Dyadic Trade,Exit Costs,and Conflict

Abstract

Most studies of the link between dyadic trade and militarized conflict examine the extent of trade interaction. However, interaction measures do not account for the impact of cutting off trade (i.e., exit costs). In this article, I highlight the link between exit costs, the cost of conflict, and “the spoils of conquest,” arguing that one state’s exit costs are associated with higher incidence of dyadic conflict when its trade partner’s exit costs are low. However, its exit costs become less aggravating—and eventually pacifying—as its trade partner’s exit costs increase. I test this argument by estimating import demand and export supply elasticities, developing yearly exit cost measures for directed dyads, 1984–2000. Statistical tests confirm that unilaterally high exit costs are aggravating, but that jointly high exit costs are pacifying, a pattern most prominent for trade in strategic commodities.

Keywords

trade and conflict vulnerability trade elasticities

With record levels of trade flowing across state borders, scholars advocating the “peace through trade” hypothesis predict that militarized conflict should become increasingly rare. Yet, history has shown that this relationship is not so simple. World War I occurred despite then—unprecedented levels of trade, and amid predictions that such costly conflicts would be unthinkable (or at least unprofitable; Angell 1913). Similarly, conflicts have raged in recent years despite high and growing levels of trade. Improving our understanding of when and how trade precludes or encourages conflict is crucial, particularly given the rise of China as a trading state and military power. Yet, despite significant advances in recent years, theoretical and empirical obstacles continue to preclude a clear understanding of the relationship between trade and conflict (Mansfield and Pollins 2001).

In this article, I account for the potentially differential vulnerability inherent in dyadic trade (e.g., Hirschman 1945; Keohane and Nye 1977; Wagner 1988). To operationalize vulnerability, I expand on Crescenzi’s (2003, 2005) concept of exit costs that capture explicitly the ability of states to adjust to the interruption of trade. I take this concept one step further, considering the impact of relative exit costs within dyads. I posit complementarily aggravating influences of unilaterally high exit costs; for the less dependent state, they suggest higher incentives for coercion, while, for the more dependent state, they suggest higher gains from conquest. As such, unilaterally high exit costs suggest an elevated likelihood of conflict initiation. Conversely, given high exit costs for one state, higher exit costs for its trade partner encourage peace because mutually high exit costs are associated with relatively higher benefits from cooperation and higher costs for conflict. I illustrate this relationship with the case of Japan and the United States prior to World War II, demonstrating that the potential for states to recoup trade losses by force leads to conclusions differing from extant research in this area (e.g., Crescenzi 2003, 2005). I contrast the lead-up to World War II with the contemporary case of the United States and China, demonstrating that relatively high, mutual exit costs in this case may facilitate enduring peace.

Empirically, this article fills an important gap in the literature on trade and conflict as I look beyond trade interaction—the blunt measure that is typically utilized in the empirical literature—to the costs for each trade partner associated with interrupting trade. I introduce considerably expanded data, estimating relative exit costs at the directed dyad level to capture the vulnerability inherent in trade for 13,238 directed dyads, over seventeen years (from 1984 to 2000). Furthermore, extant exit cost measures are limited by their highly aggregated nature. To alleviate this problem, I create new measures at the Standard International Trade Classification two-digit (SITC2) commodity level, and then develop additive indices from these commodity-level exit cost measures.

I proceed with a discussion of extant literature examining trade and conflict, highlighting the limited explanatory power associated with a focus on trade interaction. Then, I discuss the connection between exit costs and vulnerability, linking exit costs to the costs and benefits of conflict. With exit cost measures developed from trade elasticities, and spanning directed dyad years between 1984 and 2000, I test three hypotheses, finding support for the argument that the impact of one state’s exit costs depend on those of its trade partner—particularly for trade in strategic commodities.

Exit Costs, Trade Interaction, and Conflict

International trade conveys benefits and costs to trading states. Liberal arguments tend to focus on trade gains that result from comparative advantage—economic benefits of specialization and cooperation. Trade gains are also exit costs; and because these gains terminate (i.e., exit costs are paid) at the onset of conflict, they serve as an opportunity cost to violence (Polachek 1980; Oneal et al. 1996; Oneal and Russett 1997, 2001; Russett and Oneal 2001; Polachek and Xiang 2010). Conversely, realist arguments focus on the political costs associated with trade, highlighting the potential for asymmetric trade gains to be leveraged as a means of coercion by the less vulnerable state (Hirschman 1945; Keohane and Nye 1977; Wagner 1988; Barbieri 1996). Furthermore, realists suggest that states have reason to fear the trade gains of their partners translating into military power to the extent that they may initiate conflict rather than allow adversaries to increase in relative capabilities.¹ Ultimately, exit costs have a dualistic nature. They exist when a trade relationship is at least somewhat beneficial to both parties, yet the fact that they need not be equivalent between trade partners suggests the potential for tension.

Despite the fact that exit costs are intrinsic elements of both liberal and realist theories linking trade to conflict, the vast majority of empirical studies testing this relationship measure trade as the extent of interaction (dyadic trade flows, often weighted by gross domestic product [GDP] or total national trade). This modeling decision tends to follow from practical considerations, given that measures of trade interaction are easily available.² However, because these blunt measures ignore the exit costs associated with cutting off trade relations, they are limited in their explanatory power. For example, a larger volume of dyadic trade may not equate with a larger incentive to avoid conflict (as adherents of the peace through trade hypothesis contend) if one or both dyad members can easily reroute lost trade flows to alternative markets; conversely, smaller volumes of trade may be pacifying if both trade partners cannot reap equivalent gains with third parties. Similarly, large trade volumes may not raise concerns for vulnerability if interrupted trade would be easily replaced. Given that measures of trade interaction are not well suited to answering research questions regarding the costs of cutting off trade, these measures have instead facilitated a second strand of liberal theory, which links trade to peace through increasing information flows that accompany economic interaction, reducing the information asymmetries that lead to conflict (Gartzke, Li, and Boehmer 2001; Gartzke 2003; Morrow 1999; see also Fearon 1995).

The Conditional Impact of Exit Costs on Conflict

Montesquieu ([1750] 1989) claimed famously, “Peace is the natural effect of trade. Two nations who traffic with each other become reciprocally dependent; for if one has an interest in buying, the other has an interest in selling.” This argument implies that there is a natural symmetry to trade relationships. At first glance, this assumption appears reasonable, given that individual buyers and sellers both profit from exchange and, therefore, would lose if trade were terminated. Yet, in the aggregate, there is no guarantee that exit costs are equivalent between trade partners. For example, when terminating trade of a good that is in high demand, the exporter may easily find alternate buyers in other states, whereas the importer might have to pay higher prices to obtain that same good elsewhere or produce it domestically.

Keohane and Nye (1977; see also Hirschman 1945; Wagner 1988) argue that trade may spur dyadic conflict because a state reaping trade gains—and therefore facing exit costs—is vulnerable to its trade partner, which can use the threat of terminating this trade relationship in order to coerce change in the dyadic status quo. Crescenzi (2003, 2005), through formal analysis, deduces that higher exit costs are associated with a higher likelihood of low-level conflict (typically economic or political sanctions or threats thereof—characterized by Crescenzi as economic exit), but that high exit costs for either the challenger or defender in a bargaining situation are associated with less high-level political conflict and military conflict (e.g., breaking of diplomatic relations; threats, displays, or use of force).

