1
Dynamics of Civilizational Networks
Civilizations as Dynamic Networks:
Monetization and Organizational Change from Medieval to Modern
Douglas R. White and Peter Spufford
(these are very preliminary notes for a book ms in preparation)
Filesource: FTP/Spufford/citiesNetwork2.doc
See: for additional slides
Table of Contents
Introduction
Abstract of the Argument
1. Banks, Money and Trade Imbalances
2. Commodity Chains: Raw Materials, Finished Goods, and Consumption
3. Roads, Ports, Navigable Rivers, Ships, Vehicles, and Flows
4. CapitalCities, Wealth and Investment of Rulers and Merchants, Town Size and Inflation
5. A Network Perspective on Shifts in Economic Hegemony (Flow Centrality)
6. Volume of Trade and Commercial Transformation (Institutional Change)
7. Landed and Commercial or Capitalist Hegemony (Demographic and Network effects)
8. Transformations in Agent Space – (Paris, Champagne Fairs)
9. Polities, Sociopolitical Violence, and Wars
10. Climate, Event and Agent Data
11. Modeling
12. Conclusion
13. Bibliographic Postscript: Toward Dataset Expansion for the Eurasian World System
14. Acknowledgements
15. References
1
Dynamics of Civilizational Networks
Introduction
Peter Spufford’s (2002) history of late medieval Europe inPower and Profit: the Merchant in Medieval Europe provides a new synthesis of the transition from feudalism – based on rights and services tied to land – to a monetized economy. The history is told in terms of networks – among cities, merchants, peasants, elites, states and empires, ecclesiastical and other organizations – and a rising velocity of trade that at successive thresholds transforms sites and organizations. The focus of this study on network representation allows a reformulation of explanatory concepts that can be measured and tested as to network effects on social transformation. The goal is to expand and formalize the explanations offered by Spufford for the many organizational and technological transformations that he recounts. Monetization, velocity of trade and thresholds beyond which organizations cannot perform without reorganization are crucial to Spufford’s explanations of the transformation that took place in this historical period, and are equally important in other contexts as well (Chandler 1977, 1990, Iberall and Soodak 1978, Soodak and Iberall 1978). Qualitative coding of variables also allows identification of new patterns and generalizations as well as statistical tests of hypotheses.
This paper adopts the view that pressures towards monetization in this period were not endogenous but largely driven by the dynamics of demographic and sociopolitical instability cycles, as described by Nefedov (2003), Turchin (2004), and Koratayev, White and Khalturina (2004). The demographic cycle of the ‘long 13th century’ began with population expansion in 1150 and reached a peak that provoked crisis at the end of the 13th century, as did the population peak in Braudel’s ‘long 16th century.’ Early in the 1300s famines were rife and extended (Spufford 2002:14).[i] The crisis was exacerbated by the onslaught of Black Death in successive waves, but the Late Medieval ‘economic depression of the Renaissance’ (Spufford 2002:12) did not end until 1480. Nefedov, Turchin, Goldstone (1991) and others identified a dynamic in which – in the three great 150-300 year swings of European population between 1150 and 1920 as well as other agrarian regimes – the peaking of population up against a Malthusian carrying capacity that initiates a rise in sociopolitical violence. Such violence tends to abates only long after population has fallen in the period of demographic crisis. The Nefedov-inspired and Turchin-informed model at the end of this study reviews how the stagflation (slowing growth and rising prices), crisis, and economic depression segments of the 13th century population cycle spurred the transition from feudalism to a highly monetized economy, and views the resultant monetization as a key factor in the rising velocity of trade that spurred organizational transformations leading from feudal to the modern monetized economy.
Network visualizations of historical trends and changes described by Spufford (2002) and the preliminary hypotheses that might explain them are crucial to this effort at modeling. Six initial caveats must be stated with regard to the visualizations and hypotheses offered here. (1) The network data presented here are not models but provide visualizations prior to modeling trade-system dynamics. The advantage of the visualizations is that they allow us to inspect the kinds of variables that might be coded in such models, and how these variables appear, at the time-scales for which they are coded, in time series representations. (2) The hypotheses are not models, but first-order attempts to account for patterns in the observed data. (3) Given the kinds of data available, the variables that are coded are qualitative. The codings are largely of nodes representing towns and cities and edges representing trade between cities. It is useful to be able to visualize this as a time-series network. (4) The time-series are being rescaled to generations (twenty five years each) but links at this point are coded only for changes from the 12th to 15thcenturies. (5) This is not a representation of a world-system, but only of limited aspects of one region within a larger world-system dealing with cities, politics, and religion; and ignoring for the time being other types of organizations. The larger world-system stretches through the Middle East to India and is interlinked through bulk and luxury trade to the contemporaneous Indonesian, Far Eastern and West African regions (Wilkinson 2004). (6) These representations, which proceed in stages, are not as complex as might be called for in models intended for statistical analyses of quantitative variables, which in any case are rarely available. Rather, they are simplified in order to render visually – and render judgment of adequacy – some of the main patterns of Spufford’s synthesis and analysis of medieval trade. Simplification allows lower-dimensional network analyses to be employed, e.g., assuming a few simple types of weighted edges between comparable nodes.
