War costs and public support for domestic
counterinsurgency: Evidence from Thailand
Bethany Lacina
Assistant Professor,
Department of Political Science,
University of Rochester
PO Box 270146
Rochester, NY 14627-0146
585-273-5842 (phone)
585-271-1616 (fax)
[email protected]
Spring 2016
Abstract
When does public opinion support domestic counterinsurgency? Do the costs of
conflict breed war weariness, increasing support for compromise? Or does violence
harden support for war? I examine country-wide opinion polling from Thailand concerning government policy toward insurgency in the south. The perceived severity of
insurgency predicts greater support for southern autonomy. That relationship is driven
by respondents outside of the conflict theater, however. The number of military fatalities associated with a respondent’s regional army is also positively correlated with
support for southern autonomy. As placebo tests, I demonstrate that neither perceived
conflict severity nor military fatalities predict greater support for reconciliation in the
simultaneous conflict at the capital. Together, the results suggest Thai public opinion
on domestic counterinsurgency is subject to a casualty fatigue process similar to overseas wars but rarely applied to civil conflict.
Word count: 9805
When does the public support domestic counterinsurgency? As the costs of a civil war mount,
do citizens demand the government compromise? Both scholars and policymakers believe that the
costs of overseas wars can create a popular backlash against continued military operations. This
“casualty fatigue” or “war weariness” pattern has been extensively studied in the United States and
Europe. Does a comparable process influence public opinion during a domestic conflict?
Much of the literature on civil war and counterinsurgency implies that the answer is no. Public
opinion radicalizes during civil violence, partly but not solely due to personal exposure to violence.
Deepening hatreds and intolerance are one reason civil wars tend to be much longer and more likely
to recur than international wars (Collier et al., 2003).
This article investigates public support for domestic counterinsurgency using polling from
Thailand in 2010. I find that the perceived severity of conflict in southern Thailand is positively correlated with support for greater southern autonomy. This pattern is consistent with war-weariness.
However, the correlation is only apparent among respondents living outside the conflict theater.
War weariness may be most relevant as a model of public opinion when conflict is spatially isolated and the policies at issue have limited national ramifications. I conduct two investigations to
demonstrate the credibility of the war-weariness finding. First, I create a placebo test, showing
that perceived costs of the southern conflict do not increase support for compromise in the contemporaneous conflict at the capital. Second, I take advantage of the regional structure of the four
commands of the Royal Thai Army. I show that the number of military fatalities associated with a
respondent’s regional army is positively correlated with support for compromise in the south.
In this article I make two contributions. The first is to contrast two research programs on public
opinion and war costs, one from overseas wars and one from domestic wars. By presenting and
investigating both programs, I make the first step toward determining the conditions under which
each model of public opinion is most likely to apply. Second, the case of Thailand brings the
research on public opinion during internal war to a previously-neglected type of conflict. Studies
of opinion polling in civil war disproportionately concern small countries where war fighting has
1
directly affected a large share of citizens. Here, I examine the more common scenario of an internal conflict that is geographically peripheral and personal experience of violence is relatively
rare. I find a difference in reactions to war costs between in- and out-of-theater respondents, with
only the latter group conforming to a conflict fatigue model. I gauge the plausibility of this conflict
fatigue finding with placebo tests and by examining the relationship between military fatalities
and public opinion in a subsample of my data. I conclude by arguing that the casualty fatigue hypothesis may be the best model for understanding out-of-theater populations’ support for domestic
counterinsurgency.
Conflict cost and public support for war
Two literatures study popular reaction to war: the first on overseas military campaigns, the second
on counterinsurgency and civil war. Data from overseas war show a remarkably consistent relationship between the costs of conflict and decreased support for military action. Total casualties,
adverse trends in casualties, and individuals’ social and spatial proximity to casualties all depress
support for overseas military operations. By contrast, much of the commentary and scholarship on
civil war implies that the public becomes more belligerent as war continues.
Overseas conflict: Costs create war-weariness
In the study of public support for foreign war, the implicit model is one of citizens weighing the
costs and benefits of continued conflict (Gartner, 2008a; Gartner and Segura, 1998; Mueller, 1973):
Actual or “objective” information regarding the costs and benefits of war is carefully
weighed by an individual before he or she forms attitudes regarding the legitimacy
of the war, the long-term prospects for success, the worthiness of the goals of the
intervention (Boettcher and Cobb, 2006, 838).1
1 This
updating process is subject to individual biases and framing effects (Althaus et al., 2012; Boettcher and
2
As the costs of war, especially casualties, accumulate, individuals revise their estimates of the total
costs of war upward and update their beliefs about the likely ultimate success of war (Eichenberg,
2005; Gelpi et al., 2005, 2009). Some people place such a low or high value on the outcome of
the conflict that new information about costs or the probability of success cannot shift their views.
On average, however, the accumulation of casualties and an increasing rate of casualties depress
public support for war (Gartner and Segura, 2008).2 US political elites also respond to war time
casualties, especially casualties in their constituency, by decreasing support for war (Gartner, 1997;
Murray, 2001; Myers and Hayes, 2010).
US respondents respond most to costs befalling their families and acquaintances (Davenport,
2015; Gartner, 2008b). Spatial proximity to families of the deceased also matters. For example,
individuals living closer to the families of deceased soldiers were less likely to support the Vietnam
War (Gartner et al., 1997; Gartner and Segura, 2000) and the second Iraq war (Althaus et al.,
2012; Hayes and Myers, 2009). In US elections, voters’ punishment of incumbent politicians
is conditioned by local war casualties rather than national casualties (Gartner and Segura, 2008;
Gartner et al., 2004; Grose and Oppenheimer, 2007; Karol and Miguel, 2007; Kriner and Shen,
2007).
War weariness in civil war?
In civil war, does violence impress individuals with the costs of conflict and thereby reduce their
support for war? Some theoretical and empirical literature implies it does. During the Vietnam
War, Leites and Wolf (1970) portrayed civilians making cost/benefit calculations in determining
support for rebels and governments.3 There is also some empirical evidence that internal war
Cobb, 2006; Gaines et al., 2007; Johnson and Tierney, 2006), as well as equity concerns (Johns and Davies, 2014;
Kriner and Shen, 2014; Kull et al., 1997). A critique of war weariness arguments holds that public support for war
follows cues from elected officials and the media (Baum, 2002; Baum and Groeling, 2010; Berinsky, 2007; Jakobsen
and Ringsmose, 2015; Pevehouse and Howell, 2007).
2 On casualties and voter turnout, see Koch and Nicholson (2015).
3 See also Mason (1989).
3
produces civilian war weariness. During the US Civil War, Congressional incumbents lost more
votes in districts with more local casualties (Carson et al., 2001). Blair et al. (2013) fielded an
endorsement experiment to measure support for insurgents in Pakistan. They found disaffinity for
insurgents among poor, urban Pakistanis but only in districts that suffered insurgent violence in the
last year. They interpret this pattern as evidence “that exposure to terrorist attacks reduces support
for militants” (30).
Nonetheless, treatments of internal war (Kilcullen, 2009; Mason, 2004) and counterinsurgency
doctrine (US Armed Forces, 2006) more often presume that civilians become more belligerent in
response to the costs of civil war, especially government repression. Both insurgents and their
supporters frequently report government violence as a reason for their own conversion to the militant cause (Askew and Helbardt, 2012; Mahmood, 1996; Popkin, 1979; Wood, 2003). Violence
by non-state actors receives less attention but may also be radicalizing. Some of this radicalization
reflects personal exposure to violence.4 However, the connections between personal experience of
violence and policy views are not straightforward, thanks to the role of other predispositions and
policy beliefs (Longo et al., 2014; Lyall et al., 2013).
Personal experience of violence is, in any event, only one component of public opinion during
civil war. Most civil wars only directly involve a small portion of a country’s population and
territory (Dittrich Hallberg, 2012). Yet, the polarizing effect of civil war on public opinion is
thought to be much broader. For example, the World Bank’s famous report on “conflict traps”
hypothesized that one reason civil war is more common in countries that have already had a prior
war is that conflict “intensifies hatreds” (Collier et al., 2003, 87):
. . . hatreds build up during periods of violence, leaving the society polarized. People
on both sides want vengeance for atrocities committed during the conflict and these
may supplant any prior grievances (88).
4 Bakke
et al. (2009); Bellows and Miguel (2009); Blattman (2009); Canetti-Nisim et al. (2008, 2009); De Juan and
Pierskalla (2016); Voors et al. (2012). Some research suggests community-building effects of trauma (Bellows and
Miguel, 2006, 2009; Blattman, 2009; Voors et al., 2012).
4
Sambanis and Shayo (2013) formalize this intuition, arguing that violence can reinforce conflictual
identities and undermine shared identity. Supporting that thesis, Tir and Singh (2015) use the
World Values Survey to show that social intolerance is higher in countries that have experienced
secessionist wars.5 The targets of this intolerance were all out-groups, not only the ethnic group(s)
on the opposite side of the recent civil war. Hutchison (2014) finds that civil conflict reduces public
support for civil liberties of non-conformist organizations. Multiple studies of public opinion in
Israel imply that Jewish Israelis have become more opposed to compromise with Palestinians in
response to violence or perceived threat (Berrebi and Klor, 2008; Canetti-Nisim et al., 2008, 2009;
Maoz and McCauley, 2009; Shayo and Zussman, 2011).
War costs and support for domestic counterinsurgency
In sum, the hypothesized effect of conflict costs on public support for war is very different in the
literatures on overseas and on civil conflicts. A few studies on civil conflict find evidence that war
costs push the public toward peace but, in the main, scholars have argued the opposite. Studies
of public opinion during civil war have been disproportionately drawn from Israel. Afghanistan,
Burundi, Pakistan, the Caucasus region in Russia, Nepal, Sierra Leone, and Uganda are also represented. Based on that list, conflicts over control of the central government have been studied more
often than conflicts over territorial autonomy or separation. Existing studies also come disproportionately from wars with extensive involvement by foreign militaries, from geographically small
countries, and from wars that had a relatively large conflict theater. The next section turns to the
context for this study, the war over ethnoterritorial autonomy in southern Thailand. Compared to
other civil conflicts where public opinion has been studied, this war is not very internationalized
and directly affects a limited portion of the country’s territory and citizens.
5 They
found no changes in social tolerance after non-secessionist wars.
5
Pattani Malay insurgency in Thailand
Thailand’s southern provinces of Yala, Narathiwas, and Pattani (see Figure 1) are ethnically and
religiously distinct from the rest of the country.6 About 80% of the population in these three
provinces speaks Pattani-Malay (also called Yawa) as a first language, while Thai is the dominant
language elsewhere. Almost all Pattani-Malays are Muslims, while Thailand as a whole is 90%
Buddhist. The neighboring district of Songkhla has a small population of Pattani-Malay Muslims
concentrated along the border with Pattani and Yala.
Yala, Narathiwas, Pattani, and Songkhla’s border districts were once the sultanate of Patani
(Haemindra, 1976).7 The current insurgency emphasizes this regional Pattani-Malay identity over
Muslim identity. Although there are Thai- and Chinese-ethnicity Muslims in southern Thailand,
the insurgency is exclusively oriented toward Pattani-Malay Muslims.
A separatist insurgency in the 1960s and 1970s presaged a new phase of southern violence
beginning in the early 2000s (Funston, 2008). A new phase of southern violence began in the early
2000s. In 2004, militants made a successful surprise attack on an army base, gaining national
and international attention. The insurgents were highly fractionalized and, through 2010, had no
identifiable organization or leadership with credible control over rebel violence (Funston, 2008;
Liow and Pathan, 2010).