However, I contend that the relationship between trade and conflict is dependent on the conditional extent of exit costs, and specifically, that unilaterally high exit costs are aggravating to dyadic relations. For the United States, its allies, and Japan, the lead-up to World War II illustrates the potentially aggravating effect of unilaterally high exit costs. Prior to the war, Japan relied on the United States for a variety of vital commodities, most importantly oil and steel. Conversely, although the United States benefited from its exports to Japan, it risked relatively little economic harm if it terminated this trade, given the universally high demand for these commodities. Opposing Japan’s imperialist agenda in East Asia, the United States attempted to leverage Japan’s higher exit costs as political power. Beginning in July 1940, the United States began imposing harsh sanctions on Japan with the threat of more stringent consequences for Japan’s continuing hostility (Feis 1950; Hufbauer et al. 2007). This attempt at coercion reached a breaking point when, in July 1941, Roosevelt froze Japanese assets in the United States and severely limited petroleum exports to Japan. The Japanese response, far from submitting to US demands, was to attack Pearl Harbor in an attempt to eliminate US resistance, and then to capture oil and mineral producing territory from the Dutch and British, taking by force that which was denied to them in trade.³

The example above highlights the dual nature of vulnerability as, on one hand, a liability, and yet also a prize potentially to be won. The latter half of this dualism is addressed by Liberman (1993, 1996), who examines the “spoils of conquest,” demonstrating that conquest can be profitable if the value of captured resources outweighs the costs associated with controlling those resources. With regard to exit costs and conflict, there exists potentially an even stronger motivation to initiate conflict, given that, in addition to gains from conquest, there may be a high cost associated with the status quo: continued dependence and possible coercion. Yet, recent theoretical and empirical work examining exit costs and conflict (e.g., Crescenzi 2003, 2005) does not account for this aspect of trade dependence. Specifically, in Crescenzi’s formulation, higher exit costs for either the challenger or target are associated with reduced likelihood of high-level conflict because such conflict would be more costly. Additionally, Crescenzi’s model does not allow states to recoup exit costs by force from defeated adversaries. For example, even when victorious in the crisis equilibrium of Crescenzi’s game (in which political and/or military conflict occurs), a challenger gains only the value of its original demand, paying exit costs as well as additional costs associated with conflict (Crescenzi 2003, 814–15).

Conversely, I contend that high exit costs suggest that spoils of conquest would be valuable, ceteris paribus. Indeed, Japan’s high exit costs prior to World War II were connected directly to the benefits of conflict as well as its costs, as Japan had both economic and strategic incentives to capture the commodities on which it so depended. Although Crescenzi’s parameterization of expected outcomes is reasonable for the majority of political disputes, in which military conflict is not a serious consideration, it is less suited to cases where militarized conflict is a realistic possibility. By accounting for the benefits of conquest in the relationship between exit costs and conflict, my model applies to a broader set of cases.

Starting with the assumption that exit costs affect both the costs and the benefits of conflict, I contend that the events leading up to the attack on Pearl Harbor follow from a predictable relationship.⁴ In an important departure from recent literature, I conceptualize an interactive causal mechanism connecting exit costs to conflict. First, in accordance with Crescenzi’s theory, I contend that the opportunity cost of military conflict, and, hence, the incentive to avoid military consequences of political disputes is lower for a state that faces lower exit costs. However, given that a state facing low exit costs has a trade partner with high exit costs, the differing ability of states to endure lost trade may lead to coercion attempts by the state facing relatively little harm if trade is interrupted. Furthermore, a state’s relatively low exit costs may spur it to initiate militarized disputes against its trade partner in an attempt to gain concessions because it has relatively little to lose and the ability to impose significant economic harm on its adversary. Although the potential initiator could simply terminate trade as a punishment for the target’s refusal to submit to its demands, prior research shows that militarized conflict is likely to follow this type of economic coercion (Drezner 1998; Lektzian and Sprecher 2007).

Were military force not an option, the target of attempted economic coercion would be rational to concede if the cost thereof is lower than the cost of economic exit. Yet, the vulnerable state also has the option to use force in an attempt to capture as the spoils of war any paid exit costs. Indeed, the very fact that exit costs are high suggests that these potential spoils of war are extremely profitable, as, for example, the capture of strategic resources in the short term would preclude future coercion attempts; the victorious state in this case would eliminate its vulnerability. Even within dyads maintaining relatively good relationships, if a state with high exit costs perceives the potential for future coercion, then this vulnerability may itself serve as a source of tension. As such, there is a complementarily aggravating impact of unilaterally high exit costs, facilitating the initiation of militarized disputes by both the less vulnerable and the more vulnerable trade partner. This argument leads to my first two hypotheses:

Hypothesis 1: Higher exit costs for a state’s trade partner are associated with a higher likelihood that it initiates militarized conflict against that trade partner, given that its own exit costs are low.

Hypothesis 2: Higher exit costs for a state are associated with a higher likelihood that it initiates militarized conflict against its trade partner, given that the trade partner’s exit costs are low.

However, given that one state faces high exit costs, the associated conflict propensity will decrease as the other state’s exit costs increase because, in this case, the complementarily aggravating influences of dependence cancel out. A state facing high exit costs is less likely to fear coercion when its trade partner also has high exit costs, given that it holds equivalent leverage over its trade partner. Similarly, in this case, the conquest-tempting nature of high exit costs is reduced because the potential spoils must be balanced against potential losses accrued if a state’s trade partner is the one to capture the resources on which it is dependent. In fact, given high exit costs for one state, similarly high exit costs for its trade partner suggest that both states have a lot to lose by terminating trade and little reason to coerce the other side, such that prospects for peace are higher than if neither state faced any exit costs at all. Although attempted conquest of strategic commodities is always an option, there exists, with mutually high exit costs, a peaceful alternative to attempted conquest, as trade in this case provides mutual gains without spurring political coercion, and without provoking the uncertainty and costliness inherent in military conflict (e.g., Rosecrance 1986). When both states face high exit costs, even trade in commodities lacking strategic value, if terminated, involves a loss of income and wellbeing for both sides. As such, Montesquieu’s argument regarding buyers and sellers extends to state relationships. In short, peace is the natural effect of trade when both trade partners rely on dyadic trade. This argument leads to my third hypothesis:

Hypothesis 3: The aggravating influence of each state’s exit costs diminishes as its trade partner’s exit costs are held at higher levels, becoming associated with lower levels of conflict when its trade partner's exit costs are very high.

