Religion and Violence — Parsing the Actual Numbers
Claims about religiously motivated conflict measured against terrorism databases, historical records, war encyclopedias, and peer-reviewed conflict research. How researchers define and count religious motivation — and what happens when you control for poverty, state weakness, and political exclusion.
Religion accounts for roughly 7% of historical wars under narrow definitions, but the share of modern civil conflicts with religious dimensions has risen from 3% in 1975 to 55% in 2015. Yet when you control for poverty, state weakness, and political exclusion, religious diversity loses statistical significance as a conflict predictor. Popular claims about historical religious atrocities — Crusade death tolls, Inquisition executions, witch trial numbers — are inflated by orders of magnitude relative to documented evidence. Religion functions as a powerful amplifier, mobilization tool, and identity marker that makes conflicts more lethal and harder to resolve — while also providing some of history’s most effective peacebuilding resources. Anyone offering a simple number is misleading you.
Wars — The “7% of Wars” Claim and the Modern Surge
The most commonly cited statistic — that only about 7% of all wars were religious — originates from the Encyclopedia of Wars by Charles Phillips and Alan Axelrod (Facts on File, 2005). The encyclopedia catalogs 1,763 wars spanning roughly 5,000 years; its index lists 121 entries under “religious wars,” yielding approximately 6.87%. The widely circulated figure of “123 religious wars” actually traces to Theodore Beale’s The Irrational Atheist (2008), who added two wars he believed the editors had omitted.
As historian Andrew Holt (Florida State College, 2018) documented, Phillips and Axelrod never explain their classification criteria — the “religious wars” label appears only in the index, with no methodological discussion anywhere in the three-volume work. The classification is binary, offering no spectrum for wars with partial religious dimensions. A corroborating BBC-sponsored audit rated major conflicts on a 0-to-5 religious motivation scale and found over 60% had no religious motivation, with fewer than 7% scoring above 3.
The more important story is what has happened since 1975. Three independent, peer-reviewed datasets converge:
Monica Duffy Toft (God’s Century, W.W. Norton, 2011) found religious civil wars rose from roughly 19% of all civil wars in the 1940s to approximately 50% by 2010. Svensson and Nilsson’s RELAC dataset (2018, Journal of Conflict Resolution, 420 conflict dyads from UCDP) found the share rose from 3% in 1975 to 55% in 2015. Basedau at GIGA Hamburg found at least 50% of armed conflicts in sub-Saharan Africa were “theological” by 2017. Critically, 75% of religiously defined conflicts in 2015 involved groups with self-proclaimed Islamist aspirations.
Toft further established that religious civil wars are more deadly, last on average two years longer, and are twice as likely to recur. Svensson (2007, Journal of Conflict Resolution) found conflicts with explicit religious claims are significantly less likely to be settled through negotiated agreements.
03 – The definition problemWilliam Cavanaugh (The Myth of Religious Violence, Oxford University Press, 2009) argues there is no transhistorical essence of “religion” — the boundary between “religious” and “secular” motivation is a modern Western construction. Catholic France fighting on the Protestant side in the Thirty Years’ War, and Protestants slaughtering Protestants in the same conflict, illustrate the problem. Neither the Correlates of War dataset nor the core UCDP dataset directly codes for religious causation.
A standardized coding scheme for religious motivation — applied consistently across all conflicts with inter-coder reliability testing — that yields a substantially different percentage than 6–7% for historical wars or substantially different from 50–55% for modern civil conflicts. Evidence that the RELAC coding inflates or deflates the trend.
Terrorism — A Lethality Gap, Not a Frequency Gap
Europol’s TE-SAT reports consistently show that separatist attacks outnumber jihadist attacks in raw numbers. In 2023, separatist attacks (70) constituted 58% of all EU attacks, mostly from Corsican groups, while jihadist attacks numbered just 14. Yet jihadist attacks caused all six fatalities that year. In 2024, jihadist attacks (24) killed five and injured 18; all other categories combined produced only two injuries. Arrests tell a parallel story: in 2024, 289 of 449 terrorism-related arrests (64%) were jihadist-related.
02 – The lethality gap in global dataBruce Hoffman’s analysis of the RAND-St. Andrews Chronology documented dramatic growth from zero identifiable modern religious terrorist groups in 1968 to 26 of 56 groups (46.4%) by 1995. In 1995, religious terrorists committed only 25% of recorded international incidents but were responsible for 58% of total fatalities (Inside Terrorism, Columbia University Press, 2006).
Victor Asal and R. Karl Rethemeyer (Journal of Politics, 2008) found religious ideology was a statistically significant predictor of organizational lethality even after controlling for group size and territorial control. James Piazza (2009) found in Iraq that al-Qaeda-affiliated groups committed one-third of Islamist attacks but were responsible for 95.9% of all casualties. Ido Levy (2023) confirmed that suicide terrorism — disproportionately associated with jihadist groups — largely accounts for the lethality gap.
03 – Most victims are MuslimThe Fondation pour l’innovation politique (Fondapol) compiled all Islamist terrorist attacks worldwide from 1979 to 2024 and found that 86.3% of attacks and 88.9% of deaths occurred in Muslim-majority countries. The CSIS analysis (Cordesman, 2017) using GTD data found 90% of deaths from top Islamist groups occurred in Muslim-majority countries.
