1) The document presents a theory that mass media can have non-linear effects on civil war onset, strengthening states above a critical threshold of media density but increasing risks below that threshold.
2) Analysis of cross-national data finds a positive correlation between media density and civil war onset for countries below the 20th percentile of media density, but a negative correlation above.
3) Changes in global TV density are positively associated with the number of civil war onsets in a given year, helping to explain the rise of civil wars since the 1960s.
1. Mass Media and
Civil War Onset
Article: j.mp/media_civil_war
Slides: j.mp/media_war_slides
Justin Murphy, PhD
Assistant Professor
Department of Politics and IR
University of Southampton
jmrphy.net
@jmrphy
2. Overview
• The Puzzle
• A Theory of Non-Linear Media Effects
• Hypotheses
• Data and Method
• Findings
• Implications and Conclusion
3. The Puzzle
• It is widely argued that mass media play a role in
the outbreak of civil wars (Brass 1997; Des Forges
1999; Gagnon Jr 1994; Kellow and Steeves 1998;
Metzl 1997; Tambiah 1996).
• But there is also evidence that mass media
strengthen the state and pacify potential
challengers (Anderson 1983; Deutsch 1953; Warren
2014).
• How can we square these contradictory findings?
4. A Theory of Non-Linear
Media Effects
• A system of mass communications will strengthen a
state because it generates a unique economy of scale
(Warren 2014).
• The key of my argument is that this should only take
effect beyond some critical threshold of media density.
• Before that threshold, each increase in media density
encourages insurgency by increasing the payoffs of
taking state power before mass communications are
achieved.
5. Hypotheses
At the cross-national historical level (unit is the country-year):
1. There should exist some threshold of mass media density
below which there is a positive correlation between mass
media and civil war onset and above which there is a negative
correlation.
At the international level (unit is global system in each year):
2. The mean level of media density should have a negative
correlation with civil war prevalence but year-to-year changes
in media density should be associated with an increase in civil
war prevalence.
6. Data and Method
• Cross-national time-series data on civil war onset (Sambanis 2004) and
mass media density from Arthur Banks and World Bank (Warren 2014).
• Unbalanced panel of 175 countries over a maximum of 55 years in
the period from 1945 to1999.
• Semi-parametric logistic regression (a general additive model) fits
mass media density to civil war onset with a nonparametric smooth.
• Traditional parametric logistic regressions for more convenient effect
estimates.
• For the international level hypothesis, negative binomial models for the
count of total onsets in each year, using TV as the key independent
variable.
7. The Non-Linear Effect of Media Density on Civil War Onset
0 1 2 3 4 5
−4−3−2−101
logmdi
s(logmdi,2.56)
8. Effect of Mass Media Density on Civil War
Onset for All Countries
9. Effect of Mass Media Density on Civil War
Onset for Countries < 20th Percentile of Media Density
10. Effect of Change in the Global Mean of TV Density
on Number of Civil War Onsets
11. Conclusion
• These findings suggest a crucial revision to our current understanding of the role
played by media in civil war onset.
• The introduction and early growth of mass media appears to significantly increase
the probability of observing a civil war within a particular country.
• The global growth of mass media technologies since the 1960s helps explain the
dramatic rise of civil wars in this period.
• Thank you and feel free to get in touch.
• Website: http://jmrphy.net
• Twitter: @jmrphy
• Article: j.mp/media_civil_war
• Slides: j.mp/media_war_slides