The Effects of Mergers

The Effects of Mergers
Burcin Yurtoglu
University of Vienna
Department of Economics
1
Three sets of consequences of mergers
1) They can affect the performance of the merging
firms such as
– Profits, growth rates, markets shares, productivity, ...
2) They can affect industry and aggregate
concentration levels.
3) They can affect social welfare
-
As a result of (1) and (2)
2
Profitability
• Major Studies
– Ravenscraft and Scherer (1987) analyze 6000
mergers between 1950 and 1977.
– Meeks (1977) 1000 mergers after WW II in the UK.
– Gugler, Mueller, Yurtoglu, and Zulehner (2003)
analyze 2753 mergers from around the world.
3
Ravenscraft and Scherer (1987)
• They regressed the profits of individual lines of business
in the years 1975-77 on industry dummies and a variable
that measured the fraction of the line of business that
had been acquired since 1950.
• They also controlled for other aspects of mergers
– Hostile vs friendly
– Market share
– Accounting convention: pooling vs purchase accounting
• The profit rates of the acquired lines of business were
2.82 % below those of non-acquired units
4
A typical regression equation
Profit rate 75-77 = [257 Industry dummies] + 0.68 POOL
- 2.82* PURCH + 0.84 NEW + 1.46 EQUALS
+ 30.15* SHR - 3.65 HOSTILE - 3.77 WHITE
-2.23 OTHER
Adj. R2 = 0.182
N = 2,732
5
GMYZ (2003): Prediction of Sales
S C t n  S G t 1
S Ct n  S Gt 1
S IG t n
S IG t 1
S IG t n
S IG t 1
 S Dt
 S Dt
S ID t n
S ID t
S ID t n
S ID t
 S Dt 2
S ID t n
S ID t 2
 S S t 3
S IS t n
S IS t 3
6
GMYZ (2003): Prediction of Profits
 IG t 
1,t 
n 
C t n
C t n
IG t n
K IG t n

IG t 
1
K IG t 
1
K IG t n
K ID t n
K G t
K D t  ID t , t n
G t 
1
1  IG t 
1, t 
n D t 
K IG t 
K ID t
1
K IG t n
K ID t n
K

