MARITIME PIRACY IN MALACCA STRAIT AND SOUTH CHINA …MARITIME PIRACY IN MALACCA STRAIT AND SOUTH...
Transcript of MARITIME PIRACY IN MALACCA STRAIT AND SOUTH CHINA …MARITIME PIRACY IN MALACCA STRAIT AND SOUTH...
MARITIME PIRACY IN MALACCA STRAIT AND SOUTH CHINA SEA: TESTING THE DETERRENCE AND REACTANCE MODELS BO JIANG Department of Criminology, University of Pennsylvania (August 2013 to May 2014) Department of Criminology and Criminal Justice, University of Maryland (June 2014 - ) KEYWORDS: Southeast Asia, maritime piracy, deterrence, reactance, series hazard model, logistic model ABSTRACT: In this article, I used series hazard modeling and multivariate logistic models to test the relative strengths of deterrence and reactance models for the risk of piracy attacks under a military intervention and several major events. I found significant evidence for the conclusion that the rate of piracy attacks reduced significantly when direct controls are implemented to reduce the environmental opportunities of crime, and when certainty effect becomes stronger. The only exception is when pirates are highly motivated. Also, my results support the reactance perspective when the odds of successful attacks significantly increase after anti-piracy military intervention was introduced. No. of words: 7354 No. of figures: 2 No. of tables: 5
1
INTRODUCTION
Of the many piracy prone areas in the world, the two prominent piracy hotspots in Asia are
the South China Sea and the Malacca Straits which are contiguous to each another. The South
China Sea is bounded on the north by China, on the south by Indonesia, on the east by the
Philippine islands, and on the west by the Malay Peninsula and Vietnam. It encompasses an area
of around 3.5 million square kilometers, and contains more than 230 scattered islets, reefs, and
shoals. It is home to the world’s most perplexing problems of jurisdiction and maritime
management. Linking the Indian Ocean with the South China Sea, Malacca Strait is one of the
world’s busiest shipping lanes. Its strategic importance lies in it being the shortest trade route
between the Indian and the Pacific Ocean. Prior to 2007, Malacca Strait was notorious for being
the global hot spot for piracy but with the increase of piracy off the coast of Somalia, it no longer
holds that designation. Currently, about 45% of total world trade carried by 70,000 merchant
vessels over 300 gross tons ply the South China Sea and about 20,000 oil tankers carrying one-
third of the world’s crude oil passes through Malacca Strait annually.
Other than commercial vessels, numerous smaller vessels such as barges, fishing boats, and
passenger vessels also transit the area. Ships passing at low speed close to shore, with the
equivalent of several years’ wages in cash in the ship’s safe and with foreign crews and
timetables militating against full reporting of incidents, made them obvious and tempting targets
2
as some of these ships can feed a whole Indonesian village1. These types of attacks can be
conducted by a small group of men or by a ring of people attacking ships on behalf of the entire
villages. In these later cases, “whole villages”, located predominantly on remote islands, attack
vessels for food, cigarettes, and other small goods. Stuart (2002) noted that while the attack itself
is conducted by a small number of able-bodied men from the community, the whole village
might share the booty, benefit from the attack, and admire and support the perpetrators.
Pervasive poverty, lack of legal employment opportunities combined with the proximity to
shipping lines nourished piracy as a ‘thinkable’ alternative way of earning a living (Vagg, 1995).
Frecon (2006), who conducted research in the Riau Archipelago and interviewed pirates and
members of the community they live in, concludes that locals tolerate piracy as a means to
living. Piracy appears almost the “logical choice” for such fishers as they have the necessary
maritime skills, local knowledge of the area, and the required equipment, including boats and
long knives.
The occurrence of piracy trended upward. By 2005, the IMO estimated that pirate attacks in
the Malacca Strait accounted for 40% of overall global incidents. Indeed, maritime attacks in the
region have caused an estimated annual $3 billion in economic loss.2 Piracy poses a danger not
only to shipping companies for their financial loss but also to the safety of 1.4 million seafarers
who are transiting the area. Stanley Weeks notes that “…piracy raises insurance rates, restricts
1 Sethuraman Dinakar and Harry Maurer, “The Jolly Roger Flies High, as Piracy Feeds the Hunger”, Business Week International Edition, 24 May 1999. 2 Brandon, J.J. 2003. Piracy as Terrorism, Journal of Commerce., issue: June 3.
3
free trade, and increases tensions between the affected littoral states, their neighbors and the
countries whose flagged ships are attacked or hijacked.” In addition, since the terrorist event of
September 2001 in the U.S., security alerts of potential maritime attacks by terrorist groups had
been issued. Malacca Strait was added to the “Hull, War, Strikes, Terrorism and Related Perils”
list by the Joint War Committee (JWC) in mid 2005 to reflect the possibility of this doomed
scenario (Reinhart, 2012).
Notwithstanding the numerous literatures on maritime piracy, I was not able to identify any
studies of maritime piracy that explicitly examined the deterrence and reactance effects. There is
also a lack of prior studies that quantify the effectiveness of anti-piracy measure and policies in
the context of regional socioeconomics conditions using statistical tests. Moreover, there is scant
literature employing empirical methods in studying the phenomenon of piracy in terms of
typology, trend, location, vessel details, modus operandi etc. Most of the prior works on
maritime piracy were qualitative and theoretical in their approach, focusing mainly on the
historical context (Mueller and Adler, 1985; Eklof, 2006; Elleman et al., 2010); socioeconomic
conditions that nourish piracy (Vagg, 1995; Johnson and Valencia, 2005; Young, 2007), the
connection between organized crime, terrorists, guerrilla movements and maritime piracy
(Chalk, 2000; Ong-Webb, 2006) and responses to piracy (Liss, 2010; Beckman, 2012). Many of
the authors conducted field work through interviews with the pirates, officers from the
enforcement agencies, crews who are victims of piracy and the village chief where the pirates
originate. Although such literatures provided a qualitatively revealing and comprehensive
4
discourse on the retrospect and prospect of maritime piracy, they did not offer systematic
statistical tests of their hypothesis.
MILITARY INTERVENTION AND MAJOR EVENTS
After an extensive literature review, I identified one military intervention and four major
events that could potentially have huge impacts on the risk of piracy in Malacca Strait and South
China Sea. The military intervention is called MALSINDO (which is an acronym based on the
participating countries of MALaysia, SINgapore and InDOnesia). Started in 2004, MALSINDO
is by far the most elaborated and extensive effort to combat piracy in the Malacca Strait. It
requires “…the navies of the participants to commit five to seven ships to patrol the strait, and to
establish a hotline that allowed the commands of the three navies to coordinate operations”. The
effort is reinforced by joint aerial surveillance through the ‘Eye in the Sky’ Initiative that involve
“regular reconnaissance sorties linked to a web-based information sharing network to allow
better information sharing” (Schuman, 2009). Under this framework, patrol vessels from one
country are permitted to enter the soverign waters of another country when under hot pursue of a
pirate ship.
Of the four major events, the first is Cooperation Afloat Readiness and Training (CARAT),
which comprises a series of annual bilateral military exercises created in 1995 and conducted by
United States Pacific Fleet with Singapore, Thailand, Brunei, Indonesia, the Philippines,
Malaysia and Indonesia. During each stop, the U.S. naval forces exercised with the host nation's
5
air, sea and land forces. The second event is the Indonesian Tsunami that occurred on December
26 2004. With a total death toll estimated around 230,000 including 10,000 foreign tourists
(mostly Europeans), the event prompted a worldwide humanitarian response. During the search
and rescue phase of the relief operations the U.S. Navy dispatched 17 ships including an aircraft
carrier, as well as 75 aircrafts including a P-3 patrol aircraft to assist the relief operations. In
addition, more than 11,600 military personnels were deployed within one month of the disaster. I
included these two events because the sudden influx of U.S. military presence in the regions
could potentially have a large deterrence effect on piracy.
