Coding decisions on the Nagorno-Karabakh conflict, South Ossetia, Abkhazia, and more

Published on: 2 March 2023 | Last updated: 29 November 2023

The Central Asia and the Caucasus regions (referred to as CAC collectively) are made up of eight now-independent former Soviet republics spanning across a large area of central Eurasia. Despite geographically expanding across a large and culturally diverse region, the states of the CAC share a number of political and institutional similarities including the ongoing effects of their sudden independence from the Soviet Union. These effects include, among others, armed border conflicts, corruption, varying levels of authoritarianism, and both internal and external pressure to reform.

The CAC is an important economic and geopolitical region. Central Asia has significant economic potential in the form of vast energy and mineral reserves and is therefore vulnerable to how swift changes in markets can impact domestic stability. A drop in commodity prices in 2014 demonstrated a need for the region to reform and diversify its economy (RUJEC, 9 September 2019). Meanwhile, the CAC region borders a number of powerful states with mixed diplomatic allegiances, nuclear capabilities, and ideological insurgencies, namely Turkey, Russia, Iran, Afghanistan, and China. These factors have made it a target of external interference, for better or for worse.

Countries within the CAC region operate with varying levels of authoritarianism. According to the Economist Intelligence Unit (EIU) 2018 Democracy Index, Azerbaijan and Turkmenistan ranked very low in pluralism, civil liberties, functioning of government, political participation, and political culture. The other CAC nations ranked either as authoritarian or hybrid regimes, with Georgia, Armenia, and Kyrgyzstan generally offering their populations more political freedom (The Economist, 2018). Similarly, journalistic freedom is fairly limited across the region, with Azerbaijan and Turkmenistan ranking amongst the lowest in the world (Reporters Without Borders, 2019). While ceasefire violations make up the majority of events in ACLED’s CAC dataset, riots and protests have been increasing across the region, particularly in countries with less authoritarian regimes where demonstrations are a means of discourse in which local populations feel they can engage. Demonstrations typically center around demands for increased political, economic, and institutional reform, or against the interference of border nations, such as China and Russia (NY Times, 22 June 2019).

ACLED’s coverage of political violence and demonstrations in CAC captures a number of ongoing border conflicts and ‘de facto’ states. The most active of these is the continuing violence between Armenia and Azerbaijan, which engage in daily ceasefire violations along three fronts – two of which border de facto or autonomous states (Artsakh and Nakhchivan, respectively). Conflicts with breakaway states (South Ossetia and Abkhazia) are present in nearby Georgia to a lesser degree, and are essentially limited to non-violent ‘tensions’ since the Russo-Georgian war ended in 2008. In

Central Asia, several areas of the Fergana Valley – where Kyrgyzstan, Tajikistan, and Uzbekistan meet – are also disputed; however, those conflicts are most often fought amongst local border communities.

2018 included the Armenian Revolution: a series of peaceful protests and marches led by now-President Nikol Pashinyan in response to actions by the ruling Republic Party and then-President Serzh Sargsyan. Following the resignation of the latter, the Pashinyan government began a number of reforms and crackdowns on corruption, although change has been slow (Eurasianet, 7 May 2019).

How does ACLED deal with de facto states?

The CAC region includes a number of seperatist states which exist nominally within the borders of an internationally-recognized country, but which have become institutionally independent. Moreover, these states have their own armed forces which operate like other state forces and assert de facto control over the territory in which they are active. Therefore, while not recognized by a majority of states internationally, these regions generally meet the requirements for statehood according to declarative theory.¹ Declarative theory is one of two leading theories which outline the circumstances under which a state should be recognized as sovereign. The conditions are: a) a defined territory; b) a permanent population; c) a government; and d) a capacity to enter into relations with other states (for example, see: Organization of American States).Nevertheless, ACLED chooses to recognize only the boundaries of states which have significant international recognition; therefore, these are the names represented within the ‘Country’ variable designation in the dataset. However, ACLED codes the armed forces of states which assert de facto control of a space with an interaction term of “1” (State Forces) as the actions of such actors tend to mimic the actions of other state forces active in their home regions. Coding actors as such does not denote legitimacy but rather acknowledges the fact that distinct governing authorities co-exist and may exercise de facto control over different portions of territory in a country. In these cases, ACLED codes the de facto state forces using the name of the separatist territory. For example, elsewhere in the ACLED dataset, Somaliland’s military is coded as the actor “Police Forces of Somaliland (2010-)”.

