The Myanmar
Conflict Monitor

1Executive summary

Following the 1 February 2021 coup, the conflict in Myanmar rapidly transitioned from a period of high mass participation to one defined by lethal crackdowns. MCARI recorded 4,085 events during this period, resulting in at least 796 fatalities and 6,233 detentions. The distribution of fatalities was highly concentrated, rather than being spread evenly across the record. Over 51.3% of all deaths occurred on just five days, with 176 people killed on 27 March alone. Lethal force was concentrated in March.

Measuring the quarter's severity solely by fatalities obscures the primary mechanism of the conflict. Arrests constituted the overwhelming majority of recorded impacts, outpacing fatalities by a factor of 7.8 (6,233 arrests compared to 796 deaths). Events were recorded in 268 of Myanmar's 330 townships.

The conflict is characterized by two distinct layers. The first involves protests and crackdowns in urban areas, where 553 out of 561 deaths were civilians. The second layer involves armed conflict in border regions, where 142 of the 218 deaths were military and pro-coup forces. While in urban areas nearly all of the dead were civilians, border regions saw a steady rate of engagement between established armed groups.

2The quarter in review

February saw 1,940 events, the majority of which were peaceful demonstrations and strikes. While 76 people died in February, the situation escalated sharply in March. Although the event count moved only slightly to 2,145, the death toll increased ninefold to 720. A major turning point occurred on 14 March in Hlaingtharya, Yangon, where 49 people were killed in a single night. On 27 March, Armed Forces Day, 176 people died across 41 townships, 135 of whom were civilians.

In the armed conflict layer, KIO/KIA and KNU/KNLA engaged in clashes with the military throughout the period. In Kachin State, 100 deaths were recorded, with only 14 being civilians. The end of March also marked the use of airstrikes, signifying a further escalation. During the quarter, 44,116 recorded displacement events occurred, with over 95% taking place in March.

By late March, the record also captured the earliest signs of localized armed resistance. Residents in Kale and Tamu in Sagaing Region, and Gangaw in Magway Region, began defending themselves against military crackdowns. Where the record names their weapons they are locally made: airguns and percussion rifles in Kale, hunting rifles in Gangaw.

3Scale and composition

Most of what MCARI recorded this quarter was not lethal violence (Table 1).

Assessing conflict intensity merely by counting events is misleading, as lethality varies drastically by incident type. Violence against civilians demonstrated the highest lethality rate: 52.8 of every 100 events recorded a death (Table 1's 147.5 counts deaths per 100 events, not events with a death). In contrast, armed conflict yielded a death in 18 of every 100 events, and coercion in 0.4. MCARI records a death at a protest as the security forces' attack on the protesters (Violence against Civilians), not as part of the protest, so the protest category itself carries no deaths. An event count is not a measure of violence.

Protests and civil unrest accounted for 72.0% of the record, and coercion, including arrests and detention, accounted for 12.1%. Together, these categories make up 84.2% of all events but carry only 0.3% of the fatalities. In contrast, violence against civilians and armed conflict represent only 13.8% of the record but account for 97.6% of the total death toll. The most marked shift between February and March was the drop in protest activity from 87.0% to 58.5% of the record, while violence against civilians rose from 1.9% to 16.0% (Figure 1).

While protests significantly decreased both in absolute volume and as a proportion of total events, dropping from 1,688 (87.0%) in February to 1,254 (58.5%) in March, the overall number of recorded events expanded from 1,940 to 2,145. This reduction of 434 protest events did not indicate a de-escalation. Instead, the resulting gap was entirely offset by a surge in crackdowns and arrests, changing the composition of recorded events during this period.

4Fatalities

Fatalities this quarter were highly concentrated by date (Figure 2).

The daily volume of events does not dictate the scale of fatalities. The surge in deaths during March was not driven by a mere increase in the number of incidents, but by a sharp rise in the lethality of each one.

This disconnect is evident when comparing peak days: February 9 recorded the second-highest event count of the quarter (125 incidents) but resulted in only a single fatality. By contrast, March 27, the quarter's highest (129 incidents), saw a nearly identical event volume but yielded 176 fatalities. Overall, the statistical correlation between daily event frequencies and fatalities is weak (a correlation of 0.23).

The rise is measurable week by week. In the first week of March, 7.0% of events where the Myanmar Military is the recorded acting party resulted in a fatality. By the fourth week (March 22–28) that rate had reached 40.2%, a 5.7-fold rise in the likelihood that an event with the military as acting party recorded a death.

Of the 796 people recorded killed, 76.1% (606) were civilians. Among the dead, 160 were military and pro-coup personnel and 12 were members of armed opposition groups. Civilian deaths rose from 59 in February to 547 in March, a ninefold increase that demonstrates the severity of the crackdowns (Table 2).

Although total fatalities escalated ninefold between February and March, the demographic profile of the victims remained static. Civilians accounted for 77.6% of the recorded victims in February and 76.0% in March. The volume of violence rose ninefold; the civilian share of the dead did not change.

