Trade, Institutions and Ethnic Tolerance:

Evidence from South Asia

Saumitra Jha∗

April 4, 2012

Abstract

I provide evidence that the degree to which medieval Hindus and Muslims could providecomplementary, non-replicable services and a mechanism to share the gains from exchange hasresulted in a sustained legacy of ethnic tolerance in South Asian towns. Due to Muslim-specificadvantages in Indian Ocean shipping, inter-ethnic complementarities were strongest in medievaltrading ports, leading to the development of institutional mechanisms that further supportedinter-ethnic exchange.

Using novel town-level data spanning South Asia’s medieval and colonial history, I findthat medieval trading ports, despite being more ethnically mixed, were five times less proneto Hindu-Muslim riots between 1850-1950, two centuries after Europeans disrupted Muslimoverseas trade dominance. I provide evidence that these differences were transmitted via thepersistence of institutions that emerged to support inter-ethnic medieval trade. I characterisethese institutions and the lessons they yield for reducing contemporary ethnic conflict.

JEL codes: N25, O18, Z12, F10Keywords: Trade, Institutions, Culture, Religion, Cities, Ethnic Conflict, Social Norms, Peace

1 Introduction

On February 27th, 2002, a carriage of the Sabarmati Express carrying Hindu activists caught fire at

Godhra railway station in the western Indian state of Gujarat. At least 58 people were burnt alive.

In the weeks that followed, towns throughout Gujarat succumbed to ethnic violence that claimed

at least a thousand lives and forced 98,000 people into refugee camps. The city of Ahmadabad,

particularly in its medieval precincts, experienced 24 days of rioting that took the lives of more

∗Stanford Graduate School of Business. Address: saumitra@stanford.edu, 655 Knight Management Center, Stan-ford CA94305. I owe particular thanks to Susan Athey and Avner Greif, as well as Aprajit Mahajan, Kenneth Arrowand David Laitin. I am grateful for the suggestions of Ran Abramitsky, Amrita Ahuja, Roger Bolton, Feysal Devji,Claudia Goldin, Emeric Henry, Asim Khwaja, Michael Kremer, Prakash Kannan, Kimuli Kasara, Anjini Kochar,Timur Kuran, Jenny Kuan, Jessica Leino, Pedro Miranda, Monika Nalepa, Rohini Pande, Matthias Schuendeln, TNSrinivasan, Ashutosh Varshney, Steven Wilkinson, Gavin Wright, Joanne Yoong and seminar participants at Chicago,Clio, Davis, Harvard, Northwestern, Stanford, LiCEP, LSE, MIT, Princeton, NEUDC, NYU, UBC, UWO, the all-UC economic history group and the World Bank. Sangick Jeon and Astasia Myers provided much valued researchassistance.

1

than 324 people. Once Mahatma Gandhi’s headquarters for non-violent teaching, Ahmadabad has

since become notorious for the frequency and intensity of its ethnic conflict.

The patterns of violence in Ahmadabad stand in sharp contrast to those that prevailed in Surat,

a historic port city just 140 miles south. Surat too, is divided into an old medieval city and newer

settlements. Surat’s proportion of Muslim residents, at 12.3 percent, mirrored Ahmadabad’s 13.0

percent (Census of India, 2001). Yet, over the course of the 20th century, Surat witnessed virtually

no religious violence. Even in 1969, when most other large Gujarati cities, particularly Ahmadabad,

succumbed to religious rioting, Surat was unaffected (Engineer, 1995, Varshney, 2002).1 It took

the destruction of the Babri Mosque in Ayodhya in December 1992 to end Surat’s unusual record

of religious peace, when six days of rioting left 175 dead. Following the Godhra incident ten years

later, Surat also experienced six days of violence that led to at least nine deaths. Yet, unlike in

Ahmadabad, where riots were concentrated in the old city, Surat’s riots were largely focused in

newer migrant settlements (Varshney, 2002). Despite possessing many attributes that commonly

correlate with religious violence, Surat’s reputation as an “oasis of peace” persists.2

The medieval-era precincts of India’s cities, which are often relatively poor and house Hindus

and Muslims in close and crowded proximity, are often cited in the popular press and by academics

as natural repositories for religious tensions that routinely explode into violence.3 Indeed, the

potent relationship between poverty, ease of insurgency and ethnic violence found in studies at the

cross-country level (eg Fearon and Laitin, 2003, Miguel, Satyanath and Sergenti, 2004) seem to find

ample validation in the crowded precincts of India’s towns. Hindu-Muslim violence has resulted in

more than 40,000 deaths or injuries since 1947, the overwhelming majority in towns and cities.4

However, the aggregate statistics of religious violence in India mask a great diversity in Hindu-

Muslim relations. Even within Gujarat, one of India’s wealthiest states, the contrast between

the relatively peaceful old city of Surat and its newer settlements and the violence throughout

1Between 1885 and 1985 Surat experienced two incidences of Hindu-Muslim violence, with one death in each case.In 1985-1986, when another wave of violence combining religious and inter-caste tension hit Gujarat, Surat’s riotswere between castes rather than religions (Varshney, 2002)[p228].

2This term was used to describe Surat during the Gujarat riots of 2002 by the Times of India editorial page,February 1st, 2007.

3See Engineer (1988) on Delhi, Vadodara and Hyderabad, Das (1990)[p.65] on Delhi. For Aligarh, Brass(2003)[p.162-3] cites Upar Kot–the walled medieval city or “the upper fort” as the location for all 29 mohallas(neighbourhoods) prone to religious violence in 1978.

4Wilkinson and Varshney and Wilkinson (2004) calculate that 93 percent of religious riot deaths occurred in urbanareas (settlements with above 5000 in population that possess a municipality) between 1950 and 1995. As the authorsdiscuss, this figure may be subject to reporting bias, due to the greater likelihood of counting urban deaths.

2

Ahmadabad city suggests that old towns in India exhibit patterns of religious conflict that are

not completely explained by factors commonly associated with ethnic violence: the incidence of

poverty, ethnic composition, state or town-level electoral incentives or even rioters’ ability to avoid

the police. Understanding how communities that have all the ingredients for conflict still succeed

in maintaining tolerance may provide insight into how other communities may do the same.

In this paper, I provide evidence that the degree to which Hindus and Muslims could provide

one another with complementary, non-replicable services prior to European colonization has gener-

ated a legacy of reduced ethnic conflict even in towns that might otherwise seem likely locations for

violence.5 I use the fact that due to the coordination of overseas markets through Muslim pilgrim-

ages such as the Hajj, Muslims enjoyed exogenous complementary and non-replicable advantages

in accessing Indian Ocean trade routes in the medieval era. I find that medieval overseas trading

ports, which served as the geographical focuses of Hindu-Muslim complementarities in trade, were

five times less prone to Hindu-Muslim riots and around twenty-five percentage points less likely

than otherwise similar towns to experience a religious riot between 1850 and 1950, two centuries

after Europeans disrupted Muslim advantages in overseas trade. Between 1850 and 1950, medieval

port towns were around ten times less likely to experience their first outbreak of Hindu-Muslim

rioting in any given year. Though the differences in patterns of conflict between medieval ports and

other towns appear to diminish over time, Gujarati medieval trading ports continued to experience

fewer riots and exhibited less widespread religious rioting in the aftermath of the burning of the

Sabarmati Express in 2002.

The paper draws upon a novel town-level dataset that combines data on medieval trade and

political structures with colonial-era indicators of demography and development to provide evidence

that the differences between erstwhile ports and other towns are not the result of factors, such as

a more mixed ethnic composition or poverty, that are often suggested to explain ethnic violence.

Instead, I show that medieval ports are more ethnically mixed and poorer than similar towns,

both factors that are commonly associated with greater violence. In fact a medieval trade legacy

appears to reduce violence the most in larger towns that are more ethnically mixed. Further, the

results do not appear to be driven by the selection of medieval ports by Muslim traders based upon

5Technically, two actions are complements if 1) adopting one does not preclude adopting the other, and 2) wheneverit is possible to implement them separately, the sum of each return cannot be greater than doing them together.

3

unobserved pre-existing factors that might have fostered a more tolerant local population. Because

of the severity of Indian Ocean storms, sheltered harbours were historically prized as locations for

medieval ports. Thus access to natural indentations on the historical coastline, or “medieval natural

harbours”, provide an exogenous determinant of medieval trade, one that I find later diminished

in determining the location of modern ports. Yet, those towns that became ports because of their

location at medieval natural harbours exhibit similar reductions in Hindu-Muslim violence to other

medieval port towns.

I then evaluate the potential mechanisms through which a legacy of medieval trade might have a

lasting effect on ethnic violence over two centuries later. I confirm that the differences in violence in

medieval trading ports and other towns arise specifically from overseas trade in the medieval era–

where Muslims enjoyed non-replicable complementarities– rather than medieval land-based trade–

where Hindus could locally replicate Muslim trading networks– or modern trade– where European

intervention eroded Muslims’ overseas advantages. I next show that continued colonial-era trade

in medieval ports does not explain the results. Due to heavy silting of inlets and river mouths

during the monsoon rains, a number of medieval ports have increasingly ceased to be accessible

to shipping over time. Those medieval ports that subsequently became inactive or inaccessible to

overseas shipping show a remarkably consistent legacy of reduced modern religious violence, while

modern overseas ports, a useful placebo comparison, do not.

A puzzle these results raise is why historical patterns of inter-religious complementarity might

have lasting effects on colonial and contemporary religious tolerance even after the decline of Mus-

lim advantages in trade. I draw upon qualitative fieldwork conducted in Gujarat in 2006-7 and

historical case studies to interpret the quantitative evidence as consistent with the development of

‘institutions’, systems of complementary elements that include cultural norms, beliefs and organi-

zations, that are subsequently taken as given by individuals (Greif and Laitin, 2004, Greif, 2005,

Milgrom and Roberts, 1990). These complementary institutional elements appear to have emerged

to reinforce the gains from peaceful exchange between Hindus and Muslims in the medieval era and

to have persisted even after increased European involvement in the Indian Ocean precipitated a de-

cline of Muslim advantages in trade. While the institutional elements that persist differ in different

medieval ports, they share an economic logic. In some communities, additional ethnic specialisation

in complementary roles occurred during the medieval period– these communities appear to have

4

maintained inter-ethnic complementarity even after the decline of the original complementarity in

trade. In other communities, organisations and cultural norms emerged that appear to have miti-

gated the incentives for ethnic violence by allowing the gains from inter-ethnic trade to be shared

more equitably between groups and by building inter-ethnic trust.

Such mechanisms explain the main empirical findings. With continued incentives that support

inter-ethnic trust and exchange, mobile minority groups are more likely to stay, even in towns that

are relatively poor. Indeed, medieval ports, despite being poorer, appear to have retained more

ethnically diverse populations, even during extreme shocks, such as the Partition of the subcontinent

along religious lines in 1947, which led to mass migration and ethnic cleansing elsewhere. Further,

with continued ethnic specialisation in complementary activities, ethnic minorities are less likely

to compete with other groups, leading to less violence aimed at reducing competition. In fact,

with ethnic specialisation, larger ethnic minorities are likely to compete more within themselves,

dissipating oligopolistic rents and lowering incentives for violence against minorities to seize wealth.

Thus, unlike in other towns where more mixed populations may lead to more political and economic

inter-ethnic competition and violence, the persistence of community level institutional elements

that foster inter-ethnic complementarity and facilitate inter-ethnic redistribution may have lowered

the incentives for violence even in poorer towns with more ethnically-mixed populations. This is

precisely what we observe in erstwhile medieval ports.

I conduct a series of further auxiliary tests. I find evidence that medieval-era human capital

investments and the institutional legacy of inter-ethnic trade complementarities are themselves

complements in reducing subsequent violence. I also find that workers in medieval ports, while no

longer more involved in transport and shipping, are more likely to show higher proportions engaged

in trade and commerce in 1991, consistent with higher levels of trust. However, the reverse is true

for workers in colonial era ports and towns on medieval internal trade routes, where incentives

to develop reinforcing institutions supporting trust were also lacking. Medieval port voters also

show large and systematic swings against the ruling party, the BJP, in the elections immediately

following the Gujarat riots in 2002. Along with mitigating incentives for ethnic violence, the insti-

tutional legacy of historical inter-ethnic complementarities appears to continue to shape economic

and political decision-making in other fundamental ways as well.

This paper follows in a rich intellectual tradition evaluating the long-term effects of historical in-

5

stitutions (eg Acemoglu, Johnson and Robinson, 2001, Banerjee and Iyer, 2005, Guiso, Sapienza and

Zingales, 2008, Nunn and Wantchekon, forthcoming). By stressing the central role of inter-ethnic

complementarities in encouraging cooperation and discouraging conflict between ethnic groups,

this paper introduces a new dimension into an important set of studies that have sought to un-

derstand the role of ethnicity and inter-ethnic inequalities as a determinant of civil conflict and

public goods provision(eg Horowitz, 1985, Alesina and La Ferrara, 2005, Fearon and Laitin, 2003,

Miguel et al., 2004, Habyarimana, Humphreys, Posner and Weinstein, 2009, Baldwin and Huber,

2010, Jha and Wilkinson, 2011). The paper suggests that if the structure of economic incentives for

exchange, mobility and violence between ethnic groups are not considered, factors may be omitted

that can dramatically alter the impact of more proximate causes such as ethnic heterogeneity and

inter-ethnic inequalities on modern indices of peace, public goods provision and growth.6

This paper most closely builds upon important studies that have recognised the vital role of in-

stitutions in enhancing cooperation and mitigating ethnic violence, both in theory (Kandori, 1992,

Greif, 2005, Fearon and Laitin, 1996) and in the South Asian context in particular. Based on de-

tailed fieldwork, Brass (2003) finds the existence of what he terms “institutionalised riot systems”:

concerted action by local elites to maintain fissures along religious lines, for local, sometimes elec-

toral, gain. Wilkinson (2004) focuses upon the state-level incentives for fomenting and responding

to religious violence. He finds that when elections are close, riots are used by politicians to reaffirm

religious identities and thus to sway votes. Varshney (2002)’s work provides a very useful counter-

point to the work of Wilkinson and Brass. Varshney focuses on the importance of cross-religious

social capital or “civic engagement” in defusing religious tension. “Peace committees” develop from

existing cultural, political or business groups, that cross religious lines.7

These works provide extremely valuable insights into the proximate causes of ethnic violence

in contemporary South Asia and beyond. This paper furthers these studies by analysing the

6This paper also speaks to an emerging literature that explores the reasons why poor societies become and remainethnically diverse, despite the social costs. As we shall discuss, the role played by exogenous complementarities andreinforcing institutions is common to a number of ethnically diverse societies, one that adds to intriguing worksthat stress the roles of geographical constraints to migration and misguided governmental policies in fostering ethnicdiversity (Michalopoulos, forthcoming, Fearon and Laitin, 2011, Dippel, 2011).

7Varshney’s use of pair-wise case studies is especially illuminating. Each pair consists of one town where religiousriots are rare: Calicut, Lucknow and Surat–and one where they are common: Aligarh, Hyderabad and Ahmadabad. Itis reassuringly consistent with the theory outlined here that two of the three cities that Varshney identifies as enjoyinghigh levels of civic engagement–Calicut and Surat–were once major medieval trading ports and the three cities wherecivic engagement between Hindus and Muslims ultimately failed–Aligarh, Hyderabad and Ahmadabad–were centresof Muslim political patronage, where Muslim clients acted as substitutes for Hindu clients.

6

historical structure of incentives that led to the contemporary “institutional” environment.8 This

paper argues that contemporary inter-ethnic civic engagement, close ethnically-delineated elections,

inter-ethnic inequalities and ethnic violence are all in part the legacy of more than a thousand years

of the presence or absence of exogenous complementarities between ethnic groups.

Finally, this paper resolves an important paradox that concerns the effects of trade on tolerance.

At least as early as Montesquieu, philosophers have argued that commerce encourages civility be-

tween individuals, as the loss of trade makes conflict more costly (Hirschman, 1977)[pg.78,95]. Yet,

historical and contemporary examples abound of even the most commercially-oriented minority

groups around the world becoming repeated targets of ethnic violence and expropriation (Benbassa

and Rodrigue, 2000, Landa, 1994, Chua, 2003, Jha, 2007). By outlining the conditions that favour

peaceful exchange between groups over time, this paper shows the limitations of inter-ethnic spe-

cialisation and exchange as a means of reducing ethnic tension, while also pointing to potential

policy interventions that may bolster the effects of trade in encouraging inter-ethnic peace.9

Section 2 describes the structure of medieval incentives for exchange between Hindus and Mus-

lims and provides a taxonomy of the institutions that emerged in medieval ports. Section 3 explains

the empirical strategy. Section 4 describes the construction of the data. Section 5 presents results

from religious conflict across India between 1850 and 1950, as well as for the particular case of the

Gujarat riots of 2002. Section 6 concludes.

