The Local Politics of Populism:Evidence from Duterte’s War on Drugs in the Philippines

Dotan Haim∗, Nico Ravanilla† & Renard Sexton‡

April 26, 2019

Abstract

In recent years, populist politicians have won elections in numerous countries on the strength of signature, often norm-

defying, policy proposals. Although effective at mobilizing supporters at election time, these policies can be difficult to

implement in practice. Examining the case of the Philippines, we find evidence that local politicians who are locked out

of establishment patronage networks take the lead in implementing populist policies. Taking advantage of a natural ex-

periment, we find that municipal mayors from independent or small parties execute President Rodrigo Duterte’s signature

“War of Drugs" more than 40 percent more aggressively than those from the establishment Liberal Party. These indepen-

dent mayors receive nearly 50 percent less public works appropriations than establishment-linked mayors, and struggle to

achieve re-election through traditional patronage politics. Using data from the upcoming 2019 mayoral election campaign

and ensuing results, we will find out whether these non-establishment mayors get political support in turn from Duterte

allies, and succeed at securing reelection.

∗Assistant Professor, School of International Service, American University.†Assistant Professor, School of Global Policy and Strategy, UCSD.‡Assistant Professor, Department of Political Science, Emory University. Corresponding author, e: renard.sexton@emory.edu

Work

ing pa

per -

do no

t cite

1 Introduction

Populist political movements have enjoyed substantial success in recent years, often winning through anti-elite rhetoric,

and by championing policies that are popular with their base (and potentially the public at large) but contravene long-held

social or political norms. Although social scientists, political consultants, and the media have learned quite a bit about why

populists emerge and win, we know considerably less about when these leaders are able to actually follow through and

implement anti-establishment policy platforms.

When politicians enter office, they must contend with political and bureaucratic impediments that limit their ability to

enact policy change, especially for high-profile, signature policies. This process is likely to be especially difficult for populist

leaders because established elites have a strong incentive to prevent changes to "politics-as-usual," especially when they

are norm-defying or of questionable legality. How do populist national leaders overcome institutional opposition?

In this paper, we argue and find evidence that local politicians who are locked out of establishment patronage

networks take the lead in implementing populist policies, assisting outsider leaders to circumvent, rather than defeat, the

“deep state." These politicians, expecting a quid pro quo from the emerging outsider network, gamble that political support

from a popular national politician can rescue their political ambitions, substituting for their lack of access to the traditional

patronage network.

Exploiting a natural experiment in the Philippines, we find that mayors from independent or small parties execute

President Rodrigo Duterte’s signature “War of Drugs" more than 40 percent more aggressively than those from the estab-

lishment Liberal Party.1 On other peace and security measures, such as total crime rates, homicides, or property crime,

these mayors’ records look indistinguishable.2

This dramatic effort on behalf of the President’s pet policy fits well with these non-establishment mayors’ incentives.

These outsider mayors are at a distinct disadvantage when it comes to attaining government funds that would allow them to

cement support through patronage or improved public goods provision. We find that establishment (Liberal Party) mayors

receive more than 40 percent more public works procurement funds, the primary source for patronage and pork at the

municipal level, than non-establishment mayors. This difference is driven largely by two categories of spending — roads

projects and flood abatement — that are well known as prime sources of kickbacks and nepotistic contracting.

Note: Paragraph here about the electioneering results from early May and the election results of 13 May.

1The Liberal Party, aligned with the previous President Benigno Aquino III, made up more than 80% of mayors after the 2016 election.2The differences in drug-related crime reports we observe could be the product of a true difference in activities by municipal police stations or instead

strategic changes in reporting in order to appear that they are doing greater anti-drug activities. We carry out several diagnostics to adjudicate which it ismore likely to be, as well the implications of each for local politics.

1

Work

ing pa

per -

do no

t cite

To produce these results, we take advantage of 189 very close mayoral elections between an establishment (Liberal

Party) candidate and a non-establishment candidate during the 2016 election campaign. Using a difference-in-differences

approach with geo-coded data from Philippine National Police (PNP) crime blotter reports, procurement contracts and

election campaign behavior in 2019, we estimate the difference between having an establishment versus non-establishment

mayor during the ensuing Duterte presidency.

These findings speak to an enduring and recently revitalized debate about populist strategies and their implications

for national and local democratic politics (Riker, 1982; Mudde, 2007; Inglehart and Norris, 2016). Much of the discussion

has revolved around the faultlines that give rise to populist movements, and why they succeed or fail at attaining power.

(Sachs, 1989; Seligson, 2007; Mudde, 2013; Pepinsky, 2019). Our study fills an important gap in the literature by exploring

how the implementation (or failure to launch) of populist policies ultimately depends on pre-existing political structures and

networks. In short, populist leaders need to secure the cooperation of local political elites (counterparts) to successfully

implement their populist agenda, but must rely on those that are outside of the previous political establishment.

The results also address a complicated thematic issue — policies around illegal drug enforcement — at a time

when violent crackdowns are on the rise. From El Salvador (Holland, 2013) to Bangladesh (Quackenbush, 2018) and

Cambodia (Prak Chan Thul, 2017), harsh government responses to drug trafficking and usage are often brutally popular

among voters. Our study complements important work by Dell (2015) that documents how crackdowns, driven by the

partisanship of local officials, impact violence between competing cartels in Mexico. We build on existing scholarship by

showing that beyond ideological comportment, concern for human rights, or partisanship, local electoral incentives and

political “quid pro quo" play a critical role in dictating to what extent violent drug crackdowns occur.

Where might the insights from this paper generalize? In contemporary cases as diverse as the United States, Brazil,

Hungary, India, Indonesia and the Philippines, populist leaders have won elections running on signature policies and faced

the dilemma of how to implement them within an entrenched political system. Within highly developed bureaucratic institu-

tions, it may be possible to make pronouncements from on high, but even in those cases procedural or legal impediments

can derail implementation (e.g. asylum crackdown in the US). Within countries with less developed bureaucratic and/or

command and control structures, it is even more difficult to get the policy implemented throughout the country. Instead,

one must rely on allies throughout the many (de-facto and/or de-jure) decentralized parts of government. The basic logic

of marginalized lower-level officials independently driving forward signature policies in order to build connections with a

populist leader, and the leader rewarding them with political cover, cuts across contexts.

