Example Question (Econometrics)

Example Question (Econometrics):
1) Describe the identification strategy of Acemoglu et al. (The Colonial Origins of Comparative
Development: An Empirical Investigation).
In their base setup the authors are interested in the relationship between GDP per capita in
1995 (gdp1995) and current institutions (CI). They argue that differences in institutions cause
differences in GDP. (with the following regression equation: Gdp1995= β*CI+ ε)
However, there exist potential problems regarding the identification assumption [cov(CI, ε)=0].
First, there might be the problem of reverse causality (rich countries can afford better
institutions). Second, there might exist potential confounders. Particularly, there might lurk
variables in the error term that affect both GDP and current institutions. An example might be
human capital. Higher average human capital (HK) in the country leads to better institutions as
well as higher GDP [then the error term as well as institutions are functions of human capital,
ε(HK) and CI(HK), which leads to cov(CI, ε)≠ 0; note the error term ε is the unexplained rest of
GDP differences across countries that is not explained by our right hand side variables, but
explained by unobserved factors like human capital.]
The author solve this issue by relying on a IV strategy and using settler mortality in 1900
(mor1900) as instrument (first stage: CI=*mor1900+u which yields the fitted values
CI(mort1900) which we use in the second stage: Gdp1995= β* CI(mort1900)+ ε). The argument
here is that mortality affected the European settlement extent and consequently the institutions
that the Europeans introduced in the respective colonies. Through persistence we have that
past institutions explain today’s institutions. The identification assumption is now
[CI(mort1900), ε)=0]. That is mortality affects todays GDP only through today’s institution but is
not correlated to any other factor in the error term which affects todays GDP. This assumption,
however, is likely to be violated, because Europeans also introduced infrastructure as well as
European education systems (universities) in the colonies. The extent of these (infrastructure)
investments, however, depended on the extent of European settlement intensity. Thus,
mortality in 1900 not only explains institutions but also human capital (ε(HK(mort1900))) and
infrastructure. And we have again cov(CI(mort1900), ε(HK(mort1900)))≠ 0.