2 Incentives for exchange in the medieval Indian Ocean

A simple theoretical model can be used to illuminate why Hindus and Muslim traders enjoyed

an enduring environment of peaceful co-existence in the Indian Ocean region when many other

commercial ethnic minorities have not. It is useful to provide an intuitive sketch of the model and

its relevant predictions (please see Author (n.d.) for details of the formal theory).

The model focuses on settings where there are two types of agents: local and non-local. Non-

locals differ from locals only in that they have better outside options. In the case of medieval India,

8This paper also adds to other work on the proximate causes of Hindu-Muslim violence has stressed income shocksdue to changes in rainfall (Bohlken and Sergenti, forthcoming).

9This problem has some parallels with the literature on trade and conflict between nations (see Polachek andSeiglie (forthcoming) for an overview). A crucial distinction is that unlike most trading nations, individuals livingin the same town tend to have similar or replicable factor endowments and thus a key challenge is to maintaincomplementarity over time. The comparative advantages of each nation are generally assumed.

7

“non-local” Muslim traders enjoyed resources, including information and networks, that linked

them to the Middle East and the rest of the Islamic World. These external resources made it less

costly for Muslims to leave a particular town and go elsewhere. In contrast, the resources of “local”

Hindus tended to be also concentrated locally.

In the model, individuals from either group have the following choices every period: to stay or

leave town, to produce a good for exchange, and to attack any other agent that they encounter.

Such violence is destructive, but violence may be useful for seizing the victim’s property and to

deter or punish the victim’s actions.“Strong” individuals may exist who are more likely to prevail

in a violent attack against weaker opponents. The model can be used to find conditions that favour

“peaceful co-existence” over time: an equilibrium with a mixed population of locals and non-locals,

full production, no out-migration and no violence.

In the environment above, an important condition that favours peaceful co-existence is that

non-locals and locals produce complementary goods or services. To see this, consider first the

alternative: that locals and non-locals provide substitute goods and thus are competitors. Then,

with repeated interactions, a strong local will have an incentive to attack weak non-locals, as this

allows that local not only to seize the non-local’s property but also to encourage non-locals to

leave, reducing the future competition the local faces. In fact, non-local competitors provide more

attractive targets of violence than weak locals, as local competitors are harder to encourage to

leave due to their lower outside options. Thus rather than class violence, societies where local and

non-local groups compete are likely to exhibit greater ethnic violence.10

In contrast, when ethnic groups provide complementary goods or services to one another, then

the incentive to attack non-locals falls over long time horizons. If non-locals leave if attacked, locals

will face reduced supply and higher future relative prices for goods that only non-locals can provide.

The more that non-local goods increase the value of local goods, the more valuable the presence of

non-local suppliers and the lower the incentives for ethnic violence.

Thus inter-group complementarity can support peaceful co-existence over time. However, even

with repeated interactions, peaceful co-existence will fail if members of one group are able to cheaply

replicate or violently seize the resources that make members of the other group desirable trading

10An alternative theory for why ethnic violence is more likely to occur than class violence is that mobilizationrequires resources available to the wealthy, who prefer ethnic violence to class violence (Esteban and Ray, 2007). Forevidence on economic competition and ethnic violence in the United States, see Olzak (1992).

8

partners. Therefore, to maintain inter-group complementarity over time, it is necessary for the

source of each group’s complementarity to be difficult for others to acquire. Complementarities

emerging from expropriable assets such as wealth, machines or land, or even artisanal skills and

other forms of human capital that may be replicated, can therefore fail to sustain peaceful co-

existence in the long term.

Furthermore, in settings where the non-local group is a small minority, even such “robust”

complementarities may be insufficient to maintain peaceful co-existence. With few competitors,

members of a minority group that provide non-replicable complementary services can enjoy high

relative prices and substantial profits. This has been a common feature in the histories of many

ethnic minority trade networks. Particularly in times of resource shock or crisis, when strong locals

discount the future more highly relative to the present, such high profits may result in a temptation

for agents to engage in violence to seize these profits even at the cost of losing future gains from

trade.11 Thus the maintenance of peaceful co-existence over time will benefit from mechanisms that

redistribute the surplus between groups and thereby reduce incentives to violently expropriate.12

From the seventh century to the seventeenth century, Muslim traders involved in transoceanic

commerce satisfied all three of the conditions outlined above for sustaining peaceful co-existence.

First, there were Islam-specific advantages to trade across the Indian Ocean. Pilgrimages, particu-

larly to Mecca, coordinated the development of the world’s largest textile market during the Hajj

(Lombard, 2000). The Hajj was supplemented by pilgrimages (ziyaret) to other sites, such as Fustat

(Cairo), Kerbala, Basra and in Hadramaut (Yemen), that all fostered regional trade. Muslims had

strong preferential access to these pilgrimage routes, and the markets they induced.

Second, Muslim advantages in oceanic trade stemmed from preferred access to trade networks,

11See e.g. Landa (1994) for a discussion of how high relative prices and the resultant wealth led to an expropriationof the Chinese in Southeast Asia, and Chua (2003) for many other examples around the world. This is also consistentwith the findings of Baldwin and Huber (2010) on the role of between group inequality.

12As discussed in more detail in Author (n.d.), a rational response by minority members is to engage in greatertransfers to the strongest locals, for whom the constraint not to engage in violence binds first– often the local ruler.“Protection” by dictators has been a common feature of the histories of many market-oriented ethnic minorities inboth medieval and contemporary developing country settings (Chua, 2003, Benbassa and Rodrigue, 2000). Howeverin these settings, equilibria can exist where elites demonstrate their “protection” of minority groups and thus extractgreater long-term transfers by intermittently allowing pogroms by weaker locals.

With sufficient organisation within the minority community, a more efficient equilibrium may also exist wherenon-local minorities provide public goods that benefit the local group as a whole, as well as providing lower transfersto the elite. Such a pattern of transfers reduces the incentive by non-elites to engage in pogroms, as well as reducingthe ability of rulers to extract transfers over the long term. Thus they are incentive-compatible for minority members.Such strategies have historically been employed by highly organised minority groups, including the Bohra Muslimtrading community in India and the Ismaili community in East Africa (Penrad, 2000).

9

which are difficult to steal or replicate. A key characteristic of trade networks is that they enjoy

increasing returns to scale. The remarkable scale of the Hajj in particular was such that it was

prohibitively costly for even a substantial number of Hindus to replicate. Since trade networks

are also intangible, they were also impossible for Hindus to steal. Oceanic trade differed from

land-based or riverine trade routes in this manner, as most long distance land-based trade can be

divided into relays of shorter distances, each of which can be replicated by a member of the local

group.13 Most sea trade routes, however, cannot be replicated in relays. It is therefore at towns

with direct access to the Indian Ocean that Muslim advantages in Middle Eastern trade became

most relevant and gains from exchange between Hindus and Muslims were most pronounced.14

Third, Muslims had access to a natural, decentralized mechanism of redistribution of the surplus

from trade to the local population: increased intra-Muslim competition due to the relative ease of

entry by any Muslim into Indian Ocean trade. Unlike most kin-based trade networks that have high

barriers to entry (Rauch and Casella, 2002), entry into Islamic trade networks was relatively cheap

for all Muslims. Pilgrimages provided a clear coordination device, so even non-merchant and newly

converted Muslims could enter trade. Family or community ties were not necessary to follow estab-

lished pilgrimage routes, and indeed many pilgrimages were financed through trade (Ibn Battuta,

1355, di Verthema, 1503, Lombard, 2000). Though trading “communities” did emerge, members of

these communities often were in economic competition either within their own communities or with

other Muslim trading communities (Penrad, 2000, Subrahmanyam, 2000).15 Incipient and actual

entry by Muslim competitors could improve the terms of trade for the local population whenever

relative prices for non-local goods became too high.16

13In the Indian context, for example, a Hindu trader in inland Aligarh who sought to sell to markets in the MiddleEast would not be forced to rely on a Muslim in Aligarh, but could hire a Hindu merchant to transport his goods tothe coast.

14Sizeable colonies of Hindu and Jain traders were established in some parts of the Middle East - particularlyneighbouring territories of Persia and Iraq. The French trader, Jean de Thevenot (1633-1667) noted the presence ofbania moneylenders in Isfahan, Basra and Hormuz (Mehta, 1991). However, shipping was dominated by Muslims,and the great textile mart at Mecca remained exclusively Muslim. The fate of Ludovico di di Verthema (1503), whovisited Mecca disguised as a Muslim, illustrates this. Exposed as a Christian in the Holy City, he was sold as a slaveto a merchant travelling to India. The chronicles of his journeys provide useful contemporary detail on the port townsof early 16th century India.

15Though Muslims dominated shipping, other Middle Eastern trading groups, including those of Jews and Ar-menians, were also involved in the Middle Eastern trade and cooperated in the Karimi convoys across the IndianOcean (Goitein, 1966). It is likely that the presence of Muslim competition made Indian ports less profitable butmore tolerant destinations for these groups as well.

16Though this issue has hitherto not attracted much attention in existing work, there appears a systematic rela-tionship between the strength of different Middle Eastern empires (e.g. Fatimids versus Abbasid) and conversion ofvarious sects of Islam on India’s coasts. For example there was a wave of conversion to Shaf’ii Islam (more common

10

Trading ports in the Indian Ocean were thus well-favoured to provide geographical loci for

peaceful co-existence and trade between Hindus and Muslims. From the seventh century onwards,

Muslims, both immigrants to India and indigenous converts, dominated the shipping trade in

the Indian Ocean and Muslim trading networks expanded along both coasts (Arasaratnam, 1994,

Dasgupta, 2004) (see Figure 1[b]). Muslim dominance of overseas trade continued for close to a

thousand years. Though the Portuguese discovery of routes to the Indian Ocean in 1498 did not

entirely disrupt the flow of trade, the Portuguese did destroy the commerce of a number of key

trading ports, often via blockade. The end of Islamic trade dominance was further expedited

in the face of increased competition by the Dutch and English, and the disintegration of the

Mughal empire. Mughal ports, such as Masulipatam, Surat and Hughli, gave way to competition

from Madras, Calcutta and Bombay (Dasgupta, 2004). Muslim trading networks continued to be

important in trade with Southeast Asia and Zanzibar, but the expansion of colonial rule to these

regions brought competition from non-Muslim traders operating under colonial protection (Bose,

2006). By the beginning of the eighteenth century, the era of Muslim trade dominance in the Indian

Ocean was over, and many medieval trading ports ceased to be commercially important.

Thus, for close to a thousand years, inter-group complementarities existed between Hindus and

Muslims in medieval trading ports. However, even during this period, peaceful co-existence could

still be threatened by shocks. Examples include resource or political shocks that threatened the

survival of strong locals, such as emerged with the increased instability of regional kingdoms and

the Mughal empire, or the exogenous development of new substitutes to Muslim shipping, such

as occurred with the European discovery of routes to the Indian Ocean. Thus, higher mutual

incentives existed in medieval ports than other towns for residents to invest in and develop com-

plementary mechanisms to maintain the incentives for peaceful co-existence even in the presence of

such shocks. Insofar as these mechanisms, once developed, were costly to reverse by any individual

agent, they can be considered part of the “institutional” environment that shape an agent’s subse-

quent incentives for peaceful co-existence. Such institutional mechanisms appear to have persisted

in shaping Hindu-Muslim interaction at medieval ports long after the decline of Muslim advantages

in the Arabian peninsula) during Islam’s early centuries, while Shia conversion in Indian ports, such as that of theBohra community, coincides with the expansion of the Fatimid caliphate, and the centrality of Fustat (Cairo) andYemen to trade.

11

in trade that encouraged their initial development.17

Table 1 summarises evidence gleaned from both fieldwork conducted by the author in 2006-07

and the historical record for the different institutional mechanisms that emerged in medieval ports.

Medieval-era organisations appear to have fulfilled two distinct, but complementary roles. One set

of organisational mechanisms encouraged group specialisation and raised the costs of replicating

the services provided by another ethnic group. Specialisation in skilled activities was encouraged

through a system of apprenticeships that were often exclusively limited to members of the same

ethnic group (Campbell, ed, 1899, Haynes, 1991). Norms and own-group social sanctions also

emerged that raised the costs of replicating another group’s activities. A prominent example of

this was the norm of Kaala-paani (“black water”): that Hindus that sailed offshore would be

outcaste by their own community.18

A second set of mechanisms helped reduce the incentives for violence, whether by coordinating

responses to crises or by sharing the gains from exchange. In Gujarat and Malabar, merchant guilds

and inter-religious organisations helped organise both boycotts and joint petitions to political figures

to seek redress when members of one religious group were threatened by strong individuals (al Mal-

ibari, 1528, di Verthema, 1503). Organisations also emerged to encourage repeated interactions

between members of different religious groups, which encouraged trust and the formation of joint

ventures (Dasgupta, 1994). Muslim traders around India provided commercial taxes and explicitly

endowed local public goods, including water projects and even Hindu temples (Risley, Meyer, Burn

and Cotton, eds, 1909, Bayly, 1989).19 Relative to other areas, conversion to Islam and immigration

17Greif and Laitin (2004) provide a general theory of how equilibria can be self-reinforcing and self-undermining,altering ‘quasi-parameters’ that individuals take as given. I wed their work to a central insight from the theory ofmonotone comparative statics– that the presence or absence of complementarities underlie most robust conclusionson how optimizing behavior changes in response to shocks (Topkis, 1998, Milgrom and Roberts, 1990). An intriguingpuzzle that Greif and Laitin (2004) raise is on the role of foresight– if individuals are rational, they should foresee thattheir cumulative actions may lead to the undermining or reinforcing of the equilibria that they play. The presence ofcomplementary or substitute investments in institutional elements can resolve this puzzle. If individuals are perfectlyrational, the critical sufficient condition under broad range of assumptions for such self-reinforcing equilibria to emergeis that all such investments be mutual complements (Milgrom and Roberts, 1990).

18This cultural norm, though common to many Indian sea ports, was particularly prevalent in Calicut and otherports in Malabar (Bouchon, 2000). However, Kaala-paani proscriptions on Hindu travel overseas were not widelyfollowed in Gujarati ports until Muslim dominance of overseas trade began to decline (Mehta, 1991). Thus it may bethat these institutions were established as a response to growing competition between groups.

19An important example is that of the trading port of Veraval, adjacent to the Hindu pilgrimage centre of Somnath.Temple authorities were willing to provide land on their own grounds to a Muslim trader to build a mosque, not longafter Mahmud of Ghazni sacked the temple in 1026, a event long perceived to have resulted in a massive deteriorationin Hindu-Muslim relations. Hindu temple authorities encouraged Muslim traders as taxes on commerce provided amajor component of their revenues. Such inter-ethnic transfers were a common practice in many medieval ports inGujarat (Thapar, 2004).

