trauma and conflict risks
Hertie, October 2026
The question
Can psycho-social accounts explain conflict cycles (and what can be done about it)?
Can psycho-social accounts explain conflict cycles (and what can be done about it)?
Conflicts do seem to re-occur
Conflict exposure can have enormous psychological effects on features that could themselves matter for between group relations
The question is whether distress, threat perceptions, or other psychological orientations can explain the wars that come back.
Can psycho-social accounts explain cycles?
The macro literature does not provide initial support for the idea:
- Civil wars do recur. Between 1945 and 1996, 36 percent of civil wars were followed by another war (Walter 2004). But:
renewed war is likely to have less to do with the attributes of a previous war […] than with current incentives individual citizens have to rejoin a rebel group
And:
- More deadly wars: not associated with reoccurrence
- Longer wars: reduced likelihood of reoccurrence
Can psycho-social accounts explain cycles?
Much of the macro literature does not really think about psychology at all
Exposure and militancy
On the other hand a lot of micro evidence suggests possible pathways:
Gábor Máté accounts of psychological effects and suggestions about these scaling (Pierre Boulez Saal 2024)
Hirsch-Hoefler et al. (2016) find that exposure to political violence makes Israelis and Palestinians less likely to support peace efforts
Hadzic et al. (2020) find that communities with more wartime casualties gave more support to ethnic parties
Grossman et al. (2015) show that for Israeli soldiers of the Second Intifada, combat is associated with a less conciliatory index, about \(-0.3\) standard deviations, and with more hawkish voting (and credible design).
Exposure and militancy
MacManus et al. (2013) show among UK personnel, post-deployment PTSD is associated with subsequent violent offending.
Taylor et al. (2020) use a meta-analysis of ten observational studies to show higher odds of violent arrest among veterans with PTSD, about 1.6
Complex patterns:
- Lindsey (2022) finds that exposure to armed violence in eastern Congo raises support for punishment for rape and lowers it for domestic violence (protective masculine norms)
Contrary logics?
- Also: Lindsey and Koos (2025) find that households of survivors of wartime sexual violence in eastern Congo do more social and political activity, not less (related to stigma, self-blame, and reconnection).
Famine effects
Narciso and Severgnini (2023) link family exposure to the Great Irish Famine, 1845–50, to participation in the Irish Revolution, 1916–21
- A one percent higher potato-crop failure rate in the county of birth raises the probability of being a rebel by 0.033 percentage points.
- A one percent higher excess mortality is associated with about 0.33 percentage points.
They track surnames, and they instrument blight with wind direction.
- Socio-cultural persistence? Something deeper?
Major theme of this talk
- Something of a convergence on the micro side:
- Exposure to wartime violence leaves political traces.
- However, they are small, and they are specific to the parties and groups of that war.
Multiple imaginable mechanisms
Interventions in this space (not always theory informed)
Micro-macro linkages missing
given the rich work that exists and importance of the topic, this should be a priority
A map
What the evidence shows
Positive effects of exposure to violence?
Bauer et al. (2016) concluded that although the human costs of war are horrific, there are reasons for optimism once the violence ends.
In case after case, people exposed to war violence go on to behave more cooperatively and altruistically, which we will generally call “prosocial” behavior.
(based on review of 23 papers)
They also note that the prosocial pattern may be strongest inside the village or the identity group.
Cooperation after violence
Trust stands out. It is also conceptually different.
In-group and out-group
in-group pattern read as ‘parochial altruism.’ if so, it could harden social divisions and contribute to conflict cycles
Lessons from the larger record
Barceló (2026) harmonizes 172 studies, across more than 50 countries, on 22 outcomes in four families
Not inconsistent with Bauer et al. (2016), though the negative implications come through much more strongly.
Lessons from the larger record
Most studies have no causal design. The headline models are bivariate.
Participation
Social-group participation is about 0.07 standard deviations and holds in the quasi-experimental subset. Political participation is weaker once the design is quasi-experimental. Altruism is small, and about \(-0.01\) in that subset.
