supervisory board dismissal in dax30 companies a ... · supervisory board dismissal in dax30...

27
Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks Supervisory Board Dismissal in DAX30 Companies – a Proportional Hazard Analysis Benjamin Balsmeier 1 , Alexander Dilger 2 , J¨ org Lingens 2 1 University of Leuven 2 University of M¨ unster October 24, 2011

Upload: others

Post on 31-May-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Supervisory Board Dismissal in DAX30 Companies

– a Proportional Hazard Analysis

Benjamin Balsmeier1, Alexander Dilger2, Jorg Lingens2

1University of Leuven 2University of Munster

October 24, 2011

Page 2: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Motivation

Many large cooperations in the developed world are organized

as public companies.

This implies that ownership (capital owners) and leadership

(management) are effectively separated.

The implication is that public companies are confronted with

a severe principal-agent problem.

Trying to solve this problem lies at the heart of the many

corporate governance codices around the world.

Page 3: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

The corporate governance codex is a framework which

determines how to organize the company AND (sometimes)

the behavior of the management board (coming in the form of

a legal requirement).

In Germany e.g. the executive board (’Vorstand’) is overseen

by the supervisory board (’Aufsichtsrat’).

The composition of the supervisory board reflects the

stakeholder approach (favored in continental Europe) where

e.g. employer representatives are members of the board.

Page 4: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Other countries have different approaches, e.g. the US

monistic board structure with only one board that runs AND

oversees the company.

But also in this approach, outside directors monitor inside (i.e.

executive) directors and have the right to fire them.

The duties of outside directors are hence very similar to that

of the supervisory board.

Page 5: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

With corporate governance being so important to solve the

principal-agent problem (being a prerequisite for efficient

allocations) there are some questions that need to be

considered/answered:

1 What are the effects of corporate governance on firm

performance (i.e. return on investment, employment

stability...)?

2 Which variables affect the ’success’ of corporate governance

and what is the direction of this effect?

Page 6: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

This paper analyzes the latter question.

The focus here is how personal interdependencies of

supervisory board members affect the success of corporate

governance.

Why? Because people have the notion of the ’Deutschland

AG’ and we would like to know whether we find this notion in

the data.

Page 7: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Identification

is really hard...We will analyze (everything else equal) how personal

interdependencies shield against dismissal. Based on this analyzes

we would argue that shielding against dismissal might be

associated with bad governance.

Page 8: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Literature

Loads of literature...

however mainly focussing on the anglo-saxon (i.e. US style)

corporate governance system.

Concerning the effects of networks and personal

interdependencies: mixed evidence.

Page 9: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Fich/Shivdasani JoF (2006): show firm underperformance if

board members have multiple external mandates.

Core et al. JoFE (1999): report that CEOs have higher

salaries if they have multiple mandates.

Fahlenbrach et al. JoF (2010): CEOs with multiple mandates

have no effect on firm performance.

Page 10: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Results

An additional supervisory board membership in another

DAX30 company substantially decreases the probability of

being dismissed.

Other variables measuring interconnectedness have no

significant effects.

We find substantial differences in the dismissal probability

between capital and employee representatives.

Page 11: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Data

Data

We have collected data on firm level variables and information

concerning the persons on supervisory boards using

Hoppenstedts Aktienfuhrer.

For 2001-2005 we collect data on all companies that have

been listed on the DAX30 at least once in this period.

This gives us information on 35 companies and the persons

who are on the supervisory and the executive board of the

respective company.

Page 12: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Data

We generate a number of covariates such as firm

performance, number of employees, fraction of widely held

stock (to name a few), but also average salary of supervisory

and executive board members.

Moreover, knowing the board members’ names, we can

construct ’employment’ spells.

Additionally, we can construct a measure for the number of

multiple spells one board member has in other DAX30

companies.

Page 13: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Data

Variable N Average Std. Dev. Min Max

Return on Equity (in %) 2,383 19.545 23.561 -136.113 66.629

Employees 2,383 131.368 124.954 0.038 502.545

Compensation Supervisory Board (log) 2,383 11.040 0.706 8.942 12.273

Compensation Executive Committee (log) 2,383 14.158 0.770 6.988 15.772

Dividend Paid (1=yes) 2,383 0.904 - 0 1

Chair of Supervisory Board (1=yes) 2,383 0.055 - 0 1

Employee Representative (1=yes) 2,383 0.469 - 0 1

Size of the Supervisory Board 2,383 18.743 2.905 5 22

Free Float(=widely held stock) 2,383 56.136 26.102 1 100

Rate of Investment (in %) 2,383 7.220 4.387 0.014 25.395

Page 14: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Data

Variable N Average Std. Dev. Min Max

Outside Mandates in SBs (individual) 2,383 0.463 1.115 0 6

Outside Mandates in ECs (individual) 2,383 0.067 - 0 1

Outside Mandates in SBs by Shareholder Rep. 2,383 8.572 5.899 0 23

Outside Mandates in SBs by Employee Rep. 2,383 0.432 0.796 0 4

Outside Mandates in ECs (cumulated) 2,383 1.287 1.099 0 4

When measuring the impact of networks, we consider the

individual effects of multiple mandates, but also that of aggregate

membership of all peers in the board.

