A Bargaining Model of Domestic Politics and the Cost of War
Unformatted Document Text:
in bargaining and war.
The second problem the bargaining model poses lies in empirical testing.
historical data sets provide some, but certainly not enough, information to fully evaluate
the propositions of the bargaining model. The relative dearth of empirical investigations
of bargaining models is due, in part, to the fact that many of the key conceptual elements
of the bargaining model lack direct (or even indirect) empirical measures. To address the
observability problem, we turn to laboratory experiments to shed light on how the bargaining
structure shapes decision making, and to evaluate when and why decision makers diverge
from equilibrium strategies. The use of experimental methods to evaluate predictions from
the bargaining model is important because they move us beyond measurement issues that
hinder non-experimental approaches. In the laboratory we can specify and measure the
eﬀects of variables that are diﬃcult or impossible to assess empirically. In the context of
bargaining, we can set and evaluate the eﬀects of initial shares and probability of success on
players’ decisions. This aspect gives us a direct test for the bargaining model’s predictions.
Moreover, suﬃcient replications of each treatment permit one to see causal relationships and
gauge variability due to individual eﬀects. While experiments with ﬁnancially motivated
subjects intentionally sacriﬁce some richness in context, this is done for gains in replication
and control (e.g. Davis & Holt 1993).
These advantages can be quite important, since
actual political conﬂicts are surrounded by a rich and unique array of parallel events and
personalities which sometimes make it hard to spot underlying factors that may inﬂuence
outcomes in a predictable manner.
Our goals in this paper are aimed at addressing the two persistent problems identiﬁed
above. First, we produce a formal analysis of the bargaining model that allows deviations
from equilibrium due to noise surrounding beliefs about the other side’s rationality. We
construct a quantal response model that characterizes the distribution of behavior around
a Nash equilibrium depending on those beliefs. Second, we report results of laboratory
experiments designed explicitly from the structure and predictions of the bargaining model