K D t  ID t , t n 
G t 1 
D t 
K IG t 1 G t 1 IG t 1, t n
K ID t
K ID t n
K IS t n
K D t 2  ID t 2, t n S t 3 
K S t 3  IS t 3, t n
D t 2 
K ID t 2
K IS t 3
7
GMYZ (2003): Main results
Profits
Years
after
the
merger
Number
of
Mergers
Difference
in Mn $
t+1
2,704
t+2
Sales
p-value
%
Positive
Difference
in Mn $
pvalue
%
Positive
5.91
0.062
57.0%
-214.16
0.000
51.5%
2,274
11.11
0.009
57.2%
-382.81
0.000
49.5%
t+3
1,827
10.79
0.056
54.8%
-549.59
0.000
46.4%
t+4
1,517
19.68
0.007
57.8%
-633.46
0.000
46.3%
t+5
1,250
17.81
0.046
57.6%
-714.04
0.000
44.6%
8
Analysis of variance in year t+5 by country
categories
Country/country group
Profits
Difference in Mn $
t-value
Sales
Difference in Mn $
t-value
Average
17.8
2.00
-714.0
6.63
USA
-0.4
0.33
-16.2
0.70
UK
6.3
0.38
168.3
1.13
Continental Europe
24.5
0.37
47.6
0.55
Japan
-59.4
0.85
-1615.0
1.83
Aus/NZ/Can
-51.4
1.32
-91.4
0.45
Rest of the world
98.1
1.26
432.5
0.63
Adjusted R²
-0.0006
0.0003
Number of Observations
1,250
1,250
9
Analysis of variance in year t+5 in the
manufacturing sector by merger categories
Category
Profits
Difference
in Mn $
t-value
Sales
Difference
in Mn $
t-value
Average
3.1
0.27
-660.0
5.19
Horizontal
38.7
2.07
464.1
2.25
Vertical
-91.4
1.82
-329.1
0.59
Conglomerate
-9.0
1.13
-164.7
1.87
Adjusted R²
Number of
Observations
0.0066
0.0045
702
702
10
Market Power or Efficiency
S 0
S 0
0
0
1
Efficiency
Increase
3
Market Power
Reduction (?)
2
Market Power
Increase
4
Efficiency
Decline
11
S > 0
S < 0
Π > 0
Π < 0
1
3
Small
34.7
17.5
Large
23.4*
12.7*
All
29.1
15.1
2
4
Small
20.4
27.4
Large
34.8*
29.1
All
27.6
28.2
12
Effects on Market Shares
• Mueller (1985, 1986) examined the effects of mergers for
a sample of 209 acquired firms from the 1,000 largest
companies of 1950 in the USA.
• The methodology compared the market shares of firms
acquired between 1950 and 1972 with those of nonacquired firms of similar size in the same industries
• Typical regression equation:
mi 72 = 0.011* + 0.885* mi 50-0.705* Dmi50
N =313
R2 = 0.940
D : a dummy variable for acquired firms.
13
Mergers Effects on Productivity
• Lichtenberg and Siegel (1987)
– between 1972 and 1981 productivity fell in plants
before an ownership change and rose afterward 
spin-offs of plants obtained in previous mergers
• Baldwin (1981)
– similar results for Canada
• McGuckin and Nguyen (1995)
– Plant productivity increases following mergers in the
USA
14
Mergers Effects on Firm-level Employment
• If a merger results in an optimal employment level different from the
total employment of the merging firms, then a profit maximising firm
will adjust its labour force via an active policy of hiring or firing.
• Dynamic models of labour demand place great emphasis on the
costs of this adjustment process.
• In general, the a priori predictions on the effects of mergers on
labour demand are ambiguous:
– A merger may lead to a reduction in output, e.g. because the merger
increased market power or the technologies of the acquiring and
acquired company exhibit increasing returns to scale, and a consecutive
reduction in employment.
– A merger may, on the other hand, lead to an increase in output, e.g.
because the merger significantly increased the efficiency of the
combined firm or led to product improvements and demand shifts.
– Thus, the employment of the combined entity may rise or fall relative to
the sum of the pre-merger employment levels
15
Summary of Empirical Studies on the Employment Effect of mergers
Study
Country
Brown and Medoff
(1988)
USA
Bhagat et al. (1990)
USA
Lichtenberg and
Siegel (1992)
McGuckin et al.
(1995) and
McGuckin and
Nguyen (2000)
USA
USA
Sample and time
period
Form of ownership
change
Large sample of firms "assets only"
acquisitions:
including also very
small, unlisted firms in
Michigan; 1978-1984. "true mergers":
62 takeovers;
hostile takeovers
1984-1986
All ownership
Auxiliary
changes
Establishment Reports
of the 1977 and 1982
economic censuses
Longitudinal Research
Database (LRD);
All ownership
changes
1977-1987
Main results
Wages 5% higher,
employment 5% lower
Wages 4% lower,
employment 2% higher
In 28/62 cases workers
laid off involving 5.7% of
work force
-15.7 % in employment
in auxiliary
establishments, but only
-4.5% in production
establishments
Positive, but
insignificant effect on
employment at firm
level; +16.1% at plant
level
442 mergers;
Conyon, Girma,
Thompson and
Wrigth (2002a, b, c)
UK
related:
-19% in employment
unrelated:
-8% in employment
277 listed firms;
hostile:
-17% in employment
1975-1996
friendly:
-9% in employment;
foreign:
+13% in labour
productivity
16
Gugler and Yurtoglu (2004)
• If mergers are used as a general device to restore a firm’s optimal
employment level, we would expect differential effects depending on
labour market institutions.
• High labour adjustment costs make hiring a worker a somewhat
irreversible decision.
– Therefore, it appears likely that in countries with very rigid labour
markets some firms carry excess labour with them.
– Fewer such firms are expected to exist in countries whose labour
markets allow quick adjustment of the workforce.
– Mergers and acquisitions are an effective means to achieve a desired
restructuring, since the managerial team is likely to be new and
therefore less likely to be committed to upholding past contracts with
stakeholders (Shleifer and Summers, 1988).
– Since Europe has more rigid labour markets than the USA, mergers
may be used as a device to reduce excess labour. Thus, we expect that
the demand for labour is reduced by more after a merger or acquisition
in Europe than in the USA.
17
The labour demand function
Firms are output constrained, have a Cobb-Douglas technology, and face quadratic
adjustment functions.
Firms determine the path of their future employment by
minimizing the expected discounted value of their costs
Lc K
E M
N

