Power and Size:

Urbanization and Empire Formation in World-Systems Since the Bronze Age

Christopher Chase-Dunn

Alexis Álvarez

Daniel Pasciuti

V 4-13-04 (8041 words) Revised version.

World-systems are intersocietal interaction networks in which culturally different peoples are strongly linked together by trade, political-military engagement and information flows. This chapter presents an overview of research on city and empire growth/decline phases and new evidence on the relationship between urban growth and the rise and fall of empires in six regions that once contained substantially separate world-systems.[1] We find that empires and cities grow and decline together in some regions, but not others, and we examine the temporal correlations between growth/decline phases of largest and second largest cities and empires within regions. Do large empires grow at the expense of other large states within a region or are there periods of regional growth in which states (and cities) are growing together?

Earlier research has demonstrated the utility of studying settlement systems and networks of interacting polities as windows on the historical development of social complexity and hierarchy (Chase-Dunn and Hall 1997). By knowing the population sizes of settlements and the approximate territorial sizes of states and empires we can compare rather different time periods and regions in order to discover both regularities and uniquenesses.

This chapter summarizes the results of earlier studies using city and empire sizes and presents new results on the relationships between changes in urban populations, city-size distributions and the territorial sizes of states and empires. Archaeologists often assume that the concentration of political power can be inferred from the rise of a size hierarchy of settlements – increases in the steepness of the settlement size distribution (e.g. Kowalewski 1982). Existing data can be used to test this hypothesis, though more certain results await the improved accuracy and greater temporal resolution of estimates of city and empire sizes.[2]

Chase-Dunn and Willard (1993) examined urban growth and city-size distributions[3] in nine different regional political/military networks (PMNs) [4] using data on city sizes from Tertius Chandler’s (1987) compendium. Political/military networks (PMNs) are interstate systems – systems of adjacent conflicting and allying states. David Wilkinson (1987) bounds these expanding and contracting systems of states as they merge or become incorporated into what Wilkinson calls the “Central Civilization.” Chase-Dunn and Willard (1993) plotted changes in the Standardized Primacy Indices (a measure of the steepness of the city-size distribution) over time, and read descriptions of what was happening in nine different PMNs to examine the hypothesis that changes in the city-size distribution are related to changes in the centralization of state power in regional state systems. They also accidentally discovered a synchrony of changes in city size distributions and phases of urban growth/decline in the East Asian and the West Asian-Mediterranean PMNs over a long period from 650 BCE to 1500 CE.[5]

This latter discovery led to further research using data on the territorial sizes of empires gathered by Rein Taagepera (1978a, 1978b, 1979,1997). That analysis (Chase-Dunn, Manning and Hall 2000) found additional evidence for synchrony between the East Asian and the West Asian-Mediterranean PMNs over this same 2150-year period, and confirmed what had also been indicated by scant city size data from India, that the Indic PMN was marching to a different drummer.

These synchrony results were further confirmed by additional analysis of the city data by Chase-Dunn and Manning (1998). That study examined inter-regional synchrony by comparing constant regions rather than PMNs. PMN boundaries change over time because of the expansion of the Central PMN, whereas specified regions that are held constant over time constitute a different, but related, unit of analysis. Chase-Dunn and Manning found support for the East/West synchrony phenomenon using constant regions, and so this result is not likely to be an artifact of the way in which units of analysis have been constructed.

Power, Urban Growth and Urban Size Hierarchies

This chapter returns to the question asked in the Chase-Dunn and Willard (1993) study about the relationship between urban growth, city-size distributions and the rise and fall of empires. What is the relationship between the size of settlements and power in intergroup relations? Under what circumstances does a society with greater population density have power over adjacent societies with lower population density, and when might this relationship not hold? Population density is often assumed to be a sensible proxy for relative societal power. Indeed, Chase-Dunn and Hall employ high relative population density as a major indicator of core status within a world-system (Chase-Dunn and Hall 1997). But Chase-Dunn and Hall are careful to distinguish between “core/periphery differentiation” and “core/periphery hierarchy.” Only the latter constitutes actively employed intersocietal domination or exploitation, and Chase-Dunn and Hall warn against inferring power directly from differences in population density.