The third event is the Hainan Incident. On April 1, 2001, a mid-air collision between a
United States Navy P-3 signals intelligence aircraft and a People's Liberation Army Navy
interceptor fighter jet resulted in an international dispute between the U.S. and China. As
tensions mount between US and China and subsequently spill over to other regional states in the
South China Sea, heighten sense of mutual distrust and the fear of losing sovereignty over
national territory hampered international cooperation. As such, coordinated patrols and cross
border hot pursuits of pirate ships were forbidden. When the certainty of apprehension falls, the
deterrence perspective predicts that piracy is likely to become more active.
The last event is Lebaran (Celebration Day), which takes place during the first three days
immediately after Ramadan. Lebaran is the biggest holiday in Malaysia and Indonesia and is one
of the largest temporary human migrations globally where workers, particularly unskilled
laborers, return to their hometown to celebrate with their families. Thus, during the month of 6
Ramadan pirates might be highly motivated because they needed extra cash to bring home for
celebration. The stronger is the drive of the pirates, the more likely piracy will occur.
DETERRENCE AND REACTANCE EFFECTS
Kenny (1929) summarized that writers such as Beccaria, Blackstone, Romilly, Paley,
Feuerbach define deterrence by punishment as a method of retrospective interference after crime
occurs; by holding out threats that, whenever a wrong has been actually committed, the
wrongdoer shall incur punishment. Punishment is usually concurrent with reinforcement – that
is, the behavior which is punished is also rewarded. For example, a pirate attacks a ship for
money (reward), for which, if caught, he is placed in prison (punishment). If the strength of the
rewarding stimulus is greater than that of the punishing stimulus, the utilitarian pirate will most
likely continue to behave in his old way even if punished. The deterrence effect could also be
achieved when incapacitation theory (Ehrlich, 1972) is interpreted together with rational choice
theory (Becker, 1968). As the perceived probability of apprehension increases, deterrence effect
is achieved when utilitarian criminals choose legal avenues of economic gain over illegal ones.
To be effective, the punishing stimulus must be unavoidable and in order to do so, apprehension
must be certain. Zimring and Hawkins (1973) note that a substantial deterrence effect is achieved
when the probability that a particular type of offender will be apprehended is greatly increased.
Also, Marvell and Moody (1996) suggest that additional police presence deters crime in this
7
manner by making criminals believe arrests and subsequent sanctions are more likely3. For
piracy, an increase in the presence of the costal enforcement machinery would increase the
objective probability of apprehension. Therefore my first deterrence-related hypothesis is:
H1: The hazard of piracy attacks is inversely proportional to the objective probability of
apprehension.
Consistent with Maslow and Murphy’s (1954) theory on motivation, I hypothesize in H2b
that deterrence measures operating on potential pirates may be sufficiently influential to deter
piracy that is weakly motivated but will fail to deter piracy that is strongly motivated. For certain
people in certain situations, the drive toward piracy may be so strong that no threat will stop
them. My second deterrence-related hypothesis is:
H2: The strength of the drive that motivates piracy has a negative effect on amenability to
deterrence.
The second decisive factor in crime control is dealing with future consequences of
punishment. There is an extensive literature (Braithwaite, 1989, 2005; Dugan et al., 2003; Nevin,
2003; Siqueira and Sandler, 2007; LaFree et al., 2009) concluding that punishment has some
undesirable side effects, such as backlash and psychological reactance (Brehm, 1966; Tyler,
1990). The reactance perspective suggests that for a given person at a given time, there is a set of
behaviors any one of which he could engage in either at the moment or at some time in the
3 In New York City the rate of street crime decreased after the saturation of a small high crime area with extra police. New York City Police Department, Operation 25 (1955).
8
future. This set may be called the individual’s “Free Behaviors”. For specific behaviors to be
free, the individual must have the relevant physical and psychological abilities to engage in them.
Given that a person has a set of free behaviors, he will experience reactance whenever any of
those behaviors is eliminated or threatened with elimination. The greater is the importance and
proportion of the free behaviors which are eliminated or threatened with elimination, the greater
will be the magnitude of reactance predicted. I relate the reactance theory to my current focus of
piracy through the following hypothesis:
H3: Pirates would re-establish their freedom of sea robberies through more aggressive attacks
when their free behaviors are threatened with elimination.
DATA AND METHOD
DATA
The analysis is based on 438 and 1678 piracy attacks from 1995 to 2010 in Malacca strait
and South China Sea, respectively, drawn from the Piracy and Armed Robbery Reports in the
Global Integrated Shipping Information System (GISIS) developed by the International Maritime
Organization (IMO). The database was supplemented with Piracy Reports compiled by the IMO.
Information provided in the data set included the date of the attack, ship name, ship type, area of
the attack, incidence details, consequences for crews, where the incident was reported to and
costal state actions. Ancillary data on the GDP, import (cif) and export (fob) values,
9
unemployment rate, inflation rate were obtained from CEIC Data Manager. The number of
terrorism incidents was drawn from the Global Terrorism Database (GTD) developed by the
National Consortium for the Study of Terrorism and Responses to Terrorism (START).
METHOD
Following Dugan (2011)4, I use the series hazard model, which is an extension of the Cox
(1972) proportional hazard model, to estimate the impact of the military interventions on the
hazard of an additional piracy attacks controlling a set of relevant variables. The hazard model
used in the analysis is specified as follows:
h(Y) =𝜆𝜆0(Y) exp (𝛽𝛽1Military Interventions + 𝛽𝛽2Other Events + 𝛽𝛽3Context)
where β is a vector of parameters to be estimated, and 𝜆𝜆0(Y) is the unspecified baseline hazard
function. The baseline hazard function corresponds in my case to the probability of pirate attack
when all the covariates have values of 0. In this study, I assume the unspecified baseline hazards
for Malacca Strait and South China Sea are the same since they are contiguous and share many
unobserved characteristics. I will present the estimated coefficients β in each geographical area
under the section ‘RESULTS’ before discussing their relevance under the section
‘DISCUSSIONS AND CONCLUSIONS’.
4 Other studies that use the series hazard model include 1) the study of temporal frequency of aerial hijacking (Dugan et. al., 2005), 2) the estimation of the impact of British counterterrorist strategies on political violence in Northern Ireland (LaFree et. al, 2009).
10
“Days Until Next Attack” is the dependent variable used in the series hazard model, which
measures time (days) to each event from the time of the previous event. The clock is reset to zero
after each attack. In Malacca Strait, the first attack occurred on January 5, 1995 and this date was
chosen to be the time origin as it marks the onset of continuous exposure to risk of piracy
attack.5 Next I observe that in Malacca Strait, there are 14 ties and 59 repeated attacks while in
South China Sea, there are 114 ties and 320 repeated attacks. For example in South China Sea,
attacks occurred on June 1st, 1995, June 2nd, 1995, June 3rd, 1995 and June 4th, 1995, thus giving
rise to three ties. In the same geographical area, three repeated attacks occurred on April 11th,
1995. Given that 19% (320/1679) of the attacks in South China Sea are repeated in nature, it is
crucial to correct for the dependence among the multiple attacks on the same day.
Dependence among observations can be thought of as arising from unobserved
heterogeneity. Second attacks tend to be like first attacks because there are unmeasured, stable
factors affecting both attacks. For example, repeated attacks occur when the same group of
pirates attacks multiple ships of the same type in the same region of the sea using very similar
methods in the same day. Therefore, pooling these observations without taking the dependence
into account can lead to standard error estimates that are biased downward and test statistics that
are biased upwards.