The paper proceeds with a section on theory and hypotheses, before discussing the Philippine context, as well

as the regional debate about methamphetamine enforcement in Southeast Asia. We then explain our research design,

2

Work

ing pa

per -

do no

t cite

including the identification strategy, data and estimation. Finally we present the results, before concluding.

2 Theory

Across a wide range of countries, populist movements have gained significant traction in recent years, including the elec-

tions that swept into power Andrés Manuel López Obrador in Mexico, Jair Bolsonaro in Brazil, and Rodrigo Duterte in the

Philippines. While the nature of these leaders’ policy platforms differs significantly, their political ideologies and accompa-

nying rhetoric center around bucking "entrenched," "elite," or "establishment" political systems and ways of doing business

(Akkerman, Mudde and Zaslove, 2014).3 One important way that populist leaders do this is by championing policies that

are popular with their base (and potentially the public at large), but contravene long-held social or politics norms (Mudde,

2004; Spruyt, Keppens and Van Droogenbroeck, 2016). Examples include Donald Trump’s 2016 campaign promise to

“build a wall" between the United States and Mexico, Hungarian Prime Minister Victor Orban’s crusade against the Central

European University and Rodrigo Duterte’s ongoing “War on Drugs" in the Philippines.

After being elected, a critical question is how populist national leaders, who rely on signature policies and personal

charisma, can work within the existing system to implement their agenda. In many areas of the developing world, including

Latin America and Southeast Asia, "business-as-usual" politics often involve a layered system of patronage, clientelism,

and pork distribution whereby national politicians need strong local allies to carry out their policy agendas (Wantchekon,

2003; Kitschelt, Wilkinson et al., 2007; Stokes et al., 2013).

Local politicians with ties to national figures get preferential access to funds that allows them both to win electoral

support for themselves as well as consolidate support for the national politicians to whom they have ties (Calvo and Murillo,

2013; Hicken, 2011). As a result, those local politicians from establishment political parties or family "dynasties" with ties

to the center are likely to fight tooth and nail to maintain the importance of existing political networks, undermining outsider

national politicians in the process (Fergusson, Larreguy and Riaño, 2015). By bucking the traditional system, how do

populist leaders incentivize local elites to implement their preferred policies full-force?

We posit that populist politicians without ties to existing party machines4 or a robust informal network of local allies5

can instead rely on local politicians who themselves do not have access to these same clientelist, establishment means of

politics. Because local politicians without strong ties to established political networks cannot rely on their ability to procure

3It is this pro-people, anti-elite ideology that many scholars use to define populism. Typically elected democratically, periods of populism — where‘majoritarian’ democracy trods on ‘civil rights’ democracy — are common features of unconsolidated democracy, what Dan Slater (2013) calls democratic“careening."

4For example, the pre-2000 PRI in Mexico (Magaloni, 2006; Fergusson, Larreguy and Riaño, 2015)5Relevant in the Philippines Hicken (2014), Indonesia(Pepinsky, 2009), and India (Auerbach, 2016)

3

Work

ing pa

per -

do no

t cite

the pork needed to bolster their legitimacy through the distribution of patronage, they may instead choose to wholeheartedly

carry out the policies espoused by populist national figures. Even if they did not run on these policies in the first place,

jumping aboard a national populist movement can allow these politicians to tap into an alternative source of legitimacy,

allowing them improved chances to compete with well-connected politicians down the line.

One might think that riding the populist wave would be the optimal strategy for all local politicians, especially during

times when the policies espoused by the center are broadly popular. For example, in the Philippines, Rodrigo Duterte’s

violent drug war is broadly popular, garnering support from over 80% of the population in most surveys (See Figure 1).

While this may be true to a certain extent, fully carrying out a policy of this nature carries significant risk for local politicians.

First, even when a policy like the Duterte drug war is broadly popular, implementing the policy to its full extent may not

be. While people seem to express strong support for the "campaign against illegal drugs," many civilians in the Philippines

express significant fear of being victim of its more violent manifestations (Table 1). In addition, support for extrajudicial

killings is much lower, hovering below 50%. Instead of fully implementing policies espoused by populist national politicians,

local politicians may be better off implementing a more moderate version, allowing them to gain the political benefits without

alienating a significant portion of the population.

Second, implementing populist policies espoused by anti-establishment politicians may lead to estrangement from

the stable political parties and power-brokers that enable sustained political careers. This is perhaps especially salient in

countries like the Philippines, where political dynasties often control politics for several generations (Cruz, Labonne and

Querubin, 2017; McCoy, 2009)). Again, in this case, it may be worth it for local politicians to hedge, only implementing

policy initiatives from the national leadership to the extent that it does not step on the toes of other important political

players. Finally, fully implementing policy initiatives that involve illegal practices or human rights violations can leave local

elites vulnerable to future prosecution if the political winds change.

Which local elites, then, will find the benefits of fully implementing the policies espoused by populist national leaders

to be worth the risk? We argue that the benefits are most likely to outweigh the risks for local politicians who have

weak existing ties to establishment political structures. Especially in political systems that are highly reliant on clientelist

networks linking local politicians to national-level patrons, the local politicians who are not part of these networks are at

a distinct disadvantage when it comes to winning re-election. Winning the support of the local population (either through

patronage or effective programmatic politics) is highly dependent on having access to funds that stream down from the

central government. However, legislators and agency heads who control budgeting may block non-aligned local politicians

from getting access to funds in favor of their political allies at the local level who can help them campaign during the next

election. As a result, local politicians who are "locked out" of establishment political networks may choose to implement

4

Work

ing pa

per -

do no

t cite

populist policies more aggressively to gain attention and ride the leader’s popularity to re-election. Stemming from this

logic, we put forward the following linked hypotheses:

H1: Local politicians who are not aligned with establishment political networks are more likely to aggressively

implement policies supported by national-level populist politicians.

H2: Local politicians who are not aligned with establishment political networks get less “pork."

Observing the effort by non-establishment mayors to implement their signature policies, we would expect the pop-

ulist leader and their allies to reward these mayors with political attention. This could occur by sending prominent supporters

to campaign for these mayors, making campaign contributions or allowing popular political branding to take place.

H3: Local politicians who are not aligned with establishment political networks get greater attention at election

time from the populist leader

H4: Local politicians who implement signature policies with particular vigor get particularly high attention

We expect that the above hypotheses are most likely to hold in situations where local politicians have significant

leeway over policy implementation. Even if local politicians wanted to more aggressively prosecute populist-supported

policies (like Duterte’s drug war), they would be unable to do so if the bureaucracy in charge of implementation was entirely

independent. In other words, there must be some leeway for political influence over the bureaucratic implementation in

order for these hypotheses to hold.

