How to get the value of coeffiences minus one ?

[Olivia Li]

Hello everyone , I got a problem when doing the parallel trend exam. The pre_periods ssame to be above the "Y=0" line. But there is a decrease after pre2 till post1,post2 ...
So could you please tell me, if I want to get the value of regressors (i.e. Y ) minus 1 and draw a new coefplot, how could I do ?

Thanks !!

[Jared Greathouse]

I am confused. You don't really tell us how you've made the plot- in fact, if possible, please provide your data and code to make this plot to begin with.


My question though is why do you want to... subtract one from the coefficients? Why are these coefficients not good enough for this particular situation? Olivia Li

[Olivia Li]

I am confused. You don't really tell us how you've made the plot- in fact, if possible, please provide your data and code to make this plot to begin with.


My question though is why do you want to... subtract one from the coefficients? Why are these coefficients not good enough for this particular situation? Olivia Li

Hi, the reason why I want to do this is that I think the trend doesn't perfectly pass the parallel-trend-exam because it seems like before the shock, the coefficient is above the Y=0 line, while after that the trend is around Y=0.
IN a word ,it is not like the typical kind of parallel-trend-exam plot, which is zero before shock and negative/positive after the shock.

I am not sure if I get this right, but if it is not necessary to change my plot, could you please explain why ? Thanks sooo much !

[Jared Greathouse]

So let me just understand you, you're asking if this graph implies a violation of the parallel trends assumption needed for difference-in-differences?

[Olivia Li]

So let me just understand you, you're asking if this graph implies a violation of the parallel trends assumption needed for difference-in-differences?

Yes, that's right !

[Jared Greathouse]

I don't buy the parallel trends assumption here. Looks like the outcomes were trending downward before the policy, implying parallel trends don't hold for your comparison units.

So now I've got two questions for you: is only one unit treated? And, how many comparison units do you have? How many units are never treated?

[Olivia Li]

Thank you so much , Greathouse!
No, all the 30 firms are treated at different time points. Some of them are treated at 2016m6, some at 2016m12, some at 2017m7,and some at 2017m11

[Jared Greathouse]

Wait, all 30 firms are treated at some point?

[Olivia Li]

Wait, all 30 firms are treated at some point?

Yes, they are all treated.^_^