more than 23%. Columns 3 and 4 augment the model with measures for customs unions and trade preferences within empires.33 The two variables take a value of 1 if country/colony i and j were both members of the same customs union or had a preferential trade agreement with each other. The results show that two trade policy variables both signicantly predict empire but do not substantially improve the t of the model. Column 5 adds a currency union variable for empire countries.34 The currency union variable substantially improves the t of the model. Despite the admittedly parsimonious specication, the trade policy and transactions costs variables explain more than 50% of the variation in the empire indicator variable. Since the transactions costs and trade-policy variables are highly correlated with empire and also seem to capture a signicant amount of the variation in empires across colonies, we now consider the extent to which these variables mattered for bilateral trade during the period 18701913. We replace the empire indicator variable used in our earlier empirical models with the transactions costs and trade-policy variables described above. Table 6 reports country and dyad xed-effects models. Consistent with the results shown in earlier Tables, countries that are more distant trade less. Bilateral trade between two countries and/or colonies is increasing with miles of railway track and population, and countries that border each other or are members of the gold club have greater bilateral trade ows.
Customs unions and trade preferences are based on information from US Tariff Commission (1922). The currency union variable for empire countries was coded using Pick and Sedillot (1971), http:// www.dollarisation.org, and The Banker (1950).
34 33
1826
[NOVEMBER
Table 6 The Effects of Empire on Trade, 18701913