12

Tab

le1:

Taxon

om

yof

inst

itu

tion

sin

Ind

ian

med

ieval

port

s

Coa

stM

ajor

med

ieva

l po

rtsM

uslim

trad

ing

grou

psSt

rong

co

mm

unity

or

gani

satio

n

Med

ieva

l co

mpl

emen

tary

se

rvic

es

Med

ieva

l ins

titut

ions

19th

cen

tury

/ co

ntem

pora

ry

com

plem

enta

ry

serv

ices

19th

cen

tury

/ con

tem

pora

ry in

stitu

tions

Con

tem

pora

ry

resi

dent

ial

inte

grat

ion

Add

ition

al

barr

iers

to

repl

icat

ion

Inte

r-re

ligio

us

orga

nisa

tions

Tran

sfer

m

echa

nism

s

Add

ition

al

barr

iers

to

repl

icat

ion

Inte

r-re

ligio

us

orga

nisa

tions

Tran

sfer

m

echa

nism

s

Guj

arat

Bro

ach,

C

amba

y,

Dw

arka

, Po

rban

dar,

Sura

t,

Som

nath

-V

erav

al

Ara

bs, D

audi

B

ohra

s, M

emon

s, N

izar

i Ism

ailis

Yes

7Tr

ans-

ocea

nic

ship

ping

App

rent

ices

hip

rest

rictio

ns5

Mer

chan

t G

uild

s, Po

litic

al

dele

gatio

ns2

Com

mer

cial

ta

xatio

n3 , Joi

nt

vent

ures

2

Aga

te,

Car

nelia

ns1 ,

Silv

er th

read

w

eavi

ng5 , Y

arn

cutti

ng, D

iam

ond

cutti

ng, (

Gul

f / S

E A

sia

netw

orks

)4

App

rent

ices

hip

rest

rictio

ns5 ,

Adm

inis

trativ

e sa

nctio

ns,

Soci

al sa

nctio

ns(K

aala

-paa

ni)1

Peac

e co

mm

ittee

s, B

usin

ess

asso

ciat

ions

4 , N

atio

nal

polit

ical

par

ty

``m

inor

ity

win

gs''7

Polit

ical

do

natio

ns,

Join

t ve

ntur

es6 ,

Loca

l pub

lic

good

s, D

isas

ter

relie

f7

Yes

4,7

Mal

abar

/ C

entra

l W

est

Bha

tkal

, C

alic

ut,

Cra

ngan

ore,

C

ochi

n,

Man

galo

re,

Qui

lon

Ara

bs, B

eary

s,

Koy

as, M

appi

las,

Naw

aiya

ts

Non

e ev

iden

tTr

ans-

ocea

nic

ship

ping

Soci

al sa

nctio

ns

(Kaa

la-p

aani

)8Po

litic

al

dele

gatio

ns8

Com

mer

cial

ta

xatio

n, Jo

int

vent

ures

, Eas

e of

con

vers

ion,

Lo

cal p

ublic

go

ods8

(Gul

f net

wor

ks),

Com

mod

ities

tra

ding

4,9

Soci

al sa

nctio

ns(K

aala

-paa

ni)2

Pe

ace

com

mitt

ees,

Cha

mbe

rs o

f co

mm

erce

, C

lubs

4,9

Loca

l pub

lic

good

s9Y

es4,

9

Cor

oman

del

(Eas

t)K

ilakk

arai

, M

asul

ipat

nam

, N

egap

atna

m,

Pulic

at,

Tutic

orin

, V

izag

apat

nam

Mar

raik

ayar

s, Pe

rsia

ns, L

abba

isY

es10

,11

Tran

s-oc

eani

c sh

ippi

ngN

one

evid

ent

Non

e ev

iden

tC

omm

erci

al

taxa

tion,

Join

t ve

ntur

es10

, V

olun

tary

do

natio

ns to

H

indu

-spe

cific

pu

blic

goo

ds11

pear

l div

ing,

co

asta

l shi

ppin

g,

(Gul

f/ SE

Asi

a ne

twor

ks)10

Non

e ev

iden

tR

egio

nal

polit

ical

pa

rties

10

No10

,11

Sour

ces:

1: M

ehta

(199

1), 2

: Das

gupt

a (2

000)

3: T

hapa

r (20

04),

4: V

arsh

ney

(200

2), 5

: Gaz

ette

er o

f the

Bom

bay

Pres

iden

cy (1

899)

, Hay

nes (

1991

), 6:

Con

cern

ed C

itize

ns T

ribun

al (2

002)

, 7:

per

sona

l int

ervi

ews,

Bla

nk (2

001)

, 8: a

l Mal

ibar

i (15

28),

di V

erth

ema

(150

3), B

ouch

on (2

000)

, 9: O

sella

(200

3), 1

0: M

ore

(199

7), 1

1: S

. Bay

ly (1

989)

13

from the Middle East was encouraged by local populations in Malabar ports (al Malibari, 1528),

reducing costs of entry into trade and further increasing within-Muslim competition. The sharing

of the gains from trade, whether through increased intra-group competition, explicit inter-group

transfers, or joint ventures between groups are likely to have provided Hindus and Muslims in

medieval ports reduced incentives for inter-ethnic violence in times of crisis.

As Table 1 indicates, a number of these mechanisms have persisted and evolved through the 19th

and 20th centuries. A tradition of inter-ethnic participation in organisations continues to flourish

in a number of towns that were once trading ports in the medieval period, including in business

organisations, clubs and even political parties (More, 1997, Varshney, 2002). These organisations

may have also facilitated the maintenance of complementarities between groups. In contemporary

Surat, for example, Muslims and Hindus have continued to explicitly adopt complementary roles in

production, long after the demise of Surat’s trade. Diamond-cutting and silver-thread weaving are

almost exclusively conducted by Muslim workers, while complementary roles in both production

processes are handled by Hindus and Jains (Varshney, 2002).20

Though the institutional mechanisms that emerged in medieval trading ports share an economic

logic, relations between Hindus and Muslims do differ across ports. For example, Muslim traders in

Calicut and Surat showed (and continue to show) evidence of social and residential integration.21

These cities are also notable for the presence of contemporary mechanisms, such as inter-religious

peace committees, for defusing conflict (Varshney, 2002). On the other hand, Muslim traders

in the Coromandel Coast lived in more segregated communities; instead they engaged in explicit

transfers to the majority community by endowing Hindu temples and pursued joint trading ventures

with the local rulers of Ramnad (More, 1997, Eaton, 1993, Bayly, 1989). Thus, the presence of

inter-ethnic complementarities has not necessarily resulted in widespread social and residential

integration between members of different religions. Instead, a common feature of these ports is

the development of institutional mechanisms that helped reduce the effect of shocks on violence,

either explicitly or by sharing the gains from trade between groups. These institutional mechanisms

also appear to have fostered inter-ethnic complementarities in new sectors following the decline of

20For other examples, including on restrictions on Hindu entry into the agate trade in Cambay, see Mehta (1991).21As one respondent from the Bohra (traditional Muslim trading) community in Surat told the author in 2007:

“When we went to our apartment complex in Nanpura [a predominantly non-Muslim neighbourhood], they asked us“are you ‘H-Class’ [Hindu] or ‘M-Class’ [Muslim]? When I said I am ‘M-Class’, they refused to rent to us. But thenI said I was [a] Bohra, and they said ‘in that case, you are welcome.’”

14

Muslim advantages in trade.

In direct contrast to the robust complementarities visible at medieval trading ports were the

incentives present in towns that were the centres of Muslim political authority, where Hindus and

Muslims acted as substitutes for one another and competitors for patronage. Concurrent with

the spread of medieval trade, Muslim political control began to expand into India (Figure 1(b)).

With the conquest came control of patronage and land revenue systems, based upon control of the

surplus from India’s mainly agricultural wealth. These patronage systems were concentrated in

towns, many of which were established by the fiat of the Muslim rulers themselves (Raychoudhari,

1998). Mints were established to monetize this wealth.

In medieval Indian kingdoms, political and religious patronage played a very important role.

Though not necessarily members of the royal household themselves, the majority of the city’s

population was often tied by client relations to people who were.22 Once flourishing cities that lost

their roles as political centres rapidly became ghost towns.

Following the Muslim conquest, it is likely that Muslim clients, both converts and immigrants,

substituted for and competed with Hindu clients for patronage. Though “vertical” inter-ethnic

links existed between Hindu artisans and Muslim patrons, such ties were often in competition

to intra-Muslim patron-client relations. Though the Hindu and Muslim artisans that constituted

the majority of the populations of these cities lived side by side, there was limited incentive for

inter-ethnic exchange between these groups.23 Thus, despite the fact that, like medieval ports,

political centres provided historical incentives for conversion to Islam and enjoyed historical wealth,

patronage centres were the historical focus of inter-ethnic competition rather than inter-ethnic

complementarity. Thus patronage centres were unlikely to develop institutions to support religious

tolerance.

22Thus, when the Mughal Emperor Aurangzeb went South to campaign in the late 17th century, it was not unusualthat four of five of Delhi’s 400,000 residents left with him (Blake, 1991).

23The walled city of Delhi, Shahjahanabad, founded in 1638 as the capital of the Mughal empire, still appears toexhibit these features. In my fieldwork there, I found Muslims largely concentrated in separate mohallas near theJama Masjid, the Fatehpuri Masjid and Ballimaran. Distinct styles of clothing, the focus of community commonspace around the mosque and stalls that advertise using Arabic script (Urdu) provide overt indicators of the ethniccomposition of the mohalla. In Ahmadabad, gated pols separated ethnic groups within the medieval core. See alsoNoe (1993) on Delhi and Brass (2003)[p.150-160] on Aligarh.

15

3 Empirical strategy

As described above, the “robust” complementarities between Hindus and Muslims in India’s over-

seas ports were largely created by exogenous features, particularly the Hajj, that were inherent to

Islamic doctrine. Such complementarities made medieval trading ports conducive to inter-ethnic

exchange and favourable for further investment in institutional mechanisms that bolstered such

exchange.

My empirical approach uses towns that became medieval trading ports as an indicator of the

“treatment” of historic incentives for inter-ethnic trade.24 The ideal comparison would measure

the difference in Hindu-Muslim relations between a town that enjoyed such incentives and the same

town that did not. In the absence of such a counterfactual, I construct a series of control functions

that mimic such a counterfactual town under two sets of assumptions.

First, I assume that the selection of locations for medieval trade was uncorrelated with subse-

quent religious interaction. This assumption will be violated if medieval ports had different initial

conditions that might also have had an effect on religious violence, for example, through congenial

geography that provided increased opportunities for subsequent wealth. Thus, I add a rich set of

controls for initial conditions that might have impacted subsequent religious interactions. These in-

clude polynomial controls for longitude and latitude, propensity for natural disasters (which might

lower or enhance cooperation (Wade, 1988)), proximities to navigable rivers (which may raise town

wealth independently) as well as to the Ganges, which due to Hindu sacred geography is an ex-

ogenous driver of the proportion of upper castes. I also subset the data to only consider towns

proximate to the modern coast. Conditional on these factors, I can estimate the average treatment

effect of medieval trade on religious conflict in those towns that enjoyed medieval trade.

The experience of medieval ports can be compared not only to otherwise initially similar towns

that were not medieval ports, but also to other medieval towns which were historically rich but where

the theory suggests robust complementarities should not exist. These include geographically similar

towns on internal trade routes, where Hindus could locally replicate Muslim trade networks, and

towns that were centers of political patronage, where Hindus and Muslims were likely competitors.25

24Insofar as non-medieval port towns-e.g. Banaras- also developed inter-religious institutions over time, the mea-sured medieval port effect will under-estimate the effect of institutions in those ports on religious tolerance.

25A related approach would be to construct a propensity score for medieval trade based upon nearest neighbourmatching, both geographically and on medieval characteristics, and compare medieval ports to towns that had similar

16

Looking at effects over long periods of history raises a separate challenge that deviates from

a canonical experiment: even controlling for initial conditions, towns under study were subject to

different external political influences both during and after the treatment that might also influence

subsequent religious relations. Some component of these influences– e.g. the expansion of Muslim

or European political rule– might partly result from a desire to occupy regions with active trade.

To account for such political channels, I compare the effect of a medieval trading legacy both with

and without a rich set of controls for these political factors, including fixed effects for different

provinces and native states, as well as the timing of a district’s colonial annexation.26 As we shall

see, these controls do not greatly alter the measured treatment effects.

A second potential concern to the above approach is that Muslim traders may have chosen to

trade at geographically similar ports for unobservable reasons, such as having a local population

with a proclivity for peace independently of trade. This historically peaceful population might

continue to be inclined towards peace in modern times. To assess this hypothetical challenge to the

results, I relax the assumption that the selection of medieval ports was uncorrelated with subsequent

religious violence, and instead uses the presence of natural harbours on the historical coastline as

an instrument for medieval port location.

Given the severity of the monsoon winds, medieval ports– more so than their modern counterparts–

needed to be located in naturally-protected inlets. As the extract from the Portuguese cartographer,

Diego Homem’s 1558 portolan reveals (Figure 1a), these harbourages were typically located at small

inlets formed by indentations in the medieval coastline (Dasgupta, 2004). In fact, an overwhelming

share– 81%– of medieval ports in the sample were located at locations that possessed likely medieval

natural harbours (Table 2). Insofar, as seems plausible, that towns with an indentation in their

medieval coastline were not any more likely than otherwise geographically similar towns to have

attracted a more peaceful pre-existing population, this comparison allows us to assess the degree

to which medieval traders’ selection of locations in which to trade may have biased the results. I

provide evidence that suggests the medieval era coastal indentations do seem to have little effect

other than to drive medieval port location, including demonstrating that these indentations are

uncorrelated with the location of colonial era ports. I then exploit this exogenous determinant of

propensities to become such ports. This approach leads to similar results.26The timing of annexation had a number of effects, including different land tenure systems (Banerjee and Iyer,

2005).

17

medieval port location directly, providing instrumental variables estimates that compare modern

religious relations in towns that became medieval ports because of their historical natural har-

bours and geographically similar towns that would have become medieval ports had they had such

harbours.

Another natural process allows a further robustness check: the coast itself has moved over

time. The massive flow of water from the hills during India’s monsoon rains regularly pushes large

amounts of silt to the mouths of rivers and inlets. Over time, silting has meant that towns that were

at harbours in the medieval period have become increasingly inaccessible to shipping (Arasaratnam,

1994).27 By providing natural variation in the viability of trade in towns over time, the silting

process allows us to assess whether it is continued colonial era overseas trade and shipping in

medieval ports or other mechanisms that explain the legacy of medieval trade.

4 Data

The dataset on pre-Independence Hindu-Muslim violence, drawn from newspaper reports and offi-

cial sources, is largely based upon that compiled by Wilkinson (2005b). In this dataset, a religious

riot was defined as a violent confrontation by two communally-identified groups. Data on historical

trade in India’s ports came from a number of sources. The Periplus Maris Erythraei (Casson, ed,

1989) provided the locations of a number of pre-Muslim and early Muslim ports. The records of con-

temporary Muslim, Christian and Chinese observers, including Chau Jukua (1225), Ibn Ibn Battuta

(1355), Ludovico di Verthema (1503), Duarte Barbosa (1519) and Zayn al-Din al Malibari (1528)

provide supplementary evidence. These contemporary accounts were augmented by secondary

sources (Yule, ed, 1866, Subrahmanyam, 1990, Chaudhuri, 1995, Chakravarti, 2000). Finally, every

town listed in the Imperial gazetteers of India from 1907 was examined for evidence of contemporary

and medieval trade.

I identified a town as a medieval trading port if it exhibited substantive evidence of direct

overseas trade, prior to the 18th century and independent of European involvement. This definition

eliminates most river ports and those ports either founded by Europeans (including the Presidency

27Even contemporary dredging techniques (which of course were unavailable during the period of Muslim tradingdominance) are unable to contend with the volumes of silt generated. Even the modern port of Calcutta is no longeraccessible to most shipping, which has been diverted to the downriver town of Haldia.

18

towns of Bombay, Calcutta and Madras) or those that became overseas trading ports as a result

of European establishments (e.g. Hughli, Tranquebar). I found a total of 68 confirmed medieval

trading ports in undivided India, of which 59 were distinct towns in 1901.

These medieval ports were linked to their geographical location using ArcGIS. To categorize me-

dieval era “natural harbours,” I used the US Geological Survey Digital Atlas of South Asia 2001 to

identify water bodies that were within 10km of the modern Indian coastline, including non-perennial

ponds and streams and those without an outlet to the sea. If these water bodies intersected the

coast in the medieval period, they would have produced minor inlets, or sheltered harbours. I

define towns within 10km of those water bodies as having had access to a “medieval natural har-

bour.” This approach identifies major irregularities and inlets that are likely to have existed in the

medieval period.28

I constructed a GIS of trade routes, mints, crafts and towns in medieval India using the Mughal

census commissioned by Emperor Akbar, the Ain-i-Akbari and supplemental medieval sources

documented in Habib (1982) (see Figure 1b). I matched these towns to districts in colonial India

by manually tracing and then georeferencing pre-Independence district maps. This GIS was used

to link towns across district and state changes throughout history and across periods of British and

Muslim rule. I was able to match all towns to their British district and Native state.

In addition, I collected data from the Imperial gazetteers on a number of different natural dis-

asters from 1850 to 1900, including droughts, earthquakes, locust infestations, floods and cyclones.

The cross-district patterns in propensity to face these natural risks are likely to have persisted up

until India’s dramatic population gains beginning in the twentieth century. The Imperial gazetteers

provided information on the revenue, religious composition and political histories of each district

and many of India’s towns. The decennial censuses and Imperial gazetteers yielded data, mainly

at the district level, but also for larger towns, on religious demography, municipal income as well

as political histories. Municipal income per capita provides a town-specific measure of the average

wealth of the town–this measure was based mainly upon a tax on internal commerce (octroi) and

a poll tax with minimum wealth requirements.

