Maritime flows have been carrying the bulk of global trade for centuries. They can be seen as a crucial benchmark of the vitality of the world economy as a whole and of its components such as port cities and maritime regions at various scales (Vigarié, 1968). Being one of the oldest forms of human interaction, maritime flows are good indicators of economic circulation and a useful tool to "take the pulse of world trade and movement" (Ullman, 1949). It is well-known that most global cities are or have been global ports (Dogan, 1988; Keeling, 1995). However, the few existing empirical works providing a clear snapshot of global maritime flows at different historical periods were more interested by other aspects such as past climatic conditions (Herrera et al., 2003), the diffusion of knowledge through shipping (Dedieu et al., 2011), the environmental impacts of shipping (Halpern et al., 2008), bioinvasions (Kaluza et al., 2010), and the measurement of a composite global urban accessibility index including maritime flows (Nelson, 2008).
Applying a maritime perspective to world history and world regionalization thus remains an unaccomplished project (Lewis and Wigen, 1999) although the maritime foundations of the so-called world systems (Wallerstein, 1979; Braudel, 1985) need not being demonstrated any further. It has been somewhat forgotten with the radical competition of air transport for passenger flows, the drastic social and urban impacts of changing port and shipping technologies, and the growing preference for other flows in academic research (Hall and Hesse, 2012). More likely in this perspective are studies of other global networks such as those shaped by airlines, multinational firms, trade, migration, knowledge, and communication flows (Van Hamme and Patris, 2011; Ducruet and Lugo, 2013).
Although there is only one data source providing detailed information, on a weekly basis and on a global level, about flows of merchant vessels between ports of the world, it has, surprisingly, never been used or mapped systematically. There is only one mention of this source in the whole academic literature, by a geographer (Rees, 1955), where the author promoted the usefulness of Lloyd's List "as a source for port study in school". It is only in the late 1990s that such a large-scale approach was proposed in order to reveal the polycentric organization of global maritime flows, based on Lloyd's Voyage Record (Joly, 1999). This pioneering work was pushed further a decade later when mapping the changing spatial pattern of North Korea's port activities and overseas connections over the 1985-2005 period (Ducruet et al., 2009), the liner shipping network and port hierarchy of Northeast Asia (Ducruet et al., 2010a), the Atlantic (Ducruet et al., 2010b) and the world (Ducruet and Notteboom, 2012), the analysis of regional subsystems in this network (Ducruet and Zaidi, 2012), its combination with airline flows forming a global urban hierarchy (Ducruet et al., 2011), and the intermingling of various commodity types in the multigraph (Ducruet, 2013). Other studies of the kind have emerged in parallel, mostly focusing on graph properties (Deng et al., 2009; Hu and Zhu, 2009) and using newly available sources to study recent flows.
Undoubtedly, the potential offered by Lloyd's List to analyze the long-term evolution of cities and their global interactions remains by far unexplored. It is, perhaps, the only source capable of documenting in concrete terms globalization, regionalization, and urbanization processes over nearly three hundred years. The current paper is a continuation of recent attempts to map and analyze global maritime flows in the late nineteenth and early twentieth centuries (Ducruet, 2012; Ducruet and Marnot, 2013). Four complete paper issues of about 300 pages each have been extracted manually into a global database for the years 1890, 1925, 1961, and 2004. One first approach is to apply conventional graph theoretical measures to the respective global maritime networks in order to reveal the topological structure of flows and its evolution over time. Secondly, we map such flows at two levels, continents and ports, to analyze major shifts of hierarchy and centrality in the network. How is the current distribution of port activity path-dependent, and can we identify key evolutions at local and global levels? How does port activity reflect wider aspects of economic life under different contexts? Have some port cities been more resilient than others over time?
