Price and Probability: - PowerPoint Presentation

Slide1

Price and Probability: Decomposing the Takeover Effects of Anti-Takeover Provisions

Vicente Cuñat

The London School of Economics

Mireia Gine

University of Pennsylvania, WRDS & IESE Business School

Maria Guadalupe

INSEAD and CEPR

Slide2

(Internal) Anti-Takeover ProvisionsAnti-Takeover Provisions (ATPs): Charter amendments that explicitly or implicitly delay or preclude hostile takeovers (staggered boards, poison pills, supermajority…)Evidence on their negative effect on firm value.

Cross sectional (Gompers, Ishii,

Metrick

, 2003)

Causal evidence (Cuñat,

Gine

, Guadalupe, 2012)

Value effects

are

indirect evidence of

the effect

of ATPs on

takeovers

Research Question

: Study the

direct effects

of

ATPs

on takeovers. Identify

channels

through which they create/destroy value for firms.

Slide3

Value Creation from ATPs through TakeoversThree channels for ATPs affecting shareholder value: Takeover probability Takeover premium Disciplining effect of a takeover threat

Popular

wisdom (textbook!):

Negative

effect of

ATPs

on

takeover probability

Positive effect on premium (improved bargaining)

Existing Evidence

Limited/contradicting evidence

Challenge to establish causality

Slide4

Research GoalEstablish the causal effect of anti-takeover provisions on value creation through the merger/takeover channelProvide a framework that decomposes the overall value premium into: probability + price+ selection effectsProbability

:

Change on the probability of a takeover due to ATPs

Price:

Effect of ATPs on the premium paid for a

given firm

Selection Effects:

Are the premiums of the additional firms taken over with low ATPs inherently different from those taken over with high ATPs?

Slide5

Summarizing a Takeover Market

f(Y)

Y

Takeover Premium

Density

Probability of

no

takeover

Probability of a takeover

Premium =0

Distribution of observed premiums

Slide6

Summarizing a Takeover Market

f(Y)

Y

Expected premium

conditional

on a takeover

Slide7

Summarizing a Takeover Market

f(Y)

Y

Expected overall gains via takeover premiums

“

unconditional

premium”

Slide8

f(Y)

Y

f(Y)

Y

Twin Populations

Slide9

c

Control

Keep ATP

D=0

Treatment

Drop ATP

D=1

f(Y)

Y

f(Y)

Y

Apply a Randomized Treatment

Treatment (D) = “Drop an Anti Takeover Provision (ATP)”

What can we identify in this setting?

Slide10

c

f(Y)

Y

f(Y)

Y

Effect on Takeover Probability

How does dropping an ATP affect the likelihood of a takeover?

(Probability Effect)

Control

Treatment

Slide11

c

f(Y)

Y

f(Y)

Y

Effect on Unconditional Premium

How does dropping an ATP affect the expected shareholder

gains from future takeovers? (Unconditional Premium)

Control

Treatment

Slide12

c

f(Y)

Y

f(Y)

Y

Effect on Conditional Premium

This is what the literature normally computes, the effect of a treatment on the premiums of the takeovers that we observe

However, this is not very meaningful, because these are different firm populations.

I

t is the composition of

two effects:

Slide13

f(Y)

Y

f(Y)

Y

Causal Effect on Premium

Effect 1- Pure premium effect:

Effect of the treatment on the premium of takeovers that would have happened even under no treatment

Slide14

f(Y)

Y

f(Y)

Y

Causal Effect on Premium

How does dropping an ATP affect the premium

of a given

t

arget? (Causal Effect on

Price

)

Control

Treatment

Slide15

c

f(Y)

Y

f(Y)

Y

Selection Effect on Premium

Effect II - Selection Effect:

Characteristics of the premium of takeovers that would not happen under no treatment

Slide16

c

f(Y)

Y

f(Y)

Y

Selection Effect on Premium

How does dropping an ATP affect the

selection (in terms of premium) of the firms that are taken over?