28A placebo test of proximity to water bodies that are not near the coast reveals no effect on violence.

19

Table 2: Summary Statistics

Riots, 1850-1950 Obs Mean SD Obs Mean SD Obs Mean SD# of Hindu-Muslim Riots 476 1.116 3.416 53 0.925 5.487 59 0.136 0.472Any H-M Riot 476 0.418 0.494 53 0.170 0.379 59 0.102 0.305# Killed in H-M Riots 476 23.277 242.361 53 88.906 639.995 59 0.136 0.571Total Days of H-M Riots 476 1.630 11.301 53 3.000 20.598 59 0.051 0.289

Initial ConditionsMedieval Natural Harbour 476 0.132 0.339 53 1.000 0.000 59 0.814 0.393<10km from Modern Coast 476 0.111 0.315 53 0.528 0.504 59 0.898 0.305Log. Dist. Modern Coast 476 11.998 2.031 53 8.946 1.739 59 7.358 1.709Log. Dist. Navigable River 476 12.755 1.641 53 12.684 2.358 59 13.364 1.704Natural Disasters, 1850-1900 476 1.536 2.512 53 0.906 2.133 58 1.603 3.201Log. Dist. R. Ganges 476 11.846 1.960 53 12.823 2.091 59 13.292 1.537

Medieval Era CharacteristicsMedieval Town 476 0.592 0.492 53 0.566 0.500 59 1.000 0.000Mughal Mint 476 0.084 0.278 53 0.000 0.000 59 0.051 0.222Other Muslim Patronage Ctr 476 0.132 0.339 53 0.057 0.233 59 0.220 0.418Mughal Internal Trade Route 476 0.153 0.361 53 0.038 0.192 59 0.051 0.222Mughal Skilled Crafts in Town 476 0.048 0.215 53 0.075 0.267 59 0.169 0.378Major Shi'a Dynasty 476 0.200 0.400 53 0.208 0.409 59 0.186 0.393Centuries Muslim Rule 476 4.073 2.286 53 2.727 2.691 59 2.097 2.126

Colonial Era Outcomes and Covariates% Muslims 1901 244 29.879 17.732 20 18.596 14.884 22 32.449 22.101Mun. Income per Capita 316 1.805 3.092 28 2.155 2.6382 28 1.580 1.103Colonial Overseas Port (1907) 476 0.038 0.191 53 0.170 0.379 59 0.356 0.483Log. Population 1901 476 9.672 1.129 53 9.420 1.209 59 9.170 1.315

Natural Harbours, Not Medieval Ports

Towns, Not Medieval Ports

Medieval Ports

20

5 Results

Table 2 presents summary statistics comparing medieval ports to other towns that existed in 1901,

and to the subset of towns at natural harbours that did not become medieval ports. Medieval

ports exhibit strikingly lower incidences of religious violence compared to both these classes of

other towns, as well as to towns geographically close by (Figure 2). Medieval ports experienced

around five times fewer riots on average. The proportion of medieval ports experiencing at least one

outbreak of religious violence between 1850 and 1950 was around 10%; close to 40% of other towns

faced a riot. The intensity of the riots also appears to be lower: on average, five medieval ports

together experienced a single death due to religious violence, but in other towns, religious violence

claimed an average of nearly 23 lives per town. Natural harbours that did not become medieval

ports also have a greater incidence of religious riots and more intense violence than medieval ports.

However, as indicated by municipal income, medieval ports were on average poorer than both

harbour towns and other towns. Medieval ports also have a more mixed religious population (see

also Figure 2). These indicators are commonly associated with higher rather than lower incidences

of ethnic violence.

Table 2 also shows how medieval port towns compare along a range of geographical and medieval

characteristics. An overwhelming share, 81% of medieval ports, were located at natural harbours.

Medieval port towns are also more likely to be near the modern coast. Yet, apart from these

distinctions, medieval port towns actually appear similar in their geographical and medieval-era

characteristics to other towns.

The similarities between medieval ports and other towns are confirmed in Table 3, which shows

the determinants of medieval port location. Along a range of initial geographical conditions (Col

1) medieval-era measures of trade and human capital (Col 2), restricting the data to towns within

200 and 100 km of the modern coast, and adding colonial era fixed effects for native state, province

and timing of annexation (Col 5-7), there are only two key robust determinants of medieval port

location: coastal towns and medieval era natural harbours. This does not change if we do not control

for natural harbours (Cols 8-9). In contrast, colonial-era port location seems to be unrelated to the

presence of medieval era natural harbours (Col 10-11). The F-tests of the strength of the natural

harbour relation with medieval ports are sufficiently strong to avoid weak instrument pathologies

21

Tab

le3:

Regre

ssio

n:

Dete

rmin

ants

of

Med

ieval

an

dC

olo

nia

lO

vers

eas

Port

Locati

on

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

Out

com

eTo

wn

at M

edie

val N

atur

al H

arbo

ur0.

187*

**0.

176*

**0.

180*

**0.

210*

*0.

263*

**0.

299*

**0.

368*

**-0

.040

-0.0

33[0

.054

][0

.051

][0

.060

][0

.072

][0

.064

][0

.059

][0

.086

][0

.091

][0

.116

]To

wn

<10k

m fr

om M

oder

n C

oast

0.45

7***

0.43

4***

0.42

7***

0.39

9***

0.45

3***

0.41

4***

0.35

6**

0.52

4***

0.50

1***

0.56

4***

0.59

3***

[0.0

83]

[0.0

93]

[0.0

87]

[0.0

94]

[0.1

14]

[0.1

11]

[0.1

26]

[0.0

91]

[0.1

06]

[0.1

34]

[0.1

53]

Log.

Dis

tanc

e to

Mod

ern

Coa

st-0

.014

-0.0

16-0

.010

-0.0

08-0

.012

-0.0

08-0

.004

-0.0

24-0

.030

-0.0

05-0

.002

[0.0

15]

[0.0

14]

[0.0

38]

[0.0

45]

[0.0

29]

[0.0

33]

[0.0

38]

[0.0

15]

[0.0

28]

[0.0

21]

[0.0

31]

Log.

Dis

tanc

e to

Nav

igab

le R

iver

0.01

30.

008

0.01

40.

173*

**-0

.004

-0.0

600.

253*

**0.

004

-0.0

18-0

.002

0.01

6[0

.008

][0

.008

][0

.046

][0

.052

][0

.023

][0

.089

][0

.072

][0

.007

][0

.021

][0

.009

][0

.013

]N

atur

al D

isas

ters

, 185

0-19

000.

014

0.01

20.

034

0.03

70.

011

0.02

60.

028

0.01

40.

014

-0.0

02-0

.002

[0.0

12]

[0.0

12]

[0.0

26]

[0.0

30]

[0.0

17]

[0.0

41]

[0.0

45]

[0.0

12]

[0.0

18]

[0.0

02]

[0.0

05]

Log.

Dis

tanc

e to

Gan

ges

-0.0

01-0

.001

-0.0

37-0

.196

***

-0.0

010.

083

-0.2

48**

*-0

.006

-0.0

05-0

.005

-0.0

10[0

.006

][0

.008

][0

.031

][0

.038

][0

.005

][0

.079

][0

.080

][0

.007

][0

.009

][0

.007

][0

.007

]M

edie

val T

own

0.05

10.

118

0.19

7*0.

024

0.11

10.

203*

0.03

40.

020

[0.0

33]

[0.0

70]

[0.0

92]

[0.0

31]

[0.0

97]

[0.1

11]

[0.0

29]

[0.0

23]

Mug

hal M

int i

n To

wn

-0.0

39-0

.046

-0.0

29-0

.026

-0.0

040.

069

-0.0

62-0

.046

[0.0

33]

[0.0

73]

[0.0

97]

[0.0

35]

[0.0

89]

[0.1

14]

[0.0

48]

[0.0

30]

Oth

er M

edie

val P

atro

nage

Cen

ter

0.02

60.

053

0.02

90.

028

-0.0

06-0

.055

0.03

90.

067*

*[0

.035

][0

.074

][0

.069

][0

.053

][0

.114

][0

.117

][0

.052

][0

.030

]M

edie

val I

nlan

d Tr

ade

Rou

te-0

.010

-0.0

75-0

.138

-0.0

01-0

.046

-0.1

37-0

.028

0.00

4[0

.036

][0

.093

][0

.115

][0

.042

][0

.119

][0

.149

][0

.037

][0

.025

]M

edie

val S

kille

d C

rafts

in T

own

0.04

50.

058

0.04

60.

020

-0.0

37-0

.091

0.07

2*-0

.104

*[0

.035

][0

.105

][0

.128

][0

.039

][0

.059

][0

.078

][0

.040

][0

.054

]M

ajor

Shi

'a st

ate

befo

re 1

857

0.00

70.

018

0.00

3-0

.026

0.11

60.

111

-0.0

100.

044

[0.0

38]

[0.0

94]

[0.1

06]

[0.0

59]

[0.0

96]

[0.0

91]

[0.0

94]

[0.0

89]

Cen

turie

s Mus

lim R

ule

-0.0

14-0

.050

-0.0

71**

*-0

.028

-0.0

26-0

.101

-0.0

250.

019

[0.0

15]

[0.0

31]

[0.0

21]

[0.0

39]

[0.1

07]

[0.0

75]

[0.0

38]

[0.0

23]

F-te

st (N

atur

al H

arbo

ur)

12.1

712

.03

8.97

8.61

17.1

25.4

918

.32

0.19

0.08

Prob

>F0.

000.

000.

010.

010.

000.

000.

000.

670.

78Sa

mpl

eFu

llFu

llC

oast

al,

<200

kmC

oast

al,

<100

kmFu

llC

oast

al,

<200

kmC

oast

al,

<100

kmFu

llFu

llFu

llFu

ll

Prov

ince

/ N

S x

Ann

exat

ion

FEN

oN

oN

oN

oY

esY

esY

esN

oY

esN

oY

esO

bser

vatio

ns24

824

811

089

248

110

8924

824

824

824

8R

-squ

ared

0.62

0.63

0.63

0.64

0.68

0.69

0.71

0.59

0.64

0.55

0.63

__O

vers

eas P

orts

190

7_

M

edie

val O

vers

eas P

orts

(7th

- 17t

h C

entu

ry)

Not

es:

All

regr

essi

ons

incl

ud

equ

adra

tic

pol

yn

om

ials

inL

on

git

ud

ean

dL

ati

tud

e.R

ob

ust

stan

dard

erro

rs(c

lust

ered

at

Nati

veSta

tex

An

nex

ati

on

leve

l):

*si

gnifi

cant

at10

%;

**5%

;**

*1%

22

in most specifications.

The average effect of medieval trade on the number of Hindu-Muslim riots faced by medieval

ports between 1850 and 1950, is assessed in Columns 1-5 of Table 4, which present incidence ratios

from negative binomial regressions appropriate for count data.29 Towns with medieval trade legacies

experienced around five times fewer religious riots than similar towns between 1850-1950. This

result is robust and remarkably consistent even after controlling for initial conditions that might

shape port selection (Col 1), and medieval factors influencing trade and Hindu-Muslim relations

(Col 2). The effect actually strengthens when we add colonial era provincial and annexation fixed

effects (Col 3), and select finer coastal samples (Cols 4-5). OLS also provides consistent results

(Cols 6-7).

It may be that outlying towns that were highly riot-prone are driving these results. Columns

8-11 addresses this by instead examining the probability that a town experienced any religious riots

between 1850-1950. The effect is again remarkably consistent effect across specifications– medieval

ports are around 25 percentage points less likely to experience a religious riot.

I can evaluate a number of alternative explanations. Otherwise similar coastal towns do not

appear more ‘cosmopolitan’, and the propensity for natural disasters also does not appear affect

ethnic tolerance by fostering risk-mitigation institutions (Wade, 1988). It may be that medieval

ports that are still towns in 1901 are drawn from successful “survivors” that experienced less

religious violence. However controlling for whether a town was mentioned in the Ain-i-Akbari

or other medieval sources does not affect the results, and towns with (often crowded and poor)

medieval precincts that were not medieval ports actually appear somewhat more prone to violence.

Other useful medieval comparison groups include towns where the Mughals established Mints

to monetize wealth– as the theoretical framework suggests, these towns despite being historically

wealthy, arguably provided incentives for inter-ethnic competition between Hindus and Muslims

rather than complementarity. Indeed, Mint towns were 1.5 to two times more riot-prone. Likewise,

towns on inland trade routes, where Hindus could locally replicate Muslim networks via relays, also

show an increased probabilities of subsequent ethnic violence. Thus, rather than historical trade

29This table uses the sub-sample of 248 towns for which complete data exist. Using the full sample for each setof covariates yields similar results, but suffers from the problem of under-reporting of rioting in non-descript towns,including many medieval ports. Towns that are well-documented by official sources also tend to be those where thereligious rioting is well-documented.

23

Tab

le4:

Regre

ssio

n:

Hin

du

-Mu

slim

Rio

ts,

1850-1

950

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

Neg

. Bin

. (I

. Rat

ios)

Neg

. Bin

. (I

. Rat

ios)

Neg

. Bin

. (I

. Rat

ios)

Neg

. Bin

. (I

. Rat

ios)

Neg

. Bin

. (I

. Rat

ios)

OLS

OLS

Prob

it,

dF/d

XPr

obit,

dF

/dX

Prob

it,

dF/d

XO

LS

Med

ieva

lOve

rsea

sPor

t0

209*

**0

291*

**0

074*

**0

019*

**0

013*

*-1

330*

-117

6*-0

237*

**-0

223*

**-0

194*

**-0

266*

**

A

ny H

-M R

iot,

1850

-195

0?

# o

f Hin

du-M

uslim

Rio

ts in

Tow

n, 1

850-

1950

Med

ieva

l Ove

rsea

s Por

t0.

209*

**0.

291*

**0.

074*

**0.

019*

**0.

013*

*-1

.330

*-1

.176

*-0

.237

***

-0.2

23**

*-0

.194

***

-0.2

66**

*[0

.101

][0

.125

][0

.050

][0

.023

][0

.028

][0

.679

][0

.555

][0

.071

][0

.060

][0

.046

][0

.071

]To

wn

>10k

m fr

om M

oder

n C

oast

6.76

8*3.

670

3.57

13.

705

5.22

80.

589

0.29

50.

467*

**0.

456*

*0.

116

0.27

8*[6

.999

][3

.313

][4

.384

][3

.270

][5

.428

][0

.701

][0

.424

][0

.159

][0

.184

][0

.132

][0

.143

]Lo

g. D

ista

nce

to M

oder

n C

oast

1.16

71.

147

0.94

80.

937

0.95

40.

005

-0.0

280.

042

0.03

9-0

.002

0.00

6[0

.256

][0

.246

][0

.278

][0

.151

][0

.373

][0

.153

][0

.185

][0

.048

][0

.048

][0

.041

][0

.038

]Lo

g. D

ista

nce

to N

avig

able

Riv

er1.

200*

**1.

272*

**1.

409*

**8.

736*

1.01

70.

523

-0.1

180.

028

0.06

20.

133

0.10

8***

[0.0

83]

[0.0

67]

[0.1

56]

[11.

115]

[1.0

05]

[0.3

67]

[0.1

44]

[0.0

34]

[0.0

39]

[0.1

17]

[0.0

31]

Nat

ural

dis

aste

rs, 1

850-

1900

1.09

31.

054

1.04

30.

778

0.75

90.

037

0.03

60.

024*

0.02

40.

008

0.02

9[0

.066

][0

.043

][0

.073

][0

.173

][0

.227

][0

.069

][0

.026

][0

.014

][0

.015

][0

.017

][0

.023

]Lo

g. D

ista

nce

to G

ange

s0.

838*

**0.

891

0.95

20.

220

1.81

1-0

.393

0.70

8***

-0.0

54*

-0.0

44-0

.095

-0.0

17[0

.050

][0

.065

][0

.086

][0

.231

][1

.735

][0

.358

][0

.210

][0

.032

][0

.035

][0

.150

][0

.020

][0

.050

][0

.065

][0

.086

][0

.231

][1

.735

][0

.358

][0

.210

][0

.032

][0

.035

][0

.150

][0

.020

]M

edie

val T

own

1.19

61.

414*

4.11

3*1.

685

0.21

10.

128

-0.0

24-0

.049

-0.0

61[0

.270

][0

.260

][3

.162

][1

.780

][0

.239

][0

.297

][0

.067

][0

.100

][0

.058

]M

ugha

l Min

t in

Tow

n2.