Research Design

To test my hypotheses, I code exit costs for each member of a directed dyad conditional on those of the other dyad member. My unit of analysis is the directed dyad year, and my analysis spans from 1985 to 2001.⁵ In order to facilitate comparison of my results to Crescenzi’s (2003, 2005), and to address disagreement regarding the appropriateness of using militarized interstate dispute (MID) initiation to address the extent of dyadic conflict (see Pevehouse 2004), I code dependent variables (DV) using events data⁶—specifically the World Events Interaction Survey (WEIS) and the Integrated Data for Events Analysis (IDEA) data sets.⁷ Using Crescenzi’s (2005) classification, I divide dyadic events into (1) high conflict, which is roughly analogous to MID initiation; (2) low conflict, which captures economic exit or threats thereof; and (3) status quo, associated with cooperation and routine dyadic events.⁸ I aggregate these events data such that I have one observation of high-conflict initiations and one observation of low-conflict initiations per directed dyad year. Given that my DVs are counts, I use zero-inflated negative binomial models.⁹

In all models, I cluster standard errors by the nondirected dyad to account for nonindependence by country pairs. Furthermore, I include explanatory variables for peace years, peace years squared, and peace years cubed to account for duration dependence inherent in trade and conflict studies (Carter and Signorino 2010), and I lag all explanatory variables by one year (except for the aforementioned peace years variables) to mitigate simultaneity bias. Given that lagged explanatory variables may not be sufficient to preclude endogeneity with regard to the trade–conflict relationship (Keshk, Pollins, and Reuveny 2004), I also ran several robustness checks using Keshk’s (2003) simultaneous equation model package for Stata. Results of these models (which are available by request from the author) suggest that my research design does not suffer from endogeneity bias.

Operationalizing Exit Costs

To capture the conditional extent of exit costs associated with trade, I create yearly measures of state A’s exit costs with respect to trade with state B and state B’s exit costs with respect to trade with state A, as well as an interaction of each state’s exit costs. Each state’s exit cost measures incorporate trade elasticities (both import demand and export supply, for imports and exports, respectively) along with an indicator of trade interaction.¹⁰

A few notable studies have attempted to measure exit costs directly, typically utilizing trade elasticities—and, specifically, the price elasticity of demand for imports—to do so (Polachek and McDonald 1992; Polachek and Seiglie 2006; Crescenzi 2003, 2005; Maoz 2009). A state’s price elasticity of demand for imports is the response in quantity demanded of a given commodity given a change in its price. If, for example, a 1 percent increase in the import price for a given commodity leads to a 1 percent decrease in imports, then the commodity is considered to have a unit elastic demand. Given this change in price, a less than 1 percent decrease in imports suggests that demand is inelastic while a greater than 1 percent decrease in imports suggests that demand is elastic.¹¹ The more elastic the demand, the easier it is to redirect trade to alternate markets, produce the good domestically, or simply endure lower availability.¹² Importantly, a more inelastic demand is a double-edged sword, suggesting that a given trade tie is valuable, and yet also conveying that the exporter holds economic means of political coercion against the importer. Both the positive and the negative aspects of inelastic demand increase as trade volume increases; hence, the attribution of various terms of positive or negative connotation (e.g., trade gains or vulnerability) to this situation.

However, the price elasticity of import demand provides only half of the information necessary to determine exit costs for dyadic trade. I also utilize export supply elasticities that capture the change in quantity supplied of a given commodity given a change in its price. Extant studies typically employ the unrealistic assumption that export supply is perfectly elastic, meaning that exporters can always redirect lost trade essentially with zero cost.¹³ However, if export supply elasticities vary, then we are unable to understand exit costs looking at import elasticities alone.

This concept can be illustrated using Ricardo’s ([1817] 1981) famous example in which England exports cloth to Portugal and Portugal exports wine to England. If England’s import demand for wine is more elastic than Portugal’s import demand for cloth, extant models would suggest that, holding trade volumes for cloth and wine equal, England will have lower exit costs for dyadic trade. However, if Portugal’s export supply for wine is elastic and England’s export supply for cloth is inelastic, then total exit costs may in fact be equal, because England’s ease of obtaining wine elsewhere is matched by Portugal’s ease of finding alternate markets for its wine exports, while, simultaneously, both states face relatively more difficulty replacing their lost cloth trade.

To illustrate further the relationship between elasticities and exit costs, Figure 1, presenting two partial equilibrium graphs of export supply and import demand for a commodity exported by state A to state B, demonstrates that exit costs can vary considerably given some level of trade interaction. Quantity traded is on the x-axis, while price is on the y-axis. The upward sloping lines reflect state A’s export supply while downward sloping lines reflect B’s import demand. Price elasticities are reflected in the slopes of each line; more elastic import demand and export supply are represented by flatter (more horizontal) lines. Trade gains—and hence exit costs—for state A (the exporter) are represented by the triangle below the price line, while B’s gains/exit costs are represented by the triangle above the price line. The graph on the left demonstrates similar elasticities, and, therefore, balanced exit costs (for trade in this one commodity). Conversely, in the graph on the right, import demand is more elastic while overall price and quantity is unchanged, leading to a case in which state A’s (the exporter’s) gains, and, therefore, exit costs, are higher than state B’s.

Figure 1.

Illustration of elasticities and exit costs.

Estimation of Trade Elasticities

For each state, I create import demand and export supply elasticities by commodity, using bilateral trade data provided by Feenstra et al. (2005), which is disaggregated to the Standard International Trade Classification four-digit (SITC4) commodity level. Although there are over six million observations spanning 1984–2000, there are relatively few observations by dyad, year, and commodity necessary to estimate elasticities in a manner consistent with the typical methods of economists.¹⁴ My theoretical model holds elasticities to vary by trading state and commodity, irrespective of the trade partner, such that I capture the effects of supply and demand for commodities rather than the confounding influence of dyadic political relationships. As such, I utilize fixed effects regressions for each state (both as importer and as exporter), for each commodity (at the SITC2-digit level).¹⁵ I run a total of 19,459 regressions (10,146 for states as importers and 9,313 for states as exporters), representing 157 states and an average of 64 SITC2-level commodities traded per state.¹⁶ The specification of these models follows.

To calculate the price elasticities of supply and demand for a given state, and for a given commodity, I regress the (natural log of the) value of dyadic trade flow (imports or exports) on the (natural log of the) unit value¹⁷ from the given trade partner and the (natural log of the) average unit value of the commodity from all other countries, as well as a year counter to account for trending.¹⁸ I include only two explanatory variables because, with my fixed effects design, I can include only variables that vary by commodity and trade partner. Although precluding the inclusion of important determinants of trade—such as dyadic distance, relative development, and regime types—the use of fixed effects accounts for unit heterogeneity by trade partner, returning estimates for trade elasticities reflecting the behavior of the actual commodity for each trading state, rather than some confounding political variable.

Each trade partner is both an importer and an exporter, so both the import and the export equations by commodity are run on all states. The fixed-effects regression for import elasticities, run for each importer (state i within dyad ij), by commodity, is specified as

l n i m p o r t v a l u e_{j} = β_{0} + β_{1} \times ln u n i t v a l u e_{j} + β_{2} \times ln u n i t v a l u e_{j} + β_{3} \times y e a r + a_{j} + u_{j} .

Where import value_j , the DV, is the value of state i’s imports from the exporter j. The primary independent variable is unit value_j , the unit value of the commodity from the exporter j; third-party value_j is the average unit value of the commodity from all exporters other than j—as such, I capture the response of quantity imported from a given country of origin due to price controlling for the price of alternatives; year_t is a trend counter spanning from 1984 to 2000; a_j is the unit (exporter) specific error; and u_j is random error.

