Lives lost: Xinjiang-related violence vs. mass incarceration (in numbers)

Police station following an attack on June 26, 2013, in Pichan County's Lukchun Municipality.
Police station following an attack on June 26, 2013, in Pichan County's Lukchun Municipality.

Note: This article is real-time adaptive, with many of the results and computations therein based on the latest available data, which is injected automatically each time you open the page. While the general structure and conclusions are unlikely to change much over time, the specific numbers will.

Update on September 9, 2024: This article was updated with more rigorous sentencing statistics, taking into additional consideration that a significant portion of court verdicts result in suspended sentences that are unlikely to lead to prison time.

A primary argument by which the Chinese authorities have justified the mass incarcerations and crackdowns in Xinjiang is that of “counter-terrorism”, claiming that widespread criminal prosecutions and extrajudicial camp detentions were a justified instrument to counter the violent “terrorism” and “religious extremism” that have long plagued the region (“separatism”, the third element that adds to these to form the “three evil forces“, has been less common in recent rhetoric).

These claims have often been polarizing: just as the Chinese government and its supporters are unlikely to publicly postulate that decades of repression and injustice could have contributed to violent expressions of resistance, so do many Uyghurs and their supporters outright reject the association of Uyghurs with anything that could objectively be defined as “terrorism” (often seeing in it a dangerous bridge to giving the Chinese authorities a carte blanche on their actions). Regardless of one’s position, there is not much disagreement regarding the existence of the violence itself, however, and the fact that much of the violence has been rooted in ethnic tensions and political dissatisfaction.

In an attempt to make better sense of things, this page seeks to maintain a detailed list of all the known (95) violent incidents related to Xinjiang since 2000 – specifically, those where civilian and/or police casualties (deaths or injuries) have been documented, and which cannot be attributed to more standard crimes (e.g., gang warfare, psychopathic behavior, drunken fights, or robbery). Additionally, the characteristics of these incidents will be brought together and summarized statistically, as a means of presenting a more complex picture that avoids common narratives. Finally, we use the obtained statistics to calculate an upper estimate on the deaths caused by the violence in Xinjiang, and then compare it to a lower estimate of the equivalent death toll from the mass-incarceration campaigns undertaken by the Chinese authorities in response to this violence.

List of violent incidents/clashes

Since many incidents are alleged but not reported, the list presented is not comprehensive, but strives to be. It will be updated whenever we learn of a new incident, and we encourage those with knowledge of such incidents to submit them to us (preferably with references and evidence). The details regarding specific cases may originate from various sources – from China’s state media to independent investigations by Radio Free Asia – and unfortunately must be viewed as approximate, very likely biased, and potentially unreliable. However, this is the nature of data – in Xinjiang especially – and we are happy to refine on a case-by-case basis to improve precision, while relying on averaging and statistics to provide some clarity for the general picture.