Independent Variables Gold Standard Distance Border Number Landlocked Railway Track Population Common Language Years in Empire Customs Union Trade Preferences Empire Currency Union Non-Empire Currency Union War War(-1) War(-2) War(-3) War(-4) War(-5) War(-6) Year Dummies Observations R-squared Country Fixed Effects 0.137*** (0.03) 0.747*** (0.02) 0.498*** (0.04) 0.875*** (0.10) 0.101*** (0.01) 0.208*** (0.02) 0.198*** (0.04) 0.010*** (0.00) 0.364*** (0.12) 0.852*** (0.07) 1.490*** (0.06) 0.665*** (0.08) 0.07 (0.21) 0.403 (0.28) 0.252 (0.21) 0.203 (0.21) 0.035 (0.24) 0.012 (0.26) 0.106 (0.23) No 21,630 0.7 Country Yr. Fixed Effects 0.135*** (0.519) 0.777*** (0.0153) 0.602*** (0.044) 0.912*** (0.134) 0.365 (1.094) 0.274** (0.203) 0.270*** (0.040) 0.008*** (0.001) 0.828*** (0.148) 0.986*** (0.917) 1.571*** (0.518) 0.656*** (0.092) 0.058 (0.364) 0.254 (0.354) 0.267 (0.356) 0.210 (0.351) 0.163 (0.358) 0.058 (0.350) 0.656 (0.092) Yes 21,630 0.7 Dyad Fixed Effects 0.360*** (0.02) Dyad Yr. Fixed Effects 0.155*** (0.02)
0.143*** (0.00) 0.219*** (0.01) 0.001 (0.00) 0.293*** (0.10) 0.320*** (0.06) 0.199*** (0.07) 0.244** (0.10) 0.181 (0.16) 0.211 (0.16) 0.077 (0.16) 0.017 (0.16) 0.061 (0.16) 0.078 (0.16) 0.037 (0.16) No 21,630 0.17
0.030*** (0.01) 0.064*** (0.01) 0.004*** (0.00) 0.163 (0.10) 0.235*** (0.05) 0.282*** (0.07) 0.066 (0.10) 0.162 (0.15) 0.23 (0.16) 0.127 (0.16) 0.023 (0.16) 0 (0.16) 0.003 (0.16) 0.043 (0.15) Yes 21,630 0.22
The results in Table 6 also show that many of the channels we have quantied help to account for the observed variation in bilateral trade across countries and colonies. First, the empirical estimates suggest that membership in a preferential trade agreement raised bilateral trade ows by 26% to 168% (depending on the econometric specication). The coefcient on the customs union variable is also economically and statistically signicant in three out of the four specications. Membership in a customs union increased bilateral trade ows by 18% to 129%. This result provides a reason
2008 ]
TRADE AND EMPIRE
1827
why, as shown in Table 2, the empire effect differed for Germany, France and Britain. Metropoles differed in the nature and extent to which they set up explicit trade policies for their colonies.35 Table 6 also suggests that empire boosted trade by lowering transactions costs. Being part of an empires currency union signicantly increased trade by 22% to 381%, depending on which xed-effects model was used. The currency union effect in colonial empires may also be capturing the impact of monetising many African colonies that historically traded very little with the rest of the world and largely relied on barter to exchange goods prior to colonisation. Countries and colonies that transacted in a common tongue also exhibited greater bilateral trade approximately 20% more than those that did not. This result suggests that a lingua franca was another way in which empires may have lowered transactions costs and boosted trade. The number of years that a colony has been part of an empire is also statistically signicant in two of the four bilateral trade regressions, although the sign varies according to which specication is used. One interpretation of the positive relationship exhibited in the country-xed-effects specications is that trade in British Dominions, such as Australia, Canada and New Zealand beneted most from strong social networks and shared customs, whereas newer colonies (many of which were in Africa) had a different cultural heritage and therefore beneted less; these would constitute the remaining colonies included the two dyad xed-effects regressions. It may be impossible to test empirically all of the ways in which empire impacted trade. The channels that we have identied nevertheless account for a signicant amount of the cross-sectional variation in empire and help to shed additional light on the empire effect reported earlier in the article.36
4. Conclusion
What is the impact that empires have on trade ows? Do metropoles use their political control to tilt the balance of trade through preferential trade policies or customs unions? Do they lower transactions costs by standardising language and currency and creating social networks? We provide some perspective on these questions using a new database of over 21,000 bilateral trade pairs collected from primary sources for the period 18701913. We nd strong empirical evidence that membership in an empire more than doubled bilateral trade during the Age of High Imperialism. Moreover, it was quantitatively more important in explaining bilateral trade ows during the rst era of globalisation than either membership in the gold standard or a non-empire
35 It is possible that preferential trading agreements may have changed the way in which countries oriented their economies towards producing goods for the metropoles. If this were the case, then colonial tariff policies may have lowered the prospects of long-term economic development, as suggested by Lewis (1970). As we have indicated throughout, the focus of this article is on analysing the determinants of bilateral trade ows and the impact of empire on international exchange rather than on constructing a counterfactual of trade in the absence of empire. 36 We also tested for the possibility that metropoles may have increased bilateral trade by building out or nancing infrastructure development in their colonies. The infrastructure variable, dened as the interaction between empire and railways was only statistically signicant in the xed-effects model. However, the coefcient on infrastructure in the xed-effects model was not economically signicant; empire railways increased bilateral trade ows by approximately 0.2%.
1828
[NOVEMBER