Ct
s
t
s
t
s0
b g b gO
P
Q
d

w
L

 Lt s
t
s
t
s
t
s
2
2
e
  K t s
2
2

t
subject to
Qt s g( K t s , L t s )

s
ct : the user cost of capital,
wt : the wage rate
d and e define the quadratic adjustment costs.
E t , , and  denote the mathematical expectation given the information set
available to the firm at period t, the discount rate, and the first-differences operator,
respectively.
18
The first-order conditions (Euler equations) for each input are
F
G
H b
Fd bK
E G
H
g b
g
g b
g
Et 
d Lt 
t s
Lt s d Lt s 
Lt 
wt s 
s 1
s 1 
t
t 
s 1

t s
K t s d K t s 
K t 
ct s 
s 1 
IJ
K

g I
J0
K K

g
0
Lt s

s

s
for L
for K
t
s
Solve this model by working in the neighbourhood of the long-run equilibrium, i.e.,
where e=d=0. The Euler equations can then be written as
F

g
G
Lt
H

g
K t
IJ* wt
K ct
After linearization of g around the long-run equilibrium,

g

g
*
*
*
Qt 
Qt 
Lt 
Lt 
Kt 
Kt
L t
K t
d
i
d
i
*
Q t 0 ,
we obtain given that at the long-run equilibrium Q t 
*
Fw IdL L i.
G
c J
H
K
K t K t 
t
t
*
t
t
19
the optimal path for employment

b g
b
g
 1


Lt Lt 

 L*t s
1 1 
s
s 0
log approximation
bg

b gb g
b
g
c h
Log Lt 
Log Lt 
 1



 Log L*t s
1 1 
s
s 0
*
The expression of the desired levels of employment, L t s , can then be derived from
the solution of the firm's optimisation program without adjustment costs.
In our
context, this depends on the expected production, Q*t , on the expected factor price
b g
*
ratio, wt ct , and on technological progress represented by time dummies, Dt,
Log L
*
t s
 a 1Log Q
*
t s
 
*
w
 a 2Log
c
 a 3,t  s D t  s  a 4   t  s
t s
b g follow an AR(2) process
Assuming that the exogenous factors Qt s and w c
t
s
b gbw cg D f

1 qit 2 q it 1 
1 w c
l it l it 
1
it
2
it 
1
t
t
i


it
20
Differences in Employment Effects of Mergers
USA vs. Europe
Dependent variable: ln(Employment) (t)
Coef
ln(Employment) (t-1)
0.341
ln(Output) (t)
0.347
ln(Output) (t-1)
-0.045
Merger (t)
-0.029
Divest (t)
-0.050
Merger USA (t)
Merger Europe (t)
Divest USA (t)
Divest Europe (t)
Pseudo R²
0.131
Sargan test
0.41
AR(1)
0.00
AR(2)
0.77
z-value Coef z-value
8.61
0.345
8.69
43.41 0.346 43.16
-3.07 -0.047 -3.17
-2.74
-6.21
0.010
0.73
-0.100 -5.39
-0.062 -6.50
-0.022 -1.46
0.130
0.45
0.00
0.79
21
Mergers Effect on Share Prices
Event Study Methodology:
– announcement  new information  changes in share
price reflect the market’s expectations of the effects of
mergers
Problems:
– when does the share price change occur?
– How does one separate it from other events?
 select a control group and assume that the acquiring
firm’s share price would have performed over the
chosen period exactly as the control group
 relative to the control group (start earlier)
22
Time Line for an Event Study
estimationO F
event O F
post event O
F
G
G
G
P
P
window P
window
window
H
H
H
Q
Q
Q
_____________|_____________________|__________ |__________|____________________|______________
To
T1
0
T2
T3
Let  = 0 be the event date
 = T1+1 to  = T2 is the event window
 = T0+1 to  = T1 is the estimation period
 = T2+1 to  = T3 is the post-event period
L1 = T1 - T0 is the length of the estimation period
L2 = T2 - T1 is the length of the event period
23
Capital Asset Pricing Model
Rit   t   t   it   it
Rit : the return on firm i’s shares in period t,
 it : the CAPM’s measure of systematic risk for i in t,
µit : the error term for the equation.
 Use time series observations on Rit to estimate a βit for each firm.
 Use these along with the returns for each firm in a given period
t, say a month, to estimate γt and δt for that month.
 These estimated parameters plus the βit it for firm i are then used
to predict the return on firm i’s shares in t.
ˆ it  ˆ t  ˆ t   it
R
ˆ it , and the actual return
The difference between this predicted return, R
is the error of the prediction
e it  R it  Rˆ it
24
The Market Model
Assumption: a stable linear relation between the market return (R Mt)
and the security return
R jt   j   j R M ,t  
E ( jt )  0
jt
var( jt )   2

j
In applications a broad based stock index is used for the market
portfolio, with the S&P 500 Index, the CRSP Value weighted Index
being popular choices.
The predicted return for a firm for a day in the event period is the
return given by the market model on that day using these estimates.
That is:
 R
 
