Methodology

The Myanmar Conflict Map’s dashboard allows users to visualise violent events reported in Myanmar since 1 July 2020. The underlying data is provided by the Armed Conflict Location & Event Data Project (ACLED), a global event-based conflict monitoring system.

Like other monitoring systems, ACLED aggregates conflict data pulled from open sources. For the Myanmar context, it draws primarily on international and local media, including Burmese and ethnic-language sources. A team of researchers regularly monitors these sources to collect information about individual events that take place in Myanmar.

Following a verification and review process, information about an event is ‘coded’ into the dataset according to the guidelines and definitions established by the ACLED Codebook. An entry in the dataset includes the event type and subtype (e.g., ‘battle’ and ‘armed clash’) as well as details associated with the event, such as time, location and the actors involved. An explanation of how the Myanmar Conflict Map reclassifies and displays the ACLED data is provided below.

Understanding event-based conflict data

Event-based conflict data provides a critical, albeit incomplete, picture of armed violence. The use case for conflict data can vary depending on the nature of the conflict, the scope and quality of information available, and the methodology employed by the monitoring system. While rigorous methodology can help ensure the best data quality possible, all conflict monitoring systems are prone to some bias and information gaps. Analysing conflict data therefore requires careful consideration and an understanding of its limitations.

The nature of armed conflict and violence changes over time as new actors, tactics, technologies or other variables emerge. The evolution of conflict dynamics may necessitate an adaptation of monitoring or coding practices. The rise of drone warfare, for example, not only changes battlefield dynamics but also introduces unique challenges for researchers attempting to document and analyse patterns of violence.

Because ACLED collects data on conflicts taking place around the world, its methodology necessarily takes account of a wide range of variations in context and dynamics. The advantage of this approach is that it allows researchers to compare different conflicts and evaluate trends in violence at a global level. However, it may not provide the same granularity as a dataset designed to take account of the unique dynamics of a single conflict, or what is known as a context-specific monitoring system.

Still, ACLED’s Myanmar dataset reliably captures broad conflict trends, like conflict acceleration and deceleration over months or years, as well as conflict frequency and distribution. Data tools on this site allow the user to visualise many of the conflict’s defining trends, including the rapid spread of violence following the 2021 coup d’état, the proliferation of localised militias or People’s Defense Forces, the newfound reliance on airpower by the Myanmar military (or Tatmadaw), and the use of violence against civilians. The ACLED data also captures key spatial patterns over time, such as the prevalence of violence near strategic objectives like roads, border crossings and resource-extraction sites.

The visualisation of quantitative conflict data, however, cannot convey every aspect of armed violence, irrespective of methodology or the quality of the underlying information. While trend-lines or maps can demonstrate the frequency and spatial distribution of conflict, they do not readily communicate the intensity or brutality of armed conflict, nor the psychological, societal or economic effects of that violence.

With that said, certain quantifiable variables can sometimes provide a proximate indication of qualitative dimensions of conflict. In Myanmar, for example, the duration of an armed clash is often suggestive of the intensity of the combat engagement. In any case, the utility of conflict data ultimately depends on both the quality of the methodology and source material, as well as how that data is interpreted and incorporated into broader analytical frameworks.

Robust conflict analysis should therefore take account of data limitations and always utilise a combination of quantitative and qualitative sources of information, analysed within the broader social, political, economic and historical context. For this reason, the Myanmar Conflict Map offers both data-visualisation tools and in-depth written analysis to help readers understand the general nature, scope, intensity and trajectory of armed violence in present-day Myanmar.

Reclassifying event types

ACLED defines 25 sub-event types and classifies them under six broader event types, both violent and non-violent. Definitions of sub-event and event types are available in the ACLED Codebook.

The Myanmar Conflict Map excludes sub-event types in two instances: where the event is non-violent (e.g., an agreement between two actors) and when a sub-event type is likely to have been significantly underreported. In particular, sexual violence and threats of sexual violence are widespread but vastly underreported. The research team has acknowledged the prevalence of sexual violence in its Warscape series.

The Myanmar Conflict Map reclassifies the remaining ACLED sub-event types into five event types, as per the table below.

Mapping geospatial conflict data

ACLED records the location (latitude, longitude and administrative unit) of all conflict events. The lowest-level administrative unit in Myanmar is the village (or in cities, the ward). Above this lies the township level, roughly akin to a county in the United Kingdom or United States. There are 330 townships in Myanmar, ranging in size from 5 square kilometres in cities to 12,000 km² in rural peripheries. Above this lies the district level, which usually includes several townships. At the highest administrative level there are seven states and seven regions.

Geospatial data is coded by ACLED according to three levels (or ‘codes’) of spatial precision. Code 1 is ascribed to events for which there is precise geospatial information available. In the Myanmar context, this means that open sources mention at least the village or ward where an event took place. In this case, the event is typically coded to the geographical centre of the village or ward, unless more specific information about the location is available, such as a landmark or address.

