Help interpreting the results

[therock]

I am investigating the effect of 2002 Regulation requiring independent board of directors on CEO compensation. The regulation affected only some firms, since many other firms already had independent board prior to the regulation. Thus, I am using difference-in-difference methodology.

However, there are two ways of calculating the dependent variable, CEO compensation. The first way is simply using the natural logrithm of Total Compensation. The other way is using the natural logrithm of Excess Compensation, compensation adjusted for size and industry. The following model is used:

CEO Compensation = Inside Board + Post Regulation + Inside Board*Post Regulation + Control Variables  

The Inside Board (IB) is a dummy variable indicating companies that had to change board structure to comply with the regulation. The Post Regulation (PR) is a dummy variable indicating the years following the regulation; years after 2001. My data goes from 1996-2012.

I also performed the test that Inside Board + Inside*Post Regulation = 0. All results from SAS are presented in the table below:

	Natural Log of Total Compensation	Natural Log of Excess Total Compensation
Inside Board	0.6042***	-0.2558***
Post Regulation	0.5569***	-0.1286***
Inside Board * Post Regulation	-0.3975***	0.1431***
IB + IB*PR = 0	0.2067 (P-Value 0.0007)	-0.1127 (P-Value: 0.0161)

I need help interpreting the results. And determining which variable to use for the analysis. I cannot use both dependent variables in my paper.

Any help would be much appreciated! Thank you so much in advance!

[PaigeMiller]

Can you show us the SAS code you used? In fact, it would also be good to show us the actual SAS output rather than this table you typed in.

We can't really comment on which model to use as you have not stated any criteria that could be used to choose between the models of the two dependent variables.

[therock]

I understand if you cannot comment on which model to use. I just need help interpreting this numbers. Previous posts have helped guide the code I used.

[PaigeMiller]

---

@therock wrote:
I understand if you cannot comment on which model to use. I just need help interpreting this numbers. Previous posts have helped guide the code I used.

---

But we don't know what SAS code you used. Please tell us. And we would also like to see the actual SAS output instead of these tables that you typed yourself.

[ballardw]

---

@therock wrote:
I understand if you cannot comment on which model to use. I just need help interpreting this numbers. Previous posts have helped guide the code I used.

---

Please interpret:  42

Your question is analogous to mine. If we have no idea what created the numbers interpretation is pretty difficult if not impossible.

[therock]

Hi Everyone,

Sorry for getting back so late. I had some internet issues.

For my analysis, I use proc surveyreg. I have many control variables and therefore, I am not including in the question. 

The goal of this question is to ask somebody who has enough knowledge to interpret the results. 

For example, for the Excess Compensation variable this is what I wrote. I am hoping somebody can either say it is accurate or can correct it:

My findings indicate that excess CEO compensation decreased for all firms following the law. However, excess CEO compensation did not significantly decreased for firms that had to change board structure to comply with the law, indicating that CEOs of non-compliant firms were paid significantly more than CEOs of compliant firms following the mandate.

[mkeintz]

It is the effect of the regulation (for companies that were impacted by it) that you are most interested in.

If some industries (with high compensation growth rates) already had 80% independent boards, and others (with low compensation growth rates) only had 20% pre-regulation independent boards, I think using compensation as the dependent variable would confound the effect of industry and regulation.  (Your doing this model across all industries, right?).

And given the (probably) large number of industries in your data, addressing the above by including interaction terms of industry*regulation would be, to say the least, very messy.

But if you looked at excess compensation, because it is adjusted for industry, I would expect it to be more suited to testing for the effect of regulation.

However, I am speaking as a person who hasn't done this sort of work for a long time, and always was uncomfortable without checking my methods with others.  So you may want to confirm my advice with more authoritative sources.

[therock]

Hi,
I forgot to mention that I use industry and size as control variables in both models. The model that I am using is confirmed by my professor. What I need advice is regarding the interpretation. Please see my new post regarding interpretation. Am I interpreting it correctly? Thank you so much!