The following is a list of de facto states within the CAC region:

The Republic of Artsakh:

Also known as the Nagorno-Karabakh Republic, it is a breakaway state existing within the borders of Azerbaijan. Formerly the Nagorno-Karabakh Autonomous Oblast, it declared independence from the Azerbaijan Soviet Socialist Republic (SSR) in 1991 following a referendum. This led to the three-year Nagorno-Karabakh War, which ended with a ceasefire. The majority population of Artsakh are ethnically Armenian, and were supported by Armenia during the war with Azerbaijan.

Armenia continues to support the independence of Artsakh, although official Armenian military deployment within the territory seems to be rare.

For areas under the control of Artsakh, ACLED uses the internal administrative divisions of the de facto state rather than those of the de jure state (Azerbaijan), given that sources will most often reference the former when attributing an event location. Currently, the Republic of Artsakh divides itself into the following Admin 1 regions: Shahumyan, Martakert, Askeran, Hadrut, Martuni, Shushi, Kashatagh, Stepanakert. These regions roughly correspond to the following official Admin 1 regions of Azerbaijan: All of the Kalbajar, Lachin, Qubadli, Zangilan, Jabrayil, Shusha, Khojali districts, Khankendi city, Shusha city, as well as parts of the Tartar, Agdam, Khojavend and Fizuli districts.

The government and armed forces of Artsakh are represented by the following actors:

• Government of Artsakh (2007-)
• Military Forces of Artsakh (2007-)
• Police Forces of Artsakh (2007-)

Republic of South Ossetia – the State of Alania:

South Ossetia is a breakaway state existing nominally within the borders of Georgia, and recognized by Russia and several other states. Similarly to Artsakh, South Ossetia declared independence from the Georgian SSR in 1991. This declaration led to a series of wars between Georgia and South Ossetia / Russia, the last of which being the five-day Russo-Georgian War in 2008. Many consider South Ossetia to be under Russian military occupation (Belgium UN, 8 August 2019).

For areas under the control of South Ossetia, ACLED uses the internal administrative divisions of the de facto state rather than those of the de jure state (Georgia), given that sources will most often reference the former when attributing an event location. Currently, the Republic of South Ossetia – the State of Alania divides itself into the following Admin 1 regions: Dzau, Leningor, Tskhinvali, and Znaur districts. These regions roughly correspond to the following official Admin 1 regions of Georgia: Small parts of the Racha-Lechkhumi and Kvemo Svaneti, Imereti, Mtskheta-Mtianeti regions, and the northern half of Shida Kartli region.

The government and armed forces of South Ossetia are represented by the following actors:

• Government of South Ossetia (2017-)
• Military Forces of South Ossetia (2017-)
• Police Forces of South Ossetia (2017-)
  

Republic of Abkhazia:

Abkhazia is another breakaway state nominally within northwestern Georgia, which first declared independence in 1992, leading to a war where it was also backed by Russian forces. Abkhazia is named after its primary ethnic group, the Abkhazians; the region experienced a large-scale ethnic cleansing of Georgians following the conflict.

Unlike South Ossetia, Georgia initially recognized the autonomy of Abkhazia, but not its independence. An official pro-Georgian ‘Government of the Autonomous Republic of Abkhazia’ was established to govern the region. However, following fighting with separatist groups, the pro-Georgian government was exiled to Tbilisi, where it remains today. For this reason, Georgia has an official Admin 1 region known as the Autonomous Republic of Abkhazia, which outlines the borders of the de facto state.