The regression for export elasticities, run for each exporter (again, state i within dyad ij, as each state is typically both an importer and an exporter), by commodity, is specified as:

\begin{aligned} l n e x p o r t v a l u e_{j} & = β_{0} + β_{1} \times ln u n i t v a l u e_{j} + β_{2} \times ln t h i r d p a r t y v a l u e_{j} + β_{3} \\ \times y e a r + a_{j} + u_{j} . \end{aligned}

Where export value_j , the DV, is the value of state i’s exports to the importer j. The primary independent variable is unit value_j , the unit value of the commodity to the importer j; third-party value_j is the average unit value of the commodity to all importers other than j; year_t is a trend counter spanning from 1984 to 2000; a_j is the unit (importer) specific error; and u_j is random error.

The raw elasticity measures are simply β₁ from each regression. However, because these variables have large ranges, I standardize them from 0 to 1, with zero as the most elastic (i.e., demand adjusts down and supply adjusts up the most as price increases) and 1 as the most inelastic (i.e., demand and supply adjust the least due to price changes). Furthermore, the largest and smallest elasticities represent severe outliers, so I set raw elasticities smaller than the 5th percentile to 0, and raw elasticities larger than the 95th percentile to 1, with raw elasticities between standardized appropriately. As such, my standardized measures avoid extreme skewness at both low and high extremes. Ultimately, my standardized elasticities serve as weights for measures of trade interaction; when supply or demand is elastic, trade is weighted down—potentially to zero (although this is rare), while, at the most inelastic supply and demand, trade interaction retains its full value.

Creating the Commodity-Level Exit Costs Measure

The elasticities derived above capture the rate at which exit costs accrue. These elasticities must be combined with measures of trade interaction in order to capture the cost associated with terminating existing dyadic trade. To ensure the robustness of my results, I utilize two operationalizations of exit costs. The first uses trade share as the measure of interaction, consistent with Crescenzi’s operationalization. The second utilizes trade as a percentage of GDP (a.k.a. dependence), to provide additional confidence that resulting measures of exit costs are valid (see Gartzke and Li [2003] Barbieri and Peters [2003] for a useful exchange on measuring trade).¹⁹ By importer, exporter, commodity, and year, I multiply the elasticities (which vary by state and commodity, but are constant over time) with the importer’s (exporter’s) trade interaction measure (share or dependence, respectively).²⁰ I add each state’s commodity-level import and export exit costs with a given dyadic partner together,²¹ obtaining two variables per directed dyad, per commodity: A’s exit costs and B’s exit costs. Specifically,

C_{i j c t} = (e_{m i c} \times m_{i j c t}) + (e_{x i c} \times x_{i j c t}),

where C_ijct is the exit costs of state i on state j, for commodity c, and at time t; e_mic is the import price elasticity of demand of importer i, for commodity c; m_ijct is the interaction measure (share or dependence) with regard to state i, from exporter j, for commodity c, and at year t; e_xic is the export elasticity of supply for exporter i, for commodity c; and x_ijct is the interaction measure (share or dependence) with regard to state i, to importer j, for commodity c, and year t.

Operationalizing Aggregate Dyadic Exit Costs

To create yearly measures of exit costs for state i with regard to trade with state j and vice versa, I sum the dyadic, commodity-level exit costs measures presented above. The final measure takes into account the total gains that each state receives from its dyadic trade partner in a given year. Specifically, for each dyad member,

E_{i j t} = \sum_{1}^{k} C_{i j c t},

where E_ijt is the final measure of a state’s exit costs with a given dyadic partner; and C_ijct is the commodity-level exit cost measure (for commodities 1 to k) calculated above.²² For the final measures of exit costs, I include variables for A’s (the potential initiator’s) exit costs and B’s (the potential target’s) exit costs.²³ Furthermore, given the conditional nature of my hypotheses, I create an exit cost interaction. The coefficients for the components represents the effect of each state’s increasing exit costs given that the other state’s exit costs are held at zero, addressing Hypotheses 1 and 2. The interaction effects, which must be interpreted using the components and interaction term, address Hypothesis 3 (Braumoeller 2004; Brambor, Clark, and Golder 2006; Kam and Franzese 2007).

Figure 2 illustrates the construction of the share-based and dependence-based exit cost measures. In both measures, commodity-level exit costs are added together to create aggregate measures; however, the implications of this coding decision vary by measure. Mathematically, the dependence-based measure is a weighted version of the typical interaction measure of trade/GDP. In fact, if all import and export elasticities for a dyad member were equal to 1 (the most inelastic), then the exit cost measure would be identical to the common interaction measure. The trade share-based measure is distinct, however, because I do not divide the summed value by the number of commodities traded within dyads. To do so would generate an average exit costs measure, whereas I am more interested in total exit costs.²⁴

Figure 2.

Construction of trade share and trade/GDP (i.e., dependence) exit cost measures.

This coding can be illustrated with an example in which state A imports steel and aluminum from state B while exporting wheat to B. Assume that each of these three traded commodities represents equal trade share for both states. If all three of these commodities have equally inelastic demand but perfectly elastic supply, then state A should, all else equal, have twice the exit costs of j because it imports twice value of costly-to-replace commodities. However, if I divide by the number of commodities traded when calculating share-based exit costs, the resulting measure of average exit costs would rate these two importers as having identical exit costs. The share-based index presented here records state A as having exit costs equal to twice those of state B.

Finally, given that commodity-level data provide detailed information regarding what states trade as well as how much they trade,²⁵ I run additional models on two alternate exit cost variables: one capturing exit costs for trade only in “strategic” commodities, and one capturing exit costs only for trade only in “nonstrategic” commodities. Strategic commodities are defined as fuels, iron and steel, industrial machinery, and arms.²⁶ This distinction between strategic and nonstrategic commodities is preliminary, yet it provides some insight into whether commodities that are more easily transferable into military and industrial power are more prone to the causal mechanisms illustrated above. The case of the United States and Japan was almost certainly one where the strategic nature of commodities (oil and steel used to build and fuel warships, armored vehicles, and planes) was critical in determining whether those states would resort to militarized conflict. Yet, high exit costs for trade in nonstrategic commodities may nonetheless represent significant potential for lost income, even if this trade consists of seemingly innocuous commodities (e.g., in textiles or consumer electronics, as constitutes a large portion of United States—China trade today).

Additional Explanatory Variables

In models utilizing exit costs measures derived from trade shares, I control for the extent of dyadic trade in order to offset bias that might result if a state conducts a high share of its trade in a number of commodities with a given partner, yet conducts little trade with that partner overall. Specifically, I control for the (logged) dyadic trade flow (from Feenstra et al. 2005). I control also for the (natural log of the) lower GDP (from Gleditsch 2002) within the dyad.²⁷

In all models, I control for typical correlates of conflict, which may also influence trade levels and the overall frequency of dyadic events. Specifically, I control for democracy, distance, alliance similarity, and the dyadic capability ratio. I code democracy as an interaction of each state’s combined democracy–autocracy score (rescaled from 0 to 20) from the Polity IV project (Marshall and Jaggers 2009). Democracy is associated with higher likelihood of conflict and larger trade volumes. I code the log of distance (plus one) from EUGene (Bennett and Stam 2000a), given that the opportunity for both trade and conflict increases with proximity. Furthermore, I use this variable to control for zero inflation. I code alliance similarity using Signorino and Ritter’s (1999) global weighted S score, to account for the extent to which the relationship between trade and conflict may actually result from similar foreign policy preferences (Gartzke 1998). Finally, I code the dyadic capability ratio—defined as A’s composite indicator of national capabilities (CINC) score divided by the sum of A’s and B’s CINC scores—from EUGene.