167*

**1.

553

1.62

81.

561

0.67

81.

726*

*0.

138

0.19

50.

038

[0.6

10]

[0.4

17]

[0.9

41]

[1.7

17]

[0.7

15]

[0.5

76]

[0.1

33]

[0.2

42]

[0.1

08]

Oth

er M

edie

val P

atro

nage

Cen

ter

0.84

61.

254

1.08

92.

380

0.65

90.

063

0.12

80.

141

0.15

6[0

.244

][0

.434

][0

.186

][1

.928

][0

.647

][0

.185

][0

.114

][0

.139

][0

.112

]M

edie

val I

nlan

d Tr

ade

Rou

te1.

420*

1.19

80.

837

1.66

50.

400

1.33

9**

0.21

4***

0.28

70.

148*

*[0

.283

][0

.221

][0

.658

][1

.706

][0

.419

][0

.532

][0

.072

][0

.182

][0

.056

]M

edie

val S

kille

d C

rafts

in T

own

1.89

1*2.

322*

*5.

934*

**6.

164*

1.61

31.

134

-0.0

440.

246

0.10

4[0

.616

][0

.812

][3

.214

][6

.710

][1

.002

][0

.916

][0

.125

][0

.307

][0

.111

]M

ajor

Shi'a

stat

ebe

fore

1857

1.54

00.

391

0.46

30.

165

-1.8

43**

-1.0

22*

-0.1

63-0

.176

***

-0.3

20**

Maj

or S

hia

stat

e be

fore

185

71.

540

0.39

10.

463

0.16

5-1

.843

-1.0

22-0

.163

-0.1

76-0

.320

[0.6

67]

[0.4

32]

[0.6

88]

[0.2

39]

[0.8

74]

[0.5

09]

[0.1

12]

[0.0

52]

[0.1

40]

Cen

turie

s Mus

lim R

ule

1.19

61.

023

0.50

61.

819

-0.0

56-0

.248

0.06

1*0.

106

0.02

8[0

.130

][0

.130

][0

.236

][1

.063

][0

.237

][0

.409

][0

.034

][0

.081

][0

.058

]Sm

ith-B

lund

ell E

xoge

neity

Tes

t: c2 (1

)0.

012

0.05

30.

216

Prob

> c2 (1

)0.

914

0.81

80.

642

Prob

c(1

)Sa

mpl

eFu

llFu

llFu

llC

oast

al,

<200

kmC

oast

al,

<100

kmFu

llC

oast

al,

<100

kmFu

llFu

llC

oast

al,

<100

kmFu

ll

Prov

ince

/ N

S x

Ann

exat

ion

FEN

oN

oY

esY

esY

esY

esY

esN

oN

oN

oY

esO

bser

vatio

ns24

824

824

811

089

248

8924

824

889

248

R-S

quar

ed0.

440.

590.

42

Not

es:

All

regr

essi

ons

incl

ud

equ

adra

tic

pol

yn

om

ials

inL

on

git

ud

ean

dL

ati

tud

e.R

ob

ust

stan

dard

erro

rs(c

lust

ered

at

Nati

veSta

tex

An

nex

ati

on

leve

l):

*si

gnifi

cant

at10

%;

**5%

;**

*1%

24

per se, it appears that it is the exogenous and non-replicable inter-ethnic complementarities present

in medieval overseas ports that have lasting effects on ethnic tolerance.

It has also been argued by Bayly (1985) and Prior (1993) that pre-Independence religious

violence tended to occur when major (Shia) Muslim festival processions–Urs and Muharram–tended

to coincide with Hindu festivals. Since both religions follow the lunar calendar, these processions

coincide every three decades. According to Bayly (1985), such religious festivals were used as a

display of wealth and power by an emergent Hindu middle class in the wake of the decline of Muslim

political power. If it is the case that the coincidence of timing of processions played an important

role in pre-Independence riots, then it is likely that riots should occur more often in areas with

long-term Shiite traditions. Table 4 also examines whether riots are more likely in regions that

experienced long periods of Shiite rule and thus where there was a greater tendency for Shiite

conversion and immigration. However, the medieval port effect is robust to matching by their Shia

histories, and towns with long histories of Shia rule actually appear to less riot-prone and less likely

to experience any religious rioting.

An important debate exists between whether ‘institutions’– considered narrowly to be for-

mal rules or executive constraints– or human capital are responsible for beneficial social out-

comes (Glaeser, La Porta, Lopez-de Silanes and Shleifer, 2004, Rodrik, Subramanian and Trebbi,

2004). Tables 4 also assess whether historical human capital accumulation– as measured by the

medieval presence of skilled craftsmen– subsequently reduces ethnic violence. Not only is the effect

of a medieval legacy robust to controlling for the presence of such skills, it appears that skilled

towns are more riot-prone.

Yet, as the interaction terms in Table 5 suggests, the legacy of medieval skills differs dramatically

depending on whether they were accumulated in medieval port towns– where skills accumulation by

Hindus and Muslims were more likely to be institutionally organized to maintain complementarity–

and in other towns. In fact, medieval port towns that accumulated skills experienced between two

and four fewer religious riots than otherwise similar towns, in contrast to non-port towns which

experienced two additional riots.

These effects are consistent with a more general interpretation– that the returns to human

capital rise in environments where there is greater cooperation, just as investments that coordinate

and support greater cooperation are cheaper when there is greater human capital available. In

25

Table 5: Regression: Complementarities and Human Capital

OLS (1) (2) (3) (4) (5) (6)(A) # of Hindu- Muslim Riots in Town (1850-1950)Medieval Overseas Port -0.532 -0.543 -0.307 -0.902* -0.899** -0.532*

[0.454] [0.436] [0.338] [0.529] [0.356] [0.262]Medieval Skilled Crafts in Town 2.015* 2.360* 1.906* 1.927* 2.380* 2.055*

[1.165] [1.296] [1.070] [1.108] [1.248] [1.107]Medieval Port x Skilled Town -2.734*** -4.004*** -4.181*** -2.201* -3.678** -4.020**

[0.978] [1.240] [1.128] [1.263] [1.492] [1.436]R-squared 0.270 0.510 0.640 0.440 0.600 0.710

(B) Probability of Any Hindu-Muslim Riot in Town (1850-1950)Medieval Overseas Port -0.206** -0.244** -0.180** -0.250*** -0.237** -0.126

[0.077] [0.084] [0.076] [0.084] [0.107] [0.115]Medieval Skilled Crafts in Town -0.017 0.216 0.213 0.115 0.306* 0.345

[0.113] [0.160] [0.211] [0.127] [0.169] [0.209]Medieval Port x Skilled Town 0.002 -0.158 -0.150 -0.081 -0.173 -0.217

[0.269] [0.244] [0.254] [0.292] [0.275] [0.237]

SampleFull Coastal,

<200kmCoastal, <100km

Full Coastal, <200km

Coastal, <100km

Province/NS x Annexation FE No No No Yes Yes YesObservations 248 110 89 248 110 89R-squared 0.27 0.33 0.40 0.42 0.45 0.58

Notes: All regressions include quadratic polynomials in Longitude and Latitude and Log. Distances fromthe Modern Coast, Navigable Rivers and the Ganges, Coastal Town and Natural Disasters. Medieval-eracontrols add: Medieval Town, Mughal Mint, Other Patronage Center, Inland Trade Route, Skilled Crafts inTown, Major Shi’a State, Centuries Muslim Rule. All interactions are demeaned. Robust standard errors(clustered at Native State x Annexation level): * significant at 10%; ** 5%; *** 1%

26

other words, rather than being substitute explanations for economic development, as sometimes

mooted in the literature, at the very least in the context of peaceful coexistence in South Asian

towns, the evidence suggests that human capital and institutions were complements.

If institutional mechanisms that foster inter-ethnic complementarity, cooperation and trust

persist from the medieval period, they should also impact the types of economic activity in which

individuals engage. However, as the case evidence outlined in Table 1 suggests, while sharing

an economic logic, there is heterogeneity in how such institutions have supported cooperation in

different medieval ports over time. Table 6 examines the long-term effects of medieval trade legacy

in both skilled and unskilled towns on the composition of modern employment in 1991. Notice

that while there is some weak evidence that the initial economic rationale for ports and trade

routes– as locales for transport and communications– continues to influence employment patterns

in colonial era ports and medieval internal trade routes, medieval ports do not appear to have

greater proportions of their labour force in this sector (Cols 1-3). Also in contrast to colonial ports

and medieval inland trade routes, residents of medieval ports do to be employ significantly more

in (mainly domestic) trade and small commerce, a sector which often relies on the presence of

greater trust (Cols 4-6). This contrast once again appears to confirm that historic non-replicable

inter-ethnic complementarities, rather than other forms of historic trade, appear to have lasting

effects on trust.30

A further intriguing aspect of Table 6 is that historical human capital accumulation- as mea-

sured by skilled crafts– does not, by itself, show much persistence in fostering either trade or

manufacturing. However historic skills appear to influence the coordination of employment activity

in medieval port towns, favouring non-household manufacturing over trade. As the case evidence

in Table 1 suggests, in some medieval ports, non-household manufacturing has been organized and

coordinated to maintain inter-ethnic complementarity in new areas. Because of the potential for

one group to ‘hold up’ the other when both groups invest in complements, such arrangements, once

again, are consistent with inter-ethnic trust.

30The latter resonates with Nunn and Wantchekon (forthcoming)’s finding that trade can actually have negativeeffects on contemporary trust when it is coercive, as in Africa’s slave trade.

27

Tab

le6:

Eff

ects

on

Mod

ern

Em

plo

ym

ent

Com

posi

tion

,1991

Indi

vidu

als p

er 1

000

(199

1) in

:(1

)(2

)(3

)(4

)(5

)(6

)(7

)(8

)(9

)

(A) M

edie

val P

orts

(OL

S)M

edie

val O

vers

eas P

ort

0.05

60.

029

0.01

70.

287*

**0.

188*

**0.

353*

**0.

161*

0.15

2*0.

093

[0.0

47]

[0.0

30]

[0.0

60]

[0.1

03]

[0.0

62]

[0.0

94]

[0.0

90]

[0.0

88]

[0.1

39]

Med

ieva

l Por

t x S

kille

d To

wn

-0.0

320.

001

0.02

9-0

.161

*-0

.144

-0.5

03**

*0.

076

0.03

70.

155

[0.0

67]

[0.0

44]

[0.0

59]

[0.0

87]

[0.1

19]

[0.1

18]

[0.0

73]

[0.0

71]

[0.2

17]

Med

ieva

l Tow

n-0

.006

-0.0

09-0

.053

-0.0

69-0

.021

-0.1

73-0

.053

-0.0

520.

036

[0.0

08]

[0.0

10]

[0.0

34]

[0.0

67]

[0.0

52]

[0.1

05]

[0.0

52]

[0.0

54]

[0.1

19]

Mug

hal M

int i

n To

wn

-0.0

14-0

.020

*-0

.057

-0.0

58-0

.074

0.18

5-0

.067

*-0

.092

**-0

.197

***

[0.0

11]

[0.0

12]

[0.0

35]

[0.0

53]

[0.0

60]

[0.1

19]

[0.0

38]

[0.0

43]

[0.0

57]

Oth

er M

edie

val P

atro

nage

Cen

ter

0.00

60.

014

0.06

7***

-0.0

77*

-0.0

85*

-0.2

190.

017

0.02

6-0

.003

[0.0

10]

[0.0

09]

[0.0

20]

[0.0

42]

[0.0

45]

[0.1

43]

[0.0

34]

[0.0

51]

[0.1

01]

Med

ieva

l Inl

and

Trad

e R

oute

-0.0

01-0

.002

0.09

7**

-0.1

01**

-0.1

15*

-0.1

90-0

.085

*-0

.109

*-0

.343

*[0

.014

][0

.017

][0

.036

][0

.047

][0

.063

][0

.175

][0

.043

][0

.056

][0

.176

]M

edie

val S

kille

d C

rafts

in T

own

-0.0

010.

003

-0.0

07-0

.011

-0.0

15-0

.008

0.00

80.

041

0.04

9[0

.008

][0

.007

][0

.017

][0

.041

][0

.045

][0

.162

][0

.034

][0

.045

][0

.096

]R

-squ

ared

0.15

0.36

0.62

0.17

0.38

0.61

0.17

0.38

0.56

(B) C

olon

ial P

orts

(Pla

cebo

- OL

S)Po

rt w

ith fo

reig

n tra

de, 1

907

0.06

20.

047*

0.03

60.

074

-0.0

100.

088

-0.0

82-0

.047

0.02

2[0

.041

][0

.025

][0

.023

][0

.062

][0

.045

][0

.067

][0

.148

][0

.138

][0

.095

]R

-squ

ared

0.16

0.37

0.62

0.16

0.37

0.58

0.15

0.37

0.54

Sam

ple

Indi

aIn

dia

Indi

a,

Coa

stal

(<

100k

m)

Indi

aIn

dia

Indi

a,

Coa

stal

(<

100k

m)

Indi

aIn

dia

Indi

a,

Coa

stal

(<

100k

m)

Con

trols

Med

ieva

lM

edie

val

Med

ieva

lM

edie

val

Med

ieva

lM

edie

val

Med

ieva

lM

edie

val

Med

ieva

lPr

ovin

ce/ N

S x

Ann

ex F

EN

oY

esY

esN

oY

esY

esN

oY

esY

esO

bser

vatio

ns21

621

676

216

216

7621

621

676

Trad

e an

d C

omm

erce

Non

-Hou

seho

ld M

anuf

actu

ring

Tran

spor

t and

Com

mun

icat

ions

Not

es:

All

regr

essi

ons

incl

ud

equ

adra

tic

pol

yn

om

ials

inL

on

git

ud

ean

dL

ati

tude

an

dL

og.

Dis

tan

ces

from

the

Mod

ern

Coast

,N

avig

ab

leR

iver

san

dth

eG

ange

s,C

oast

alT

own

and

Nat

ura

lD

isas

ters

.M

edie

val-

era

contr

ols

ad

d:

Med

ieva

lT

own

,M

ugh

al

Min

t,O

ther

Patr

on

age

Cen

ter,

Inla

nd

Tra

de

Rou

te,

Skille

dC

raft

sin

Tow

n,

Ma

jor

Sh

i’a

Sta

te,

Cen

turi

esM

usl

imR

ule

.R

ob

ust

stan

dard

erro

rs(c

lust

ered

at

Nati

ve

Sta

tex

An

nex

ati

on

level

):*

sign

ifica

nt

at10

%;

**5%

;**

*1%

28

5.1 Placebos and Robustness

So far the location of medieval ports has been assumed to be exogenously determined. I now test

and relax this assumption. First, it is possible to use the historical natural harbour instrument

to test for this exogeneity assumption under the probit assumptions (Smith and Blundell, 1986) .

Although natural harbours are a strong determinant of medieval port location (Table 3), I fail to

reject the null hypothesis of exogeneity with close to 90% confidence (Table 4). In other words,

medieval ports that were chosen for trade due to their natural harbour locations do not appear

significantly different from other medieval ports in unobserved ways that might be relevant for

modern religious conflict. Thus it seems reasonable to interpret the coefficient of medieval port

as the average treatment effect of medieval trade on religious violence in those towns that enjoyed

medieval trade.

Even though the exogeneity tests suggest that selection does not appear important in this

context, for completeness, Table 7 subjects the main results to a battery of placebo checks as well

as providing local average treatment comparisons. As I have argued, inter-ethnic complementarities

between Hindus and Muslims in overseas trade were largely disrupted during the colonial period.

If my interpretation is true, the effect colonial era port status on ethnic tolerance should be weaker

than the effect of medieval ports. If instead, the effect of medieval ports is coming from some

unobserved characteristic of port towns that make them more tolerant, such as simply increased

wealth or human capital accumulation, we should expect colonial-era overseas ports to have lower

colonial-era violence. As Panel A suggests, however, status as a colonial era port seems to have no

relationship with the propensity for colonial-era violence.

It still might be the case that medieval ports enjoy unobserved differences in wealth stemming

from their location at harbours that are not captured by colonial port records, such as profits from

smuggling. Panel B compares the sample of medieval ports that subsequently silted up and became

inaccessible to overseas shipping. Though there are only thirteen silted medieval ports, the effect of

silted ports on the effect on probability of religious violence is remarkably consistent in magnitude

with that of all medieval ports. Thus, the transmission mechanism that links a medieval trading

legacy to contemporary religious relations does not appear to be through unobserved modern trade.