datelocationdescriptionnumber of attackerscivilian/police deaths target planned? ref.
2005
JAN 20Maytagh District, Karamay (approx.)Bus bombing that occurred on a bus traveling from Maytagh to Wusu, resulting in at least 12 deaths. Motivations and responsibility unclear.?12civilianlikely
2008
MAY 7flight from Urumqi to BeijingInflammable material was found in the toilet of a plane heading from Urumqi to Beijing, after which the plane made an emergency landing in Lanzhou, with 2 people being taken into custody. No further details were made available, though state media called the incident an “attempted terrorist attack”.?0civilianlikely
AUG 4Kashgar City2 men reportedly drove a truck into a group of 70 jogging police officers, proceeding to attack them with grenades and machetes, with 16 officers killed.216policelikely
AUG 10Kucha County15 Uyghurs attacked prominent government buildings, supermarkets, and hotels, with explosives and gunfire reported. A security guard was killed, as well as at least 10 of the suspects.151civilianlikely
AUG 12Yamanyar Township, Yengisheher CountyAssailant(s) jumped off a car at a checkpoint, stabbing and killing 3 police officers, and injuring one.?3police?
AUG 27Qizilboyi Township, Peyziwat CountyPolice were attacked while investigating a case and passing by a cornfield, with the attackers stabbing them from behind. 2 police officers died, while 5 were injured.?2police?
2009
JUL 5UrumqiArguably the largest and most symbolic incident in post-2000 Xinjiang, when protests by Uyghurs following an earlier incident at the factory in Shaoguan exploded into violent ethnic clashes and rioting around the city, with 197 deaths, mostly Han Chinese, officially reported. Retaliation by Han on July 7 has been alleged to have brought more (unreported) deaths, and intense crackdowns followed, with the internet being cut for close to a year. Thousands disappeared or were arrested throughout the year following the incident, with the fates of many still unknown. 100s or 1000s197mixedunlikely
2010
APR 26near Hal Village, Mughalla Municipality, Keriye CountyA clash took place between 2 Uyghur men and the local police when the men went to take down posters that they found unacceptable because of the partial nudity in them. 1 police officer was killed with a knife, with the attacker also killed. The other man fled to a friend’s home, with the friend being killed by police in the doorway and the man arrested and sentenced to life imprisonment. Other friends and relatives of the men were detained and threatened afterwards. 21policeunlikely
AUG 19Egerchi Township, Aksu CityAn attacker drove up to a patrol group of 1 auxiliary police officer and 15 members of a joint-defense team on a three-wheeler and threw an explosive, resulting in the immediate death of 5, later death of 2, and the injury of 14.17mixedlikely
2011
JUL 18Nurbagh Subdistrict, Hotan City18 Uyghur men occupied a police station in Nurbagh, killing 2 guards and taking 8 hostages, in what was allegedly a response to the government’s restrictions on burqas. A firefight ensued, with 14 attackers killed and 4 captured. 2 hostages were also killed during this time. The Turkistan Islamic Party claimed responsibility for the attack.184mixedlikely
JUL 30Kashgar City 2 Uyghur men hijacked a truck, killed the driver, and then drove the vehicle into a crowd of pedestrians, getting off and proceeding to attack people with knives, killing 6. One of the attackers was killed by the crowd and the other taken into custody. The Turkistan Islamic Party claimed responsibility for this attack and the one that took place a day later.27civilianlikely
JUL 31Kashgar CityA follow-up attack to the one from the day before. A group attacked a shopping mall, killing 2 people in a restaurant and 4 outside, in addition to injuring 15. Five of the suspects were shot dead by police, with two more killed later. 116civilianlikely
DEC 28Qoshtagh Village, Guma CountyA group of Uyghurs from Mokoyla Township, including women and children, were attempting to flee China by crossing the Pamirs (state media has portrayed them as a terrorist group seeking to receive training abroad). After police found them and tried to stop them, a clash occurred, with the police officer Adil Abduweli being killed, with 7 from the group killed thereafter, including 2 women. State media has also alleged that the group had taken 2 hostages, though this is inconsistent with what local police told Radio Free Asia during the latter’s investigation.?1policeunlikely
 2012
FEB 28urban area of Qaghiliq County A group of 9 people, led by Abdukerim Mamut, attacked pedestrians with axes and knives on Qaghiliq’s Happiness Road. 15 pedestrians and 1 police officer were killed, with 7 of the attackers killed by police and 2 captured.916civilianlikely
JUN 29flight between Hotan and Urumqi6 people attempted to hijack an airplane flying from Hotan to Urumqi, but were overpowered by the passengers and crew, being taken into custody after the plane returned to Hotan. There were some injuries but no deaths.60civilianlikely
SEP 7Mekit CountyNurmemet Yasin, a butcher from Mekit County, allegedly stabbed and killed a prostitute after she and other sex workers tried to pull him in. He also stabbed and injured her boss when the latter came at him, being arrested by police soon after.11civilianunlikely
 2013
MAR 7Golden Triangle area of Korla CityA fight broke out at a video arcade, allegedly between Uyghur and Han, resulting in the alleged deaths of 4, 13 people being injured, and the city being put under security lockdown.?4civilianunlikely
MAR 9Gujanbagh Township, Hotan CityA gasoline-bomb raid was allegedly carried out on a police station, without any details regarding the number of casualties being reported.?0policelikely
MAR 21Dighar Municipality, Pichan County, TurpanAround 150 residents marched in anger on Han residential areas in response to a young Uyghur boy being killed by a Han Chinese brick-kiln worker. Only some windows were allegedly damaged, with 10 of the leaders taken into custody but later released, and a case being opened against the worker.~1500civilianunlikely
APR 23Maralbeshi CountyA clash occurred between local authorities and residents, allegedly as a result of a local inspection group ordering a woman to remove her face veil. 21 people died as a result, though the exact identities are disputed (based on what has been reported by two different sources, 6 of the dead were among the alleged assailants).1415mixedunlikely
APR 25Yengi’awat Township, Hotan CountyA clash resulted in 2 community police personnel being killed.?2police?
MAY 9Uchar Municipality, Yengisar CountyAccording to local and official reports, a man named Alim Ebey was on the run (suspected of involvement in the April 2013 Maralbeshi incident), and killed 2 village officials during a house search. He himself was killed in retaliation, with his wife later dying in police custody.12civilianunlikely
MAY 20Qaraqash County5 locals armed with knives allegedly went to the Qaraqash River early in the morning and attacked a group of Han Chinese dam workers, killing 5 and injuring 4, with 2 of the injured later dying at the hospital.57civilianlikely
MAY 30Dadamtu Township, Ghulja City3 Uyghur high-school students allegedly caused the death of a Han classmate following a fight between students. The court, which sentenced the three to terms of 5-6 years, stated that they pelted stones at the Han student as he tried to cross a stream while fleeing from them, resulting in him drowning.31civilianunlikely
JUN 26Lukchun Municipality, Pichan County, TurpanAlleged rioting led to the deaths of 35, including 9 police and security guards, with 11 attackers being shot dead by police.1724mixedlikely
JUN 28HotanAn unstated number of Uyghurs, displeased with religious restrictions, allegedly traveled from Han’eriq Township to Tuanjie Square (in Hotan City, most likely), where they rioted and started fires on a pedestrian street. Police allegedly killed 2 people in the incident.many0civilianunlikely
JUN 28Qashboyi Neighborhood, Uchturpan CountyA religious affairs worker accompanied by two police officers asked an Uyghur man to shave his beard, which resulted in a conflict and the man stabbing the worker and one of the police officers, before being shot dead. No other deaths were reported.10policeunlikely
JUN 30Atush City, Kizilsu PrefectureA man stabbed and killed a police officer during a house search, in addition to stabbing and injuring another, before being shot dead.11policeunlikely