Analysis

I report the results of twelve models. Table 1 contains models 1 through 6, which present coefficients for zero-inflated negative binomial regressions on high-level and low-level conflict initiations, utilizing trade share-based exit costs measures. Table 2 presents models 7 through 12, replicating models 1 through 6 with exit cost measures derived from trade as a percentage of GDP. Results of these models support all three hypotheses, suggesting that higher exit costs for one state are associated with higher propensity for conflict initiation when its trade partner’s exit costs are held at zero, but that this aggravating influence diminishes as the trade partner’s exit costs increase. Furthermore, I find the greatest support for Hypothesis 3 for exit costs associated with trade in strategic commodities. When these costs are the most asymmetric, they are associated with a large increase in conflict initiations, whereas the marginal effect for each state’s exit costs becomes negative and significant when its partner’s exit costs are held at high values. Substantively, when both states face high exit costs in strategic goods, the likelihood of conflict initiation falls below the baseline case (of no exit costs for either state) by more than 90 percent.

Table 1.

Exit Costs and Conflict Initiation Counts 1985–2001, Trade Share-based Exit Cost Measure.

	All commodities		Strategic commodities		Nonstrategic commodities
	1: count of “high conflict” events	2: count of “low conflict” events	3: count of “high conflict” events	4: count of “low conflict” events	5: count of “high conflict” events	6: count of “low conflict” events
A’s exit costs	0.693***	0.514***	1.163***	1.085***	0.806***	0.593***
A’s exit costs	(0.088)	(0.073)	(0.111)	(0.078)	(0.092)	(0.071)
B’s exit costs	0.657***	0.562***	1.023***	1.064***	0.770***	0.640***
B’s exit costs	(0.086)	(0.071)	(0.110)	(0.075)	(0.085)	(0.071)
A exit costs × B exit costs	−0.138***	−0.062*	−0.437***	−0.345***	−0.168***	−0.083**
A exit costs × B exit costs	(0.034)	(0.029)	(0.069)	(0.054)	(0.034)	(0.030)
ln trade flow	0.044	0.061	−0.034	−0.033*	−0.017	0.022
ln trade flow	(0.041)	(0.037)	(0.019)	(0.014)	(0.043)	(0.038)
Minimum GDP	0.136*	0.336***	0.263***	0.469***	0.180**	0.369***
Minimum GDP	(0.055)	(0.044)	(0.052)	(0.044)	(0.056)	(0.047)
ln distance	−0.035	−0.124**	−0.105**	−0.170***	−0.033	−0.127***
ln distance	(0.032)	(0.039)	(0.037)	(0.037)	(0.031)	(0.038)
Capability ratio	0.371**	−0.134	0.464***	−0.038	0.369**	−0.143*
Capability ratio	(0.138)	(0.070)	(0.130)	(0.071)	(0.139)	(0.069)
Polity IV in A	−0.010	0.028	−0.024	0.017	−0.004	0.034*
Polity IV in A	(0.019)	(0.017)	(0.018)	(0.016)	(0.019)	(0.017)
Polity IV in B	−0.006	0.021	−0.018	0.006	0.000	0.026
Polity IV in B	(0.019)	(0.017)	(0.017)	(0.016)	(0.019)	(0.017)
Polity IV A × Polity IV B	−0.002	−0.002*	−0.001	−0.001	−0.002	−0.003**
Polity IV A × Polity IV B	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
Alliance similarity	−2.121***	−2.061***	−2.140***	−2.005***	−2.161***	−2.083***
Alliance similarity	(0.289)	(0.216)	(0.297)	(0.200)	(0.289)	(0.213)
Peace years	−0.127***	−0.101***	−0.116***	−0.091***	−0.126***	−0.101***
Peace years	(0.012)	(0.012)	(0.012)	(0.011)	(0.012)	(0.012)
Peace years²	0.002***	0.001***	0.002***	0.001***	0.002***	0.001***
Peace years²	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Peace years³	−0.000***	−0.000***	−0.000***	−0.000***	−0.000***	−0.000***
Peace years³	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Constant	−3.266***	−7.104***	−3.800***	−8.000***	−3.571***	−7.341***
Constant	(0.974)	(0.682)	(1.018)	(0.719)	(0.981)	(0.700)
Inflation parameters
ln distance	0.321***	0.381***	0.338***	0.436**	0.321***	0.392***
ln distance	(0.044)	(0.102)	(0.052)	(0.169)	(0.044)	(0.114)
Constant	−2.185***	−6.072*	−2.441***	−6.906	−2.203***	−6.229*
Constant	(0.308)	(2.659)	(0.387)	(3.641)	(0.311)	(2.838)
ln alpha	2.777***	2.448***	2.892***	2.455***	2.783***	2.457***
ln alpha	(0.172)	(0.110)	(0.169)	(0.0953)	(0.171)	(0.106)
Observations	165,006	165,006	165,006	165,006	165,006	165,006
Prob χ²	≤0.0001	≤0.0001	≤0.0001	≤0.0001	≤0.0001	≤0.0001

Note: GDP = gross domestic product.

Robust standard errors in parentheses.

*p < .05, **p < .01, ***p < .001; two-tailed tests.

Table 2.

Exit Costs and Conflict Initiation Counts 1985–2001, Trade “Dependence”-based Exit Cost Measure.