3 Context

3.1 Philippine politics

We test our theory in the Philippines, where we look at differential municipal implementation of Rodrigo Duterte’s "War on

Drugs" after his election in May 2016. While he was a longstanding mayor in Davao City, the Philippines’ second largest

population center, Duterte had relatively few ties to the traditional sources of national power. He is the first Philippine

president from the large southern island of Mindanao and declared that he was running on the last possible day for legal

filing of a campaign. Duterte’s ideology has been described as populist by a number of scholars and his policy platform

centered on a campaign against illegal drugs that followed in the footsteps of the harsh crackdown he oversaw as Davao’s

mayor. Duterte won with a plurality of just 39 percent of the vote, but this is fairly typical for presidential elections in the

5

Work

ing pa

per -

do no

t cite

Philippines. Although Duterte won with the smallest winning share of the vote since 1992 (when Fidel Ramos won with

less than 24 percent of the vote), no presidential candidate has won more than 42.1% of the popular vote since the 1987

Constitution was inaugurated.

The Philippines is a presidential system. Presidents serve a six-year term with no reelection, while mayors and

congressmen can serve up to two consecutive three-year terms. At the local level, there are 81 provinces in the country

that encompass 18 municipalities on average (for a total of 1,489 municipalities). Provinces are overseen by governors

and municipalities by mayors. Because of the high degree of decentralization, mayors are perhaps collectively the most

important political power brokers in the country. Especially because of the long-standing entrenchment of local political

dynasties (Querubin, 2016), mayors exert a great deal of control over local policy implementation. However, these local-

level politicians are also highly reliant on national-level congressmen when it comes to the allocation of government funds.

By controlling national level budgeting and procurement of funds, congressmen have significant control over the resources

available to certain localities.

Formal political parties in the Philippines are very weak and campaigns are centered strongly around individual per-

sonalities (Hicken, 2009). The political party structure changes from election to election, candidates often switch between

parties in different election cycles, and parties are not usually associated with a well-defined programmatic platform. That

being said, the parties that do coalesce during each election cycle often represent strong, though short-lived, alliances

between national legislators and local mayors who rely on each other to sustain the clientelistic exchange of pork for a

ready corps of supporters.

In the absence of stable national party organizations, national legislators often need to rely on mayors as local

political intermediaries (Abinales and Amoroso, 2005; Hutchcroft and Rocamora, 2003). For their part, mayors rely on their

network of political brokers at the grassroots level to shore up votes. Mayors find it in their best interest to form alliances

with national legislators, because in exchange for the “command votes” that they can mobilize, they are able to reap the

benefits of a quid pro quo in the determination of the legislators’ district spending priorities (De Dios, 2007).

Alliances between politicians are often formalized on "sample ballots" that are coordinated by congressmen and

passed out to voters by brokers organized by mayors. Therefore, during a single election, candidates’ party is often a strong

signal of their alignment with a particular political machine. In the 2016 election, the most well-coordinated political alliance

was the "Liberal Party" (LP), which is the same ticket the prior president Aquino III ran under. Liberal party politicians were

characterized by being aligned with the well-entrenched political power brokers in Filipino politics.

Another important scope condition that is met in the Philippine case is the ability of mayors to influence the Philippine

National Police (PNP) prosecution of the drug war. Mayors, Governors, and other prominent local politicians have long had

6

Work

ing pa

per -

do no

t cite

a hand in influencing the assignment of local military and police leadership and, through this mechanism, have significant

power over local security policy (Sidel 1999). When local politicians are unhappy with the activity of municipal and provincial

police leadership, they can use their networks of influence to initiate police chief replacements or block promotions to higher

positions.6 As a result, local security leadership can end up beholden to the preferences of local political leadership. In

addition, policing in the Philippines is highly decentralized, with local police leadership having significant leeway over policy

implementation.

3.2 War on Drugs in Southeast Asia

Illegal drug use in Southeast Asia is a highly salient and politically charged topic, as it is in many parts of the world. From

2013 to 2018, methamphetamine seizures across Southeast Asia tripled from about 40 million tons to just under 120 million

tons (United Nations Office on Drugs and Crime, 2019). Among then ten ASEAN member states, all but Vietnam currently

report that methamphetamine is the “primary drug of concern" in their society (United Nations Office on Drugs and Crime,

2019).

Ten years ago, only three of those countries listed methamphetamine as their top concern.7 One of those three is

the Philippines, where methamphetamine production and usage has been prevalent for nearly two decades(McKetin et al.,

2008). Over the past five years, more than 90 percent of admitted for drug treatment in the Philippines have been related

to crystalline methamphetamine, a.k.a. crystal meth (United Nations Office on Drugs and Crime, 2019). This has led to

public outcry about the public cost of drug use, and turned drugs into a powerful political tool for enterprising politicians like

now President Duterte. Known as “the Punisher" during his time as mayor of Davao City in southern Mindanao, Duterte

repeatedly won office by promising to deal with criminality and the drug economy in the harshest terms possible: “When I

said I’ll stop criminality, I’ll stop criminality. If I have to kill you, I’ll kill you. Personally."8

The Duterte administration’s crackdown, though brutal and of often questionable legality 9, continues to enjoy broad

public support. Table 1 shows that across socio-economic classes the Drug War, including a escalation in 2017 known

as “Double Barrel Reloaded", is quite popular. This popularity is stable over time, as seen in Figure 1 — with the notable

exception of late 2016, when a scandal about the Philippine National Police broke. The drug crackdown’s popularity

returned after a matter of months, following a short “pause" demanded by the President.

6This sentiment was expressed to the authors in interviews with several municipal police chiefs in the Bicol Region of the Philippines7In this paper we use ‘meth,’ ‘shabu’ and ‘methamphetamine’ interchangeably, referring to the category of “Amphetamine-type Stimulants (ATS)," as

defined by the UNODC. This includes methamphetamine in crystalline, pill and powder form, MDMA (commonly known as “ecstasy"), and amphetamine.Where more specific terms are used, particular subgroups of ATS are indicated.