Control

Treatment

Slide17

Estimating all the Elements: Decomposition Total shareholder gains = ΔY = = Causal effect on Probability

=

E[Y

| D=0 ,

Pop

0

] Δ

P

+ Casual effect on Premium

+ Pr[T=1 |

D=1] Causal Effect

+ Selection Effects +Pr[T=1

| D=1]{ E[Y | D=1, Pop1] - E[Y | D=1 , Pop0] }

Slide18

Estimating all the Elements: Decomposition Total shareholder gains = ΔY =

=

Causal effect on Probability

=

E[Y

| D=0 ,

Pop

0

] Δ

P + Causal effect on Premium

+ Pr[T=1 | D=1] Causal Effect

+ Selection Effects +Pr[T=1 | D=1]{ E[Y | D=1, Pop1] - E[Y | D=1 , Pop0] } Identification

Strategy:Need to solve TWO issues:Exogeneity of the treatment DDeal with selection

Slide19

Estimating all the Elements of the DecompositionExogeneity: Causal Estimate - Unconditional Premium and

Probability

RDD

votes

on

shareholder-sponsored

proposals

to drop ATPs. Extrapolate to the whole

population with a matching model

CIA validated by the RDD design (Angrist and Rokkanen, 2015)Selection: causal effect on premium and selection effect: Bound the best-case and worst-case selection scenario (Lee, 2009).

Slide20

Advance Results ATPs have a negative effect on price, probability and selection Accounting for selection is important (25% to 50% of the effect)No trade-off between probability of a takeover and premium paid

Drivers of the negative effect on premium

Lower competition among bidders (less contested deals)

Worse matching between bidders and target (synergies)

Ambiguous results on bidder’s returns (bargaining)

Slide21

Data DescriptionRiskmetrics + ISS Tapes: 1994-2013Shareholder proposals on ATPs voted on at annual meetingsType of proposal, votes in favor of dropping an ATP

SDC Platinum:

Data on firm

takeovers/mergers

Define takeover within 5 years of a vote

Takeover premium: cumulative return from 4

weeks before announcement until completion

Slide22

Merger Probability

RDD estimate:

9

%

Slide23

Unconditional Premium

RDD estimate: 4%

Slide24

Extrapolation: Angrist and Rokkanen (2015)Can we extrapolate the RDD results to the full population of firms?Running variable, (Votes) - only source of unobserved heterogeneityIf V, randomly assigned: Equivalent to a randomized trialIf not, condition on variables that remove the indirect relationship between vote and outcomes. Conditional Independence Assumption (CIA)

Use the RD setting to test the CIA -

Angrist

and

Rokkanen

(

2015)

Find a model (set of controls) for which the vote becomes irrelevant

Estimate effect independently

of v

otes Matching estimator using the same model validated by the RDD.

Slide25

A Simple ModelVariables that predict mergers and vote outcomesLog Sales (t-1)Total Market Value (t-1)Profit Margin (t-1)Cash & Liquid Assets (t-1)Av. Industry

Tobins’Q

Av. Industry Market Value

% ownership of institutional investors

Firm’s E-index

Plus - Year Dummies

Slide26

Conditional Independence Assumption - Test

Slide27

Propensity Score Estimates

Panel A: Propensity Score Weighting

(1)

(2)

(3)

(4)

Takeover Probability

Unconditional Premium

yes

0.045**

0.046*

2.77***

2.75**

(0.0208)

(0.025)

(1.03)

(1.16)

t stat

2.18

1.82

2.69

2.38

Model

Y

N

Y

N

Obs

2.063

2.063

2.063

2.063

Panel B: Nearest Neighbor Matching with clustering

(1)

(2)

Takeover Probability

Unconditional Premium

yes

0.0344*

2.505***

(0.0209)

(0.910)