5 This criterion excludes earlier periods, i.e. January 1, 1995 to January 4, 1995, when the hazard is necessarily 0. 11
To correct for dependence among repeated attacks, the Robust Variance Estimator method6
(rather than the Observed Information Matrix) was used (Allison, 2010). Because observations
are correlated, some of the apparent information in the sample is redundant. Therefore, repeated
attacks on the same day are given the same Strata id so that the estimation procedure is not
‘fooled’ into thinking that it has more information than it actually does. Consequently, in South
China Sea the standard errors are adjusted for 1358 clusters. However, the tradeoff is that when
the Robust Variance Estimator method was invoked, the EXACT method could not be used since
the partial likelihood function is no longer valid. Ties were thus handled by an approximation
proposed by Efron (1977). This tradeoff is justified because in South China Sea ties constitute a
much smaller fraction (114/1679) as compared to repeated attacks (320/1679).
Table 1 provides the summary statistics of variables included in the analysis. I examine
separately all attacks in Malacca Strait from January 1995 to March 2008 and South China Sea
from January 1995 to December 2010. The mean number of days between successive attacks is
around twelve days in Malacca Strait and four days in South China Sea.
<Insert Table 1>
MILITARY INTERVENTION
6 It was developed for Cox regression by Lin and Wei (1989) and described in some detail in Therneau and Grambsch (2000).
12
For the list of variables under Military Intervention and Other Events, the dummy variable
“1” is coded if the piracy attack occurred during the period and “0” otherwise. Since
MALSINDO officially began in July 2004 but was only fully operational with the launch of the
Eyes in the Sky in July 2005, the midpoint of those two dates was taken and piracy attacks
recorded from January 2005 till the end of the sampling period were coded as “1” and “0”
otherwise. According to Table 1, MALSINDO was in effect for about twelve percent and thirty
percent of the total attacks in the Malacca Strait and South China Sea, respectively.
OTHER EVENTS
The start date for CARAT in Malacca Strait is proxy by the CARAT start date in Singapore7
and the end date is taken to be ten to fourteen days after the start date. In South China Sea,
CARAT starts on either May or June each year8. The end date is taken to be around two months
after the start date. Approximately fourteen and seventeen percent of the piracy attacks were
perpetuated during the periods of CARAT in Malacca Strait and South China Sea, respectively.
The next variable is “Indonesian Tsunami”. Piracy attacks perpetuated in Malacca Strait
during the one year period after the tsunami were coded as “1” and “0” otherwise. About four
percent of the total attacks were recorded during this period. The second variable “During
7 May 13, 1995; July 13, 1996; July 21, 1997; July 20, 1998; July 12, 1999; July 11, 2000; July 2, 2001; July 2, 2002; July 14, 2003; June 1, 2004; May 31, 2005; May 30, 2006; July 17, 2007 and June 24, 2008 8 The dates are May 12, 1998; June 28, 1999; June 14, 2000; May 17, 2001; May 1, 2002; May 28, 2003; June 12, 2004; June 15, 2005; June 14, 2006; May 12, 2007 and May 3, 2010.
13
Ramadan” is a yearly recurring event9 and piracy attacks perpetuated during these periods were
coded as “1” and “0” otherwise. Lastly, piracy attacks perpetuated within one year of the Hainan
Incident were coded as “1” and “0” otherwise. Altogether, they accounted for about nine percent
of the total attacks according to Table 1.
CONTROL VARIABLES
The existence and activities of guerrilla and terrorist groups in Southeast Asia affect piracy
in two ways. First, it is believed that members of some of these politically motivated groups
conduct pirate attacks to finance their operations. Second, many scholars suggested that pirates
may cooperate with, or be of assistance to, terrorist organizations (see Koknar, 2005 and Low,
2008 for a discussion). As expected, the monthly number of piracy attacks is significantly
correlated with that of terrorist attacks in Malacca Strait (r=0.196, p=0.013). Therefore in order
to maximize confidence in the results, I need to control for changing rates of terrorist related
piracy attacks over time. This was done by including the variable “Lagged Regional Terrorist
Incidents” which captures the one-month lagged monthly aggregate counts of terrorism incidents
that occurred in countries along the Malacca Strait from 1995 to 2008. The littoral countries are
9 In the sampling period from 1995 to 2010, Ramadan began on the following dates: February 1, 1995, January 22, 1996, January 10, 1997, December 31, 1997, December 20, 1998, December 9, 1999, November 28, 2000, November 17, 2001, November 6, 2002, October 27, 2003, October 16, 2004, October 5, 2005, September 24, 2006, September 13, 2007, September 2, 2008, August 22, 2009, August 11, 2010 and lasted for twenty nine or thirty days.
14
Singapore, Malaysia and Indonesia. Lagging the terrorism incidents by one month is crucial to
avoid reverse causality.
To test the possibility that it was not military interventions and other events but the increased
availability of “preys” in the regions that explain the hazard of piracy attacks, I included the
variable “Trade”. It captures the cumulative value of trade passing through the Malacca Strait on
a quarterly basis for the years 1995 to 201010.
I included the variable “Regional Military Expenditure” to control for the lower/higher
hazards of piracy due to the larger/smaller presence of enforcement at sea. The military
expenditure for countries located along the Malacca Strait, i.e. Malaysia, Indonesia and
Singapore were added and divided by the sum of the value of GDP in these countries to derive
the regional military expenditure as a percentage of GDP.
To guard against the possibility that it was the strain/anomie theory that predicted the hazard
of piracy attacks rather than military interventions or specific events, I included the control
variable “Regional Misery Index”. It was computed as the sum of the unemployment rate and
inflation rate in the regions concerned.
10 In the Malacca Strait cumulative trade is measured as the sum of export (fob) from South Korea, Japan, Macau, Hong Kong, China, Taiwan, Vietnam, Philippines, Thailand, Malaysia, Indonesia, Brunei and Singapore to the European Union, plus import (cif) from the European Union to these thirteen Asian countries/territories. The unit of measurement is USD millions. A similar method was used to calculate the value of cumulative trade passing through the South China Sea less Indonesia and Singapore since trades from these two countries do not pass through South China Sea.
15
Lastly, I use the temporal ordering of piracy attacks to create a success density measure.
This measure is constructed by taking the current and two previous attacks, and calculating the
probability of success for the current and two previous attempts divided by the duration of the
current event from the second previous event, weighted by 365 days. A large success density
would indicate a high success probability for most of the recent incidents. Lastly, I included a
measure of officially reported attacks to account for the momentum of previous attacks. I
observed from Table 1 that more than seventy percent of the attacks were reported to the relevant
costal authorities. Also, the variables “Tankers”, “Container and Cargo Ships” and “Non-
Commercial Ships” were included to study the hazards and the likelihood of success against the
different types of ships.
RESULTS
Figure 1 shows four time series illustrating the trends in piracy in the Malacca Strait and
South China Sea from 1995 to 2010. Overall, it can be seen that Southeast Asia as a whole has
experienced a drop in the number of attacks in recent years. The share of attacks committed in
the region fell to approximately 10 percent of the worldwide total in 2009, compared to 18
percent in 2008 and more than 38 percent in 2003. The drop has been especially marked in the
Malacca Strait, where there were two incidents in 2009, the same number as in 2008, and a
drastic drop compared to the early 2000s (for instance, there were 28 incidents in 2003 and 38 in
2004). As expected, there is a significant correlation between yearly piracy attacks (r=0.664,
p=0.007) in Malacca Strait and South China Sea.
16
<Insert Figure 1>
The use of both vertical axes in Figure 1 allows the comparison of two trends in one graphic.