In many world-systems military superiority is the key dimension of intersocietal relations. Military superiority is generally a function of population density and the proximity of a large and coordinated group of combatants to contested regions. The winner of a confrontation is that group that can bring the larger number of combatants together quickly. This general demographic basis of military power is modified to some extent by military technology, including transportation technologies. Factors such as better weapons, better training in the arts of war, faster horses, better boats, greater solidarity among soldiers and their leaders, as well as advantageous terrain, can alter the simple correlation between population size and power. Ironically, George Modelski’s (2003) important study of the growth of world cities completely ignores the phenomenon of state and empires sizes, though Modelski is himself an astute scholar of international relations and geopolitical power. Modelski contends that cities are the most important driving force of world system evolution and that we may conveniently ignore states and empires. We think that the relationship between political power and settlement systems may have changed over the millennia, so that the explicit analysis of the relationships between size and power needs to be directly examined.

The most important general exception (in comparative evolutionary perspective) to the size/power relationship is the phenomenon of semiperipheral development (Chase-Dunn and Hall 1997:Chapter 4). The pattern of uneven development by which formerly more complex societies lose their place to “less developed” societies takes several forms depending on the institutional terrain on which intersocietal competition is occurring. Less relatively dense semiperipheral marcher chiefdoms conquer older core chiefdoms to create larger chiefly polities (Kirch 1984). Likewise, semiperipheral marcher states, usually recently settled peripheral peoples on the edge of an old region of core states, frequently are the agents of a new core-wide empire based on conquest (Mann 1986; Turchin 2003).[6] Another exception is the phenomenon of semiperipheral capitalist city-states – states in the interstices between tributary empires that specialized in long-distance trade and commodity production. Though these were rarely the largest cities within the world-systems dominated by tributary empires, they played a transformational role in the expansion of production for exchange and commodification in the ancient and classical systems. And less dense semiperipheral Europe was the locus of a virile form of capitalism that condensed in a region that was home to a large number of unusually proximate semiperipheral capitalist city-states. This development, and the military technology that emerged in the competitive and capitalist European interstate system, made it possible for less dense Europe to erect a global hegemony over the more densely populated older core regions of Afroeurasia (Chase-Dunn and Hall 1997). The more recent hegemonic ascent of formerly semiperipheral national states such as England and the United States are further examples of the phenomenon of semiperipheral development.

The phenomenon of semiperipheral development does not totally undermine the proposition that societal power and demographic size are likely to be correlated. What it implies is that this correlation can be overcome by other factors, and that these processes are not entirely random. Denser core societies are regularly overcome or out-competed by less dense semiperipheral societies, but it does not follow that all semiperipheral or peripheral regions have such an advantage. On the contrary, in most world-systems most low-density societies are subjected to the power of more dense societies. Semiperipheral development is a rather important exception to this general rule.

Why should a city system have a steeper size distribution when there is a greater concentration of power? The simple answer is that large settlements, and especially large cities, require greater concentrations of resources to support their large populations. This is why population size has itself been suggested as an indicator of power (Taagepera, 1978a: 111). But these resources may be obtainable locally and the settlement size hierarchy may simply correspond to the distribution of ecologically determined resources. People cluster near oases in a desert environment. In such a case it is not the political or economic power of the central settlement over surrounding areas that produces a centralized settlement system, but rather the geographical distribution of necessary or desirable resources. In many systems, however, we have reason to believe that relations of power, domination and exploitation do affect the distribution of human populations in space. Many large cities are as large as they are because they are able to draw upon far-flung regions for food and raw materials. If a city is able to use political/military power or economic power to acquire resources from surrounding cities it will be able to support a larger population than the dominated cities can, and this will produce a hierarchical city size distribution.

Of course the effect can also go the other way. Some cities can dominate others because they have larger populations, as discussed above. Great population size makes possible the assembly of large armies or navies, and this may be an important factor creating or reinforcing steep city size distributions.