currency union. This empire effect appears to be robust to a variety of econometric specications, including instrumental variables regressions, xed-effects specications that account for multilateral resistance, propensity score estimates, and a number of selection models. We suggest two broad channels through which empire may have boosted trade during this period: transactions costs and trade policies. Our empirical ndings suggest that membership in an empire-currency union, sharing a common language, preferential trading agreements and customs unions were important in accounting for the observed variation in bilateral trade ows. Moreover, variation in colonial trade policies and currency-union arrangements helps to explain why the effect that empires have on trade differ. Empire currency unions were especially prevalent in the British, French and German Empires, while preferential trading agreements were widely used by France and some of the British Dominions. Although we have not fully accounted for all the channels through which empire may have impacted trade, transactions costs and trade policies account for over 50% of the cross-sectional variation in empire and appear to have played a signicant role in boosting trade within empires during the Age of High Imperialism. An interesting avenue for future research would be to analyse how the positive relationship between empire and trade impacted productivity and economic growth.37 For example, metropoles may have increased productivity by creating free trade zones that promoted competition and intercolonial trade. The free trade systems established by empires may have promoted specialisation within colonies and increased their productivity. This would be consistent with a model developed by Alcala and Ciccone (2004) where the greatest impact of trade on productivity occurs in the traded goods sector rather than the non-traded, service sector through a Balassa-Samuelson effect.38 In creating the machinery for trade (that sometimes included a new production technology and an enhanced market with a quasi-monopsonist buyer of exports), empires may have also imparted institutions that either fostered or undermined productivity and growth. Some scholars have argued that trade can transform political institutions and foster the development of property rights, which in turn can lead to greater investment and growth as it did in the North Atlantic economies between 15001850 (Acemoglu et al., 2005). Some settler colonies institutions, for example, may have benetted favourably from the trade and openness that the British empire promoted. On the other hand, empires may have undermined long-run productivity and growth by leaving extractive institutions, such as those suggested by Acemoglu et al. (2001). Understanding the relationship between empire and trade and long-run outcomes such as productivity and growth is complicated by the fact that the institutional
37 For an analysis of the impact of trade and growth or income levels, see Frankel and Romer (1999) and Irwin and Tervio (2002). 38 The Balassa-Samuelson effect can cause changes in the relative price of non-tradable goods, which introduces a bias in the use of nominal openness to measure the productivity gains from trade. Alcala and Ciccone employ a measure of real openness, measured as imports plus exports in exchange rate US$ divided by GDP in purchasing power parity US$, to control for cross-country differences in the relative price of nontradable goods.
2008 ]
TRADE AND EMPIRE
1829
footprints of empires varied, not only across empires and colonies but even within colonies. For example, some colonies exports were produced in very controlled plantations systems whereby colonists owned and controlled the land and capital that coffee, sugar, rubber or other crops were grown on and employed low-wage, local labour in the production of these commodities. Other tradables sectors of the same colony may have been left untouched. This raises questions about how such parallel specialisation impacted the growth prospects of these economies. To assess the long-run impact of trade and empire on productivity and growth thoroughly, future research will need to examine the institutional variation within and across colonies as well as the changes in trade relations that took place after independence.
Appendix
Table A1 The Effects of Empire on Trade including GDP Estimates
Independent Variables Empire Membership Distance Gold Standard Border Number Landlocked GDP Population Railway Non-Empire Currency Union War War(1) War(2) War(3) War(4) War(5) War(6) Year Dummies Observations R-squared Pooled OLS 0.952** (0.41) 0.510*** (0.10) 0.169 (0.14) 0.861*** (0.20) 0.909** (0.44) 0.450*** (0.14) 0.012 (0.11) 0.204*** (0.05) 1.214** (0.50) 0.421* (0.22) 0.283** (0.12) 0.333* (0.18) 0.237 (0.20) 0.029 (0.24) 0.06 (0.30) 0.003 (0.29) Yes 6,671 0.49 Dyad Yr. Fixed Effects 0.961*** (0.10) 0.115*** (0.02) Country Fixed Effects 2.155*** (0.09) 0.437*** (0.02) 0.051 (0.04) 0.856*** (0.05) 0.630*** (0.07) 0.199 (0.13) 0.072*** (0.02) 2.005*** (0.12) 0.281 (0.26) 0.287 (0.21) 0.363 (0.22) 0.253 (0.25) 0.096 (0.28) 0.005 (0.29) 0.107 (0.26) No 6,671 0.68 Country and Yr. Fixed Effects 2.175*** (0.09) 0.445*** (0.02) 0.034 (0.04) 0.852*** (0.05) 0.291*** (0.10) 0.189 (0.13) 0.034* (0.02) 2.023*** (0.12) 0.254 (0.25) 0.253 (0.21) 0.368 (0.23) 0.248 (0.25) 0.084 (0.29) 0.016 (0.30) 0.118 (0.26) Yes 6,671 0.68
0.368*** (0.05) 0.02 (0.07) 0.072*** (0.01) 0.116 (0.23) 0.02 (0.14) 0.115 (0.15) 0.266* (0.15) 0.07 (0.15) 0.007 (0.15) 0.006 (0.15) 0.047 (0.14) Yes 6,671 0.41
Robust standard errors in parentheses. * Signicant at 10%; ** Signicant at 5%; *** Signicant at 1%. The Author(s). Journal compilation Royal Economic Society 2008
1830
[NOVEMBER
Table A2 Propensity Score