The left vertical axis shows the number of attacks recorded for each region in a given year while
the right vertical axis shows the percentage change of the number of attacks per year. The solid
line in Figure 1 shows the trend in the number of attacks in Malacca Strait. For the years 1995-
1998 and 2007-2008, the number of attacks was below the median of seventeen cases. In fact in
2008, only one attack was recorded and there were no further attacks till the end of the sampling
period in 2010. The highest number of attacks recorded are in the years 2000 (One hundred and
eighteen), 2004 (sixty) and 2004 (fifty-eight). Measured along the right vertical axis, the solid
line with circle shows the yearly percentage change in the number of attacks in Malacca Strait.
For the years 1997, 2001, 2002, 2005 and 2007 to 2010, the rate of change of attacks was below
the average rate of minus thirty seven percent in the sampling period. The lowest rate of change
occurred from 2007 to 2009, in which it dropped drastically by ninety two percent and one
hundred percent respectively. There was an especially sharp rise in the rate of change of attacks
from 1998 to 2000, where it increased by seven hundred and forty percent and one hundred and
eighty one percent respectively.
Turning over to South China Sea, the dash line in Figure 1 shows the trend in the number of
attacks. For the years 1995, 1997, 1998, and 2006 to 2009, the number of attacks was below the
median of ninety-eight cases. The region was relatively ‘calm’ during 2006 to 2009, with an
average of only around sixty-nine cases. In the opposite end, the region was the most infested 17
with piracy during the period 1999 to 2004, with an average almost twice that of the ‘safest’
period. The peak occurred in the year 2000 with one hundred and forty five attacks recorded
during the year. Measured along the right vertical axis, the dash line with cross shows the yearly
percentage change in attack in South China Sea. For the years 1997 to 1998, 2001, 2004 to 2006,
the rate of change of attacks was below the average rate of one point four percent in the sampling
period. During the ‘safest’ period, the rate of change of attacks dropped sharply by twenty five
percent, thirteen percent and thirty two percent. On the opposite end, the rate of change of attack
reached its peak in the period 1998 to 1999, where there was a sharp increase of forty eight
percent.
Comparing the two regions, Figure 1 reveals that from 1995 to 1998, Malacca Strait had a
lower year on year percentage change compared with South China Sea. From 1999 till the end of
the sampling period, the volatility of the percentage change in Malacca Strait is much higher than
that of South China Sea. Specifically, both time series peak in the year 1999 but Malacca Strait
had a peak value that is about fifteen times larger than that in the South China Sea.
SERIES HAZARD MODELS RESULTS
<Insert Table 2>
<Insert Table 3>
18
Table 2 and 3 shows the series hazard model results in Malacca Strait and South China Sea
respectively for all attacks, attacks perpetuated in territorial water, international water and at
port. Overall, the results provide strong evidence for the deterrence perspective. In Hypothesis 1,
I predict that the hazard of piracy attacks will decrease when the certainty of apprehension
increases. Consistent with H1, in Table 3 it is shown that during the period when CARAT takes
place every year in South China Sea, the hazard of ships being attacks falls by around 12%11
and the reduction is more prominent in international water, where ships have 26% lower hazard
of being attacked. Secondly, Table 2 shows that ships have a 48% lower hazard of being attacked
during the one-year period after the Indonesian Tsunami.
The effects of MALSINDO in Malacca Strait also confirm the certainty of apprehension
hypothesis H1. Compared with ships that were attacked before the introduction of MALSINDO,
ships have 78% lower hazard of being attacked after MALSINDO was introduced. The biggest
hazard reduction occurs at territorial water, where ships have nearly 95% lower hazard of being
attacked. Next I examined the effects of Hainan Incident in South China Sea. According to Table
3, the hazard of ships being attacked is 43% greater in the one-year spell after the Hainan
Incident. Therefore there is significant evidence in both regions to support Hypothesis 1.
My second deterrence-related Hypothesis 2 predicts that the strength of the drive that motivates
piracy has a negative effect on amenability to deterrence. In support, Table 2 shows that during
11 The hazard ratio is calculated by taking the exponential of the coefficient estimate. 19
Ramadan, ships have 38% greater hazard than non-Ramadan period of being attacked. In
particular, piracy at the port area is 4.668 times higher during the month of Ramadan.
Under Context, there are several significant results in Table 2 and 3. Firstly, it is shown that
the value of trade passing through the Malacca Strait and South China Sea has a neutral impact
on the hazard of ships being attacked. Secondly, the fact that every percentage increase in
Regional Military Expenditure decreases the hazard of being attack by 71% is also another
significant support for H1. The strain/anomie effects are also present since every percentage rise
in Regional Misery Index increases the hazard of port piracy by 27%.
The next two variables, Success Density and Reported Attack, are benefit related and based
on the premise that pirates will be more likely to attempt piracy when the expected benefits of
piracy increase. For every reported attack, the hazard of the next attack increases by 37%. My
results support the conclusion that new piracy attacks will be more likely shortly after earlier
attacks. Moreover, the hazard of a new piracy attack will increase by 1.3% in the Malacca Strait
and 1% in the South China Sea when the most recent three attacks were successful and close
together. By comparing these two coefficients, contagion effects are found to be driven mainly
by all events instead of only successful events.
Lastly, it is also noted that the hazard of attack in South China Sea drops by 10% if the ship
type is container and cargo ships. However, the effects are reversed in the Malacca Strait, where
container and cargo ships have 35% higher hazard. Perhaps this is because container ships need
20
to slow down in the busy but narrow and shallow Malacca Strait to avoid collision, making them
vulnerable targets.
LOGISTIC MODELS RESULTS
The significance of the variable Success Density in Tables 2 and 3 indicate the existence of
contagion effect. Given that the frequency of piracy attacks will increase following successful
attacks, in this section I use the logistic models estimating success to study the characteristics
and determinants of successful attacks. In the logistic regression, the dependent variable is
“Success” which is a dummy variable that takes a value of “1” if the attack is successful and “0”
otherwise12. In Table 1, it is shown that approximately fifty four percent and sixty six percent of
the attacks in Malacca Strait and South China Sea are successful. Similar to the findings in
Figure 1, Figure 2 shows that during the sampling period, the yearly number of successful
attacks in Malacca Strait and South China Sea are significantly correlated (r=0.561, p=0.024).
According to both Figure 1 and 2, during 1996 to 2001 although there is an increasing number of
a piracy attack in South China Sea, the number of those attacks which are successful is in fact
declining. And, as we have seen, during the sampling period the volatility of the yearly
percentage in both attacks and successful attacks in Malacca Strait is much greater than that of
the South China Sea.
12 I use the column “Incident Details” in the data set to track trends in successful and non-successful piracy attacks from 1995 to 2010 in Malacca Strait and South China Sea. I defined success as when the pirates managed to board the ship and escaped with some valuables, managed to kidnap any crew(s) or when the ship is hijacked.
21
<Insert Figure 2>
In Table 4 and 5 I report the estimated conditional odds ratio from a logistic regression
analysis of variables measuring Military Intervention, Other Events and Context. The odds ratio
reported measures the ratio of the odds of an event occurring in the successful group to the odds
of it occurring in the non-successful group, holding fixed the values of other variables. All
coefficients greater than one are risk factors of piracy attacks.
<Insert Table 4>
<Insert Table 5>
Overall, the logistic model results provide support for the reactance effects. In Hypothesis 3
I examine whether pirates would re-establish their freedom of sea robberies through more
aggressive attacks when their free behaviors are threatened with elimination. Indeed as shown in
Table 2 and 4, there are significant evidences supporting H3. Although ships have 78% lower
hazard of being attacked following the introduction of MALSINDO, the odds that a piracy attack
was successful increased by 288.4 percent (388.4-100). Thus, the probability of a successful
piracy attack increases from 0.511 to 0.77313, indicating more aggressive attacks.
