Event study plots are increasingly popular in applied research. esplot is a new command for stata allowing researchers to quickly and easily create event study plots.
Note that this package is under active development and is frequently updated. As a result, new features are frequently added; sometimes bugs come along with these new features, but I am often able to patch them quickly when they are reported.
-
Install from within stata
net install esplot, from("https://raw.githubusercontent.com/dballaelliott/esplot/pkg/") replace
OR 2. Download/clone this repository
The example folder contains a .do file and a .csv with example data. Users may find it useful to download the folder and run (or simply browse) make_examples.do for sample syntax.
These data can also be used to experiment with the examples provided on the help site. Note that example.csv uses the same variable names for ease of use, but that the values have been randomized, so plots may not align with the example images.
See site for introduction and overview.
Type help esplot after installation for internal stata help documentation.
esplot 0.10.0 (late May 2021). Continuous treatment is now allowed when using panel data syntax. That is, the values of event and compare can be other than 0 and 1.
esplot 0.9.5 (early March 2021). Lets users specify saturate or bin to control how endpoints are treated.
esplot 0.9.1 (late Feb 2021). Adds support for alternate syntax when data cannot be tsset (i.e. are not panel data).
esplot 0.7.1 (late Nov 2020). Adds support for quantile regression.
As of esplot 0.9.0, leads and lags outside of the specified window are estimated, but not displayed. This is a change to the default; in earlier versions, periods outside the window were binned together. The old functionality is available by specifying the suboption bin to window with window(start end, bin). But think before you bin! This forces the dynamic effects to be constant outside of the specified window. If you're comfortable with that restriction, it can help with the identification of time fixed effects, but if this assumption is violated, it can cause the dynamic effects within the window to be misidentified. See the docs for more information on these suboptions.
Currently, this project nominally supports Stata versions as early as 11. However, I've mostly developed and tested this program, especially the graphics, in Stata 16 and later. I've tried to test that core functionality works in older versions, but I can't guarantee that I haven't missed anything.