Network properties of global maritime flows
Lloyd's List materials contain crucial information on vessels and their movements. The analysis of flows is thus based on traffic data by port pairs. Each time a vessel operated a voyage between two ports, one call was added to each pair and each individual port. The first result is thus a global O-D matrix of inter-port maritime flows basing the analysis of a non-planar and weighted (i.e. calls, tons) network. Directionality was ignored for the sake of simplicity, while other aspects such as vessel registry, flag (nationality), dates of departure and arrival, and type of vessel (e.g. steamer, sailing vessel, tanker, containership) are kept for further analyses in ulterior researches.
As seen in Table 1, the number of vessels and ports remains comparable between 1890, 1925, and 1961 but reaches unprecedented levels in 2004. However, the number of vessel calls and links already witnessed a steady growth by 1961. Those figures are good indicators of the acceleration of global trade activity over the last century, fuelled by technological innovations and revolutions in maritime transport and a parallel reduction of maritime transport costs. The world fleet in 1890 still comprised a majority of sailing vessels, but the proportion of steamer vessels has increased from 38% in 1890 to 96% in 1925. Postwar trends affecting world shipping are well-known, such as, since the early 1960s, the gradual spread of containerization, the shift of manufacturing towards less-developed countries acquiring their independence and becoming "new" trade partners.
Increasing connectivity is revealed in various ways, such as the average number of links by port (degree centrality and also Beta index) and the completeness of the network (Gamma and Alpha indices). Ports have become more interconnected on average compared with the maximum possible connectivity. The total kilometric length of the network has grown tremendously (measured in orthodromic distance), with the effect of placing ports farther topologically (eccentricity), although in reality, each port is more easily accessible from any other port due to the increased speed and frequency of vessel trips. The average clustering coefficient presents an increasing proportion of cliques between 1890 and 1925, but this proportion has decreased since then in favor of more concentration. This is well reflected in the growing concentration of the port hierarchy of vessel calls, measured in various ways, although Gini and Herfindhal indices point at a decreasing concentration. In fact, concentration and de-concentration dynamics are not incompatible: larger ports increasingly dominate the global hierarchy but there is an increasing number of larger ports over time.
Table 1: Network properties of global maritime flows, 1890-2004
The spatial distribution of global maritime flows has also changed drastically when looking at the proportion of flows over links of certain length (Figure 1). In particular in 1890, most traffic concentrates over links up to 3500 kilometers or less (80%), while in 1925, the same proportion is reached over links up to 5500 kilometers or less. This is, again, a good illustration of the expansion of world trade routes, backed by the diffusion of steamers and the success of Suez (1869) and Panama (1914) canals enabling direct access among distant regions. The trend is not linear, however, because shorter links have regained importance in 1961 and 2004; yet, 2004 is the year where longest links have the heaviest share compared with previous periods. The explanation is a concomitant process of globalization and regionalization, whereby long-range links and proximity links both increase their importance.
Figure 1: Vessel traffic distribution over distance, 1890-2004
Interregional flows: from a core-periphery to a polycentric structure
The most salient feature of the flows' architecture is the overwhelming and rather stable dominance of Europe over the whole period (Figure 2), concentrating 48, 53, 39, and 43% of vessel movements at respective years. Latin and North America have witnessed a gradual decline from 21 and 15% respectively in 1890 to 11 and 10% respectively in 2004. Asia has been the fastest growing region, increasing its share from 7% in 1890 to 22% in 2004, followed by Africa, which shifted from 4 to 9%.
The geographic distribution of flows has thus clearly moved from a system centered upon the Atlantic (Konvitz, 1994) to a truly global system with a more homogenous coverage. While the Europe-Latin America route was the strongest link in 1890, due to the commercial importance of the West Indies and Argentina for European trade, the only important route not connected with Europe was the inter-American link, in a context of growing U.S. interests for Latin American resources (Marnot, 2012). One major difference in 1925 is the increase of all Asia-related routes, and the decrease of Latin America that reaches an equal weight with Asia, compared with its threefold superiority in 1890. Reinforced colonial interests of Europe towards Asia are the main cause of this shift, backed by the opening of Suez Canal (1869) and the deployment of steamer vessels along the Europe-Asia route. Already in 1961, Asia has become the second largest pole after Europe, and it is still the case nowadays when measured by the number of vessel calls. Tonnage figures even place Asia at the first rank in recent years, but vessel calls provided the most comparable unit for measuring traffics over the period. Another important change visible in 1961 is the shift of Oceania under Asian polarization, but while the weight of this link is just above the one with Europe in 1961, it has become almost four times heavier with Asia than with Europe in 2004. Africa's second largest link has been with Asia since 1925 (see also Appendix 1).
