Oil, Warfare and Diplomacy in the Nagorno — Karabakh Conflict
Background
The Nagorno — Karabakh Autonomous Region was established in 1923 by the Soviet Union, as a part of Azerbaijan Soviet Socialist Republic. According to a census taken in 1926, 89.2% of the 125,000 population of the autonomous region was comprised of ethnic Armenians. By 1989, this proportion dropped to 79.6% as a result of resettlement programs implemented by the Azerbaijan Soviet Socialist Republic. Despite historical ethnic tensions between the local Azerbaijani and Armenian populations, peace and neighborly relations were maintained under the Soviet umbrella.
Relations became strained as the Soviet Union weakened, and the Nagorno — Karabakh legislature passed a resolution in 1988 to become a part of Armenia. This followed a period of relatively low — intensity fighting between separatist Armenians and the Azerbaijani armed forces. In 1991 the autonomous region officially declared independence and full-scale war ensued between the two countries, resulting in roughly 30,000 casualties and hundreds of thousands of displaced civilians. By 1993, Armenia controlled Nagorno-Karabakh and approximately 20 percent of the remaining Azerbaijani land. In 1994, the Russian Federation brokered a ceasefire with the mediation of the Russian Federation.
The conflict froze over for more than two decades, with both sides becoming entrenched along a line of contact, which became a theater for regular sniper, mortar and artillery fire, as well as other tactical maneuvers on both sides. April 2016 saw drastic escalation, as the most intense fighting broke out between the countries since 1994, resulting in more than three hundred casualties. After four days of fighting, Russia mediated a new ceasefire agreement.
Four years later, a breakdown in negotiations led to full-scale military confrontation starting in late September 2020. Forty four days of continued combat resulted in military and civilian casualties surpassing thousands. As a result, Armenia lost control over 75% of Nagorno-Karabakh and the surrounding “buffer zone”.
Russia brokered another ceasefire on November 9th, but without successful mediation efforts and a peace treaty between the two countries, ceasefire violations still threaten to reignite military conflict and destabilize the region.
This article presents a quantitative analysis of the relationship between hydrocarbon markets and the dynamics of warfare and diplomacy in the Nagorno — Karabakh conflict. It cross — examines conventional economic indicators with unconventional data concerning diplomacy and warfare, obtained by applying Natural Language Processing models on multi — language online news media.
The section entitled Quantifying Diplomacy and Warfare describes the data used to measure the intensity of diplomacy and warfare in the Nagorno Karabakh conflict. The section entitled Economic Trends in Azerbaijan will explore economic conditions in Azerbaijan over the last decade. The section entitled Economy, Diplomacy and Warfare will cross — examine data and trends described in the previous two sections. Finally, the section entitled Conclusion will summarize insights and outline relevant future research topics.
Quantifying Diplomacy, Warfare and Domestic Unrest
To obtain granular proxy indicators for concepts such as diplomacy, warfare and domestic unrest, we implement a methodology based on extracting structured knowledge from online news media. Global Database of Events, Language, and Tone (GDELT) crawls over hundreds of thousands of news articles from around the world on a daily basis and uses Natural Language Processing models to extract events, themes, locations, organizations, people and sentiment. For each event, GDELT provides attributes such as event type, nations involved in the event, event location, event date, etc. To conduct our analysis, we query the GDELT event database for all events recorded between Armenia and Azerbaijan, grouping them into three groups — diplomacy, warfare and domestic unrest.
Our metric for the level of diplomacy, warfare and domestic unrest is the proportion of daily events between Armenia and Azerbaijan that fall into each of these categories. The graphs below illustrate the time series data obtained using this methodology.
The trends depicted in the graphs above are in line with the expert — level intuition for the relative dynamics of the conflict and domestic politics in Azerbaijan. Furthermore, they provide more nuanced detail that may elude even professional intuition. Evidently, media coverage of warfare, diplomacy and domestic unrest is a viable proxy indicator to quantify concepts as abstract as warfare, diplomacy and domestic political unrest. One of the reasons that this is significant, is that media coverage has hourly granularity, whereas conventional metrics to measure these phenomena have at best annual or bi-annual granularity.