Voting and trust
Voting and generalized trust are about zero. Normative prosociality, where it is estimated at all, leans negative, and no quasi-experimental estimate is in the figure.
Hardening toward the old enemy
Antagonism, threat, and intergroup distrust move by about 0.06 to 0.11 standard deviations. Several of these hold in quasi-experimental studies.
In-group identification
Not a general disposition to fight
Little movement supporting general dispositions.
But: a lot of heterogeneity
How big is 0.06 standard deviations?
Take a binary outcome with control mean \(0.5\).
An effect of 0.06 standard deviations moves the rate from
0.50to0.53The probability that a bad outcome is due to the exposure is anywhere between
6%and94%(Dawid et al. 2016). The data barely constrain individual attribution.
Big lessons but four limits
Very very impressive aggregation work and a rich body of studies.
Still:
The estimates are micro, not macro.
They are small.
They are strongest for specific outcomes not generalized psychological orientations.
Most are not causally identified. Those that are, are often weaker
Mechanisms
Three ways a mark could travel
An abundance of mechanisms:
Biological mechanisms: Developmental programming, in utero or in childhood.
Social transmission, through families, narratives, or local practice.
Effects on conditions: e.g. segregation, inequality
Amazing mice study
Dias and Ressler (2014) fear-conditioned mice to an odor before conception
Offspring and grand-offspring were more sensitive to that odor (social pathways blocked).
Cue-specific sensitivity is inherited.
But this is in line with alertness, not clearly with adverse processing effects associated with trauma.
Epigenetic evidence not very clear for this account
Holocaust methylation results:
Bierer and coauthors compare 125 offspring with 31 controls, and a pooled sample of 147 and 40 (Bierer et al. 2020).
- FKBP5 site-6 methylation is lower in offspring, and expression is higher. The methylation difference lines up with maternal exposure in childhood, and with lower offspring anxiety.
So, that suggests a developmental-programming result. But not fear or obvious tie to conflict propensity.
- Yehuda and Lehrner (2018) are also reserved in claims
Interventions
Aid aimed inside the community
A lot of aid interventions focus on strengthening trust, or social cohesion:
Fearon et al. (2015) study community-driven reconstruction in Liberia. Positive effects but the target is collective-action capacity inside communities
Humphreys, Sánchez de la Sierra, and van der Windt study Tuungane in eastern Congo. The target is the same: practices inside the village (Humphreys et al. 2019).
BMZ TDA programming, two interventions are underway in conflict settings.
One, in Nigeria, is a counselling intervention. It appears to generate prosocial activity.
One is a local-engagement intervention. It suggests some gains also, at a local level.
Neither addresses future conflict risk.
Reintegration in Aceh
Barron, Humphreys, Paler, and Weinstein evaluate BRA-KDP, a community-driven reintegration program in Aceh (Barron et al. 2009).
Like the Liberia and Tuungane projects, it targets practice inside communities.
If anything, it led to less acceptance of former combatants.
PTSD interventions 1
Robjant et al. (2019) randomize 92 female former child soldiers with PTSD in eastern Congo to adapted narrative exposure therapy or treatment as usual
At nine months, Cohen’s \(d\) is about \(-2.0\) for PTSD symptoms, \(-1.0\) for appetitive aggression, and \(-1.1\) for violent behavior in the past three months.
This is a large effect among people who had perpetrated violence.
The outcome is local violence (not the onset of a war).
Excellent study.
PTSD interventions 2
Pham et al. (2004) in four Rwandan communes, find that PTSD symptom criteria go with less support for national trials, less belief in community, and less interdependence across ethnic groups
- Reconciliation interventions may be most difficult among those that need them most
Back to war: A thought
Aggregation logics
The path from individual responses to social outcomes is often not direct Humphreys and Scacco (2020)
Simple aggregation: Walter style simple aggregation (Walter 2004) — if individuals are more conflict prone society or more hostile to out there conflict will occur directly
Social agggregation: Social norm change: Canetti et al. (2017)
Indirect aggregtation or effects via leaders
Strategic logics
Strategic accounts of conflict often do not focus on psychological factors (or individual orientations broadly).