Variable Gold Standard Number Landlocked Distance Border Railway Track Population Common Language Number of Observations Pseudo R-squared Note: Propensity score estimated using a probit model. Estimation includes a constant term. *** Indicates signicant at the 1% level. Full Sample 0.347*** (0.023) 0.800*** (0.079) 0.439*** (0.015) 0.223*** (0.060) 0.073*** (0.004) 0.193*** (0.006) 0.661*** (0.033) 21,686 0.301
Table A3 Reduction in Bias Using Propensity Score

Full Panel Variable Number Landlocked Distance Border Railway Track Population Gold Standard Common Language Sample Unmatched Matched Unmatched Matched Unmatched Matched Unmatched Matched Unmatched Matched Unmatched Matched Unmatched Matched Treated 0.009 0.009 8.072 8.092 0.044 0.044 12.422 12.722 14.884 15.102 0.634 0.631 0.256 0.243 Control 0.057 0.008 7.779 8.031 0.099 0.045 15.745 12.036 18.244 15.102 0.489 0.637 0.090 0.206 Bias 27.1% 1.7% 32.1% 7.3% 21.5% 0.5% 87.6% 18.1% 120.1% 0.0% 29.5% 1.3% 44.9% 9.9% Reduction in bias 97.5% 77.4% 97.8% 79.3% 100.0% 95.6% 78.0%
2008 ]
TRADE AND EMPIRE
1831
Table A4 Non-parametric Estimates of the Empire Effect

Matching Method Nearest Neighbour Stratication Radius (Maxium Distance of Control)
Full Panel 1 3 0.03 0.01 0.005 Average Treatment on the Treated 1.274 1.247 1.188 1.291 1.301 1.302 (Standard Error) (0.120) (0.086) (0.051) (0.084) (0.076) (0.075) (t statistic) 10.630 14.55 23.443 15.410 17.180 17.320 Percentage Expansion 258 248 228 264 267 268 of Trade due to Empire Number of Observations 21,686 21,686 21,686 21,559 21,505 21,492 Treated 5,213 5,213 5,016 5,086 5,032 5,019 Untreated 16,473 16,473 16,670 16,473 16,473 16,473
Table A5 Selection Models of Trade

