Dear all,

I am currently analyzing the link between board gender diversity and firm performance. I have a panel data of 264 firms from the S&P500 over the 2006-2015 period. I have 3 periods: before (2006-2007), during (2008-2009) and after the Subprime Crisis (2010-2015)

It's my first time looking at the difference in differences analysis, I want to see the impact of firms which have more women compare to the firms with less women on their board of directors (before VS during the Crisis).

I am trying first to do 1) a difference in differences without fixed effects and without control variables, 2) differences in differences with fixed effects but without control variables and finally 3) differences in differences with fixed effect and control variables.

My main dependent variable is TOBINS'q (firm value) and my main independent variable is WOB (% of women on board). I create a dummy variable called Crisis (=1 if YEAR = 2008 & 2009, = 0 otherwise).

Questions:

1) If i am doing a difference in differences, it will show me the difference outside the crisis (pre and post crisis) and during the Crisis right? if i want to see only before and during, how should i do it and do you think it is better ?

2) Can i use a continuous variable in interaction with my Crisis dummy variable (c.WOB##i.Crisis) ? for now i used the command xtile to create a new variable WOB_quart which gives me 1 (firms with the less women),2 ,3 and 4 (which have the highest percentage of women), i used this new variable WOB_quart to create my dummy variable (WOB_dummy) which gives = 1 if WOB_quart = 4 (top quartile) and 0 if WOB_quart = 1 (bottom quartile)

I used the following command without FE and control variables :
Code:
 diff Q2ln_w, t(WOB_dummy) p(Crisis)

DIFFERENCE-IN-DIFFERENCES ESTIMATION RESULTS
Number of observations in the DIFF-IN-DIFF: 1436
            Before         After    
   Control: 508            166         674
   Treated: 645            117         762
            1153           283
--------------------------------------------------------
 Outcome var.   | Q2ln_w  | S. Err. |   |t|   |  P>|t|
----------------+---------+---------+---------+---------
Before          |         |         |         |
   Control      | 0.535   |         |         |
   Treated      | 0.577   |         |         |
   Diff (T-C)   | 0.042   | 0.024   | 1.78    | 0.075*
After           |         |         |         |
   Control      | 0.329   |         |         |
   Treated      | 0.428   |         |         |
   Diff (T-C)   | 0.099   | 0.048   | 2.05    | 0.040**
                |         |         |         |
Diff-in-Diff    | 0.057   | 0.054   | 1.06    | 0.290
--------------------------------------------------------
R-square:    0.04
* Means and Standard Errors are estimated by linear regression
**Inference: *** p<0.01; ** p<0.05; * p<0.1

.
I don't really know how to interpret the results:

Is it right to say that "before" the Crisis, the firms from the top quartile "treated" compare to the control group, have a positive effect and significant coefficient (better performance), and same for during the Crisis ("After"). However in general, the results from Diff-in-Diff : it's positive but insignificant effect ?


Code:
If i want to include fixed effect first and then adding control variable, should I use xtreg ?
xtreg Q2ln_w WOB_dummy##Crisis i.YEAR#SIC_group, fe
&
xtreg Q2ln_w WOB_dummy##Crisis [Control variables] i.YEAR#SIC_group, fe
Thanks a lot for your help!

Stephan