that dyad, rather than due to disagreement over the regional status quo. To allow for this difference, I
include a variable “Previous Unification” that equals one whenever the dyad members were previously
unified. The correlation between this variable and contiguity is so modest that it is reasonable to include
them as separate indicators.
The historical sources listed above also indicate when treaties of alliance were signed among
autonomous political units in the Rio de la Plata. I code an alliance as present whenever one or more
historians claim a treaty of alliance existed among actors. I have found such evidence substantiating the
existence of eleven alliances over the years from 1812 to 1862. Some of these were simple dyadic
agreements (such as between Entre Rios and Corrientes in 1827), others involved as many as fourteen
actors (the Argentine Confederation of 1853-62). The shortest alliance (Entre Rios and Corrientes) lasted
only one year, the longest (Rosas’s Federal Pact) lasted twelve. I include alliance as a control variable in
my analyses both because its inclusion is standard in IR conflict research, and because the existence of an
alliance could be a manifestation of similar attitudes toward the Rio de la Plata status quo. In the analyses
to follow, the variable “Alliance” indicates that the dyad members shared an agreement on military
matters codified in a written treaty of alliance.
The control variables included below are “Log of Distance,” “Contiguity,” “Previous
Unification,” and “Alliance.” Ideally I would have also included variables representing whether each
dyad was jointly democratic and how much trade existed between them. But democracy was so rare
within the region during the time period under study that there was only one jointly democratic dyad
(Corrientes and Uruguay for part of the 1830s and 40s). That this dyad did not fight produces perfect
identification between joint democracy and peace. This lack of variation prevents me from controlling
for the presence of joint democracy. Similarly, trade data are simply unavailable for the actors in the
region in the first half of the 19
th
century. Some province-specific trade data with European states exist,
but I have found none among Rio de la Plata actors writ large.
Finally, I also correct for duration dependence with the natural cubic spline technique described
by Beck, Katz and Tucker (1998). In the analyses to follow I do not report the estimates of the spline’s
effects, although all four peace years variables are statistically significant in every analysis. In
constructing the spline function I placed the knots at each quartile in the distribution of peace years.
Quantitative Results
If power transition theory is correct the interactive effect of Parity and SQ Dissatisfaction will be positive
and statistically significant. Existing research suggests “Log of Distance” and “Alliance” will be
negatively associated with war, while “Contiguity” will be positively associated.
The dataset described above features annual observations of each Rio de la Plata dyad. There
were from four to seventeen autonomous political units in the region each year from 1810 to 1862, and a
resulting universe of 5261 annual dyadic observations. Reflecting on Bennett and Stam’s (2000)
demonstration that some relationships are not robust across different model specifications, I report the
results in a variety of ways. First, since the decisions to start a war and to continue a war may differ, I
present analyses in which the dependent variable is coded 1 whenever a war begins within a dyad, and
alternatively coded 1 for any year in which a war starts or is ongoing. Power transition theory’s war
hypothesis has traditionally been treated as one about war onset, and thus I will stress the War Onset
analyses below. The alternate analyses including ongoing war are presented to demonstrate robustness.
A second source of variation in results about international conflict is differences in the case selection
rules followed. Not surprisingly, the sample of cases employed can influence the results reached.
Consequently I analyze three different sets of cases.
The first set of results are reported in Table 1a. I have argued elsewhere (Lemke 2002:112) that
power transition theory is most appropriately analyzed for dyads that include the system’s dominant
power. Since power transition theory suggests wars are fought for control of the systemic status quo, it
seems plausible that wars pertinent to evaluation of the theory must include the actor that created and
maintains the status quo. If that argument is valid, then the results in Table 1a are the most appropriate
for analysis of power transition theory’s applicability in the Rio de la Plata region, because Buenos Aires
province was almost always the most powerful “state” in the region. In addition to being the perhaps
most theoretically pertinent test, it is also the most stringent in that the fewest cases, and thus the least
9
Convention