	All commodities		Strategic commodities		Nonstrategic commodities
	7: count of “high conflict” events	8: count of “low conflict” events	9: count of “high conflict” events	10: count of “low conflict” events	11: count of “high conflict” events	12: count of “low conflict” events
A’s exit costs	0.520***	0.641***	1.115**	1.645***	0.658***	0.856***
A’s exit costs	(0.123)	(0.096)	(0.366)	(0.346)	(0.171)	(0.137)
B’s exit costs	0.365***	0.588***	0.562**	1.432***	0.491***	0.779***
B’s exit costs	(0.083)	(0.082)	(0.194)	(0.337)	(0.121)	(0.112)
A exit costs × B exit costs	−0.080***	−0.106***	−0.787*	−1.364***	−0.124***	−0.170***
A exit costs × B exit costs	(0.018)	(0.016)	(0.306)	(0.349)	(0.031)	(0.026)
ln distance	−0.109**	−0.165***	−0.092**	−0.150***	−0.107**	−0.173***
ln distance	(0.036)	(0.034)	(0.035)	(0.032)	(0.035)	(0.035)
Capability ratio	0.328**	−0.129*	0.309**	−0.104	0.290**	−0.140*
Capability ratio	(0.104)	(0.057)	(0.097)	(0.057)	(0.104)	(0.056)
Polity IV in A	−0.016	0.014	−0.021	0.009	−0.010	0.020
Polity IV in A	(0.017)	(0.016)	(0.017)	(0.016)	(0.017)	(0.016)
Polity IV in B	−0.005	0.007	−0.012	0.002	0.001	0.013
Polity IV in B	(0.017)	(0.017)	(0.017)	(0.017)	(0.017)	(0.017)
Polity IV A × Polity IV B	0.001	0.001	0.002	0.003*	−0.000	0.001
Polity IV A × Polity IV B	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
Alliance similarity	−2.325***	−2.412***	−2.399***	−2.462***	−2.347***	−2.441***
Alliance similarity	(0.242)	(0.206)	(0.243)	(0.208)	(0.237)	(0.199)
Peace years	−0.114***	−0.087***	−0.108***	−0.082***	−0.116***	−0.087***
Peace years	(0.011)	(0.010)	(0.011)	(0.010)	(0.011)	(0.010)
Peace years²	0.002***	0.001***	0.002***	0.001***	0.002***	0.001***
Peace years²	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Peace years³	−0.000***	−0.000***	−0.000***	−0.000***	−0.000***	−0.000***
Peace years³	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Constant	1.577***	1.343***	1.723***	1.433***	1.609***	1.394***
Constant	(0.416)	(0.380)	(0.418)	(0.373)	(0.406)	(0.373)
Inflation parameters
ln distance	0.457***	0.534***	0.478***	0.515***	0.462***	0.547***
ln distance	(0.0697)	(0.0794)	(0.0611)	(0.0621)	(0.0717)	(0.0849)
Constant	−2.670***	−4.313***	−2.544***	−3.798***	−2.700***	−4.458***
Constant	(0.561)	(0.680)	(0.481)	(0.512)	(0.580)	(0.738)
ln alpha	2.818***	2.435***	2.680***	2.339***	2.813***	2.459***
ln alpha	(0.167)	(0.0983)	(0.162)	(0.103)	(0.169)	(0.102)
Constant	165,026	165,026	165,026	165,026	165,026	165,026
Prob χ²	≤0.0001	≤0.0001	≤0.0001	≤0.0001	≤0.0001	≤0.0001

Note: Robust standard errors in parentheses.

*p < .05, **p < .01, ***p < .001; two-tailed tests.

Table 1 presents coefficients for zero-inflated negative binomial regressions estimating conflict initiation counts utilizing exit cost measures derived from trade share.²⁸ Models 1 and 2 test Hypotheses 1 through 3 on the count of high-level and low-level conflict initiations, including measures of exit costs aggregated from all trade within the dyad. Models 3 and 4 replicate these tests, looking only at exit costs in strategic commodities, while models 5 and 6 examine whether results hold for trade in nonstrategic commodities.

At first glance, the coefficients for each state’s exit costs, as well as for the interactions, look comparable across all six models presented in Table 1. Specifically, the coefficient for each state’s exit costs is positive and highly significant, suggesting that, for each state, higher exit costs are associated with higher counts of conflict initiation when trade partner exit costs are held at zero. These results provide support for Hypotheses 1 and 2. In each model, the interaction term is negative and significant, suggesting that the aggravating impact of each state’s exit costs diminishes at higher levels of trade partner exit costs. However, interaction coefficients alone are limited in explanatory power; an examination of marginal effects (e.g., Braumoeller 2004; Brambor, Clark, and Golder 2006; Kam and Franzese 2007) reveals important distinctions.

In model 1, which looks at the impact of exit costs aggregated from all trade on high-level conflict initiation, an examination of marginal effects suggest that the aggravating impact of each state’s exit cost diminishes as the exit costs of the other state rise, with the marginal effect eventually becoming negative (but statistically indistinguishable from 0). Model 2, looking at low-level conflict initiations, shows a similar pattern; however, marginal effects remain positive for both the initiator and the target, even when exit costs for the other state are at the maximum. In other words, mutually high exit costs appear associated with higher levels of low-level conflict, even relative to cases in which exit costs are equal to zero. Although consistent with my argument that the marginal effect of each state’s exit costs becomes smaller as the other state’s exit costs increase, this result suggests only partial support for Hypothesis 3, which expects a negative and significant marginal effect of one state’s exit costs when the other’s are held at a very high level. Ultimately, this result supports Crescenzi’s (2003) finding that exit costs are associated with higher levels of economic and minor political conflict.

Models 3 and 4 replicate models 1 and 2 looking only at exit costs for strategic commodities (e.g., fuel, iron and steel, industrial machinery, and arms). In model 3, the effect of exit costs is much more striking than in the baseline model. Specifically, the marginal effect of both the initiator’s and the target’s exit costs becomes negative and statistically significant when the exit costs of the other dyad member is held at higher values, fully supporting Hypothesis 3. Model 4 looks similar to model 3, suggesting that even low-level conflict initiation is less likely as one state’s exit costs for strategic commodities are held at higher values, conditional on high exit costs for its trade partner. Models 5 and 6 again replicate models 1 and 2, this time for exit costs associated with nonstrategic commodities. In these cases, the impact of exit costs on conflict initiation looks more like the baseline case (with support for Hypotheses 1 and 2, but only partial support for Hypothesis 3), suggesting that trade in strategic commodities has a uniquely strong impact on dyadic conflict.

Figure 3 show graphically the substantive impact of exit costs on high-level conflict initiation for trade in strategic and nonstrategic commodities, respectively. In each three-dimensional graph, the x-axis represents the potential initiator’s exit costs, the z-axis represents the potential target’s exit costs, and the y-axis represents the expected count of high-level conflict initiations. The left-hand graph, derived from the results of model 3 (as discussed earlier), highlights the aggravating influence of exit costs for strategic commodities, and also shows a stark difference in the effect of A’s exit costs and B’s exit costs, conditional on the exit costs of the other dyad member. Specifically, this graph shows that the highest expected conflict occurs when exit costs are most asymmetric. When the potential initiator’s exit costs are held at the maximum, while the potential target’s exit costs are held at zero, the expected count of high-level conflict initiations is equal to 5.2. Conversely, when the asymmetry is reversed (and it is the potential conflict target facing high exit costs), the expected count of high-level conflict initiations is equal to 2.2. This distinction in impact fits with the case of World War II, in which Japan—the state with high exit costs, initiated violent conflict against the United States and its allies, rather than vice versa. Importantly, however, this graph shows that the expected count of high-level conflict initiations is equal to 0.0003 when both states’ exit costs are held at the maximum. This expected count is 94 percent lower than when both states face no exit costs (in which case the expected count is equal to 0.005), providing support for Hypothesis 3 that each state’s exit costs become pacifying as the other state’s exit costs increase.

Figure 3.

Exit costs and expected high-level conflict counts (from models 3 and 5)

The right-hand graph of Figure 3, derived from model 5, shows that exit costs for trade in nonstrategic commodities have a much smaller substantive impact, as the expected count of high-level conflict initiations is approximately 0.4 when exit costs are most unbalanced, regardless of whether the initiator or target faces high exit costs. This expected count falls to 0.1 when both states’ exit costs are held at the maximum; yet, this value is larger than the expected count when neither state faces any exit costs (0.005)—illustrating that, with regard to nonstrategic commodities, I find support for the direction of the interaction effect posited in Hypothesis 3, but not for the expectation of a pacifying impact of one state’s exit costs when partner exit costs are very high. Importantly, however, this graph suggests, in accordance with Hypothesis 3, that the aggravating impact of one state’s high exit costs for nonstrategic commodities declines as its partner’s exit costs increase.

