Probing Conflict Contagion and Casualties in Mindanao, Philippines

ABSTRACT

Subnational and local hostilities are regarded as a distinguishing feature of Asian conflicts. Like the violent conflicts in Indonesia, Myanmar, Sri Lanka, those in Mindanao, Philippines are noted for their persistence and enormous welfare and economic costs. Extending recent empirical studies on Asian conflicts, here we examine the importance of spatial contagion as a factor behind the local incidence of conflicts and the associated casualties in the Autonomous Region of Muslim Mindanao. Estimating negative binomial regression models on a panel of cities and municipalities for period 2011–2016, we find strong evidence of contagion: both the frequency of violent hostilities and their related casualties in a locality are affected by past conflicts, especially those related to crime, in its contiguous neighbors. No such effect is found in the placebo tests involving random neighbors in the same province, thus ruling out that the estimated contagion effects is just a correlated reaction to a common environment.

KEYWORDS:

• Conflict casualties
• spatial contagion
• negative binomial models
• Mindanao
• Philippines

JEL Codes:

• R59
• H79

Acknowledgments

I gratefully acknowledge The Asia Foundation (Bangkok Office), the UPecon-Health Policy Development Program, and the Philippine Center for Economic Development for the institutional and financial support; International Alert Philippines for the data; Xylee Javier, Erlinda Ranchez, Sylvia Nachura and Angelo Gabrielle Santos for their excellent research assistance; and, for their comments and suggestions, an anonymous referee and the participants in a workshop held on 1–2 June 2016 in Bangkok, Thailand, and in a seminar at the UP School of Economics held on 28 September 2016. I remain responsible for any and all errors.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. UCDP defines an armed conflict as a ‘contested incompatibility that concerns government and/or territory where the use of armed force between two parties, of which at least one is the government of a state, results in at least 25 battle-related deaths in a calendar year.’ Data downloaded on 1 December 2018 from http://ucdp.uu.se/downloads/.

2. For example of estimates, see Molato (2015), Tigno (2006), Adriano and Parks (2013), and Pettersson and Eck (2018).

3. Unlike those in UDCP, here the conflicts may or may not involve the government and are closer in definition to subnational and local conflicts given in Asia Foundation (2017).

4. This section and the next (Data) adapt and update up to 2016 the relevant sections in Capuno (2017). (Note that as of this writing (March 2019), the ARMM, after two plebiscites held in early 2019, was superseded by the Bangsamoro Autonomous Region of Muslim Mindanao, which now includes Cotabato City, some new barangays, and all the ARMM local government units.)

5. The IRA is the share of the LGU in the national government revenues from income and consumption taxes. The IRA is apportioned following a fixed formula based on the LGU’s level, its population share, its land area share, and the total number of LGUs in the same level (Diokno 2012). It is the single most important central fiscal transfer and revenue source for most LGUs (Llanto 2012).

6. For news accounts, see ‘Maguindanao Massacre – How it happened’ by Matikas Santos, dated 21 November 2014. http://www.inquirer.net/143183/maguindanao-massacre-how-it-happened. Or, ‘Maguindanao Massacre: One Year After’ by Ed Lingao, dated 28 December 2010. http://pcij.org/tag/maguindanao-massacre/.

7. Before 2016, Conflict Alert was named as the Bangsamoro Conflict Monitoring System. https://conflictalert.info.

8. As of 2 October 2018, Conflict Alert reports cases up to 2017. However, the conflict data for 2017 are not included in the regression analysis since the corresponding fiscal and election data for 2017 are unavailable.

9. Note that municipal-level population projections in ARMM may not be stable due to the displacement or migration following conflicts.

10. Symbols in italics and bold face refer to scalars and vectors, respectively.

11. Lagged explanatory variables, including the civil wars in neighbors (states), are also used in Sambanis (2001), Elbadawi and Sambanis (2002), and Carmignani and Kler (2016). See Sambanis (2002) for a review of conflict studies that address the reverse-causality and endogeneity issues.

12. According to Greene (2008, 209), the Hausman specification test ‘does not guarantee the differences of the two covariances derived from the fixed-effects model and random-effects model will be positive definite’. This is the certainly the case in all the estimates here.

13. If X₁ is a dummy variable, the change in X₁ is the change in its value from 0 to 1.

14. While possibly these non-ARMM neighbors could be the sources of contagion, this issue cannot be pursued here because lack of comparable conflict data for the non-ARMM LGUs.

15. For this variable (interior LGU) alone, the surrounding LGUs include those outside ARMM.