lagged explanatory variable

[Lisa Space]

Dear statalist community,

this is more a comprehension question than it is technical, however I could not find an answer online so if you guys could help it would be really nice.

I'm using a conditional fixed effects ordered logit model (blow-up and clusete namely) and a panel dataset to estimate the effect of a life-event (x) on ones preferences (y).
My explanatory variable is a life-event that occured in t-1. It is therefore a dummy taking value 1 if the event occured and 0 if it did not.

I have trouble interpreting what the model really does. I know that the fixed effect model looks at within variation so looks if one person changed when this event occured but does it look only at the effect of this one year (t-1 till t)?

I've also estimated another model using the same estimator, there I've regressed not the life-event (let's say getting married in t-1) but the complete state (married and non-maried) with each other and I get completely different results. What I thought that the model does here is to look at a longer period of time as my panel dataset is 13 years long. But I'm really unsure if my interpretations are right here so it would be really appreciated if somebody could help.

Thank you so much in advance

Lisa

[Clyde Schechter]

I'm not sure I understand what you have done. It would have been better to show the actual code you used.

But here's what I think. I think in your first situation, you are estimating the immediate shock produced on the outcome in the time period immediately following the event. The event is, in this model, assumed to have no lingering effect beyond that first time period. In your second situation, you are estimating the average long-term persistent effect of being in the post-event state.

So your first model estimates the immediate impact of getting married. The second model estimates the long-run effect of being married.