Code 2 denotes cases where open sources only reported an approximate event location. In the Myanmar context, events that happened either between two villages or somewhere within a given township are commonly categorised as Code 2. Events reported only at the township level are usually coded to the main town, unless other identifying information is provided, such as proximity to a road or natural feature. The precision range within Code 2 therefore varies.

ACLED ascribes Code 3 to an event in the rare instance where open-source reports have cited only the state or region where it took place, in which case the coordinates for the capital city are used. Of the events listed on the Myanmar Conflict Map, 56.5% are Code 1 while 43.4% are Code 2. Therefore, the inclusion of events classified as Code 3 has no meaningful impact on the overall quality of the conflict data.

The Violent Events Map allows users to display the conflict data in two ways. By selecting ‘True coordinates’, violent events are plotted according to the exact coordinates input by ACLED. Depending on the data parameters input by the user, the map will likely show some events coded to approximate locations (i.e., Code 2 events). The map may depict a concentration of events around a main town if numerous Code 2 events were plotted to those coordinates as an approximation, as per ACLED’s geospatial coding practices.

Despite this limitation, the ACLED data is on aggregate able to reliably capture the pattern of armed conflict over time and within a given area. This is especially true when examining patterns of violence in or around strategic areas and corridors.

Code 1 and 2 Precision

Code 1 Precision only

Alternatively, the user can display event data by selecting the ‘Township scatter’ option. The feature displays conflict data by assigning each event a randomised geolocation within the township in which it occurred. This mapping method better visualises the prevalence or frequency of armed violence across a given township but cannot capture more granular geospatial event patterns. In townships where conflict is highly concentrated in or around specific areas, randomised geodata may suggest that armed violence is more spatially widespread than it really is. For this reason, the map is set to ‘True coordinates’ by default.

Fatalities

Both the Violent Events Map and the Charting Tool allow the user to sort data by the number of fatalities associated with a single violent event. ACLED’s fatality estimates are based on a triangulation of the most reliable open sources available. Where variations occur or where fatalities are reported in vague terms (e.g., ‘dozens killed’) ACLED defers to a more conservative estimate. ACLED reviews and updates fatality estimates when new or better information becomes available. ACLED estimates the total number of deaths arising from a conflict event, but does not disaggregate the data based on which group inflicted or suffered the deaths. A detailed description of ACLED’s fatality coding practice and discussion of its limitations is available here.

Several methodological limitations inherent in open-source conflict monitoring are particularly pronounced in the Myanmar context. Most of the sources that meet ACLED’s reliability criteria demonstrate a strong bias in favour of forces opposed to the Tatmadaw. In turn, many of these same sources rely on claims made by opposition forces about the number of Tatmadaw soldiers killed during a conflict event. Opposition forces and their supporters have a strong incentive to exaggerate Tatmadaw casualties, and this information then makes its way into mainstream reports. As a result, figures originally reported as Tatmadaw deaths likely account for an inflated portion of the overall fatality estimate in Myanmar.

In contrast, opposition forces, especially ethnic armed organisations, regularly downplay or omit their own casualty figures when providing information to the media. Although pro-regime social-media outlets post large volumes of photographic and video evidence of opposition casualties, these sources do not meet ACLED’s reliability threshold and therefore are omitted from the dataset. Some pro-opposition media, including mainstream outlets that ACLED relies on, have demonstrated a tendency to omit or obscure the number of civilians killed by opposition forces. For these reasons, the total number of opposition fighters killed, as well as the total number of civilians killed by opposition forces, are likely underrepresented in ACLED’s Myanmar dataset.

Fatality estimation in Myanmar is subject to methodological and data limitations and therefore requires careful interpretation. Still, the aggregate fatality figure recorded by ACLED constitutes the most reliable estimate based on open sources that is publicly available. The overall fatality estimation since 2021 is broadly commensurate with qualitative reporting and analysis of the intensity and scale of the violence in Myanmar, and thus serves as a useful benchmark for comparison with other conflicts.

Areas of operation

The Areas of Operation tool allows the user to visualise the operational reach of select armed groups in Myanmar. The composite heat map is based on violent-event data only. It does not capture the presence of a group in areas where its activities are exclusively non-violent (i.e., administrative, economic or clandestine).

The tool allows the user to plot areas of operation over time. Spatial variation from year to year may indicate that a group’s operational reach has expanded or contracted, but this is not always the case. For example, if a group has fully captured a territory and fighting there has subsided, then there may be no violent-event data to represent that group’s continued presence. Stated differently, the absence of violent data could also reflect a consolidation of territorial control.

The Area of Operations map is best used to understand the general operational scope and reach of a given armed group by year since 2021. It does not attempt to measure territorial control, except in the cases of the United Wa State Army and National Democratic Alliance Army, which have maintained stable ceasefires and front lines since 1989. Brief descriptions of the territorial reach of each group are provided.