Obs

2,296

2,296

Slide28

So Far…Total shareholder gains = ΔY =

=

Causal effect on Probability

=

E[Y | D=0 , Pop

0

] Δ P

+

Causal effect on Premium

+ Pr[T=1 | D=1] Causal

Effect + Selection Effects

+Pr[T=1 | D=1]{ E[Y | D=1, Pop1] - E[Y | D=1 , Pop0] }

Slide29

So Far… 2.7% = 2.7% =

=

1.33%

=

29.6%

*

4.5%

+ Causal effect on Premium

+ 13.5% Causal Effect

+ Selection Effects +13.5% { E[Y | D=1, Pop1] - E[Y | D=1 , Pop0] }

Slide30

Recovering the Causal Premium EffectLee’s Sharp Bounds (Lee, 2009)Assume effect of ATPs on takeover probabilities is monotonic Then: Takeovers under with an ATP are a subset of takeovers without an ATPCreate best an worst case selection scenarios (lower and upper bounds), by the worst/best observations in the sample of takeover without ATP

Slide31

D=0

T=1



10% of firms

Y | T=1

f(Y | T=1)

Trimming: Example

D=1

T=1



14% of firms

Y

| T=1

f(Y | T=1)

28%

additional firms

Slide32

D=0

T=1



10% of firms

Y | T=1

f(Y | T=1)

Trimming: Example

D=1

T=1



14% of firms

Y

| T=1

f(Y | T=1)

28% Highest Y* selected

Y | T=1

f(Y | T=1)

28% Lowest Y* selected

Y | T=1

f(Y

| T=1)

28%

additional firms

Slide33

Trimming: Example

28% Highest Y* selected

Y | T=1

28% Lowest Y* selected

Y | T=1

D=0

T=1



10% of firms

Y

| T=1

f(Y

| T=1)

D=1

T=1



14% of firms

Y | T=1

f(Y | T=1)

Trimmed Populations

f(Y | T=1)

f(Y

| T=1)

Slide34

Trimming: Example

Lower causal bound

Y | T=1

Y | T=1

D=1

Y | T=1

f(Y | T=1)

Y | T=1

f(Y | T=1)

Upper causal bound

f(Y | T=1)

f(Y

| T=1)

T=1



14% of firms

D=0

Slide35

Causal Effect on Premium

(1)

(2)

(3)

(4)

(5)

(6)

 

Premium 4weeks before Announce

.

to Completion

Premium 1 week before Announce

. to CompletionCAR(-5,5) FFM

CAR

(Vote,Ann+1)

Runup (-42,5)

FFM

Runup

(-

42, Completion) FFM

Lower Bound Estimation of the Premium Effect

 

 

yes

0.29

6.29*

6.85***

19.66

3.66

0.59

(4.051)

(3.45)

(2.05)

(21.08)

(3.13)

(4.44)

Upper Bound

Estimation of the Premium Effect

 

 

 

 

yes

5.46**

9.94***

9.44***

42.11**

8.49**

16.02***

(2.87)

(3.22)

(2.08)

(19.65)

(3.26)

(4.21)

 

 

 

Obs

2296

2296

2296

2296

2296

2296

Slide36

Decomposition…Split the observed premium into its causal effect and firm selection

effect

2.7%

=

2.7%

=

=

1.33%

= 29.6% * 4.5%

+ Causal effect on Premium + 13.5% Causal Effect + Selection Effects +13.5% { E[Y | D=1, Pop1

] - E[Y | D=1 , Pop0] }

Slide37

Decomposition…Split the observed premium into its causal effect and firm selection

effect

2.7%

=

2.7%

=

=

1.33%

=

29.6% * 4.5%

+ [ 0.04% , 0.73%] + 13.5% [ 0.3% , 5.4%]+ [ 1.34% , 0.65%] +13.5% [ 10.3% , 5.0%]

Slide38

Decomposition

(1)

(2)

(3)

(4)

Change in Shareholder Value

Premium Effect

Takeover Probability Effect

Selection Effect

Panel A: Lower Bound

Estimation of the Premium

Effect

2.7%

0.04%1.33%

1.34%

(1%)

(49%)

(49%)