The next series of findings relate to Other Events. As shown in Table 4, during the month of
Ramadan, the odds that a piracy attack in Malacca Strait was successful dropped by 42.5 percent,
13 Calculated by holding all other variables at their mean. 22
indicating that the probability of a successful attack drop from 0.562 to 0.437. This finding is
hardly surprising because fasting is obligatory for the Muslim pirates and consequently, attacks
are less likely to be successful as they are physically weaker.
Under Context, there is significant evidence in Table 4 to support the argument that the
value of trade passing through Malacca Strait has no impact on the odds of successful attacks.
Next, Tables 4 and 5 provide significant evidence to support Hypothesis 2, where the strength of
the drive that motivates piracy has a negative effect on amenability to deterrence. It is shown that
when livelihoods in the regions become more miserable, pirates are more motivated to succeed.
In particular, for every one percent increase in the Misery Index, the odds of a successful attack
increased by 4.90 percent in South China Sea. Strangely, the odds of a successful attack
increased by 89.8 percent and 35.9 percent in the Malacca Strait and South China Sea
respectively when it was reported to the coastal states.
There is significant evidence indicating that non-commercial ships are much more prone to
successful attacks as compared to tanker, containers and cargo ships. The odds of mounting a
successful attack towards non-commercial ships is approximately 4.66 times (8.49/1.82) more
than that of container and cargo ships in the Malacca Strait and twice (2.23/1.15) in South China
Sea. The greatest disparity occurs in the international water along Malacca Strait, where non-
commercial ships are 5.75 times (13.978/2.429) more likely to be victims of successful attacks as
23
compared to container and cargo ships. This region is the most dangerous ‘hotspot’ of successful
attacks, since the odds ratios of successful attacks among all ship types are the highest.
DISCUSSION AND CONCLUSIONS
In this article, I identified a military intervention and four major events related to piracy and
tested their impacts on subsequent attacks from 1995 to 2010. Overall, I conclude that the
military intervention and three of the major events produced deterrence effects and reactance
effect was observed in the military intervention.
First, and most policy relevant, there is strong evidence in the results to support the
hypothesis that new piracy attacks are less likely to be undertaken when the certainty of
apprehension increases. The reduction in piracy in the Malacca Strait is often attributed to
heightened anti-piracy efforts by littoral countries and improved cooperation between them since
2004 (Ong-Webb, 2006; Schuman, 2009). As a result, the military intervention MALSINDO
greatly increased the objective probability of piracy apprehension in the Malacca Strait as it
provided the means for each coastal state to conduct individual patrols in waters under its own
jurisdiction whilst maintaining communications with each other for better coordination in hot
pursuits. The results showed that the two major events, CARAT and Indonesian Tsunami, have
strong deterrence effects on piracy. In these two events, the dominant presence of the
international community, especially US military ships, aircrafts and personnel greatly increased
the perceived probability of apprehension. They impose strong direct controls over the
24
environmental conditions in Malacca Strait and South China Sea by greatly increasing the
perceived probability of apprehension.
Moreover, deterrence effect alone might not fully justify the profound 48% lower hazard of
attack along the Malacca Strait during the one-year period after the Indonesian Tsunami. Perhaps
an additional effect, the strain/anomie effect, is at work here. It is possible that shortly after the
tsunami, many impoverished local fishermen are hired by Non-Profit-Organizations (NGOs) to
help out in the rescue operations. Since pervasive poverty is the main driver that pushes local
fishermen into becoming common sea robbers to begin with, gainful employment opportunities
would have eradicated a huge potential pool of pirates.
Conversely, my findings indicate that the hazard of piracy attacks increases when deterrence
effects are weaker. The Hainan Incident is a case in point. Without cross border hot pursuits, the
objective probability of apprehension is much lower. Thus, the incidence of piracy increases
after the Hainan Incident.
The second policy relevant conclusion pertains to the determination of the pirates. I found
that deterrence measures operating on potential pirates may be sufficiently influential to deter
piracy that is weakly motivated but will fail to deter piracy that is strongly motivated. In the
Ramadan month piracy is strongly motivated because pirates need extra money to get things
ahead of Lebaran as well as to compensate for the loss in ‘income’ during the Lebaran period.
25
The third policy relevant conclusion is that the success of deterrence of piracy in Malacca
Strait is undermined by the presence of reactance. While a significant deterrence effect is
attributed to MALSINDO, my results indicate an adverse reactance effect to this policy. It could
be argued that piracy in Southeast Asia is a form of “Free Behavior” because the impoverished
local fishermen have both the relevant physical and psychological abilities to engage in it.
Moreover, they know by general custom that, they may engage in it without any moral restraint
and sanction. Therefore according to the reactance perspective, when their free behaviors are
punished, a hostile, aggressive response on the part of the punished person is created, resulting in
the higher probability of success in their next attack.
Another significant finding is that the values of trade passing through the two areas have no
effect on both the hazard of piracy attacks and the odds of successful attacks. This result could be
reconciled by the fact that on one hand, increased trade meant the sheer frequency with which
situations present themselves make crime both tempting and easy. On the other hand, increased
trade also has a deterrence effect when the increased presence of ships, particularly those
travelling in a convoy, prevents pirates from attacking. Both effects could have neutralized each
other.
There is no significant evidence to show that the existence and activities of guerrilla and
terrorist groups in Southeast Asia affects the hazard of piracy in both regions. The pirates seem
to be of a very different mindset to terrorists, and many of these pirates may not believe that
26
killing a large number of “innocent” people to further political aims is justifiable. Overall,
terrorists, unlike pirates, aim at disrupting society, while criminals, especially organized crime
criminals, tend to benefit from current political and economic arrangements. For example, if a
major maritime terrorist attack was to occur in the Malacca Strait, the area would be heavily
patrolled in the aftermath of the event to prevent further such acts. Subsequently, pirate attacks in
these waters would become difficult to conduct, particularly if vessels were forced to choose
different transit routes.
Strong support exists for the conclusion that military expenditure leads to lower hazard of
piracy. Higher military expenditure means greater presence of coastal enforcement. From the
deterrence perspective, greater presence of coastal enforcement acts as a form of direct control
on piracy as it reduces the environmental opportunities for crime. From the deterrence
perspective, the greater presence of costal enforcement would indicate a higher certainty effect,
thereby reducing crime.
The results also reveal the presence of strain/anomie effects at work in piracy when higher
Misery Index leads to greater hazard of being attack and more successful attacks. With the lack
of legal employment opportunities and widespread poverty in some communities escalating
especially in the aftermath of the 1997 Asian financial crisis, illegal activities are an alternative
way to earn a living. As fish stock depletes in the region, some of the more desperate fishers turn
to piracy as a source of income in a time of need, while more opportunistic fishers may use it to
27
earn extra cash or to supplement low catches. Unemployed and desperate fishers are also, in
some cases, recruited by organized crime gangs to attack or hijack merchant vessels or tugs. For
jobless and impoverished fishers, “employment” as a pirate by an organized gang may, therefore,
be one of the few options left to earn an income.
Lastly, my results indicate that non-commercial ships are much more susceptible to
successful attacks than commercial ships. As fish stock depletes in the region, fishing boats need
to venture farther into secluded ocean areas in order to increase their catch, thus making them
vulnerable preys. Compared to commercial ships, they have lower side walls which make
boarding by the pirates easy. They are also rarely equipped with pirate deterrent equipment such
as high-pressure water hose or a safe room. Moreover, commercial ships are tracked by the
automatic identification system (AIS) while most fishing boats are not. Thus, when commercial
ships suddenly stop in their intended course when they encounter pirates, the coastal enforcement
agencies would be notified. The lack of these pirate-prevention installations is probably another
reason why non-commercial ships appear riskier than commercial ships.