The government and armed forces of Abkhazia are represented by the following actors:

• Government of Abkhazia (2014-)
• Military Forces of Abkhazia (2014-)
• Police Forces of Abkhazia (2014-)

De facto state police “kidnapping”:

People living in contested regions (South Ossetia, Abkhazia, and Artsakh) are often taken into custody by the local de facto state police for crossing the dividing line. Sources tend to refer to these events as “kidnappings”; however, ACLED codes these events with the ‘Sub-event type’ of “Arrests” given the de facto legitimacy of these state forces within their own territory. This is in line with the way in which ACLED codes arrests by legitimate state forces across the dataset (see page 15 of the ACLED Codebook). As such, most of these “kidnappings” will not be represented in the dataset as they do not meet the threshold of being significantly ‘strategic’ to be included under the ‘Event type’ of ‘Strategic developments’ — those events which can impact trends in disorder.

How does ACLED code ceasefire violations between Armenia/Artsakh and Azerbaijan?

Ceasefire violations along multiple fronts have been ongoing since the end of the Nagorno-Karabakh War and the signing of the Bishkek Protocol in 1994. However, while there are a large number of daily violations reported (an average of over 21 as of 19 October 2019), the lethality of these events is exceptionally low, about two fatalities reported each month. The two main sources that cover the vast majority of ceasefire violence are the Ministry of Defence of Azerbaijan and the Ministry of Foreign Affairs of Armenia.

Covering official reports from both states allows ACLED to triangulate the information and to form a more complete picture. The following notes explain how each circumstance is sourced:

Firing from Armenia/Artsakh into Azerbaijani-controlled territory

The primary source for this type of event is the Azerbaijan Ministry of Defense (MoD) website, which updates each day with ceasefire violations from the previous day. These events are sourced using a script and translation tool, described in the Appendix at the end of this document.²Note that the Azerbaijan MoD stopped reporting violations into Nakhchivan Autonomous Republic specifically around 2017.

In addition, traditional and new media sources also report on ceasefire events, though these sources tend to largely duplicate the information from the MoD, except in the following cases:

• they report on fatalities or injuries, and/or damage to civilian housing.
• they report events which are more detailed than those of the MoD (i.e: “the ceasefire was violated at X date and X location”). Examples include descriptions of drone attacks, infiltration missions, larger clashes, or the use of heavier weaponry such as missiles and tanks.
• they report events which might not be covered by MoD reports, such as firing into Nakhchivan.

Data from these sources which meet the above criteria are cross-checked with MoD events and additional information is added to coverage of the event where applicable.

Firing from Azerbaijan into Armenian/Artsakhi-controlled territory

On a monthly basis (occasionally with a one or two month delay) the United Nations Security Council (UNSC) releases a document submitted by the Armenian Ministry of Foreign Affairs (MFA) containing a detailed table of ceasefire violations.

As is the case when coding firing from Armenia/Artsakh into Azerbaijani-controlled territory, traditional and new media are used to supplement data with more detailed coverage. Basic weekly coverage of these supplemental sources is also used to cover the frontline in real-time per ACLED’s publication schedule in order to ensure the delay in UNSC releases impacts trends in the data as minimally as possible. Per standard ACLED processes, a duplication check with the published data is done before submitting new information from UNSC reports.

The Armenian MoD also has a weekly report of the number of violations along the frontline, however these reports are often too vague to be coded as they lack details.

General assumptions and decisions when coding frontline violence between Armenia/Artsakh and Azerbaijan

Actors:

• For ceasefire violations along the frontline between Armenia and Azerbaijan, the default assumption is that the event is an interaction between the Military Forces of Armenia and the Military Forces of Azerbaijan.
• For ceasefire violations along the frontline between Azerbaijan proper and territory controlled by the Republic of Artsakh, the default assumption is that the event is an interaction between the Military Forces of Azerbaijan and Military Forces of Artsakh.
  • Azerbaijan reports by default claim that they are being fired upon by ‘the Military Forces of Armenia’ as they do not recognize the forces of Artsakh. However, per ACLED methodology, Artsakh forces operate independently in that region – albeit with substantial aid from Armenia – and only in exceptional cases are the ‘Military Forces of Armenia’ actually operating directly from Artsakh-controlled territory.