Given that Figure 3 is based on abstract examples, I use the case of the United States and China as a concrete demonstration of the impact of exit costs. My data show that China faces higher exit costs than the United States overall (4.86 relative to 3.96 on the share-based index) as well as with regard to trade in strategic commodities (3.14 relative to 1.85 on the share-based index). Model 3 (looking at trade in strategic commodities) suggests that the expected count of high-level conflict events for the United States and China (with the United States as the initiator) is 1.8. However, if US exit costs were decreased to 0, this expected count would increase by 50 percent, to 2.7. Similarly, if China’s exit costs doubled while US exit costs remain at current levels, the expected count of high-level conflict events would nearly double, to 3.5. Conversely, if US exit costs increased to match China’s current values, the expected count of high-level conflict events would fall by approximately one-third, to 1.3.

The results of Table 2, replicating Table 1 using exit costs derived from trade as a percentage of GDP are essentially identical to those obtained from Table 1. As such, I omit a detailed discussion of this table due to space considerations. These results are available by request.

The fact that my results look similar to Barbieri’s (1996) finding that asymmetric trade share is aggravating suggests the question: Do my results support the argument positing a conditional impact of exit costs, or is it simply uncovering a conditional impact of trade interaction (relative to total trade or GDP)? Given that rational individuals should trade specifically when they stand to gain from it, one can imagine a scenario in which exit costs correlate with measures of trade interaction. However, further analysis suggests that this is not the case. Correlations of my exit cost index to typical measures of trade interaction—specifically, trade share and dependence—are quite low (e.g., my share-based exit costs index correlates at 0.21 with trade share, and at 0.16 with trade dependence). Furthermore, in models 1 through 6, there is little evidence that trade flow is pacifying. Finally, I find null results in models replacing exit cost interaction with an interaction of each state’s trade share or trade as a percentage of GDP. Overall, my results suggest that trade interaction alone is not associated with dyadic conflict—particularly violent conflict.

Space limitations preclude detailed analysis of control variables. However, one notable pattern arises: democracy variables are not statistically significant while the coefficient for alliance similarity is negative and highly significant. This result supports prior studies suggesting that preference similarity is more pacifying than regime type (Gartzke 1998). However, another explanation for this finding is that missing data for exit costs cause the unusual results for democracy (e.g., Dafoe 2011), bias that could also affect alliance similarity. Future research may benefit by exploring in detail these relationships.

Conclusion and Discussion

In this article, I find strong evidence of a conditional relationship between exit costs and the initiation of dyadic conflict. Unbalanced exit costs—particularly when the potential conflict initiator has high exit costs—are associated with a higher count of conflict events; yet, this aggravating influence disappears—and in some cases, reverses—as joint exit costs increase. Furthermore, the impact of exit costs on conflict initiation is more prominent when looking specifically at trade in strategic commodities.

Given my results, how might one explain the perceived pacifying effect of trade that is found in so many studies? The simplest explanation is that states attempt to structure trade relationships to avoid unilateral dependence, such that the conditions associated with the highest expectation of conflict (according to my mechanisms) are rare. Yet, given that states are not always immediately willing or able to control trade by their citizens, the conditions favoring conflict are bound to arise at least occasionally.²⁹

This article demonstrates that there are many opportunities in which to examine the impact of trade vulnerability on international politics. First, this article only begins to assess the relationship between exit costs and conflict. For example, future research can better assess the impact of individual commodities on dyadic conflict, perhaps by better distinguishing commodities that are considered “strategic,” as this classification likely varies by dyad. Additionally, future research can utilize the commodity-level exit cost measures developed for this article to address a wide array of research questions. For example, studies looking at protectionism can control for the gains from trade associated with imports. Also, research on sanctions can utilize these data to better determine the true cost of sanctions, both to the sender and to the target, potentially better isolating cases in which sanctions or threats thereof, through higher imposed costs to the target, are more likely to succeed in changing policy of sanctioned states.

Footnotes

Acknowledgments

This article was presented previously at the 2009 International Studies Association—Midwest meeting in Saint Louis. I would like to thank the panel participants for their helpful comments. I would also like to thank Cooper Drury and JCR’s anonymous reviewers for invaluable suggestions. Replication data and do files are available at .

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Angell

Norman

. 1913. The Great Illusion: A Study of the Relation of Military Power to National Advantage. London: Heinemann.

Bahmani-Oskoee

Mohsen

. 1998. “Cointegration Approach to Estimate the Long-run Trade Elasticities in LDCs.” International Economic Journal 12 (3): 89–96.

Bahmani-Oskoee

Mohsen

Niroomand

Farhang

. 1998. “Long-run Price Elasticities and the Marshall–Lerner Condition Revisited.” Economic Letters 61 (1):101–109.

Barbieri

Katherine

. 1996. “Economic Interdependence: A Path to Peace or a Source of Conflict?” Journal of Peace Research 33 (1): 29–49.

Barbieri

Katherine

Peters II

Richard Alan

2003. “Measure for Mis-measure: A Response to Gartzke and Li.” Journal of Peace Research 40 (6):713–19.

Barbieri

Katherine

Keshk

Omar

Pollins

Brian

. 2008. Correlates of War Project Trade Data Set Codebook, Version 2.0. Accessed March 5, 2010. http://correlatesofwar.org.

Bennett

Stam

. 2000a. “EUGene: A Conceptual Manual.” Interaction Interactions 26 (2): 179–204.

Bennett

Stam

. 2000b. “Design and Estimator Choices in the Analysis of Interstate Dyads: When Decisions Matter.” Journal of Conflict Resolution 44 (5): 653–85.

Brambor

Thomas

Clark

William Roberts

Golder

Matt

. 2006. “Understanding Interaction Models: Improving Empirical Analysis.” Political Analysis 14 (1): 63–82.

10.

Braumoeller

Bear

. 2004. “Hypothesis Testing and Multiplicative Interaction Terms.” International Organization 58:807–20.

11.

Carter

David B.

Signorino

Curtis S.

. 2010. “Back to the Future: Modeling Time Dependence in Binary Data.” Political Analysis 18 (3): 271–92.

12.

Crescenzi

Mark J. C.

2003. “Economic Exit, Interdependence, and Conflict.” Journal of Politics 65 (3): 809–32.

13.

Crescenzi

Mark J. C.

2005. Economic Interdependence and Conflict in World Politics. Lanham, MD: Lexington Books.

14.

Dafoe

Allan

. 2011. “Statistical Critiques of the Democratic Peace: Caveat Emptor.” American Journal of Political Science 55 (2): 247–62.

15.

Dorussen

Han

. 2006. “Heterogeneous Trade Interests and Conflict: What you Trade Matters.” Journal of Conflict Resolution 50 (1): 87–107.

16.

Drezner

Daniel W.

1998. “Conflict Expectations and the Paradox of Economic Coercion.” International Studies Quarterly 45 (4): 709–31.

17.

Fearon

James D.

1995. “Rationalist Explanations for War.” International Organization 49 (Summer): 379–414.

18.

Feenstra

Robert

Lipseym

Robert E.

Deng

Haiyan

Alyson C.