Power Vacuum
McCargo (2008) and Mahakanjana (2006) argue that the proximate cause of conflict was a power
vacuum created by Thaksin Shinawatra, Thailand’s elected prime minister between 2001 and 2006.
Thaksin is notable for being a member of the country’s Sino-Thai nouveau riche, as opposed to the
traditional political elite mostly descended in some way from the royal family.
6 Figure
1 based on GADM (2012) and census region definitions from National Statistical Office (2010).
district, to the west of Songkhla, is 70% Malay-speaking Muslims but was not a part of the Patani sultanate
(Funston, 2008, xiii).
7 Satun
6
Figure 1: Thailand’s four census regions—north, northeastern, central and southern—and its southernmost provinces
N
NE
C
Bangkok
S
Songkhla
Pattani
Yala
Narathiwat
Thaksin attempted to build a political following in the far south by dislodging bureaucrats and
military personnel affiliated closely with the monarchy. In 2002 Thaksin dissolved the Southern
Border Administration Center (SBPAC) and Civil-Military-Police Task Force 43 (CMP-43), unelected bodies that coordinated military and civil service activity in the south (Funston, 2008, 25).
Responsibility for southern security was transferred from the military to the national police. Neither the SBPAC nor CMP-43 had provided regional political autonomy, but they did serve as a
means for central officials to consult with and coopt Pattani Malay elites (Mahakanjana, 2006).
The military also maintained a network of informants inside separatist outfits in the south. When
the police took control of southern security, approximately 20 of these informants were killed,
allowing the insurgency space to grow (McCargo, 2008, 115).
The second piece of Thaksin’s plan to build a following in the south was to forge a relationship
with the largest Pattani Malay political party, called Wadah, headed by Wan Nor (Wan Muhammad Nor Matha). In 2002, Wadah joined Thaksin’s party and Wan Nor became interior minister.
However, alignment with the government undermined Wadah’s popularity by implicating it in
state repression, especially a 2004 raid on the Kru-Ze mosque in Pattani and dozens of protestor
deaths in military custody after a 2004 demonstration in Tak Bai, Narathiwat. Thus, both pieces of
Thaksin’s strategy to gain a political following in southern Thailand—police-led security and the
Wadah alliance—faltered.
Counterinsurgency policy
Between 2004 and 2010, successive Thai governments had a consistently hardline response to insurgency. Thaksin increased force deployments in the south and rejected any talk of political compromise (Albritton, 2006). After a 2006 coup, the military government reestablished the SBPAC
and a joint civil-military-police command for the south, but with greater military dominance of
each (Funston, 2008, 25). Emergency laws gave the military impunity from prosecution for deaths
in custody and torture. The military’s leading role was not countered by the next two civilian
8
governments: a Thaksin-aligned government elected in 2007 or the anti-Thaksin Democrat Party
government that took power in 2008 (Abuza, 2011; Dalpino, 2011). Instead, both governments
were preoccupied by the “colored shirt” protests in Bangkok, described below.
Throughout the 2000–2010 period, the Thai central elite steadfastly opposed autonomy for
southern Thailand, even in limited forms such as an elected governor. (Thailand is a unitary state
with unelected provincial level governments).8 Bangkok’s resistance to autonomy was embodied
in a National Reconciliation Council (NRC) appointed in 2005:
Reporting in June 2006, [the NRC] made a series of modest proposals for improving the quality of justice, security, and governance in the deep South . . . The NRC’s
proposals were considered too progressive by most government officials but did not go
nearly far enough for most Malay Muslims. The NRC refused to engage seriously with
ideas of substantive decentralization . . . let alone any proposals for different forms of
autonomy (McCargo, 2008, 10).9
Gunaratna and Acharya (2013), rightwing analysts who oppose southern autonomy (174), also
admit that there was no political will in the capital to implement the NRC’s recommendations and
that powerful conservatives in the government immediately moved to quash the NRC’s tangible
recommendations, such as the recognition of Malay as a “working language” in the south (156–
157).
While central elites have taken a hardline against southern autonomy, the public has been disengaged from the southern conflict, according to most analysts. Albritton (2006) wrote that in 2005
“A majority of the population . . . is largely oblivious to the ongoing conflict” (172). As turmoil in
the capital increased, Prasirtsuk (2009) observed that “The conflict in Thailand’s south continues
but has been eclipsed by the political crisis [in Bangkok]. The southern insurgency rarely made
headlines in 2008.” (183). The same indifference was reported in 2010: “Although violence in the
8 The
9 See
exceptions are Bangkok and Pattaya City, both in central Thailand.
also McCargo (2012, chp. 4).
9
Muslim South, targeting teachers and other community leaders, did not diminish in 2010, the lack
of progress had little impact on the government’s standing in other regions of Thailand” (Dalpino,
2011, 160). Quotes like these underline that civil conflicts can be spatially isolated and of limited
nationwide salience. These circumstances suggest an analogy to overseas war and the possibility
that casualty fatigue is a useful means of understanding public support for ongoing COIN. On the
other hand, the salience of Thailand’s southern conflict was apparently so low that its costs may
not register in public opinion.
Survey of Thailand, 2010
The Asia Foundation (TAF) fielded a national survey with 1,800 respondents from September 17
to October 23, 2010 (Asia Foundation, 2011). The study included 300 respondents from Narathiwas, Pattani, and Yala, 100% of whom were self-reported Pattani Malays.10 Narathiwas, Pattani,
Yala, and Bangkok provinces were designated for inclusion in the survey prior to further sampling. Investigators drew a random sample of the remaining provinces in the country, stratifying
based on the four regions used in the census of Thailand (Figure 1): north, northeast, central, and
south.11 Within provinces, an equal number of rural and urban respondents were sampled, except
in Bangkok, which is entirely urban. Thus, respondents’ sampling probabilities depended on their
census region, province, and rural versus urban location.12
My dependent variable taps public support or opposition to the government’s counterinsurgency policies with a question about autonomy as a conflict-resolution measure. TAF asked respondents: “Some have suggested that political decentralization or limited autonomy (but not territorial
10 Some
villages in the three southern provinces were omitted from the sampling frame due to insecurity.
A.1 in the Supplemental Materials shows the surveyed provinces.
12 Sampling weights provided by TAF also reflect sex and respondents’ age cohort, presumably to adjust for nonresponse. Throughout the analysis in the main text, I use these survey weights and cluster observations by sampling
strata when calculating standard errors. I report results based on unweighted and unclustered data in the Supplementary
Materials (Tables A.8 and A.11). Pattani Malays are over represented and rural respondents are under represented in
the unweighted sample. See the summary statistics for weighted and unweighted data in Tables A.1–A.7.
11 Figure
10
separation) might help resolve the long-term conflict in southern Thailand. Do you agree or disagree?”13 Given the entrenched government resistance to autonomy, respondents who agreed that
autonomy might resolve conflict were deviating sharply from Bangkok’s counterinsurgency strategy. Outside of the three deep south provinces, 63% of respondents endorsed southern autonomy
in response to this question, indicating substantial dissent from the central elites’ military-driven
COIN strategy.14 75% of Pattani respondents endorsed autonomy as a means to end conflict.
I measure the key independent variable, respondents’ beliefs about the costs of the conflict
in the south, based on replies to an open-ended question regarding Thailand’s biggest problems.
Respondents were asked, “In your view, what is the biggest problem facing Thailand?” I code High
perceived conflict costs as a 0/1 variable, with a value of 1 for respondents who described conflict
in southern Thailand as the country’s biggest problem. Not surprisingly, given central government
turmoil at the time, only a small number of respondents—4%—listed conflict in the south as the
country’s gravest problem. The most common answer was political conflict in the capital (43%),
followed by economic problems (33%). A plurality of Pattani Malay respondents put the conflict
at the center (35%) at the top of their list of problems but conflict in the south was the second most
frequent answer (24%).
The wording of this question is unfortunate because it does not capture the full range of variation in the perceived costs of southern conflict. Instead, the question only identifies respondents
who thought those costs were high enough to eclipse other problems in Thailand. In the appendix,
I use an alternative measure of perceived conflict severity (Tables A.10 and A.12); the results support the findings in the main text. Survey respondents were asked whether the country was headed
13 The
full text of all survey questions is provided in the Supplementary Materials.
This question taps both respondents’ willingness to see autonomy in the south and beliefs about its efficacy as a
conflict-resolution measure. Studies of casualty fatigue using observational data have difficulty distinguishing beliefs
about the desirability of outcomes from beliefs about what is likely to happen if the government tries one policy versus
another. Only experimental studies have thus far been able to parse that distinction (e.g., Gelpi et al., 2005, 2009).
14 A 2009 TAF survey of the same population found 48% support for autonomy (Meisburger, 2009, 50). Gunaratna
and Acharya (2013) report two national surveys from 2005 in which a majority of respondents opposed martial law in
the south and the punitive security zone system the government was implementing at the time (153).
11
in a positive or negative direction and then asked to explain why they thought so. The latter question was open-ended and the enumerators recorded up to six responses. I used respondents’ lists of
problems to code a second measure of perceived conflict costs: whether the respondent listed the
southern conflict as one reason for pessimism about Thailand. This variable is available only for
respondents who thought Thailand was headed in the wrong direction, although this is the majority
of respondents (61%). In that subset, about 10% cited the conflict in the south as a reason Thailand
was headed in the wrong direction, including 8% of non-Malays and 33% of Malays.
Perceived costs and support for southern autonomy
Table 1 examines the relationship of perceived conflict costs with support for government compromise in a domestic COIN context. Model 1 reports a bivariate logistic regression of Thai respondents’ perceptions of the costs of conflict in the south as a predictor of support for southern
autonomy.15 The correlation is positive and statistically significant. The estimated coefficient (1.0)
translates to an odds ratio of 2.7. To interpret that figure, consider that the average rate of support
for southern autonomy in the data is about 60%. If the corresponding odds are multiplied by 2.7,
the probability of support goes up to about 80%.
Table 1 continues with multivariate models of the relationship between perceptions of the severity of southern conflict and endorsement of southern autonomy. Model 2 includes a dummy variable for Pattani Malay ethnicity and measures of respondents’ age, sex, education, income, and
whether they live in an urban or rural area.16 After adding these controls, the magnitude of the
15 In order to estimate these models and the models in Table 2 on the same sample, I excluded respondents who
answered the southern autonomy question but not the questions on protestor pardons described below. In the Supplementary Materials, I include these respondents in an analysis of support for southern autonomy and find results similar
to those in the main text (Tables A.9 and A.15).
16 Age is a continuous scale ranging from 2 (18-19 years old) to 11 (Above 80 years old). Education is captured as
dummy variables for secondary/vocational education and for post-secondary education. TAF used different income
scales for urban and rural areas and for Bangkok, in each case ranking a respondent from one to nine. I use these
rankings as a continuous measure, reflecting respondents’ incomes relative to their communities.
12
Table 1: Logistic regressions of respondent support for southern autonomy as a conflict resolution
measure and perceived costs of the southern conflict
Model 1
Model 2
Model 3
Model 4
Model 5
1.0∗
(0.37)
1.0∗
(0.37)
1.1∗
(0.39)
1.5∗
(0.39)
1.6∗
(0.42)
0.37
(0.33)
0.23
(0.36)
0.63
(0.35)
0.50
(0.40)
-1.6
(0.86)
-1.6
(0.90)
High perceived
conflict costs
Pattani Malay
Costs*Pattani Malay
Favors
decentralization
0.91∗
(0.15)
0.92∗
(0.15)
Military role too
large
-0.084
(0.16)
-0.092
(0.16)
Satisfaction w
national government
0.13
(0.11)
0.12
(0.11)
Age
0.0070
(0.031)
0.0019
(0.033)
0.0074
(0.032)
0.0026
(0.034)
Female
0.087
(0.14)
0.11
(0.15)
0.097
(0.14)
0.12
(0.15)
Secondary education
0.35∗
(0.17)
0.37∗
(0.19)
0.36∗
(0.17)
0.37∗
(0.19)
Post-secondary
education
-0.32
(0.23)
-0.29
(0.24)
-0.32
(0.23)
-0.29
(0.25)
Income
0.025
(0.043)
0.028
(0.046)
0.024
(0.043)
0.027
(0.046)
Rural
0.25
(0.13)
0.13
(0.14)
0.26
(0.13)
0.14
(0.14)
10
1419
10
1334
Sampling strata
10
10
10
Sample size
1419
1419
1334
Ln likelihood
Hypothesis tests (F-statistics from Wald tests with weighted data):
Perceived costs + Costs*Pattani = 0
Standard errors in parentheses.