In case of more specific reports, such as the targeting of civilians or other groups, the above assumptions are adjusted accordingly.

Locations:

• Some reports refer to “positions on unnamed hills in the X region.” A combination of satellite imagery and known military positions from records has been used to determine the most likely position of these hills in reference to the region given and established front line. The ‘Location’ variable in the dataset is coded in such cases as Nameless Hills (X Region), with a geoprecision of 2, which references a specific set of geo-coordinates.
• Similarly, some reports refer to non-specific “positions along the line of contact” (LoC) in a region. We code these as a single event located at a predetermined point in Martuni province, on the Artsakh side, along the frontline (LoC) between Artsakh and Azerbaijan proper, with a geoprecision of 3. The Location name used for these events is Artsakh-Azerbaijan LoC (Unspecified).

If additional, more specific, incidents are mentioned within the same reports which outline firing on vague locations (hills; unspecified location), these are added as separate events in the data.

• Reports of an event containing both the origin point and the target of fire for the same incident are coded as a single event at the target’s location. These reports typically cover the northern Armenia-Azerbaijan frontline.
• Some reports mention only the origin location of fire, and not the location that was targeted. This happens in particular with Azerbaijan MoD reports referring to fire from Artsakh controlled territory. In such cases, the event is coded not at the origin location, but at a derived target location. This target location is coded at a nearby named location, such as a town, on the direct opposite side of the frontline – i.e. in Azerbaijan controlled territory. The target locations are typically within a few kilometers of the point of origin and coded with geoprecision 2.
• In some cases, no named locations or landmarks are nearby on the other side of the frontline from the point of origin. In such cases, the event is coded to the nearest named location or landmark on the side of origin. For example, firing coming from “near Karvend” (Artsakh side) is coded to the town of Karvend, with a geoprecision of 2.

Event and weapon types:

• As per general ACLED rules, the use of small arms – alone or in combination with other weapon types – during a single event is coded as ‘Sub-event type’ variable ‘Armed clash’. When heavy weapons are used alone in a specific location, without the use of small arms, the event is coded ‘Sub-event type’ variable ‘Shelling/artillery/missile attack’.
• Unspecified weapons are interpreted as small arms and coded as ‘Sub-event type’ variable ‘Armed clash’.
  • Investigations of incidents and reported weapon types from both Armenian and Azerbaijani sources over 2018 revealed that small arms were by far the dominant weapon type used.
  • Reports from the Azerbaijan MoD occasionally mention the use of heavy weapons, but these are reported in addition to unspecified small arms use. Due to the lack of specificity on locations, we cannot assume that any particular location was targeted with primarily heavy weapons. Events from these unspecific reports of heavy weapons use tend to span over several locations, and are therefore coded as ‘Sub-event type’ variable ‘Armed clash’.
• Armenian MFA reports to the UNSC outline the distribution of weapon types for the entire set of locations for that day, not for each individual location. Weapons information is hence applied to all the locations/events for that day equally.
  • When the vast majority of weapons used are heavy weapons (defined here as ⅔ or greater), the events are coded as the ‘Sub-event type’ variable ‘Shelling/artillery/missile attack’ instead of ‘Armed clash’, with the assumption that most of the events that day were solely heavy weapons use.

A note on two gaps in the ceasefire violation data from MFA and MoD:

There are two separate gaps in reporting from both the Armenian MFA and the Azerbaijani MoD throughout the 2018-2019 dataset. The first is a missing report from the Armenian MFA to the UNSC in July 2018. It is unclear why this report is missing, although it is believed to be the result of a clerical error. Non-MFA sources continued to report events during this period, so this gap does not represent a gap in the conflict. The second gap occurs between 22 June and 16 July 2019, when the MoD of Azerbaijan reported that there were no ceasefire violations – though the Armenian MFA and a number of other sources detailed specific reports of firing along the Line of Control during that period. ACLED continues to investigate whether these missing reports are retrievable.

