USIP: PeaceTech Analyzing relationships between violence, financial corruption, and public opinion Jackson Institute of Global Affairs Capstone Project December 21, 2016
Presentation contents 1. Background 2. Public opinion on governance and conflict 3. Nigerian Financial Intelligence Unit reform 4. Final Takeaways 12/21/16 Jackson Institute for Global Affairs | 2
BACKGROUND Third party pollsters track media usage, demographics, 1a and state-by-state public opinion, compiled in Book 1 % Nigerians with access to channels of Percent of residents living comfortably information and communication on present income • Recent surveys provide valuable 100 information about 80 Family demographics, Radio policy approval 60 TV ratings, and top Newspaper 40 issues on a state- Internet 20 by-state basis 0 • This information is static and cannot Percent of residents finding is very be analyzed over Top issues for constituents in certain groups difficult to live on present income time or in relation to and states (i.e. Northeast) 1 drivers like violence 8% Poverty 11% Corruption 35% 35% Terrorism 17% Food & water Other 29% 29% 1. ACLED and Global Terrorism Database 2. Internal Displacement Monitoring Centre 12/21/16 Jackson Institute for Global Affairs | 3
BACKGROUND PeaceTech and its clients require analyses with broader 1b scope and ability to evaluate drivers of opinion Numerous questions from governors Example of changing sentiment toward a and NGOs remain unanswered government official over time • Existing literature on public opinion 1 How does public opinion about the is insufficient to federal government, military, and state 0.8 answer key governors change over time? questions for 0.6 governors or NGOs 0.4 like PeaceTech 0.2 How does violence change public 0 • We need perceptions of governors and good 1/1/15 7/1/15 1/1/16 7/1/16 innovative governance? techniques to aggregate and Example of negative opinions of the federal analyze public government in a given year, by state How do military interventions change sentiments public opinion about state and federal officials? To what extent do citizens associate different kinds violence with the government—positively or negatively? 1. ACLED and Global Terrorism Database 2. Internal Displacement Monitoring Centre 12/21/16 Jackson Institute for Global Affairs | 4
BACKGROUND Social media can augment third party assessments of 1c public opinion on violence, issues, actors, and events Example: social media tracks changes in opinion of Buhari over time 1 • Public opinion analysis normally President Buhari requires SMS, in- % Nigerians using internet for news weekly person, or paper surveys 40 • Social media can collect millions of 35 opinions and can be filtered by date 30 and topic based on lexicons built with expert input and 25 media analysis 5.9 million posts 20 • Although the about Buhari in Nigeria 2012 2014 2016 demographics of 2015-2016 social media users is not perfectly Fluctuations in total posts about Buhari, 2014-16 representative, as media penetration 1M improves, so does Post representativeness 2 volume 7/2015 1/2016 7/2016 1. Crimson Hexagon social media post aggregator 2. Regression analysis of BBG/Gallup data shows small but significant differences in education, income, and urbanicity of social media users 12/21/16 Jackson Institute for Global Affairs | 5
BACKGROUND Preliminary social media analysis illuminates shifts in 1d sentiment related to federal government, military, governors Ex: Differences in opinion on federal military Positive Negative Neutral • Comparing sentiments from local and federal National level Lagos state-level Benue state-level levels illuminates opinion of military opinion of military opinion of military significant differences in opinion, trends, 1 1 1 and spikes among 0.8 0.8 0.8 constituent groups 0.6 0.6 0.6 0.4 0.4 0.4 • Posts can be 0.2 0.2 0.2 classified as 0 0 0 positive, negative, 10/1/15 10/1/16 10/1/15 10/1/16 10/1/15 and neutral based on tested, scientific algorithms 2 Change in negative posts about military, 2015-16 Change in positive posts about military, 2015-16 1. Crimson Hexagon social media post aggregator 2. Based on the National Research Council - Word-Emotion Association Lexicon, 2011 12/21/16 Jackson Institute for Global Affairs | 6