But they do focus on decision making by individuals (leaders)
They focus often on logics of bargaining and bargaining failure.
And here there are at least four connections
Strategic logics
Four connections:
Grievance, in Gurr’s sense, is a group’s reading of relative deprivation and of its capacity to act (Gurr 1970). Trauma would have to change that reading, or that capacity.
Fearon (1995) highlights commitment problems. Possible connections here to trauma, e.g. in the work of Máté (Pierre Boulez Saal 2024).
Fearon (1995) also highlights uncertainty; a psychological state would matter if it stopped actors from absorbing information, or gave them a reason to conceal what they know.
Jervis (1976) highlights misperceptions: a psychological state would matter if it changed what leaders believe the other side wants, or what they believe their own actions signal.
Strategic logics
So can we address this at the level of a leader?
Could trauma, guilt, or an inherited historical obligation shift how a leader handles misperception, a commitment problem, or uncertainty?
I don’t know studies focusing on this; but sounds promising
Close
Where we are
Theoretically and empirically rich agenda
- Plenty of micro studies
- Much weaker understanding of aggregate outcomes
To treat trauma as a cause of future conflict, one would need:
- evidence about trauma itself rather than about having lived through violence
- a mechanism that aggregates: is the effect via leaders directly or indirectly via aggregate demand or via social norms
- exclusion of the many social pathways
Extra slides
A heartfelt position
On 11 October 2023, then Vice Chancellor Robert Habeck connected the Hamas attacks immediately to German violence against Jews and underlined Germany’s responsibility (Habeck 2023).
He said Germany would support Israel wherever that support is needed: wir werden es dabei unterstützen, wo immer es unsere Unterstützung braucht.
The position is heartfelt. It is disconnected from recognizably ethical principles. It is about a debt.
His own account of why
Welt reports Habeck’s account: coming to terms with the guilt of his great-grandfather and his grandfather still places him under a political obligation (Welt 2024).
Memory moves emotion. Attitudes are imprecise.
Balcells and Voytas find strong emotional responses to the Troubles and Beyond exhibit, in the museum and in a survey experiment (Balcells and Voytas 2026).
Attitude estimates sit near zero. The largest cleavage effect they highlight is \(d = 0.12\).
Figure 2 intervals run into several tenths of a standard deviation. That range includes the effects they found in a Chile museum study, about \(0.21\) to \(0.27\) (Balcells et al. 2022), and it is larger than Barceló’s pooled estimates (Barceló 2026).
The attitude result is an imprecise null. It is not a precise zero.
Stolpersteine: emotion, or a vote, not a mechanism from trauma
Türkoglu, Ditlmann, and Firestone estimate about a 1 percentage point lower AfD vote share after local placement (Türkoglu et al. 2023). They do not estimate a psychological mediator. They discuss visibility, moral emotion, and local social practices.
Ruipérez Núñez and Sauter show photographs of the stones. Anger, disgust, sadness, hostility, and fear rise. Pride and joy fall, from about \(0.17\) standard deviations for fear to \(-0.91\) for joy. Tolerance and most attitudes do not move. Right-leaning respondents become more negative toward refugees (Ruipérez Núñez and Sauter 2026).
Contact reaches peers
Scacco and Warren, in Kaduna, find no change in prejudice after sixteen weeks in mixed classes. Mixed-class subjects discriminate less. Homogeneous classes may discriminate more (Scacco and Warren 2018).
Mousa assigns displaced Iraqi Christians to mixed or all-Christian soccer teams. Tolerant behavior rises toward Muslim peers on the team. It does not extend to Muslim strangers, and attitudes do not move consistently (Mousa 2020).