Independent Variables Empire Membership Distance Gold Standard Border Number Landlocked Railway Track Population Non-Empire Currency Union War War(1) War(2) War(3) War(4) War(5) War(6) Observations Two-Step 4.113*** (0.14) 0.813*** (0.02) 0.146*** (0.03) 0.517*** (0.06) 0.11 (0.07) 0.226*** (0.01) 0.532*** (0.01) 0.767*** (0.11) 0.045 (0.39) 0.325 (0.40) 0.163 (0.40) 0.088 (0.40) 0.092 (0.40) 0.151 (0.40) 0.129 (0.39) 21,630 Maximum Likelihood 2.166*** (0.28) 0.637*** (0.07) 0.019 (0.10) 0.628*** (0.21) 0.176 (0.23) 0.199*** (0.03) 0.418*** (0.03) 0.701** (0.33) 0.118 (0.20) 0.392* (0.23) 0.214 (0.15) 0.146 (0.16) 0.032 (0.19) 0.097 (0.25) 0.047 (0.23) 21,630
Standard errors in parentheses. * Signicant at 10%; ** Signicant at 5%; *** Signicant at 1%.
1832
[NOVEMBER
Table A6 The Effects of Empire on Trade with Restricted Sample, 18701913 (Sample includes only countries/colonies that changed empire status during sample period)
Independent Variables Empire Membership Distance Gold Standard Border Number Landlocked Railway Track Population Non-Empire Currency Union War War(1) War(2) War(3) War(4) War(5) War(6) Year Dummies Observations R-squared 0.156 (0.13) 0.045** (0.02) 0.775*** (0.11) 1.106*** (0.17) 0.422 (0.65) 3.767*** (0.65) 0.152 (0.65) 0.191 (0.65) 0.505 (0.65) 0.505 (0.65) 0.599 (0.65) No 552 0.4 0.314** (0.14) 0.048*** (0.02) 0.770*** (0.14) 0.861*** (0.18) 0.477 (0.64) 3.593*** (0.64) 0.464 (0.64) 0.37 (0.64) 0.995 (0.63) 0.968 (0.63) 1.106* (0.63) Yes 552 0.53 Dyad Fixed Effects 0.766*** (0.09) Dyad Yr. Fixed Effects 0.726*** (0.08) Country Fixed Effects 0.755*** (0.10) 0.615*** (0.16) 0.182 (0.12) 0.045** (0.02) 0.764*** (0.14) 1.088*** (0.19) 0.405*** (0.14) 3.785*** (0.14) 0.134 (0.14) 0.172 (0.14) 0.512*** (0.10) 0.512*** (0.10) 0.606*** (0.10) No 552 0.93 Country and Yr. Fixed Effects 0.717*** (0.09) 0.677*** (0.14) 0.335** (0.13) 0.049*** (0.02) 0.769*** (0.15) 0.847*** (0.16) 0.471** (0.22) 3.598*** (0.21) 0.458* (0.25) 0.363* (0.21) 1.006*** (0.19) 0.978*** (0.17) 1.112*** (0.18) Yes 552 0.94
Standard errors in parentheses. * Signicant at 10%; ** Signicant at 5%; *** Signicant at 1%.
Santa Clara University and NBER Claremont McKenna College and NBER Submitted: 5 March 2007 Accepted: 12 January 2008
References
Accominotti, O. and Flandreau, M. (2005). Does bilateralism promote trade? Nineteenth century liberalization revisited, CEPR Discussion Paper No 5423. Acemoglu, D., Johnson, S. and Robinson, J. (2001). The colonial origins of comparative development: an empirical investigation, American Economic Review, vol. 91(5) (December), pp. 1369401. Acemoglu, D., Johnson, S. and Robinson, J. (2005). The rise of Europe: Atlantic trade, institutional change, and economic growth, American Economic Review, vol. 95(3) (June), pp. 54679. The Author(s). Journal compilation Royal Economic Society 2008
2008 ]
TRADE AND EMPIRE
1833
Alcala, F. and Ciccone, A. (2004). Trade and productivity, Quarterly Journal of Economics, vol. 119(2) (May), pp. 61346. Alesina, A. (2002). The size of countries: does it matter?, Harvard University, Working Paper. Anderson, J. and Wincoop, E. (2003). Gravity with gravitas: a solution to the border puzzle, American Economic Review, vol. 93(1) (March), pp. 17092. Bae, K. and Bailey, W. (2003). The Latin Monetary Union: some evidence on Europes failed common currency, Cornell University, Working Paper. Baier, S. and Bergstrand, J. (2007). Do free trade agreements actually increase members international trade?, Journal of International Economics, vol. 71(1) (March), pp. 7295. Bairoch, P. (1989). European trade policy 18151914, in: (P. Mathias and S. Pollard, eds.), The Cambridge Economic History of Europe, vol. 8: The Industrial Economies: The Development of Economic and Social Policies. pp. 558, Cambridge: Cambridge University Press. Baldwin, R. (2005). The euros trade effects, European Central Bank, Working Paper. Banks, A. (1976). Cross-national time series: 18151973, computer le. Inter-university Consortium for Political and Social Research [producer and distributor], Ann Arbor, MI. Barbieri, K. (1996). Economic interdependence and militarised interstate conict, 18701985, Ph.D. Dissertation, Binghamton University. British Board of Trade. Various years. Statistical Abstract for the Principal and Other Foreign Countries, London: Board of Trade. British Board of Trade. Various years. Statistical Abstract for the Several British, Colonies, Possessions, and Protectorates, London: Board of Trade. British Board of Trade. Various years. Statistical Abstract for the United Kingdom. London: Board of Trade. Butlin, S. (1986). The Australian Monetary System, Sydney: Ambassador Press Pty Ltd. Clemens, M. and Williamson, J.G. (2004). Wealth bias in the rst global capital market boom, ECONOMIC JOURNAL, vol. 114(495) (April), pp. 30447. Deardorff, A.V. (1998). Determinants of bilateral trade: does gravity work in a neoclassical world?, in (J. Frankel, ed.), The Regionalization of the World Economy, pp. 728, Chicago: University of Chicago Press. Edison, H., and Melvin, M. (1990). The determinants and implications of a choice of exchange rate regime, in: (W. Haraf and T.D. Willett, eds.), Monetary Policy for a Volatile Global Economy, pp. 144, Washington: AEI Press. Estevadeordal, A., Frantz, B. and Taylor, A.M. (2003). The rise and fall of world trade, 18701939, Quarterly Journal of Economics, vol. 118(2) (May), pp. 359407. Evenett, S.J. and Keller, W. (1998). On theories explaining