Hi,
My question is only partly related to stata but I would really appreciate any help here.
Research goal:
I am researching the effect of the Chief Financial Officers influence in a company on the companys acquisitions. This managerial influence is also referred to as "managerial power" in the management literature. My goal is to research how the managerial power of a CFO impacts certain characteristics of acquisitions made by the company.
Sample:
My sample consists of around 4,300 Acquisitions announced from U.S. companies between 1996 and 2018 made by 1,380 unique acquirors (firms). For each acquisition I have data on the CFO that was in place while the deal was announced.
I implemented four variables that are used for proxying a managers influence in the acquiring company (reflecting the four dimensions of managerial power introduced by Finkelstein (1992)1).
Those variables are:
I am controlling for effects suggested in the literature which include Deal characteristics (Payment method, Deal value...), Acquiror characteristics (Free Cash Flow, Size (Market value)...) as well as for other CFO characteristics (Age, Gender, tenure...).
Example of my data (due to dataex variable limits I only include a subset of variables):
Data preparation:
I prepared the data for panel data analysis using
Question 1:
I did not include a -timevar- as this would reduce my sample due to duplicates (there are acquirors that announced more than one dear in a given year) and I am not planning to use time series commands
Most paper that are related to my research question make use of industry and year fixed effects in their analysis. As to my understanding it's not possible to use the fe approach when the key explanatory variables are constant across time as the fe will wipe out those variables. How is it possible however, that stata calculates coefficients for those variables that are constant over time like for example the gender variable?
I am wondering why there isn't any problem with this code for example:
My empirical model looks like this:
Question 2:
I would really appreciate if the model specification looks alright considering my data structure. I included firm fixed effects (and not industry) because when implementing firm fixed effects it also control for differences across industries as a firm is typically within the same industry across time. Would you agree with this argumentation?
The output when running the regression on my overall sample is:
Question 3:
Accordingly, there doesn't seem to be an effect on either of the power variables on the CAR. Before moving on with my research I just want to make sure whether there could be a fundamental problem with the model causing the potential effects to be insignificant.
If you made it up to this point and read the entire post: Thank you very much!
Kind regards
_____________________________________
1 https://www.jstor.org/stable/256485
My question is only partly related to stata but I would really appreciate any help here.
Research goal:
I am researching the effect of the Chief Financial Officers influence in a company on the companys acquisitions. This managerial influence is also referred to as "managerial power" in the management literature. My goal is to research how the managerial power of a CFO impacts certain characteristics of acquisitions made by the company.
Sample:
My sample consists of around 4,300 Acquisitions announced from U.S. companies between 1996 and 2018 made by 1,380 unique acquirors (firms). For each acquisition I have data on the CFO that was in place while the deal was announced.
I implemented four variables that are used for proxying a managers influence in the acquiring company (reflecting the four dimensions of managerial power introduced by Finkelstein (1992)1).
Those variables are:
- Pay Slice: Compensation of the CFO relative to the sum of compensation of the 5 highest paid managers in the company.
- M&A Experience: Number of Deals the CFO was involved in other companies up to 10 years before the deal announcement.
- Number of Boardseats: Number of Boards the CFO sits on during the deal announcement (excluding the boardseat he may holds in the acquiring firm)
- Ownership: Shares owned by the CFO relative to shares owned by all managers of the company (in percentage)
I am controlling for effects suggested in the literature which include Deal characteristics (Payment method, Deal value...), Acquiror characteristics (Free Cash Flow, Size (Market value)...) as well as for other CFO characteristics (Age, Gender, tenure...).