8Ressa (2015).9e.g. Human Rights Watch (2017)

7

Work

ing pa

per -

do no

t cite

We note that more than two thirds of respondents say they fear being injured or killed inadvertently by drug war

operations, especially among lower class respondents. At the same time, less than forty percent of respondents believe

that the rights of suspects are being respected. That citizens are willing to countenance such risks suggests that many

people believe that drug pushing and usage is a major problem and an aggressive approach is the only way to stop it. That

said, we suspect that social desirability bias may be pushing the figures for the anti-drug campaign and President Duterte

somewhat higher than their ‘true’ values.

Table 1: Support for the Duterte Drug War by Social Class

Pulse Asia Percent of % Support % Fear % Support % ApproveSocio-economic Class sample Drug War Drug War Double Barrel Duterte

A, B & C Wealthy, upper middle, middle 9 89 47 78 84D1 Lower middle A 53 90 73 73 80D2 Lower middle B 21 88 69 72 78E Impoverished 17 91 66 70 86

Total 90 68 73 81

Even if the Philippines case is perhaps the most extreme example, governments in Southeast Asia have primarily

responded to similar drug problems using violent crackdowns against smugglers, distributors and even users. Replicating

what has become known in El Salvador in the 1990s as the ‘Mano Dura’ (iron fist) approach, these police-led clearing

operations are intended to beat back drug trafficking networks and give confidence to law-abiding citizens. Severe drug

crimes are liable for capital punishment in most ASEAN countries – in recent years traffickers and distributors have been

executed in Myanmar, Thailand, Malaysia, Singapore, Vietnam, Indonesia and the Philippines. Although primarily carried

out through formal state processes, vigilantism and extrajudicial killings are on the rise across the region.

8

Work

ing pa

per -

do no

t cite

Figure 1: Public Approval for the Drug War and Pres. Duterte (Pulse Asia)

0.2

.4.6

.81

July 2016 Jan 2017 July 2017 Jan 2018 July 2018

Duterte Approval Support for Drug War

9

Work

ing pa

per -

do no

t cite

4 Research Design

We seek to answer two main empirical questions in the this study: (1) How do political incentives affect how local politicians

implement the drug crackdown policy?10 And (2), what are the implications of aggressive drug crackdowns on other socio-

economic outcomes of interest, including crime, health and economic activity?11 In this draft, we focus on the first of these

questions.

To do so, we leverage very close mayoral elections that occurred simultaneously with the 2016 presidential election

that saw President Duterte elected. Among the nation’s more than 1,600 cities and municipalities, 189 had mayoral

elections with a margin of victory of less than five percentage points.12 We characterize these as very close elections,

where, in the context of Philippines politics, the partisanship of the winner is effectively random. We carry out a series

of diagnostics and placebos to demonstrate that these very close elections approximate a natural experiment, and for

robustness also estimate the results using a regression discontinuity design (Caughey and Sekhon, 2011; Eggers et al.,

2015).

We primarily employ a two period difference in differences estimator, which has three important advantages.13 First,

it allows us to take advantage of pre-treatment data for statistical power purposes. Second, a difference in differences has

the important benefit of relaxing the as-if-random assumption, requiring instead only parallel trends for identification, in

addition to exchangeability, positivity and SUTVA (Bertrand, Duflo and Mullainathan, 2004; McKenzie, 2012).14 Lastly,

unlike a regression discontinuity design, we can effectively estimate heterogeneous treatment effects without giving away

a great deal of statistical power.

4.1 Data

Bantay Krimen

The key dependent variable in our analysis (intensity of drug war implementation) is operationalized by the "Bantay

Krimen" dataset published online by the Philippine National Police (PNP).15 We scraped the data underlying the maps dat-

ing back to late 2015, six months prior to Duterte’s election. Data are posted in real time based on blotter reports collected

10Note: We recently obtained personnel data for Region V of the PNP, so will be adding these results in a later draft.11In addition the children’s health results we show below, we recently obtained data on local business incorporation that we will add to the final section

of our results.12Margin of victory is defined as the absolute value of the vote percent margin between the Liberal Party and the top performing non-establishment

candidate. The handful of cases where the Liberal Party did not have a finisher in the top two are excluded. For robustness, in the appendix we alsoshow the results with a two percentage point margin.

13We use also employ ANCOVA and RDD in the Appendix.14Each of our plots indicates that in addition to having close to identical means in the pre-period, the pre-trends are quite similar.15https://www.bantaykrimen.com/

10

Work

ing pa

per -

do no

t cite

by municipal police stations and organized by provincial police offices. Data are broken down into 7 main categories: 1)

drug-related crimes, 2) theft, 3) assault, 4) homicide, 5) robbery, 6) rape, and 7) vehicular accidents. Over the time period

of this study, 994,212 crimes were recorded in the dataset, including 120,768 drug-related incidents. We use the number of

drug-related incidents as a measure of the degree to which local police offices implemented Duterte’s drug war. Reporting

drug-related incidents follows a set guidelines articulated in the Revised PNP Manual on Anti-Illegal Drugs Operations and

Investigation (2014).16

One concern with using data from official police reports is that there may be an opportunity for municipal police

to manipulate reported crime rates. Thus, while an increase in the rate of drug-related crimes may represent actual

aggressive implementation of the drug war, it may also represent police chiefs (or mayors who pressure police chiefs) who

falsely "display" more aggressive implementation. At this point, we remain agnostic to which of these scenarios is actually

being played out, as both are consistent with the logic we present in the theory section. Even if the effects are driven

by mayors who put pressure on the police to publicly signal that they are prosecuting the drug war more intensely, this is

likely still driven by the same political incentive to display alignment with Duterte’s policy. In future iterations of the project,

we plan tease out which mechanisms is at play by using data on drug-related deaths and arrests reported in local news

outlets.17

2010 Census

The Philippine Statistical Authority (PSA) carried out the “2010 Census of Population and Housing" over the course

of 2010. The survey collects information at the household level for more than 92 million Filipinos. We obtained this data

at a municipal level from the PSA, allowing us to compare, within our sample of close mayoral elections, the municipalities

that narrowly elected a Liberal Party (LP) mayor and those that elected a minor party/independent (non-establishment)

mayor. Table 2 show the summary statistics for 187 of the 189 municipalities within our sample.18 Importantly, we find that

there are no substantive differences between the LP and non-establishment municipalities in terms of their background

characteristics.

2016 Election Data

To measure our key independent variable (mayoral alignment with establishment politics), we use precinct-level

data (aggregated to the municipal level) from the 2016 Philippine elections. Voter turnout in the election was around 82%.