One counter-intuitive finding is that when an attack is reported to the coastal state, the
probability of the next successful attack increases. This last point highlights one limitation in this
study, which is the controversial nature of reporting a pirate attack. There are basically two
issues, which are self-selection in reporting and under reporting. In the former case which
explains the counter-intuitive finding, only successful attacks are reported because by reporting
ship owners could file an official record with the coastal law enforcement which will later be
28
used as evidence to make an insurance claim. Thus, it will be misleading to use such data as the
results will be consistently biased upwards. With regards to the problem of under-reporting, it
was estimated that fifty per cent of all pirate attacks remain unreported for a variety of reasons.
Some ship owners are reluctant to report attacks as they fear an investigation will delay their
vessel’s operation, resulting in additional costs. Many also do not want to be branded as
unreliable carriers of freight, or fear rising insurance costs. Additionally, attacks on fishing boats
and other smaller craft are rarely reported to the IMO.
Although this study is a first attempt at applying the deterrence and reactance framework to
piracy in Southeast Asia, which is the most “pirate infested” region in the world between 1992 –
2006, much remains unknown in other pirate hotspots of the world. I envision an additional
project to study the trends of piracy in the waters of Somalia and the Horn of Africa, which in
2008 has taken over the notorious title from Malacca Strait as a “war-risk” zone given by the
Lloyd’s Joint War Committee. As no prior study of which I am aware has collected data and has
applied formal statistics tests that compare the deterrence and reactance models, much more
research assessing the magnitude of anti-piracy military interventions in the region are needed.
29
REFERENCES
Allison, Paul D. 1995. Survival Analysis Using the SAS system: A Practical Guide. Cary, NC: SAS Institute Inc. Allison, Paul D. 2010. Survival Analysis Using SAS system: A Practical Guide, Second Edition. Cary, NC: SAS Institute Inc. Andenaes, Johannes. 1974. Punishment and Deterrence. Ann Arbor: University of Michigan Press. Azrin, N.H.1960. Sequential effects of punishment. Science. 131:605–606. Becker, Gary S. 1968. Crime and punishment: An economic approach. Journal of Political Economy 76: 169–217. Beckman, Robert C. and Roach, Ashley, J. 2012. Piracy and international maritime crimes in ASEAN: prospects for cooperation. Northampton, MA: Edward Elgar. Benson, Bruce L., Iljoong Kim, and David W. Rasmussen. 1994. Estimating deterrence effects: A public choice perspective on the economics of crime literature. Southern Economic Journal 61:160-168. Braithwaite, John. 1989. Crime, Shame and Reintegration. Cambridge, UK: Cambridge University Press Braithwaite, John. 2005. Pre-empting terrorism. Current Issues in Criminal Justice 17(96-114) Brantingham P.L. and Brantingham P.J. 1981. Notes on the geometry of crime, In P.J. Brantingham and P.L. Brantingham (Eds.), Environmental Criminology, Sage Publications, Beverly Hills, CA. Brehm Jack W. 1966. A Theory of Psychological Reactance, Academic Press, New York. Brehm, Stevens Sharon and Brehm Jack W. 1981. Psychological Reactance: A Theory of Freedom and Control. Academic Press. Chalk, Peter. 2000. Non-military security and global order: The impact of extremism, violence and chaos on national and international security. New York: St Martin’s Press.
30
Clarke, Ronald V. and Pat Mayhew.1980. Designing out Crime. London: H.M.S.O. Crowe, T. 2000. Crime Prevention through Environmental Design: Applications of Architectural Design and Space Management Concepts, 2nd Edition. Boston: Butterworth-Heinemann. Cox, D.R. 1972. Regression models and life tables, J.R. Statistical Society, Series B 34(184-220). Dugan, Laura, Nagin, Daniel S., Rosenfeld, Richard. 2003. Exposure reduction or retaliation? The effects of domestic violence resources on intimate partner homicide. Law Society Review 27 (169-198). Dugan, Laura, LaFree, Gary and Piquero, Alex R. 2005. Testing a rational choice of airline hijackings. Criminology 43(1031-1065). Dugan, Laura. 2011. The series hazard model: An alternative to time series for event data. Journal of Quantitative Criminology 27(379-402). Efron, B. (1977). The efficiency of Cox’s likelihood function for censored data. Journal of the American Statistical Association 72(557–565). Ehrlich, Isaac. 1972. The deterrent effect of criminal law enforcement. Journal of Legal Studies 1:259-276. Ehrlich, Isaac.1973. Participation in illegitimate activities: A theoretical and empirical investigation. Journal of Political Economy 81:521-567. Elleman, Bruce A., Forbes, Andrew and Rosenberg, David. 2010. Piracy and maritime crime: historical and modern case studies, Newport, R.I. : Naval War College Press.
Eklöf, Stefan. 2006. Pirates in Paradise: A Modern History of Southeast Asia’s Maritime Marauders, Nordic Institute of Asian Studies (NIAS) Press, Copenhagen, Denmark.
Farley, M.C.1993. International and Regional Trends in Maritime Piracy, 1989-1993. Naval Postgraduate School.
Frecon, Eric. 2006. Piracy and Armed Robbery at Sea Along the Malacca Straits: Initial Impressions from Fieldwork in the Riau Islands, in Graham Gerard Ong-Webb (ed.), Piracy, Maritime Terrorism and Securing the Malacca Straits, Singapore: ISEAS Publishing.
31
Gibbs, Jack P. 1975. Crime, Punishment, and Deterrence. Amsterdam: Elsevier Scientific Publishing. Gottfredson, Michael R. and Travis Hirschi.1990. A General Theory of Crime. Stanford, Calif.: Stanford University Press International Maritime Bureau. 2012. Reports on acts of piracy and armed robbery against ships, London, UK. Jeffery, C. Ray. 1971. Crime Prevention through Environmental Design. Beverly Hills, CA: Sage Publications. Johnson, Derek and Valencia, Mark. 2005. Piracy in Southeast Asia: Status, Issues, and Response, jointly published by International Institute of Asian Studies, The Netherlands and Institute of Southeast Asian Studies, Singapore. Kenny, Courtney Stanhope. 1929. Outlines of criminal law, Cambridge: Cambridge University Press. Koknar, Ali M. 2005. “Corsairs at Starboard: Jihad at Sea”, Journal of Counterterrorism & Homeland Security International, Vol. 11, No. 1, pp36–41. LaFree, Gary, Dugan, Laura and Korte, Raven. 2009. The impact of British counterterrorist strategies on political violence in Northern Ireland: Comparing deterrence and backlash models, Criminology, 47(17-45). Liss, Carolin. 2010. Oceans of crime: Maritime piracy and transnational security in Southeast Asia and Bangladesh. Institute of Southeast Asian Studies, Singapore. Loughran Thomas A., Pogarsky Greg, Piquero Alex R. and Paternoster Raymond. 2012. Re-examining the functional form of the certainty effect in deterrence theory. Justice Quaterly 29: 712-741 Low Desmond.2008. Global Maritime Partnership and the Prospects for Malacca Straits Security. Pointer, Journal of the Singapore Armed Forces, Vol. 34 No. 2 Maslow, A.H. and Murphy, Gardner. 1954. Motivation and personality, New York: Harper & Brothers.
