Tables
Year

Number of companies

Friendly takeovers

Failed friendly bids

Hostile takeovers

Failed hostile bids

Bid targets overall

Bankruptcies

1988

579

26

0

13

5

44

2

1989

550

26

1

10

6

43

10

1990

526

21

3

7

3

34

11

1991

508

19

0

2

3

24

8

1992

495

7

0

3

3

13

6

1993

483

6

1

0

0

7

1

1994

462

6

0

2

2

10

3

1995

438

12

0

6

1

19

1

1996

59

6

0

1

0

7

0

Totals


129

5

44

23

201

42

Notes:

Observations on accounting years are allocated to the current calendar year if the accounting year ends in JulyDecember, and to the preceding calendar year if the accounting year ends in JanuaryJune. See also the Data Appendix for a description of the ‘year’ variable.

At time of sampling, only 59 companies had company accounts reported by Datastream for 1996.
Table 2: Takeover likelihood model: Friendly versus Hostile Targets
Marginal effects from multinomial logit regression.
Sample period is 198996.
Number of observations = 4100.
^{2} (26) = 146.6; PValue= 0.0000. Log Likelihood = 669.1
Likelihood Ratio Index = 0.1186

Friendly Targets

Hostile Targets

Variable

Marginal effect

Standard error

PValue

Marginal effect

Standard error

PValue

Log Real Capital Stock

.0024

.0012

0.049

.00051

.00054

0.34

ReturnonSales

.0312

.0233

0.18

.0099

.0122

0.41

Q

.0020

.00081

0.01

.0017

.00062

0.006

Sectoral Q

.0015

.0013

0.25

.00020

.00072

0.77

Leverage

.0055

.0028

0.05

.0012

.0036

0.73

Sectoral Leverage

.0096

.0130

0.46

.0038

.0059

0.52

Income Gearing

.00025

.00094

0.78

.000015

.000036

0.67

Sectoral Income Gearing

.00032

.00095

0.73

.00010

.000048

0.03

Takeover Rumours

.0074

.0029

0.01

.0043

.0012

0.00

Age 15 Years Dummy

.0530

.0080

0.00

.0094

.0048

0.05

Age 69 Years Dummy

.0238

.0047

0.00

.0075

.0057

0.19

Aggregate Takeover Activity

.00020

.000071

0.004

.000069

.000035

0.05

Real GDP Growth

.0013

.00083

0.10

.001163

.0004459

0.009

Notes:

Time dummies are excluded.

Industry dummies are insignificant and thus excluded.

Regressors scaled so that means lie on the unit interval. Marginal effects evaluated at means of variables.

^{2} statistic is for a Wald test of the null hypothesis that the marginal effects are jointly insignificantly different from zero.

The Likelihood Ratio Index—which is also known as McFadden's (pseudo) Rsquared—has been recommended by CameronWindmeijer (1997) as a measure of goodnessoffit for the logit model.
Table 3: Summary Statistics on Actual Sample Frequency and Predicted Probabilities
Sample period 198996. 4100 observations.
Probability

Mean

Std. Dev.




Sample Frequency of Alltarget takeover bids

.0380

.1913

Recursive Alltarget Predicted Probabilities

.0391

.0544




Sample Frequency of Friendlytarget takeover bids

.0263

.1601

Recursive Friendlytarget Predicted Probabilities

.0260

.0399




Sample Frequency of Hostiletarget takeover bids

.0117

.1075

Recursive Hostiletarget Predicted Probabilities

.0130

.0290

Table 4: Prediction Tables. Recursive Model. Hostile targets. Table 4.1: Hostiletarget predicted probabilities for 1989

Actual Bids in Subsequent Year

ProbabilityRanked Observations

Top

Bottom

Top/Bottom

10

3

0


20

6

0


30

7

0


40

7

0


50

7

0


100

9

0


200

12

0


290

14

2

7

Total Number of observations: 579
Total number of hostile bids: 16
Table 4.2: Hostile predicted probabilities yearbyyear
Year of bid

Number of observations in preceding year

Number of actual bids

Number of bids in top half of ranked observations

Number of bids in bottom half of ranked observations

Top/bottom

1989

579

16

14

2

7

1990

550

10

8

2

4

1991

526

5

5

0

—

1992

508

6

4

2

2

1993

495

0

0

0

—

1994

483

4

2

2

1

1995

462

7

6

1

6

1996

438

1

0

1

0

Table 5: Production Function Model
Dependent variable is y_{it} (log real sales).
Sample period is 1989 to 1996.
551 companies; 3413 observations.
Independent Variable