16To be classified as drug-related, the reported incident must fall under one of the following seven categories: (1) buy-bust operations, (2) search andseizure by virtue of a warrant, (3) marijuana eradication, (4) financial investigation, (5) controlled delivery, (6) clandestine laboratory, and (7) inflagrantedelicto cases (i.e. unplanned operations.

17We are in talks with the people who run the website "https://drugarchive.ph/", a project directed by the Columbia school of journalism. The projectaims to collect independently verified incidents related to the drug-war, including those not reported by the police.

18Boundary changes between 2010 and 2016 impacted the matching for two municipalities

11

Work

ing pa

per -

do no

t cite

Figure 2: Drug-related Police Blotter Reports (Nationwide)

Duterte- Transition

Begins

Jee Ick-joo- Kidnapping

(pause)

- DoubleBarrel

Reloaded

Pause of- Double Barrel

010

020

030

0

Jun 2016 Oct 2016 Apr 2017 Oct 2017 Apr 2018

Drug-related Crimes via Police Blotter

12

Work

ing pa

per -

do no

t cite

Table 2: 2010 Census Data (balance test among close mayoral elections)

Variable LP Non-establishment Diff. % Diff. T-testmayor mayor p value

Population 25749 26388 639 0.02 0.88% Male 0.51 0.51 0.00001 0.00 0.99Household Size 3.64 3.64 0.006 0.00 0.95% working overseas (OFW) 0.026 0.024 -0.001 -0.04 0.63Education Index 1.84 1.85 0.01 0.01 0.86% aged 18-39 0.56 0.56 0.004 0.01 0.59% Single (unmarried) 0.28 0.28 -0.001 0.00 0.16% Roman Catholic 0.76 0.79 0.027 0.04 0.46% Muslim 0.07 0.06 -0.01 -0.14 0.70% Other Religion 0.17 0.15 -0.015 -0.09 0.55% Indigenous People 0.21 0.15 -0.05 -0.24 0.26% Home: good quality 0.36 0.36 -0.0001 0.00 0.98% Home: bad quality 0.46 0.46 0.0004 0.00 0.83

N (total: 187) 98 89

We create two main categories of mayors based on their political party. First, we designated all mayors who ran under the

LP party label to be aligned with establishment elements of Filipino politics. Approximately 50% of all congressmen and 75

% of mayoral winners ran under the LP label. Duterte ran under PDP-Laban, a party label revived from the late 1980s. For

the purposes of this study we treat all non-LP candidates who were "non-aligned" with the political establishment. Aside

from the LP, all the remaining parties were highly localized or disjointed.

Among the approximately 1500 mayoral elections, we code all elections decided by less than a 5% vote margin

as being "closely contested" elections where we assume the winner is as-if random. 189 elections fall into this category.

Among those, slightly more than half were won by LP mayors and slightly less than half were won by non-aligned mayors.

The following charts display the density of mayoral races by the vote share margin for the non-LP candidate. We

see that there appears to be no heaping along the ‘discontinuity,’ that is, neither LP nor non-LP are able to “sort" into

winning (Barreca, Lindo and Waddell, 2016).

13

Work

ing pa

per -

do no

t cite

02

46

810

Perc

ent

-1 -.5 0 .5 1unaligned_vote_share_margin

Density of Unaligned Mayor Margins vs. LP

05

10Pe

rcen

t

-.5 0 .5unaligned_vote_share_margin

Density of Unaligned Mayor Margins vs. LP

Public Procurement Data

We use the newly available, granular data on civil works contracts from the online civil works monitoring system

of the Department of Public Works and Highways (DPWH). This data details where the contract was implemented, which

contractor firm won the contract, the date the contract was awarded, value of the contract, type of public works, and whether

the spending is for repairs of existing infrastructure or for new construction.

4.2 Estimation

To implement the two-period difference-in-differences estimator, we collapse the crime blotter reports and other time series

data into ‘pre’ and ‘post’ periods, divided by President Duterte’s first day in office: 1 July 2016.19 Then, using OLS, we

estimate:

Yit = β0 + β1Mi + β2Pt + β3Mi ∗ Pt + εit

where Y is the outcome of interest, say drug-related blotter reports per capita, in municipality i in period t. M is an

indicator for ‘non-establishment’ mayors, set to one when the mayor elected in 2016 is not from Liberal Party. P is an

indicator for the post-Duterte period, which becomes 1 after 1 July 2016, 0 before. β1 and β2 are equivalent to municipality

and period fixed effects, and ε is the error term. Municipality fixed effects are absorbed through demeaning and standard

errors are clustered by municipality.

19We show in the Appendix that the results are robust to using the first day of the Duterte transition (1 June 2016) as an alternative binning date.

14

Work

ing pa

per -

do no

t cite

5 Results

In this section we present the primary results of our study, and an analysis of the political mechanisms driving the effects

we find, followed by several placebo tests. Lastly, we present prima facie evidence that the Duterte drug war has not

generated any improvements in family heath, the local economy or crime.

We begin by illustrating the broad patterns in the data, before carrying out our regression analysis. In Figure 3

we observe that following Duterte’s accession to the presidency, there is no divergence in police blotter reports of total

crime between narrowly winning Liberal Party and non-establishment mayors. Overall crime rates remain stable through

Duterte’s first year and a half in office, with the exception of a slight increase during Operation Double Barrel Reloaded in

mid 2017.

In contrast, Figure 4 shows an immediate and widening gap in drug-related blotter reports between non-establishment

and Liberal party mayors. Although both experience an uptick during Double Barrel Reloaded, drug-related crime reports

are significantly higher among non-establishment mayors.

Our difference-in-differences estimates confirm a large and significant difference between non-establishment and

LP-led municipalities in their reporting of drug-related crime. Table 3 shows that in municipalities with an non-establishment

mayor there is a 0.36 per 1,000 increase in the number of drug-related crimes, a more than 38% increase over the Liberal

Party post-period average.20 We do not observe statistically significant differences between LP and non-establishment

mayors for any other kind of crime recording in police blotter reports. That includes theft, assault, homicide, robbery, rape,

car accidents, car theft or combined total crime.