Example of my data (due to dataex variable limits I only include a subset of variables):
Code:
* Example generated by -dataex-. For more info, type help dataex
clear
input long(Acq_ID CFO_ID) str26 CFO_Name byte(CFO_Age CFO_MBA) float CFO_PaySlice byte CFO_No_Boardseats int CFO_Tenure byte CFO_No_Deals str52 Acq_Name int(Deal_Announced Deal_Year_Ann) double Deal_Value float(CFO_Perc_Own_Dir Acq_CAR_1_1_ES2)
19966 206565 "Jeff Campbell" 61 1 .2097557 2 2638 0 "McKesson Corp" 18800 2011 90000000 12.200355 -.02181801
6908 32837 "Larry Carter" 78 0 .18803322 0 477 0 "Cisco Systems Inc" 13261 1996 4833530000 .9654859 -.00428218
82823 56737 "Bob Dykes" 73 0 .23726204 1 722 6 "Verifone Systems Inc" 18506 2010 9000000 7.022498 -.0419592
1715277 206147 "John Austin" 60 0 .20607075 0 528 0 "The Chefs' Warehouse Inc" 19703 2013 29900000 2.1689453 -.00999988
19839 334420 "Joyce Schuldt" 54 0 .14923082 0 5 0 "Maverick Tube Corp" 16579 2005 1.860e+08 0 .04028475
2621 334012 "Paul Reilly" 65 0 .1490617 0 521 0 "Arrow Electronics Inc" 17078 2006 80000000 75.45046 .010976512
30738 141599 "Brian McKeon" 59 1 .1580198 1 1286 4 "Timberland Co" 16747 2005 82000000 4.217062 .025647774
1979174 482279 "Randy Broaddrick" 44 0 .2631579 0 157 3 "Ring Energy Inc" 19761 2014 6450000 2.2242737 .
25116 321843 "Dennis Klaeser" 64 1 .1742429 1 744 0 "PrivateBancorp Inc" 16540 2005 64000000 .59503466 .06500283
1064229 360075 "James Budge" 56 0 .250058 0 1938 5 "Rovi Corp" 18618 2010 741761000 10.470184 -.006451557
806647 507495 "Cesar Ribeiro" 52 0 .26558557 0 1121 0 "Lincoln Educational Services Corp" 17923 2009 3000000 .8498709 -.0029008
7637 322193 "Gary Haroian" 70 0 .04092294 0 80 0 "Concord Communications Inc" 14629 2000 104391000 12.99716 -.2070821
32498 206690 "Dana Evan" 63 0 .199771 0 1836 1 "Verisign Inc" 16812 2006 30000000 11.331512 .016551858
30437 206292 "Greg Beecher" 64 0 .2313652 1 1292 1 "Teradyne Inc" 18884 2011 5.800e+08 4.894503 .00418798
428 33484 "Tom Freyman" 67 0 .14662687 0 844 0 "Abbott Laboratories" 15910 2003 1.600e+08 5.97643 -.0427234
4081 532517 "Kevin Riley" 62 0 .2105263 0 355 0 "Berkshire Hills Bancorp Inc" 18618 2010 114915000 8.721497 .001181409
29943 321145 "Craig Cohen" 63 0 .3952577 0 782 0 "TALX Corp" 15147 2001 12100000 3.8284714 -.002068691
21553 205616 "Steve Krablin" 71 0 .15972222 0 1536 7 "National-Oilwell Inc" 14685 2000 328136000 1.2829663 .05274526
5860 320196 "Martin Durant III" 69 0 .1481551 0 2119 0 "Carmike Cinemas Inc" 16545 2005 66000000 2.0207245 .
30483 343439 "Joseph Abell III" 67 1 .15521003 0 1773 0 "TETRA Technologies Inc" 16869 2006 50000000 0 .04393317
29526 203470 "Mike Lehman" 72 0 .1679528 0 2029 0 "Sun Microsystems Inc" 14479 1999 545053000 .14703006 -.06867583
21585 60411 "Lewis Holland" 78 0 .12717396 0 701 0 "National Commerce Bancorp, Memphis,Tennessee" 14367 1999 78251000 1.7000953 .028819654
324 34064 "Neal Arnold" 61 1 .09556917 0 197 0 "Fifth Third Bancorp" 14411 1999 2120738000 0 -.06448339
122662 346602 "Kevin Phillips" 61 0 .1666305 0 1450 0 "ManTech International Corp" 18252 2009 2.420e+08 19.83208 .05128356
3169 537867 "Kevin O'Shea" 56 1 .2171365 1 1189 0 "AvalonBay Communities Inc" 21063 2017 76800000 2.75876 -.005469032
1941706 457702 "Mark Wetzel" 63 1 .2186312 0 112 3 "Aviv REIT Inc" 19785 2014 48500000 .05218355 .028736543