The relationship between power and settlement systems is contingent on technology as well as political and economic institutions. Thus we expect to find that the relationship between urban growth and decline sequences and the growth/decline sequences of empires varies across different systems or in the same regional system over time as new institutional developments emerge. We know that the development of new techniques of power, as well the integration of larger and larger regions into systems of interacting production and trade, facilitate the emergence of larger and larger polities as well as larger and larger cities. Thus, there is a secular trend at the global level and within regions between city sizes and polity sizes over the past six millennia. But the question we are asking here is about finer temporal and spatial relationships. Do cities and empires rise and fall together? Are there important exceptions to this pattern? What are the causalities involved?

We may also ask whether or not the causal relations are stable over time within regions? We expect that there may be periodic changes in the relationship between power and size as new institutions develop. The rise of capitalism as an alternative source of power to military might and changes in the relationship between military power and demographic factors most likely change the nature of the connections between size and power. We know that empires ceased to increase in territorial size with the demise of the modern colonial empires. And the contemporary world city system may be unique in the extent to which some of the largest cities are located in the semiperiphery rather than in the core. The exponential growth of cities after 1800 CE makes it more difficult to study growth/decline phases because the largest cities no longer decline in size.

We will further examine the relationship between power, urban growth and settlement size hierarchies by comparing trends in the growth/decline sequences of city populations and the territorial sizes of empires. The main unit of analysis in this study will be Constant Regions[7] rather than political/military networks (interstate systems). We shall also consider studying individual polities (states and empires), because the simplest form of the hypothesis of a causal relationship between power and urbanization is that larger states can afford to create larger cities. But available data are not sufficient for studying the power/size relationship within individual polities except for relatively recent states.

And we will examine the temporal relations between the sizes of the largest cities and empires in each region as well as size distributions of cities and empires and the relations between the largest and the second largest cities and empires when data are available.

Measurement of the population sizes of cities and the territorial sizes of empires is not without difficulties, especially for early periods. How can we know the number of people who reside in Los Angeles today? We use the most recent census, a survey of “residents” conducted by the U.S. federal government. What are the spatial boundaries of “Los Angeles”? Do we mean the city of Los Angeles, Los Angeles County, the contiguous built-up area that constitutes “greater Los Angeles,” or a definition based on the proportion of the local population that is employed in “Los Angeles”? Does “Los Angeles” include San Diego? Nighttime satellite photos of city lights reveal a single unbroken megalopolis from Santa Barbara to Tijuana (see Figure 1).

Figure 1: Southern California / Northwestern Mexico conurbation (city lights from satellite photographs).

So where is Los Angeles? We want to use the contiguous built-up area as our main way of spatially defining cities, another example being the New England urban agglomeration (see Figure 2). For early cities we do not have official, and ostensibly complete, census figures. Thus we rely on methods that archaeologists and students of early urbanization have developed to estimate the population sizes of cities.

Figure 2: New England conurbation (city lights from satellite photographs).

These involve, for example, determining the spatial size of the city and then estimating the population density per unit of area and so estimating the total population (city size * density = city population). Population density varies depending upon the size of families, the nature of dwellings, the amount of non-residential area within settlements, and cultural differences. Anthropologists and archaeologists have made an important effort to produce reliable methods for estimating population sizes from residential areas (e.g. Brown 1987), and the famous historical demographer Paul Bairoch (1988: 21-4) has examined the problem of urban population densities in comparative perspective.

Tertius Chandler (1987) used reports about the number of soldiers to estimate city sizes, assuming that an army represents, on the average, about ten percent of the population of the city in which the army resides. Such estimates are obviously error-prone. Another problem with existing data on both city and empire sizes is that they were produced from surveys of both secondary and primary sources that are now, in many cases, obsolete because more recent and better research has been published by archaeologists, epigraphers and historians. Chandler’s compendium was mainly based on his thorough survey of the contents of the main library at the University of California, Berkeley over the four decades prior to its publication in 1987. We also use the new estimates of city population sizes produced by George Modelski’s (2003) important study. A new project to improve upon existing compendia of city sizes is under way at the Institute for Research on World-Systems at the University of California, Riverside (Pasciuti 2003).

