Hi,
I am trying my best to describe my problem as precise as possible without getting lost in details. That being said, please let me know if something is not clear.
I am working on a paper which examines the effect of a Chief Financial Officers' (CFOs) influence in a firm on certain characteristics of Mergers and Acquisitions (M&A) made by the firm in which the CFO is employed.
Data example showing selected variables whereas each entry represents one acquisition:
Based on management and financial literature I apply four measurements to capture the influence (often referred to as managerial power) of a CFO in the acquiring firm.
These measurements are:
1. CFO_PaySlice: CFOs’ annual compensation (Salary + Bonus) divided by the annual compensation of all top management team (TMT) members
2. CFO_No_Boardseats: Number of board seats a CFO holds outside the acquiring company
3. CFO_Perc_Own_Dir: CFO ownership in the company, which is defined as the number of common shares the CFO holds divided by the total number of shares held by all other TMT members.
4. CFO_No_Deals: Acquisition experience of a CFO, measured as the number of deals a CFO has participated in as a manager or director outside the acquiring company during the ten years prior to each deal.
My dependent variables are M&A related measurements that reflect certain characteristics of a deal, e.g., whether the deal was considered to be a success, the negotiation length of a deal or the premium paid by the acquirer to the target.
I have to admit that it is really hard to make causal inference for at least two reasons (which are also often discussed in the literature on this topic):
Omitted Variable Bias: The firms (acquirer) in my sample are likely to differ with respect to a number of (unobservable) characteristics. I control for firm, CFO and deal level characteristics which have been found to impact my IVs. Of course, controlling for these will not solve the problem entirely as there will be some sort of unobservable firm-level heterogeneity.
My baseline dataset has (pooled) cross-sectional rather than panel characteristics so that I am not able to apply firm fixed effects (at least as to my understanding) using the entire sample to rule out any differences that differ across firms but is stable across time.
A paper related to mine examines the effect of social connections between the acquirer and target board and has a data structure similar to mine. To further control for any other unobservable or omitted acquirer characteristics, the authors apply a firm fixed effects model and state:
My sample consists of 3,167 deals made by 1,142 acquirers, suggesting that many of the companies in my sample made more than one acquisition in the sample period (1996-2018).
In order to apply a similar analysis to my data I need to consider companies that occur more than once in my sample, i.e. those that made more than one acquisition. Do I need to drop those single acquisition companies in my dataset by hand or will the firm fixed regression command take care of these observations?
Any help regarding the commands to run this kind of analysis is much appreciated.
The second reason is likely to introduce endogeneity into my analysis because of sample selection bias: It's unlikely that CFOs are randomly assigned to companies. I am aware of this problem and may address it in another post.
Thanks
I am trying my best to describe my problem as precise as possible without getting lost in details. That being said, please let me know if something is not clear.
I am working on a paper which examines the effect of a Chief Financial Officers' (CFOs) influence in a firm on certain characteristics of Mergers and Acquisitions (M&A) made by the firm in which the CFO is employed.
Data example showing selected variables whereas each entry represents one acquisition:
Code:
* Example generated by -dataex-. For more info, type help dataex
clear
input long Acq_ID int Deal_Announced str52 Acq_Name str73 Targ_Name double Deal_Value str7 CFO_ID byte CFO_Age float CFO_PaySlice byte CFO_No_Boardseats float CFO_Perc_Own_Dir byte(CFO_No_Deals Acq_Listed) float Acq_TobinsQ
29526 16589 "Sun Microsystems Inc" "Storage Technology Corp" 4261718000 "35644" 57 .12773155 0 .003914469 0 1 1.5520844
430 20362 "Ameris Bancorp" "Jacksonville Bancorp Inc,Jacksonville,FL" 58872000 "335855" 46 .15081033 0 .15556154 0 1 1.0793786
23985 20149 "PCTEL Inc" "Nexgen Wireless Inc" 22500000 "340737" 60 .1666754 0 .11837476 1 1 1.3441194
17580 14955 "Juniper Networks Inc" "Micro Magic Inc" 2.600e+08 "206644" 48 .20138165 0 .02049373 0 1 34.53374
884336 18533 "Green Plains Renewable Energy Inc" "Global Ethanol LLC" 1.187e+08 "540821" 53 .1642229 0 .05355228 0 1 1.0793169
5191 18562 "Brookline Bancorp Inc" "First Ipswich Bancorp,Ipswich, Massachusetts" 18992000 "204279" 67 .1243596 0 .2439337 0 1 1.0355493
2382 16925 "Applied Materials Inc" "Applied Films Corp" 460746000 "330577" 56 .1073884 1 .031921104 1 1 2.5402186
3424 19813 "Balchem Corp" "Performance Chemicals & Ingredients Co" 5.670e+08 "347472" 54 .12672943 0 .01011314 0 1 4.828643