937594 44196 "Byron Pollitt Jr" 70 1 .23381294 0 1057 3 "Visa Inc" 18667 2011 1.900e+08 7.633737 .015262982
13057 497919 "Tim Gallagher" 60 1 .1641626 0 1094 0 "Genesee & Wyoming Inc" 17651 2008 97000000 5.313417 -.033886403
14372 36363 "Dave Zwiener" 67 1 .2044825 1 288 0 "The Hartford Financial Services Group Inc" 13803 1997 184722000 6.702297 .015409133
755706 509385 "Judy Bjornaas" 59 1 .28308672 0 700 0 "NCI Inc" 17344 2007 64800000 2.5869734 -.005903121
end
format %tdDD/NN/CCYY Deal_Announced
format Deal_Value %15.0fc
I prepared the data for panel data analysis using
Code:
xtset Acq_ID
xtsum CFO_PaySlice CFO_No_Boardseats CFO_No_Deals CFO_Perc_Own_Dir
Variable | Mean Std. dev. Min Max | Observations
-----------------+--------------------------------------------+----------------
CFO_Pa~e overall | .2030756 .0961471 0 3.606383 | N = 4348
between | .1157044 0 3.606383 | n = 1380
within | .0484867 -.108415 1.262201 | T-bar = 3.15072
| |
CFO_No~s overall | .3199172 .7994125 0 14 | N = 4348
between | .6833354 0 6 | n = 1380
within | .4903563 -2.537226 11.90325 | T-bar = 3.15072
| |
CFO_No~s overall | .9544618 2.452258 0 29 | N = 4348
between | 2.163101 0 21 | n = 1380
within | 1.409748 -10.64554 24.7878 | T-bar = 3.15072
| |
CFO_Pe~r overall | 5.382116 6.925948 0 100 | N = 4348
between | 6.745748 0 100 | n = 1380
within | 4.065572 -30.33086 43.11768 | T-bar = 3.15072
I did not include a -timevar- as this would reduce my sample due to duplicates (there are acquirors that announced more than one dear in a given year) and I am not planning to use time series commands
Most paper that are related to my research question make use of industry and year fixed effects in their analysis. As to my understanding it's not possible to use the fe approach when the key explanatory variables are constant across time as the fe will wipe out those variables. How is it possible however, that stata calculates coefficients for those variables that are constant over time like for example the gender variable?
I am wondering why there isn't any problem with this code for example:
Code:
xtreg Acq_CAR_1_1_ES2 CFO_Gender, fe
Code:
xtreg Acq_CAR_1_1_ES2 CFO_PaySlice CFO_No_Boardseats CFO_No_Deals CFO_Perc_Own_Dir Deal_Controls Acquiror_Controls CFO_Controls i.Deal_Year_Ann, fe (vce cluster Acq_ID)
I would really appreciate if the model specification looks alright considering my data structure. I included firm fixed effects (and not industry) because when implementing firm fixed effects it also control for differences across industries as a firm is typically within the same industry across time. Would you agree with this argumentation?
The output when running the regression on my overall sample is:
Code:
. xtreg Acq_CAR_1_1_ES2 CFO_PaySlice CFO_No_Boardsitze CFO_No_Deals CFO_Perc_Own_Dir CFO_Board CFO_Age CFO_Gender CFO_MBA CFO_CPA CFO_Tenure Deal_Value Targ_Listed Deal_Structure Deal_No_Bidders Deal_Div_FF12 Acq_MktValue A
> cq_Leverage Acq_ROA Acq_Cash_holdings Acq_TobinsQ Acq_FCF Acq_No_Deals i.Deal_Year_Ann, fe vce(cluster Acq_ID)
Fixed-effects (within) regression Number of obs = 2,521
Group variable: Acq_ID Number of groups = 980
R-squared: Obs per group:
Within = 0.0467 min = 1
Between = 0.0063 avg = 2.6
Overall = 0.0220 max = 75
F(39,979) = .
corr(u_i, Xb) = -0.1598 Prob > F = .