∗
p < 0.05
0.019
0.0049
coefficient on perceived costs is unchanged and significant at the 95% confidence level.17
Model 3 includes three controls that capture respondent views of conflict in Bangkok. In 2010
nationwide decentralization was one of the political reforms being considered to resolve the crisis
at the center. Respondents’ views on southern autonomy may have been informed by their support
for national decentralization. I control for whether a respondent told the TAF that they favored
decentralization throughout Thailand (Favors decentralization).18 Second, respondent attitudes
toward policy in the south may primarily reflect their view of the military’s role in politics—recall
that there was a coup in 2006. Military role too large indicates that a respondent told TAF that
the military was too prominent in Thai politics. Finally, Satisfaction w national government is a
ranking from one to four of how satisfied the respondent was with the Bangkok regime.
The relationship between conflict costs and support for southern autonomy is more stark after
controlling for Thailand-wide decentralization, the military’s role in politics, and satisfaction with
the national government (Model 3). The coefficient on perceived costs increases by 10% and
remains significant.
Pattani Malays and conflict costs
Was the relationship between perceived conflict costs and support for southern autonomy the same
for Pattani Malays and other Thai citizens? In the TAF survey, all of the respondents in Yala,
Narathiwat and Pattani were Pattani-speakers. The sole Pattani-speaker in the survey who was
not in the three southernmost provinces lived in adjacent Songkhla, which has seen some insurgent activity. By virtue of both their ethnicity and their location, the Pattani Malay respondents
had more stake in southern autonomy and a greater likelihood of personal exposure to conflict
17 This
is the only result reported here and in subsequent tables that is overturned if survey weights are not used
(see Table A.8). If Model 2 is reestimated without weighted data, the coefficient on perceived costs is 0.55 and has a
p-value of 0.070 in a two-tailed test of the null hypothesis.
18 After an explanation of decentralization, respondents were asked “Which statement is closer to your point of
view? (1) Better to decentralize to local government. Or (2) The central government having the authority for decision
making, as at present, would be more effective and efficient.”
14
violence. As noted above, Pattani Malays were much more likely to rate the conflict as Thailand’s most important problem. The policy solution to conflict at issue here, southern autonomy,
would directly change the relationship between in-theater respondents and the government but
would have only secondary effects on respondents elsewhere. Given the profound implications,
in-theater respondents’ opinions on autonomy should be relatively stable compared to respondents
in other provinces. Small changes in the perceived costs of the southern conflict could shift the
views of Thais elsewhere, who do not attach much intrinsic importance to southern autonomy. By
contrast, Pattani Malays are likely to hold strong views on the merits of autonomy. Therefore,
only dramatic changess in the perceived severity of southern conflict would outweigh preferences
regarding southern autonomy.
Models 4 and 5 add an interaction between the Pattani Malay dummy variable and the perceived
costs variable. Model 4 includes demographic control variables, Model 5 contains the full set of
controls. The implications of these models are best captured in Figure 2. The figure plots the
predicted difference in the probability that a respondent supported southern autonomy in case s/he
believed conflict there was Thailand’s greatest problem. The predicted differences are plotted
with 95% confidence intervals and displayed separately for non-Pattani and Pattani respondents.19
For non-Pattani respondents, predicted support for southern autonomy is 10–20% higher among
those who perceived the costs of conflict as high. By contrast, among Pattanis, there is essentially
no difference in support for autonomy between those who rated the conflict as Thailand’s major
problem and others. The predicted change is small in magnitude and switches signs between
Models 4 and 5, consistent with a null relationship. At the bottom of Table 1, I report Wald tests
of the hypothesis that among Pattani Malays the combined coefficient on perceived costs equals
zero. The tiny numbers reported there are actually F-statistics and not p-values. It is impossible
to reject the hypothesis of a null effect in either Model 4 (p =0.89) or Model 5 (p =0.94). On the
other hand, the wide confidence intervals on the Pattani/costs interaction term also means that the
19 Control
variables are set to the mean value for continuous measures and modal value for indicator variables.
15
Difference in support for autonomy if
perceived costs of southern conflict high
Figure 2: Predicted difference in the probability of support for southern autonomy if respondent
perceived high costs to the southern conflict, plotted for non-Pattani and Pattani respondents, with
95% confidence intervals
.3
.2
.1
0
−.1
−.2
Model 4
Model 5
Respondent was:
Non−Pattani
Pattani
Predicted changes in probability plotted with 95% confidence intervals
predicted effect of perceived conflict severity among Pattani Malays overlaps the predicted effect
on other respondents. Even with this uncertainty, the results in Models 4 and 5 reveal that the
relationship between perceived costs and support for autonomy in earlier models is being produced
by out-of-theater, non-Pattani respondents.
Summary of main results
Thai-speaking respondents, all of whom lived out-of-theater, display a war-weariness pattern familiar from studies of overseas wars. Those who thought the southern insurgency was a major
problem wanted the government to compromise on autonomy. People who rated the costs of conflict lower were more content for the government to stay the course. Among Pattani Malays in-
16
theater, ratings of the conflict’s severity do not predict support for autonomy as a peacebuilding
measure. With a large personal stake in the ultimate outcome of the conflict, Pattani Malay support
for autonomy is similar regardless of how respondents rated the costs of conflict. This result does
not rule out major effects of war costs on Pattani Malays in other respects such as interethnic trust.
Also, an important limitation of the study here is that ethnic identity and geographic proximity
to conflict are perfectly correlated in the survey sample. These factors need to be disentwined in
future work.
Nonetheless, the results suggest that in civil conflicts that are spatially isolated and involve
issues of limited scope, there is an important distinction in how to understand public opinion among
in- versus out-of-theater populations. That insight has not emerged from previous studies that
have tended to focus on wars for control of the central government and wars in which the theater
of violence encompassed most of the country. Yet, the majority of domestic counterinsurgency
takes place in conflicts that, like the war in Thailand, are comparatively isolated both in terms of
geography and the issues at stake.
The remainder of this essay tests the robustness of the war-weariness model as an explanation
for support for southern autonomy. The next section compares respondents’ perceptions of the cost
of southern conflict to their support for compromise between competing factions in Bangkok. The
aim is to conduct a placebo test. If the costs-of-southern-war variable predicts support for conciliatory policies in a different conflict, the costs variable may be measuring a general propensity for
compromise. The final section of the paper looks for evidence that objective measures of military
costs in the south are correlated with greater support for southern autonomy among out-of-theater
respondents.
17
A placebo test: Political reconciliation in Bangkok
One means to judge the credibility of the results in Table 1 is to compare perceptions of the costs
of the southern conflict to survey questions about policies not directly related to those costs. In
this section, I compare perceptions of the costs of southern conflict to respondent support for
conciliatory policies toward clashing factions at the capital. It is possible that some people see
all conflict as very high cost and have a strong preference for conciliation or compromise. In that
case, there might be a strong positive relationship between perceived costs of conflict in the south
and support for compromise in the national arena.20 On the other hand, the absence of a positive
correlation would imply that the connection between the perceived costs of southern conflict and
support for autonomy reflects opinions specific to events in the south.
For this placebo test, I use questions about the competing “colored shirt” protest movements
in Bangkok, which dealt with control of the central government. To review, in 2001 and 2005,
Thaksin Shinawatra’s party won large victories in national elections. In 2006, Thaksin was overthrown in a military coup and went into exile. His allies won new elections in 2007 but their
government was destabilized by opposition “yellow shirt” demonstrations in Bangkok. When the
anti-Thaksin Democrat Party took power in 2008, the pro-Thaksin response was a corresponding campaign of “red shirt” demonstrations. Red shirts characterized Thaksin opponents as antidemocratic for supporting military and judicial coups. Yellow shirts argued Thaksin was the real
threat to democracy, pointing to his weakening of oversight agencies and the free press and crediting his electoral success to corruption and vote buying.
In its 2010 survey, TAF asked respondents whether charges of terrorism against red shirt protest
leaders “should be dropped to promote political reconciliation” and posed the same question regarding yellow shirt protest leaders. I coded a dummy called Support central political reconciliation that takes a value of one if the respondent supported dropping the charges against both yellow
20 Or
resolving conflict at the capital might be seen as a necessary step to resolving conflict in the south.
18
and red shirt leaders. Respondents who supported maintaining the charges against either or both
sets of leaders are scored as a zero. About 60% of respondents supported dropping charges against
both red and yellow shirt protest leaders; Pattani Malays and non-Pattani Malays were equally
likely to support this form of political reconciliation.
Table 2 reports logistic regressions of the perceived costs of the southern conflict and support
for central political reconciliation. The bivariate regression (Model 6) shows a negative and statistically insignificant correlation between these measures. Adding control variables (Models 7 and 8),
the coefficient on southern conflict severity remains negative but becomes statistically significant.
The negative correlations here suggest that the results above—i.e., the positive correlation between
perceived severity of southern conflict and support for autonomy (Table 1)—are not driven by a
general taste for conciliatory policies. However, the presence of a statistically significant rather
than a null relationship in Models 7 and 8 is unexpected.21
The results in this section refute the hypothesis that the relationship between perceived costs of
southern conflict and support for southern autonomy reflects a general preference for compromise
in the face of violence. The next section investigates the plausibility of a war-weariness effect
in Thailand by asking a different question. Is there evidence that differential exposure to the
“objective” costs of the southern conflict translates into support for southern autonomy? To gauge
the answer, I take advantage of the geography of the Royal Thai Army (RTA).
21 A
possible interpretation of this result is that it is an artifact of the wording of my question for gauging perceived
costs of the southern conflict. Recall that the question asked for the biggest problem facing Thailand. Respondents
whose response was the southern conflict implicitly revealed that they considered central conflict secondary. These
respondents might also have seen little need for reconciliation in Bangkok. A test of this interpretation is to check
whether my alternative measure of perceived costs (described above) relates to support for central reconciliation; this
measure is based on a survey item asking respondents to list Thailand’s problems rather than asking them to rank those
problems. In Table A.12, the alternative measure of perceived costs does not have a large or statistically significant
correlation with support for central political reconciliation. The sign on the correlation is also unstable. These null
results suggest the negative correlations in Table 2 are due to the limitations of my main measure of perceived costs.
19
Table 2: Logistic regressions of perceived costs of the southern conflict and respondent support for
dropping charges against leaders of Bangkok protests
Model 6
Model 7
Model 8
-0.63
(0.33)
-0.84∗
(0.33)
-0.82∗
(0.33)
0.50
(0.31)
0.53
(0.33)
High perceived
conflict costs
Pattani Malay
Favors
decentralization
-0.096
(0.15)
Military role too
large
0.55∗
(0.16)
Satisfaction w
national government
-0.31∗
(0.10)
Age
-0.029
(0.031)
-0.045
(0.032)
Female
-0.0047
(0.13)
0.037
(0.14)
Secondary education
-0.26
(0.17)
-0.28
(0.18)
Post-secondary
education
-0.57∗
(0.23)
-0.62∗
(0.25)
Income
-0.088
(0.045)
-0.11∗
(0.047)
Rural
0.27∗
(0.13)
0.28∗
(0.14)
10
1419
10
1334
Sampling strata
Sample size
Standard errors in parentheses.
