[In This Economy] Rodrigo Duterte vs drug war statistics

Whether through their TVs or smartphones, many Filipinos are glued to their screens this week to watch the confirmation of charges hearing of former president Rodrigo Duterte at the International Criminal Court (ICC) in The Hague.

I’m an economist, not a lawyer, but it’s clear to me that Duterte is not yet undergoing a full-blown trial regarding his actions as president that led to the death of thousands of Filipinos. Instead, the prosecution is convincing the court that there’s enough evidence on hand for the ICC to proceed to a full-blown trial.

I leave all the legal arguments and nuances to experts of international law. But here I just want to remark on the use of statistics and data in the proceedings, as well as the causal link that lies at the heart of Duterte’s trial.

I was particularly interested in the intervention of Robynne Croft, a member of the prosecution, who presented a bunch of graphs and maps that show the scale of the deaths that resulted from the drug war. These data are supposed to provide the court “contextual elements” regarding the alleged crimes against humanity of Duterte.

Croft presented a bar chart showing the number of killings in Davao City (at least 274) from 2012 to 2015, before Duterte’s ascendance to the presidency. These include alleged criminals in Davao, including “users and pushers” of illegal drugs. The thing is, Duterte himself has been repeatedly caught on camera boasting about the hundreds or even thousands of people killed under his watch. He inadvertently made the job of the prosecution a lot easier. 

Even more striking is Croft’s map showing a timelapse of the incidence and distribution of deaths related to the war on drugs from 2016 onward, including at least “5,823 killings.” The Duterte government itself admitted to the thousands of drug war-related deaths, and added them in the former president’s “Key Accomplishments” report as well as #RealNumbersPH. (I wrote about this a lot back in 2016 and 2017, in my early days as a regular Rappler columnist.)

Causal link

In economics research, pieces of information that give context about a research topic are called stylized facts. But the real question, legally and analytically, is causation: did Duterte’s actions systematically cause the surge in killings? How can we quantitatively say that X caused Y, as opposed to Y causing X, X being just correlated with Y, or X and Y being caused by some other cause Z. (I happen to teach a course this semester on causal inference in economics, that’s why I’m invested in this.)

In real life, there are plenty of spurious correlations we should watch out for. Just because ice cream consumption is correlated with the occurrence of the summer months doesn’t mean that ice cream caused summer, just as a rooster doesn’t cause the sun to rise every morning because they go together. Correlation isn’t causation.

Duterte’s eventual innocence of guilty before the ICC seems to be hinging on establishing the causal effect of his actions on the drug-related deaths in the Philippines during his term; that is, how systematic the deaths were at the behest or orders of Duterte himself.

Here again, Croft cited statistics including findings in Bulacan province showing a sudden jump in the fatality rate there compared to other countries and cities in the world. Based on the unnamed expert they cited, there was a reported 590% jump in suspects killed nationwide, and a 1,600% increase in “police use of deadly force” in Bulacan.

I argue there are even more compelling statistics that can be cited.

I managed to crunch data from ACLED, or Armed Conflict Location & Event Data Project, a comprehensive dataset of political violence and protest activity globally. They have comprehensive data related to drug-related fatalities before, during, and after Duterte’s term.

The first graph below shows a breakdown of drug-related fatalities in the Philippines from 2016 onward. Note the sudden spike of “state killing of suspects” after Duterte won the May 9, 2016 polls, and especially as he took office on June 30, 2016. There was also a dramatic spike of “vigilante killings.” To the extent that these events were near zero before Duterte entered the scene, that might point to his pivotal role in the systematic killing of civilians. (Note that state killings didn’t really disappear throughout Duterte’s term, although they definitely peaked in 2016 to 2017.)

Zooming in on 2016, the second graph below shows, as clear as day, that drug-related deaths jumped right when Duterte took office by end-June 2016. Even before that, the deaths started to rise incrementally after Duterte won the May 9 polls that year. Before these landmark events, drug campaign fatalities were at a minimum.

The fact that drug war deaths already rose in the few months after Duterte won the presidency and before his inauguration suggests not just an increased focus of reporting of drug-related deaths but an anticipatory effect, where law enforcement agencies were emboldened to follow the lead and words of Duterte.

This analysis is especially compelling because political violence events unrelated to the drug war did not show a similar break in trend. In the language of causal inference, this functions as a kind of placebo test: the spike appears specific to drug-related fatalities.

A skeptic might argue that reporting simply improved, or that police were already killing at similar rates but these incidents were previously undercounted. But that explanation doesn’t account for the sharp timing of the increase. If this were merely better documentation, we would expect broader changes in reported political violence. Instead, the discontinuity appears concentrated specifically in drug-related killings, while other categories remain largely unchanged.

From a technical standpoint, this analysis can be further sharpened, and I hope more researchers will join this effort. The difficulty is that the drug war occurred all at once in the Philippines, and a cleaner analysis could be done if the drug war was implemented in some regions but not in others — resembling a “natural experiment.”

In particular, a stronger case could examine variation across regions: for example, whether drug-related killings spiked disproportionately following the appointment of particular police generals or drug war implementors. Such regional timing differences could resemble a natural experiment and allow for a more rigorous analysis.

In any event, the decision in Duterte’s ICC case will depend on the strength of legal arguments, not statistical ones. But while courts ultimately decide on legal standards, careful statistical analysis can illuminate patterns of responsibility, and help ensure that justice, if it comes, rests on evidence rather than rhetoric or “political demagoguery.” – Rappler.com

You can check out the ACLED data here.Dr. JC Punongbayan is an assistant professor at the UP School of Economics and the author of False Nostalgia: The Marcos “Golden Age” Myths and How to Debunk Them. In 2024, he received The Outstanding Young Men (TOYM) Award for economics. Follow him on Instagram (@jcpunongbayan).

Click here for more In This Economy by JC Punongbayan articles.