The Clash of Civilizations: A Cliometric Investigation Erin Fletcher & Murat Iyigun University of Colorado March 2010 Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 1 / 34
Motivation Measures of religious and ethnic fractionalization are causal factors of economic growth and development. Ethno-linguistic di erences have detrimental e ects on social stability, institutions and economic growth. Easterly and Levine (1993), Alesina et al. (1999, 2003), La Porta et al. (1999) and Mauro (1995). Religious fractionalization, in contrast, has ambiguous or slightly positive e ects. Alesina et al. (2003). Despite the impact of fractionalization on economic outcomes, there isn t much supporting evidence that fractionalization leads to internal con ict. Fearon and Laitin (2003), Collier and Hoe er (2005, 2007), Ray (2005). Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 2 / 34
Motivation This serves as an implicit refutation of the "Huntington hypothesis". That di erences of ethnic, religious and cultural identities are an important determinant of con ict. But fractionalization in all its demographic forms is endogenous in the long run. Thus, empirical work adopts an identi cation strategy which focuses on the medium run. Among the three fractionalization measures, there is some evidence that the religious kind is the most responsive to the environment. Alesina et al. (2003), Campos and Kuzeyev (2007). Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 3 / 34
The Objective We explore the determinants of religious, ethnic and linguistic fractionalization in the long run. Our main emphasis is on the role of con icts historically in religious fractionalization. We utilize some novel data that cover 953 con icts which took place in Europe, the Middle East, North Africa and the Near East between 1400 and 1900 CE. We have information on the actors involved which allows us to categorize them by type (power wars, ecclesiastical con icts, peasant revolts, etc.) We also identi ed the geographic location of these con icts which enables us to allocate them to one of 59 contemporary countries. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 4 / 34
Findings The frequencies and types of con ict in uenced contemporary levels of religious and to some extent ethnic and linguistic fractionalization too. Frequency of Muslim on Christian wars within a country s borders is a signi cant and positive predictor of its religious homogeneity. An additional incidence of violent con ict between Muslim and Christian parties within a country is linked with roughly 4 percent lower religious fractionalization. So did Muslim on Muslim con icts in particular, those between Sunni and Shi as. An additional con ict among the Muslims within a country is associated with between 4 to 7 percent less religious fractionalization. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 5 / 34
(More) Findings Countries history of con ict based on the religious identity of the parties involved have no e ects on linguistic or ethnic fractionalization. Muslim on Christian confrontations and intra-islam con icts also show some positive e ects on political fragmentation. Thus, con icts between religious groups led to religious homogenization by altering not only demographics but also borders. Long-run con ict histories between the 15th and 19th centuries also have strong explanatory power in cross-country di erences in polity measures. This is true even when current levels of ethnic, linguistic and religious fractionalization are controlled for. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 6 / 34
(More) Findings Similar results are obtained if data on 502 con icts over the period between 1400 and 1600 CE is analyzed. Other factors related to religious fractionalization include population density now and in 1500 CE (-); equatorial distance (+); Muslim-Christian bu er zones, such as Eastern Europe and the Balkans (+); Distance to holy sites, such as Jerusalem (-) and Mecca (+). Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 7 / 34
Some Implications If the types and frequencies of con ict mattered for fractionalization across countries, then there would be a shorter-term attenuation in the impact of fractionalization on con icts. In fact, Huntington (1993, p. 137, 208) himself acknowledged as much: Deep divisions within a country can lead to massive violence or threaten the country s existence. This latter threat and movements for autonomy or separation are most likely to arise when cultural di erences coincide with di erences in geographic location. If culture and geography do not coincide, they may be made to coincide through either genocide or forced migration, (emphasis added). Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 8 / 34
Some Relevant Work In political science and political economy, we have empirical work on the role of religious di erences or a nities on the propensity for con ict. Richardson (1960), Wilkinson (1980). The liberal peace view by Montesquieu, Kant & Angell; Jha (2008), Clingingsmith et al. (2009). In political economy, an established literature on incorporating appropriation and con ict into models of production. Haavelmo (1954), Hirshleifer (1991), Grossman (1994), Grossman and Kim (1995), Grossman & Iyigun (1995, 1997), Skaperdas (1992). The economics of religion Supply side: Barro & McCleary (2005), Ekelund et al. (1996), Iannaccone (1992). Demand side: Glaeser and Sacerdote (2003), Inglehart and Baker (2000). Other in uences of religion: Botticini & Eckstein (2005, 2007), Abramitzky (2008) and Iyigun (2007, 2008a, b). Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 9 / 34
Some Relevant History In the late 15th and early 16th centuries, the geographic areas of interest were typically more homogenous than they are today. The [European] continent... did have some potential bases of unity... Religion, language, and the residues of the Roman occupation probably made the European population more culturally homogeneous than any other comparable world area outside of China, Tilly (1992, p. 4, 5). In Central and Northern Europe: The Great Schism (1054 CE) and Coptic, Nestorian splits had also occurred in 5th CE. But these were primarily segmented markets. Precedents of the Protestant Reformation had been set, but Protestant-Catholic split did not become o cial until the Peace of Westphalia in 1648. In England, the Church of England splits in 1534 during Henry VIII s reign. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 10 / 34