10
1419
∗
p < 0.05
Military fatalities and support for southern autonomy
Studies of casualty fatigue frequently compare the number of deaths from a respondent’s geographic jurisdiction to support for war. US public opinion has been compared to state and countrylevel fatalities, for example. The theory is that “in assessing these [war] costs, a respondent cannot
help but weight proximate experiences more heavily, if for no other reason than this information
is both salient and readily accessible” via both media channels and informal information networks
within communities (Gartner et al., 1997, 674).
Unfortunately, I do not have access to information on the home provinces of soldiers killed in
southern Thailand. However, I can take advantage of the geographies of recruitment by Thailand’s
four regional armies. Each of the four armies draws its membership primarily from a particular
census region of Thailand. Also, in October, 2007, each army was assigned to a different southern
district. Thus, for each respondent, I know which army’s region they live in. I can also observe
military casualties in the southern district where that army was posted between October 2007 and
the TAF survey.
By 2010, about 30,000 soldiers had been deployed to the south, including the regular army and
about 10,000 paramilitary rangers (ICG, 2010, 3).22 Units are rotated out of the south on an annual
basis, so the share of soldiers with past, current, or future exposure to the southern conflict is even
higher than these figures imply. The RTA deployments include both volunteers and conscripts. In
Thailand, males ages 21 to 30 are required to participate in the annual draft lottery for two year
terms in the army. Gonwong et al. (2014) explain the scope:
The RTA uses a lottery system to select ≈ 60, 000 young Thai men at the district
level of their family residence for enlistment annually. The men enlisted comprise
approximately 10% of young men at the district level in Thailand (1531).
22 McCargo
(2008, 102) estimates that 60% of soldiers in the south were RTA conscripts circa 2008. Rangers are
often ex-conscripts who have finished their tour of duty. Other branches of the military played minimal roles in the
south.
21
Volunteering for the army can result in a shorter stint. Students may delay the draft for up to
five years. Relatively affluent Thais are frequently able to avoid the draft through education or
corruption.
Based in part on the census region where they lived (see Figure 1), recruits and conscripts are
enrolled in one of the four commands of the RTA. The first army draws from the central region, the
second from the northeast, the third from the north, and the fourth from the south, including the
Pattani-majority provinces (Pimonpan, 2009). The correlation between recruits’ area of origin and
the regional army they join is not perfect but available data indicates it is very high, especially for
conscripts.23 Also, there is minimal migration between census regions in Thailand (ICG, 2010),
reinforcing the regional character of each army.24
The geography of deployment, 2007–2010
A 2007 redeployment of the four armies in the south makes it possible to match military fatalities
in the southern conflict to a regional army (see Table 3).25 Initially, the 4th army took the lead in
southern counterinsurgency. In 2007, General Anupong took over command of southern military
operations and reassigned the four RTA commands. The first army was put in charge of Narathiwat,
the second in Pattani, and the third in Yala. The fourth army was given responsibility for five
districts in Songkhla province, marginalizing that command in future counterinsurgency, “a mark
23 McQueen et al. (1996) reported that 100% of the conscripts at the 3rd Army base in Phitsanuloke at the time of
their study were from the northern region. However, in a study of the 1st Army’s conscripts, Hatairat (2010) reported
that the central region was overrepresented but so was the northeastern region (22–23).
24 A 2008 government survey found less than 3% of Thais live outside the census region where they were born
(National Statistical Office, 2009, Table 1).
25 Military fatalities data are from the Global Terrorism Database (GTD), which is based on internationally-available
news sources (National Consortium for the Study of Terrorism and Responses to Terrorism, 2013). The GTD categorizes government fatalities in terms of military, police, and civilian government officials. Reports from Deep South
Watch, an NGO in southern Thailand, imply GTD understates the scope of the conflict. For example, the Deep South
Watch data record 3,471 total deaths between January 2004 and May 2009 (Srisompob, 2009). In the same period,
GTD records 844 deaths. Given the GTD’s sources, it is probably undercounting fatalities. However, missed fatalities
are unlikely to be military personnel.
22
Table 3: Military fatalities and deployment of the Royal Thai Army by southern province
Province
Narathiwat
Pattani
Songkhla
Yala
Military
fatalities,
1/04–9/07
RTA deployment
after 10/07
Military
Fatalities,
10/07–9/10†
16
9
0
4
1st (Central)
2nd (Northeast)
4th (South)
3rd (North)
26
8
0
1
†TAF survey fielded in September, 2010.
of Anupong’s displeasure at its ineffectiveness in tackling the violence” (ICG, 2008, 6).26 As
Table 3 shows, the number of military fatalities is southern Thailand between 2007 and 2010 was,
by some standards, modest. The objective risk to a non-southern respondent or the repondent’s
family was minimal. However, studies of overseas war suggest that small numbers of casualties
can have dramatic impacts on public opinion. For example, Thai military fatalities per capita in
the south from 2007–2010 were an order of magnitude larger than US deaths per capita during
the 1993 Battle of Mogadishu (a.k.a., “Black Hawk Down”), which inspired a debate on public
casualty aversion.
With these pieces of information, I can ask whether the number of military fatalities associated
with a respondent’s regional army is correlated with support for southern autonomy. However,
a difficulty in comparing respondents across regional army areas is that there are also profound
regional divides in Thai politics. To address this problem, I focus on six provinces located near
the intersection of the 1st, 2nd, and 3rd army regions (Figure 3).27 The six provinces lie in three
different army regions but are otherwise similarly situated. Importantly, the census/army regions
have no additional administrative role.28 Other national ministries use alternative regional defini26 At
the same time, Anupong announced a plan to recruit a new infantry division that would be drawn heavily from
the three southernmost provinces. However, progress on recruiting and deploying that division was slow. At the end
of 2010, soldiers in the south were overwhelmingly from elsewhere in Thailand (ICG, 2010, fn. 12).
27 None of the TAF survey provinces were close to the border between the central and southern regions (Figure A.1).
28 Lop Buri and Nakhon Ratchasima are potential outliers because they have important RTA bases; these provinces
can be dropped without overturning the results below (Table A.16).
23
tions. Local government bodies are at the provincial level or below. The army regions are also
not equivalent to political jurisdictions. In the 2007 elections (the last before the TAF survey),
parliament was selected through (1) single member districts (SMD) nested within provinces and
(2) seats rewarded proportionally based on election zones comprised of several provinces.29 The
six provinces in Figure 3 spanned three different election zones in 2007 elections. Provinces that
shared an army region did not necessarily share an election zone and vice versa. The table accompanying Figure 3 also shows that in all six provinces the 2007 single-member races were mostly
won by the pro-Thaksin party—the PPP—or by one of the parties that joined the PPP government
after the elections. In other words, the core party of opposition to Thaksin, the Democrat Party,
was weak in all of these provinces.
A further objection to using fatalities from a respondent’s regional army command to predict
views on southern autonomy is that RTA assignments in 2007 may have been based on public
opinion. Perhaps armies from parts of the country relatively sympathetic to southern grievances
were sent to more dangerous districts in an effort to win hearts and minds. If so, there might be a
spurious positive relationship between more military fatalities and more support for southern autonomy. However, the 2007 assignment strategy seems to have accomplished the opposite. The
1st army, from central Thailand, was assigned to Narathiwat, the province that incurred the most
military casualties both before and after the 2007 redeployment (Table 3). Central Thailand, compared to north and northeast Thailand, was more friendly to the military and the Democrat Party.
A reasonable expectation would be that the public in this region was less inclined to contradict the
sitting Democrat central government’s policy, ceterus paribus. Also, according to the TAF survey
data, decentralization for any part of Thailand was less popular in the central zone (roughly 60%
of respondents supporting) than in the north (72%) and northeast (70%).
29 Election
data from Carr (2016).
24
Figure 3: Survey districts on northern, northeastern, and central army region borders
N
Khon Kaen
NE
Petchabun
Nakhon
Ratchasima
Lop Buri
Sing Buri
C
Prachin Buri
Province
Petchabun
Khon Kaen
Nakhon Ratchasima
Lop Buri
Prachin Buri
Sing Buri
Army region
Election
zone, 2007
Province SMD
seats won by
PPP, 2007
Province SMD
seats in PPP
coalition, 2007
3rd (N)
2nd (NE)
2nd (NE)
1st (C)
1st (C)
1st (C)
2
2
5
2
5
7
4 of 6 (66%)
11 of 11 (100%)
6 of 16 (38%)
3 of 5 (60%)
2 of 3 (66%)
0 of 1 (0%)
5 of 6 (83%)
11 of 11 (100%)
16 of 16 (75%)
4 of 5 (80%)
3 of 3 (100%)
1 of 1 (100%)
Regional army fatalities and support for southern autonomy
Table 4 reports logistic regressions comparing military fatalities in a respondent’s army region to
support for southern autonomy.30 Model 9 is the bivariate regression of logged military fatalities
and endorsement of southern autonomy. Military fatalities are positively signed and statistically
significant. The substantive interpretation of this result appears in Figure 4. Predicted support for
southern autonomy is plotted against unlogged military fatalities ranging from one to 26, the span
of the data. The predicted level of support quadruples over this range: from 20% in case of one
military fatality to 80% at the maximum.
The coefficient on logged fatalities remains statistically significant and similar in magnitude
as basic demographic control variables are added (Model 10). Note that there were no Pattani
Malay respondents in these six provinces; that control variable is omitted. The positive relationship
between army region fatalities and support for southern autonomy also persists after controlling
for attitudes about countrywide decentralization, the military, and the sitting government (Model
11).
Finally, I repeat the placebo test performed above, comparing the fatalities from a respondent’s
army area to support for dropping charges against Bangkok protest leaders. The correlations reported in Models 12–14 (Table 4) are not statistically distiguishable from null effects and change
signs between models. Again, this non-result implies that southern conflict costs are specifically
relevant to understanding attitudes about southern autonomy.
In sum, respondents living at the border of three recruitment regions for the RTA were more
likely to support southern autonomy if their region’s army was posted in a deadlier part of the
south. These results need to be interpreted conservatively because the military casualty data is very
aggregated and may proxy for differences in regional politics, despite the steps taken to address
that problem. On the other hand, these results can be combined with the earlier analysis in which
30 Unweighted
results are reported in Table A.13. Table A.14 reestimates the models in Table 4 without survey
weights but with standard errors clustered by province.
26
Table 4: Logistic regressions of respondent support for southern autonomy and for dropping
charges against central protestors compared to military fatalities in southern conflict associated
with the respondent’s army region
Southern autonomy
Central political reconciliation
Model 9
Model 10
Model 11
Model 12
Model 13
Model 14
0.84∗
(0.17)
0.85∗
(0.18)
0.82∗
(0.20)
-0.064
(0.14)
-0.031
(0.14)
0.039
(0.16)
Ln regional army
fatalities
Favors
decentralization
1.6∗
(0.32)
-0.62∗
(0.31)
Military role too
large
-0.47
(0.35)
0.45
(0.31)
Satisfaction w
national government
-0.26
(0.28)
-0.14
(0.23)
Age
-0.032
(0.074)
-0.038
(0.075)
0.087
(0.066)
0.054
(0.066)
Female
0.14
(0.30)
0.030
(0.33)
-0.41
(0.27)
-0.34
(0.28)
Secondary education
-0.13
(0.40)
-0.38
(0.42)
-0.21
(0.34)
-0.18
(0.36)
Post-secondary
education
-1.2∗
(0.57)
-1.5∗
(0.64)
-0.080
(0.52)
0.048
(0.54)
Income
0.13
(0.092)
0.12
(0.092)
-0.098
(0.092)
-0.13
(0.093)
Rural
-0.15
(0.26)
-0.11
(0.29)
0.30
(0.26)
0.28
(0.27)
6
317
6
299
6
317
6
299
Sampling strata
Sample size
Standard errors in parentheses.
