Hi,
The setting:
In a research project of mine I am researching the effect of a CEO's influence in the acquiring firm on the success of an acquisition made by the firm. My sample consists of around 3,000 transactions between 1996 and 2018 whereas each observation represents one Deal. For each deal I have data on acquiring firm level (e.g., CEO in place, size, profitability, listing status), CEO level (e.g., sex, age, tenure) and deal level (e.g., transaction volume, payment method).
I have 4 continuous variables of interest that work as proxies for a CEOs influence in the acquiring firm.
Question 1
Fixed effect models rely on sufficient within variation of the variables of interest and - at least as to my understanding - insufficient within variation can cause some problems when running those models. While there are tests out there to help one to decide on whether run fe models I was wondering how you decide if the observed within variation is large enough. Here is the xtsum output for my four variables of interest (I dropped singleton observations as suggested by Correia 2015):
Related literature in my research field doesn't refer to any tests to specify whether fe would be appropriate so I was wondering if one could tell simply by looking at the xtsum output?
Question 2
How do I observe the within variance if I apply multiple fixed effects?
My firm fixed effects model takes on the following form:
With Firm_Controls, CEO_Controls and Deal_Controls containing variables that have been found to impact merger success (the DV) on the firm (e.g., size), CEO (e.g., age) and Deal (e.g., payment method) level.
I run this firm fixed effects model to adjust for unobservable firm level characteristics that are likely to influence my DV and IVs. I also include Year Fixed effects.
A related concern is that unobservable CEO level characteristics bias the findings so that I would also like to check the robustness of my findings by including CEO firm fixed effects. In the firm fixed effects model above I compare the same firm that made several acquisitions in my sample but with varying values of my IVs (i.e., observing the within-firm variance).
In a CEO firm fixed effects model I would compare the deals made by the same firm and same CEO but with varying IV (observing the within CEO-firm variance).
Before conducting this analysis I would like to check the within variance (as in the firm effects model). I tried to observe the desired variation using xtsum and sumhdfe but didn't really get what I am looking for. Any help is much appreciated.
The setting:
In a research project of mine I am researching the effect of a CEO's influence in the acquiring firm on the success of an acquisition made by the firm. My sample consists of around 3,000 transactions between 1996 and 2018 whereas each observation represents one Deal. For each deal I have data on acquiring firm level (e.g., CEO in place, size, profitability, listing status), CEO level (e.g., sex, age, tenure) and deal level (e.g., transaction volume, payment method).
I have 4 continuous variables of interest that work as proxies for a CEOs influence in the acquiring firm.
Question 1
Fixed effect models rely on sufficient within variation of the variables of interest and - at least as to my understanding - insufficient within variation can cause some problems when running those models. While there are tests out there to help one to decide on whether run fe models I was wondering how you decide if the observed within variation is large enough. Here is the xtsum output for my four variables of interest (I dropped singleton observations as suggested by Correia 2015):
Code:
. xtsum Var1 Var2 Var3 Var4
Variable | Mean Std. dev. Min Max | Observations
-----------------+--------------------------------------------+----------------
Var1 overall | .1624156 .0471193 .0495121 .3118844 | N = 2670
between | .0373512 .0532582 .3118844 | n = 645
within | .0291219 .0314556 .2888065 | T-bar = 4.13953
| |
Var2 overall | .3191011 .7216935 0 4 | N = 2670
between | .5993148 0 4 | n = 645
within | .4214965 -1.680899 3.845417 | T-bar = 4.13953
| |
Var3 overall | .9621723 2.320448 0 13 | N = 2670
between | 1.942231 0 13 | n = 645
within | 1.230062 -6.837828 11.79551 | T-bar = 4.13953
| |
Var4 overall | .0544385 .063733 0 .3119208 | N = 2670
between | .0546248 0 .3119208 | n = 645
within | .0377405 -.2184922 .2865345 | T-bar = 4.13953
Question 2
How do I observe the within variance if I apply multiple fixed effects?
My firm fixed effects model takes on the following form:
Code:
//Fe Model with Firm and Year fixed effects reghdfe DV var1 var2 var3 var4 $Acquiring_Firm_Controls $CEO_Controls $Deal_Controls, vce(cluster Acquiring_Firm_ID) absorb(Acquiring_Firm_ID Year) (output omitted)
I run this firm fixed effects model to adjust for unobservable firm level characteristics that are likely to influence my DV and IVs. I also include Year Fixed effects.
A related concern is that unobservable CEO level characteristics bias the findings so that I would also like to check the robustness of my findings by including CEO firm fixed effects. In the firm fixed effects model above I compare the same firm that made several acquisitions in my sample but with varying values of my IVs (i.e., observing the within-firm variance).
In a CEO firm fixed effects model I would compare the deals made by the same firm and same CEO but with varying IV (observing the within CEO-firm variance).
Before conducting this analysis I would like to check the within variance (as in the firm effects model). I tried to observe the desired variation using xtsum and sumhdfe but didn't really get what I am looking for. Any help is much appreciated.
Comment