Relying on a statistical average obscures the true nature of lethal violence during this period. While the data shows an average of 0.19 fatalities per event, this figure does not reflect actual ground realities. The distribution of lethality was highly skewed: the vast majority of incidents, 3,842 events representing 94.1% of the total, resulted in zero fatalities. Conversely, lethal violence was extremely concentrated, with a mere 20 events accounting for 40.8% of the total death toll. Fatalities were not spread across the conflict. They came through a tiny fraction of mass-casualty incidents occurring alongside a massive volume of non-lethal events.

The 14 March crackdown in Hlaingtharya was the deadliest single event with 49 civilian deaths. 27 March was the deadliest day, with 176 fatalities across 41 townships, including 135 civilians and 40 military and pro-coup personnel killed in various parts of the country (Table 3).

An analysis of peak fatality days reveals two distinct operational patterns driving lethal violence. March 27, which recorded 176 fatalities across 41 townships, shows a nationwide, simultaneous application of lethal force, with armed conflict accounting for only 25.0% of that day's deaths. March 14 sits at the same end of the scale, with 97 fatalities across 14 townships and no armed conflict at all.

Conversely, March 30 recorded 43 fatalities concentrated in just 11 townships, driven overwhelmingly by armed conflict (72.1% of the day's deaths). The two profiles answer different questions: days like March 27 demonstrate the geographic breadth of coordinated crackdowns, while days like March 30 highlight the acute lethality of localized armed engagements.

The deadliest events were split between attacks on civilians and armed engagements (Table 4).

An analysis of the top 10 most lethal events reveals a clear dichotomy in the conflict, effectively capturing two distinct operational theaters within a single dataset. The demographic impact exhibits absolutely no middle ground. Four of these events, security forces' attacks on protesters recorded as attacks on civilians, resulted in 110 fatalities, all of whom were civilians.

Conversely, the remaining six events, which comprised armed clashes and airstrikes, produced 130 fatalities with zero recorded civilian deaths. This complete polarization indicates that the table is tracking two parallel forms of violence: unilateral actions against unarmed populations and active combat engagements between armed actors.

5Geography

The conflict map shows a clear divergence between where events occurred and where people died. While events were recorded in 268 townships, deaths occurred in only 106. In 162 townships (60.4%), no deaths were recorded (Map 1).

When interpreting the map, the size of the circles, representing event volume, naturally draws the eye, but the color gradient indicating fatalities conveys the actual severity of the violence. A high density of events does not inherently equate to high lethality. The top 20 townships by event volume, the largest circles on the map, accounted for just 22.7% of the total fatalities. In contrast, the top 20 townships by fatality volume concentrated 54.6% of the death toll. The locations where incidents occur most frequently are not the locations where they are most lethal.

For example, Taikkyi in Yangon recorded 82 events but zero deaths, whereas Kyaukkyi in Bago-East recorded only 9 events but 30 deaths. Fatalities were heavily clustered in specific areas, such as Hlaingtharya and Dagon Myothit (South) in Yangon, Hpakant and Mohnyin in Kachin, and Myingyan and Chanmyathazi in Mandalay (Table 5).

Comparing the top 10 townships by event frequency against the top 10 by fatality volume reveals a sharp geographic disconnect, almost as if analyzing two separate countries. Only two townships, Hpakant and Monywa, appear on both lists, meaning these two rankings encompass 18 distinct townships. The lethal risk associated with individual events varies drastically between these areas: the lethality rate in the fatality-heavy townships (0.85 deaths per event) is 4.4 times higher than in the event-heavy townships (0.19).

6Actors

Although 29 distinct acting parties were recorded, 94.2% of events (3,849) involved only three groups: protesters, the Myanmar military, and the police (Table 6).

Analyzing the actors involved reveals two key points. First, protesters participated in the most events (2,360) but were not linked to a single civilian death. That follows from how the record is built: a death at a protest is recorded against the party that attacked the protesters, not on the protest row.

Second, the military and police appeared in a similar number of events (763 and 726), but the military was far more lethal. Civilian deaths occurred in 24.0% of military events, 8.7 times the rate in police events (2.8%). Grouping them together under a single "security forces" label hides how much more often a civilian death is recorded where the military is the acting party.

Of the 606 civilians killed, 93% (566) died in events where the acting party was either the Myanmar military or the police. In the armed conflict layer, KIO/KIA, RCSS/SSA-S and KNU/KNLA remained the primary actors (Annex C, Parties on armed conflict events).

The data reveals two concurrent but distinct layers of violence during the reporting period. While protests and subsequent crackdowns dominated the overall volume, recording 3,817 events across all 59 days, armed conflict ran alongside them throughout. Armed conflict accounted for merely 4.5% of total events (184), yet they were recorded on 47 of the 59 days, about four days in five, at fewer than four a day. This was not a sporadic reaction to the coup: the first armed conflict event is recorded on 4 February. Despite its minimal share of the event volume, this layer drove 27.4% of all recorded fatalities.