Panel C presents instrumental variable results, comparing towns that became medieval ports

29

Table 7: Regression: Placebos and Robustness

(1) (2) (3) (4) (5) (6) (7)A) Coefficient on Overseas Port in 1907 (OLS)# H-M Riots, 1850-1950 -0.346 0.326 0.085 0.201 0.306 0.591 0.690

[0.542] [0.549] [0.616] [0.573] [0.919] [0.807] [0.770]R-squared 0.16 0.26 0.36 0.48 0.43 0.46 0.57

Any H-M Riot, 1850-1950 0.066 0.084 0.034 0.057 0.148 0.082 0.095[0.133] [0.117] [0.119] [0.116] [0.142] [0.120] [0.101]

R-squared 0.21 0.26 0.30 0.39 0.41 0.43 0.57B) Coefficient on Medieval Port Silted by 1901 (OLS)# H-M Riots, 1850-1950 -0.410 -1.308*** -1.245** -1.298** -1.187* -1.439** -1.375*

[0.371] [0.417] [0.533] [0.564] [0.635] [0.671] [0.716]R-squared 0.16 0.26 0.39 0.51 0.44 0.49 0.59

Any H-M Riot, 1850-1950 -0.217** -0.272** -0.233* -0.198** -0.203 -0.201 -0.096[0.106] [0.111] [0.111] [0.086] [0.126] [0.144] [0.150]

R-squared 0.21 0.27 0.31 0.40 0.41 0.44 0.57C) Coefficient on Medieval Port (2SLS)# H-M Riots, 1850-1950 -3.734 -3.938 -3.550* -2.056 -3.363* -2.374** -2.118**

[2.323] [2.531] [2.005] [1.421] [1.979] [1.034] [0.966]Any H-M Riot, 1850-1950 -0.472 -0.253 -0.657 -0.359 -0.240 -0.637 -0.648*

[0.553] [0.543] [0.526] [0.298] [0.370] [0.415] [0.333]Sample Full Full Coastal,

<200kmCoastal, <100km

Full Coastal, <200km

Coastal, <100km

Province / NS x Annex FE No No No No Yes Yes YesObservations 248 248 110 89 248 110 89

Notes: Each cell represents a regression. All regressions include quadratic polynomials in Longitude andLatitude and Log. Distances from the Modern Coast, Navigable Rivers and the Ganges, Coastal Townand Natural Disasters. Medieval-era controls add: Medieval Town, Mughal Mint, Other Patronage Center,Inland Trade Route, Skilled Crafts in Town, Major Shi’a State, Centuries Muslim Rule. Robust standarderrors (clustered at Native State x Annexation level): * significant at 10%; ** 5%; *** 1%

30

because they were close to indentations on the medieval coastline to towns that would have been

medieval ports had they enjoyed such proximity. This comparison yields results that are broadly

consistent with the average treatment effect in magnitude and direction.

An alternative transmission mechanism linking medieval trade to contemporary religious rela-

tions may be that medieval ports, which were historically wealthy, continue to be wealthy, and it is

the persistence of wealth that explains the connection between a medieval port legacy and a reduced

modern incidence of religious conflict. Conversely, it could be that following their decay, medieval

trading ports lost their Muslim population in the colonial period, and thus had less potential for

Hindu-Muslim conflict. Indeed, an argument with a venerable tradition in the popular press and

among academics is that Hindu-Muslim violence is more likely in areas with more mixed religious

populations. According to this argument, the more likely a member of one group is to interact or

compete with another, the more likely there is to be conflict.31

Table 8 estimates the effect of a medieval trading legacy on indices of income and religious

demography.32 As the table demonstrates, a medieval trade legacy has a negative effect on the

log. municipal income per capita. However, medieval trading ports, like other medieval towns and

Mughal mint towns, continue to have a greater proportion of Muslims than the towns and villages

in their surrounding district. However, unlike those other, more riot-prone ethnically mixed towns,

medieval ports exhibit lower violence, despite also being poorer. Once again, as the placebo checks

reveal, these effects are specific to medieval, not colonial, ports.

Though the effect of medieval trade on contemporary religious violence may operate partially

through population size and religious demography, it may be of interest to make these results

more comparable to existing studies and assess the residual effect of medieval trade on violence,

controlling for the ethnic mix and size of population. As Table 9 reveals, medieval ports with similar

populations and religious demography as other towns continue to be half as riot-prone. Further,

while the linear and quadratic terms on the proportion Muslim confirm the common finding in

the literature that more ethnically mixed towns tend to be more prone to riots in general, the

interactions of the linear and quadratic terms with a medieval legacy have exactly the opposite sign:

more religiously- mixed medieval ports actually are less prone to religious violence. Further the

31See Wilkinson (2005a) for an overview.32This sample is confined to those towns that possessed municipal corporations and thus enjoyed municipal income.

These were mainly in areas of British administration.

31

Table 8: Effect of Medieval Trade on Colonial Income and Religious Demography

(1) (2) (3) (4) (5) (6)A) OLS Medieval Overseas Port 0.140** 0.126* 0.103 -0.385** -0.384* -0.424*

[0.058] [0.072] [0.079] [0.168] [0.209] [0.211]Medieval Town 0.058** 0.074** -0.358** -0.218

[0.028] [0.032] [0.143] [0.158]Mughal Mint in Town 0.056** 0.062** 0.056 0.059

[0.025] [0.028] [0.198] [0.158]R-squared 0.38 0.42 0.55 0.11 0.22 0.5B) IVMedieval Overseas Port 0.299** 0.420** 0.355** -1.899** -2.463** -2.678***

[0.135] [0.187] [0.174] [0.832] [1.166] [0.914]C) Placebo (OLS)Port with foreign trade, 1907 0.042 0.041 0.018 -0.425* -0.142 0.236

[0.066] [0.079] [0.101] [0.235] [0.226] [0.350]R-squared 0.37 0.41 0.54 0.11 0.21 0.49SampleObservations 237 237 237 195 195 195

Controls Initial Medieval Medieval Initial Medieval MedievalProvince / NS x Annexation FE no no yes no no yes

|%Muslims in Town - District| Log Mun. Income Per Capita

British MunicipalitiesSubcontinental

Notes: All regressions include quadratic polynomials in Longitude and Latitude and Log. Distances fromthe Modern Coast, Navigable Rivers and the Ganges, Coastal Town and Natural Disasters. Medieval-eracontrols add: Medieval Town, Mughal Mint, Other Patronage Center, Inland Trade Route, Skilled Craftsin Town, Major Shi’a State, Centuries Muslim Rule. Robust standard errors (clustered at Native State xAnnexation level): * significant at 10%; ** 5%; *** 1%

32

Table 9: Interactions with Contemporaneous Population and Religious Demography

(1) (2) (3) (4) (5) (6) (7) (8)Neg. Bin. (I. Ratios)

Neg. Bin. (I. Ratios)

OLS OLS OLS OLS OLS OLS

Medieval Overseas Port 0.479*** 0.534** -1.213** -0.882* -0.472 -0.455 -1.292** -0.893*[0.127] [0.154] [0.544] [0.462] [0.526] [0.617] [0.538] [0.466]

Log Population, 1901 2.884*** 2.793*** 0.487* 0.610 1.652*** 1.443*** 0.432 0.540[0.249] [0.352] [0.273] [0.372] [0.412] [0.427] [0.274] [0.368]

% Muslim in town, 1901 1.398*** 1.334** 0.070 0.021 0.390*** 0.356* 0.097 -0.003[0.158] [0.164] [0.136] [0.177] [0.124] [0.178] [0.168] [0.185]

% Muslim in town2, 1901 0.872** 0.859*** 0.041 0.086 -0.219*** -0.220*** -0.139 -0.126[0.048] [0.046] [0.043] [0.061] [0.077] [0.074] [0.135] [0.125]

Medieval Port x Log Population -1.248** -1.601*** -1.490* -1.712[0.559] [0.543] [0.830] [1.221]

Medieval Port x % Muslim -0.428* -0.111 -0.026 -0.076[0.247] [0.311] [0.299] [0.368]

Medieval Port x % Muslim2 0.282* 0.151 0.060 0.086[0.139] [0.118] [0.104] [0.105]

Observations 248 248 110 89 248 248 110 89R-squared 0.54 0.64 0.38 0.41 0.58 0.69F-test (Medieval Port Variables) 3.95 3.84 1.46 1.39Prob>F 0.01 0.01 0.27 0.29Sample Full Full Coastal,

<200kmCoastal, <100km

Full Full Coastal, <200km

Coastal, <100km

Controls Initial Medieval Medieval Medieval Initial Medieval Medieval MedievalProvince / NS x Annex FE No No Yes Yes No No Yes Yes

Notes: All regressions include quadratic polynomials in Longitude and Latitude and Log. Distances fromthe Modern Coast, Navigable Rivers and the Ganges, Coastal Town and Natural Disasters. Medieval-eracontrols add: Medieval Town, Mughal Mint, Other Patronage Center, Inland Trade Route, Skilled Craftsin Town, Major Shi’a State, Centuries Muslim Rule. Orthogonal polynomials in the Proportion Muslim areused. All interactions are demeaned. Robust standard errors (clustered at Native State x Annexation level):* significant at 10%; ** 5%; *** 1%

33

medieval port effect is greater in larger towns.33 These results are consistent with the persistence

of inter-ethnic complementarity in medieval ports: when a minority population specialises in a

complementary service, increases in its population tend to increase intra-minority competition and

improve the terms of trade for the members of the majority, reducing incentives for expropriative

violence.

5.2 Persistence and Change, 1850-1995

I now examine the relationship between a medieval trade legacy and the time series variation in

ethnic riots. Figure 3 compares the timing of the first failure of inter-religious peace among the

same sample of towns, using data on religious violence through 1995 from Varshney and Wilkinson

(2004).

Notice first that most towns in the sample analysed above experienced at least one riot over the

145 year period, including a majority of medieval trading ports. Though tragic, the latter finding

adds confidence that the sample of medieval ports were also at risk for violence earlier in time

as well, and not different in unobserved ways that simply made conflict impossible or unreported.

However, at all times ethnic tolerance in medieval ports have enjoyed a better survival probability

than non-ports. The difference is most remarkable in such periods as the emergence of mass religious

politics in the wake of the Khilafat agitations on the 1920s and the turmoil of Partition in 1947

that steadily resulted in the failure of religious tolerance in other towns.

One important shock also experienced just prior to Partition was the shock to private organiza-

tional skills gained by India’s combat troops during the Second World War. Yet, Jha and Wilkinson

(2011) find that though districts that raised troops that acquired more combat experience tended

to have greater ethnic cleansing of their ethnic minorities, districts with medieval ports that ex-

perienced such organizational shocks reveal significantly less ethnic cleansing during the Partition.

This is once again consistent with the presence of persistent inter-ethnic complementarities, as both

members of the majority and minority would lose from ethnic cleansing in such an environment.

Table 10 provides the results of Cox proportional hazards regressions on survival of religious

tolerance for the same sample used in Table 4, between 1850 and 1950 to control for other factors

that might be driving the unconditional differences in Figure 3. A medieval trading legacy reduces

33This also reassures that our results are not driven by possible under-reporting in small medieval port towns.

34

Table 10: Survival of religious tolerance in towns of India, 1850-1950: Cox proportionalhazards regression providing hazard ratios of the time till first incidence of Hindu-Muslim rioting in asample of towns over 5000 in population in 1901.

Hazard Ratios (1) (2) (3) (4) (5)Medieval Port 0.165*** 0.118*** 0.075*** 0.047*** 0.004***

[0.058] [0.064] [0.020] [0.049] [0.007]Mughal Mint in Town 1.598 6.723* 1.488 3.632

[0.504] [6.932] [0.416] [6.452]Medieval Inland Trade Route 1.706*** 5.645** 1.655*** 5.534

[0.245] [4.291] [0.286] [7.488]Medieval Skilled Crafts in Town 1.910*** 1.292 2.197*** 1.867

[0.468] [0.996] [0.583] [3.061]Centuries Muslim Rule 1.314** 2.490*** 1.148 3.801**

[0.159] [0.639] [0.179] [2.158]Sample Full Full Full,

Coastal (<100km)

Full Full, Coastal

(<100km)Controls Initial Medieval Medieval Medieval MedievalProvince/ NS x Annex FE No No No Yes YesLog-Likelihood -1088.98 -1046.26 -86.13 -1005.69 -81.01Observations 408 408 107 408 107

Notes: All regressions include quadratic polynomials in Longitude and Latitude and Log. Distances fromthe Modern Coast, Navigable Rivers and the Ganges, Coastal Town and Natural Disasters. Medieval-eracontrols add: Medieval Town, Mughal Mint, Other Patronage Center, Inland Trade Route, Skilled Craftsin Town, Major Shi’a State, Centuries Muslim Rule. Robust standard errors (clustered at Native State xAnnexation level): * significant at 10%; ** 5%; *** 1%

the risk of the first outbreak of religious rioting in each year by around and ten times, with the

coefficient strengthening with the addition of controls (Cols 1-5). In contrast, tolerance was likely

to fail significantly more rapidly in towns that were on medieval inland trade routes, and in towns

with skilled human capital in the medieval era.

While it appears that a medieval trading legacy slowed the failure of religious tolerance in towns,

as Figure 3 suggests, the effect of institutions– even those forged over a millennium– can erode over

time. Carefully unpacking the time series variation in religious relations in India’s towns in the

post-independence period is beyond the scope of this paper, but remains a topic for future research.

However it is possible make a few observations. First, while the failure of tolerance in non-port

towns seems to have followed a gradual deterioration, the legacy of tolerance in medieval ports

appear to have been challenged more abruptly, particularly at the end of 1970s, timing that seems

to follow the oil shocks and the subsequent relative increase in funding by certain Gulf states for

religious schools and institutions that promote a version of Islam less adapted to domestic norms

35

Table 11: Effects on religious violence:, February-April 2002: Effects of medieval port legacyon medieval ports on days of rioting and probability of violence (including isolated incidents), in towns ofGujarat.

Dependent variable (1) (2) (3) (4)OLS OLS OLS OLS

Religious riot occurred?

-0.229*** -0.302*** -0.293*** -0.329**[0.049] [0.077] [0.099] [0.114]

R-squared 0.46 0.56 0.58 0.61

Days of rioting -1.754* -2.805** -2.733* -2.903*[0.858] [1.316] [1.418] [1.447]

R-squared 0.51 0.63 0.65 0.67

Any religious violence? -0.104 -0.129 -0.141 -0.18**[0.113] [0.139] [0.093] [0.082]

R-squared 0.61 0.62 0.65 0.67

Observations 199 153 148 137Historic controls no yes yes yesContemporary controls no no yes yes% Muslim(²) controls no no no yes

Notes: All regressions include linear and quadratic controls for longitude and latitude and log. distancesfrom coast, navigable rivers, coastal town, Natural Disasters, log. population (1991), Class of town (I,II,III),distance to Godhra (100,200,300km), district fixed effects. Historical controls include: Centuries Muslimrule, Muslim-founded or capital, Town in British India, Land revenue 1901. Contemporaneous controlsinclude: Proportion SC/ST, Municipal income and expenditures per capita, Modern port. Robust standarderrors in brackets (clustered at 1991 district): * significant at 10%; ** 5%; *** 1%

and traditions.34

5.3 Results from the Gujarat riots, February-April 2002

While the subcontinent-wide analysis allows me to evaluate the generalisability of the analysis

across regions that would later house three countries and one-fifth of the world’s population, it

leaves open the possibility that controls for unobserved geographical and historical variation are

inadequate. Further, as Figure 3 suggests, tolerance in medieval ports does appear to have weakened

over time. I therefore supplement the subcontinent-wide analysis using data from the two months

of religious violence in the towns of one particular state of India, Gujarat, following the burning of

the Sabarmati Express in Godhra in 2002.

34Such new externally-funded schools were among the only new buildings in relatively economically stagnantmedieval port towns like Veraval as well playing an increasing role even in economically vibrant port towns like Suratas well [author’s field observations, 2006-7].

36

To construct the variables used in this analysis, I went through news reports on Gujarat from

the day of the burning of the Sabarmati Express, February 27th until April 15th.35 Following

Varshney and Wilkinson (2004), I coded a riot as occurring in a town if there was evidence of

violence by communally-identifiable “mobs” or other large groups in that town. I also coded a day

of “violence” as having occurred in a town if there was an isolated incident, such as a stabbing,

without any evidence of broader groups being involved.

