welfare.
Example 1 In figure 3 we set θI = .1, θR = .5, α = .5, p = .5, τ = .5, λ = .7, a = .4, B = 2
and Γ = 0.8. With these parameters π* = 0.44 and br(π*) = 0.47 .
Example 2 In figure 4 we keep the same data of example 1 but assume that both the
stakeholder ability at affecting the replacement decision, and the incumbent control benefits
are higher (i.e. we set a = .8 and Γ = 2). In this case: π* = 0.038 and br(π*) = 0.063.

Figure 3: The continuous (dotted) curve represents shareholder value when entrenchment is
(not) to be countered. Shareholders preempt entrenchment: π* = 0.44 and xbr(π*) = 0.47.
Notice that in both examples shareholder value is indeed maximized by countering man-
agerial entrenchment (i.e., it is not optimal to set π and xr below the xbr(π) locus). Therefore,
we can conclude:
Proposition 2 There is an open set of parameters for which shareholder value is maximized
by countering managerial entrenchment, and hence by providing a minimal level of explicit
stakeholder protection xbr (π*) > 0.
Remark 2 — Welfare effects — To be written.
17
More intriguing information
1. A model-free approach to delta hedging2. The Dynamic Cost of the Draft
3. The Economics of Uncovered Interest Parity Condition for Emerging Markets: A Survey
4. Correlation Analysis of Financial Contagion: What One Should Know Before Running a Test
5. The name is absent
6. The name is absent
7. The name is absent
8. Technological progress, organizational change and the size of the Human Resources Department
9. The Impact of EU Accession in Romania: An Analysis of Regional Development Policy Effects by a Multiregional I-O Model
10. Improvements in medical care and technology and reductions in traffic-related fatalities in Great Britain