Col. (1)

Col. (2)

Col. (3)

Col. (4)

Col. (5)

y_{it1}

0.5993
(0.1042)^{a}

0.7935
(0.0536)^{a}

0.6716
(0.0766)^{a}

0.8656
(0.0487)^{a}

0.6630
(0.0798)^{a}

l_{it}

0.7429
(0.1044)^{a}

0.7675
(0.0816)^{a}

0.7236
(0.1009)^{a}

0.7555
(0.0929)^{a}

0.7654
(0.0919)^{a}

l_{it1}

0.5723
(0.1147)^{a}

0.6066
(0.0846)^{a}

0.5540
(0.1192)^{a}

0.6409
(0.1028)^{a}

0.5811
(0.1056)^{a}

k_{it}

0.3495
(0.1344)^{a}

0.1793
(0.1078)^{c}

0.3241
(0.1301)^{b}

0.1588
(0.1210)

0.2635
(0.1191)^{b}

k_{it1}

0.2233
(0.1252)^{c}

0.1397
(0.1014)

0.2096
(0.1204)^{c}

0.1409
(0.1099)

0.1792
(0.1145)^{c}

p^{F}_{it1}


0.9275
(0.4674)^{b}


0.7519
(0.4779)


p^{H}_{it1}



0.1369
(0.2436)

0.0858
(0.2723)


p^{A}_{it1}





0.4747
(0.2108)^{b}

m1

0.00

0.00

0.00

0.00

0.00

m2

0.86

0.53

0.73

0.48

0.78

Sargan

0.81

0.17

0.54

0.09

0.58

Notes:

Time dummies are included in all specifications. Industry dummies (in the levels equations) are insignificant and thus excluded.

Predicted probabilities generated by a recursivelyestimated takeover likelihood model.

All equations estimated with onestep GMM system estimator. Instruments for equations in firstdifferences are y_{it2}, l_{it2}, k_{it2}, p_{it2} and further lags. Instruments for equations in levels are l_{it1}, k_{it1} and p_{it1}. Instruments y_{it1} are rejected by Sargan test, and thus excluded.

In parentheses are asymptotic standard errors robust to general crosssection and timeseries heteroskedasticity. The superscript ‘a’ indicates that the coefficient is significantly different from zero at the 0.01 level, ‘b’ at the 0.05 level and ‘c’ at the 0.10 level.

m1 and m2 are test statistics, distributed standard normal, for first and secondorder serial correlation in the firstdifferenced residuals. Sargan statistic is that for the corresponding twostep GMM estimator, distributed chisquared. Pvalues are reported.
Table 6: Production Function Model: Robustness
Dependent variable is y_{it} (log real sales).
Sample period is 1989 to 1996.
551 companies; 3413 observations.
Independent Variable

Col. (1)

Col. (2)

Col. (3)

Col. (4)

Col. (5)

y_{it1}

0.6814
(0.0744)^{a}

0.6395
(0.0765)^{a}

0.7134
(0.0669)^{a}

0.6319
(0.0855)^{a}

0.8292
(0.0594)^{a}

l_{it}

0.7879
(0.0805)^{a}

0.7685
(0.0888)^{a}

0.7671
(0.0890)^{a}

0.7289
(0.0959)^{a}

0.7779
(0.0836)^{a}

l_{it1}

0.5823
(0.0965)^{a}

0.5709
(0.1025)^{a}

0.5877
(0.1082)^{a}

0.5240
(0.1088)^{a}

0.6361
(0.1021)^{a}

k_{it}

0.1521
(0.1092)

0.2624
(0.1118)^{b}

0.2384
(0.1149)^{b}

0.2562
(0.1187)^{b}

0.0893
(0.0894)

k_{it1}

0.0907
(0.1161)

0.1691
(0.1079)

0.1538
(0.1109)

0.1584
(0.1156)

0.0546
(0.0911)

p^{A}_{it1}

0.3742
(0.2119)^{c}

0.4618
(0.2061)^{b}

0.7212
(0.2993)^{b}

0.4140
(0.2063)^{b}

0.5339
(0.2775)^{b}

Q_{it1}

0.0017
(0.0031)




0.000025
(0.0037)

LEV_{it1}


0.0246
(0.0081)^{a}



0.0302
(0.0104)^{a}

Rum_{it1}



0.0209
(0.0133)


0.0164
(0.0122)

Age_{it1}




0.0019
(0.0019)

0.0004
(0.0012)

m1

0.00

0.00

0.00

0.00

0.00

m2

0.65

0.76

0.72

0.70

0.46

Sargan

0.61

0.61

0.50

0.59

0.23

Notes:

Time dummies are included in all specifications. Industry dummies (in the levels equations) are insignificant and thus excluded.