AUG 7Aykol Municipality, Aksu CityPolice were allegedly on the way to a mosque to make arrests, and were met with an angry mob hurling stones and bricks. Reinforcements, including SWAT, were called in, opening fire and killing at least 3 from the mob and injuring many others. 10 police officers were reported to be wounded.?0policeunlikely
AUG 14Turpan CityA pro-government imam was stabbed and killed in front of his home, with three suspects being taken into custody.31civilianlikely
AUG 20Yilqichi Township, Qaghiliq CountyA house raid by authorities resulted in a confrontation where at least 15 were shot dead, with 1 police officer also being killed.?1policeunlikely
OCT 28Tiananmen Square, BeijingA car with an Uyghur man, his wife, and his mother ran over pedestrians and crashed, bursting into flames. All 3 people inside were killed, as were 2 tourists, with 38 people reported as injured. 3 other people were later executed for allegedly planning the attack.32civilianlikely
NOV 16Maralbeshi County9 people armed with knives and axes attacked a police station, killing 2 auxiliary police officers and injuring 2 others.92policelikely
DEC 15Saybagh Township, Konasheher CountyViolent clashes broke out while police were allegedly on their way to make arrests, leading to the deaths of 2 police officers and to police shooting 14 people dead. Some of the people involved would be executed later, while in prison.?2policeunlikely
DEC 30Yarkand CountyMen armed with knives and explosives allegedly attacked a police station, with 8 people being shot dead. It is not clear if there were any police casualties.?0policelikely
 2014
JAN 15Yengi’eriq Township, Awat County, Aksu3 men approached a police station and asked to meet the chief, attacking the auxiliary police outside when the latter asked to search them. 2 of the police officers were injured, 1 severely, and the 3 men were shot dead. 30policelikely
JAN 22Ghaldir Village, Qaratal Municipality, Aksu CityA village Party secretary was stabbed to death in his home by 2 men, who were allegedly taking revenge for friends who had been detained.21civilianlikely
JAN 24Toqsu County, AksuA group of at least 17 attacked police with explosives, with 6 being shot dead by police, 6 dying when their explosives went off, and 5 being taken into custody. One police officer was slightly injured.17+0policelikely
FEB 14Uchturpan County, AksuA police patrol unit was attacked by men on motorbikes, wearing explosive devices and carrying knives. 2 civilians and 2 police officers were injured, while 8 attackers were shot dead, 1 was captured, and 3 blew themselves up.120mixedlikely
MAR 1Kunming Train Station, Kunming City, Yunnan Province6 men and 2 women armed with knives attacked civilians at random in the Kunming Train Station, killing 31 and wounding 143 others.831civilianlikely
APR 4Qarasay Municipality, Qaraqash CountyA “terrorist bombing” incident with at least 1 person involved, specifics not clear.????
APR 14Araghol Village, Laskuy Municipality, Hotan CityThe relatives of 2 women allegedly got into a fight with 2 police officers, when the latter demanded that the women take off their hijabs. The police officers were then stabbed dead in the confrontation.?2policeunlikely
APR 27Janggili’eski Township, Qaghiliq County4 senior officials were on a fishing expedition at Kokkolyar Lake, with 3 of them disappearing and later found to be brutally murdered. Over 150 people were interrogated as part of the investigations. The killings took place on the first day of Xi Jinping’s visit to the region.?3civilianlikely
APR 29Qarasay Municipality, Qaraqash CountyA bombing incident with at least 14 people sentenced to prison after (1 for “abetting criminals”, 13 for “separatism”), specifics not clear.13+???
APR 30Urumqi Train Station, Urumqi2 men carried out a bomb-and-knife attack at the Urumqi Train Station, both dying in a suicide bombing. 1 bystander was killed, while 79 were wounded.21civilianlikely
MAY 13Muji Municipality, Guma CountyA group of 3 men attacked a police station with explosives in the middle of the night, injuring 2 police officers. The police then started a hunt for the suspects, shooting 2 dead and arresting the third.30policelikely
MAY 20Alaqagha Municipality, Kucha CountyHundreds of people protested in front of government buildings because of the detention of several women and middle school girls for wearing headscarves, with the protest turning violent and protesters beating the school principal and a township official, in addition to pelting stones at the buildings. Up to four people are believed to have been killed after special police opened fire on protesters.100s (protesters)0civilianunlikely
MAY 22UrumqiAttackers drove two cars and threw explosives into shoppers at a vegetable market, killing 31 and injuring over 90. A later report by Radio Free Asia specified the number of deaths as 39, noting that it was carried out by 5 suicide bombers.539civilianlikely
MAY 23Yarkand CountyA bomb went off in front of the county public security bureau, with a police vehicle reported as damaged but no reported casualties. Other bombs were later disarmed.?0policelikely
MAY 29Towen Tarsha Village, Misha Township, Peyziwat CountyFour attackers allegedly staged a motorcycle accident to lure out a high-ranked police officer, whom they then proceeded to stab to death.41policelikely
MAY 29Qaratash Village, Ich’eriq Municipality, Toqsu CountyA police raid on an alleged “separatist” hideout resulted in the stabbing and death of 1 police officer, with 1 suspect and his brother being shot dead. A third person present was later sentenced.31policeunlikely
JUN 4Seleychong Village, Lengger Municipality, Konasheher CountyVillage officials allegedly angered local residents after lifting a woman’s veil, prompting her 2 brothers to attack the officials with sticks and knives and causing them to flee, while calling the police. A confrontation followed, during which 4 people were shot and killed, with 1 police officer being killed as well.41policeunlikely
JUN 14Qumqisar Village, Qumqisar Township, Mekit CountyA suspect already in custody, believed to be responsible for planning attacks on police stations, was being taken to his home by 8 police officers to help identify the instruments of the crime. When left with two of the officers, he managed to secure a weapon and stab one of them, before being shot dead. The stabbed officer died, while another was mistakenly shot dead by the remaining police.11policeunlikely
JUN 15Hotan CityAt least 3 attackers assaulted people with knives in a crowded chess hall, injuring 4, prior to being stopped by police and some of the chess players. Two of the attackers died from their injuries, while a third was arrested.3+0civilianlikely
JUN 20Ishqul Township, Yarkand CountyA suspect attacked police while being released from his second interrogation at the township police station, stabbing 1 officer to death and seriously injuring another.11policeunlikely
JUN 20Qayash Village, Manglay Township, Qaraqash County2 police officers were stabbed and killed at a checkpoint in the early morning hours, with assailants setting on fire the outpost with 3 others while the latter were sleeping, leading to them being burned alive.?5policelikely
JUN 21Qaghiliq CountyA large group of attackers allegedly rammed a car into the local public security bureau and detonated explosives, with police shooting 13 of the attackers dead. Three police officers received slight injuries.13+0policelikely
JUN 27Aksu CityAn Uyghur man stabbed several people on a pedestrian street, killing 1 and injuring 3.11civilianlikely
JUL 9Towenki Mekit Village, Imamlirim Municipality, Uchturpan County6 Han Chinese farmers were stabbed to death, with a woman being hospitalized with stab wounds. The murders were believed to be linked to similar murders in Aksu City and Uchturpan’s Aqtoqay Township. 7 suspects were identified, with 3 of them arrested and 1 shot dead.76civilianlikely
JUL 12Aksu City5 Han Chinese traders were tricked into stopping and getting out of their car, after which they were brutally murdered by 5-6 people. The murder was linked to the later murder of a judicial official.5-65civilianlikely
JUL 18Jigdilik Village, Aqtoqay Township, Uchturpan CountyA judicial official was murdered outside the mosque in Jigdilik Village, being stabbed to death, with 2 people identified as suspects.21civilianlikely
JUL 18Beshkent Municipality, Yarkand CountyA police officer was injured during a confrontation with a family, allegedly as a result of police screening women wearing headscarves. The family was allegedly shot dead as a result, with the incident potentially contributing to the larger incident on July 28.?0policeunlikely