Hengyong

. 2005 “World Trade Flows: 1962-2000.” Working Paper No. 11040. Accessed March 5, 2010. http://www.nber.org/papers/w11040

19.

Feis

Herbert

. 1950. The Road to Pearl Harbor. Princeton, NJ: Princeton University Press.

20.

Gartzke

Erik

. 1998. “Kant We All Just Get Along? Opportunity, Willingness, and the Origins of the Democratic Peace.” American Journal of Political Science 42 (1): 1–27.

21.

Gartzke

Erik

Quan

. 2003. “Measure for Measure: Concept Operationalization and the Trade Interdependence: Conflict Debate.” Journal of Peace Research 50 (5): 553–71.

22.

Gartzke

Erik

Quan

Boehmer

Charles

. 2001. “Investing in Peace: Economic Interdependence and International Conflict.” International Organization 55 (2): 391–438.

23.

Gleditsch

Kristian

. 2002. “Expanded Trade and GDP Data.” Journal of Conflict Resolution 46 (5): 712–24.

24.

Goldstein

Joshua S.

1992. “A Conflict-Cooperation Scale for WEIS Events Data.” The Journal of Conflict Resolution 36 (2): 369–85.

25.

Hirschman

Albert O.

1945. National Power and the Structure of Foreign Trade. Berkeley: University of California Press.

26.

Houthakker

H. S.

Magee

Stephen P.

. 1969. “Income and Price Elasticities in World Trade.” The Review of Economics and Statistics 51 (2): 111–25.

27.

Hufbauer

Gary Clyde

Schott

Jeffrey J.

Elliott

Kimberly Ann

Oegg

Barbara

. 2007. Economic Sanctions Reconsidered: History and Current Policy. 3rd ed. Washington, DC: Institute for International Economics.

28.

Kam

Cindy D.

Robert J. Franzese

. 2007. Modeling and Interpreting Interactive Hypotheses in Regression Analysis. Ann Arbor: The University of Michigan Press.

29.

Kee

Hiau Looi

Nicita

Alessandro

Olarreaga

Marcelo

. 2008. “Import Demand Elasticities and Trade Distortions.” The Review of Economics and Statistics 90 (4): 666–82.

30.

Keohane

Robert O.

Nye

Joseph S.

. 1977. Power and Interdependence: World Politics in Transition. Boston: Little, Brown.

31.

Keshk

Omar M. G.

2003. “CDSIMEQ: A Program to Implement Two-stage Probit Least Squares.” The Stata Journal 3 (2): 1–11.

32.

Keshk

Omar M. G.

Pollins

Brian M.

Reuveny

Rafael

. 2004. “Trade Still Follows the Flag: The Primacy of Politics in a Simultaneous Model of Interdependence and Armed Conflict.” Journal of Politics 66 (4): 1155–79.

33.

Lektzian

David J.

Sprecher

Christopher M.

. 2007. “Sanctions, Signals, and Militarized Conflict.” American Journal of Political Science 51 (2): 415–31.

34.

Liberman

Peter

. 1993. “The Spoils of Conquest.” International Security 18 (2): 125–53.

35.

Liberman

Peter

. 1996. Does Conquest Pay? The Exploitation of Occupied Industrial Societies. Princeton, NJ: Princeton University Press.

36.

Mansfield

Edward D.

Pollins

Brian M.

. 2001. “The Study of Interdependence and Conflict: Recent Advances, Open Questions, and Directions for Future Research.” Journal of Conflict Resolution 45 (6): 834–59.

37.

Maoz

Zeev

. 2009. “The Effects of Strategic and Economic Interdependence on International Conflict Across Levels of Analysis.” American Journal of Political Science 53 (1): 223–40.

38.

Marshall

Monty G.

Jaggers

Keith

. 2010. “Polity IV Project: Political Regime Characteristics and Transitions, 1800-2009.” Accessed March 5, 2010. http://www.systemicpeace.org/polity/polity4.htm.

39.

Marquez

Jaime

. 1990. “Bilateral Trade Elasticities.” The Review of Economics and Statistics 72 (1): 70–77.

40.

Marquez

Jaime

. 2002. Estimating Trade Elasticities. Dordrecht, the Netherlands: Kluwer Academic Publishers.

41.

Morrow

James D.

1997. “When do ‘Relative Gains’ Impede Trade?” Journal of Conflict Resolution 41 (1): 12–37.

42.

Morrow

James D.

1999. “How Could Trade Affect Conflict?” Journal of Peace Research 36 (4): 481–489.

43.

Montesquieu

Charles de Secondat

. (1750) 1989. The Spirit of the Laws. New York: Cambridge University Press.

44.

Oneal

John R.

Russett

Bruce M.

. 1997. “The Classical Liberals Were Right: Democracy, Interdependence, and Conflict, 1950-1985.” International Studies Quarterly 41 (2): 267–93.

45.

Oneal

John R.

Russett

Bruce M.

. 2001. “Clear and Clean: The Fixed Effects of the Liberal Peace.” International Organization 55 (2): 469–85.

46.

Oneal

John R.

Russett

Bruce M.

. 2005. “Rule of Three, Let it Be: When More Really is Better.” Conflict Management and Peace Science 22 (4): 293–310.

47.

Oneal

John R.

Oneal

Frances H.

Maoz

Zeev

Russett

Bruce

. 1996. “The Liberal Peace: Interdependence, Democracy, and International Conflict, 1950-85.” Journal of Peace Research 33 (1): 11–28.

48.

Orcutt

Guy

. 1950. “Measurement of Price Elasticities in International Trade.” The Review of Economics and Statistics 32 (2): 117–32.

49.

Pevehouse

Jon C.

2004. “Interdependence Theory and the Measurement of International Conflict.” The Journal of Politics 66 (1): 247–66.

50.

Polachek

Solomon W.

1980. “Conflict and Trade.” Journal of Conflict Resolution 24 (1): 55–78.

51.

Polachek

Solomon W.

McDonald

Judith

. 1992. “Strategic Trade and the Incentive for Cooperation.” In Disarmament, Economic Conversion, and Management of Peace, edited by Chatterji

Manas

Forcey

Linda Rennie

, 273–84. New York : Praeger.

52.

Polachek

Solomon W.

Xiang

Jun

. 2010. “How Opportunity Costs Decrease the Probability of War in an Incomplete Information Game.” International Organization 64 (1): 133–44.

53.

Ricardo

David

. 1817 1981. On the Principles of Political Economy and Taxation. Cambridge, MA: Cambridge University Press.

54.

Rosecrance

Richard

1986. The Rise of the Trading State: Commerce and Conquest in the Modern World. New York: Basic Books

55.

Russett

Bruce M.

Oneal

John R.

. 2001. Triangulating Peace: Democracy, Interdependence, and International Organizations. New York: W. W. Norton.

56.

Signorino

Curtis S.

Ritter

Jeffrey M.

. 1999. “Tau-b or Not Tau-b: Measuring the Similarity of Foreign Policy Positions.” International Studies Quarterly 43 (1): 115–44.

57.

Wagner

R. Harrison

. 1988. “Economic Interdependence, Bargaining Power, and Political Influence.” International Organization 42 (3): 461–83.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

13.71 MB

0.00 MB