More Relevant History In Eastern Europe and the Balkans: Ottoman territorial gains did not fully accrue until late-15th century to early-16th century. Conversions seems to have been limited, despite the head tax, cizye. In the Iberian Peninsula: The Spanish Inquisition under Isabella I and Ferdinand I, 1578. By the turn of the 17th century, the Iberian Peninsula is religiously purged. What about in Anatolia and the Middle East? Anatolia became predominantly Muslim starting with the 1071 Manzikert battle. The Middle East was probably the most religiously fractionalized, due to hysteresis & proximity to the Holy Lands. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 11 / 34
The Data The con ict data come mainly from Con ict Catalog dataset by Brecke (1999, in progress). For Europe, Middle East and North Africa, between 1400 CE and 1900 CE, they contain (i) when, (ii) who, (iii) where, (iv) how long. Patchy info on casualties. Richardson s magnitude 1.5 and higher (i.e. log of 32 or more deaths). MacEvedy & Jones (for population levels and crude proxies of income). POLITY IV for institutional score proxies and democracy measures. City Distance Tool by Geobytes. Alesina et al. (2003) for the three fractionalization measures. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 12 / 34
Con icts by Location & Century Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 13 / 34
Christian vs. Christian Con icts by Country Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 14 / 34
Muslim vs. Muslim Con icts by Country Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 15 / 34
Christian vs. Muslim Con icts by Country Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 16 / 34
The Empirical Analysis We estimate FRAC ij = λ 0 + λ 1 CHRISTIANMUSLIM i λ 2 CHRISTIANCHRISTIAN i + λ 3 MUSLIMUSLIM i + λ 4 X i + ε i, (2) Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 17 / 34
Empirics (cont.) FRAC i will in turn be religious, ethnic or linguistic fractionalization within country i. X t the control variables include Nine geographic region controls: WESTERNEU, CENTRALEU, EASTERNEU, NORTHERNEU, BALKANS, AFRICA, ASIA, MIDEAST and ISLAND; Population density in 1994, 1500 and 1000 CE, POPDEN, POPDEN1500, POPDEN1000; Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 18 / 34
Empirics (cont.) In less parsimonious speci cations, X t also includes: EQUATOR, LANDLOCK, LANDAREA; Distances to JERUSALEM, MECCA and ROME; Whether country i is a Christian or Muslim majority; Duration an average years in which the types of con ict took place: YRCM, YRMM, YRCC, DURCM, DURMM, DURCC. We rely on two di erent cuts of the con ict data (but primarily the rst one): Cross-country data covering 1400 to 1900 CE; Cross-country data covering 1400 to 1600 CE. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 19 / 34
Descriptive Statistics Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 20 / 34
(More) Descriptive Statistics Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 21 / 34
Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 22 / 34
Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 23 / 34
Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 24 / 34
Conjectures Why would Christian on Muslim and intra-islam confrontations depress religious fractionalization, whereas those among Christians have no e ect? There is no easy answer: one possibility is that di erences in the nature of interactions and the outcome of the history of con icts between Muslims versus Christians, Protestants against Catholics and Sunni versus Shi a. How? Two Abrahamic, monotheistic faiths with One-God/One True Religion duality inherent in both. At least thus far, there is a contrast between the inter-denominational historical dynamics within Christianity and Islam. Catholic vs. Protestant violence lasted centuries, but the Treaty of Westphalia in 1648 put an end to it. The Sunni-Shi a divide is as fundamental, but remains practically unresolved and a source of religious strife (e.g., Iraq). Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 25 / 34
Conjectures (cont.) If this is what is going on, then disaggregating con icts within Islam and Christianity further (to the denominational level) could be informative. With this in mind, we estimate FRAC ij = λ 0 + λ 1 CHRISTIANMUSLIM i + λ 2 CATHOLICPROTESTANT i + λ 3 CATHOLICORTHODOX i + λ 4 ORTHODOXPROTESTANT i +λ 5 SHIASUNNI + λ 6 X i + ε i, (2 ) Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 26 / 34
Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 27 / 34
Identi cation, Robustness & Other Specs. Reverse causality is a legitimate empirical issue. But channels of reverse causality strongly maintain that fractionalization produces more con icts. Our key ndings show a negative relationship. Hence, if there are reverse causality e ects, they are attenuating. Still, as an alternative we estimated equation (2) with con ict data running through 1600 CE instead of 1900 CE. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 28 / 34
Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 29 / 34
Identi cation, Robustness & Other Specs. Impact of history of con icts vs. fractionalization on institutions? Endogenous borders? Omitted variables? Endogeneity of con icts? Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 30 / 34
Fractionalization vs. Con ict History: Polity E ects Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 31 / 34
Endogenous Borders? Borders are clearly endogenous and it is highly likely that patterns of con ict had to do something with today s borders. With this in mind, we estimated STATEFORM it = λ 0 + λ 1 STATEFORM it + λ 2 CHRISTIANMUSLIM it 1 +λ 2 CHRISTIANCHRISTIAN it 1 + λ 3 MUSLIMUSLIM it 1 + γ c I c c 2000 i + j=1400 ρ j I j t + ε i, (4) where STATEFORM it is one of three alternative political fragmentation variables. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 32 / 34
Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 33 / 34
Final Thoughts The economic growth and development literature pays heed to the role of fractionalization in under-development. But all relevant demographic fragmentation is endogenous in the long run. In particular, given the historical context, we emphasize the e ects of con icts on sorting due to out-migration, redrawing of borders and atrocities. Back to Huntington. Fletcher & Iyigun (Colorado) Endogenous Fractionalization 03/13 34 / 34