Not All Peace Years Are Created Equal: Trade, Imposed Settlements, and Recurrent Conflict

Abstract

The peace through trade hypothesis receives extensive support from a large empirical literature. However, extant research does not isolate whether this relationship holds for states that have fought in the past, or whether its influence following armed disputes is dependent upon the ways in which states settle their conflicts. Additionally, although recent research finds that imposed settlements are more pacifying than other forms of political settlement, these studies tend not to isolate factors associated with variation in the stability of imposed settlements. In this article, we examine how settlements condition the influence of trade on conflict recurrence, both to overcome a limitation in extant studies of trade and conflict, which tend to ignore the way states settle prior disputes, and to further an understanding of how post-conflict state interaction varies by (and within) settlement type. Looking at dyadic trade and recurrent conflict from 1885 to 2000, we find that imposed settlements foster a pacifying effect of trade, while negotiated settlements and failures to reach settlement lead to relationships in which trade has crosscutting effects on the stability of peace, resulting in an overall null effect of trade on conflict recurrence.

Keywords:

• deterrence
• imposed settlements
• recurrent conflict
• trade

A previous version of this article was presented at the 2008 annual meeting of the International Studies Association-Midwest, St. Louis, MO. We would like to thank Sara Mitchell, Cameron Thies, Cooper Drury, and Paul Bellinger for helpful comments and suggestions.

Notes

¹This argument arises from claims by Morrow (1999) that opportunity cost arguments regarding trade and conflict are invalid, given that one state's decreased incentive to challenge the status quo following from its desire to maintain trade gains should be cancelled by its trade partner's increased incentive therein, using these same gains as leverage. Morrow argues that opportunity costs therefore have ambiguous effects on conflict (but see Polachek and Xiang 2010 for a counter argument).

²See Hegre (2000, 2004) for analyses that do distinguish between dyads that have and have not fought.

³The key distinction between an imposed settlement and a negotiated settlement, both of which are generally agreed to, is that an imposed settlement occurs when one side dictates the post-conflict status quo (Maoz 1984).

⁴For much more detail on the formalization of this argument, see Senese and Quackenbush (2003, especially pages 698–703).

⁵See Table A in the appendix for a summary of trade flow and dependence by settlement type.

⁶We identify time-varying covariates in the discussion of explanatory variables.

⁷See Table B in the appendix for descriptive statistics of all explanatory variables.

⁸The analyses presented below do not change significantly if we include unclear settlements.

⁹This percentage refers to the use-of-force sample. In the all-MID sample, imposed settlements are slightly less prevalent, at 16.2% of 24,623 dyad years.

¹⁰We choose to focus on trade as a function of GDP because this measure captures trade salience to a state's economy. Alternately, Barbieri (1996) constructs measures of salience and symmetry from dyadic trade flow as a percentage of each state's total trade. Gartzke and Li (2003), referring to this measure as trade share, find that it captures a different aspect of the trade-conflict relationship. Specifically, states with higher trade shares tend to have overall lower trade as a percentage of GDP; whereas higher trade share implies vulnerability to one trading partner, higher trade openness (and, therefore, dependence) implies reliance on trade for economic prosperity.

¹¹However, care must be taken in the interpretation, as, for example, the component term for lower trade/GDP is, in isolation, not meaningful, because it represents its effect when higher trade/GDP is held at zero—which can occur only when lower trade/GDP is also held at zero. Technically, neither side's trade/GDP measure can equal zero if there is any trade flow.

¹²Results are consistent when we remove this variable.

¹³Furthermore, for both the non-directed and directed relative capability variables, we subtract 0.5 from the raw measure, such that parity takes the value of 0. This transformation facilitates analysis of survival over time (see Figures 1, 2, and 3) and does not affect any coefficients.

FIGURE 1 Survival of peace by recipient trade (from Model 6). Imposer trade/GDP held at its mean.

FIGURE 2 Survival of peace by imposer trade (from Model 6). Recipient trade/GDP held at its mean.

FIGURE 3 Survival of peace by trade (from Model 6). Both states' trade/GDP held at equivalent values.

¹⁴In robustness checks available from the authors, we estimate baseline models excluding trade variables, finding as expected (and in accordance with the extant literature) that imposed settlements are more pacifying than other settlement types.

¹⁵Interestingly, the lower trade/GDP and interaction variables violate the proportional hazards assumption in some cases, and as such, we correct for this violation by including additional variables for these components interacted with the log of analysis time, thereby modeling the changing effect of each variable over time. However, all trade variables—including these proportional hazards corrections—fail to reach statistical significance at the .05 level.

¹⁶As we mention above, neither side's trade/GDP can equal zero if there is any trade, so this coefficient should be interpreted skeptically. See Braumoeller (2004) for a discussion of the interpretation of interaction terms.

¹⁷Once again, it is not technically possible for the recipient's trade as a percentage of its GDP to be zero if there is any trade flow, but there are many cases in which the imposer relies more on trade for income than does the recipient (specifically, there are 1,327 dyad years when this is the case, relative to 2,058 dyad years in which the recipient is more dependent than the imposer, and 308 dyad years in which each state's trade/GDP is equal).

¹⁸We calculate survival probabilities using the “stcurve” post-estimation command in Stata 10. All probabilities come from Model 6. These figures are derived from point estimates rather than marginal effects, although an examination of marginal effects for the interaction falls in line with expectations. Specifically, the marginal effect for the imposer's trade/GDP suggests that it is aggravating when the recipient's trade/GDP is held at zero, but this effect diminishes, eventually becoming increasingly pacifying as the recipient's trade/GDP increases. The recipient's trade/GDP appears to have no significant effect regardless of the level at which the imposer's trade/GDP is held. Marginal effects are calculated in accordance with Braumoeller (2004; see also Kam and Franzese 2007).

¹⁹Although, increasing trade does not increase the likelihood of recurrent conflict either.

²⁰We calculate this value by dividing the coefficient for both democratic by its proportional hazards correction, and then exponentiating this quotient.

²¹Thanks to an anonymous reviewer for pointing this out.

²²Thanks to an anonymous reviewer for highlighting this concern.

²³We use Keshk's (2003) simultaneous equation model package for Stata to estimate simultaneous equations models.

²⁴It should be noted, however, that the example of the U.S. and Germany in the immediate postwar period does not influence our results, given that Germany (and Japan) drop from the international system, and therefore, from our sample, following the imposed settlement.

²⁵There may be a connection between our results and the “peace by empire” hypothesis (Aron 1966; Maoz 1984), as the major commercial empires of recent centuries—the Dutch, British, and U.S.—did not hesitate to use force to gain access to markets when necessary.