Page 15: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Data

How were descriptive statistics calculated?

817 individuals are observed in every year of their spell: in

sum this makes N = 2383 observations.

using this N as the basis for descriptive results is debatable:

individuals with long spells and firms with large supervisory

boards are over-represented.

Note, however, that we did some robustness checks which

show that the descriptives are all in the same ballpark range.

Page 16: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Method

Method

Our identification strategy requires the estimation of a

conditional probability of leaving the supervisory board

(dismissal).

The data at hands, however, (potentially) do not provide

enough information using standard estimation techniques

(OLS).

Why? The spell data is (left and right) censored.

Page 17: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Method

One solution to this problem would be to dump censored

observation.

Result: Loads of information would be lost + severe bias in

the remaining data.

Our approach: use proportional hazard (Cox regression)

techniques.

Advantages: stochastic model to cope with censoring +

non-parametric approach.

Page 18: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Method

Brief intuition of the method:

Let T denote the random variable ’length of supervisory board

spell’, let t be a realization.

Cumulated distribution function is denoted by F (t) which

defines the hazard rate λ(t) = F ′(t)/(1− F (t)).

λ(t) measures the probability of leaving the board during the

next time instant dt conditional on being in the board for a

’spell’ length t.

Page 19: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Method

The hazard rate is our endogenous variable and we estimate a

function of the form λ(t,X) = λ0(t)eβX.

λ0 is the baseline hazard rate (which is not explicitly modeled

here) and X is the vector of covariates.

Using the Cox-Regression approach, the parameter vector β is

estimated using maximum-likelihood (without specifying the

prob. distribution for the general dynamics of spells).

Page 20: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Method

Note the interpretation of β:

d log[λ]

dXi=

dλ/dXi

λ= βi

The parameter hence measures the relative marginal change in

the hazard caused by a marginal change in Xi .

Page 21: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Variable Coefficient

Return on Equity 0.016***

Employees 0.132***

Employees squared -0.000192***

Compensation SB (log) -1.530***

Compensation EC (log) -0.709

Dividend Paid -0.384

Chair of SB 0.019

Employee Representative -0.095

Size of the SB -0.534***

Free Float -0.047***

Rate of Investment 0.081***

∗ ∗ ∗ 1% ∗∗ 5%

Page 22: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Higher return on equity increases the probability of being

dismissed (1.6% if return increases by 1%point)

Employment has non-linear effect on hazard rate.

1% increase in the (average) salary (of the supervisory board)

decreases hazard rate by 1.55% (causality!).

Page 23: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Variable Parameter

Outside Mandates SBs (ind.) -0.174**

Outside Mandates ECs (ind.) -0.173

OM in SBs by Shareholder Rep. -0.023

OM in SBs by Employee Rep. -0.006

OM in ECs (cumulated) -0.077

Page 24: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Only individual interdependencies affect the hazard of leaving

the board.

We do not find peer effects, i.e. the number of outside

mandates of the peers does not affect dismissal.

Significant effect only for individual mandates in other DAX30

supervisory boards: 1 additional mandate decreases hazard by

17%

Page 25: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Distinguishing employee and shareholder representatives shows

(not reported here):

employees’ representatives spells’ are not affected by outside

mandates.

structural covariates affect dismissal probability differently

(e.g. compensation of executive board).

Page 26: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Robustness

Variations at the individual and the firm level → robust

standard errors

Heterogeneity of the Baseline Hazard Rate → firm-level

stratification

Is the impact of covariates proportional → check

Page 27: Supervisory Board Dismissal in DAX30 Companies a ... · Supervisory Board Dismissal in DAX30 Companies { a Proportional Hazard Analysis Benjamin Balsmeier1, Alexander Dilger2, J org

Motivation Literature Results–Preview Econometric Approach Results Concluding Remarks

Conclusion

We constructed a data set with spell information on

supervisory board membership.

Using this data we estimated a Cox model explaining the

conditional probability of being dismissed (=hazard rate)

We have shown that personal networks only affect the hazard

in rare cases (individual simultaneous membership).

Based on the identification strategy we would be tempted to

argue that the notion of the ’Deutschland AG’ seems to be

exaggerated.