Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Board declassifications and firm value: Have shareholders and boards really destroyed billions in value?
Emiliano M. Catan & Michael KlausnerGCGC - June 2, 2018
0
Background•Over the past 15 years, several hundred publicly-traded firms
destaggered their boards of directors (often in response to shareholder proposals).
2
Background
3
0.0
5.1
.15
.2.2
5.3
.35
.4.4
5.5
.55
.6.6
5
1995 2000 2005 2010 2015Year
Fraction of firms with Staggered Boards
6 5 46
2 31
13
23
30
36
31 32
20
42
29
39
42
31
16
010
2030
40
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
Number of Board Declassifications
Background•Over the past 15 years, several hundred publicly-traded firms
destaggered their boards of directors (often in response to shareholder proposals).• The declassification wave has received overwhelming
support from shareholders in general, and from the big institutional investors in particular.
4
Background• Recent scholarly work (e.g., Cremers, Litov & Sepe (2017),
Cremers & Sepe (2016)) suggests that these declassifications led to economically significant drops in firm value.• On the basis of their empirical findings, Cremers & Sepe propose
1. Amending 14a-8 to make declassification proposals inadmissible.
2. Adopting a staggered board structure as a “quasi-mandatory” rule.
• But is that what the evidence really tells us?
5
• If indeed the declassification wave was so harmful…• Directors should be blamed for being
ignorant/spineless.• It would raise fundamental questions about the
increasingly institution-centric corporate governance system adopted by the US over the past decades.
Why the answer matters…
6
Caveat: spirit of the exercise• We do not necessarily…• endorse the use of Q as the best/single outcome variable to determine the
value implications of governance.• believe that the presence of staggered boards/the adoption of
declassifications is “exogenous”.• However, most of the recent debate on the effect of staggered boards
has done both.• The spirit of the exercise is to show that, even under the (heroic?)
assumptions that (i) destaggerings are plausibly “exogenous” and that (ii) Q is a good proxy for shareholder value, the results do not withstand closer scrutiny.
7
Sample• Collected data on board structure over 1996-2015 for all firms ever in
S&P1500 (over 2200 firms, 28K firm-years, after excluding financials, utilities, firms with dual-class shares).• Matched with:• Compustat-CRSP for accounting data• CRSP for stock returns• Shark Repellent for precatory proposals
8
Results 1
Firm FEYear FE “Years since Public” FE
6.5% of average Q
9
(or around $350B in value)
0.0
5.1
.15
.2.2
5.3
.35
.4.4
5.5
.55
.6.6
5
1995 2000 2005 2010 2015Year
Fraction of firms with Staggered Boards
Potential source of concern
10
Potential source of concern
11
Potential source of concern
12
0.1
.2.3
.4
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
Year
Very Large Large Small
Fraction of firms with staggered boards targeted by SH proposals
1.5
22.
53
3.5
1995 2000 2005 2010 2015Year
Very Large Large Small
Sample: firms that always had annual boardAverage value of Q
Potential source of concern
13
11.
52
2.5
33.
5
1995 2000 2005 2010 2015Year
Very Large Large Small
Average value of Q
Potential source of concern
14
1.5
22.
53
3.5
1995 2000 2005 2010 2015Year
Very Large & Staggered Very Large & AnnualLarge & Staggered Large & AnnualSmall & Staggered Small & Annual
Average Value of Q
11.
52
2.5
33.
5
1995 2000 2005 2010 2015Year
Very Large Large Small
Average value of Q
Potential source of concern
1 5
1.5
22.
53
3.5
1995 2000 2005 2010 2015Year
Very Large & Staggered Very Large & AnnualLarge & Staggered Large & AnnualSmall & Staggered Small & Annual
Average Value of Q1.
52
2.5
33.
5
1995 2000 2005 2010 2015Year
Very Large & Staggered Very Large & AnnualLarge & Staggered Large & AnnualSmall & Staggered Small & Annual
Average value of Q
~10% of the sample~20% of the sample~70% of the sample
If our conjecture is right…
•Naïve regression should suggest that most of the ostensible effect of declassifications is driven by the very large firms (and show substantial pre-treatment trends).• That statistical association between staggered boards
and Q should be attenuated in a regression that compares apples to apples.
1 7
Results 2 (naïve)
18
Results 2 (naïve)
1 9
Sum=0.156p-value<.05(or 7% of firm value)
Sum=0.62p-value <.001(or 29% of firm value)$ 1 trillion!!!
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Very Large Firms
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Large Firms
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Small Firms
Tobin's Q - Within-firm Dynamics Around Declassification
Lags & Leads (naïve)
20
Results 2 (apples-to-apples)
21
Results 2 (apples-to-apples)
22
Lags & Leads (apples-to-apples)
23
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Very Large Firms
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Large Firms
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Small Firms
Tobin's Q - Within-firm Dynamics Around Declassification-1
-.50
.51
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Very Large Firms
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Large Firms
-1-.5
0.5
1
-6-5-4-3-2-1 0 1 2 3 4 5 6Years around destaggering
Small Firms
Tobin's Q - Within-firm Dynamics Around Declassification
Event study: management declassification proposals
• Event date: date of filing of proxy statement• Model: four factor
2 4
Event study: first precatory declassification proposal targeting firm• Event date: date of filing of proxy statement• Model: four factor
2 5
Robustness + Additional Details
• Results are robust:•More accurate construction of size-buckets by using
“stacked cohort approach” (Gormley & Matsa, RFS 2011)•NN-matching• Calendar-time portfolio analysis
2 6
Conclusion• The sky is probably not falling.•We should have some healthy skepticism about recent
claims suggesting that declassifications destroyed hundreds of billions of dollars in value.•Methodologically, scholars running panel regressions where Q is the outcome variable should be cautious about differential secular trends in stock prices.