21677 17643 "Navigant Consulting Inc" "Chicago Partners LLC" 73000000 "334379" 57 .04742795 0 .007434811 5 1 1.3638552
9002 19114 "Denbury Resources Inc" "Thompson Field,Fort Bend County,Texas" 3.600e+08 "339938" 45 .1878194 0 .07283341 3 1 1.1047777
4311 16224 "Bio-Rad Laboratories Inc" "MJ GeneWorks Inc" 47000000 "336782" 46 .14681797 0 .00055602356 0 1 1.990272
11505 14710 "Fairchild Semiconductor International Inc" "QT Optoelectronics" 97800000 "106407" 53 .15830627 1 .2185668 1 1 3.134584
10137 15775 "EDO Corp" "Darlington Inc" 28500000 "274644" 56 .14105119 0 0 0 1 1.3925874
27977 16667 "Silicon Laboratories Inc" "Silicon MAGIKE Inc" 16000000 "332855" 51 .18263084 0 .006097347 1 1 4.002823
14967 21396 "Hologic Inc" "Faxitron Bioptics LLC" 85000000 "1391225" 50 .13746095 0 .04871405 0 1 1.916813
4865 16421 "Boston Scientific Corp" "Advanced Stent Technologies Inc" 1.200e+08 "35339" 55 .2055593 2 .01886221 0 1 5.827407
2086 17307 "AMN Healthcare Services Inc" "Rx Pro Health Inc" 17300000 "333848" 50 .2492932 0 .06924654 0 1 2.1377676
27242 14641 "Schlumberger Ltd" "CellNet Data Systems Inc" 2.350e+08 "32959" 49 .0837891 0 .0045382758 0 1 2.594161
17915 15874 "Kimco Realty Corp" "Mid-Atlantic Realty Trust" 446316000 "204730" 44 .14793916 0 .02837317 0 1 1.3454137
1942976 19904 "NV5 Holdings Inc" "Owner's Representative Services Inc" 1400000 "895192" 48 .14182062 0 .008522737 0 1 1.3564024
end
format %tdDD/NN/CCYY Deal_Announced
label values Acq_Listed Acq_Listed1
label def Acq_Listed1 1 "Public", modify
format Deal_Value %15.0fc
Based on management and financial literature I apply four measurements to capture the influence (often referred to as managerial power) of a CFO in the acquiring firm.
These measurements are:
1. CFO_PaySlice: CFOs’ annual compensation (Salary + Bonus) divided by the annual compensation of all top management team (TMT) members
2. CFO_No_Boardseats: Number of board seats a CFO holds outside the acquiring company
3. CFO_Perc_Own_Dir: CFO ownership in the company, which is defined as the number of common shares the CFO holds divided by the total number of shares held by all other TMT members.
4. CFO_No_Deals: Acquisition experience of a CFO, measured as the number of deals a CFO has participated in as a manager or director outside the acquiring company during the ten years prior to each deal.
My dependent variables are M&A related measurements that reflect certain characteristics of a deal, e.g., whether the deal was considered to be a success, the negotiation length of a deal or the premium paid by the acquirer to the target.
I have to admit that it is really hard to make causal inference for at least two reasons (which are also often discussed in the literature on this topic):
Omitted Variable Bias: The firms (acquirer) in my sample are likely to differ with respect to a number of (unobservable) characteristics. I control for firm, CFO and deal level characteristics which have been found to impact my IVs. Of course, controlling for these will not solve the problem entirely as there will be some sort of unobservable firm-level heterogeneity.
My baseline dataset has (pooled) cross-sectional rather than panel characteristics so that I am not able to apply firm fixed effects (at least as to my understanding) using the entire sample to rule out any differences that differ across firms but is stable across time.
A paper related to mine examines the effect of social connections between the acquirer and target board and has a data structure similar to mine. To further control for any other unobservable or omitted acquirer characteristics, the authors apply a firm fixed effects model and state:
This specification will not rule out all remaining omitted variables problems, but it will help control for time-invariant acquirer characteristics. Specifically, we compare the deals in which the acquirer has a board connection to the target with those deals by the same acquirer in which the acquirer has no board connection to the target. Put differently, keeping the identity of the acquirer fixed, we compare the connected and non-connected deals made by the same acquirer. Our sample size reduces significantly to 318 in this specification since we focus only on the deals made by those acquirers which undertake at least one acquisition where they have a board connection to the target.
In order to apply a similar analysis to my data I need to consider companies that occur more than once in my sample, i.e. those that made more than one acquisition. Do I need to drop those single acquisition companies in my dataset by hand or will the firm fixed regression command take care of these observations?
Any help regarding the commands to run this kind of analysis is much appreciated.
The second reason is likely to introduce endogeneity into my analysis because of sample selection bias: It's unlikely that CFOs are randomly assigned to companies. I am aware of this problem and may address it in another post.
Thanks
Comment