BACKGROUND Can sentiment analysis of social media correlate changes in opinion about government or military 1e with terrorist, riot/protest, and state violence? Mapping opinion of governors and government after terrorist violence, riots and protests, and government interventions Determine correlation Map the shifts in public between terrorist violence By date, determine ratio of sentiment onto incidents and changing public Collect social media positive, negative, and of violence or protest data to capture raw sentiment about … . neutral posts about state public sentiment - Federal government and federal governments - State governors - Military 12/21/16 Jackson Institute for Global Affairs | 7
Presentation contents 1. Background 2. Public opinion on governance and conflict 3. Nigerian Financial Intelligence Unit reform 4. Final Takeaways 2. According to the Internal Displacement Monitoring Centre 12/21/16 Jackson Institute for Global Affairs | 8
SENTIMENT ANALYSIS Violence is primarily linked to four actors and concentrated 2a in specific states, providing clear case study examples • Four actors account Total violent incidents, by identifiable perpetrator 2 for the majority of incidents, and a Riots/protests Nigerian military Boko Haram Fulani Herdsmen Other 1 handful of states bear the brunt of this Riots/Protests violence Nigerian military Boko Haram Fulani Herdsmen • Boko Haram is Other concentrated in the Northeast states, 0 200 400 600 800 1000 1200 1400 1600 1800 2000 particularly Borno • Fulani Herdsmen are Distribution of Fulani attacks 3 Distribution of riots/protests 2 Distribution of Boko attacks 3 prevalent in the Middle Belt states, especially % of total, 2016 % of total, 2016 % of total, 2016 Higher % Higher % Higher % Benue • Riots and protests concentrate in urban centers like Lagos 1. 380 observations by 80 different actors, 129 observations by police forces 2. ACLED, 2015-2016 3. Global Terrorism Database 12/21/16 Jackson Institute for Global Affairs | 9
SENTIMENT ANALYSIS Social media analysis can determine significance of 2b shifts in opinion after incidents of violence PROCESS MAP ILLUSTRATIVE 1 • Sentiment analysis 0.8 Map social classifies each Positive 0.6 media 1 social media post Negative sentiment in 0.4 as positive, negative, given state Neutral 0.2 or neutral (example) 0 1/1/15 4/1/15 7/1/15 10/1/15 1/1/16 4/1/16 7/1/16 10/1/16 • Analyze shifts in opinion relative to ILLUSTRATIVE 1 incidents or violence Overlap 0.8 and regress to incidents of Positive 0.6 riots and identify significant 2 Negative 0.4 protests (or trends other event Neutral 0.2 type) Riot/protest • A significant result 0 1/1/15 4/1/15 7/1/15 10/1/15 1/1/16 4/1/16 7/1/16 10/1/16 could indicate that citizens blame % change in % change in state or federal positive posts negative posts 1 government for Aggregate % 0.8 Day of change in 12.7 -3.4 lack of security or 3 0.6 sentiment after have little faith in 1 Day 22.6 -5.4 an incident 1 and 0.4 governance evaluate pattern 0.2 2 Days 5.2 -0.4 0 1. Using an OLS regression 12/21/16 Jackson Institute for Global Affairs | 10
SENTIMENT ANALYSIS Borno does not exhibit significant changes in sentiment 2c toward the state government following any violent incidents Changes in sentiment toward the governor and state government 861 posts 1 2015-2016 0.8 Incidents Group Positive 0.6 Negative 376 Military 0.4 170 Neutral Boko Haram Borno Take-aways 0.2 10 Riots/Protests 0 • Violence has no 583 Attacks1 1/1/15 7/1/15 1/1/16 7/1/16 statistically significant impact on public opinion about % increase of positive posts about the governor % increase of negative posts about the governor the state governor or federal government significant at p<0.05 = significant at p<0.05 = MIL 1 BH R/P ATT MIL 1 BH R/P ATT • Violence by non-state actors leads to a spike -2.5 -9.0 15.8 -5.3 3.5 5.5 -9.3 7.8 Day of Day of in negativity about the military 1 Day -8.2 -1.1 27.1 -5.7 7.7 6.1 -18.2 6.8 1 Day 2 Days 4.4 -0.8 0.0 3.4 2 Days -6.3 8.0 2.6 -2.2 1. MIL = military intervention, BH = Boko Haram attack, R/P = riot or protest, ATT = attack 12/21/16 Jackson Institute for Global Affairs | 11
Recommend
More recommend