Additional Materials
Do results depend on when the size buckets were formed?
• Cohort-based approach (Gormley & Matsa, RFS 2011)• For each year c, we could separately estimate the effect of
declassifications of firm value for firms that destaggered during c.• Treatment firms: firms that destaggered in year c• Control firm-years: firms that were part of the sample in c and c-1 and• Never switched board structure (all firm-year observations)• Switched board structure at some year c’>c (firm-year observations until c’-1)
• We sort treatment and control firms into size buckets according to their decile of market capitalization as of c-1• Call all those firm-year observations “cohort c”
2 9
Do results depend on when the size buckets were formed?
• Although we could estimate one separate estimate reflecting the effect of declassifications for firms in cohort c, we want to aggregate all those estimates into one.• To do that, we form a database that ”stacks” the different cohorts,
and estimate
3 0
Not subindexed by c
Subindexed by c
Do results depend on when the size buckets were formed?
31
Do results depend on when the size buckets were formed?
32
Do results depend on when the size buckets were formed?
33
Another robustness check: N-N matching• For every firm that destaggered, we look (with replacement) for the
nearest neighbor that within the same 2-digit SIC industry that did not declassify in that year or any earlier year along the following dimensions:• Qt-1
• Qt-2
• Qt-3
• Qt-4
• Qt-5
• Market capt-1
• (R&D/Sales)t-1
• Firm age
• We follow both members of the matched pair between year -5 and year +5 surrounding the destaggering (or pseudo-destaggering).• If either firm drops out of the sample (or the control firm destaggers), the
matched pair is also dropped at that point in time. 3 4
Some examplesTreatment Matched control Cohort
RO B ERT H A LF IN T L IN C H EN RY (JA C K ) & A SSO C IAT ES 2 0 0 3
B R ISTO L-M Y E RS SQ U IB B CO SC H ER IN G -P LO U G H 2 0 0 3
CO C A -CO LA CO P E P SICO IN C 2 0 0 3
O M N ICO M G RO U P ELEC T RO N IC D ATA SYST EM S CO R P 2 0 0 3
G R EAT LA K ES C H EM IC A L CO R P FER RO CO R P 2 0 0 3
H A SB RO IN C K ID B R A N D S IN C 2 0 0 3
T EN ET H EA LT H C A R E CO R P H EA LT H SO U T H CO R P 2 0 0 3
P FIZE R IN C LILLY (ELI) & CO 2 0 0 3
R EEB O K IN T ER N AT IO N A L LT D A P TA RG RO U P IN C 2 0 0 3
D E LL T EC H N O LO G IES IN C N ETA P P IN C 2 0 0 3
M ID W AY G A M ES IN C A N SYS IN C 2 0 0 3
B ELLSO U T H CO R P D IR EC T V 2 0 0 4
D O W C H E M IC A L D U P O N T (E I) D E N E M O U RS 2 0 0 4
ED O CO R P C U B IC CO R P 2 0 0 4
FED EX CO R P SO U T H W EST A IR LIN ES 2 0 0 4
G ER B ER SC IEN T IF IC IN C M ILA C RO N IN C 2 0 0 4
STA RW O O D H O T E LS& R ESO RTS W R LD H ILTO N W O R LD W ID E H O LD IN G S 2 0 0 4
M E RC K & CO A B B O T T LA B O R ATO R IES 2 0 0 4
SA FE W AY IN C K RO G E R CO 2 0 0 4
AT & T IN C V ER IZO N CO M M U N IC AT IO N S IN C 2 0 0 43 5
NN-Matching
37
910
1112
Lagg
ed L
og(M
ktca
p) fo
r Tre
atm
ent fi
rms
7 8 9 10 11 12Lagged Log(Mktcap) for Control firms
3 8
910
1112
Lagg
ed L
og(M
ktca
p) fo
r Tre
atm
ent fi
rms
7 8 9 10 11 12Lagged Log(Mktcap) for Control firms
Pfizer & Eli Lilly
Potential source of concern
39
Have management/SHs reacted to the recent findings?
40
82.93 82.7378.04 78.76
74.75 75.91 75.53 76.300
2040
6080
2010 2011 2012 2013 2014 2015 2016 2017
Percent of outstanding shares voting 'for'Management Proposals to Declassify
97.73 97.63 97.11 97.01 97.51 96.60 96.55 97.360
2040
6080
100
2010 2011 2012 2013 2014 2015 2016 2017
Percent of votes cast voting 'for'Management Proposals to Declassify
Puzzle: what is going on?
41
Did declassifications affect R&D-intensive firms more intensely?
4 2
1 if firm had R&D/Sales in top quartile as of 2002 (or the closest year available)
R&D Intensive
Low RD
Small
Large
VeryLarge X
R&D Intensive
Low RD
Small
Large
VeryLarge X
R&D Intensive
Low RD
Small
Large
VeryLarge X
Are these results also confounded by differential secular trends in value?
43
Are these results also confounded by differential secular trends in value?
44
Staggered boards & R&D intensity: comparing apples to apples
45
R&D Intensive
Low RD
Small
Large
VeryLarge X
R&D Intensive
Low RD
Small
Large
VeryLarge X