6
317
∗
p < 0.05
6
317
Predicted probability of
support for southern autonomy
Figure 4: Predicted probability of support for southern autonomy and military fatalities from a
respondent’s army region. Based on Model 9 (Table 4) and plotted with 95% confidence interval.
.8
.6
.4
.2
0
0
5
10
15
20
Military fatalities from respondent’s army region
25
respondents gave their own assessment of the severity of southern conflict. Taken together, the
evidence is consistent with conflict costs eroding public support for a purely coercive response to
southern insurgency.
Conclusion
In 2010, the Thai public was asked about the desirability of using regional autonomy to end conflict
in the south of that country. My analysis of this polling data concerns how war costs influenced
the respondents’ view of southern autonomy. Perceived severity of the southern conflict predicts
greater support for southern autonomy but not for reconciliatory policies in central conflict. Respondents living in regions associated with more military casualties in the south were also more
likely to endorse southern autonomy. These findings suggest a casualty fatigue model of public
opinion similar to that reported in studies of overseas war rather than the hardening of antagonisms stressed in much of the civil war literature. A relationship between perceived war costs and
willingness to support policy compromises does not rule out other forms of polarization, such as
increasing social intolerance. However, in this case, any such animosities did not translate directly
into increased public appetite for war or unwillingness to compromise in order to end war.
It is also notable that the casualty fatigue finding is limited to survey respondents who lived
out-of-theater. The in- and out-of-theater distinction in the data is likely to be important in most
civil conflicts, which involve geographically isolated fighting and policy demands that have limited direct implications for many citizens. Earlier studies of public opinion and civil war have
considered cases that are not representative in these respects.
This study takes a first step toward broadening our base of knowledge about public opinion
and domestic war. The next step for future research will be to further specify contexts in which
war costs increase public support for compromising with insurgents’ aims. If a casualty fatigue
model is borne out in other data, that finding would raise a number of questions about how public
29
opinion shapes domestic counterinsurgency, for example whether anticipation of an adverse public
reaction to war costs influences central military tactics or prompts political compromise.
30
References
Abuza, Zachary. 2011. The Ongoing Insurgency in Southern Thailand: Trends in Violence, Counterinsurgency Operations, and the Impact of National Politics. Institute for National Strategic
Studies Strategic Perspectives, No. 6, Washington, DC: National Defense University Press.
Albritton, Robert B. 2006. “Thailand in 2004: The “Crisis in the South”.” Asian Survey 45 (1):
166–173.
Althaus, Scott L., Brittany H. Bramlett, and James G. Gimpel. 2012. “When War Hits Home: The
Geography of Military Losses and Support for War in Time and Space.” Journal of Conflict
Resolution 56 (3): 382–412.
Asia Foundation. 2011. The 2010 National Survey of the Thai Electorate. Bangkok: The Asia
Foundation.
Askew, Marc and Sascha Helbardt. 2012. “Becoming Pattani warriors: Individuals and the Insurgent Collective Action in Southern Thailand.” Studies in Conflict & Terrorism 35 (11): 779–809.
Bakke, Kristin M., John O’Loughlin, and Michael D. Ward. 2009. “Reconciliation in ConflictAffected Societies: Multilevel Modeling of Individual and Contextual Factors in the North Caucasus of Russia.” Annals of the Association of American Geographers 99 (5): 1012–1021.
Baum, Matthew A. 2002. “The Constituent Foundations of the Rally Round the Flag Phenomenon.” International Studies Quarterly 46 (2): 263–298.
Baum, Matthew A. and Tim J. Groeling. 2010. War Stories: The Causes and Consequences of
Public Views of War. Princeton, NJ: Princeton University Press.
Bellows, John and Edward Miguel. 2006. “War and Institutions: New Evidence from Sierra
Leone.” American Economic Review 96 (2): 394–399.
Bellows, John and Edward Miguel. 2009. “War and Local Collective Action in Sierra Leone.”
Journal of Public Economics 93 (11–12): 1144–1157.
Berinsky, Adam J. 2007. “Assuming the Costs of War: Events, Elites and American Public Support
for Military Conflict.” Journal of Politics 69 (4): 957–997.
Berrebi, Claude and Esteban F. Klor. 2008. “Are Voters Sensitive to Terrorism? Direct Evidence
from the Israeli Electorate.” American Political Science Review 102 (03): 279–301.
Blair, Graeme, C. Christine Fair, Neil Malhotra, and Jacob N. Shapiro. 2013. “Poverty and Support
for Militant Politics: Evidence from Pakistan.” American Journal of Political Science 57 (1):
30–48.
Blattman, Christopher. 2009. “From Violence to Voting: War and Political Participation in
Uganda.” 103 (2): 231–247.
Boettcher, William A. and Michael D. Cobb. 2006. “Echoes of Vietnam?: Casualty Framing and
Public Perceptions of Success and Failure in Iraq.” Journal of Conflict Resolution 50 (6): 831–
854.
Canetti-Nisim, Daphna, Gal Ariely, and Eran Halperin. 2008. “Life, Pocketbook, or Culture: The
Role of Perceived Security Threats in Promoting Exclusionist Political Attitudes toward Minorities in Israel.” Political Research Quarterly 61 (1): 90–103.
Canetti-Nisim, Daphna, Eran Halperin, Keren Sharvit, and Stevan E. Hobfoll. 2009. “A New
Stress-Based Model of Political Extremism: Personal Exposure to Terrorism, Psychological
31
Distress, and Exclusionist Political Attitudes.” Journal of Conflict Resolution 53 (3): 363–389.
Carr, Adam. 2016. Psephos: Adam Carr’s Election Archive. Melbourne: Adam Carr, accessed
January 15, 2016 at http://psephos.adam-carr.net/countries/t/thailand/.
Carson, Jamie L., Jeffery A. Jenkins, David W. Rohde, and Mark A. Souva. 2001. “The Impact
of National Tides and District-Level Effects on Electoral Outcomes: The U.S. Congressional
Elections of 1862–63.” American Journal of Political Science 45 (4): 887–898.
Collier, Paul, V. L. Elliott, Håvard Hegre, Anke Hoeffler, Marta Reynal-Querol, and Nicholas
Sambanis. 2003. Breaking the Conflict Trap: Civil War and Development Policy. Washington,
DC: The World Bank.
Dalpino, Catharin. 2011. “Thailand in 2010.” Asian Survey 51 (1): 155–162.
Davenport, Tiffany C. 2015. “Policy-Induced Risk and Responsive Participation: The Effect of a
Son’s Conscription Risk on the Voting Behavior of His Parents.” American Journal of Political
Science 59 (1): 225–241.
De Juan, Alexander and Jan Henryk Pierskalla. 2016. “Civil War Violence and Political Trust:
Microlevel Evidence from Nepal.” Conflict Management and Peace Science 33 (1): 67–88.
Dittrich Hallberg, Johan. 2012. “PRIO Conflict Site 1989–2008: A Geo-Referenced Dataset on
Armed Conflict.” Conflict Management and Peace Science 29 (2): 219–232.
Eichenberg, Richard C. 2005. “Victory Has Many Friends: U.S. Public Opinion and the Use of
Military Force, 1981–2005.” International Security 30 (1): 140–177.
Funston, John. 2008. Southern Thailand: The Dynamics of Conflict. Washington, DC: East-West
Center.
GADM. 2012. GADM Database of Global Administrative Areas, v 2.0. Davis, CA: University of
California, Davis.
Gaines, Brian J., James H. Kuklinski, Paul J. Quirk, Buddy Peyton, and Jay Verkuilen. 2007.
“Same Facts, Different Interpretations: Partisan Motivation and Opinion on Iraq.” Journal of
Politics 69 (4): 95774.
Gartner, Scott S, Gary M. Segura, and Michael Wilkening. 1997. “All Politics are Local: Local
Losses and Individual Attitudes toward the Vietnam War.” Journal of Conflict Resolution 41 (5):
669–694.
Gartner, Scott Sigmund. 1997. Strategic Assessment in Wartime. New Haven, CT: Yale University
Press.
Gartner, Scott Sigmund. 2008a. “The Multiple Effects of Casualties on Public Support for War.”
American Political Science Review 102 (1): 95–106.
Gartner, Scott Sigmund. 2008b. “Ties to the Dead: Connections to Iraq War and 9/11 Casualties
and Disapproval of the President.” American Sociological Review 73 (4): 690–695.
Gartner, Scott Sigmund and Gary M. Segura. 1998. “War, Casualties, and Public Opinion.” Journal
of Conflict Resolution 42 (3): 278–300.
Gartner, Scott Sigmund and Gary M. Segura. 2000. “Race, Casualties, and Opinion in the Vietnam
War.” Journal of Politics 62 (1): 115–146.
Gartner, Scott Sigmund and Gary M. Segura. 2008. “All Politics are Still Local: The Iraq War and
the 2006 Midterm Elections.” PS: Political Science & Politics 41 (1): 95–100.
Gartner, Scott Sigmund, Gary M. Segura, and Bethany Barratt. 2004. “War Casualties, Policy
32
Positions, and the Fate of Legislators.” Political Research Quarterly 53 (3): 46777.
Gelpi, Christopher, Peter D. Feaver, and Jason Reifler. 2005. “Success matters.” International Security 30 (3): 7–46.
Gelpi, Christopher, Peter D. Feaver, and Jason Reifler. 2009. Paying the Human Costs of War:
American Public Opinion and Casualties in Military Conflicts. Princeton, NJ: Princeton University Press.
Gonwong, Siriphan, Thippawan Chuenchitra, Patchariya Khantapura, Dilara Islam, Narongrid
Sirisopana, and Carl J. Mason. 2014. “Pork Consumption and Seroprevalence of Hepatitis E
Virus, Thailand, 2007–2008.” Emerging Infectious Diseases 20 (9): 1531–1534.
Grose, Christian R. and Bruce I. Oppenheimer. 2007. “The Iraq War, Partisanship, and Candidate
Attributes: Variation in Partisan Swing in the 2006 U.S. House Elections.” Legislative Studies
Quarterly 32 (4): 531–557.
Gunaratna, Rohan and Arabinda Acharya. 2013. The Terrorist Threat from Thailand: Jihad or
Quest for Justice? Washington, DC: Potomac Books.
Haemindra, Nantawan. 1976. “The Problem of the Thai-Muslims in the Four Southern Provinces
of Thailand (Part One).” Journal of Southeast Asian Studies 7 (2): 197–225.
Hatairat, Kaoaiem. 2010. The Effect of Squad Leader Mentors through Short Message Services on
Mobile Phone in Promoting Safe Sex among First (Central) Army Area Conscripts of Thailand.
Bangkok: Doctoral dissertation, College of Public Health Sciences, Chulalongkorn University.
Hayes, Andrew F. and Teresa A. Myers. 2009. “Testing the “Proximate Casualties Hypothesis”:
Local Troop Loss, Attention to News, and Support for Military Intervention.” Mass Communication and Society 12 (4): 379–402.
Hutchison, Marc L. 2014. “Tolerating Threat? The Independent Effects of Civil Conflict on Domestic Political Tolerance.” Journal of Conflict Resolution 58 (5): 796–824.
ICG. 2008. Thailand: Political Turmoil and the Southern Insurgency. Asia Briefing, No. 80,
Bangkok/Brussels: International Crisis Group.