JUL 19Purchaqchi Township, Qaraqash CountyAssailants armed with knives and axes broke into the home of the local Party secretary, killing his wife and significantly injuring him. The attack was seen as revenge for a mosque raid that had taken place earlier.?1civilianlikely
JUL 28Yarkand CountyArguably the largest incident since the 2009 clashes in Urumqi, with various sources outside of official China media reporting hundreds or thousands dead as a result of a harsh response by authorities to local protests. State media has blamed the attack on terrorists, claiming that 37 innocent people died.?37mixed?
JUL 30Kashgar CityJume Tahir, a pro-Party imam of Kashgar’s Id Kah Mosque, was stabbed to death by 3 Uyghur men.31civilianlikely
AUG 1Purchaqchi Township, Qaraqash CountyPolice, allegedly assisted by 30000 volunteers (presumably very indirectly), tracked down suspected “terrorists” to an abandoned house near a corn field, with the group resisting arrest and throwing explosives at pursuers. No one was injured, while 9 suspects were shot dead and 1 was arrested.10+0policeunlikely
AUG 3Ustunki Yarbagh Village, Aqtoqay Township, Uchturpan County2 Uyghur suspects on the run killed 2 Han farmers. They were part of the group believed responsible for the attacks on July 9 and July 12.22civilianlikely
AUG 3Yakowruk Township, Uchturpan CountyThe 2 Uyghur suspects of the incident above killed the deputy director of a brick factory.21civilianlikely
AUG 4Toqquzaq Village, Yakowruk Township, Uchturpan CountyThe 2 Uyghur suspects of the 2 incidents above were stopped by 3 local farmers, who had been recruited by the authorities to help with the search. One of the farmers was wearing a guard uniform, and was stabbed to death by the suspects, while the other two were spared.21civilianunlikely
SEP 21Bugur CountyA series of bomb and knife attacks left 6 civilians, 2 police officers, and 2 auxiliary police officers dead, with another 54 civilians injured, according to state media. 40 “rioters” were killed in the violence, and another 2 arrested.42+10mixedlikely
OCT 10Guma County2 Uyghurs carried out multiple knife attacks, stabbing to death 3 police officers and 3 government officials before being killed themselves.26mixedlikely
OCT 12Maralbeshi County4 Uyghurs attacked a farmers’ market with knives and explosives, killing a total of 18 civilians and police officers before being gunned down.418mixedlikely
NOV 28Yarkand CountyA large group attacked a food street with knives and explosives, killing 4 and wounding 14. 11 of the attackers were shot dead.11+4civilianlikely
2015 
JAN 12Yengisheher Municipality, Yengisheher CountyA suspect with an explosive attacked police with an axe at an urban commercial center, being shot dead. Police then discovered 5 more people with explosives, shooting them dead also. No police or civilians were injured.60civilianlikely
JAN 28Lengger Township, Keriye County3 Uyghurs were passing through an inspection point and got into a fight with the police, stabbing and killing 3 officers and security guards. 2 of them were later shot dead by a SWAT unit, with the third being caught later.33policeunlikely
FEB 13Guma CountyAn attacker set off a bomb, killing at least 7 people and injuring at least 7 others.17?likely
FEB 16Bash Quduqla Village, Purchaqchi Township, Qaraqash CountyA man drew a knife in response to police allegedly being rough with him during a check, prompting them to shoot him dead. No casualties were reported among the police.10policeunlikely
FEB 16Qaghiliq Municipality, Qaghiliq CountyA father and two sons set fire to a police vehicle close to the county’s No. 1 Primary School. Some police appear to have been injured.30policelikely
FEB 17Yaqa’eriq Municipality, Bay CountyPolice were carrying out house searches when they entered a house with 10 men, who proceeded to attack them with knives and axes, killing 4 police officers. 9 of the attackers were killed by police, with 4 civilians allegedly shot dead by police by accident.104policeunlikely
MAR 6Toqsun CountyA clash between Uyghurs and police led to the death of 1 police officer and 5 Uyghurs (not clear if assailants or civilians).?1police?
MAR 8Tagharchi Municipality, Yarkand CountyA group of 7 Uyghurs attacked a commander of the local armed forces, his wife, and her uncle, stabbing them to death, in addition to killing a security guard who tried to help. All 7 were shot to death when police arrived.74mixedlikely
MAR 9Jume Bazar Village, Purchaqchi Municipality, Qaraqash CountyAt least 7 Uyghurs were subjected to a police search while in a restaurant, resisting and attacking the police, with 1 police officer being stabbed to death. Two of the men were shot dead, two were arrested, and three fled.7+1policeunlikely
MAR 11Xinshi District, Urumqi2 young Uyghur men attacked Han migrant laborers with knives, injuring three. One of the attackers was shot dead, while the other was captured.20civilian?
MAR 12Shamalbagh Municipality, Kashgar CityA group of 6 Uyghurs attacked people outside a local casino, injuring eight and putting two of them in critical condition, in addition to damaging vehicles. Four of the six attackers were shot dead by police.60civilianlikely
MAY 11Lop Municipality, Lop CountyTwo bomb attacks were carried out on a police checkpoint, the first in the evening of May 11 and the second in the morning of May 12. A total of 3 police officers were killed and 4 more were injured, with 3 suicide bombers also dying in the incidents.33policelikely
MAY 21Layqa Township, Hotan CountyThe deputy head of the township was taking a 19-year-old Uyghur youth to the local police station when the latter pulled out a knife and stabbed him, causing the deputy head to die in the station soon after. The youth was later detained.11civilianunlikely
MAY 25Zawa Municipality, Qaraqash CountyA group of 6 young men attacked a local police patrol with bombs, with 2 of the men being shot dead and 4 fleeing. No police casualties have been reported.60policelikely
JUN 19Yengisheher County4 young attackers stabbed and killed a police officer at a checkpoint. (Date approximate, and corresponds to the reported date when the city was put on high alert.)41police?
JUN 22Tahtakowruk Village, Nezerbagh Municipality, Kashgar CityA police checkpoint was attacked with knives and bombs by a number of people who were allegedly from the same “religious family”, leading to the death of 3 police officers and 16 of the attackers.16+3policelikely
JUL 13Shenyang City, Liaoning ProvincePolice were allegedly attacked by 4 people wielding knives when they attempted a raid on “Xinjiang terrorist suspects”, retreating and calling for backup, which resulted in a major operation where 16 suspects were detained and one woman was injured. No police casualties were reported, with the local report later deleted.40policeunlikely
SEP 7Zhengzhou City, Henan ProvinceA shootout between police and “Xinjiang Uyghur extremists” allegedly took place, with a police officer and an Uyghur alleged to have died. Authorities were quick to delete the information from social media.?1police?
SEP 18Soghan Coal Mine, Bay CountyA group attacked the coal mine with knives, killing a number of security guards, police officers, and others, with at least 50 deaths reported.9+50mixedlikely
OCT 13AksuThis incident followed the earlier one above, when the Aksu Prefecture Public Security Bureau’s Deputy Director, Memetjan Tohniyaz, took some police officers and herdsmen to search mountain caves where the suspects were believed to be hiding. The herdsmen were captured by the 7-8 suspects, with Memetjan allegedly confronting them and sacrificing himself, being killed in their place. He was later declared a martyr by the authorities.7-81mixedunlikely
 2016
DEC 28Qaraqash County3 assailants drove a vehicle into the courtyard of a government building, setting off an explosive and attacking people with knives. An official and a security guard were killed, 3 others were injured, while the 3 attackers were shot dead.32civilianlikely
2017
FEB 14Guma Municipality, Guma County3 young Uyghurs attacked civilians with knives, killing 5 and injuring 5 more, with one of the latter dying in the hospital. The three attackers were shot dead. According to the local sources, it is unlikely that the attack was planned, and appears to have been a result of frustration following the authorities pressuring the family of two of the attackers to stop praying.36civilianunlikely