The proportion of intraregional flows in the total of flows by region shows that Europe has always been the most integrated area, if one considers such figures as accurate indicators of regional integration levels. This occurs despite the exclusion of coastal shipping flows in 1890 and 1925 by Lloyd's List: deep-sea vessel movements still reveal strong internal ties among European ports more than anywhere else. Despite some fluctuations, Latin America and Asia also witnessed important intraregional activity, probably due to the ease of circulation between different sub-regions such as the Caribbean, Brazil/Argentina, East Asia, and the Middle East/Indian Ocean.
All in all, the analysis at continent level confirms the evolution of the global maritime network from a mono-centric to a polycentric structure. There is a strongly path-dependent evolution by which Europe maintains its prominence despite profound reconfigurations of global geopolitics.
Figure 2: Interregional maritime flows, 1890-2004
Hierarchy and growth trajectory of port cities
On the level of port cities, the aforementioned global structure becomes clearer. British cities as well as New York and Buenos Aires are the largest ports by far in 1890, followed by a handful of other large ports in Europe (Hamburg, Antwerp, Marseilles, Genoa), North America (Pensacola, San Francisco), Latin America (Rio de Janeiro, Barbados, Montevideo), Asia (Calcutta, Bombay, Rangoon, Hong Kong), and Oceania (Melbourne, Sydney). The Rio de la Plata is at the time an emerging rival of Western powers economically and culturally during this era of prosperity (1880-1930), fuelled by massive European immigration, railroad development, and the export of meat and other products. Its enormous importance has somewhat faded away already in 1925, while Asia as a whole has grown substantially, with the reinforcement of European colonial comptoirs (Indian Empire, Indonesia) and the rise of Japan as seen with the now equivalent activity of Yokohama and Kobe. The rapid industrialization of Japan has resulted in very high traffic growth rates between 1890 and 1925 (Yokohama 340%, Kobe 270%), far beyond the more traditional Asian port cities (Bombay 69%, Singapore 32%, Batavia 14%). Internally, there has been very drastic shifts, such as from the British Isles towards the European continent, with the emergence of the so-called "North European Range" from Le Havre to Hamburg (Vigarié, 1964), and the concentration of the U.S. East Coast traffics around the Bostwash megalopolis (Gottmann, 1961) notwithstanding the growth of Los Angeles and Seattle-Vancouver on the West Coast.
Decline is apparent at Europe's Iberian Peninsula as well as throughout Latin America, partly due to their continued reliance on sailing vessels and to the Asia-Pacific shift (Arrault, 2008). Those trends are prolonged in 1961 but one can observe, in Asia, the fast-growing position of Hong Kong and Singapore, Asia's most modern ports at the time together with Japan (Lee et al., 2008). The activity of Chinese ports remains highly constrained by internal political tensions and a preference for selected trade partners within the socialist block, as only Shanghai remains visible on the map (Murphey, 1974; Wang and Ducruet, 2013), losing 41% of its traffic between 1925 and 1961. This has directly affected Hong Kong, China's traditional gateway, as seen with its 39% decline between 1890 and 1925, but as an emerging Newly Industrialized Country (NIC), its port traffic has grown up to 368% (1925-1961) and 116% (1961-2004). In comparison, Singapore has been more successful in terms of growth rates (33, 453, and 138%), resulting in the first rank of world ports in 2004 (see also Appendix 2). Although Shanghai is not yet the world's largest port in 2004, its traffic has grown 1195% since 1961. The emergence of Middle Eastern ports based on oil extraction and shipping in 1960s also largely explains the position of Aden in the Red Sea for transit (e.g. mail), owned by British Petroleum at the time. The traffic boom of large Western ports has occurred much earlier, as seen with the fast growth of main gateways during the 1890-1925 period (Rotterdam 375%, Antwerp 284%, Amsterdam 172%, Hamburg 126%), followed by moderate growth and slowdown. In fact, British cities as well as New York had already reached a peak of activity around 1890, given their limited subsequent growth (New York 19, -4, -40%; London 52, -40, -52%). Overall, the acceleration of world trade since the 1960s resulted in an increased connectivity of world ports up to nowadays, as in 2004 the map gives the impression that all coasts enjoy high levels of port activity.