Panel B: Upper Bound

Estimation of the Premium Effect

2.7%

0.73%

1.33%

0.65%

(27%)

(49%)

(24%)

Slide39

Drivers of Positive Premium EffectLower ATPs may attract more competitive bidsPotentially better matching between bidder and target (synergies)Better selection of bidders With ATP, the manager decides (balance shareholder value + private benefits of control)Without ATP the market and shareholders decide (maximize shareholder value)Better bargaining position due to more competitors, worse bargaining position due to lower barriers (ambiguous effect)

Slide40

Drivers of Positive Premium Effect (1)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Acquirer Premium

Competition

 

Acquirer Premium

Acquirer CAR(-5,5) FFM

Runup

Acquirer

(-42,5) FFM

Runup

Acquirer (-

42,Comp) FFM

Number of Bidders

Unsolicited Deal

Challenged Deal

Stock

Percent

Lower Bound Estimation

 

 

 

 

 

yes

-8.46**

-4.44***

-6.05**

-3.30

0.15**

0.037

0.11***

-26.87***

(3.19)

(1.35)

(2.86)

(4.44)

(0.06)

(0.03)

(0.03)

(7.22)

Upper Bound Estimation

 

yes

2.38

1.03

5.87**

17.64***

0.26**

0.09

0.17*

-3.15

(2.35)

(1.38)

(3.26)

(3.78)

(0.10)

(0.09)

(0.09)

(5.39)

Obs

2296

2296

2296

2296

2296

2296

2296

2296

Slide41

Drivers of Positive Premium Effect (I1)

(1)

(2)

(3)

(4)

Matching

Same 2Digit SIC

Size Target Rel. to

Acquiror

Total

Synergies

FFM

Total Synergy/ Total Mkt Cap

Lower Bound Estimation

 

yes

0.175**

-1.33**

-306,949

-0.03

(0.06)

(0.51)

(2,428,514)

(0.03)

Upper Bound Estimation

yes

0.23***

-0.69

7,209,041***

0.14***

(0.06)

(0.48)

(1,844,029)

(0.03)

Obs

2296

2296

2296

2296

Slide42

Drivers of Positive Premium EffectMore competitionMore biddersMore challenged dealsMore cash dealsLower/ambiguous effect on acquirer premiumBetter matchingLarger biddersMore strategic biddersMore total synergies

Slide43

ConclusionsAfter rejecting an ATP:Merger probability increases by 4.5% (0.9 % per year)Expected merger unconditional premium increases by 2.7%Decomposing the unconditional premium:Takeover probability effect responsible for 49% of value effectCausal Premium Effect: Between 1% and 26% of value effect

Selection Effect: Between 24% and 49

% of value

effect

Causal Premium Effect explained by

More bidding competition

Better matching between bidder and target

No effect on bidder’s premiu

m

Slide44

ConclusionsWe identify several channels through which ATPs on average destroy valueLower likelihood of a deal happeningWorse selection of targetsWorse selection of bidders The

division of gains

from dropping an ATP seem to accrue almost exclusively to target shareholders. No apparent trade-off between price and probability

Procedure

to estimate the effects of

ATPs causally, beyond the discontinuity

We deal with the inherent selection problems when price and probability are co-determined

Slide45

THANKS!

Slide46

END

Slide47

Heterogeneous Effects: Extrapolation

Merger Probability

Merger Probability

Unconditional Premium

Unconditional Premium

Slide48

Beyond the Discontinuity: CIA

Takeover Probability

Unconditional Premium

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

D=0

D=1

D=0

D=1

 

 

 

 

 

 

 

Vote

-0.00113

0.000615

-0.00349***

-0.00167

-0.0715*

-0.0319

-0.113***

-0.0458

(0.00103)

(0.00105)

(0.00101)

(0.00114)

(0.0385)

(0.0398)

(0.0392)

(0.0446)

 

 

 

 

 

 

 

ln Sales

-0.0151

 

-0.0253

 

-1.491***

 

-0.351

(0.0152)

 

(0.0154)

 

(0.575)

 

(0.602)

Profit Margin

0.216***

 

0.0463

 

4.572

 

0.171

(0.0763)

 

(0.0833)

 

(2.893)

 

(3.260)

Ln Market Value

-0.0176

 

-0.0161

 

-0.572

 

-0.844

(0.0141)

 

(0.0143)

 

(0.536)

 

(0.560)

Cash Liquidity

0.0933

 

0.248**

 

2.724

 

10.47**

(0.115)

 

(0.106)

 

(4.356)

 

(4.133)

Percent Inst. Own.