These caveats stated, a considerable range of patterns of attribute data and network relationships are simultaneously and successively visualized in a network time-series. Attributes are mapped onto the nodes in a series of static or temporal comparisons, using size of nodes and color (type) or scalar shading (or color intensity) of nodes, shape of nodes, and position of nodes to capture attribute variation. Thus, up to four aspects of attribute variation of nodes may be simultaneously represented. Edge relationships are mapped using width of edges, color (type) or scalar shading (weighting) of edges, solidity (versus various kinds of dotted or broken lines) and direction of edges. Again, up to four aspects of edge variation may be simultaneously represented. Further, because edges (or arcs as directed edges) can be specified of different types, each type can have its own weightings. Even these simple representations deal with multidimensional attribute and relational complexes. Table 1 shows some of features of network representation for the cities network.
Table 1A: Information Conveyed in Graphing the Network
NODES / LINESSize of nodes / Position of nodes / Color or Shading / Shape of Nodes / Width of lines / Solid / Broken / Color or Shading / Direction of lines
Population / Geographic / Polity Capitals / as needed / Importance / Trade / as needed / Export
Centralities, Wealth / Equivalence Scaling / Cohesion / As needed / Flow / War / as needed / Import
Table 1B: Information Conveyed in GIS (Geographic Information Systems)Formats
POLYGONS / PIXELS / NODES / PATHSSize / Shape Boundaries / Color / Radius / Length / Color or Shading
Territory / Polities / Terrain / Cities / Distance
Religion / Water / Attribute / Sea Route
The problem is not to overburden our visual representations so that they are too complex to comprehend and thus carry the ‘story’ in Spufford’s book. The requisite networks are built out of the maps and descriptions of changing relationships and attributes of the medieval players (cities, polities, religious and commercial organizations, families). By analyzing the city network, for example, some dimensions are simplified into spatial positions, like that of geographic location.[ii] Similarly, relative distances in a more abstract representational space can be computed in terms relativeequivalence of positions within the network (White and Reitz 1983). Such representations supplement the historical text in important ways: The network allows measurements made according to the configuration of relationships in trading networks. Examples are centrality of nodes (Freeman 1977, Freeman, Borgatti and White 1991) and nested levels of structural cohesion of nodes (Moody and White 2001). The results of these computations may be shown on the same spatial network by the scalar shading of nodes. In some cases, the results of a network analysis are easier to visualize when the spatial configuration of nodes reflects an x-y (or 3d) grid of distances determined directly by the analysis. Such is the case, for example, with structural positions calculated in a multidimensional similarity space (Smith and White 1992). If needed, we can represent 3d outcomes of positional analysis. Hence, visualization allows some flexibility in the ways that we view the ensemble of attributes and relationships that constitute the network and we have similar flexibility in how to construct the network for purposes of visualization.
Choice of network visualizations and preparation of data for network measurement evolved out of the way that Spufford (2002:12-14) organized and began his book with transformations of trade: the increase in population and money supply; and the capital cities and commercial centers that were most transformed by the expansion and then the contraction of trade. A strategy for mathematical modeling will be suggested for one aspect of change, namely, the effects of monetization and network flows on episodic change in organizational complexity.
Spufford’s (2002:26) first map presents a weighting of edges in terms of the flows of credit among cities as banking places and the concomitant courier routes involved in credit transfers. Coding began with that network, noting the different types of cities and flows, and added to it the maritime trading links which are so important to the physical transport of bulk commodities in addition to the flow network of credits. The maritime network was not presented by Spufford (p. 398) until the concluding chapter – exemplified only for the Venetian trade –and required that we supplement from outside sources the maritime trade routes of the Venetians with those of the Genoese. These initial data, after Wehbe’s compilation of the cities and populations in the figure on page 26, required only a few days of work to assemble into a network of 91 cities, displayed on the web ( in a series of introductory slides.
Results of this first stage of network construction led to initial network analyses that can be briefly summarized. Computation of betweenness centralities of the nodes in this network, ignoring the weights of edges, showed Genoa to be the most central city, commensurate with its role in the trade network by the early 13th C. Cohesion analysis showed a biconnected core, that is, a set of cities in which each pair are connected by two or more independent trade routes.[iii] Positional equivalence analysis showed that this core could be visualized as an overall trading “circle” reminiscent of the anthropological ‘kula ring’ rather than the core-periphery structure of the modern international world-system (Smith and White 1992). These, however, were only suggestive and not definitive results, i.e., provisional rather than final conclusions about network structure.
The harder work then began to decide how to treat the actual flow of commodities, including credit and the metals used for coinage. The areas of wool production and the towns or cities that had or imported wool and had a woollens industry are well described by Spufford (see pp. 230-233 and 328-329 for raw wool production) as were the transport routes and intermediary agents of trade. It is not difficult to envision coding, if only in cursory form, the flows of raw wool and then of finished woollens in the network, and the flows of revenues produced; and similarly for other major commodities and commodity flows. Revenue flows for commodities, however, were facilitated by coinage and the scarcity of coins was offset by the flows of precious metals out of mining areas and towns that served as mining centers. Constructing this part of the network was the next step.
