Tutorial 3: Stata Applications of IV method The University of Queensland ECON7360
Instructor: Rigissa Megalokonomou
The Effects of Institutions on Economic Development
Background The data set for this exercise comes from the paper by Daron Acemoglu, Simon Johnson and James Robinson ¡±The Colonial Origins of Comparative Development: An Empirical Investigation¡±, published in the American Economic Review, December 2001, vol. 91, p. 1369-1401. This can be downloaded from JSTOR using http://www.jstor.org/stable/2677930
Idea They want to test the idea that ¡®institutions matter¡¯ for economic growth, in particular that countries with more secure protection of property rights have an advantage in encouraging capital accumulation. But a simple regression of GDP per capita on some measure of the strength of property rights is vulnerable to the critique of omitted variables and reverse causality. They use an IV approach, instrumenting property rights with a measure of settler mortality from the 19th century. They argue this is a good instru- ment, because parts of the world in which settler mortality was high tended to introduce institutions designed to exploit the area¡¯s resources, while in areas where the settler mortality was low they tended not to do this and they tended to build sound institutions. And it is argued that these early institutions are correlated with current institutions.
Use the data in institution.dta for the following questions. (i) Describe the variables in the data.
1
(ii) How many countries are there in the sample?
(iii) Draw two scatter plots (i) between lgdp and prot, and (ii) between
prot and logmort.
(iv) Regress lgdp on logmort and regress prot on logmort (hint: use robust standard errors). Provide the IV estimate of the effect of property rights (prot) on GDP per capita (lgdp) using these two OLS estimates.
(v) We want to estimate the effect of the strength of property rights on GDP per capita. Implement the IV method by using 2SLS.
(vi) Now regress prot on logmort (hint: use robust standard errors). Gen- erate predicted value and call this prothat. Now regress lgdp on prothat. Show standard errors will be wrong.
Another researcher argues that settler mortality also had an affect on the population today who are of European descent and that the ¡®Neo-Europes¡¯ (those countries like the US, Canada etc that, through a combination of guns, germs and immigration, have a high proportion of their population who are of European descent) have better economic performance even given their institutions.
(vii) If this argument is correct, what is the consequence for the consis- tency of the IV estimate of the effect of prot on lgdp?
(viii) How would you deal with the problem?
(iX) Implement your proposed solution and comment on the results.
2