ICG. 2010. Stalemate in Southern Thailand. Asia Briefing, No. 113, Bangkok/Brussels: International Crisis Group.
Jakobsen, Peter Viggo and Jens Ringsmose. 2015. “In Denmark, Afghanistan is Worth Dying For:
How Public Support for the War Was Maintained in the Face of Mounting Casualties and Elusive
Success.” Cooperation and Conflict 50 (2): 211–227.
Johns, Robert and Graeme A. M. Davies. 2014. “Coalitions of the Willing? International Backing
and British Public Support for Military Action.” Journal of Peace Research 51 (6): 767–781.
Johnson, Dominic D. P. and Dominic Tierney. 2006. Failing to Win: Perceptions of Victory and
Defeat in International Politics. Cambridge, MA: Harvard University Press.
Karol, David and Edward Miguel. 2007. “The Electoral Cost of War: Iraq Casualties and the 2004
U.S. Presidential Election.” The Journal of Politics 69 (3): 633–648.
Kilcullen, David. 2009. The Accidental Guerrilla. New York: Oxford University Press.
Koch, Michael T. and Stephen P. Nicholson. 2015. “Death and Turnout: The Human Costs of War
and Voter Participation in Democracies.” American Journal of Political Science forthcoming.
Kriner, Douglas and Francis Shen. 2014. “Responding to War on Capitol Hill: Battlefield Casualties, Congressional Response, and Public Support for the War in Iraq.” American Journal of
33
Political Science 58 (1): 157–174.
Kriner, Douglas L. and Francis A. Shen. 2007. “Iraq Casualties and the 2006 Senate Elections.”
Legislative Studies Quarterly 32 (4): 507–530.
Kull, Steven, I.M. Destler, and Clay Ramsay. 1997. The Foreign Policy Gap: How Policymakers
Misread the Public. Washington, D.C.: Center for Strategic and International Studies.
Leites, Nathan and Charles Wolf, Jr. 1970. Rebellion and Authority: An Analytic Essay on Insurgent Conflicts. Santa Monica, CA: RAND.
Liow, Joseph Chinyong and Don Pathan. 2010. “Confronting Ghosts: Thailand’s Shapeless Southern Insurgency.” Lowy Institute for International Policy 30.
Longo, Matthew, Daphna Canetti, and Nancy Hite-Rubin. 2014. “A Checkpoint Effect? Evidence
from a Natural Experiment on Travel Restrictions in the West Bank.” American Journal of Political Science 58 (4): 1006–1023.
Lyall, Jason, Graeme Blair, and Kosuke Imai. 2013. “Support for Combatants during Wartime: A
Survey Experiment in Afghanistan.” American Political Science Review 107 (4): 679–705.
Mahakanjana, Chandra-nuj. 2006. “Decentralization, Local Government, and Socio-political Conflict in Southern Thailand.” East-West Center Working Papers 2.
Mahmood, Cynthia Keppley. 1996. Fighting for Faith and Nation: Dialogues with Sikh Militants.
Philadelphia, PA: University of Pennsylvania Press.
Maoz, Ifat and Clark McCauley. 2009. “Threat Perceptions and Feelings as Predictors of JewishIsraeli Support for Compromise with Palestinians.” Journal of Peace Research 46 (4): 525–539.
Mason, T. David. 1989. “Nonelite Response to State-Sanctioned Terror.” The Western Political
Quarterly 42 (4): 467–492.
Mason, T. David. 2004. Caught in the Crossfire: Revolution, Repression, and the Rational Peasant.
Lanham, MD: Rowman & Littlefield.
McCargo, Duncan. 2008. Tearing the Land Apart: Islam and Legitimacy in Southern Thailand.
Ithaca, NY: Cornell University Press.
McCargo, Duncan. 2012. Mapping National Anxieties: Thailand’s Southern Conflict. Copenhagen:
Nordic Institute of Asian Studies.
McQueen, Kathleen M., Taweesak Nopkesorn, Michael D. Sweat, Yothin Sawaengdee, Timothy D.
Mastro, and Bruce G. Weniger. 1996. “Alcohol Consumption, Brothel Attendance, and Condom
Use: Normative Expectations among Thai Military Conscripts.” Medical Anthropology Quarterly 10 (3): 402–423.
Meisburger, Tim. 2009. Constitutional Reform and Democracy in Thailand: A National Survey of
the Thai People. Bangkok: The Asia Foundation.
Mueller, John E. 1973. War, Presidents and Public Opinion. New York: Wiley.
Murray, William S. 2001. “A Will to Measure.” Parameters XXXI (3): 134–47.
Myers, Teresa A. and Andrew F. Hayes. 2010. “Reframing the Casualties Hypothesis:
(Mis)Perceptions of Troop Loss and Public Opinion about War.” International Journal of Public
Opinion Research 22 (2): 256–275.
National Consortium for the Study of Terrorism and Responses to Terrorism. 2013. Global Terrorism Database [Data file]. College Park, MD: University of Maryland, Retrieved from
http://www.start.umd.edu/gtd.
34
National Statistical Office. 2009. The Migration Survey Report 2008. Bangkok:
National
Statistical
Office,
Thailand,
accessed
November
6,
2015
at
http://service.nso.go.th/nso/nsopublish/service/migration/migrantRep51.pdf.
National Statistical Office. 2010. The 2010 Population and Housing Census.
Bangkok:
National Statistical Office, Thailand, accessed November 6, 2015 at
http://web.nso.go.th/en/census/poph/cen poph 10.htm.
Pevehouse, Jon C. and William G. Howell. 2007. While Dangers Gather: Congressional Checks
on Presidential War Powers. Princeton, NJ: Princeton University Press.
Pimonpan, Ukoskit. 2009. “The Internal Culture of Military Units and its Impact on the Conflict
Resolution in Thailand’s Far South.” In Imagined Land? The State and Southern Violence in
Thailand, ed. Satha-Anand Chaiwat, 93–120, Tokyo: Research Institute for Languages and Cultures of Asia and Africa.
Popkin, Samuel L. 1979. The Rational Peasant: The Political Economy of Rural Society in Vietnam. Berkeley, CA: University of California Press.
Prasirtsuk, Kitti. 2009. “Thailand in 2008: Crises Continued.” Asian Survey 49 (1): 174–184.
Sambanis, Nicholas and Moses Shayo. 2013. “Social Identification and Ethnic Conflict.” American
Political Science Review 107: 294–325.
Shayo, Moses and Asaf Zussman. 2011. “Judicial Ingroup Bias in the Shadow of Terrorism.” The
Quarterly Journal of Economics 126 (3): 1447–1484.
Srisompob, Jitpiromsri. 2009. Five Years, Five Months: Analysis of the 65 Months of Unrest
in the Southern Border Provinces. Pattani: Deep South Watch, accessed January 15, 2016 at
http://www.deepsouthwatch.org/node/369.
Tir, Jaroslav and Shane P. Singh. 2015. “Get Off My Lawn: Territorial Civil Wars and Subsequent
Social Intolerance in the Public.” Journal of Peace Research 52 (4): 478–491.
US Armed Forces. 2006. Counterinsurgency. Army Field Manual 3-24. Marine Corps Warfighting
Publication 3-33.5. Washington, DC: Department of the Army.
Voors, Maarten J., Eleonora E. M. Nillesen, Philip Verwimp, Erwin H. Bulte, Robert Lensink,
and Daan P. Van Soest. 2012. “Violent Conflict and Behavior: A Field Experiment in Burundi.”
American Economic Review 102 (2): 941–64.
Wood, Elisabeth Jean. 2003. Insurgent Collective Action and Civil War in El Salvador. New York,
NY: Cambridge University Press.
Supplementary materials for “War costs and public
support for domestic counterinsurgency: Evidence from
Thailand”
Intended for online publication only
Bethany Lacina
Assistant Professor,
Department of Political Science,
University of Rochester
PO Box 270146
Rochester, NY 14627-0146
585-273-5842 (phone)
585-271-1616 (fax)
[email protected]
Spring 2016
1
Figure A.1: Provinces in Asia Foundation Survey, Fall 2010
3
15
11
6
21
5
16
4
29
23
7
8
2
27
25
18
13
12
1
22
19
20
26
9
17
28
24
14
30 10
1 - Bangkok
2 - Buri Ram
3 - Chiang Rai
4 - Kamphaeng Phet
5 - Khon Kaen
6 - Lamphun
7 - Lop Buri
8 - Nakhon Ratchasima
9 - Nakhon Si Thammarat
10 - Narathiwat
19 - Ratchaburi
11 - Nong Khai
20 - Rayong
12 - Nonthaburi
21 - Sakon Nakhon
13 - Pathum Thani 22 - Samut Prakan
14 - Pattani
23 - Sing Buri
15 - Phayao
24 - Songkhla
16 - Petchabun
25 - Suphan Buri
17 - Phuket
26 - Surat Thani
18 - Prachin Buri
27 - Surin
28 - Trang
29 - Uthai Thani
30 - Yala
Survey items for variables in the main text and supplementary materials
The following are TAF-provided translations of their survey, which was conducted in Thai.
• Support for southern autonomy (0/1): “Some have suggested that political decentralization
or limited autonomy (but not territorial separation) might help resolve the long-term conflict
in southern Thailand. Do you agree or disagree?”
• Support central political reconciliation (0/1): Respondents answered two items related
to charges against the main political factions in protests at the capital: “And for the leader
of red shirts [who] has been accused for terrorism, do you think the lawsuit and charges
should be dropped to promote the political reconciliation?” and “And for the leader of yellow shirts [who] has been accused for terrorism, do you think the lawsuit and charges should
be dropped to promote the political reconciliation?” Respondents who supported dropping
charges against both yellow and red shirt leaders were coded as supporting central reconciliation. Respondents who said the charges should stand against the yellow shirt leaders, the
red shirt leaders, or both sets of leaders were coded as a zero.
• High perceived conflict costs (0/1): Respondents were asked “In your view, what is the
biggest problem facing Thailand?” I code high perceived costs for respondents who described conflict in southern Thailand as the country’s biggest problem.
• High perceived conflict costs II (0/1): Respondents were also asked “Generally speaking,
do you think things in Thailand today are going in the right direction, or do you think they
are going in the wrong direction?” and an open-ended follow-up question “Why do you say
that?” Enumerators listed up to six answers to the open-ended follow-up question. I coded
High perceived conflict costs II as a 1 if a respondent who thought Thailand was going in
the wrong direction listed conflict in the south as one of the reasons for that opinion. This
variable is coded as a zero if the respondent did not include the southern conflict in their list
of reasons for pessimism. This variable is coded only for the 866 (61%) of respondents who
thought Thailand was going in the wrong direction.
• Pattani Malay (0/1): “Normally, what language do you speak at home?” Respondents who
reported that they spoke Pattani Malay or Yawa are coded as Pattani Malay.
• Favors decentralization (0/1): “Sometimes people in Bangkok and upcountries have different interests and different points of view; therefore, some people say that the government
could work more effectively and efficiently if the power of decision making were decentralized to local governments. In addition, the governor should be directly elected (like in
Bangkok). However, some people believe that the government will be more fair, more effective and more efficient if centrally controlled. Which statement is closer to your point of
view? (1) Better to decentralize to local government. Or (2) The central government having
the authority for decision making, as at present, would be more effective and efficient.”
If the respondent was uncertain of the meaning of decentralization, the following text was
also read: “Decentralization means to decentralize power to local areas under democratic
3
governance. It’s aimed to reduce central government’s role so that they can focus on key
missions as necessary as possible [sic] and to enhance the role of local government to take
over duties. The law is required to allow local government to have authority in managing
works, money, and labor to solve its problems and to develop its own areas; it means some
local laws can follow local interests, custom and tradition. This is based on the fundamental
in which local people best know their own local problems and interests. The familiar forms
of local government are Sub-District Administration Office (SAO), municipal, Provincial
Administration Office (PAO), and such special districts as Bangkok and Pattaya.”