Figure 3: Evolution of the global port hierarchy, 1890-2004
A useful way to differentiate port cities according to their growth trajectory is to make a typology (clustering) based on the number of vessel calls by port and by year. Such an approach takes its inspiration from previous attempts to classify ports and port cities based on container traffics since the 1970s (Ducruet and Lee, 2006; Guerrero and Rodrigue, 2013). Main results provide six possible trajectories among the 381 ports for which traffic was recorded for each year under study (Figure 4). Only one trajectory shows a continuous decline (E) of vessel calls over the period, compared with ports having witnessed growth between the last two years only (F). All other trajectories are enjoying continuous growth but in diverse ways: fast growth from an initial larger size (A) or smaller size (B); slowdown after a period of continuous growth (C) or early growth (D). The declining ports are the least numerous type and concentrate at specific locations, such as New York and Boston, Glasgow, Liverpool, London, and Swansea, Calcutta and Yangon among the largest ports of the sample. This evolution thus concerns the main poles of the British Empire, which have not recovered from geopolitical change and trade reorientation, although they still handle noticeable traffic volumes nowadays. The other declining type, differing from the first by a recent recovery, mostly concentrates in Latin America and the US Gulf Coast but also in Europe: Cadiz, Batumi, and Saint Nazaire. Those port cities have found difficult to adapt new technological standards, such as from wind to steam, or have gone through deep territorial change, such as Batumi.
Figure 4: Typology of port trajectories, 1890-2004
The fastest growing port cities (A) are the most numerous and can be found dominantly in certain regions such as South Asia (Karachi, Bangkok, Hong Kong, Singapore, Indonesia) and Oceania, West and Southern Africa, Western Europe (Le Havre, Rouen, Dunkirk, Amsterdam, Marseilles-Fos, Barcelona, Lisbon), and the Caribbean (Kingston). Such locations have been more resilient to change due to the lesser impact of changing local and global circumstances, such as the current global hub port cities of Hong Kong and Singapore (Lee and Ducruet, 2009). A similar trajectory (B) Chinese port cities and a number of medium-sized European ones, such as in the Baltic. Although it is irrelevant to draw a direct line between such local contexts, those cities have in common to have experienced partial international trade closure during several decades, before a rapid recovery. Early rapid growth and slowdown (D) mostly characterizes larger ports such as Bombay, Hull, Edinburgh, Bristol, Antwerp, Bremen, Hamburg, Bordeaux, Genoa, Baltimore, Philadelphia, New Orleans, Mobile, San Francisco, Seattle, and Sydney. In comparison, some port cities have attained a comparable size (C) but thanks to a mid-period rather than early growth, followed by slowdown: Yokohama and Kobe, Manila, Aden, Santos, Vancouver, Portland, but also Odessa, Piraeus, Naples, Rotterdam, Helsinki, Stockholm, Gothenburg, Malmo, and Copenhagen. Indeed, such cities have developed through different waves of industrialization in the contemporary period, but all C and D types have in common a stabilization of traffics in the recent period.