Statistical summary

As is readily seen, a great number of incidents were reported in 2013-2015, centered around the start of the “People’s War on Terror” in May 2014. It is not clear how much of this is due to reporting gaps or bias and how much is linked directly to the policies (i.e., a rise in incidents in 2013 prompting the declaration of the “War”, a spike in incidents following as the initial reaction and opposition in 2014, and prolonged opposition also leading to more incidents in 2015).

In terms of whom the attacks targeted, one sees an almost even split for the documented incidents – with attacks purely on civilians versus attacks purely on police having almost equal portions, with around 1/7 of the incidents targeting or involving a mix of police and civilians (at times, unarmed government staff). This appears to underline the complexity of the issue.

One also sees that the majority of the documented incidents (around 2/3) were likely premeditated, suggesting some degree of organization and planning. A third, however, appear to have been spontaneous in nature, with common cited causes being pent-up frustrations or police searches/checks, which then led to violence and clashes.

Specifically, premeditated attacks were more common in the incidents targeting civilians, with 28 of 38 (73.68%) likely being planned. For attacks targeting police, only 15 of 34 (44.12%) appear to have been planned, suggesting a higher incidence of clashes initiated by police checks, searches, or raids.

Finally, a frequency chart of the reported civilian/police death counts shows that the great majority of the documented incidents had few to no (0-5) reported victims (73 of 93, or 78.49%), with no deaths reported in 25 (26.88%) of the incidents. Large incidents – such as the July 5, 2009 clashes (197 reported deaths), the Kunming Train Station attack (31 reported deaths), the Urumqi vegetable market attack (39 reported deaths), and the Yarkand County clashes (37 deaths reported by state media, not including authorities’ actions against locals) – are outliers in this sense, and cannot be seen as representative of the majority based on the reports available.

