To probe the mechanism on attention, I use causal mediation analysis. I find the relationship is largely mediated by media and online attention surrounding attacks/victims. Other mechanisms, such as family pressure on government and media frames, are (in as much as it’s possible) ruled out.

I also consider some extensions, such as individual effects, temporal dynamics, demands for concessions, and whether strategy/reputation are important. For example, concessions are more likely for NGO staff and foreigners of allied nationalities.

In a victim-level analysis, I find the government are more likely to concede for the release of high-profile victims than non high-profile victims. There are small differences across the two main groups (ASG and NPA).

I test the theory with granular data on the Philippines. The data track hostage-taking activities by six insurgent groups from 1975 to 2018, detailing important victim-level characteristics and information on which attacks result in concessions.

There remains increasing evidence of governments conceding to rebel organizations, particularly for the release of hostages. But it’s unclear when and why this occurs.