Centers of Gravity: Regional Powers, Democracy, and Trade

ABSTRACT

Classic studies on hegemonic stability and power transition suggest that concentration of capabilities favoring a single state can promote economic cooperation and discourage militarized conflict. However, tests of these arguments have been primarily limited to examining temporal variation in global capability distributions and corresponding levels of system-wide cooperation; few have examined the impact of capability concentration at the region level. In this article, we contend that concentration of regional military capabilities corresponds to lower trade costs for states throughout a region and to an incentive for weaker states to de-prioritize expenditure on the military, freeing resources that can be used to promote trade. As a result, this condition promotes higher levels of trade, particularly within the region. We also argue that democratic regional powers are better able to foster confidence in the sustainability of cooperation; thus, the trade-enhancing impact of concentrated regional capabilities is stronger when the predominant state is more democratic. We find evidence in support of our expectations in statistical models examining state trade between 1960 and 2007.

KEYWORDS:

• Power
• trade
• regional cooperation

Supplementary data

Supplemental data for this article can be accessed on the author’s website.

Notes

¹ Power cycle theory (for instance, Doran 1989; Doran and Parsons 1980) is unique among the cyclical theories of conflict and cooperation in that the explanation of major power war centers on the rate of change in states’ capabilities, relative to states’ role or prestige in the international system.

² Specifically, the hegemon could guarantee fixed (or at least stable) exchange rates, lead in the coordination of macroeconomic policies, and provide liquidity during financial crises.

³ Kindleberger (1973) blames the Great Depression and Second World War on the lack of a hegemon—due to the UK’s inability and the US’s unwillingness to take this role.

⁴ See Pahre (1999) for an extensive review of HST critiques. Examples include Conybeare (1984), who challenged HST on the basis that it violated principles of standard international trade theory since any state that is large enough to inuence its terms of trade would impose an optimum tariff (but see Gowa [1989], who argues that this argument violated the logic of game-theoretic models). Other debates have centered on the measurement of hegemony (for instance, Webb and Krasner 1989) and the range over which a state must wield inuence in order to be considered a hegemon Baldwin (1979).

⁵ Notably, lack of militarized conflict is not necessarily synonymous with increased cooperation (for example, economic openness). However, as we will note, reduced risk of conflict suggests reduced barriers to economic cooperation.

⁶ Fearon (1995), and most notably, Powell (1999), consider the distribution of capabilities to be separate from state beliefs over this distribution. However, we contend that, as the regional distribution of capabilities increasingly favors one state, all states should become increasingly confident that this state would win an armed conflict, thus decreasing the likelihood thereof.

⁷ Using bargaining theory, previous research has found support for a pacifying impact of power preponderance at the dyadic level following from clearer information (Reed, Clark, Nordstorm, and Hwang 2008). This finding can be extended to the regional level if regional powers are likely to become involved in conflicts in their sphere of influence, as we will discuss.

⁸ The natural extension of power transition theory to lower levels of analysis could explain why scholars have been able to subject the theory to rigorous empirical testing.

⁹ Perhaps the most notable advance has been in containerization, which reduced shipping costs dramatically.

¹⁰ We assume that there are uniform trade costs across industries and territory within a given region. While this assumption is almost certainly not true, it is useful for purposes of parsimony. Although a regional preponderance of military capabilities may only result in lower trade costs for some commodities, we will see increased trade of those commodities and, therefore, increased aggregate trade. Future research could benefit from examining whether a regional preponderance of capabilities fosters sharply increased trade in some commodities but not others.

¹¹ Indeed, most, if not all, of the other actors they list could be linked at least indirectly to information.

¹² This expectation fails if both sides of a conflict believe the regional power will intervene on their side. However, we expect that the regional power will make clear which side it supports. Given that it faces no regional threats, yet seeks to prevent erosion of regional stability that could weaken it relative to extraregional competitors, regional powers should very rarely favor conflict among other states in their region.

¹³ This is not to say that these states simply choose to disarm and not maintain a military. They should, however, pursue those policies with less fervor than otherwise would be the case and redirect their efforts toward more cooperative policies.

¹⁴ Critically, data on policy barriers are considerably limited. While some indicators exist for tariffs and nontariff barriers, these data preclude a time-series cross-sectional analysis on a large scale. More problematic is the fact that some barriers to trade (for example, currency manipulation, quality standards) are difficult to measure. Aggregate policy barriers to trade are infeasible for this study. However, looking at trade flows, and controlling for the confounding influence of distance and economy size, allows us to capture the influence of concentrated capabilities on trade costs.

¹⁵ In theory, GATT/WTO rules limit the ability of members to raise external barriers. Nonetheless, preferential liberalization within a region could lead to diversion of trade away from states outside the region even if the outsiders are more efficient producers.