Table 3: Effect on Post-Duterte Crime Rates

(1) (2) (3) (4) (5) (6) (7)Dependent variable: Drug-related Theft Assault Homicide Robbery Rape Total

Non-establishment Mayor * Post-Duterte .36* -.04 -.13 .13 .04 -.01 .41(.18) (.13) (.24) (.10) (.06) (.03) (.47)

LP Post-Duterte Mean .95 .94 1.57 .60 .39 .33 4.93N 378 378 378 378 378 378 378Clusters 189 189 189 189 189 189 189

Notes: Municipality and period fixed effects, SE clustered by municipality, * = p<.05. Sample: Municipalities with veryclose (<5% margin of victory) for the 2016 mayoral election.

20In the Appendix we show that this holds for a two percentage point margin for the ‘close election’ sample, and close to identical estimates are obtainedfrom an RDD with local polynomial regression, as well as ANCOVA with pre-treatment covariates and region fixed effects.

15

Work

ing pa

per -

do no

t cite

Figure 3: Total Police Blotter Reports

- Duterte Takes Office

05

1015

2025

Tota

l Crim

e pe

r 1,0

00 re

gist

ered

vot

ers

Jul 2016 Jan 2017 Jul 2017 Jan 2018 Jul 2018Month

Unaligned Mayor Liberal Party Mayor

Total Crime in Police Blotter (close elec.)

Notes: Among 189 municipalities with very close (<5% margin of victory) for the 2016 mayoral election. Excludes “pure"auto accidents (those that are not classified as homicide or other crime, e.g. hit and run).

What explains why non-establishment mayors appear to prosecuting the drug war more aggressively than estab-

lishment mayors? We hypothesized that non-establishment mayors are less able to generate pork for their municipalities

through public procurement, a critical part of patronage politics in the Philippines (cite Nico’s working paper). Because they

are outside the establishment political network they are both less connected and potentially less likely to be well-versed in

securing funds from the center. Indeed, as shown in Figure 5, we find that in the post-Duterte period non-establishment

mayors are on average not able to secure nearly as much public procurement for their municipalities than Liberal Party

mayors.

Table 4 shows the effects in greater detail. On average, establishment (Liberal Party) mayors are able to secure

about 7.8 million pesos per 1,000 population in public works spending during the post-Duterte period, nearly 40 percent

16

Work

ing pa

per -

do no

t cite

Figure 4: Drug-related Police Blotter Reports

- Duterte Takes Office

01

23

45

Crim

es p

er 1

,000

regi

ster

ed v

oter

s

Jul 2016 Jan 2017 Jul 2017 Jan 2018 Jul 2018Month

Unaligned Mayor Liberal Party Mayor

Drug-Related Crime in Police Blotter (close elec.)

Notes: Among 189 municipalities with very close (<5% margin of victory) for the 2016 mayoral election. Drug-relateddefined as blotter reports where primary designation is RA-9165, referring to the “Comprehensive Dangerous Drugs Act of2002."

higher than the 5.5 million that municipalities with non-establishment mayors attract. As seen in columns (2) and (3) of

Table 4, these differences come almost entirely from spending on roads and flood abatement, two categories of projects

that are notorious for the scale of their kickbacks and patronage (Ramos, 2019) (also Nico citation). In contrast, we see

no difference in procurement levels for schools and health facilities, which although helpful public goods are very difficult to

steer toward the type of kickbacks and clientelistic spending that characterizes local politics in the Philippines.

17

Work

ing pa

per -

do no

t cite

Figure 5: Public Works Procurement Amounts per month

2.00

0e+0

84.

000e

+08

6.00

0e+0

88.

000e

+08

$$

670 675 680 685 690 695 700 705Month

Unaligned Mayor Liberal Party Mayor

Procurement $$ (close elections)

Notes: Among 189 municipalities with very close (<5% margin of victory) for the 2016 mayoral election. Procurement forpublic works includes roads, schools, health facilities, repairs.

18

Work

ing pa

per -

do no

t cite

Table 4: Effect on Public Works Procurement

Easy patronage/kickbacks Difficult for patronage(1) (2) (3) (4) (5)

Dependent variable: Total Public Works Roads Flood-abatement Schools Health Facilities

Non-establishment Mayor * Post-Duterte -2.22* -1.05+ -.93* -.00 -.13(.90) (.55) (.38) (.38) (.11)

LP Post-Duterte Mean 7.75 2.45 1.87 2.13 0.17N 378 378 378 378 378Clusters 189 189 189 189 189

Notes: Outcome is millions of Philippine Pesos (about 50 Pesos per USD) per 1,000 population. Municipality and periodfixed effects, SE clustered by municipality, * = p<.05 + = p<.1. Sample: Municipalities with very close (<5% margin ofvictory) for the 2016 mayoral election.

19

Work

ing pa

per -

do no

t cite

5.1 A political quid pro quo?

In this section we will use data that were are currently collecting regarding campaign activities by the Duterte administration,

PDP Laban party and allied Senators, and then will record the results of the May 2019 mayoral elections. We should have:

• Data on whether candidates in each of our 189 close elections from 2016 received a visit from Duterte-aligned

Senators in the three months leading up to the May elections (collected by research assistants in the Philippines via

public and campaign Facebook pages).

• Information from the final “Sample Ballots" that are issued the day before the election regarding whether each

candidate was listed on the pro-Duterte slate (collected by research assistants in the Philippines)

• Electoral results by municipality, showing the vote shares and re-election results for each of the 189 municipalities.

5.2 Diagnostics and Placebo tests

In this subsection we show pre-treatment information regarding the outcomes and covariates in an effort to bolster the

research design. In Table 5 we carry out a placebo test, demonstrating that the alignment of the mayor after 2016 does

not impact pre-2016 PNP blotter reports. Table 6 extends this to pre-treatment covariates collected from the 2010 census.

We find substantive balance across treatment and control municipalities on a range of politically-important characteristics,

including demographics, ethnic and religious background, and household assets.

In order for our difference-in-differences specification to be valid, we must assume parallel trends. Although this

assumption cannot be proven, evidence of parallel pre-trends is an important factor in favor. In Figures 6, 7 and 8 we

show that our main outcomes of interest — drug-related crime reports, total crime reports and public works procurement

amounts — are moving in parallel during the period before Duterte takes office. In fact the trends are close to identical.

Taken together, we feel more confident that the effects we describe above are well-identified.

