Difference between two coefficents

[Omar Shaher]

Dear Researchers, I have an unbalanced panel dataset for a list of firms for the period extending from 2000-2015.

Set of standards issued in 2003, and the adoption is voluntary. Firms started to adopt in different years. I want to examine what's called timely loss recognition, and to do that, I need to use the below model:

ACC= DCFO + CFO + DCFO*CFO

So I will measure it one time for the adopters and the other time for the none adopters, and accordingly, I have applied the following model:

gen DCFO_CFO= DCFO* CFOt1
xtreg ACC DCFO CFOt1 DCFO_CFO i.Year if Event==1, fe
xtreg ACC DCFO CFOt1 DCFO_CFO i.Year if Event==0, fe

The variable of interest is DCFO_CFO. So, I need to get the difference between the coefficient of DCFO_CFO of adopters in the first model and the coefficient of non-adopters DCFO_CFO in the second model, and I need to show the significance for the difference. So, could you please tell me what is the code for this, please?

Many thanks in advance.

[Omar Shaher]

Dear researchers,
Let me make it clearer:
If I have this model:

xtreg ACC DCFO CFOt1 DCFO_CFO i.Year if Event==1, fe
xtreg ACC DCFO CFOt1 DCFO_CFO i.Year if Event==0, fe

Can I get the difference between the coefficients for the first and second regression?
If yes, what is the code for this, please.

[Felix Bittmann]

Here a simple example with interaction effects

sysuse nlsw88, clear
reg wage i.union##(c.grade c.hours)

The interaction effects give the group differences (in the example, the groups are defined by union status, for you this is the event). As you see, the interaction is negative for hours (-0.0915). This means that the effect of hours is smaller in the union group than in the non-union group.

[Omar Shaher]

Dear Felix,

I can't thank you enough for the answer. I have used the code that you mentioned but I think the coefficient of hours is for the non-union group, while the interaction is the effect of the union group on the relationship between grade and wage, and to get the relationship between grade and wage for the union group, we can run the following code:

margins union , dydx ( grade hours )

But I think what I am looking for is to find the difference between the coefficient of grade before the interaction and the coefficient after the interaction?

[Felix Bittmann]

I am not sure I understand this correctly. With margins, you get the coefficients of, say, hours, separately for union and non union members, which are 0.05589 and -0.03561. The difference is .0915107, but this is exactly the value of the interaction term, where you also find the p-value of this difference. If you need a different statistic I am not sure what you mean, especially with "after the interaction". This makes it sound as there is a stepwise inclusion, but here there is no model without the interaction term included.

[Omar Shaher]

Dear Felix,

Now the image is so clear to me. Greatly appreciated.

Many thanks for the answer.