As is evident in the above graphs, media coverage also provides a viable proxy indicator for the type of weaponry used. The graphs indicate significant increases in the use of heavy weaponry and aerial weapons during the war in late 2020, which is in line with the reality on the ground. The graph depicting use of small arms correctly depicts increasing activity after 2014. However, the decline in 2020 may be misleading. Although use of small arms drastically escalated during the war in late 2020, media coverage was preoccupied with the use of unmanned drones and ballistic and rocket artillery. This has the effect of reducing the relative scale of small arms use.
Economic Trends in Azerbaijan
The Azerbaijani economy is heavily reliant on oil revenues.
The green line in the graph above denotes the spot price for Brent crude. The blue line denotes the breakeven fiscal price of oil that the government of Azerbaijan requires to balance its budget. The orange line denotes the difference between the green and blue lines. According to this data, the government of Azerbaijan did not face extended periods of negative breakeven differences until after 2014, after which Azerbaijan entered a period of financial crises and budget deficits.
The economic indicators described below illustrate the effect of falling oil prices on the economy of Azerbaijan. Economic outlook projections by major credit rating agencies summarize expectations concerning the economy. Unlike the time period before 2014, after 2014 most economic outlook ratings were negative and Azerbaijan did not receive a single positive economic outlook rating.
According to the World Bank, the economic decline resulted in increasing poverty levels between 2014 and 2018. Record — low oil prices and COVID — 19 related quarantine measures in 2020 pushed poverty headcounts higher still and the trend is expected to persist in the foreseeable future. Migration rates to Russia, which are a commonly cited proxy for poverty levels in the South Caucasus, indicate that migration from Azerbaijan to Russia increased by 17% between 2016 and 2019.
Higher poverty levels in turn coincide with higher levels of domestic unrest, as coverage of domestic protests, demonstrations and other similar events became more common. The graph below illustrates the rise in the level of domestic political unrest immediately after the breakeven difference in oil prices plunged below 0.
Economy, Diplomacy and Warfare
The graphs below illustrate the relationship that diplomacy and warfare in the Nagorno — Karabakh conflict have with the oil price breakeven difference.
These figures illustrate that after 2014, the oil price breakeven difference has a strong positive relationship with diplomacy and a strong negative relationship with warfare. The graphs below provide correlations between said indicators for the time periods before and after 2014.
Since Azerbaijan is a price taker in the oil markets, correlations between oil prices and diplomacy or warfare in Nagorno-Karabakh indicate a one — way direct or indirect impact of oil markets on the Nagorno-Karabakh conflict. The primary insights obtained from these heat-maps are as follows:
- Correlation between oil price breakeven difference and diplomacy grew from 0.071 before 2014 to 0.74 after 2014. This shift is significant since the oil price breakeven difference went from having negligible influence on changes in the intensity of diplomatic efforts before 2014, to explaining 74% of linear change in the intensity of diplomatic efforts after 2014.
- Correlation between oil price breakeven difference and warfare flipped from 0.3 before 2014 to -0.71 after 2014. The shift from positive to negative correlation may imply strategic changes in the decision — making rationale of the Azerbaijani leadership and can be investigated further.
- Correlation between warfare and diplomacy strengthened from -0.3 before 2014 to -0.78 after 2014.
Conclusion
Empirical evidence indicates that in the period after 2014, economic conditions within Azerbaijan and international oil markets were closely related to the dynamics of diplomacy and warfare in the Nagorno Karabakh conflict. As such, the oil price fiscal breakeven difference, may be a strong predictor of geopolitical events related to the Nagorno — Karabakh conflict.
Our approach demonstrates a reliable methodology for applying Natural Language Processing and time series analysis to quantify phenomena in conflict studies, geopolitics and international relations. By employing deep learning models for topic classification and information extraction from unstructured open source news media, we are able to quantify abstract phenomena such as diplomacy and warfare and enable data — driven, evidence — based research that transcends human biases and cognitive limitations.
Further research on these methodologies may delve further into the use of topic clustering, sentiment and topic classification, and information extraction, to further explain the peaks and troughs in the time series we observed above. Furthermore, steps can be taken to integrate more conventional and unconventional variables to extract deeper insights. Use of data from social media platforms such as Facebook, Twitter and VK (Russian — В Контакте) may also enrich our dataset with unfiltered insights from every — day people and opinion makers.