Despite many challenges inherent to a long-term analysis of urban dynamics on a global scale based on harmonized measures of economic activity, this paper provided novel evidences about the respective importance of change and stability in the worldwide maritime network of port cities between 1890 and 2004. The statistical information derived from Lloyd's List every 40 years or so is rather unique and largely untapped, as it allows mapping with precision traffic volumes at city level based on vessel calls. One main result is that although the global maritime network has constantly been growing in size and connectivity, important regional shifts have occurred from one period to another. Traffic loss at certain routes and ports has been compensated elsewhere. One very clear result is the drastic opposition, in terms of port growth trajectory, between London/New York (regular decline) and Hong Kong/Singapore (fast growth), which, of course, was expectable. The analysis, however, does not neglect local dynamics since in fact, Hong Kong lost traffics between 1890 and 1925, for instance, due to political instabilities in Mainland China. The port activity of New York and London has everything but disappeared but it remains at lower levels than in the late nineteenth century. Further research shall be done based on a more complete database in order to better analyze traffic distribution dynamics and their local-global underlying factors. Dynamic clustering can be tested in order to map the changing geographic coverage of maritime forelands and functional regions, and to test their overlap with more official delineations such as empires, nations, and trading blocks. Last but not least, the analysis of traffics will benefit from being coupled with urban population statistics available over the period, in order to verify the co-evolution of cities and ports and its possible future trends.
The author would like to thank Liliane Lizzi and Robin Cura at Géographie-Cités for their help on cartography and data analysis. The research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP/2007-2013) / ERC Grant Agreement n.  "World Seastems".
Arrault J.B. (2008) Une géographie inattendue: le système mondial vu par Paul Vidal de la Blache, L'Espace Géographique, 37(1): 75-88.
Braudel F. (1985) Civilization and Capitalism, New York: Harper and Row.
Dogan M. (1988) Giant cities as maritime gateways, in Dogan M., Kasarda J.D. (eds.) The Metropolis Era, SAGE, pp. 30-55.
Ducruet C. (2012) Ports et routes maritimes dans le monde (1890-1925), Mappemonde 106: http://mappemonde.mgm.fr/num34/lieux/lieux12201.html
Ducruet C. (2013) Network diversity and maritime flows, Journal of Transport Geography 30: 77-88.
Ducruet C., Ietri D., Rozenblat C. (2011) Cities in worldwide air and sea flows: A multiple networks analysis, Cybergeo: European Journal of Geography 528: http://cybergeo.revues.org/23603
Ducruet C., Lee S.W. (2006) Frontline soldiers of globalisation: Port-city evolution and regional competition, Geojournal, 67(2): 107-122.
Ducruet C., Lee S.W., Ng K.Y.A. (2010a) Centrality and vulnerability in liner shipping networks: Revisiting the Northeast Asian port hierarchy, Maritime Policy and Management 37(1): 17-36.
Ducruet C, Lugo I. (2013) Cities and transport networks in shipping and logistics research, Asian Journal of Shipping and Logistics, 29(2): 149-170.
Ducruet C., Marnot B. (2013) Analyser les trafics portuaires mondiaux en 1890 et 1925 à partir des registres du Lloyd's (forthcoming).
Ducruet C., Notteboom T.E. (2012) The worldwide maritime network of container shipping: Spatial structure and regional dynamics, Global Networks 12(3): 395-423.
Ducruet C., Roussin S., Jo J.C. (2009) Going West? Spatial polarization of the North Korean port system, Journal of Transport Geography, 17(5): 357-368.
Ducruet C., Rozenblat C., Zaidi F. (2010b) Ports in multi-level maritime networks: Evidence from the Atlantic (1996-2006), Journal of Transport Geography 18(4): 508-518.
Ducruet C., Zaidi F. (2012) Maritime constellations: A complex network approach to shipping and ports, Maritime Policy and Management 39(2): 151-168.
Gottmann J. (1961) Megalopolis, The Urbanized Northeastern Seaboard of the United States, The MIT Press.
