Trade, Institutions and Religious Tolerance: Evidence from India
By Saumitra Jha
Introduction
• The paper tries to explain how economic incentives can foster inter-ethnic cooperation
• It uses 1000 years of history as a natural experiment
• Background– 7th to 17th century India
• Rise of Islam• Complementarity of skills
– Muslim advantage in Indian Ocean shipping• Created incentives for inter-ethnic exchange, most strongly
in medieval trading ports• Incentives for building institutions fostering inter-ethnic
cooperation
Questions
• Did medieval trade foster later religious tolerance in India?
• Is it due to the development of persistent institutions?
Outline
• Theoretical Framework– Hypotheses
• Empirics– Some econometrics tools (OVB, Bad Proxy
Control)– Saumitra’s results
• Discussion of internal validity • Omitted Variables? Bad Proxies?
Theoretical Framework
• Dynamic• Agents
– Locals (Hindus) & Non-locals (Muslims)– Non-locals have better outside options– Strong (good at attacks) and Weak agents
• Actions– Stay or leave– Produce for exchange– Attack any other agent (capture profits)
• Markets, non-perfect
Key Factor #1: Complementarity
• A low degree of complementarity between local and non-local production– then incentive for strong local to attack non-
local: ethnic violence– Why?
• Induces the non-local to leave, which reduces future competition and raises profits
• Locals not likely to leave
Key Factor #2: Replicability
• A high degree of production replicability – Lowers complementarity over time
• Increases likelihood of attack
– Necessary to have low replicability
Key Factor #3: Minority size
• A smaller non-local minority– Less competition in trade therefore higher
profits• More tempting to attack and capture• More muslims makes violence less likely
• Redistributing institutions important
• Low barriers for non-locals to enter markets
Medieval Indian Trading Ports
• All three factors beneficial– High Complementarity
• Trade through islamic pilgrimages, The Hajj • Islamic law conducive
– Low Replicability• Trade networks hard to steal
– Minority Size• High intra-muslim competition • Low barriers to muslim entry
Institutional Development
• This should induce high demand for institutions
• Two types evolved– Replicability-reducing institutions
• Ethnic-specific apprenticeships• Social sanctions (Kaala-Paani)
– Profit sharing institutions• Joint ventures• Political delegations• Commercial taxation
Empirical Strategy
• Selection-on-observables
– Assumption: conditional on initial controls (time t-1), trading ports are as good as randomly assigned
– Reduced Form: What is the link between t & t-1?• t-1 = 700-1700 BC• t = 1850-1950, 2002
• IV/2SLS– ”Natural harbours” instrument for trading port.
• What hypothesis are we testing?
titititi controlstradeportY ,1,1,,
Omitted Variable Bias
• Suppose, for example, that the true relationship is
• If the medieval (time t-1) %-of-muslims (ethnic) is unobserved
• Then the omitted variable bias is:
• Where the last coefficient comes from the regression:
• So if more muslims causes less violence and trade is correlated with more muslims, we have a downward bias (overestimate trade’s effect on peace).
et ~
1,1,01, titietti etradeportethnic
titititi ethnictradeportY ,1,1,,
tititi tradeportY ,1,,~~~
Bad Proxy Control
• If we use the contemporary (time t) %-of-muslims as a proxy for the medieval ratio
• And the proxy measures
• Then the bias will be:
• Hence, if more muslims causes more violence, and trade is positively correlated with ratio of muslims, we have a downward bias (overestimate trade’s effect on peace)
titititi ethnictradeportY ,,1,,~~~
1,21,1, tititi ethnictradeportethnic
2
1~~
IV/2SLS
• ”Natural Harbours” (protected inlets) as an instrument for medieval trading port
• Key assumption: Natural Harbours increases the likelihood of becoming a trading port, and is uncorrelated with all other factors that determine violence.
• Do we believe in the instrument?• Problems
– Consistent, but not unbiased – The ”weaker” is the instrument, the more it is biased
towards the OLS estimate
Results 1
• Main Results– Table 3 – Do we have any bad proxy controls?– Can we think of omitted variables?– Why is the 2SLS effect 80-90% larger than
OLS?– Why do we need controls when using 2SLS?
Results 2
• Alternative hypotheses (i.e. not institutions)– Table 5 Outcomes
• Contemporary Income – Lower contemporary income (insign.)
• Contemporary Trade – Lower value of trade (insign.)
• %-Muslims– More muslims
• What if %-Muslims is an omitted variable?• Bad Proxy Control
– Table 7, See OLS, % Muslims– Lower OLS point estimate (Compare to Table 3)– Why?
Concluding Interpretations
• These are reduced form results– How do we know that the mechanism is institutions?
(We don’t)
– Why not other persistent mechanisms• Changing values?• Or migration of tolerant individuals into trading ports?• Or that trade weakens religious identity?
• Is it only ethnic violence, or crime in general?• If there is an effect on the latter, then it could be that trade
induced the development of rule of law.• Then religious tolerance has nothing to do with it.