Estimating the territorial sizes of empires is also problematic. Taagepera used atlases and maps to produce his estimates of the spatial sizes of empires from 3000 BCE to the present. But the boundaries of empires are not usually formally specified, but are rather a matter of degrees of control that fall off with distance from the central region. Archaeological evidence of the presence of a core culture in a peripheral region does not prove the existence of control, because many core polities have established colonial enclaves in distant peripheries to facilitate trade (e.g. Stein 1999). So the estimation of empire sizes is also fraught with difficulties. But, as with city sizes, a significant improvement of accuracy, temporal resolution and coverage would result from a renewed effort to code empire sizes using recently published materials. This is another task that the IROWS City-Empire Research Working Group will undertake.[8]

Dating is also a major problem in studying temporal relationships in the ancient world-systems. In this paper we utilize the years originally supplied by Taagepera and Chandler. But the dating of events and city size estimations for the first millennia BCE is a matter of continuing dispute among scholars of ancient history. For ancient Western Asia the Egyptian dynastic dates are used, but these have been repeatedly revised with an error margin of around 25 years. This is a threat to any study of temporal correlations.

Mesopotamia

The first PMN we shall examine is that of Mesopotamia from 2800 to 550 BCE. It is mistaken to speak of a single West Asian/North African world-system for this whole period. Rather two core areas – Egypt and Mesopotamia – were undergoing developmental processes that were only weakly linked, especially at first. As both of these systems expanded their trade networks and political/military interaction networks they came into contact with one another. The prestige goods nets (PGNs) became linked as early as 3000 BCE (Marfoe 1987) or as late as 2250 BCE (Wilkinson 1992), while the Mesopotamian and Egyptian political/military networks became linked by the Egyptian expedition to Syria (about 1520 BCE). We examine the relationships between the population size of the largest city and the territorial size of the largest state or empire in a region as these change over time. The hypothesis of a correspondence between urbanization and the size of polities should reveal a positive correlation in these two measures over time. The data on city population sizes are especially sparse for early millennia and the time points of estimates are widely spaced, making temporal correlation risky. For Mesopotamia our data set is thus:

Year (BCE)	Empire Size (MM^2)	Empire Name	City Name	City Pop (thousands)
-2800	1	Kish	Uruk*	80
-2500	3	Kish	Uruk*	50
-2400	5	Lagash
-2300	65	Akkadian	Agade*#	36
-2200	25	Akkadian
-2100	3	Ur
-2000	10	Sumer	Ur	65
-1900	0
-1800	10	Old Assyria	Mari	29
-1700	25	Babylon
-1600	16.6	Babylon	Babylon	60
-1500	10	Kassite
-1450	10	Kassite
-1400	10	Kassite
-1360	21.7	Hittites	Khattushash (Hattusa)	45
-1350	5	Assyria	Khattushash (Hattusa)	45
-1300	10	Assyria
-1250	15	Assyria
-1200	25	Hittites	Khattushash (Hattusa)	48
-1150	5	Assyria
-1100	40	Assyria
-1050	15	Babylon
-1000	15	Babylon	Babylon	51
-950	15	Babylon
-900	15	Babylon
-850	40	Assyria
-800	57.9	Assyria	Calah	50
-750	40	Assyria
-700	90	Assyria
-650	93.3	Assyria	Nineveh	120
-600	25	Babylon
-550	50	Babylon

Table 1:Mesopotamian Largest Empires and Cities

*These estimates are from Modelski (2003). All other estimates are based on Chandler and Taagepera.

#Archaeologists have not yet decided which of the thousands of tells in Iraq is Agade, the capital of Sargon’s Akkadian empire.

Table 1 immediately demonstrates problems of missing data, especially for the third millennium. The time points for city sizes are far apart, and there are obviously missing cases. We have Modelski’s (1997) best estimate of the population size of Uruk in 2800 and 2500 BCE, but the largest empire shown in Taagepera’s data is that of Kish, a city-state that was independent of the much larger empire of Uruk in this period. This obvious error strongly demonstrates the need for upgrading the data sets we are using. In the data presented in Table 1 and in the figures and tables below we have interpolated Taagepera’s dates of changes in the sizes of empires to regular time intervals (every 50 years in Table 1 and Figures 3 and 4; every 10 years for the other regions in Table 2 and the figures in Appendix A, (see http://www.irows.ucr.edu/research/citemp/appendices/apppowsize.htm).