Predicted probabilities taken from a recursivelyestimated takeover likelihood model.

All equations estimated with onestep GMM system estimator. All independent variables except for age (which is treated as exogenous) are instrumented. Instruments for equations in firstdifferences are as in Table 4.1, with additional regressors instrumented with twice and furtherlags. Instruments for equations in levels are the lagged firstdifferences of the included independent variables, except for y_{it1} and LEV_{it1}, both of which are rejected by the Sargan test.

In parentheses are asymptotic standard errors robust to general crosssection and timeseries heteroskedasticity. The superscript ‘a’ indicates that the coefficient is significantly different from zero at the 0.01 level, ‘b’ at the 0.05 level and ‘c’ at the 0.10 level.

m1 and m2 are test statistics, distributed standard normal, for first and secondorder serial correlation in the firstdifferenced residuals. Sargan statistic is that for the corresponding twostep GMM estimator, distributed chisquared. Pvalues are reported.
Dependent variable is I_{t}/K_{it1} (Investment rate)
Sample period is 1990 to 1996
494 companies; 2697 observations.
Independent Variable

Col. (1)

Col. (2)

Col. (3)

Col. (4)

I_{t1}/K_{it2}

0.0482
(0.0535)

0.0598
(0.0492)

0.0487
(0.0524)

0.0383
(0.0489)

y_{it}

0.1938
(0.0476)^{a}

0.2338
(0.0501)^{a}

0.1800
(0.0412)^{a}

0.2210
(0.0511)^{a}

y_{it1}

0.0894
(0.0226)^{a}

0.0921
(0.0247)^{a}

0.0945
(0.0210)^{a}

0.0866
(0.0247)^{a}

(k_{it2} y_{it2})

0.0786
(0.0205)^{a}

0.0835
(0.0193)^{a}

0.0803
(0.0205)^{a}

0.0701
(0.0175)^{a}

y_{it2}

0.0110
(0.0062)^{c}

0.0293
(0.0104)^{a}

0.0104
(0.0062)^{c}

0.0321
(0.0106)^{a}

C_{t1}/K_{it2}

0.0175
(0.0111)

0.0231
(0.0122)^{c}

0.0150
(0.0105)

0.0219
(0.0122)^{c}

p^{F}_{it1}

0.9586
(0.2950)^{a}


1.0569
(0.3313)^{a}


p^{H}_{it2}


0.2045
(0.1137)^{c}

0.3019
(0.1278)^{b}


p^{A}_{it1}




0.2340
(0.1239)^{b}

p^{A}_{it2}




0.1275
(0.0905)

Wald on probabilities



0.003

0.14

m1

0.00

0.00

0.00

0.00

m2

0.30

0.47

0.25

0.00

Sargan

0.15

0.10

0.13

0.21

Notes:

Time dummies are included in all specifications. Industry dummies (in the levels equations) are insignificant and thus excluded.

Predicted probabilities taken from a recursivelyestimated takeover likelihood model.

All equations estimated with onestep GMM system estimator. Instruments for equations in first differences are (I_{t2}/K_{it3}), (k_{it2} y_{it2}), y_{it2}, C_{t2}/K_{it3}, p_{it2} and further lags. Instruments for equations in levels are (I_{t1}/K_{it2}), y_{it1} and p_{it1}.

In parentheses are asymptotic standard errors robust to general crosssection and timeseries heteroskedasticity. The superscript ‘a’ indicates that the coefficient is significantly different from zero at the 0.01 level, ‘b’ at the 0.05 level and ‘c’ at the 0.10 level.