 j
R
M ,t
jt
j
The abnormal return is then:
AR jt  R jt  E ( R jt )
 R
 
AR jt  R jt  
M ,t
j
j
25
Alternatives
The Factor Mode l
Motivation: reducing the variance of the abnormal
explaining more of the variation of the normal return.
return by
Typically the factors are portfolios of traded securities. The market
model is an example of a factor model. Other multi-factor models
include industry indexes in addition to the market.
Fama-French three-factor mode l:
r t     1 r M ,t   2 SMB t   3 HML t   t
Generally the gains from using multi-factor models for event studies
are limited. The marginal explanatory power of additional factors is
small and hence there is little reduction in the variance of the
abnormal return. The variance reduction will typically be greatest in
samples where firms have a common characteristic, e.g., they are all
concentrated in one market capitalization group or they are all
members of one industry. In these cases the use of multi-factor
models warrants consideration.
26
Findings: The first wave 1972 - 83
• Strong belief in the short horizon (announcement) effects,
i.e., strong belief in the efficiency of capital markets.
• Consequently, several studies ignored post-merger
performance of acquiring companies.
• Studies that ignored post-merger performance
– tended to find small and often insignificant changes in acquirers’
share prices around the announcements.
– The acquirers’ shareholders were judged not to have lost as a
result of the mergers,
– the acquired shareholders were clear winners,
– and thus the studies that ignored the post-merger performance of
acquiring companies concluded that M&As increased total
shareholder wealth.
27
Studies that did report post-merger returns:
• Firth (1980) and Malatesta (1983): the acquiring
companies’ shareholders had suffered
significant losses.
– Firth: all losses occurred in the announcement month
– Malatesta: they occurred over the year following the
mergers.
– They also add up the absolute wealth changes for
both groups of shareholders and find found that the
aggregate losses to the acquiring companies’
shareholders exceeded the gains in wealth of the
targets.
– The remaining studies that reported post-merger
losses for acquiring companies dismissed them as
“surprising” or “puzzling,” or simply ignored them.
28
Findings: The second wave post -1983
• Debates on
– the proper benchmark, and
– the length of the window
• The Proper Benchmark: Estimates using CAPM
imply
e it  R it  Rˆ it
Pre-merger performance
CAPM
Estimates of α
High
+ High
Lower
Low
Low or normal
Higher
29
• The natural choice for a benchmark period is some interval before
the merger announcements.
• However, several studies estimate post-merger abnormal returns
using a post-merger period.
• This choice results in much lower estimates of post-merger losses.
• Example: Magenheim and Mueller (1988)
Benchmark
Cumulative losses to acquirers
over +1 to +12 months
- 36 to - 3 months pre-merger
period
-11.3%
(significant)
post-merger benchmark
-3.2% (insignificant)
30
Long-Horizon Studies:
Agrawal, Jaffe and Mandelker (1992):
–
–
–
–
Returns over five post-merger years
1955-87: -10% significant
negative significant ARs for the 1950s, 1960s and 1980s
insignificantly positive ARs for 1970s
• Estimates by Loderer and Martin (1992) and Higson and
Elliott (1998) are also sensitive to the time period in
which M&As occurred.
• Rau and Vermaelen (1998):
–
–
–
–
2823 acquisitions over 1980-1991 period:
mergers: -4% significant
tender offers: significant and positive
acquirers with high market values relative to their book values:
-17.3% over the 3 –year post-merger period.
• Conclusion:
– The evidence compiled so far is consistent with the idea that
stock market run-ups lead to unsuccessful mergers.
31
The Returns to Acquiring Firms in the USA (1980-2001), GMY (2004)
Window
Month of Acquisition
Period of
Acquisition
All Acquisitions
N
Non-Wave
1804
Wave
1561
Difference
One Year after
Acquisition
Non-Wave
1829
Wave
1695
Difference
Three Years after
Acquisition
Non-Wave
1808
Wave
1647
Difference
Mean
0.11
(0.21)
0.56
(0.25)
-0.45
(0.33)
-4.97
(0.84)
-8.24
(1.02)
3.27b
(1.32)
-22.23
(1.98)
-32.68
(2.11)
10.45a
(2.89)
Median
-0.29
0.13
-0.42
-8.31
-10.97
2.66a
-33.47
-44.58
11.11 a
32
Additional Findings
• Managerial discretion and the gains to
acquirers
– Hubbard and Palia (1995): Managers with small
stakes tend to overpay
• Diversification:
– Diversification mergers produce losses to acquirer
shareholders at the announcement date.
– Diversification is negatively related to returns, Tobin’s
q, and market value of a company. Discount for
diversification is quite large (13%-15%).
– Spin-offs that increased focus produce positive ARs
and improvements in operating performance.
33