Another complication revealed in Table 1 is as follows: in 1350 and 1200 BCE the largest city is Khattushash (Hattusa), the capital of the Hittite empire, but the largest empire in the Mesopotamian region is the Neo-Assyrian Empire. This raises the issue of the proper unit of analysis – regions or polities – but it also raises a theoretical issue. The simplest version of the size-power hypothesis is that larger empires can afford larger cities, and to test this hypothesis we would need temporally fine-grained data on the size of the largest city within each empire. For this purpose the unit of analysis should be the polity (states and empires). But it may also be the case that regions or PMNs experience cyclical periods of growth and decline in which all the states and cities are growing, or alternatively that state and city growth is a zero-sum game in which growth in some results or is related to decline in others. By using and comparing different spatial units of analysis we can examine these competing hypotheses.

Figure 3: Largest Mesopotamian Cities and Empires

The temporal relationship between the size of the largest city and the size of the largest empire is positive for the Mesopotamian case with a positive Pearson’s r correlation coefficient of .45 based on twelve time points for which we have data for both variables (see Figure 3). This supports the hypothesis of a causal relationship between these features of the social landscape, but the positive association could also be due to other factors or to the secular trending of these characteristics. We will return to these issues when we have more and better data in the cases to be discussed below.

Egypt

As with Mesopotamia the data for Egypt are few and problematic. But using what we have produces the results displayed in Figure 4.

Figure 4: Largest Egyptian Cities and Empires

The temporal relationship between city and empire sizes in Egypt is also positive, producing a Pearson’s r correlation coefficient of .06 based on eight time points for which we have data for both variables. Though there is a secular upward trend, both city and empire sizes also reveal decline phases and these are roughly synchronous with one another, though the few estimates of city sizes makes a firm conclusion risky.

Size and Power in Regions

Table 2 (below) presents the bivariate Pearson’s r correlation coefficients between empire and city sizes for all of the regions for which we have sufficient data. It also presents the partial correlations controlling for year to remove the long-term upward trend between city and empire sizes. The Americas and Africa do not have enough city size data, though this deficiency could and should be remedied by a new coding project.

Table 2 shows that five of the six regions have statistically significant positive bivariate correlations between city and empire sizes. This lends support to the contentions discussed above of a causal interaction between power and size, but these correlations do not shed light on the question of the direction of the causal effects. Once we have improved data we plan to employ the test of antecedence to shed light on this. We should also note the two of our regional “cases” overlap with one another, Mesopotamia and West Asia.

	Correlations between 3-period moving averages of city and empire sizes
	Region	Bivariate r	Partial r	Biv. N	Years
	Egypt	0.373	0.152	8	2800 B.C.E. - 1500 B.C.E.
	Mesopotamia	0.699**	0.492*	18	2800 B.C.E. - 650 B.C.E.
	West Asia	0.542**	0.207^	44	2800 B.C.E. - 1500 C.E.
	South Asia	0.593*	0.578*	13	1800 B.C.E. - 1500 C.E.
	East Asia	0.610**	-0.191	33	1360 B.C.E. - 1800 C.E.
	Europe	0.510**	0.7324**	29	430 B.C.E. - 1800 C.E.
^	Pearson's r significant at the 0.10 level.
*	Pearson's r significant at the 0.05 level.
**	Pearson's r significant at the 0.01 level.

Table 2: Regional correlations between city and empire sizes

One problem with the bivariate results in Table 2 is that they may be due to the secular trends in which both cities and empires increase in size over the long run rather than to medium-term oscillations.[9] We are most interested in the medium term relations between cities and empires here.

There are two ways to remove the effects of the secular upward trend. The first is to compute partial correlations controlling for time. These results are presented in Table 2. Another method of detrending is to compute first differences -- the change scores from one period to the next. The irregular (and infrequent) time points of the early city size estimates make change score detrending complicated because the changes must be weighted by the amount of time that elapses between estimate intervals. We have examined the regional correlations