the success of the gravity equation, NBER Working Paper No. 6529. Feenstra, R.C., Markusen, J. and Rose, A.K. (1998). Understanding the home market effect and the gravity equation: the role of differentiating goods, NBER Working Paper No. 6804. Ferguson, N. (2002). Empire: The Rise and Demise of the British World Order and the Lessons for Global Power, New York: Basic Books. Ferguson, N. (2003). British imperialism revised: the costs and benets of Anglobalization, Development Research Institute Working Paper No. 2, New York University. Ferguson, N. (2004). Colossus: The Price of American Empire, New York: Penguin Press. Ferguson, N. and Schularick, M. (2004). The empire effect: the determinants of country risk in the rst age of globalization, 18801913, Journal of Economic History, vol. 66(2) (June), pp. 283312. Flandreau, M. (2000). The economics and politics of monetary union: a reassessment of the Latin Monetary Union, 186571, Financial History Review, vol. 7(1) (April), pp. 2543. Flandreau, M. and Maurel, M. (2001). Monetary union, trade integration, and business cycles in 19th century Europe: just do it, CEPR, Discussion Paper No. 3087. Flandreau, M. and Zu mer, F. (2004), The Making of Global Finance, Paris: OECD. Frankel, J.A. and Romer, D. (1999). Does trade cause growth, American Economic Review, vol. 89(3) (June), pp. 37999. Frieden, J.A. (2006). Global Capitalism: Its Fall and Rise in the Twentieth Century, New York: W.W. Norton & Company. Glick, R. and Rose, A.K. (2002). Does a currency union affect trade? The time-series evidence, European Economic Review, vol. 46(6) (June), pp. 112551. Glick, R. and Taylor, A.M. (2006). Collateral damage: trade disruption and the economic impact of war, NBER Working Paper No. 11565. Greif, A. (1997). Contracting, enforcement, and efciency economics beyond the law, in: (M. Bruno and B. Pleskovic, eds.), Annual World Bank Conference on Development Economics, Washington: World Bank. Hobson, J.A. (1902). Imperialism: A Study, London: Nisbet. Irwin, D. and Tervio, M. (2002). Does trade raise income? Evidence from the twentieth century, Journal of International Economics, vol. 58(1) (October), pp. 118.
1834
[ N O V E M B E R 2008 ]
Lal, D. (2001). Globalization, imperialism, and regulation, Cambridge Review of International Affairs, vol. 14(1), pp. 10721. Lal, D. (2004). In Praise of Empires: Globalization and Order, New York: Palgrave Macmillan. Lenin, V.I. (1916). Imperialism: The Highest Stage of Capitalism, Moscow: Foreign Languages Publishing House. Lewis, W.A. (1970). Tropical Development, 18801913, London: George Allen and Unwin. Lopez-Cordova, J.E. and Meissner, C. (2003). Exchange-rate regimes and international trade: evidence from the classical gold standard era, American Economic Review, vol. 93(1) (March), pp. 34453. Maddison, A. (1995). Monitoring the World Economy, 18201992, OECD: Paris. McMillan, J. (1997). Markets in transition, in (D. Kreps and K. Wallis, eds.), Advances in Economics and Econometrics: Theory and Applications, pp. 21039, Cambridge: Cambridge University Press. McMillan, J. and Woodruff, C. (1999). Interrm relationships and informal credit in Vietnam, Quarterly Journal of Economics, vol. 114(4) (November), pp. 1285320. Meissner, C. (2005). A new world order: explaining the emergence of the classical gold standard, Journal of International Economics, vol. 66(2) (October), pp. 385406. OBrien, P.K. (1991). Oxford Atlas of World History, New York: Oxford University Press. Ofcer, L. (2004). The gold standard, EH-Net Encyclopedia Entry. http://www.eh.net/encyclopedia/? articleofcer.gold.standard. Olson, J.S. (1991). Historical Dictionary of European Imperialism, New York: Greenwood Press. ORourke, K.H. and Williamson, J.G. (1999). Globalization and History: The Evolution of a Nineteenth-Century Atlantic Economy, Cambridge: MIT Press. Persson, T. (2001). Currency unions and trade: how large is the treatment effect?, Economy Policy, vol. 33 (October), pp. 43348. Pick, F. and Sedillot, R. (1971). All the Monies of the World, New York: Crown Advertising Press, Inc. Porter, A. (1999). The Oxford History of the British Empire, Vol. III, The Nineteenth Century, Oxford: Oxford University Press. Rauch, J.E. and Trindade, V. (2002). Ethnic Chinese networks in international trade, Review of Economics and Statistics 84 (February), pp. 11630. Ritschl, A. and Wolf, N. (2003). Endogeneity of currency areas and trade blocs: evidence from the inter-war period, CEPR discussion paper 4112. Rose, A.K. (2000). One money, one market: the effect of common currencies on trade, Economic Policy, vol. 15(1) (April), pp. 733. Rose, A.K. (2002). Do we really know that the WTO increases trade?, NBER Working Paper No. 9273. Rose, A.K. (2004). A meta-analysis of the effect of common currencies on international trade, NBER working paper 10373. Sarkees, M.R. (2000). The correlates of war data on war: an update to 1997, Conict Management and Peace Science, vol. 18(1), pp. 12344. Tableau General du Commerce Exterieur. Various years. France: Direction Generale des Douanes. The Banker (1950). Monetary Systems of the Colonies: A Fully Revised Reprint of Eight Authoritative Articles which Appeared Between July 1948 and February 1949 in The Banker, London: The Banker. Townsend, M. E. (1966). The Rise and Fall of Germanys Colonial Empire 18841918, New York: Howard Fertig. United States Tariff Commission. (1922). Colonial Tariff Policies, Washington: Government Printing Ofce.

