Confusion between Lags and Leads

[Anupam Ghosh]

Hi Statalisters,

I am a little confused about lags and leads. I want to construct and estimate the following model, which hosts a lagged treatment effect. Based on my current understanding, if I want to estimate the effect of an event that occurred, say, in the past year, then I would need to create an indicator variable that is lagged by one time period and use it to regress my outcome variable.

Now, the paper says that the coefficient \gamma1 reflects the mean effect on outcome in years 0–4 after the event and H_{c, 0 to 4} = 1 if a county experiences a treatment 0-4 years ago. The paper also says that leads are the corresponding years prior to the event. Thus, if I have to generate a variable like H_{c, 0 to 4}, do I have to use some variation of lags or leads?

O_{ct} = \gamma_1 H_{c, 0 to 4} + \gamma_2 H_{c, 5 to 10} + \alpha_c + \alpha_t + \varepsilon_{ct} --- Model(1)

[Clyde Schechter]

xtset country year
gen byte H_c_0_to_4 = inlist(1, event, L1.event, L2.event, L3.event, L4.event)

where the variable event is coded 1 if an event took place in that country that year, and 0 if not.

I do not understand why the paper says that leads are the corresponding years prior to the event--leads refer to things later in time, lags refer to things earlier.