20

Work

ing pa

per -

do no

t cite

Table 5: Placebo 1: No Effect on Pre-Duterte Crime Rates

(1) (2) (3) (4) (5) (6) (7)Dependent variable: Drug-related Theft Assault Homicide Robbery Rape Total

Non-establishment Mayor -.01 -.07 .02 .01 .00 .00 -.05(.03) (.05) (.06) (.02) (.02) (.01) (.13)

N 189 189 189 189 189 189 189

Notes: Bantay Krimen blotter reports from 2015 to 1 July 2016, summed by municipality. Region fixed effects, Stata robustSE. Sample: Municipalities with very close (<5% margin of victory) for the 2016 mayoral election.

Table 6: Placebo 2: No Effect on 2010 Census Characteristics

(1) (2) (3) (4) (5) (6)Dependent variable: Population % Male Household % Overseas Education % age

Size Worker 18-39

Non-establishment Mayor 104 -.00 .03 -.00 .01 .00(4378) (.00) (.07) (.00) (.03) (.01)

N 187 187 187 187 187 187

(7) (8) (9) (10) (11)Dependent variable: % Single % Roman % Muslim % Indigenous % Good

Catholic People Home Qual.

Non-establishment Mayor -.01 .03 -.00 -.03 .00(.00) (.02) (.01) (.03) (.00)

N 187 187 187 187 187

Notes: 2010 Filipino census, averages by municipality. Region fixed effects, Stata robust SE. Sample: Municipalities withvery close (<5% margin of victory) for the 2016 mayoral election.

21

Work

ing pa

per -

do no

t cite

Figure 6: Drug-related Police Blotter Reports Trends pre-Duterte

- Duterte Takes Office

05

1015

Drug

Crim

e pe

r 1,0

00 re

gist

ered

vot

ers

Jan 2016 Apr 2016 Jul 2016Month

Unaligned Mayor Liberal Party Mayor

Notes: Among 189 municipalities with very close (<5% margin of victory) for the 2016 mayoral election. Drug-relatedcrimes.

22

Work

ing pa

per -

do no

t cite

Figure 7: Total Police Blotter Reports Trends pre-Duterte

- Duterte Takes Office

05

1015

2025

30To

tal C

rime

per 1

,000

regi

ster

ed v

oter

s

Jan 2016 Apr 2016 Jul 2016Month

Unaligned Mayor Liberal Party Mayor

Notes: Among 189 municipalities with very close (<5% margin of victory) for the 2016 mayoral election. Excludes “pure"auto accidents (those that are not classified as homicide or other crime, e.g. hit and run).

23

Work

ing pa

per -

do no

t cite

Figure 8: Public Works Procurement Trends pre-Duterte

- Duterte Takes Office

010

2030

4050

Proc

urem

ent (

milli

ons)

per

1,0

00 re

gist

ered

vot

ers

Jan 2016 Apr 2016 Jul 2016Month

Unaligned Mayor Liberal Party Mayor

Notes: Among 189 municipalities with very close (<5% margin of victory) for the 2016 mayoral election. Millions of pesosof public works procurement per 1,000 population.

24

Work

ing pa

per -

do no

t cite

5.3 Implications of the Drug War

TBD after we finish cleaning data from the Department of Trade and Industry. Data analysis to date indicates no difference

between establishment and non-establishment mayors in terms of infant health or mortality.

6 Discussion

TBD after election results from May 2019 are in.

25

Work

ing pa

per -

do no

t cite

References

Abinales, Patricio and Donna Amoroso. 2005. State and Society in the Philippines. Pasig City: Anvil Publishing Inc.

Akkerman, Agnes, Cas Mudde and Andrej Zaslove. 2014. “How populist are the people? Measuring populist attitudes invoters.” Comparative political studies 47(9):1324–1353.

Auerbach, Adam Michael. 2016. “Clients and communities: The political economy of party network organization anddevelopment in IndiaâAZs urban slums.” World Politics 68(1):111–148.

Barreca, Alan I, Jason M Lindo and Glen R Waddell. 2016. “Heaping-induced bias in regression-discontinuity designs.”Economic Inquiry 54(1):268–293.

Bertrand, Marianne, Esther Duflo and Sendhil Mullainathan. 2004. “How much should we trust differences-in-differencesestimates?” The Quarterly journal of economics 119(1):249–275.

Calvo, Ernesto and Maria Victoria Murillo. 2013. “When parties meet voters: Assessing political linkages through partisannetworks and distributive expectations in Argentina and Chile.” Comparative Political Studies 46(7):851–882.

Caughey, Devin and Jasjeet S Sekhon. 2011. “Elections and the regression discontinuity design: Lessons from close UShouse races, 1942–2008.” Political Analysis 19(4):385–408.

Cruz, Cesi, Julien Labonne and Pablo Querubin. 2017. “Politician family networks and electoral outcomes: Evidence fromthe Philippines.” American Economic Review 107(10):3006–37.

Darroch, Gordon. 2017. “Netherlands ’will pay the price’ for blocking Turkish visit âAS Erdogan.” The Guardian .URL: https://www.theguardian.com/world/2017/mar/12/netherlands-will-pay-the-price-for-blocking-turkish-visit-erdogan

De Dios, Emmanuel. 2007. The Dynamics of Regional Development: The Philippines in East Asia. Massachusetts: EdwardElgar Publishing Inc. chapter Local Politics and Local Economy, pp. 157–203.

Dell, Melissa. 2015. “Trafficking networks and the Mexican drug war.” American Economic Review 105(6):1738–79.

Eggers, Andrew C, Anthony Fowler, Jens Hainmueller, Andrew B Hall and James M Snyder Jr. 2015. “On the validity of theregression discontinuity design for estimating electoral effects: New evidence from over 40,000 close races.” AmericanJournal of Political Science 59(1):259–274.

Farrell, Graham and John Thorne. 2005. “Where have all the flowers gone?: evaluation of the Taliban crackdown againstopium poppy cultivation in Afghanistan.” International Journal of Drug Policy 16(2):81–91.

Fergusson, Leopoldo, Horacio Larreguy and Juan Felipe Riaño. 2015. “Political constraints and state capacity: Evidencefrom a land allocation program in Mexico.” Development Bank of Latin America Working Paper .

Hicken, Allen. 2009. Building Party Systems in Developing Democracies. New York: Cambridge University Press.

Hicken, Allen. 2011. “Clientelism.” Annual review of political science 14:289–310.

Hicken, Allen. 2014. “Party and party system institutionalization in the Philippines.” Party System Institutionalization in Asia:Democracies, autocracies, and the shadows of the past 307.