32
Mueller, G.O.W. and Adler, Freda. 1985. Outlaws of the ocean. The complete book of contemporary crime of the high seas, Hearst Marine Books, New York Nagin, Daniel S. 1998. Criminal deterrence research at the outset of the twenty-first Century. In Michael Tonry (ed.), Crime and Justice: A Review of Research, vol. 23. Chicago: University of Chicago Press. Nevin, John A. 2003. Retaliating against terrorists. Behavior and Social Issues 12:109–28. Newman, Oscar. 1972. Defensible space; crime prevention through urban design. The Macmillan, New York. Ong-Webb, Graham Gerard. 2006. Piracy, Maritime Terrorism and Securing the Malacca Straits, jointly published by International Institute of Asian Studies, The Netherlands and Institute of Southeast Asian Studies, Singapore. Paternoster, Raymond.1987. The deterrent effect of the perceived certainty and severity of punishment: A review of the evidence and issues. Justice Quarterly 4: 173–217. Sherman, Lawrence W.1992. Attacking crime: Police and crime control. In Michael Tonry and Norval Morris (eds.), Modem Policing. Chicago: University of Chicago Press. Sherman, Lawrence W.1995. The police. In James Q. Wilson and Joan Petersilia (eds.), Crime. San Francisco: ICS Press. Siqueira Kevin and Sandler Todd. 2007. Terrorist backlash, terrorism mitigation, and policy delegation, Journal of Public Economics, 91(1800-1815). Marvell Thomas B. and Moody Carlisle E. 1996. Specification problems, police levels, and crime rates, Criminology 34(609-646). Reinhart, B. 2012. International Maritime Piracy and Armed Robbery, Civil-Military Fusion Centre. Available at www.cimicweb.org Schuman, Michael. 2009. ‘How to defeat pirates: Success in the Strait,’ Time.com. Available at http://www.time.com/time/world/article/0,8599,1893032,00.html Sparrow, Malcolm K., Mark H. Moore, and David Kennedy. 1990. Beyond 911: A New Era for Policing. New York: Basic Books.
33
Stuart, Robert. 2001. In Search of Pirates: A Modern Day Odyssey in the South China Sea, Edinburgh; Mainstream Publishing Edinburgh LTD. Tyler, Tom R. 1990. Why People Obey the Law: Procedural Justice, Legitimacy, and Compliance. New Haven: Yale University Press. Vagg, Jon. 1995. Rough Seas? Contemporary piracy in South East Asia, British Journal of Criminology, 35(63-80) Young, A. 2007. Contemporary Maritime Piracy in Southeast Asia: History, Causes and Remedies, jointly published by International Institute of Asian Studies, The Netherlands and Institute of Southeast Asian Studies, Singapore. Zimring, Franklin and Gordon Hawkins. 1973. Deterrence. Chicago: University of Chicago Press.
34
Figure 1. Piracy Attacks in Malacca Strait and South China Sea, 1995 to 2010.
-200
020
040
060
080
0%
cha
nge/
year
0
50
100
150
No.
of S
ucce
ssfu
l Atta
cks
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
Malacca Strait South China Sea% change (Malacca Strait) % change (South China Sea)
35
Figure 2. Successful Piracy Attacks in Malacca Strait and South China Sea, 1995 to 2010a.
a The attack is defined as successful if the pirates managed to obtain valuables (tins of paints, ropes, ship equipment, crews’ personal belongings, cash on board, etc), crews are wounded or hijacked, or ship is hijacked together with her cargo.
-200
020
040
0%
cha
nge/
year
0
20
40
60
80
100
No.
of S
ucce
ssfu
l Atta
cks
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Year
Malacca Strait South China Sea% change (Malacca Strait) % change (South China Sea)
36
Table 1. Descriptive Statistics, Piracy Attacks in Malacca Strait and South China Sea, 1995 to 2010
Malacca Strait
n=438 South China Sea
n=1678
Variable Mean Standard
Deviation Minimum
Value Maximum
Value Mean Standard
Deviation Minimum
Value Maximum
Value Days until next attack 11.817 22.652 1 289 4.178 4.143 1 37 Successful Attack 0.539 0.499 0 1 0.660 0.475 0 1 Military Intervention
MALSINDO 0.121 0.327 0 1 0.305 0.460 0 1 Other Events
CARAT 0.144 0.352 0 1 0.174 0.379 0 1 Indonesian Tsunami 0.041 0.199 0 1 During Ramadan 0.108 0.310 0 1 Hainan Incident 0.089 0.285 0 1
Context Lagged Regional Terrorist Incidents 3.906 6.535 0 44 17.477 17.929 0 93 Regional Trade 115389.474 34090.545 78943.284 241847.020 120146.335 54200.197 67595.484 244837.452 Regional Military Expenditure 1.058 0.252 0.792 1.469 1.832 0.116 1.637 2.037 Regional Misery Index 16.717 4.588 10.504 47.178 8.890 2.685 4.159 14.143 Success Density (omitted) (omitted) 0 ∞ (omitted) (omitted) 0 ∞ Last Success 0.541 0.499 0 1 0.660 0.474 0 1 Reported Attack 0.757 0.429 0 1 0.715 0.451 0 1 Tankers 0.307 0.462 0 1 0.230 0.421 0 1 Container and Cargo Ships 0.243 0.307 0 1 0.277 0.447 0 1 Non-Commercial Ships 0.270 0.444 0 1 0.159 0.365 0 1
37
Table 2. Coefficients and Robust Standard Errors for Series Hazard Models, Piracy Attacks in Malacca Strait, 1995 to 2010
All
n=428 Territorial Water
n=114 International Water
n=245 Port n=68
Coefficient Estimates
Robust Std. Erra.
Coefficient Estimates
Robust Std. Errb.
Coefficient Estimates
Robust Std. Errc.
Coefficient Estimates
Robust Std. Errd.
Military Intervention MALSINDO -1.506** 0.376 -2.871* 1.279 -1.869** 0.581 -0.063 0.974
Other Events CARAT 0.008 0.125 -0.480 0.338 0.146 0.186 0.506† 0.309 Indonesian Tsunami -0.659* 0.341 3.121** 0.923 -0.884† 0.468 -20.161 . During Ramadan 0.323* 0.157 -0.335 0.263 0.283 0.192 1.735** 0.638
Context Lagged Regional Terrorist Incidents 0.002 0.006 -0.018* 0.008 0.007 0.009 0.088** 0.016 Regional Trade 0.000** 0.000 0.000† 0.000 0.000† 0.000 0.000 0.000 Regional Military Expenditure -1.231** 0.330 -2.034** 0.772 -1.552** 0.440 0.654 0.899 Regional Misery Index -0.013 0.013 -0.026 0.019 -0.010 0.014 0.239** 0.074 Success Densitye 0.013** 0.001 0.021** 0.005 0.011** 0.001 0.021** 0.004 Reported Attack 0.313** 0.115 0.511* 0.233 0.387* 0.168 0.130 0.371 Tankers 0.096 0.143 0.513 0.352 -0.046 0.192 0.133 0.382 Container and Cargo Ships 0.300* 0.140 0.501 0.321 0.269 0.168 0.254 0.466 Non-Commercial Ships -0.204 0.144 0.351 0.327 -0.276 0.187 0.667 0.471
†p≤0.10 *p≤0.05 **p≤0.01, all one-tailed tests aStandard Error adjusted for 377 clusters, bStandard Error adjusted for 106 clusters, cStandard Error adjusted for 217 clusters, dStandard Error adjusted for 66 clusters,
e 𝑃𝑃(𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑓𝑓𝑓𝑓𝑠𝑠𝑐𝑐𝑐𝑐 𝑎𝑎𝑐𝑐𝑎𝑎 𝑐𝑐𝑡𝑡𝑓𝑓 𝑝𝑝𝑓𝑓𝑠𝑠𝑝𝑝𝑝𝑝𝑓𝑓𝑠𝑠𝑠𝑠 𝑎𝑎𝑐𝑐𝑐𝑐𝑠𝑠𝑎𝑎𝑝𝑝𝑐𝑐𝑠𝑠)(𝑠𝑠𝑝𝑝𝑠𝑠𝑐𝑐𝑐𝑐 𝑎𝑎𝑎𝑎𝑐𝑐𝑠𝑠𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐−𝑠𝑠𝑝𝑝𝑠𝑠𝑐𝑐𝑐𝑐 𝑎𝑎𝑎𝑎𝑐𝑐𝑠𝑠𝑠𝑠𝑐𝑐𝑐𝑐𝑠𝑠𝑐𝑐𝑠𝑠 𝑝𝑝𝑐𝑐𝑐𝑐𝑝𝑝𝑝𝑝𝑠𝑠𝑐𝑐𝑠𝑠)/365
38
Table 3. Coefficients and Robust Standard Errors for Series Hazard Models, Piracy Attacks in South China Sea, 1995 to 2010
All
n=1669 Territorial Water
n=744 International Water
n=349 Port
n=569 Coefficient
Estimates Robust
Std. Erra. Coefficient
Estimates Robust
Std. Errb. Coefficient Estimates
Robust Std. Errc.
