Last year we decided to switch from Stata to R in our Master program. This has had huge problem for the Econometrics 1 course that I teach. Econometrics 1 is an introduction to microeconometrics for economists, covering linear regression model, heteroskedasticity, panel data and endogeneity/instrumental variables.
First the positives. R is powerful and actively develop by lots of groups and individuals. It is at the forefront of statistics, data analytics and machine learning. Some packages are just wonderful. But … Stata is easier to use for teaching econometrics.
For example, econometricians tend to correct the standard errors to make them robust to heteroskedasticity. This is accomplish in Stata with a simple command:
regress y x1 x2 x3, robust
This is easy and intuitive and I can focus on the theory underlying this command. In R, this becomes:
reg <- lm(‘y~x1+x2+x3’,data)
library(sandwich)
robust_cov <- vcovHC(reg, type = “HC1”)
library(lmtest)
coeftest(reg, vcov = robust_cov)
And if I want a clean table,
library(stargazer)
robust_se <- sqrt(diag(robust_cov))
stargazer(reg, type = “text”,se = list(robust_se))
My feeling is that the students focus too much on the R command and not enough on understanding the econometrics. So something needs to be done and I already have something in mind ….