Holland, Alisha C. 2013. “RIGHT ON CRIME? Conservative Party Politics and" Mano Dura" Policies in El Salvador.” LatinAmerican Research Review pp. 44–67.

Human Rights Watch. 2017. License to Kill: Philippine Police Killings in Duterte’s War on Drugs. Human Rights Watch.

Hutchcroft, Paul and Joel Rocamora. 2003. “Strong Demands and Weak Institutions: The Origins and Evolution of theDemocratic Deficit in the Philippines.” Journal of East Asian Studies 3(2):259–292.

Inglehart, Ronald F and Pippa Norris. 2016. “Trump, Brexit, and the rise of populism: Economic have-nots and culturalbacklash.”.

26

Work

ing pa

per -

do no

t cite

Kitschelt, Herbert, Steven I Wilkinson et al. 2007. Patrons, clients and policies: Patterns of democratic accountability andpolitical competition. Cambridge University Press.

Magaloni, Beatriz. 2006. Voting for autocracy: Hegemonic party survival and its demise in Mexico. Vol. 296 CambridgeUniversity Press Cambridge.

McCoy, Alfred W. 2009. An anarchy of families: State and family in the Philippines. Univ of Wisconsin Press.

McKenzie, David. 2012. “Beyond baseline and follow-up: The case for more T in experiments.” Journal of developmentEconomics 99(2):210–221.

McKetin, Rebecca, Nicholas Kozel, Jeremy Douglas and Robert Ali. 2008. “The rise of methamphetamine in Southeastand East Asia.” Drug and alcohol review 27(3):220–228.

Mudde, Cas. 2004. “The populist zeitgeist.” Government and opposition 39(4):541–563.

Mudde, Cas. 2007. Populist radical right parties in Europe. Cambridge University Press Cambridge.

Mudde, Cas. 2013. “Three decades of populist radical right parties in Western Europe: So what?” European Journal ofPolitical Research 52(1):1–19.

Pepinsky, Thomas. 2019. “Migrants, Minorities, and Populism in Asia.”.

Pepinsky, Thomas B. 2009. Economic crises and the breakdown of authoritarian regimes: Indonesia and Malaysia incomparative perspective. Cambridge University Press.

Prak Chan Thul. 2017. “Cambodia promises harsher drug crackdown as arrests soar.” Reuters .URL: https://www.reuters.com/article/uk-cambodia-drugs-idUKKBN15N18Z

Quackenbush, Casey. 2018. “Hundreds Killed in Bangladesh Since Philippines-Style Drug Crackdown Began, RightsGroups Say.” Time .URL: http://time.com/5340677/bangladesh-philippines-drugs-death-toll/

Querubin, Pablo. 2016. “Family and Politics: Dynastic Persistence in the Philippines.” Quarterly Journal of Political Science11(2):151–181.

Ramos, Christia Marie. 2019. “House panel to call more DPWH execs over flood control scam.” Inquirer .URL: https://newsinfo.inquirer.net/1070015/house-probe-dpwh-flood-control-scam

Ressa, Maria. 2015. “#TheLeaderIWant: Leadership, Duterte-style.”. URL: https://www.youtube.com/watch?time_continue=8&v=ow9FUAHCclk.

Riker, William H. 1982. Liberalism against populism. San Francisco: WH Freeman.

Sachs, Jeffrey D. 1989. “Social conflict and populist policies in Latin America.”.

Seligson, Mitchell A. 2007. “The rise of populism and the left in Latin America.” journal of Democracy 18(3):81–95.

Sidel, John. 1999. Capital Coercion and Crime: Bossism in the Philippines. Stanford, California: Stanford University Press.

Slater, Dan. 2013. “Democratic careening.” World Politics 65(4):729–763.

Spruyt, Bram, Gil Keppens and Filip Van Droogenbroeck. 2016. “Who supports populism and what attracts people to it?”Political Research Quarterly 69(2):335–346.

Stokes, Susan C, Thad Dunning, Marcelo Nazareno and Valeria Brusco. 2013. Brokers, voters, and clientelism: The puzzleof distributive politics. Cambridge University Press.

United Nations Office on Drugs and Crime. 2019. “Synthetic Drugs in East and South-East Asia.” Global SMART Program.

Wantchekon, Leonard. 2003. “Clientelism and voting behavior: Evidence from a field experiment in Benin.” World politics55(3):399–422.

27

Work

ing pa

per -

do no

t cite

7 Appendix

Using a 2 percent cutoff for “very close" elections, we find that the results are almost identical, however with larger standarderrors due to the loss of sample size. Table 7 shows that the effect on drug-related blotter reports remain ‘statisticallysignificant’ at the p<.1 level.

Table 7: Effect on Post-Duterte Crime Rates

(1) (2) (3) (4) (5) (6)Dependent variable: Drug-related Theft Assault Homicide Robbery Rape

Non-establishment Mayor * Post-Duterte .40+ .08 .05 .25 .11 .05(.24) (.20) (.41) (.19) (.09) (.05)

LP Post-Duterte Mean .95 .94 1.57 .60 .39 .33 4.93N 164 164 164 164 164 164Clusters 82 82 82 82 82 82

Notes: Municipality and period fixed effects, SE clustered by municipality, + = p<.1 Sample: Municipalities with very close(<2% margin of victory) for the 2016 mayoral election.

Table 8: Effect on Post-Duterte Crime Rates (ANCOVA)

(1) (2) (3) (4) (5) (6)Dependent variable: Drug-related Theft Assault Homicide Robbery Rape

Non-establishment Mayor .31* -.06 -.13 .17 -.01 -.01(.14) (.11) (.23) (.16) (.05) (.03)

Liberal Party MeanN 189 189 189 189 189 189

Region fixed effects, controlling for pre-treatment crime rates * p<.05, Stata robust SE

28

Work

ing pa

per -

do no

t cite

Figure 9: Violent Crime in Police Blotter Reports

- Duterte Takes Office

05

1015

Viol

ent C

rime

per 1

,000

regi

ster

ed v

oter

s

Jul 2016 Jan 2017 Jul 2017 Jan 2018 Jul 2018Month

Unaligned Mayor Liberal Party Mayor

Violent Crime in Police Blotter (close elec.)

Notes: Among 189 municipalities with very close (<5% margin of victory) for the 2016 mayoral election. Violent crimeincludes homicides, rape, robbery, assault.

29

Work

ing pa

per -

do no

t cite