Respondents who said it was “better to decentralize to local government” (choice 1) were
coded as favoring decentralization.
• Military role too large (0/1): “Many people think the army plays too big a role in politics in
Thailand, while others see the army as an important independent institution that has helped
safeguard and stabilize the country. Which is closer to your view?”
• Satisfaction with national government (1–4): “How do you feel regarding the job the
national government is doing?” Very dissatisfied (1); Somewhat dissatisfied (2); Somewhat
satisfied (3); Very satisfied (4).
• Age (2–11): “May I know your age please?” Less than 18 (survey terminated); 18–19 years
old (2); 20–24 years old (3); 25–29 years old (4); 30–34 years old (5); 35–39 years old (6);
40–44 years old (7); 45–49 years old (8); 50–59 years old (9); 60–80 years old (10); Above
80 years old (11).
• Female (0/1): Coded by enumerator.
• Secondary education (0/1): Based on responses to the item “Could you please tell me your
highest education?” Respondents who chose “Secondary school” or “Diploma/Vocational”
school were coded as a 1. Respondents who chose “Primary school or below,” “Bachelors
degree,” “Masters degree,” or “Doctoral degree” were coded as a zero.
• Post-secondary education (0/1): Based on responses to the item “Could you please tell me
your highest education?” Respondents who chose “Bachelors degree,” “Masters degree,” or
“Doctoral degree” were coded as a 1. Respondents who chose “Primary school or below,”
“Secondary school,” or “Diploma/Vocational” school were coded as a zero.
• Income (1–9): Respondents were asked “Could you please tell me your monthly household
income which mean[s] the income of all household members (including all sources of income)?” TAF recorded household income in three separate income scales: one for Bangkok,
another for other urban respondents, and a third for rural respondents. My variable notes
each respondents’ score on their local scale; thus, this is an income measure relative to the
respondent’s locality, which may help to capture differences in purchasing power.
4
Income in 1000s of Baht†
Coding
Bangkok
Other urban
Rural
1
2
3
4
5
6
7
8
9
<15
15–24.9
25–49.9
50–59.9
60–69.9
70–79.9
80–89.9
90–99.9
≥100
<8
8–14.9
15–34.9
35–44.9
45–49.9
5–59.9
60–69.9
70–79.9
≥80
<5
5–7.9
8–19.9
20–22.49
22.5–29.9
30–39.9
40–49.9
50–59.9
≥60
†In mid-2010, 1000 Baht was worth about $31 US
• Rural (0/1): Respondents lived in either a rural or an urban district of a province. That
distinction was determined by the survey firm during sampling. All respondents in Bangkok
province are in urban districts.
5
Summary statistics and robustness tests
Table A.1: Summary statistics for national sample used in Tables 1, 2, A.8, A.11.
Weighted data
Mean
Support for southern autonomy (0/1)
Support central political reconciliation (0/1)
High perceived conflict costs (0/1)
Pattani Malay (0/1)
Favors decentralization (0/1)
Military role too large (0/1)
Satisfaction w national government (1–4)
Age (2–11)
Female (0/1)
Secondary education (0/1)
Post-secondary education (0/1)
Income (1–9)
Rural (0/1)
Observations
0.63
0.49
0.022
0
0.70
0.43
2.4
6.1
0.53
0.53
0.15
2.7
0.65
Standard error
0.023
0.024
0.0061
0
0.023
0.024
0.030
0.11
0.024
0.024
0.017
0.073
0.013
Unweighted data
Mean
0.62
0.50
0.024
0
0.70
0.44
2.3
6.1
0.52
0.53
0.17
2.8
0.44
Standard error
0.019
0.020
0.0061
0
0.018
0.020
0.026
0.098
0.020
0.020
0.015
0.065
0.020
1419
Table A.2: Summary statistics including respondents who did not answer central political reconciliation questions. See Table A.9.
Weighted data
Mean
Support for southern autonomy (0/1)
High perceived conflict costs (0/1)
Pattani Malay (0/1)
Favors decentralization (0/1)
Military role too large (0/1)
Satisfaction w national government (1–4)
Age (2–11)
Female (0/1)
Secondary education (0/1)
Post-secondary education (0/1)
Income (1–9)
Rural (0/1)
Observations
0.63
0.022
0
0.70
0.43
2.4
6.1
0.53
0.53
0.15
2.7
0.65
1515
6
Standard error
0.023
0.0061
0
0.023
0.024
0.030
0.11
0.024
0.024
0.017
0.073
0.012
Unweighted data
Mean
0.62
0.024
0
0.70
0.44
2.3
6.1
0.52
0.53
0.17
2.8
0.44
Standard error
0.019
0.0061
0
0.018
0.020
0.026
0.098
0.020
0.020
0.015
0.065
0.020
Table A.3: Summary statistics including only respondents who answered questions on alternative
measure of perceived costs of the southern conflict and question on southern autonomy. See Table
A.10.
Weighted data
Mean
Support for southern autonomy (0/1)
High perceived conflict costs II (0/1)
Pattani Malay (0/1)
Favors decentralization (0/1)
Military role too large (0/1)
Satisfaction w national government (1–4)
Age (2–11)
Female (0/1)
Secondary education (0/1)
Post-secondary education (0/1)
Income (1–9)
Rural (0/1)
0.63
0.089
0
0.70
0.43
2.4
6.1
0.53
0.53
0.15
2.7
0.65
Observations
813
7
Standard error
0.023
0.016
0
0.023
0.024
0.030
0.11
0.024
0.024
0.017
0.073
0.017
Unweighted data
Mean
0.62
0.078
0
0.70
0.44
2.3
6.1
0.52
0.53
0.17
2.8
0.44
Standard error
0.019
0.011
0
0.018
0.020
0.026
0.098
0.020
0.020
0.015
0.065
0.020
Table A.4: Summary statistics including only respondents who answered questions on alternative
measure of perceived costs of the southern conflict and questions on central reconciliation. See
Table A.12.
Weighted data
Mean
Support central political reconciliation (0/1)
High perceived conflict costs II (0/1)
Pattani Malay (0/1)
Favors decentralization (0/1)
Military role too large (0/1)
Satisfaction w national government (1–4)
Age (2–11)
Female (0/1)
Secondary education (0/1)
Post-secondary education (0/1)
Income (1–9)
Rural (0/1)
0.49
0.086
0
0.69
0.43
2.4
6.1
0.52
0.52
0.15
2.7
0.65
Observations
813
8
Standard error
0.024
0.015
0
0.022
0.024
0.030
0.11
0.024
0.024
0.017
0.070
0.016
Unweighted data
Mean
0.50
0.077
0
0.70
0.44
2.3
6.1
0.52
0.53
0.17
2.8
0.44
Standard error
0.020
0.010
0
0.018
0.019
0.026
0.096
0.020
0.020
0.015
0.063
0.019
Table A.5: Summary statistics for sample from provinces on the boundaries of the 1st, 2nd, and
3rd Royal Thai Army command regions used in Tables 4, A.13, and A.14.
Weighted data
Mean
Support for southern autonomy (0/1)
Support central political reconciliation (0/1)
Ln regional army fatalities
Favors decentralization (0/1)
Military role too large (0/1)
Satisfaction w national government (1–4)
Age (2–11)
Female (0/1)
Secondary education (0/1)
Post-secondary education (0/1)
Income (1–9)
Rural (0/1)
0.65
0.51
2.5
0.73
0.41
2.4
6.1
0.57
0.57
0.13
2.7
0.66
Observations
317
9
Standard error
0.041
0.044
0.073
0.039
0.043
0.055
0.21
0.043
0.043
0.029
0.12
0.034
Unweighted data
Mean
0.59
0.52
2.5
0.69
0.42
2.4
6.1
0.53
0.55
0.13
2.8
0.43
Standard error
0.036
0.036
0.056
0.034
0.036
0.047
0.18
0.036
0.036
0.024
0.12
0.036
Table A.6: Summary statistics for sample from provinces near 1st–3rd RTA command area borders
including respondents who did not answer central political reconciliation questions. See Table
A.15.
Weighted data
Mean
Support for southern autonomy (0/1)
Ln regional army fatalities
Favors decentralization (0/1)
Military role too large (0/1)
Satisfaction w national government (1–4)
Age (2–11)
Female (0/1)
Secondary education (0/1)
Post-secondary education (0/1)
Income (1–9)
Rural (0/1)
0.65
2.5
0.73
0.41
2.4
6.1
0.57
0.57
0.13
2.7
0.66
Observations
342
10
Standard error
0.041
0.073
0.039
0.043
0.055
0.21
0.043
0.043
0.029
0.12
0.034
Unweighted data
Mean
0.59
2.5
0.69
0.42
2.4
6.1
0.53
0.55
0.13
2.8
0.43
Standard error
0.036
0.056
0.034
0.036
0.047
0.18
0.036
0.036
0.024
0.12
0.036
Table A.7: Summary statistics for sample from provinces near 1st–3rd RTA command area borders
without Lop Buri and Nakhon Ratchasima. See Table A.16.
Weighted data
Mean
Support for southern autonomy (0/1)
Support central political reconciliation (0/1)
High perceived conflict costs II (0/1)
Pattani Malay (0/1)
Favors decentralization (0/1)
Military role too large (0/1)
Satisfaction w national government (1–4)
Age (2–11)
Female (0/1)
Secondary education (0/1)
Post-secondary education (0/1)
Income (1–9)
Rural (0/1)
0.62
0.55
0.058
0
0.66
0.43
2.4
6.0
0.53
0.58
0.11
2.6
0.69
Observations
206
11
Standard error
0.054
0.055
0.027
0
0.052
0.055
0.073
0.28
0.055
0.055
0.036
0.16
0.041
Unweighted data
Mean
0.57
0.57
0.043
0
0.60
0.46
2.4
6.1
0.50
0.57
0.11
2.7
0.45
Standard error
0.046
0.046
0.019
0
0.046
0.047
0.062
0.23
0.047
0.046
0.030
0.14
0.047
Table A.8: Logistic regressions of respondent support for southern autonomy without using survey
weights or clustering by sample strata. Robustness test for Table 1.
Model 15
Model 16
Model 17
Model 18
Model 19
0.61∗
(0.30)
0.55
(0.30)
0.72∗
(0.32)
0.72∗
(0.36)
0.94∗
(0.38)
0.37
(0.27)
0.25
(0.29)
0.48
(0.30)
0.41
(0.32)
-0.72
(0.71)
-0.92
(0.74)
High perceived
conflict costs
Pattani Malay
Costs*Pattani Malay
Favors
decentralization
1.0∗∗
(0.12)
1.1∗∗
(0.12)
Military role too
large
-0.088
(0.13)
-0.092
(0.13)
Satisfaction w
national government
0.13
(0.090)
0.13
(0.090)
Age
0.010
(0.026)
0.0065
(0.027)
0.010
(0.026)
0.0063
(0.027)
Female
-0.020
(0.11)
-0.0020
(0.12)
-0.015
(0.11)
0.0051
(0.12)
Secondary education
0.23
(0.14)
0.28
(0.15)
0.23
(0.14)
0.28
(0.15)
Post-secondary
education
-0.21
(0.19)
-0.20
(0.20)
-0.20
(0.19)
-0.20
(0.20)
Income
0.0057
(0.036)
0.012
(0.039)
0.0055
(0.036)
0.012
(0.039)
Rural
0.31∗∗
(0.11)
0.22
(0.12)
0.31∗∗
(0.11)
0.22
(0.12)
1334
-840
1419
-936
1334
-839
Observations
1419
1419
Ln likelihood
-946
-936
Hypothesis tests (F-statistics from Wald tests):
Perceived costs + Costs*Pattani = 0
Standard errors in parentheses.