(Std. err. adjusted for 980 clusters in Acq_ID)
-----------------------------------------------------------------------------------
| Robust
Acq_CAR_1_1_ES2 | Coefficient std. err. t P>|t| [95% conf. interval]
------------------+----------------------------------------------------------------
CFO_PaySlice | .0036156 .0254389 0.14 0.887 -.0463055 .0535366
CFO_No_Boardsitze | -.0006362 .0025143 -0.25 0.800 -.0055702 .0042978
CFO_No_Deals | -.0000699 .0007644 -0.09 0.927 -.0015699 .0014301
CFO_Perc_Own_Dir | -.0005353 .0002961 -1.81 0.071 -.0011164 .0000459
CFO_Board | .0262056 .0120205 2.18 0.029 .0026167 .0497944
CFO_Age | -.0005674 .0004353 -1.30 0.193 -.0014216 .0002869
CFO_Gender | -.0005629 .0078833 -0.07 0.943 -.0160331 .0149072
CFO_MBA | -.0086922 .0055608 -1.56 0.118 -.0196047 .0022204
CFO_CPA | .0046285 .0050339 0.92 0.358 -.0052499 .0145069
CFO_Tenure | 3.32e-07 1.99e-06 0.17 0.867 -3.57e-06 4.24e-06
Deal_Value | -1.70e-12 6.34e-13 -2.68 0.008 -2.94e-12 -4.53e-13
Targ_Listed | -.0104142 .0054327 -1.92 0.056 -.0210753 .0002469
Deal_Structure | .0006119 .0008242 0.74 0.458 -.0010056 .0022294
Deal_No_Bidders | .0299449 .0194579 1.54 0.124 -.0082391 .0681289
Deal_Div_FF12 | .0023041 .0043431 0.53 0.596 -.0062189 .010827
Acq_MktValue | 5.68e-11 6.87e-11 0.83 0.408 -7.80e-11 1.92e-10
Acq_Leverage | .0235281 .0299002 0.79 0.432 -.0351479 .082204
Acq_ROA | .1052179 .0336337 3.13 0.002 .0392154 .1712204
Acq_Cash_holdings | .0232797 .0233957 1.00 0.320 -.0226318 .0691912
Acq_TobinsQ | -.0010224 .0003921 -2.61 0.009 -.0017919 -.0002528
Acq_FCF | -.0411013 .0447207 -0.92 0.358 -.1288608 .0466582
Acq_No_Deals | -.0001504 .0002632 -0.57 0.568 -.0006669 .0003661
|
Deal_Year_Ann |
2000 | -.0283307 .0143435 -1.98 0.049 -.0564783 -.0001832
2001 | -.0085731 .0176251 -0.49 0.627 -.0431604 .0260142
2002 | -.0059201 .015401 -0.38 0.701 -.0361429 .0243027
2003 | -.0119448 .0164911 -0.72 0.469 -.0443068 .0204173
2004 | -.0040144 .0137251 -0.29 0.770 -.0309485 .0229196
2005 | -.0053451 .0140514 -0.38 0.704 -.0329194 .0222292
2006 | -.0147642 .0145507 -1.01 0.311 -.0433183 .0137899
2007 | .0001967 .0152248 0.01 0.990 -.0296803 .0300736
2008 | -.0152089 .0165615 -0.92 0.359 -.047709 .0172912
2009 | .0012625 .0148922 0.08 0.932 -.0279617 .0304868
2010 | .0052353 .0150384 0.35 0.728 -.024276 .0347466
2011 | .0000944 .0145346 0.01 0.995 -.0284282 .028617
2012 | .0047204 .0153463 0.31 0.758 -.025395 .0348357
2013 | -.0011319 .0164663 -0.07 0.945 -.0334452 .0311814
2014 | .0024371 .0160913 0.15 0.880 -.0291404 .0340146
2015 | -.0097286 .0167224 -0.58 0.561 -.0425445 .0230873
2016 | -.0118784 .0159008 -0.75 0.455 -.0430819 .0193251
2017 | -.0066215 .0158199 -0.42 0.676 -.0376664 .0244234
2018 | -.009364 .0179147 -0.52 0.601 -.0445197 .0257917
|
_cons | -.0052567 .0402274 -0.13 0.896 -.0841987 .0736852
------------------+----------------------------------------------------------------
sigma_u | .06233286
sigma_e | .06280384
rho | .4962363 (fraction of variance due to u_i)
-----------------------------------------------------------------------------------
Accordingly, there doesn't seem to be an effect on either of the power variables on the CAR. Before moving on with my research I just want to make sure whether there could be a fundamental problem with the model causing the potential effects to be insignificant.
If you made it up to this point and read the entire post: Thank you very much!
Kind regards
_____________________________________
1 https://www.jstor.org/stable/256485
Comment