Trade and Empire

Uploaded by

Copyright:

Available Formats

Trade and Empire

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Trade and Empire

Uploaded by

Copyright:

Available Formats

The Economic Journal, 118 (November), 18051834. The Author(s). Journal compilation Royal Economic Society 2008.

TRADE AND EMPIRE*

THE ECONOMIC JOURNAL

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

2. Analysis of Bilateral Trade Flows, 18701913

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

Table 1 The Effects of Empire on Trade, 18701913

TRADE AND EMPIRE

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

3. What Explains the Empire Effect?

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

Table 3 Colonial Tariff Systems

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

Spain United States Puerto Rico

Canary Islands Spanish Morocco American Samoa Canal Zone

Source: US Tariff Commission (1922).

See Frieden (2006, pp. 901) for discussion of this point.

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

Table 4 Empire Currency Unions, 18701913

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

18701913 18701913 19001913 18701913 18701913 18781913 18701913 18701913 18911913

18861913 19011913 18841913 19041913 19041913 18881913

18941913 18981913 18981913

Sources: Pick and Sedillot, 1971; http://www.dollarization.org; The Banker, 1950.

The Author(s). Journal compilation Royal Economic Society 2008

TRADE AND EMPIRE

Table 5 The Channels of Empire, 18701913

Empire Currency Union

1 0.16 0.18 0.46

The Author(s). Journal compilation Royal Economic Society 2008

THE ECONOMIC JOURNAL

Table 6 The Effects of Empire on Trade, 18701913