∗
p < 0.05
12
0.000026
0.00090
Table A.9: Reestimation of support for southern autonomy including respondents who did not
answer central political reconciliation questions. Robustness test for Table 1.
Model 20
Model 21
Model 22
Model 23
Model 24
0.88∗
(0.34)
0.81∗
(0.35)
1.0∗
(0.39)
1.1∗
(0.39)
1.3∗
(0.44)
0.48
(0.31)
0.32
(0.35)
0.68∗
(0.33)
0.54
(0.39)
-1.2
(0.84)
-1.2
(0.91)
High perceived
conflict costs
Pattani Malay
Costs*Pattani Malay
Favors
decentralization
0.93∗
(0.15)
0.93∗
(0.15)
Military role too
large
-0.035
(0.16)
-0.041
(0.16)
Satisfaction w
national government
0.13
(0.11)
0.12
(0.11)
Age
0.017
(0.031)
0.0048
(0.032)
0.017
(0.031)
0.0053
(0.033)
Female
0.047
(0.13)
0.042
(0.14)
0.053
(0.13)
0.050
(0.14)
Secondary education
0.30
(0.17)
0.29
(0.18)
0.30
(0.17)
0.30
(0.18)
Post-secondary
education
-0.26
(0.22)
-0.29
(0.24)
-0.26
(0.22)
-0.28
(0.24)
Income
0.021
(0.042)
0.024
(0.045)
0.021
(0.042)
0.023
(0.045)
Rural
0.25
(0.13)
0.087
(0.14)
0.26∗
(0.13)
0.092
(0.14)
10
1515
10
1396
Sampling strata
10
10
10
Sample size
1515
1515
1396
Hypothesis tests (F-statistics from Wald tests with weighted data):
Perceived costs + Costs*Pattani = 0
Standard errors in parentheses.
∗
p < 0.05
13
0.0067
0.043
Table A.10: Logistic regressions of alternative measure of perceived costs of the southern conflict
and respondent support for southern autonomy. Robustness test for Table 1.
Model 25
Model 26
Model 27
Model 28
Model 29
0.71∗
(0.33)
0.72∗
(0.35)
1.0∗
(0.40)
0.92∗
(0.38)
1.4∗
(0.42)
0.61
(0.46)
0.48
(0.54)
1.1∗
(0.55)
1.3∗
(0.58)
-1.7
(1.0)
-2.9∗
(1.1)
High perceived
conflict costs II
Pattani Malay
Costs*Pattani Malay
Favors
decentralization
1.1∗
(0.21)
1.2∗
(0.20)
Military role too
large
-0.040
(0.20)
-0.049
(0.20)
Satisfaction w
national government
0.20
(0.15)
0.17
(0.15)
Age
0.012
(0.042)
0.017
(0.044)
0.012
(0.042)
0.017
(0.044)
Female
-0.0054
(0.18)
-0.081
(0.19)
-0.021
(0.18)
-0.10
(0.19)
Secondary education
0.24
(0.23)
0.23
(0.24)
0.24
(0.23)
0.23
(0.25)
Post-secondary
education
-0.20
(0.29)
-0.18
(0.31)
-0.17
(0.29)
-0.14
(0.30)
Income
-0.050
(0.055)
-0.036
(0.060)
-0.051
(0.056)
-0.036
(0.060)
Rural
0.17
(0.18)
0.098
(0.19)
0.19
(0.18)
0.13
(0.19)
10
813
10
765
0.75
2.2
Sampling strata
10
10
10
Sample size
813
813
765
Hypothesis tests (F-statistics from Wald tests with weighted data):
Perceived costs + Costs*Pattani = 0
Standard errors in parentheses.
∗
p < 0.05
14
Table A.11: Logistic regressions of respondent support for pardons for competing central factions
in the national sample without using survey weights or clustering by sample strata. Robustness test
for Table 2.
Model 30
Model 31
Model 32
-0.48
(0.27)
-0.66∗
(0.28)
-0.67∗
(0.30)
0.41
(0.25)
0.47
(0.27)
High perceived
conflict costs
Pattani Malay
Favors
decentralization
0.021
(0.12)
Military role too
large
0.58∗
(0.13)
Satisfaction w
national government
-0.30∗
(0.087)
Age
-0.034
(0.025)
-0.048
(0.026)
Female
0.014
(0.11)
0.046
(0.11)
Secondary education
-0.33∗
(0.14)
-0.35∗
(0.15)
Post-secondary
education
-0.64∗
(0.19)
-0.62∗
(0.20)
Income
-0.074∗
(0.036)
-0.081∗
(0.038)
Rural
0.14
(0.11)
0.17
(0.12)
1419
-965
1334
-883
Observations
Ln likelihood
Standard errors in parentheses.
1419
-980
∗
p < 0.05
15
Table A.12: Logistic regressions of alternative measure of perceived costs of the southern conflict
and respondent support for pardons for competing central factions. Robustness test for Table 2.
Model 33
High perceived
conflict costs II
-0.0043
(0.34)
Pattani Malay
Model 34
Model 35
0.100
(0.32)
0.16
(0.33)
-0.072
(0.40)
-0.053
(0.44)
Military role too
large
0.71∗
(0.20)
Satisfaction w
national government
-0.26
(0.14)
Age
0.019
(0.040)
-0.022
(0.043)
Female
-0.23
(0.18)
-0.20
(0.19)
Secondary education
-0.094
(0.22)
-0.082
(0.24)
Post-secondary
education
-0.77∗
(0.30)
-0.82∗
(0.31)
Income
-0.052
(0.057)
-0.074
(0.060)
Rural
0.46∗
(0.17)
0.44∗
(0.18)
Favors
decentralization
-0.37
(0.20)
Sampling strata
Sample size
Standard errors in parentheses.
10
813
∗
10
813
p < 0.05
16
10
765
Table A.13: Logistic regressions of support for southern autonomy in provinces near 1st–3rd RTA
command area borders estimated without survey weights or clustering by sample strata. Robustness test for Table 4.
Southern autonomy
Central political reconciliation
Model 36
Model 37
Model 38
Model 39
Model 40
Model 41
0.78∗
(0.14)
0.78∗
(0.14)
0.72∗
(0.16)
-0.098
(0.11)
-0.10
(0.12)
-0.094
(0.13)
Ln regional army
fatalities
Favors
decentralization
1.5∗
(0.29)
-0.35
(0.26)
Military role too
large
-0.77∗
(0.30)
0.51
(0.26)
Satisfaction w
national government
0.059
(0.22)
-0.19
(0.20)
Age
0.016
(0.060)
-0.0054
(0.067)
0.064
(0.055)
0.044
(0.057)
Female
-0.086
(0.25)
-0.25
(0.28)
-0.37
(0.23)
-0.31
(0.24)
Secondary education
-0.056
(0.32)
-0.19
(0.35)
-0.22
(0.29)
-0.19
(0.31)
Post-secondary
education
-0.52
(0.47)
-0.64
(0.53)
-0.53
(0.44)
-0.46
(0.46)
Income
0.090
(0.081)
0.12
(0.088)
-0.071
(0.074)
-0.11
(0.077)
Rural
0.27
(0.25)
0.33
(0.28)
0.18
(0.23)
0.16
(0.24)
317
-192
299
-159
317
-214
299
-198
Observations
Ln likelihood
Standard errors in parentheses.
317
-194
∗
p < 0.05
17
317
-219
Table A.14: Logistic regressions of support for southern autonomy in provinces near 1st–3rd RTA
command area borders estimated without survey weights and with standard errors clustered by
province. Robustness test for Table 4.
Southern autonomy
Ln regional army
fatalities
Central political reconciliation
Model 42
Model 43
Model 44
Model 45
Model 46
Model 47
0.78∗
(0.35)
0.78∗
(0.35)
0.72∗
(0.33)
-0.098
(0.18)
-0.10
(0.18)
-0.094
(0.16)
Favors
decentralization
1.5∗
(0.55)
-0.35
(0.30)
Military role too
large
-0.77∗
(0.20)
0.51
(0.38)
Satisfaction w
national government
0.059
(0.14)
-0.19
(0.32)
Age
0.016
(0.054)
-0.0054
(0.044)
0.064
(0.048)
0.044
(0.046)
Female
-0.086
(0.25)
-0.25
(0.29)
-0.37
(0.27)
-0.31
(0.33)
Secondary education
-0.056
(0.30)
-0.19
(0.38)
-0.22
(0.55)
-0.19
(0.55)
Post-secondary
education
-0.52
(0.54)
-0.64
(0.53)
-0.53
(0.81)
-0.46
(0.70)
Income
0.090
(0.082)
0.12
(0.094)
-0.071
(0.047)
-0.11∗
(0.035)
Rural
0.27
(0.37)
0.33
(0.40)
0.18
(0.27)
0.16
(0.23)
317
-192
299
-159
317
-214
299
-198
Observations
Ln likelihood
317
-194
Standard errors clustered by province and reported in parentheses.
∗
18
p < 0.05
317
-219
Table A.15: Logistic regressions of support for southern autonomy in provinces near 1st–3rd RTA
command area borders including respondents who did not answer central political reconciliation
questions. Robustness test for Table 4.
Model 48
Model 49
Model 50
0.75∗
(0.17)
0.76∗
(0.17)
0.83∗
(0.20)
Ln regional army
fatalities
Favors
decentralization
1.7∗
(0.32)
Military role too
large
-0.31
(0.34)
Satisfaction w
national government
-0.19
(0.27)
Age
-0.014
(0.071)
-0.0090
(0.073)
Female
0.13
(0.29)
-0.065
(0.32)
Secondary education
-0.047
(0.38)
-0.47
(0.42)
Post-secondary
education
-0.94
(0.52)
-1.3∗
(0.59)
Income
0.100
(0.087)
0.12
(0.092)
Rural
-0.16
(0.25)
-0.18
(0.29)
6
342
6
314
Sampling strata
Sample size
Standard errors in parentheses.
6
342
∗
p < 0.05
19
Table A.16: Logistic regressions of support for southern autonomy in provinces near 1st–3rd RTA
command area borders not including Lop Buri and Nakhon Ratchasima. Robustness test for Table
4.
Southern autonomy
Central political reconciliation
Model 51
Model 52
Model 53
Model 54
Model 55
Model 56
0.72∗
(0.17)
0.73∗
(0.18)
0.68∗
(0.20)
0.12
(0.15)
0.20
(0.15)
0.26
(0.16)
Ln regional army
fatalities
Favors
decentralization
1.5∗
(0.40)
-0.64
(0.38)
Military role too
large
-0.75
(0.43)
0.050
(0.37)
Satisfaction w
national government
-0.089
(0.35)
0.23
(0.32)
Age
0.010
(0.090)
-0.050
(0.100)
0.0017
(0.083)
0.0044
(0.086)
Female
0.17
(0.37)
-0.040
(0.41)
-0.45
(0.35)
-0.37
(0.36)
Secondary education
0.11
(0.51)
-0.40
(0.55)
-1.1∗
(0.48)
-0.93
(0.48)
Post-secondary
education
-1.2
(0.69)
-1.7∗
(0.84)
-1.5∗
(0.67)
-1.2
(0.70)
Income
0.11
(0.12)
0.14
(0.12)
-0.0037
(0.10)
-0.025
(0.11)
Rural
0.068
(0.33)
0.0063
(0.37)
0.056
(0.33)
0.085
(0.35)
6
206
6
193
6
206
6
193
Sampling strata
Sample size
Standard errors in parentheses.
6
206
∗
p < 0.05
20
6
206