Comparison with equivalent death toll from mass incarcerations

As an extension, we can attempt an extreme-case analysis, in which the Chinese authorities are given the benefit of the doubt, and in which we compare the maximum possible loss of life resulting directly from the violence in Xinjiang to the minimum possible loss of life resulting from the authorities’ “counter-terrorism” campaigns of mass incarceration. In so doing, we will show that the loss of life from mass incarceration is at least many times that from the alleged violence, even under assumptions that are intentionally favorable to the Chinese authorities.

Officially, Chinese authorities have withheld the total number of violent and/or “terrorist” incidents and victims, but have quantified the number of attacks as “several thousand” in official publications and statements. For our purposes, let us interpret this conservatively as “no more than 5000”.

A brute approach would be to take the average reported deaths per incident (6.53) for the documented incidents and to multiply it by 5000, yielding an upper estimate of 32650. However, the 6.53 average is corrupted by the aforementioned outliers, and is almost certainly much higher than the actual.

Instead, it is assumed that almost all of the alleged thousands of incidents were small in scale (otherwise, the news would have broken out and been reported). Consequently, in approximating the average deaths per incident that would apply to the typical incident, we can remove “outlier” events with high reported death counts. What constitutes an outlier is subjective, but objective heuristics exist to identify them: here, we will use the interquartile method to identify those incidents whose counts are particularly high. For the reported data set, this method results in outliers being those events with reported death counts above 10, which are then left out in computing the average.

This then yields an average believed to be more representative: 1.85 deaths/incident. Applying this average to the upper estimate of 5000 yields an upper estimate of civilian/police deaths for the non-outlier incidents of 9250. Finally, adding the death counts from the outlier incidents (those with over 10 reported deaths), we get an additional 455, for a total upper estimate of 9705.

We now move to compare this upper estimate to a lower estimate on the loss of human life caused by the mass incarcerations carried out by the Chinese authorities in Xinjiang, especially in 2017 and after. Because places of detention are typically not execution grounds or “death camps”, the loss of life is generally not from direct killing but from the long years spent in incarceration. This amount is significant, but is not directly comparable to deaths that are the direct result of violent incidents because of inherently different time scales and mechanisms.

To make comparison possible, the following conversion is proposed here:

1 life = 72 years in detention

This is based on the average life expectancy in the region. It goes without saying that this conversion is imperfect and could be refined – it is too uniform and leaves out important nuances – but is believed to be sufficient for a rudimentary comparison.

To estimate the total time spent in detention in recent years, it is helpful to consider the total as a sum of the times spent in various types of detention, as they all function differently. The vast majority of the detention in Xinjiang has occurred in three forms – camps, prisons, and pre-trial detention centers – with the total detention time conservatively underestimated as:

D > D_C + D_P + D_K

where D is the total time spent in detention in Xinjiang by all those detained on political, religious, or ethnic grounds (a direct target of the “counter-terrorism” campaigns) in the recent two decades. D_C, D_P, and D_K are then the corresponding times in detention for camps, prisons, and pre-trial detention centers, respectively. That this is an underestimate and not an equality follows from there being other (lesser documented) forms of detention as well, such as strict house arrest, off-the-record police custody, residential surveillance at a designated location, hospitalization under guard, or “retention” (for officials). For simplicity, we ignore these types here, as they are not believed to contribute significantly to the total detention time.

Noting that time spent in detention can be expressed as the average period/term (T) multiplied by the number of people detained that way (N), we can rewrite the summation as:

D > T_C N_C + T_P N_P + T_K N_K

It is possible to estimate most of these elements from the available data and statistics. However, pre-trial detention presents an additional nuance: in addition to the data for this detention type being fairly limited, it often acts as a precursor to the other detention types, but sometimes does not. Specifically, a significant number of the N_C people sent to camp were held in pre-trial detention before camp transfer, as were essentially all of those N_P people sent on to prison. On the other hand, some were released after a certain amount of time in pre-trial detention (without being sentenced or sent to camp).

To account for all these scenarios, we break the pre-trial detention component into three:

T_K N_K = T_{KC} N_{KC} + T_{KP} N_{KP} + T_{KK} N_{KK}

where T_{KC}, T_{KP}, and T_{KK} are the average times spent in pre-trial detention for those who were later transferred to camp, those who were later transferred to prison, and those who were later released, respectively. Likewise, N_{KC}, N_{KP}, and N_{KK} are the total numbers of people held in pre-trial detention on political/religious/ethnic grounds who would later be transferred to camp, sent to prison, or released, respectively.

Since essentially all prisoners go through pre-trial detention, we can remove N_{KP} by noting that N_{KP} = N_P, and plug the result into the overall underestimate to get:

D > T_C N_C + T_P N_P + T_{KC} N_{KC} + T_{KP} N_P + T_{KK} N_{KK}

or

D > T_C N_C + (T_P + T_{KP}) N_P + T_{KC} N_{KC} + T_{KK} N_{KK}

At the time of writing, data for T_{KC}, N_{KC}, T_{KK}, N_{KK} are extremely limited, making any sort of reliable estimation difficult and most likely not worth the endeavor. As such, we will ignore these terms and set them to 0, which adds conservatism but does not conflict with our goal of obtaining an underestimate:

D > T_C N_C + (T_P + T_{KP}) N_P

For the terms that remain, we can start by noting that T_P+T_{KP} is simply the average prison term for the N_P prisoners, since the prison term given by the court includes the time spent in pre-trial detention before transfer to prison (i.e., it is the sum of the time spent in pre-trial detention and the time spent in the actual prison). The current data from the Xinjiang Victims Database allows us to approximate this as being around 8.53 years.