Table 11 shows the effects of medieval trading legacy on the probability of religious rioting

experienced by Gujarati towns following the burning of the Sabarmati Express at the town of

Godhra. All specifications include district fixed effects and controls for town size and distance to

Godhra. Once again, towns with a medieval trading legacy were around 20-30 percentage points

less likely to experience a riot. The magnitude of this effect is remarkably consistent with the South

Asia-wide colonial sample, and strengthens with the addition of historical and religious demographic

controls.

Table 11 also presents the medieval port effect on other outcomes: the duration of rioting

and whether the town experienced any violence (both riots and isolated incidents). As Row 2

suggests, a medieval trade legacy reduced the duration of religious rioting by around two to three

days. In contrast, medieval trading ports appear to have little effect on whether violent incidents

occurred at all. Rather the major effect of a medieval trade legacy appears to be to reduce the

escalation of religious violence into broader mob confrontation. This is consistent with the presence

of community-level incentives provided by institutional mechanisms that mitigate shocks to inter-

ethnic peace, rather than the presence of inherently peaceful individuals in medieval ports.36

These community level incentives appear to also be reflected in the electoral consequences of

Gujarat’s riots. In non-medieval port constituencies, the riots appear to have solidified support

for the ruling Bharatiya Janata Party (BJP), which was perceived to have been complicit in the

religious violence. The vote share for the BJP rose significantly from 43.43% in 1998 to 48.92%

in the elections immediately following the 2002 riots, sufficient to win the BJP an additional 13

35News sources include rediff.com and the Times of India. These were supplemented from an amicus curiae briefby the Concerned Citizens Tribunal, which provided information about less-widely publicised violence in smallertowns and villages.

36One respondent in Porbandar mentioned that attempts were made to instigate violence by outsiders, who sentwomen’s bangles to the members of the Hindu community as a symbol of their lack of masculinity in not joiningthe riots. However, he said “This did not work. We didn’t want [violence], our police and officials did not want it.Nobody wanted it.” [author’s field observations, 2006-7]

37

seats in non-medieval port constituencies. In contrast, while each of the ten Gujarati medieval port

constituencies had actually elected a BJP candidate in the 1998 elections– with an average vote

share of 55.3%– in the elections immediately following the Gujarat riots in 2002, medieval port

voters reduced the BJP vote share by 6.11 percentage points, enough to cause the BJP a loss of

three seats. Thus, relative to non-medieval ports, medieval ports showed a vote swing against the

party associated with the religious violence by a remarkable 11.6 percentage points or 6590 voters

per port. These findings may be due to the influx of refugees from the violence in other parts of

the state due to the ports’ reputations as “oases of peace”.37

6 Discussion

This paper has sought to establish that inter-ethnic medieval trade has left a lasting legacy on the

patterns of religious violence. In the ports of the medieval Indian Ocean, Islam, by making trade

accessible to all Muslims, satisfied three conditions that support peaceful co-existence over time:

the provision of a complementary service, a high cost for the majority to expropriate or replicate the

source of complementarity, and a means to redistribute the surplus from trade. These conditions

appear to have laid the basis for an enduring legacy of ethnic tolerance.

The logic underlying peaceful coexistence between Hindus and Muslims in medieval ports and

the supporting institutions that emerged can be readily applied to other historical and contem-

porary settings where non-local and local ethnic groups co-exist, both to understand why ethnic

tolerance fails, and how tolerance may be fostered. The theory and evidence described here suggests

that ethnic violence is more likely when ethnic groups compete, when the source of inter-ethnic

complementarity is easy for one group to expropriate or replicate, or when no mechanism exists

to redistribute the gains from trade. Competition between locals and immigrant groups for jobs

has often been cited as a reason for ethnic tension in the United States (Olzak, 1992). The theory

above suggests that these tensions are most likely to arise in jobs that are unspecialized and require

either few or generally-available skills or inputs, since these are the least costly for locals to enter.

Yet, even non-local minorities who do not compete, but enjoy complementarities that stem from

37At the same time, there have may also have been an element of retrospective sanctioning of the ruling party. InPorbandar, I interviewed several Muslim members of the BJP minority wing, which also maintained an active rolein several other medieval port towns. Donations to the BJP from minority groups were reportedly high prior to theviolence as well, but lowered afterwards.

38

tangible assets, such as land, machines or other forms of physical capital, may be targets of violence.

Being impossible to violently expropriate, specialized skills do provide a better basis for inter-

ethnic complementarity and tolerance, but even these can be replicated in the longer term. Mi-

norities that have specialized skills can become increasingly attractive targets of violence if locals

become able to duplicate those skills. The forced expulsion of Jews from Spain at the end of the 15th

century was precipitated in part by prior conversions, both forced and voluntary, of Jews to Chris-

tianity. These “new” Christians provided the administrative skills to Spanish rulers for which they

previously depended on the better-educated Jewish population (Benbassa and Rodrigue, 2000).

It is possible that the expansion of public education in Western Europe and the United States

may also have had the unfortunate side-effect of raising the likelihood of violence against educated

minority incumbents in skilled jobs by rendering them more replaceable by locals.38

In contrast to physical and human capital, most ethnic trading networks are both difficult to

steal– being intangible– and extremely costly to replicate. Because there are network externalities–

the value of a trading network increases with the size of its membership– there will be high costs for

any local to invest in a set of personal exchange relationships that would attain the scale necessary

to compete with an ethnic trading network. Thus non-locals can use the privileged access to goods

and services from ethnic ties elsewhere to provide the basis for sustained complementarity in mixed

communities.

Like Muslim traders in medieval Indian ports, Sephardic Jews benefited from valuable trading

networks in the 15th and 16th centuries that rendered them welcome arrivals in Ottoman ports in

the Mediterranean. With links to Spain and the Atlantic economy, their immigration was actively

encouraged by local Ottoman authorities and the city of Salonica in particular attracted a large

number of Jewish refugees. A combination of permissive immigration and religious specialisation

resulted in a long history of peaceful ethnic co-existence (Benbassa and Rodrigue, 2000). For the

next four centuries, Ottoman Salonica, sometimes referred to as the “Mother of Israel”, maintained

a remarkable degree of ethnic tolerance, with Jews specialised in commerce (Mazower, 2005).39

While the trading networks of the Chinese in modern Indonesia and South Asians in modern

38Indeed anti-Semitism in Nazi Germany was particularly pronounced among the relatively literate Protestantpopulation.

39Indeed, consistent with and building in part on an earlier version of this paper, Voigtlander and Voth (2011)find that German overseas trading ports that were part of the Hanseatic League were less likely to show persistentanti-Semitism.

39

East Africa also made them valuable to the local population, these groups have tended to lack

a general mechanism of redistribution. Chinese and South Asian ethnic trading networks, based

upon personal and community ties, were closed to competitors, and thus relatively small minority

groups were able to capture much of the gains from trade (Rauch and Casella, 2002). This ar-

guably rendered these minorities increasingly attractive targets for ethnic violence and susceptible

to expropriation by “strong” locals.

A key intuition of this paper is that by encouraging members of different ethnicities and religions

to be competitive among themselves but to assume complementary roles between groups, it is

possible to provide incentives to encourage the development of further complementary institutional

mechanisms that support inter-religious tolerance and economic integration.

These approaches may yield dividends for ethnic tolerance today, though some tradeoffs do

exist. Education systems that foster general human capital may have the adverse effect of making

ethnic groups economic substitutes, leading to increased conflict and polarization. In contrast,

developing education systems that allow minority individuals the choice of leveraging the advan-

tages of their group to engage in broader exchange, rather than promoting homogenisation of a

town’s human capital, may preserve ethnic complementarity and generate more opportunities for

exchange. Organisations that match members of different communities with complementary skills

in the creation of joint business ventures may also be effective for improving ethnic relations.40

A drawback of these approaches are, of course, that they may reinforce potentially dynamically

inefficient ethnic distinctions.

An alternative lesson does exist, however. Just as inter-ethnic business partnerships allowed

the sharing of gains in trade in the Indian Ocean, policies that assist communities in providing

explicit and well-publicised systems of joint assetholding may allow groups to share in each others’

futures in a way that reduces ethnic conflict and distinctions (Author, 2012). Trading networks

may have afforded minority groups an important source of comparative advantage that rendered

them valuable neighbours. Long after the decline of Indian Ocean trade, it may be that we can

apply some of their institutional learning to the pressing problems of inter-ethnic peace today.

40Such an organization has indeed been established by the Ismaili community in East Africa (Penrad, 2000).

40

References

Acemoglu, Daron, Simon Johnson, and James A. Robinson, “The colonial origins of com-parative development: an empirical investigation,” American Economic Review, 2001, 91 (5),1369–1401.

al-Din al Malibari, Zayn, Tohfut-ul-mujahideen, London: Oriental Translation Fund of GreatBritain, 1528. translated from the Arabic by Rowlandson, M.J. in 1833.

Alesina, Alberto and Eliana La Ferrara, “Ethnic diversity and economic performance,” Jour-nal of Economic Literature, 2005, 43 (3), 762–800.

Arasaratnam, Sinnappah, Maritime India in the seventeenth century, Delhi: Oxford UniversityPress, 1994.

Author, “Sharing the Future: Financial Innovations and Innovators in Solving the Political Econ-omy Challenges of Development,” in Pranab Bardhan and Timur Kuran, eds., Proceedings of theInternational Economics Association, Palgrave 2012.

, “A theory of trade, ethnic cronyism and tolerance.” in progress.

Baldwin, Kate and John Huber, “Economic Versus Cultural Differences: Forms of EthnicDiversity and Public Goods Provision,” American Political Science Review, November 2010, 104(4), 644–662.

Banerjee, Abhijit and Lakshmi Iyer, “History, institutions and economic performance: thelegacy of colonial land tenure systems in India,” American Economic Review, September 2005,95 (4), 1190–1213.

Barbosa, Duarte, A description of the coasts of East Africa and Malabar in the beginning of thesixteenth century, London: Hakluyt Society, 1519. translated from the Portuguese by Stanley,Henry E. J. in 1866.

Battuta, Abu Abdullah Ibn, The travels of Ibn Battuta, 1971 ed., New York: Lenox Hill, 1355.translated from the Arabic by Lee, Samuel, 1829.

Bayly, Christopher A., “The pre-history of ‘communalism’? Religious conflict in India, 1700-1860,” Modern Asian Studies, 1985, 19 (2), 177–203.

Bayly, Susan, Saints, goddesses and kings: Muslims and Christians in South Indian society 1700-1900, Cambridge: Cambridge University Press, 1989.

Benbassa, Esther and Aron Rodrigue, Sephardi Jewry: a history of the Judeo-Spanish com-munity, 14th-20th centuries, Berkeley: University of California Press, 2000.

Blake, Stephen P., Shahjahanabad : the sovereign city in Mughal India, 1639- 1739, CambridgeUniversity Press, 1991.

Bohlken, Anjali Thomas and Ernest Sergenti, “Ethnic Violence and Economic Growth: AnEmpirical Investigation of Hindu-Muslim Riots in India,” Journal of Peace Research, forthcom-ing.

Bose, Sugata, A hundred horizons: the Indian Ocean in the age of global empire, Cambridge, MA:Harvard University Press, 2006.

41

Bouchon, Genevieve, “A microcosm: Calicut in the Sixteenth century.” In Lombard and Aubin,eds (2000) chapter 4, pp. 78–87.

Brass, Paul R., The production of Hindu-Muslim violence in contemporary India, Seattle: Uni-versity of Washington Press, 2003.

Campbell, James M., ed., Gujarat Population- Musalmans and Parsis, reprinted, 1990 ed.,Vol. 9 of Gazetteer of the Bombay Presidency, Bombay: Govt. Central Press, 1899.

Casson, Lionel, ed., The Periplus Maris Erythraei, Princeton: Princeton, 1989.

Census of India, Census of India, Religion Paper, Vol. IV-B of 1 2001.

Chakravarti, Ranabir, “Nakhudas and Nauvittakas: ship-owning merchants in the West Coastof India (C. AD 1000-1500),” Journal of the Economic and Social History of the Orient, February2000, 43 (1), 34–64.

Chaudhuri, K.N., Trade and civilisation in the Indian Ocean : an economic history from the riseof Islam to 1750, Cambridge: Cambridge University Press, 1995.

Chua, Amy, World on fire: how exporting free market democracy breeds ethnic hatred and globalinstability, New York: Doubleday, 2003.

Das, Veena, “Introduction: communities, riots and survivors,” in Veena Das, ed., Mirrors onviolence: communities, riots and survivors in South Asia, Delhi: Oxford University Press, 1990,chapter 1, pp. 1–37.

Dasgupta, Ashin, Merchants of maritime India, 1500-1800, Aldershot: Variorum, 1994.

, India and the Indian Ocean world: trade and politics, Oxford, UK: Oxford U Press, 2004.

di Verthema, Ludovico, “The nauigation and voyages of Lewes Vertomannus, gentleman to thecitie of Rome, Hakluyt, 1503. transl. from the Latin by Eden, Richard, 1576.

Dippel, Christian, “Forced Coexistence and Economic Development: Evidence from NativeAmerican Reservations,” January 2011. mimeo, UCLA.

Eaton, Richard M., The rise of Islam and the Bengal frontier, 1204-1760, Berkeley, CA: Uni-versity of California Press, 1993.

Engineer, Asghar Ali, “Old Delhi too in the grip of communal frenzy,” in Asghar Ali Engineer,ed., Asghar Ali Engineer, ed., Delhi: Ajanta, 1988, chapter 10, pp. 65–74.

, Communalism in India: a historical and empirical study, New Delhi: Vikas, 1995.

Esteban, Joan and Debraj Ray, “On the salience of ethnic conflict,” February 2007. workingpaper, NYU, available at: www.econ.nyu.edu/user/debraj/Papers/EstebanRaySalience.pdf.

Fearon, James D. and David D. Laitin, “Explaining inter-ethnic cooperation,” AmericanPolitical Science Review, December 1996, 90 (4), 715–735.

and , “Ethnicity, insurgency and civil war,” American Political Science Review, 2003, 97 (1),75–90.

42

and , “Sons of the Soil, Migrants and Civil War,” World Development, February 2011, 39(2), 199–211.

Glaeser, Edward L., Rafael La Porta, Florencio Lopez de Silanes, and Andrei Shleifer,“Do institutions cause growth?,” Journal of economic growth, 2004, 9 (3), 271–303.

Goitein, Shelomo D., Studies in Islamic history and institutions, Leiden: E.J. Brill, 1966.

Greif, Avner, Institutions and the path to the modern economy: lessons from medieval trade,Cambridge: Cambridge University Press, 2005.

and David D. Laitin, “A theory of endogenous institutional change,” American PoliticalScience Review, 2004, 98 (4), 633–52.

Guiso, Luigi, Paola Sapienza, and Luigi Zingales, “Long Term Persistence,” working paper08-11, Chicago GSB 2008.

Habib, Irfan, An Atlas of the Mughal Empire, Delhi: Oxford University Press, 1982.

Habyarimana, James, Macartan Humphreys, Daniel Posner, and Jeremy Weinstein,Coethnicity: Diversity and the dilemmas of collective action, Russell Sage, 2009.

Haynes, Douglas E., Rhetoric and ritual in colonial India: the shaping of public culture in Suratcity, 1852-1928, Berkeley: University of California Press, 1991.

Hearn, Paul, “Global GIS database. Digital atlas of South Asia,” electronic resource 2001. USGeological Survey digital data series.

Hirschman, Albert O., The passions and the interests: political arguments for capitalism beforeits triumph., Princeton, NJ: Princeton University Press, 1977.

Horowitz, Donald L., Ethnic groups in conflict, Berkeley: University of California Press, 1985.

Jha, Saumitra, “Maintaining peace across ethnic lines: new lessons from the past,” Economicsof Peace and Security Journal, July 2007, 2 (2), 89–93.

and Steven Wilkinson, “Veterans, organizational skill and ethnic cleansing: evidence fromthe Partition of South Asia,” 2011. SSRN working paper 1739812.

Jukua, Chau, His work on the Chinese and Arab trade in the twelfth and thirteenth centuries,entitled Chu-fan-chi, New York: Paragon Books, 1225. translated from the Chinese by Hirth,Friedrich and Rockhill, W.W. in 1966.

Kandori, Michihiro, “Social norms and community enforcement,” Review of Economic Studies,1992, 59, 63–80.

Landa, Janet T., Trust, ethnicity and identity: beyond the institutional economics of ethnictrading networks, contract law and gift-exchange., Ann Arbor: University of Michigan Press,1994.