¹⁶ Although Mitchell (2002) examines the number of democracies at the system level, her findings could extend to subsystems as well. Given the presence of dominant power, it stands to reason that its regime type would be at least as important as the overall count of democracies in the region for the provision of mediation.

¹⁷ We assume that there is always a most powerful state, which we deem the “regional power,” in a region. The margin of difference between the first- and second- most powerful might be slight, but perfect parity seems unlikely.

¹⁸ In the main text, we present models including all regions and all of these years. However, in the supplemental appendix, we also present models testing the robustness of our results when removing notable regions (for example, Western Europe and North America) or subsetting by time period (Cold War vs. post-Cold War). All results are robust to region and time period restrictions.

¹⁹ Results look very similar if we include regional powers in our analysis. These results are available by request from the authors.

²⁰ Heckman models have been used to address the problem of zero trade flows. In the supplemental appendix, we use a more recent technique, modeling dyadic trade using a quasi-poisson distribution.

²¹ Again, results are similar if we include trade with the regional power. These results are presented in the supplemental appendix.

²² We add 1 to the raw values of intra- and interregion trade because some rare cases see trade values below 1, which would produce a negative logged value increasingly close to negative infinity as the raw trade value approaches zero.

²³ The inclusion of year fixed effects renders our results consistent with those using current dollars.

²⁴ Although we hypothesize and test expectations regarding levels of capability concentration and trade, the use of unit fixed and random effects result in a focus on variation within units such that our results provide at least some evidence for how change in our explanatory variables corresponds to change in the dependent variables (for instance, Schneider 2013).

²⁵ States are nested within regions; thus state-fixed effects imply that region effects are incorporated at the state level. We cannot also incorporate region-fixed effects because these would be time-invariant for each state.

²⁶ Given that some of Lemke’s are quite small—the smallest being the Central Highlands of Africa, which contains only two states—we replicated our analysis using larger regions as defined by the COW Project. The COW project divides the world into six regions: the Western Hemisphere, Europe, Sub-Saharan Africa, the Middle East and North Africa, Asia, Australia and Pacific Islands. However, the COW regions are largely atheoretical, generally defined by geography—primarily the division between continents. Accordingly, we prefer Lemke’s region classification. Additionally, our results are robust to some theory-driven revisions to Lemke’s coding— for example, removing Russia from the Europe region and including it with Central Asia.

²⁷ The CINC score measures six aspects of capabilities, deflating each in terms of total world capabilities (such that CINC scores can be interpreted as a state’s share of total world capabilities). The six indicators are (1) military personnel, (2) military expenditure, (3) iron and steel production, (4) energy consumption, (5) total population, and (6) urban population.

²⁸ In additional models available by request from the authors, we created an alternate measure of regional power by using the regional power’s share of total regional capabilities. This value must be greater than and can theoretically range as high as 1. This variable has the advantage of simplicity; however, interpretation is somewhat more difficult because (1) its value cannot range as low as 0, and (2) its value is dependent on the number of states in a given region. Both of these issues are less problematic as the value approaches 1. Indeed, we contend that both of these variables are useful to capture regional predominance, as their coefficients suggest a scenario where a regional power’s share of capabilities approaches 100%.

²⁹ Technically, our value would only equal 0 if all states in a given region had identical capabilities; as such, there would be no regional power. While technically possible, there are no cases of this equal dispersion of capabilities in our data. However, our NHHI does vary as low as 0.000073.

³⁰ The distribution of regional power Polity scores is bimodal, with relatively few 0s, and more values closer to –10 and 10.

³¹ In the supplemental appendix, we present models excluding this conflict variable. Our argument is that capability concentration promotes confidence that armed conflict is unlikely to occur, leading in turn to higher trade levels. Therefore, armed conflict could be considered to be a mediating variable, which is best excluded from the analysis. All results are consistent in these alternate specifications.

³² All statistical model are run in R version 3.2.3, using the plm and lme4 packages.

³³ Given that we are less interested in the conditional association between regional power democracy and trade, we exclude a detailed examination of conditional estimates from this article. However, we find in general that regional power democracy has a negative association with trade when capabilities are not concentrated but has an increasingly positive and significant association with intraregion trade as regional capabilities become more concentrated. The supplemental appendix presents figures demonstrating this “side” of the interaction.

³⁴ This value is also calculated using the formula. However, in this instance, refers to the conditional coefficient calculated in accordance with interpretation of interaction effects. For a model, we calculate the conditional coefficient for X as equal to.

³⁵ Kinne (2012) finds a pacifying impact of higher closeness centrality—a local measure of embeddedness that could correlate with concentration of military capabilities. Future work could compare the effect of closeness centrality to that of eigenvector centrality, the latter of which measures embeddedness on a global scale.