"Ratifiable Mechanisms: Learning from Disagreement," (with Thomas
R. Palfrey) Games and Economic Behavior, 10, 255–283, 1995.
In a mechanism design problem,
participation constraints require that all types prefer the proposed mechanism
to some status quo. If equilibrium play in the status quo mechanism depends on
the players’ beliefs, then the inference drawn if someone objects to the
proposed mechanism may alter the participation constraints. We investigate this
issue by modeling the mechanism design problem as a two-stage process,
consisting of a ratification stage followed by the actual play of the chosen
game. We develop and illustrate a new concept, ratifiability, that takes
account of inferences from a veto in a consistent way.
"Relational Investing and Agency Theory," (with Ian
Ayres) Cardozo Law Review, 15, 1033–1066, 1994.
This Article analyzes how, and when, corporate
governance could be improved by utilizing "relational investing." The term
relational investing is just coming into vogue and there does not yet seem to be
a consensus on what it means. Although the term has been trumpeted on the cover
of Business Week, before the Conference on Relational Investing at Columbia
University, relatively little legal writing had been published on the subject.
For the purposes of this Article, we define relational investing to encompass
commitments to buy and hold significant blocks of a corporation's stock. And it
is particularly important that the relational investors commit not to tender
their shares to hostile bidders. Using our definition, relational investing is
used to foreclose or reduce hostile takeover threats, replacing this form of
external discipline with enhanced internal discipline by the relational
investors. The long-term investment induces the relational shareholders to
invest more in acquiring information about the effectiveness of management. To
be effective internal monitors, however, relational investors must be able to
use this information to influence corporate policy. At a minimum, relational
investors must be "provocable" -- they must be
able to increase the likelihood that poor management or poor policies will be
changed. Relational investors might accomplish these changes through either
internal (informal negotiation or proxy contest) or external (tender offer)
means. Although we will often assume that relational investors are committed to
patient oversight, it is important to remember that these commitments are
usually noncontractual, suggesting that an implicit commitment to buy and hold
stock must be self-enforcing. This self-enforcement constraint might be
especially useful in determining when relational investing is likely to arise.
For example, the short-term illiquidity of large blocks of stock might make the
buy and hold commitment more credible. Moreover, large block holders often will
lack an incentive to reduce the size of their holdings. The 13(d) filing
requirements of the Securities Exchange Act could also facilitate relational
investing, because unfulfilled representations to buy and hold stock can give
rise to legal liability. Using the minimalist definition that relational
investors commit not to tender a large block of shares, it is possible that
relational investing could reduce agency costs by providing a more effective
form of corporate governance. This is far different from arguing that, as an
empirical matter, relational investing is superior to more traditional forms of
corporate governance. Indeed, some theorists suggest that with
"friendly" relational investing, there is a substantial risk of
entrenched managers and exacerbated agency costs. Without adjudicating the
ultimate efficacy of relational investing, our analysis illuminates how
relational investing might create value and highlights the contexts in which it
is most likely to be effective. We generate three main conclusions from our
analysis of relational investing: First, relational investing can reduce agency
costs, both by increasing the principal's incentive to acquire information, and
by improving the principal's ability to foster a monitoring reputation through a
long-term relationship with the firm's management. Large block holders have a
greater incentive to monitor than do "rationally ignorant" atomistic
shareholders. In addition, the commitment to hold for long periods of time
permits relational investors to enter more credibly into self-enforcing implicit
contracts that discipline poor managerial decisions and abilities. Second,
relational investing may be better suited to mitigating "moral hazard"
problems than traditional types of monitoring. In particular, potential
third-party bidders are less likely to respond to problems of moral hazard than
to problems of adverse selection. Even in a strong form efficient capital
market, external monitors may not have an adequate incentive to discipline
managers who have succumbed to moral hazard and caused the corporation to bear
an inefficient sunk cost. Relational investors, in contrast, have a multiperiod
incentive to respond. Third, relational investing can be rationalized by either
of the following theories: (1) the threat that managers will lose their jobs is
minimal; or (2) managers face too great a threat of losing their jobs.
"Cartel
Enforcement with Uncertainty About Costs," (with Thomas R. Palfrey) International Economic Review,
31, 17–47, 1990. Reprinted in Stephen W. Salant and Margaret C.
Levenstein (eds.), Cartels, Volume 1, Cheltenham,
UK: Edward Elgar, 2005.
What cartel agreements are possible
when firms have private information about production costs? For private cost
uncertainty we characterize the set of cartel agreements that can be
supported, recognizing incentive and participation constraints. If defection
results in either Cournot or Bertrand competition, the incentive problem in
large cartels is severe enough to prevent the cartel from achieving the monopoly
outcome. However, if the cartel agreement requires less than unanimous
ratification by the member firms, then the incentive problem can be overcome in
large cartels. With common cost uncertainty, perfect collusion is
possible in large cartels, regardless of the ratification rule.
"Nonrandom Mixing Models of HIV Transmission," (with Edward
Kaplan, and A. David Paltiel) in
Mathematical and Statistical Approaches to AIDS Epidemiology, edited by
Carlos Castillo-Chávez, Lecture Notes in Biomathematics Series,
Springer-Verlag, 218–239, 1989.
Models of HIV transmission and the
AIDS epidemic generally assume random mixing among those infected with HIV and
those who are not. For sexually transmitted HIV, this implies that individuals
select sex partners without regard to attributes such as familiarity,
attractiveness, or risk of infection. This paper formulates a model for
examining the impact of nonrandom mixing on HIV transmission. We present
threshold conditions that determine when HIV epidemics can occur within the
framework of this model. Nonrandom mixing is introduced by assuming that
sexually active individuals select sex partners to minimize the risk of
infection. In addition to variability in risky sex rates, some versions of our
model allow for error (or noise) in information exchanged between prospective
partners. We investigate several models including random partner selection (or
proportionate mixing), segregation of the population by risky sex rates, a
probabilistic combination of segregation and random selection induced by
imperfect information (or preferred mixing), and a model of costly search with
perfect information. We develop examples which show that nonrandom mixing can
lead to epidemics that are more severe or less severe than random mixing. For
reasonable parameter choices describing the AIDS epidemic, however, the results
suggest that random mixing models overstate the number of HIV infections that
will occur.