Estimating the average “camp term”, T_C, is tougher, given the camps’ extrajudicial nature and the terms not being set or declared (or even known) when the person is first sent there. However, as with the prison terms, we may rely on available data in the Xinjiang Victims Database and consider the reported time spent in camp for those who have been released and for whom both the date of being taken to camp and the date of being released from camp are known. While the sample size here is significantly smaller (114), it is believed to be large enough to provide a reasonable estimate.

Furthermore, the estimated average (266.78 days, or 0.73 years) is believed to be conservative and an underestimate, owing to the following biases:

  • A significant portion of the data is from Urumqi police records from 2017 and 2018, and many of those whose detention and camp release dates are reported therein are those who were released early, usually for health reasons (general mass releases did not start until late 2018). Consequently, their terms are shorter than that of the average representative camp detainee.
  • Another significant portion comes from Kazakh camp detainees in northern Xinjiang, a number of them eyewitnesses who reported their exact detention and release dates. While not “early releases”, they are “late detainees”, as camp detentions in northern Xinjiang did not really start on a great scale until late 2017, as opposed to early 2017 in the south, with some evidence also suggesting that mass releases took longer in the south also. Consequently, their camp terms ended up being shorter than the typical southerner’s by virtue of the detention trends varying by geography.

Since we are, again, searching for an underestimate, these points bolster the validity of the available estimated average camp term, as it is almost certainly lower than the true average.

Considering these two values, we can then rewrite the general underestimate as:

D > 0.73 N_C + 8.53 N_P

which leaves us with the task of estimating the detainee numbers.

For those sentenced to prison, previous analysis has yielded an estimate of at least 834517 people sentenced in the period from 2007 to 2022, as based on comparison and modeling of official statistics from the courts and procuratorates. For simplicity, we can leave out the other years, for which we do not have estimates, and work with this number.

Two issues need to be addressed before this number can be used, however:

  • the number represents the total sentenced and includes those not targeted as part of the “counter-terrorism” campaigns;
  • the number includes those sentenced but not given detention terms, and who would not contribute to the total detention time.

To address the first issue, we start by noting that the most significant differences between imprisonment in Xinjiang and imprisonment elsewhere in China (with the likely exception of Tibet) are due to the “counter-terrorism” initiatives. Therefore, a reasonable means to separate one group from the other is by taking the total sentenced and subtracting the amounts corresponding to the standard outside of Xinjiang.

The necessary statistics are compiled below:

yeareffective sentences
(national)
effective sentences
(Xinjiang)
effective sentences
(non-XJ)
population (national)population
(Xinjiang)
population
(non-XJ)
per-1000 rate
(non-XJ)
non-XJ rate applied to Xinjiang
2007933156176989154581.321B20.95M*1.300B0.7014665
20081008677168079928701.328B21.31M*1.307B0.7616153
2009997872106139872591.334B21.59M*1.313B0.7516193
2010100741999639984561.341B21.82M*1.319B0.7616583
201110516381560110360371.349B22.25M1.327B0.7817424
201211741331689711572361.359B22.53M1.336B0.8719602
201311586091643811421711.367B22.85M1.344B0.8519426
201411845622156211630001.376B23.25M1.353B0.8619995
201512326953534011973551.383B23.85M1.359B0.8820972
201612206453109411895511.392B24.28M1.368B0.8721134
201712701419978711703541.400B24.80M1.375B0.8521184
20181430091133198**12968931.405B25.20M1.380B0.9423688
20191661235100491**15607441.410B25.59M1.384B1.1328931
2020152803453092**14749421.412B25.90M1.386B1.0627413
2021171592244969**16709531.413B25.89M1.387B1.2031068
2022143158538497**13930881.412B25.87M1.386B1.0126176
Total340607
* Previously reported population numbers were modified between 2020 and 2021, resulting in older populations being raised by approximately 200000 to 300000 from their earlier reported numbers. The latest reported numbers are used throughout for 2011-2022, until the cutoff point of 2010, for which only the data from older reports are available.
** The numbers of effective sentences for 2019-2022 in Xinjiang are not available, with either reported or estimated sentence numbers being used instead. This is not expected to influence the estimates in the final column significantly, however.

Considering this data, we see that the number of sentenced individuals that would be expected for Xinjiang if the “normal” national rate (excluding Xinjiang) held is 340607. Subtracting this from the estimated 834517 sentenced yields an estimate of 834517 – 340607 = 493910 people sentenced above average in 2007-2022, as a proxy for those sentenced as part of the “counter-terrorism” campaigns.

To account for sentences that do not result in prison time, we start by reviewing the various types of sentences that a court can give (as seen from statistical breakdowns in law yearbooks):

  • innocent (无罪),
  • no criminal responsibility (不负刑事责任),
  • spared from criminal punishment (免予刑事处罚),
  • prison term of at least 5 years, including capital punishment (5年以上至死刑),
  • prison term of more than 3 years and less than 5 years (超过3年不满5年),
  • prison term of at least 1 year and no more than 3 years (1年以上3年以下),
  • prison term of less than 1 year (不满1年),
  • short detention (拘役, usually 1 to 6 months),
  • suspended sentence (缓刑),
  • surveillance (管制),
  • accessory penalty (单处附加刑).

The table below, providing the national statistics for 2017, gives an idea of the proportions:

While all of these contribute to the sentencing numbers, we are only interested in considering the categories that result in prison time.