Wald test applies if there are two probabilities included in the model, and is for the null hypothesis that the probabilities are jointly insignificantly different from zero. Pvalues are reported.

m1 and m2 are test statistics, distributed standard normal, for first and secondorder serial correlation in the firstdifferenced residuals. Sargan statistic is that for the corresponding twostep GMM estimator, distributed chisquared. Pvalues are reported.
Table 8: Investment Model: Robustness
Dependent variable is I_{t}/K_{it1} (Investment rate)
Sample period is 1990 to 1996
494 companies; 2697 observations.
Independent Variable

Col. (1)

Col. (2)

Col. (3)

Col. (4)

Col. (5)

I_{t1}/K_{it2}

0.0297
(0.0503)

0.0745
(0.0530)

0.0607
(0.0528)

0.0556
(0.0559)

0.0195
(0.0476)

y_{it}

0.1577
(0.0373)^{a}

0.1891
(0.0399)^{a}

0.1891
(0.0389)^{a}

0.1799
(0.0412)^{a}

0.1735
(0.0345)^{a}

y_{it1}

0.0842
(0.0207)^{a}

0.1051
(0.0219)^{a}

0.0950
(0.0211)^{a}

0.0998
(0.0251)^{a}

0.0729
(0.0199)^{a}

(k_{it2} y_{it2})

0.0687
(0.0201)^{a}

0.0923
(0.0220)^{a}

0.0855
(0.0213)^{a}

0.0841
(0.0233)^{a}

0.0555
(0.0179)^{a}

y_{it2}

0.0092
(0.0064)

0.0116
(0.0065)^{c}

0.0088
(0.0062)

0.0096
(0.0063)

0.0087
(0.0039)^{b}

C_{t1}/K_{it2}

0.0026
(0.0075)

0.0032
(0.0090)

0.0148
(0.0106)

0.0146
(0.0105)

0.0030
(0.0091)

p^{F}_{it1}

0.8126
(0.2799)^{a}

1.0974
(0.3481)^{a}

1.1867
(0.4141)^{a}

1.0252
(0.3119)^{a}

0.7602
(0.3398)^{b}

p^{H}_{it2}

0.3034
(0.1237)^{b}

0.3018
(0.1259)^{b}

0.2591
(0.1286)^{b}

0.3011
(0.1276)^{b}

0.2818
(0.1170)^{b}

Q_{it1}

0.0052
(0.0021)^{b}




0.0031
(0.0022)

Q_{it2}

0.0013
(0.0010)




0.0009
(0.0015)

LEV_{it1}


0.0031
(0.0047)



0.0042
(0.0067)

LEV_{it2}


0.0054
(0.0045)



0.0032
(0.0056)

Rum_{it1}



0.0095
(0.0063)


0.0091
(0.0053)^{c}

Rum_{it2}



0.0009
(0.0039)


0.0038
(0.0038)

Age_{it1}




0.0004
(0.0006)

0.0003
(0.0004)

Wald on p^{F}_{it1} and p^{H}_{it2}

0.005

0.004

0.01

0.003

0.01

m1

0.00

0.00

0.00

0.00

0.00

m2

0.30

0.20

0.25

0.23

0.23

Sargan

0.17

0.12

0.08

0.13

0.02

Notes:

Time dummies are included in all specifications. Industry dummies (in the levels equations) are insignificant and thus excluded.

Predicted probabilities taken from a recursivelyestimated takeover likelihood model.

All equations estimated with onestep GMM system estimator. All independent variables except for age (which is treated as exogenous) are instrumented. Instruments for equations in firstdifferences are as in Table 4.3, with additional regressors instrumented with twice and furtherlags. Instruments for equations in levels are as in Table 4.3; the Sargan test rejects Q_{it1}, LEV_{it1} and Rum_{it1}.

In parentheses are asymptotic standard errors robust to general crosssection and timeseries heteroskedasticity. The superscript ‘a’ indicates that the coefficient is significantly different from zero at the 0.01 level, ‘b’ at the 0.05 level and ‘c’ at the 0.10 level.

Wald test is for the null hypothesis p^{F}_{it1} and p^{H}_{it2 }are jointly insignificantly different from zero. Pvalues are reported.

m1 and m2 are test statistics, distributed standard normal, for first and secondorder serial correlation in the firstdifferenced residuals. Sargan statistic is that for the corresponding twostep GMM estimator, distributed chisquared. Pvalues are reported.
Table 9: Dividend Model
Dependent variable is (D/Y)_{it} (Dividendtosales ratio)
Sample period is 1990 to 1996
510 companies; 2821 observations.
Independent Variable

Col. (1)

Col. (2)

Col. (3)

Col. (4)

Col. (5)

(D/Y)_{it1}

0.3063
(0.1753)^{c}

0.2990
(0.1778)^{c}

0.3236
(0.1723)^{c}

0.2962
(0.1645)^{c}

0.3076
(0.1507)^{b}

(/Y)_{it}

0.0685
(0.0259)^{a}

0.0651
(0.0264)^{a}

0.0833
(0.0253)^{a}

0.0581
(0.0300)^{b}

0.0714
(0.0315)^{b}

Age_{it}

0.00064
(0.00026)^{b}

0.00068
(0.00028)^{b}

0.00046
(0.00021)^{b}

0.00056
(0.00024)^{b}

0.00041
(0.00017)^{b}

p^{A}_{it}

0.2168
(0.1484)

