Re: st: regress with dummys | difference in difference estimator

[Julian Kochan <julian.kochan@gmx.de>]

Hi Fernando,

thanks for answering. 
There are definitely omitted variables. I want to control for Tobins Q, size, industry, leverage, etc.
My idea was to create a fixed effects variable. is it possible to include all my omitted variables into one fixed effect one ? 
Such that, i can conclude later on, that under control of my omitted variables, theres is a significant influence of debt maturity on corporate investments.

Thanks so far
regards
Am 28.03.2014 um 22:03 schrieb Fernando Furquim:

> Hi Julian,
> 
> I don't think your interaction of the indicator variables tells you
> anything that the two original variables already tell you. Beyond
> that, I am not sure of the exact definitions of your model, but there
> are lots of possible reasons for the low r^2. Maybe there are temporal
> effects, so you need lags? Maybe you have lots of omitted variables
> like profitability, industry, business cycle, total debt, etc? I don't
> know much about corporate investments, but I think a review of the
> relevant literature could hint at what kind of model makes conceptual
> sense.
> 
> Also remember to -xtset- your panel.
> 
> Good luck!
> Fernando
> 
> On Fri, Mar 28, 2014 at 11:51 AM, Julian Kochan <julian.kochan@gmx.de> wrote:
>> Hi,
>> 
>> this is an update to my previous question:
>> 
>> i have a set of data with financial firm characteristics of about 10.000 firms for Year 2000 - Year 2012.
>> 
>> My goal is to use a difference in difference estimator and to do a regression with the dependent variable: corporate_investments and independent variables: dummy pre_post and a dummy for treatment and control.
>> I want to see, how investments changed from 2007 to 2008 for the treated / control firms.
>> 
>> I generated a treatment dummy, labeling a firm as treated when Short Term Debt / Total LTDebt > 20 %. All others belong to the control group, such that treatment is equal to 1 and control is equal to 0.
>> I did the same for the pre_post dummy.  generate post_2007 = (Year >= 2007)
>> 
>> Then I generated a interaction term: gen interaction = post_2007*treatment_
>> 
>> Finally I regressed: reg corporate_investments post_2007 treatment_ interaction
>> 
>> This lead to an output of R^2 = 0.003 which is pretty bad. What could be the reason ?
>> Can someone help me with the difference in difference approach ?
>> Did I generate wrong dummys ?
>> 
>> Thanks
>> 
>> Julian
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> *   http://www.ats.ucla.edu/stat/stata/
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/