Getting Started⚓︎
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.
Install⚓︎
-
Install from within stata
net install esplot, from("https://raw.githubusercontent.com/dballaelliott/esplot/pkg/") -
Download/clone from github
The github repository can be found here.
Two options for syntax⚓︎
esplot is flexible and provides the user with two alternative ways of specifying the event study.
If we don't have panel data, we simply use the syntax
esplot <outcome> <event_time>
where the event_time variable is simply the time of the observation relative to treament (or the event).
example: state minimum wage laws
Let's assume we are measuring the effect of minimum wage laws on wage. For the sake of this example, let's say that two states, CA and NY, have minimum wage increases in 2011 and 2010, respectively. Thus, an observation in CA in 2010 would have event_time = -1, but an observation in NY in 2010 would have event_time = 0.
We would then simply use the syntax:
esplot wage event_time.
| state | year | event_time | wage |
|---|---|---|---|
| CA | 2010 | -1 | 10 |
| CA | 2011 | 0 | 12 |
| CA | 2012 | 1 | 12 |
| CA | 2012 | 1 | 15 |
| CA | 2009 | -2 | 11 |
| NY | 2009 | -1 | 10 |
| NY | 2010 | 0 | 15 |
| NY | 2011 | 1 | 15 |
| NY | 2010 | 0 | 17 |
If we do have panel data, then esplot has more powerful options.
In this paradigm, we call esplot with the following syntax:
esplot <outcome>, event(<event_indicator>)
Behind the scenes, esplot will transform the data to create a vector of event-time indicators. In this paradigm, the event_indicator variable should evaluate to 1 in periods where a unit experiences an event and zero elsewhere. esplot has additional features that are reserved for panel data. Currently, the compare option is only available when using panel data.
event_time vs. event_indicator variables⚓︎
A common source of error when using esplot is the incorrect definition of and event_time or event_indicator variable.
With panel data and esplot <outcome>, event(<event_indicator>) syntax, the event_indicator variable should be defined as
event_indicator = <current_time> == <time of event>
When using esplot <outcome> <event_time> syntax, the event_time variable should be defined as
event_time = <current_time> - <time of event>
unlocking esplot with panel data⚓︎
With panel data, we have the additional option compare, allowing us to estimate a difference-in-difference event study. The syntax
esplot <outcome>, event(<event_indicator>) compare(<alt_event_indicator>)
will return an event study plot of the difference between the event coefficients and the compare coefficents. For example, this framework can be used to compare the productivity effect of assigning a unit a high-skill manager, relative to a low skill manager.
Additionally, esplot appropriately handles sources of common mistakes when manually generating a relative time variable, such as when some units have more than one event (or no event at all).
Advanced⚓︎
Accessing Regression Output⚓︎
In esplot 0.9.3 (March, 2021) and later, the internal reghdfe call is made available to be user. Typing estimates after esplot is done running will print the regression output and save the return objects to e() and r(). This allows esplot estimates to be used with post-estimation commands, like test and eststo/esttab.
See help reghdfe for more information on the information that is returned.
Note that because of the technical implementation, the estimates are stored and then reloaded; this means that e(sample) cannot be used (see help estimates use for more information.) The save_sample option is provided as a workaround.