Guerrero D., Rodrigue J.P. (2013) The waves of containerization, Paper presented at the Annual Conference of the International Association of Maritime Economists (IAME), Marseilles, France, 3-5 July.
Halpern B.S., Walbridge S., Selkoe K.A., Kappel C.V., Micheli F., D'Agrosa C., Bruno J.F., Casey K.S., Ebert C., Fox H.E., Fujita R., Heinemann D., Lenihan H.S., Madin E.M.P., Perry M.T., Selig E.R., Spalding M., Steneck R., Watson R. (2008) A Global Map of Human Impact on Marine Ecosystems, Science 319(5865): 948-952.
Herrera R.G., Wheeler D., Können G., Koek F., Jones P., Prieto M.R. (2003) Climatological Database for the World's Oceans 1750-1850, CLIWOC Final Report, UE contract EVK2-CT-2000-00090.
Joly O. (1999) La structuration des réseaux de circulation maritime, Unpublished PhD Dissertation in Territorial Planning, Le Havre University, CIRTAI.
Kaluza P., Koelzsch A., Gastner M.T., Blasius B. (2010) The complex network of global cargo ship movements, Journal of the Royal Society Interface, 7: 1093-1103.
Keeling D.J. (1995) Transport and the world city paradigm, in Knox P.L., Taylor P.J. (eds.) World Cities in a World System, Cambridge University Press, pp. 115-129.
Konvitz J.W. (1994) The crisis of Atlantic port cities, 1880 to 1920, Comparative Studies in Society and History, 36(2): 293-318.
Lee S.W., Ducruet C. (2009) Spatial glocalization in Asia-Pacific hub port cities: A comparison of Hong Kong and Singapore, Urban Geography, 30(2): 162-184.
Lee S. W., Song D.W., Ducruet C. (2008) A tale of Asia's world ports: The spatial evolution in global hub port cities, Geoforum 39: 372-85.
Lewis M.W., Wigen K. (1999) A maritime response to the crisis in area studies, The Geographical Review, 89(2): 161-168.
Lloyd's List (1890) Lloyd's Weekly Shipping Index, 21(536), April 3.
Lloyd's List (1925) Lloyd's Daily Index, April 1.
Lloyd's List (1961) Lloyd's Shipping Index Voyage Supplement, 14163, April 7.
Lloyd's List (2004) Lloyd's Voyage Record.
Marnot B. (2012) La mondialisation au XIXe siècle (1850-1914), Paris: Armand Colin.
Murphey R. (1974) The Treaty Ports and China's modernization, in Elvin M., Skinner G.W. (eds.) The Chinese City between Two Worlds, pp. 17-74. Stanford University Press, Stanford and London.
Nelson A. (2008) Estimated travel time to the nearest city of 50,000 or more people in year 2000. Global Environment Monitoring Unit - Joint Research Centre of the European Commission, Ispra Italy.
Rees H. (1955) Lloyd's List as a source for port study in schools, Geography 40: 249-254.
Ullman E.L. (1949) Mapping the world's ocean trade: A research proposal, The Professional Geographer, 1(2): 19-22.
Van Hamme G., Patris C. (2011) Territorial Impact of Globalization on Europe and its Regions (TIGER), Interim Report, The ESPON 2013 Programme, Applied Research Project 2013/1/1.
Vigarié A. (1964) Les grands ports de commerce de la Seine au Rhin, leur évolution devant l'industrialisation des arrière-pays, Paris: SABRI.
Vigarié A. (1968) Géographie de la circulation, Genin, Paris.
Wallerstein I. (1979) The Capitalist World Economy, Cambridge University Press.
Wang C., Ducruet C. (2013) Regional resilience and spatial cycles: Long-term evolution of the Chinese port system (221 B.C. - 2009 A.D.), Tijdschrift voor Economische en Regionale Geografie, DOI: 10.1111/tesg.12033
Appendix 1: World share of interregional routes (% vessel calls), 1890-2004
Appendix 2: Major ports of the world by the number of vessel calls, 1890-2004