Changes in agent space are exemplified by such processes as the takeover in the 1320s of London exports by Florentines from the earlier network of Lowland European agents resident in London (2002:238). The shift of agents-of-trade, while of utmost importance, is difficult to represent. A more complex representation would have networks of agents moving among the nodes of the network of cities. Also coded into the network for the wool trade were the English ports of export, Lowland countries’ ports of import and changes in major trade routes.
Abstract of the Argument
Spufford’s book is not a dry descriptive historical account of a historical period: It abounds in theoretical summaries of transformations that are conversant with theories of complexity. Spufford’s argument follows a complex interactional-systems logic. The empirical base for my development of this argument is Spufford (2002), supplemented by Arrighi (1994), Fischer (1996), Nefedov (2003) and Turchin (2004).
1. Networks lack causality per se; it is strategies and activities in networks that may be causal.
2. Activities transform sites (France’s wealthiest court in Champagne moves to Paris; Aalst, whose nobility marries a Burgundian ruler and who shift political membership). Decisions to move (e.g., by the Champagne nobility) might be highly influenced by changes in agent space (reconcentration of agents in Paris (2002:75).
3. Intensity of activities transforms organizations (intensity use of agents (division of labor) bypassing of markets, contracts made at arm’s length in Italian examples and Champagne fairs)
4. Changes in agent space change the logic of systems.
a. In the period of Genoese economic hegemony, the agents are family oversea diasporas.
b. The new type of agent is a commercial agent that allows the merchant to stay in one place. This might begin in one place (e.g., Florence) but ends by diffusion to other place (Examples are the Antwerp Bourse (p.50) and later rise of Dutch firms, transformed by the new possibilities of overseas organization while staying in place).
5. System logics are reversible, although their infrastructural changes are cumulative. For example, following Dutch economic hegemony, the new types of British Imperial agent include a diasporic colonist, not so much on a family model as a class model. And prior to the Genoese ‘family capitalist’ diaspora there was a period of Venetian economic hegemony in which the agents were those of the ship crews who joined Venetian corporations as a form of investment but did not form diasporas but returned home to their trades.
6. Competition determines which actors are replaced with others. (Lowlander Florentine agents in England)
7. Warfare outcomes tend to be influenced by success in other forms of economic competition (e.g., wealth more potential for successful armament) which further eliminates resources of defeated competitors (e.g., destruction of a large part of Genoese fleet)
8. Population swings (e.g., Black Death) might not be the determinants of system decline. Indeed, after the Black Death of the 1340s there was a brief period of economic boom before the European economy went to ground. Climate shift in the 13th century also had a major negative effect.
9. The general model of change is one where, while Malthusian constraints are not the determinants of system decline, since major climate fluctuation and epidemic disease may act as exogenous shocks, population cycles operate here, as elsewhere in agrarian societies (Turchin 2004), as the major dynamic driving a host of other factors, including those that tend to affect monetization. Effects of and on monetization are extensively discussed by Spufford as a main focus of the book, summarized as follows:
The thirteenth-century increase in the demand for luxury goods was backed up by newly-liberated quantities of ready cash, arising from a revolution in rents. By the end of the century landlords essentially collected their rents in money in place of a mixture of goods, services and coin, amongst which coin had been the least important. It is no wonder that this demand for distant luxuries brought about an enormous quantitative change in the volume of international trade. Moreover, as business became focused on a limited number of particular places, or rather along a limited number of routes between those places, a critical mass was reached, so that qualitative as well as merely quantitative changes in the nature of commerce began to take place. This vital transformation could only happen when the concentrated supply of money, and consequently of trade, rose beyond a certain critical point.
Up to that point, on any particular route, all that occurred was an increase in the volume of trade within the traditional framework. Italian merchants, for example, merely added extra mules loaded with goods to the mule-trains that accompanied them when they ventured northwards across the Alps. However, once the critical point was reached, the scale of enterprises allowed for a division of labour. (2002:29)
The proposal adopted here in viewing monetization not as an endogenous driver of the dynamics of change but as a secondary effect of population dynamics results from Turchin’s (2004) arguments and empirical evidence for the long-term cycles of population growth, stagflation, crisis and sociopolitical violence, depopulation and economic depression, and eventual recovery and regrowth. This evidence will not be reviewed here.
Monetization, however, can be seen as the principal driver of reorganization and of increases in the complexity of organizations when velocities of money and commodity exchanges surpass the capacity thresholds of existing organizations. This is an episodic process and operates through a variety of channels that will be reviewed at the end of this study. The primary focus here will be on network predictions that help to predict where, given overall patterns in the velocities of trade, specific reorganizations will tend to occur.
Nine topics of network coding guided construction of a network dataset. These involved rules for coding new cities, towns, or in some cases, junctions, and the difference in importance of links between them, for example, as well as changes over time.