(Page 3 of 30) - Balance of Power Dynamics and War Escalation authored by Bennett, Scott.

The study of how the balance of power and other dynamic variables affect dispute

escalation falls squarely into the set of problems where imposed temporal units pose a problem.
Because our unit of observation when we study militarized dispute (MID) escalation is normally
the MID, with a dependent variable marking whether the dispute escalates to war (1) or not (0),
we lose all ability to study dynamics within MIDs. If the balance of power changes over the
course of a MID, for instance when additional states join ongoing disputes, this provides
important information to disputants, and our independent variable of the balance of power would
be different if measured at different points. But currently, such information is lost, because it
cannot be appropriately incorporated into our research designs. We might think it a ridiculous to
think that a dispute that remains bilateral should behave the same way as a dispute where 10
states join the dispute; we might think it ridiculous that a dispute where one state joins each day
should behave like a dispute where no states join for 30 days and then 10 join on the 31

day.

Right now, though, this is exactly what we must do given our research designs.

Luckily, there is a research design and analytic solution to the problem of dynamic

variables within units (and particularly the problem of understanding the balance of power) in the
form of competing risks models of dispute escalation. Although we have a fundamental interest
in the escalation or peaceful settlement of the MID as a whole, we can construct our data with
time-varying covariates within MIDs, allow us to measure dynamic variables over time. Over
time data are then linked in analysis to provide overall estimates of how the changing variables
affect whether MIDs escalate, or are settled peacefully. In this paper I suggest that we can get
further on these issues by applying a competing risks duration model to the study of study of
dispute escalation to war. I focus on the relationship between the balance of power and dispute
escalation as the main substantive question in this paper, but the technique explored here is
applicable to any dynamic variable where change occurs at a level “under” our overall cases.

Below, I begin by discussing further the state of the evidence about the relationship

between the balance of power and war. I then discuss why our current research designs in use
actually cannot let us “find” better answers to the questions we are interested in. I present an
argument for why competing risks duration models can help us answer these questions. Finally,
I present several analyses demonstrating that in fact the balance of power does indeed affect
conflict escalation, a finding contrary to prevailing wisdom, but correct in the light of
appropriately applied methods.

Linking the “Balance of Power” and Conflict

Definitions
The term “balance of power” has been used in many ways in the conflict literature. Here,

I use the term to mean simply the measured bilateral or dyadic military balance of capabilities
between the two sides of a conflict (potential).

A related term is “balancing,” which refers to

state behavior in the context of having an option to join one side of a conflict or another. As will
be discussed below, “balancing” states may be those that join the side of their allies, join the
weaker side in a conflict, or join the side of status quo or defending state in a conflict. The
alternative to “balancing” is “bandwagoning,” which refers to states joining the stronger, or more
aggressive, side in a conflict. Again, here I use the term to mean the measured balance of
military capabilities between the two sides of a conflict. The key question that the research here

Other uses of the term have included the “balance of power” as a description of the status quo, a description of past

conflict behavior, a description of a type of international system, or a description of a state’s foreign policy.