-0.0142

 

-0.176**

 

2.533

 

-8.051**

(0.0667)

 

(0.0832)

 

(2.529)

 

(3.256)

Av. Ind.

Tobins'Q

0.00691

 

0.0243**

 

0.660

 

0.709

(0.0113)

 

(0.0119)

 

(0.428)

 

(0.466)

Av. Ind. Market Value

0.0648***

 

0.0297**

 

1.387***

 

0.315

(0.0119)

 

(0.0135)

 

(0.451)

 

(0.528)

Entrechment

Index

0.0154*

 

0.00910

 

-0.195

 

0.473

(0.00822)

 

(0.0103)

 

(0.312)

 

(0.404)

Year Dummies

Y

Y

Y

Y

1,151

1,151

1,005

1,005

1,151

1,151

1,005

1,005

R-sq

0.001

0.131

0.012

0.096

0.003

0.109

0.008

0.081

Slide49

Classic RDD: Estimates

Panel A: Probability of becoming a takeover target over the next 5 years

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(9)

(10)

 

Full

+/-10

+/-5

+/-2.5

+/-1.5

poly hl 2

poly hl 3

IK

CCT

 

D =

1

0.0508***

0.0557

0.0853**

0.106*

0.127*

0.122***

0.141***

0.095***

0.104**

(0.0195)

(0.0345)

(0.0412)

(0.0565)

(0.0702)

(0.0421)

(0.0514)

(0.029)

(0.04)

 

Obs

2,807

822

415

231

139

2,807

2,807

2,807

2,807

R-sq/Z

0.048

0.006

0.012

0.017

0.028

0.053

0.053

3.24

2.36

Panel B: Unconditional Premium

 

Full

+/-10

+/-5

+/-2.5

+/-1.5

poly hl 2

poly hl 3

IK

CCT

 

D = 1

2.601***

3.235**

3.852**

5.379**

7.237**

6.571***

6.542***

4.76***

5.36**

(0.792)

(1.266)

(1.801)

(2.641)

(3.151)

(1.758)

(2.267)

(1.43)

(2.21)

 

Obs

2,807

822

415

237

150

2,807

2,807

2,807

2,807

R-sq/Z

0.033

0.011

0.012

0.021

0.039

0.038

0.039

3.30

2.42

Slide50

Literature on Anti-Takeover ProvisionsHowever, so far, limited/contradicting evidence of direct effects of anti-takeover provisions on mergers. Probability of a Takeover (no effect) Comment and Schwert (1995), (negative effect) Bates, Becher & Lemmon (2008

) (positive)

Bange

&

Mazzeo

(2004)

Premiums:

(no effect) Comment

and

Schwert (1995), Bebchuk, Coates, and Subramanian (2002), (heterogeneous effect – delay vs. preclude) Kardyzhanova and Rhodes Kropf (2011), (poison pill vs. board independence) Cotter, Shivdasani, Zenner (1997);

Bange & Mazzeo (2004)

Threat of a takeover and disciplining effect of potential mergers: Martin and McConell (2012), Lel & Miller (2013)

Slide51

Classic RDD Specification ChecksIdentification ChecksThe vote distribution is continuous at the majority thresholdThe observable characteristics of firms are the same on both sides of the discontinuityImplementation probability jumps discretely at the

threshold

Selection Into the Sample

Local analysis: firms belong to S&P1500, have proposals

Firms in our sample tend to be larger than population firms

No selection in terms of profitability