Three formations created after the coup (Generation Z Power, Hta-Naung Task Force, and YURF) enter the record in the last week of March, each on a single event announcing its formation, and none on an event carrying a death.

7Detention, displacement and other indicators

Security forces detained 6,233 people, almost eight times the number of recorded deaths.

Regarding displacement, 44,116 recorded displacement events occurred, 95% of them in March.

While 24,251 releases were recorded, 23,314 of these were part of the 12 February general amnesty for pre-coup prisoners. Excluding this amnesty, only 937 releases were recorded against 6,233 new detentions (Table 8).

The four indicators moved in different directions. Arrests experienced a gradual 31% escalation (from 2,702 to 3,531), while the volume of displacement exploded by a factor of 20 (from 2,104 to 42,012). Conversely, recorded releases seemingly collapsed by 99%.

However, the February release figure (23,948) is skewed by a single mass amnesty on February 12, which accounted for 23,314 individuals incarcerated prior to the reporting period. Adjusting for this anomaly, the quarter recorded 6,233 arrests against just 937 relevant releases. Individuals are continuously absorbed into detention with virtually no exit mechanism.

The record holds 13 defections in Q1 2021, none in February and all of them in March (Table 8). Of the 13, 12 were soldiers who joined KNU/KNLA Brigade 5 at Hpapun on 2 March. Without them, the quarter records 1 defection: a police officer in Pyay who joined the civil disobedience movement.

Police officers also crossed into India during the same weeks, from Sagaing Region at the end of February and from Chin State through the first half of March. According to the reports, they left rather than take orders from the military. The sources name no force that received them, so MCARI records these crossings as flight and does not count them as defections.

The figure of 13 is a floor. The record cannot say how many personnel left their posts without a report reaching the sources MCARI reads.

Detentions did not always occur where the killing was highest; Tamwe in Yangon recorded 443 arrests but only 3 deaths (Figure 4).

The geography of arrests represents a third spatial pattern, largely separate from where people died, and sharing four of its ten townships with the event ranking. While the top 10 townships for arrests account for 37.9% of the national volume (2,361 out of 6,233 arrests), they recorded only 75 fatalities, a mere 9.4% of the total death toll. Furthermore, only one township from this list overlaps with the top 10 fatalities list. Tamwe shows this divergence: it recorded the highest arrest volume nationwide (443) but only three fatalities. The arrest map and the fatality map have almost nothing in common. The arrest map and the event map overlap in four townships out of ten.

The data also reveals a shift in detention and suppression tactics. While February arrests primarily targeted political figures and election officials, March saw a pivot toward striking civil servants, health workers, journalists, and neighborhood watch members to dismantle the civil disobedience movement. Furthermore, violence extended beyond daytime protests; security forces increasingly employed nocturnal terror tactics, conducting night raids, firing indiscriminately into residential wards, and destroying civilian barricades to instill fear.

8What this period does not show

Displacement is the indicator most clearly under-recorded as an event. Recorded displacement events stand at 44,116, and the whole of the figure sits on 28 rows. Of those, 6 are coded as forced displacement in its own right; the other 22 are armed clashes, shelling, attacks on civilians, airstrikes, arson and a landmine, where a source also reported people fleeing. MCARI records the figure on the act that drove people out rather than opening a separate row. By volume, the largest share, 26,000 of the 44,116, sits on the six forced-displacement rows. But two airstrikes account for 3.6 times the recorded displacement of all eleven ground clashes combined.

Geographic ambiguity presents a notable challenge in tracking arrests. For 168 events, sources provided only a broader location, such as "Yangon", "Mandalay" or "Naypyitaw", without identifying the specific township. These events represent just 4% of total recorded incidents but account for 12% of the total arrest volume, so arrests concentrate in the events the record places least precisely. Every township figure in Figure 4 is therefore a floor for the place named, and the order between them is not secure.

The recorded total of 796 fatalities is a floor rather than a count. In 91 events the sources did not say whether anyone was killed; these carry no deaths. The affiliation of 18 recorded fatalities remains unknown, precluding their categorization as either armed actors or civilians. How far above 796 the true figure lies is not knowable from the record.

Eleven of the 37 categories in the MCARI taxonomy recorded zero events. Sexual violence is the most prominent case, with zero recorded instances in February and March 2021. This most likely results from a lack of reporting and the difficulty of documentation during the initial crisis rather than an absence of such occurrences.

9Method in brief

The fatality total of 796 is a minimum-confirmed figure.

These figures are computed from the MCARI 2021 Edition (v1.0) dataset, based on reports from national, exile, and ethnic media outlets. In 91 events the sources did not say whether anyone was killed; these carry no deaths. Recorded displacement events (44,116) track the number of times displacement was reported, not the number of distinct individuals.

Q1 2021 covers 1 February – 31 March 2021. MCARI's record begins on the day of the coup.