Lombard, Denys, “Introduction.” In Lombard and Aubin, eds (2000) chapter 1.

and Jean Aubin, eds, Asian merchants and businessmen in the Indian Ocean and the ChinaSea., New Delhi: Oxford University Press, 2000.

43

Mazower, Mark, Salonica, City of Ghosts, New York: Alfred A. Knopf, 2005.

Mehta, Makrand, Indian merchants and entrepreneurs in historical perspective: with specialreference to the shroffs of Gujarat, 17th to 19th centuries, Delhi: Academic Foundation, 1991.

Michalopoulos, Stelios, “The origins of ethnolinguistic diversity,” American Economic Review,forthcoming.

Miguel, Edward, Shanker Satyanath, and Ernest Sergenti, “Economic shocks and civilconflict: an instrumental variables approach,” Journal of Political Economy, 2004, 112 (4), 725–754.

Milgrom, Paul and John Roberts, “The Economics of Modern Manufacturing: Technology,Strategy and Organization,” American Economic Review, 1990, 80 (3), 511–28.

More, J.B.P., The political evolution of Muslims in Tamilnadu and Madras, 1930-1947, OrientLongman, 1997.

Noe, Samuel V., “The walled city of Delhi,” in Hooshang Amirahmadi and Salah S. El-Shakhs,eds., Urban development in the Muslim World, New Brunswick: Rutgers University, 1993.

Nunn, Nathan and Leonard Wantchekon, “The Slave Trade and the Origins of Mistrust inAfrica,” American Economic Review, forthcoming.

Olzak, Susan, The dynamics of ethnic competition and conflict, Stanford, CA: Stanford UniversityPress, 1992.

Penrad, Jean-Claude, “The Ismaili presence in East Africa: a note on its commercial historyand community organization.” In Lombard and Aubin, eds (2000) chapter 17, pp. 223–238.

Polachek, Solomon and Carlos Seiglie, “Trade, peace and democracy: an analysis of dyadicdispute,” in “Handbook of Defense Economics,” Elsevier, forthcoming.

Prior, Katherine, “Making history: the state’s intervention in urban religious disputes in theNorth-Western Provinces in the early nineteenth century,” Modern Asian Studies, February1993, 27 (1), 179–203.

Rauch, James and Alessandra Casella, “Anonymous Market and Group Ties in InternationalTrade,” Journal of International Economics, October 2002, (58), 19–47.

Raychoudhari, Tapan, “The agrarian system of Mughal India: a review essay,” in MuzaffarAlam and Sanjay Subrahmanyam, eds., Muzaffar Alam and Sanjay Subrahmanyam, eds., Delhi:Oxford University Press, 1998, chapter 8.

Risley, H., W.S. Meyer, R. Burn, and J.S. Cotton, eds, The Imperial Gazetteer of India,Vol. various volumes, Oxford: Clarendon Press, 1909.

Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi, “Institutions rule: the primacyof institutions over geography and integration in economic development,” Journal of economicgrowth, 2004, (9).

Smith, Richard J. and Richard W. Blundell, “An Exogeneity Test for a Simultaneous Equa-tion Tobit Model with an Application to Labor Supply.,” Econometrica, 1986, 54 (4), 679–686.

44

Subrahmanyam, Sanjay, The political economy of commerce: southern India, 1500-1650, Cam-bridge: Cambridge University Press, 1990.

, “Introduction.” In Lombard and Aubin, eds (2000) chapter 1.

Thapar, Romila, Somanatha: the many voices of a history, New Delhi, India: Penguin, 2004.

Topkis, Donald M., Supermodularity and complementarity, Princeton University Press, 1998.

Varshney, Ashutosh, Ethnic Conflict and Civic Life: Hindus and Muslims in India, New Haven:Yale University Press, 2002.

and Steven I. Wilkinson, “Varshney Wilkinson Dataset on Hindu-Muslim Violence in India,Version 2,” October 2004.

Voigtlander, Nico and Joachim Voth, “Persecution Perpetuated: The Medieval Origins ofAnti-Semitic Violence in Nazi Germany,” working paper 17113, NBER June 2011.

Wade, Robert, Village republics: economic conditions for collective action in South India, Oak-land: ICS Press, 1988.

Wilkinson, Steven I., Votes and violence: electoral competition and ethnic riots in India Cam-bridge studies in comparative politics, Cambridge: Cambridge University Press, 2004.

, “Introduction,” in Steven I. Wilkinson, ed., Steven I. Wilkinson, ed., New Delhi: OxfordUniversity Press, 2005, chapter 1, pp. p.1–34.

, “Wilkinson dataset on Hindu-Muslim Violence in India, 1600-1950,” September 2005.

Yule, Sir Henry, ed., Cathay and the Way Thither, London: Hakluyt Society, 1866.

45

(a)

Natu

ral

Harb

ors

and

Port

s,1558

E EEE

EE

EEE

EEE

EE

E

E

E EEEEEEEEEEEEE

EE E EEE EEE

E

E EEEEE

EE E

E

EE E E EE EE

EEEEEE

EEEE E

E EE E

E E E E E

E

EEEEEE EE E

E

E E EEE

EEEEE

EE

EE

E

EEE

EE E

EE

E

EE

E E EE E EE EE

EEE

EE EE

E E

EEE

EEEE

EE EE

EE

EE

EEE

E E

E E E E E E

EE

EE

E EEEE

E

EEEE

EEEEEE

E E E E E E EE EE E

EE

E

E

EEEEE EEE

EEE EEE

E E EE

EEEEEEE EEEEE

EEE

E

EEEEE

EE

EE EE

EEEEEE

EEEE

E

EEEEE EE EEEE

E E E E EE

EEE

E

EEE

EE

E E

E

EE

E

EE

EE

E

EE

EE

EE

E EE EEE EEE

EEE EEEEEE

EE E E

EE E

EE E E EEEE

E

EEEE E E E

EEE E EEE EEE

EEEEEE EE

EE

E E E

E

EEEEEEE

E E EEE E EE E E EE E

E EEE

EE

EE

E

E

E

E

E E

EEE

EE E E EE

E

E E

EE E EE

EEE E

EEE

E

EEEE E

EEEEEE

EE

E E EE

E

EE

E

E EE E

EE

E EEEE E E

E

EEEE

EEEEE EEE

EE EEE E

E

EEEE

EE E E EE

E

EEE

EE

EEE

E E

EEEEEE

EE

E

E E EEEEE E

EEE

EEEEE

E

E E EE E E EE

E E EE E E E E EE E

E E EEE

EE

EE

EEE E E EEEE E E EEEE

E

E

E

E

EE

E EE EEEEEEEE

E

EE

EEE

E

EE E E E

E

EEEE

E

EE

EEE

EE

EE

EE

EE

#

#

#

#

#

#

#

#

#

#

#

# # ##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

## #

#

#

#

#

#

## #

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

XWXW

XW

XW

XW

XW

XW

XW

XW XW

XW

XW

XW

XW

XW

XW

XWXW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW XWXW

XW

XW

XW

XW

XW

XW

XWXW

XW

XWXW

XW

XW

XWXW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW XW

XW

XW

XWXW

XW

XW

XW

XWXW

XW

Lege

ndXW

Mugh

al Mi

nts#

Medie

val P

orts

EInt

ernal

Trade

Rou

teNa

vigab

le Ri

vers

Year

s Mus

lim R

ule (u

pto 1

707-

Decil

es)

1 - 178 179 - 3

86 387 - 501 502 - 689 690 - 995

025

050

075

01,0

0012

5Kil

omete

rs

(b)

Med

ieva

lP

ort

s,In

tern

al

Tra

de

Route

sand

Mughal

Min

ts,

1595

Fig

ure

1:

Med

ieval

Natu

ral

Harb

ors

,P

ort

san

dP

atr

on

age

sourc

e:a)

Dio

go

Hom

em(1

558):

this

(exce

rpte

d)

Port

ugues

enav

igato

r’s

map

show

sm

edie

val

era

shel

tere

din

lets

and

the

loca

tion

of

act

ive

port

snea

rth

est

art

of

Euro

pea

nin

terv

enti

on

inth

eIn

dia

nO

cean.

b)

Ow

nca

lcula

tions

base

dup

on

the

Ain

-i-A

kba

ri(1

595),

Habib

(1982)

and

sourc

eslist

ed.

Med

ieva

lp

ort

ssp

read

acr

oss

both

coast

s.R

epet

itiv

ew

aves

of

inva

sion

thro

ugh

the

Khyb

erP

ass

hav

em

eant

that

close

rre

gio

ns

exp

erie

nce

dM

usl

imru

lefirs

tand

most

conti

nuousl

y.

Many

tow

ns

inth

eG

anges

pla

ins,

Dec

can

and

Guja

rat

wer

efo

unded

as

centr

esof

Musl

imp

oliti

cal

contr

ol

and

patr

onage

inth

em

edie

val

per

iod,

wit

hM

ints

bei

ng

esta

blish

edto

monet

ize

wea

lth.

46

E EEE

EE

EEE

EEE

EE

E

E

E EEEEEEEEEEEEE

EE E EEE EEE

E

E EEEEE

EE E

E

EE E E EE EE

EEEEEE

EEEE E

E EE E

E E E E E

E

EEEEEE EE E

E

E E EEE

EEEEE

EE

EE

E

EEE

EE E

EE

E

EE

E E EE E EE EEE

EE

EE EE

E E

EEE

EEEE

EE EE

EE

EE

EEE

E E

E E E E E E

EE

EE

E EEEE

E

EEEE

EEEEEE

E E E E E E EE EE E

EE

E

E

EEEEE EEE

EEE EEE

E E EE

EEEEEEE EEEEE

EEE

E

EEEEE

EE

EE EE

EEEEEE

EEEE

E

EEEEE EE EEEE

E E E E EE

EEE

E

EEE

EE

E E

E

EE

E

EE

EE

E

EEE

E

EE

E EE EEE EEE

EEE EEEEEE

EE E E

EE E

EE E E EEEE

E

EEEE E E E

EEE E EEE EEE

EEEEEE EE

EE

E E E

E

EEEEEEE

E E EEE E EE E E EE E

E EEE

EE

EE

E

E

E

E

E E

EEE

EE E E EE

E

E E

EE E EE

EEE E

EEE

E

EEEE E

EEEEEE

EE

E E EE

E

EE

E

E EE E

EE

E EEEE E E

E

EEEEEE

EEE EEE

EE EEE E

E

EEEE

EE E E EE

E

EEE

EE

EEE

E E

EEEEEE

EE

E

E E EEEEE E

EEE

EEEEE

E

E E EE E E EE

E E EE E E E E EE E

E E EEE

EE

EE

EEE E E EEEE E E EEEE

E

E

E

E

EE

E EE EEEEEEEE

E

EE

EEE

E

EE E E E

E

EEEE

E

EE

EEE

EE

EE

EE

EE

#

#

#

#

#

#

#

#

#

#

#

# # ##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

## #

#

#

#

#

#

## #

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

XWXW

XW

XW

XW

XW

XW

XW

XW XW

XW

XW

XW

XW

XW

XW

XWXW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW XWXW

XW

XW

XW

XW

XW

XW

XWXW

XW

XWXW

XW

XW

XWXW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW XW

XW

XW

XWXW

XW

XW

XW

XWXW

XW

Lege

ndXW

Mugh

al Mi

nts#

Medie

val P

orts

EInt

ernal

Trade

Rou

teNa

vigab

le Ri

vers

% Mu

slim

in To

wn, 1

931 (

Decil

es)

0 - 0.0

4880.0

489 - 0

.0896

0.0897

- 0.12

40.1

25 - 0.

165 0.166

- 0.203 0.2

04 - 0.

252 0.253

- 0.312 0.3

13 - 0.

410 0.411

- 0.546 0.5

47 - 0.

999

Prop

. Mus

lim 19

42 (D

ecile

s)

0.000 0.0

01 - 0.

018 0.019

- 0.037 0.0

38 - 0.

060 0.061

- 0.083 0.0

84 - 0.

128 0.129

- 0.179 0.1

80 - 0.

494 0.495

- 0.759 0.7

60 - 1.

000

025

050

075

01,0

0012

5Kil

omete

rs

(a)

Rel

igio

us

Com

posi

tion,

1931

E EEE

EE

EEE

EEE

EE

E

E

E EEEEEEEEEEEEE

EE E EEE EEE

E

E EEEEE

EE E

E

EE E E EE EE

EEEEEE

EEEE E

E EE E

E E E E E

E

EEEEEE EE E

E

E E EEE

EEEEE

EE

EE

E

EEE

EE E

EE

E

EE

E E EE E EE EEE

EE

EE EE

E E

EEE

EEEE

EE EE

EE

EE

EEE

E E

E E E E E E

EE

EE

E EEEE

E

EEEE

EEEEEE

E E E E E E EE EE E

EE

E

E

EEEEE EEE

EEE EEE

E E EE

EEEEEEE EEEEE

EEE

E

EEEEE

EE

EE EE

EEEEEE

EEEE

E

EEEEE EE EEEE

E E E E EE

EEE

E

EEE

EE

E E

E

EE

E

EE

EE

E

EE

EE

EE

E EE EEE EEE

EEE EEEEEE

EE E E

EE E

EE E E EEEE

E

EEEE E E E

EEE E EEE EEE

EEEEEE EE

EE

E E E

E

EEEEEEE

E E EEE E EE E E EE E

E EEE

EE

EE

E

E

E

E

E E

EEE

EE E E EE

E

E E

EE E EE

EEE E

EEE

E

EEEE E

EEEEEE

EE

E E EE

E

EE

E

E EE E

EE

E EEEE E E

E

EEEE

EEEEE EEE

EE EEE E

E

EEEE

EE E E EE

E

EEE

EE

EEE

E E

EEEEEE

EE

E

E E EEEEE E

EEE

EEEEE

E

E E EE E E EE

E E EE E E E E EE E

E E EEE

EE

EE

EEE E E EEEE E E EEEE

E

E

E

E

EE

E EE EEEEEEEE

E

EE

EEE

E

EE E E E

E

EEEE

E

EE

EEE

EE

EE

EE

EE

#

#

#

#

#

#

#

#

#

#

#

# # ##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

#

## #

#

#

#

#

#

## #

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

XWXW

XW

XW

XW

XW

XW

XW

XW XW

XW

XW

XW

XW

XW

XW

XWXW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW XWXW

XW

XW

XW

XW

XW

XW

XWXW

XW

XWXW

XW

XW

XWXW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW

XW XW

XW

XW

XWXW

XW

XW

XW

XWXW

XW

Lege

ndXW

Mugh

al Mi

nts#

Medie

val P

orts

EInt

ernal

Trade

Rou

teNa

vigab

le Riv

ers# R

iots,

1850

-1949

0

1

2

3

4

56 -

9 10 - 14

15 - 24

25 - 40

Prop

. Mus

lim 19

42 (D

ecile

s)

0.000 0.0

01 - 0.

018 0.019

- 0.037 0.0

38 - 0.

060 0.061

- 0.083 0.0

84 - 0.

128 0.129

- 0.179 0.1

80 - 0.

494 0.495

- 0.759 0.7

60 - 1.

000

025

050

075

01,0

0012

5Kil

omete

rs

(b)

Hin

du-M

usl

imR

iots

,1850-1

950

Fig

ure

2:

Med

ieval

Legacie

s:R

eligio

us

Com

posi

tion

an

dH

ind

u-M

usl

imR

iots

The

patt

ern

of

moder

nre

ligio

us

dem

ogra

phy

mim

ics

patt

erns

of

Musl

imru

le,

med

ieva

ltr

ade

and

politi

cal

patr

onage.

Med

ieva

lp

ort

sand

majo

rM

usl

impatr

onage

cente

rs(s

uch

as

those

that

house

dM

ints

)co

nti

nued

tohav

egre

ate

rM

usl

imp

opula

tions

than

nea

rby

are

as

even

in1931.

47

Other Towns

Medieval Ports

Khi

lafa

t Mov

emen

t Fai

ls

Par

titio

n

Oil

Sho

cks

0.2

5.5

.75

1

0 50 100 150

years elapsed since 1850

Figure 3: Timing of the first failure of religious tolerance, 1850-1995: This Kaplan-Meiercurve compares the survival rate of towns without violence. Notice that most towns in the sample experi-enced at least one riot over the 145 year period, including most medieval ports. However, medieval portshave consistently survived for longer without religious violence, including in the aftermath of the KhilafatMovement and the Partition of South Asia. Unlike the gradual failure of tolerance in other towns, 35%of medieval ports abruptly experienced their first Hindu-Muslim riot in the years following the OPEC oilshocks (1973-79).

48