0.1110
(0.1164)




p^{A}_{it1}

0.0289
(0.0236)





p^{H}_{it}



0.5252
(0.2128)^{a}


0.3270
(0.1874)^{c}

p^{F}_{it1}




0.0720
(0.0462)

0.0519
(0.0396)

Wald on probabilities

0.31




0.07

m1

0.04

0.04

0.01

0.04

0.02

m2

0.17

0.13

0.12

0.20

0.12

Sargan

0.52

0.51

0.58

0.39

0.39

Notes:

Time dummies are included in all specifications. Industry dummies (in the levels equations) are insignificant and thus excluded.

Predicted probabilities taken from a recursivelyestimated takeover likelihood model.

All equations estimated with onestep GMM system estimator. Instruments for equations in first differences are (D/Y)_{it2}, (/Y)_{it2}, p_{it2} and further lags; age is treated as exogenous. Instruments for equations in levels are (/Y)_{it1} and p_{it1}. Sargan test rejects (D/Y)_{it1}.

In parentheses are asymptotic standard errors robust to general crosssection and timeseries heteroskedasticity. The superscript ‘a’ indicates that the coefficient is significantly different from zero at the 0.01 level, ‘b’ at the 0.05 level and ‘c’ at the 0.10 level.

Wald test applies if there are two probabilities included in the model, and is for the null hypothesis that the probabilities are jointly insignificantly different from zero. Pvalues are reported.

m1 and m2 are test statistics, distributed standard normal, for first and secondorder serial correlation in the firstdifferenced residuals. Sargan statistic is that for the corresponding twostep GMM estimator, distributed chisquared. Pvalues are reported.
Table 10: Dividend Model: Robustness
Dependent variable is (D/Y)_{it} (Dividendtosales ratio)
Sample period is 1990 to 1996
510 companies; 2821 observations.
Independent Variable

Col. (1)

Col. (2)

Col. (3)

Col. (4)

(D/Y)_{it1}

0.2651
(0.1468)^{c}

0.3271
(0.1601)^{b}

0.3943
(0.1651)^{b}

0.3284
(0.1307)^{a}

(/Y)_{it}

0.0821
(0.0244)^{a}

0.0758
(0.0230)^{a}

0.0712
(0.0182)^{a}

0.0656
(0.0164)^{a}

Age_{it}

0.00051
(0.00021)^{b}

0.00046
(0.00020)^{b}

0.00022
(0.00011)^{b}

0.00024
(0.00011)^{b}

p^{H}_{it}

0.6163
(0.2056)^{a}

0.4842
(0.1988)^{a}

0.3608
(0.1373)^{a}

0.4304
(0.1424)^{a}

Q_{it}

0.00026
(0.00034)



0.00023
(0.00028)

LEV_{it}


0.00051
(0.00096)


0.0010
(0.0010)

Rum_{it}



0.00096
(0.00051)^{c}

0.0011
(0.00054)^{b}

m1

0.01

0.02

0.02

0.02

m2

0.13

0.12

0.47

0.65

Sargan

0.31

0.44

0.58

0.21

Notes:

Time dummies are included in all specifications. Industry dummies (in the levels equations) are insignificant and thus excluded.

Predicted probabilities taken from a recursivelyestimated takeover likelihood model.

All equations estimated with onestep GMM system estimator. All independent variables except for age (which is treated as exogenous) are instrumented. Instruments for equations in firstdifferences are as in Table 4.5, with additional regressors instrumented with twice and furtherlags. Instruments for equations in levels are as in Table 4.5, with the addition of Rum_{it1} in Columns 3 and 4. Sargan test rejects (D/Y)_{it1}, Q_{it1} and LEV_{it1}.

In parentheses are asymptotic standard errors robust to general crosssection and timeseries heteroskedasticity. The superscript ‘a’ indicates that the coefficient is significantly different from zero at the 0.01 level, ‘b’ at the 0.05 level and ‘c’ at the 0.10 level.

m1 and m2 are test statistics, distributed standard normal, for first and secondorder serial correlation in the firstdifferenced residuals. Sargan statistic is that for the corresponding twostep GMM estimator, distributed chisquared. Pvalues are reported.
