Acemoglu, Johnson and Robinson (AJR) attempt to measure the effects of institutions on income differences by introducing an exogenous source of variation in institutions to measure their differing outcomes. They begin by pointing out that the history of colonisation resulted in different institutions being formed: some countries received extractive institutions (whereby the coloniser would simply extract all resources but would not build proper institutions to promote growth and sustainable living) whilst others received inclusive institutions. This depended upon the ability for colonisers to settle, if a country was full of disease then the coloniser would not wish to live in this colony, but simply take as many resources as possible and then leave. This was the fate of many African colonies where Europeans were unable to settle due to high levels of diseases which Europeans hadn’t developed immunity to. On the other hand, countries such as the Americas, Australia and New Zealand were much more hospitable to European settlers who decided to build colonies where they and their descendants would live. AJR believe these institutions then persisted, and the effects can still be felt today.If they are correct then they are able to use the mortality rate experienced by Europeans living in a country as an instrument for that countries institutions today.
Diagrammatically, their theory boils down to:
Settler Mortality -> Settlements -> Early Institutions -> Current Institutions -> Current Economic Performance
A simple regression does indeed show that colonies where Europeans faced higher death rates are a lot poorer today than countries which had lower death rates for Europeans.The authors then regress this instrument on institutions as follows:
Current Economic Performance = a(Institutions) + b(Other Controlled for Factors) + c(omitted variables and random error)
The estimated value for a tells us how important institutions are for current economic performance. The authors measure institutions using the risk of expropriation (the risk that an investment or property will be seized by the government). Because institutions is likely to be correlated with other omitted variables this introduces an omitted variable bias, whereby our estimated a is incorrect because of a correlation between institutions and other uncontrolled factors. To overcome this the authors use this settler mortality as an instrument, estimating:
Instruments = a + b(settler mortality) + u
And then substituting this into the structural regression equation.
They find that mortality rates faced by settlers over 100 years ago explains 25% of the variation in current institutions.
Overall, they find that their IV estimate of 0.94 (of the effect of institutions) is greater than their OLS estimate of 0.52 (which suffers bias) which suggests downward bias and thus that the measurement error in the institutions variable is more important than reverse causality or omitted variable bias. Acemoglu et al conclude that the results show a “large effect of institutions on economic performance” with the instrument explaining 25% of variation in today’s income.
It should be noted that Albouy criticises this use of IV though as he says there are measurement errors in the settler mortality data, particularly given that it is thin and relies on data from soldiers, bishops and labourers. He points out that out of the sample of 64 colony countries only 28 have their own mortality rates, with the rest being assigned the mortality rates of other countries which are judged to be similar in terms of disease environment.