In this analysis, I will ask you to use data from the Alvarez et al. data set on regimes and economic performance worldwide. The data are available as aclppanel.csv, and you can also access a codebook.
Using these data, try to determine whether dictatorship/democracy (REG) has a positive or negative causal effect on average education levels (EDT). Structure the analysis as a difference-in-differences setup, and deal with any problems that arise. Describe your findings.
The main contemporary R library we used that can accommodate treatments that happen at different points in time is fect, documented here. The command can be run as follows:
output.ect <- fect(OUTCOME.VARIABLE ~ TREATMENT.VARIABLE, index = c("COUNTRY", "YEAR"), force = "two-way",
data = DATAFRAME, method="fe",min.T0=1, se=1)It can be interesting to try different methods. gsynth instead of fe uses synthetic controls to estimate the diff-in-diff, for example, but only works when cases don’t switch back to control after becoming treated. mc uses a matrix completion method, and polynomial allows nonlinear time trends.