The first 3 categories (innocent, not criminally responsible, spared from criminal punishment) do not. Suspended sentences may sometimes, but are so rarely reported in practice that we do not believe them to contribute significantly to the total count (among those taken in “counter-terrorism” campaigns), and may exclude them also. Accessory penalty cases, which typically don’t result in detention time, may be excluded as well. Both short detention and surveillance terms have been documented, however, with those sentenced to “surveillance” in Xinjiang often being sent to “rectification bureaus” (矫治局) typically intended for drug rehabilitation.

Consequently, we only consider the

  • prison term of at least 5 years, including capital punishment (5年以上至死刑),
  • prison term of more than 3 years and less than 5 years (超过3年不满5年),
  • prison term of at least 1 year and no more than 3 years (1年以上3年以下),
  • prison term of less than 1 year (不满1年),
  • short detention (拘役, usually 1 to 6 months),
  • surveillance (管制)

types as relevant, and would like a lower estimate on how many of those in the sentenced statistics belong to these categories.

Taking the national level data for 2007-2022 (using the same sources as the national effective sentences in the table above), we can obtain lower estimates on the portions of sentences that result in detention time:

yeareffective sentences resulting in detention timeeffective sentences not necessarily resulting in detention timeestimated lower percentage of sentences resulting in detention time
200766397626918071.14%
200871343429524370.74%
200970525429261870.67%
201070080330661669.56%
201170004435159466.55%
201277552839860566.06%
201375721140139865.34%
201477245141211165.19%
201582606040663567.02%
201680942341122266.31%
201787683739330469.03%
2018100117842891370.02%
2019121796544327073.32%
2020110468542334972.31%
2021125584946007373.18%
2022102027641130971.28%

From these statistics, we can estimate that at least 65% of sentences lead to detention time on the national level, with the number expected to be higher in Xinjiang and in the “counter-terrorism” campaigns specifically (where sentences are more strict). Applying this to the 493910 people sentenced above average, we make the following statement regarding the total number of “counter-terrorism” prison terms:

N_P > 0.65 \times 493910 > 321041

Note that there is additional conservatism in this number as it excludes the years 2023 and after.

The overall estimate then becomes:

D > 0.73 N_C + 8.53 \times 321041

No official data are available to estimate the number of camp detainees, N_C, but it is possible to estimate it from the sentence numbers by observing the empirical ratio between the two. For 30 villages and neighborhoods where full or close to full data are available, with camp and sentenced numbers also reported (see the “detention rates” section in our primary evidence), the ratio of people reported as taken to camp versus those sentenced to prison is 0.91. Consequently, if S is the number of people who would be reported as sentenced as of 2017-2018, the number of those in camp can be estimated as:

N_C \approx 0.91S

At the time of writing, we do not have an estimate on the sentenced population for 2017-2018 specifically, and instead may estimate S by simpler means. A conservative approach stems from using official data to estimate the portion of prison sentences that are 5 years or longer for previous years.

Year% of sentences resulting in prison terms of 5 years or longerlower estimate on individuals sentenced
201312.92%19456
201424.34%27689
201526.59%41596
201634.95%39948
201787.24%127861
2018n/a133198
Percentages of longer prison terms are obtained from the statistics on effective sentences as reported in the Xinjiang Yearbook. Lower estimates on individuals sentenced have been previously obtained by consolidating various sentencing data from different sources.

On first glance, these data would imply that 12.92% of those sentenced in 2013 would still be reported as sentenced (and contribute to S) in 2017-2018, with analogous statements for 2014, 2015, 2016, and 2017. More accurately, however, it is important to account for time in pre-trial detention, which takes place before a sentence but contributes to it. If everyone spent a year in pre-trial detention, for example, then those sentenced to 5 years in 2013 would actually be released in 2017 (and not 2018). To account for this, we assume 1 year of pre-trial detention as the upper limit (empirical data supports this to be true for the vast majority of detainees), and drop the 2013 data point from consideration. The deductions that continue to hold are:

  • at least 24.34% of those sentenced in 2014 would still be reported as sentenced in 2018 (even accounting for a maximum of a year in pre-trial detention);
  • at least 26.59% of those sentenced in 2015 would still be reported as sentenced in 2018;
  • at least 34.95% of those sentenced in 2016 would still be reported as sentenced in 2018;
  • at least 87.24% of those sentenced in 2017 would still be reported as sentenced in 2018;
  • approximately everyone sentenced in 2018 would still be reported as sentenced in 2018 (mostly a tautology).

Strictly speaking, these statements are still very approximate in nature (even the last one), since the documents used to estimate the camp-to-sentenced ratio come from various points in 2017-2018, and a finer analysis would be needed to account for potential time lags (a matter of months). The percentages of those sentenced to 5 years or longer are also based on reported effective sentences, which are believed to be incomplete and also suffer from slight time lag because of potential appeal processes. However, as a first approximation and in the interest of simplicity, the statements above are believed to be largely sufficient.

Equipped with these statements, we can obtain a lower estimate on S as

S > 0.2434 \times 27689 + 0.2659 \times 41596 + 0.3495 \times 39948 + 0.8724 \times 127861 + 133198 > 276505

which finally lets us get a lower estimate on N_C:

N_C > 0.91 \times 276505 > 251619

and allows us to finalize the lower estimate on total detention years as:

D > 0.73\times 251619 + 8.53 \times 321041 > 2922161

Using the “1 life = 72 detention years” conversion, this then allows us to quantify the “equivalent death toll” from the “counter-terrorism” mass incarceration as being at least 40585 lives.

It is important to reiterate that these estimates are very conservative. As more data become available, both with regard to the violent attacks and with regard to the mass incarcerations, the disparity between the two is only expected to grow.


Sign up to be notified of new posts