Why are some parts of the world politically fragmented while others tend to be dominated by a single state? This age-old question has implications for many important topics in comparative economic development such as the origins of the Great Divergence (see Broadberry 2021) or the divergence in political institutions between China and Europe (see Jia et al. 2021).
Scholars going back at least as far as Montesquieu and Hume have attributed the rise of Western Europe to its persistent political fragmentation. More recently, Jones (2003), Mokyr (2016, 2017), and Scheidel (2019) have developed this thesis in novel ways. These authors acknowledge that a polycentric state system has static costs such as tariff barriers and more wars but argue that, on the net, it is associated with better dynamic incentives for intellectual innovation and state building.
But what determines these patterns of fragmentation? More concretely: what factors account for the prevalence of political polycentrism in Europe and the prominence of political centralisation in China? A leading explanation of this phenomenon is the ‘fractured land’ hypothesis, most famously stated by Diamond (1997). According to this view, fractured land such as mountain barriers, indented coastlines, and rugged terrain precluded the development of large empires in Europe. In comparison, China’s geographical features led to its recurring unifications.
While the fractured land hypothesis has been widely cited and much criticised (e.g. Hoffman 2015), it has not been formally modelled or tested. In Fernández-Villaverde et al. (2022), we fill this gap by providing a quantitative investigation of the fractured-land hypothesis. We do so by modelling the dynamic process of state-building and exploring how fractured land shaped inter-state competition in unexpected, non-linear ways.
Our model is very simple. We divide the globe into 65,641 hexagonal cells, each with a radius of 28 kilometres. These cells are then endowed with a vector of geographic characteristics, notably a level of agricultural productivity, ruggedness, and climate.
The model begins with each cell as an independent polity. Over time, cells come into conflict with bordering cells. The probability of conflict is increasing in the productivity of the cell. Victory depends on the aggregate productivity of the cells controlled by a polity and the geographical characteristics of the cells in conflict according to a simple contest success function. Conflicts can take place on land or span across the sea. We also allow for secession by border cells. Specifically, border cells are more likely to secede if their geographic characteristics make it harder to control them (i.e. they have more rugged terrain), if the polity that controls them is larger, and if the border of that polity is long relative to its interior.
We calibrate the parameters in the model based on variables used by military strategists (Dupuy 1979) and simulate it from 1000 BCE (i.e. roughly the start of the Iron Age) to 1500 CE (the dawn of the European voyages of discovery). Notice that the model delivers random outcomes. Therefore, inspired by Crafts (1977) and Turchin et al. (2013), we focus on pattern predictions (i.e. probability distributions) rather than replicating specific outcomes. If and when a state emerges to dominate its neighbours is neither fluke nor destiny but a balance of structure and contingency: probability distributions measure this balance.
Figure 1 shows an illustrative simulation of our model in 750 BCE for Eurasia and North Africa (we discuss the results for the rest of the world below). Nearly every cell remains independent. Nonetheless, we can start to see a process of consolidation in northern China resembling the core areas of the Shang and the Zhou dynasties. In comparison, no large polities appear in Europe.
Figure 1 Representative simulation, 750 BCE
Figure 2 shows the same simulation in 500 CE. Now, we start to see polities that roughly resemble Spain or France (including a polity very similar to the Kingdom of the Suebi in the northwest of the Iberian Peninsula, which existed between 409–585). The year 500 is around the time when the Germanic kingdoms that inherited the western Roman Empire were formed. Relative to East Asia, where a Chinese empire dominates, Europe remains politically fragmented.
Figure 2 Representative simulation, 500 CE
Finally, Figure 3 shows the end of the simulation in 1500 CE. We see how the large polity occupying China and dominating East Asia has expanded to the south toward Vietnam and Yunnan. The polity controlling India has expanded toward the south, occupying an area similar to the Delhi Sultanate (1206–1526) at its peak. In Europe, we see a nearly unified Iberian Peninsula (as happened between 1580 and 1640), polities resembling England, Scotland, and Ireland, a larger France, and the Ottoman Empire.