What types of sources does ACLED use to cover the CAC region?

The ACLED source list for the CAC region features a large number of sources in Georgian, Armenian, Kazakh, Kyrgyz, Uzbek, Tajik, Russian, and English – with the latter two covering most events. ACLED’s initial source list was created according to standard ACLED procedures when expanding into a new region; these procedures can be reviewed here, under the sub-heading: What is the process for producing country and regional sourcing strategies?

As noted earlier, countries within the CAC region typically rank poorly when it comes to journalistic freedom. In addition, much of the region is significantly rural, further limiting access options for journalists. These factors pose challenges when it comes to sourcing information, and are addressed in a number of ways. Two local media experts – a veteran reporter for Central Asia and a regional expert from an international NGO on media freedom – were consulted in this regard, and offered advice about possible source biases, the viability of expanding the use of new media to boost coverage in the region, and potential local partners collecting conflict data in the region.

Traditional media are the source of information for the majority of events in the region other than those featuring border conflicts where the information comes primarily from the conflict parties themselves, in the form of each country’s ministry of defense or foreign affairs. National news agencies are reviewed weekly for each country; coverage from international news agencies, such as Radio Free Europe / Radio Liberty, report on countries where a free press is banned by the government or not fully established, such as many in the CAC region; and regional media in neighboring countries, such as Russian news covering events in Kyrgyzstan, is also included to help fill gaps. Sourcing information directly from different language sources within a country has also helped in curbing urban bias in some areas. For example, in Uzbekistan, populations in rural areas speak Uzbek, while Russian is more commonly found in urban areas and used in business and television (Eurasianet, 19 June 2019). ACLED also reviews in-depth reporting from reputed human rights organizations like Amnesty International and Human Rights Watch.

Appendix

Example report from Azerbaijan MoD 

Azerbaijan MoD website report – English version

31 March 2018 09:52

Military units of the armed forces of Armenia violated ceasefire 101 times throughout the day, using large-caliber machine guns.

Armenian armed forces, located on nameless hills in Noyemberyan and Ijevan regions, in Chinari village and on nameless hills in Berd region, and on nameless hills in Krasnoselsk region subjected to fire the positions of the Azerbaijan Army located in Gaymagly, Gushchu Ayrim and Jafarli villages of Gazakh region, in Aghdam, Aghbulag villages and on nameless hills in Tovuz region, in Goyali village and on nameless hills in Gadabay region.

The positions of Azerbaijan Army were also fired from positions of Armenian military units located near the occupied Chilaburt village of Terter region, Shikhlar, Bash Garvand, Garagashly, Yusifjanly, Marzili villages of Aghdam region, Kuropatkino village of Khojavend region, Ashaghy Veysalli, Garvand, Garakhanbayli, Ashaghy Seyidahmadli, Gorgan, Kurdlar villages of Fuzuli region, and Nuzgar village of Jabrayil region, as well as from positions located on nameless hills in Goranboy, Terter, Aghdam, Khojavend and Fuzuli regions.

Information gained:

Event date: refers to events from the previous day (i.e. 30 March 2018)

 

Weapon type: noted as heavy machine gun, however this report does not specify at which location(s) these were used. It is clear from context (analysis of numbers and types of weapons generally used) that large caliber machine guns were very likely not the majority or only weapons used at all mentioned locations. Therefore this information cannot be considered reliable enough to code each event as ‘Sub-event type’ of ‘Shelling/artillery/missile attack’ (which is the coding used to denote the use of heavy weapons only) as opposed to ‘Armed clash’ (which is the coding used to denote the use of heavy weapons mixed with small arms).

 

Locations near Armenia: 8 locations near Armenia, coded with Military Forces of Armenia as a primary actor

 

Locations near Artsakh: 19 locations near Artsakh, coded with Military Forces of Artsakh as a primary actor

 

Point of Origin: Additional info on the point of origin to add to the Notes column