Coefficient Estimates
Robust Std. Errd.
Military Intervention MALSINDO -0.310* 0.132 -0.377 0.239 -0.858** 0.304 -0.006 0.187
Other Events CARAT -0.126† 0.065 -0.143 0.100 -0.298* 0.141 -0.045 0.103 Hainan Incident 0.357** 0.098 0.529** 0.128 0.480** 0.172 -0.057 0.152
Context Lagged Regional Terrorist Incidents 0.001 0.002 -0.001 0.003 0.004 0.003 0.003 0.003 Regional Trade 0.000 0.000 0.000 0.000 0.000* 0.000 0.000 0.000 Regional Military Expenditure 0.500 0.329 0.635 0.483 -0.152 0.649 0.439 0.605 Regional Misery Index -0.016 0.013 -0.013 0.016 -0.049* 0.024 0.015 0.027 Success Densitye 0.010** 0.001 0.010** 0.001 0.009** 0.001 0.011** 0.001 Reported Attack -0.100 0.070 -0.107 0.093 -0.180 0.147 -0.081 0.162 Tankers -0.104† 0.058 -0.107 0.087 -0.092 0.124 -0.146 0.096 Container and Cargo Ships -0.101† 0.062 -0.149† 0.089 -0.109 0.133 -0.060 0.102 Non-Commercial Ships -0.065 0.070 -0.068 0.102 -0.129 0.140 0.050 0.188
†p≤0.10 *p≤0.05 **p≤0.01, all one-tailed tests aStandard Error adjusted for 1358 clusters, bStandard Error adjusted for 664 clusters, cStandard Error adjusted for 312 clusters, dStandard Error adjusted for 511 clusters,
e 𝑃𝑃(𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑓𝑓𝑓𝑓𝑠𝑠𝑐𝑐𝑐𝑐 𝑎𝑎𝑐𝑐𝑎𝑎 𝑐𝑐ℎ𝑓𝑓𝑠𝑠𝑠𝑠 𝑝𝑝𝑓𝑓𝑠𝑠𝑝𝑝𝑝𝑝𝑓𝑓𝑠𝑠𝑠𝑠 𝑎𝑎𝑐𝑐𝑐𝑐𝑠𝑠𝑎𝑎𝑝𝑝𝑐𝑐𝑠𝑠)(𝑠𝑠𝑝𝑝𝑠𝑠𝑐𝑐𝑐𝑐 𝑎𝑎𝑎𝑎𝑐𝑐𝑠𝑠𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐−𝑠𝑠𝑝𝑝𝑠𝑠𝑐𝑐𝑐𝑐 𝑎𝑎𝑎𝑎𝑐𝑐𝑠𝑠𝑐𝑐ℎ𝑝𝑝𝑐𝑐𝑠𝑠 𝑝𝑝𝑐𝑐𝑐𝑐𝑝𝑝𝑝𝑝𝑠𝑠𝑐𝑐𝑠𝑠)/365
39
Table 4. Odds Ratios and Robust Standard Errors for Logistic Models Estimating Success, Piracy Attacks in Malacca Strait, 1995 to 2010
All
n=437 Territorial Water
n=115 International Water
n=250 Port n=67
Odds Ratio
Robust Std. Erra.
Odds Ratio
Robust Std. Errb.
Odds Ratio
Robust Std. Errc.
Odds Ratio
Robust Std. Errd.
Military Intervention MALSINDO 3.884† 2.925 15.501 27.272 5.300† 5.387 0.404 0.712
Other Events CARAT 2.128* 0.711 8.413* 7.378 1.857 0.837 0.875 0.741 Indonesian Tsunami 0.493 0.357 1.000 (omitted) 0.365 0.346 1.000 (omitted) During Ramadan 0.585† 0.195 0.985 0.576 0.603 0.278 1.000 (omitted)
Context Lagged Regional Terrorist Incidents 0.986 0.019 0.994 0.033 0.968 0.019 1.015 0.052 Regional Trade 1.000** 0.000 1.000 0.000 1.000** 0.000 1.000 0.000 Regional Military Expenditure 1.907 1.160 3.324 4.117 1.937 1.660 0.326 0.549 Regional Misery Index 1.008 0.020 0.997 0.027 1.017 0.029 1.155† 0.126 Last Success 1.260 0.276 0.789 0.358 1.625 0.487 0.640 0.449 Reported Attack 1.898* 0.535 1.247 0.602 2.064† 0.874 5.308† 5.563 Tankers 3.057** 0.957 2.469 1.536 4.063** 1.876 1.224 0.996 Container and Cargo Ships 1.820† 0.605 2.513 1.7804 2.429* 1.089 0.388 0.418 Non-Commercial Ships 8.490** 3.104 8.433** 5.958 13.978** 7.048 2.200 2.391
†p≤0.10 *p≤0.05 **p≤0.01, all one-tailed tests aStandard Error adjusted for 378 clusters, bStandard Error adjusted for 106 clusters, cStandard Error adjusted for 219 clusters, dStandard Error adjusted for 65 clusters
40
Table 5. Odds Ratios and Robust Standard Errors for Logistic Models Estimating Success, Piracy Attacks in South China Sea, 1995 to 2010
All
n=1677 Territorial Water
n=750 International Water
n=351 Port
n=574 Odds
Ratio Robust
Std. Erra. Odds Ratio
Robust Std. Errb.
Odds Ratio
Robust Std. Errc.
Odds Ratio
Robust Std. Errd.
Military Intervention MALSINDO 0.967 0.256 1.912 1.019 0.470 0.275 0.899 0.361
Other Events CARAT 1.097 0.153 0.927 0.212 1.820† 0.623 0.834 0.193 Hainan Incident 0.494** 0.089 0.706 0.197 0.205** 0.109 0.429* 0.145
Context Regional Terrorist Incidents 0.999 0.004 1.002 0.007 1.004 0.008 0.993 0.007 Regional Trade 1.000 0.000 1.000 0.000 1.000 0.000 1.000 0.000 Regional Military Expenditure 1.023 0.154 0.845 0.213 0.579 0.202 1.747† 0.502 Regional Misery Index 1.049* 0.025 1.117** 0.034 0.974 0.049 1.053 0.059 Last Success 1.004 0.111 1.357† 0.222 0.722 0.180 0.762 0.157 Reported Attack 1.359* 0.183 1.244 0.228 2.420** 0.764 0.811 0.245 Tankers 1.192 0.169 1.355 0.302 2.074* 0.648 0.897 0.214 Container and Cargo Ships 1.147 0.154 1.494* 0.301 1.330 0.426 0.954 0.225 Non-Commercial Ships 2.234** 0.391 3.464** 0.912 2.458** 0.808 2.870* 1.480
†p≤0.10 *p≤0.05 **p≤0.01, all one-tailed tests aStandard Error adjusted for 1360 clusters, bStandard Error adjusted for 666 clusters, cStandard Error adjusted for 314 clusters, dStandard Error adjusted for 515 clusters
41