Figure 3 Representative simulation, 1500 CE
While the previous simulation is illustrative, the patterns behind it are highly robust. In Figure 4, we present the evolution of the Herfindahl indices of political unification for China and Europe over 500 periods (i.e. from 1000 BCE to 1500 CE in periods of five years). We plot the average index for 30 simulations from the model (solid line) and the 95% interval computed from 10,000 bootstrap samples (coloured band). China always centralises quickly. In contrast, Europe always remains fragmented. Our model demonstrates that Europe’s political fragmentation is not an accident of history; it is structural.
Figure 4 Fan chart for 30 simulations
Our model, moreover, allows us to investigate this puzzle more deeply. We can turn on and off different variables and see how these counterfactuals change our results. Remarkably, our primary finding is that China still unifies more quickly than Europe even after we remove differences in climate and productivity. Only when we impose a uniform distribution of resources and eliminate all geographical obstacles do we observe the two ends of Eurasia centralising at a comparable pace. Thus, our model demonstrates the non-linear consequences of combining different geographic features, a surprising and unexpected discovery that only a quantitative model can deliver.
But what about the rest of the world? Intriguingly, while our model does an excellent job of predicting outcomes in Eurasia, it needs to be modified to generate patterns of state formation that resemble what we observe historically in the Americas and Africa. Specifically, in the Americas, we must consider the later development of agriculture, the slow pace at which this technology diffused in the Americas, and the unique characteristics of maize agriculture and guano. In Sub-Saharan Africa, we adapt the model to reflect lower levels of agricultural productivity in the continent due mainly to disease. For example, as Alsan (2015) established, the TseTse fly resulted in the absence of large-scale livestock in much of Africa until modern times. Once we adjust for these factors, our model does predict patterns of state formation that broadly conform to what we see historically. In that sense, our model is a ‘measurement device’ that can evaluate which factors are essential to account for observed state formation patterns. Therefore, our results substantiate the importance of fractured land in explaining global patterns of state development.
Our framework delivers several further insights. First, it highlights the importance of North China in explaining China’s precocious and persistent state development, corroborating the narrative analytics of many historians. Second, we can show that political fragmentation in Europe took a particular form: that of medium-sized states, as opposed to the high political fragmentation in South-East Asia. It is precisely this pattern of medium-sized states that Scheidel (2019) has highlighted as crucial for political competition and investments in state capacity. Our model explains why Europe satisfied the Goldilocks principle of state fragmentation: not too little, not too much.
Finally, our model can be easily adapted in numerous ways to act as a ‘sandbox’ for subsequent research in comparative state development. We can enrich the model with additional factors such as culture and religion and ask: What is the relative importance of culture and religion versus geography? We are excited about the model’s possibilities for future scholarship.
Alsan, M (2015), “The Effect of the TseTse Fly on African Development”, American Economic Review 105: 382–410.
Broadberry, S (2021), “Accounting for the Great Divergence: Recent findings from historical national accounting”, VoxEU.org, 20 April.
Crafts, N (1977), “Industrial Revolution in England and France: Some Thoughts on the Question, ‘Why was England First’”, Economic History Review 30: 429–441.
Diamond, J (1997), Guns, Germs, and Steel, New York: W.W. Norton & Company.
Dupuy, T N (1979), Numbers, Predictions and War: Using History to Evaluate Combat Factors and Predict the Outcome of Battles, London: Macdonald and Jane’s.
Fernández-Villaverde, J, M Koyama, Y Lin and T-H Sng (2022), “The Fractured-Land Hypothesis”, National Bureau of Economic Research Working Paper 27774.
Hoffman, P T (2015), Why Did Europe Conquer the World?, Princeton, NJ: Princeton University Press.
Jia, R, G Roland and Y Xi (2021) “Institutions and power structure: Imperial China and feudal Europe compared”, VoxEU.org, 17 January.
Jones, E L (2003), The European Miracle, Cambridge: Cambridge University Press, 3rd ed.
Mokyr, J (2016), A Culture of Growth: The Origins of the Modern Economy, Princeton, NJ: Princeton University Press.
Mokyr, J (2017), “The persistence of technological creativity and the Great Enrichment: Reflections on the ‘Rise of Europe’”, VoxEU.org, 2 February.
Scheidel, W (2019), Escape from Rome, Princeton, NJ: Princeton University Press.
Turchin, P, T E Currie, E A L Turner and S Gavrilets (2013), “War, Space, and the Evolution of Old World Complex Societies”, Proceedings of the National Academy of Sciences 110: 16384–16389.