Evidence that the lethality gap disappears when controlling for organizational characteristics (state sponsorship, territorial control, funding) rather than ideology. Or: GTD data showing that non-religious terrorist groups achieve comparable per-attack casualty rates when organizational factors are held constant.
Sectarian Violence — Religion as Identity Marker, Not Root Cause
The Troubles (1968–1998) are frequently miscategorized as a “religious war.” The scholarly consensus (O’Malley, McBride, Ruane & Todd) is that this was an ethno-nationalist conflict with a sectarian dimension. The Sutton Index (CAIN Archive) records 3,528 deaths: Republican paramilitaries killed 58.6%, Loyalist paramilitaries 29.2%, British Security Forces 10.0%. As Padraig O’Malley noted: “No one has seriously suggested that differences in theological beliefs are the root cause of our problems.”
02 – Iraq’s sectarian warFollowing the 2006 Samarra mosque bombing, Iraq’s sectarian violence produced an estimated 150,000–300,000 violent deaths across the full 2003–2023 period. Fawaz Gerges argued the U.S. invasion polarized the country along Sunni–Shia lines by destroying state institutions and imposing a sectarian political system — the sectarianism was catalyzed by political decisions, not theological disagreement.
03 – Myanmar and IndiaThe Rohingya crisis produced at least 6,700 violent deaths in the first month (MSF, December 2017), with over 700,000 displaced. The UN and U.S. State Department classified it as genocide. Yet the crisis involved ethnic, citizenship (1982 Citizenship Law), and colonial-legacy dimensions irreducible to Buddhism vs. Islam.
Ashutosh Varshney (Ethnic Conflict and Civic Life, Yale, 2002) compiled all Hindu–Muslim riots from 1950–1995. His central finding: eight cities containing 5% of India’s population accounted for 45% of all deaths. Cities with strong intercommunal civic engagement experienced far less violence even with similar demographics. Steven Wilkinson (Votes and Violence, 2004) showed communal violence is more likely where ruling parties don’t depend on minority votes.
04 – The theoretical frameworkJeffrey Seul (1999) articulated the now-dominant view: “Religion is not the cause of religious conflict; rather it frequently supplies the fault line along which intergroup identity and resource competition occurs.” Frances Stewart’s “horizontal inequalities” framework (Oxford, 2008) demonstrates that violence erupts when cultural differences coincide with economic and political inequalities. Toft coined “religious outbidding” — how elites compete for religious credentials to mobilize support.
A sectarian conflict where theological disagreement was the documented primary cause even in the absence of economic inequality, political exclusion, or ethnic competition. Evidence that sectarian conflicts are equally likely in societies with low horizontal inequality.
Historical Atrocities — Popular Claims vs. Documented Evidence
Modern Crusade historians (Madden, Riley-Smith, Tyerman, Asbridge) emphasize that reliable figures are essentially impossible for a 200-year period. The most commonly cited scholarly range is 1–3 million total deaths across all campaigns (Matthew White’s geometric mean: 3 million). For the sack of Jerusalem (1099), Muslim sources claimed 70,000 dead, Latin sources ~10,000, and recent Hebrew testimony cited by Asbridge suggests casualties “may not have exceeded 3,000.”
02 – The Inquisition: a 1,000× gapPopular and polemical claims range from 50 million to 100 million killed. The documented evidence is dramatically lower:
The discrepancy between popular claims and scholarly evidence is roughly 1,000× to 10,000×.
03 – Witch trialsThe widely cited “9 million burned” figure originated with Gottfried Christian Voigt in the late 18th century and was popularized by suffragette Matilda Joslyn Gage in 1893. Ronald Hutton called it a figure with “no rational basis whatsoever.” The peer-reviewed consensus from Levack (The Witch-Hunt in Early Modern Europe, 4th ed., 2016), Hutton (The Witch, 2017), and Sharpe clusters around 40,000–60,000 executions from approximately 100,000 trials. Notably, the Spanish Inquisition was actually more restrained than secular courts in witch trials.
04 – Comparing the scaleNew archival evidence substantially revising the scholarly consensus on Inquisition or witch trial death tolls upward. Discovery that the Vatican’s 1998 symposium or Monter/Kamen/Levack undercount by an order of magnitude. Evidence that the “9 million” witch-trial figure has a defensible empirical basis.
Religion as a Restraining Force — Peacebuilding and the Mechanism
The Community of Sant’Egidio mediated the 1992 Rome General Peace Accords ending Mozambique’s 16-year civil war — a peace that has held for over 30 years. In Northern Ireland, Father Alec Reid of Clonard Monastery secretly brokered the Hume–Adams talks that opened the channel to the 1998 Good Friday Agreement; Gerry Adams stated there “would not be a peace process without” Reid. Archbishop Desmond Tutu’s TRC in South Africa infused transitional justice with ubuntu theology. The Inter-Religious Council of Sierra Leone convinced rebel leader Foday Sankoh to agree to a ceasefire.
02 – The experimental mechanismR. Scott Appleby (The Ambivalence of the Sacred, 2000) articulated the core insight: the same “militance” that drives suicide bombers also drives peacemakers. Jeremy Ginges, Ian Hansen, and Ara Norenzayan (Psychological Science, 2009, four studies across six religions and six nations) found that frequency of attendance at religious services predicted support for suicide attacks, while frequency of prayer did not.
Among Palestinian Muslims, those attending mosque daily were 3× as likely to believe Islam “requires” suicide attacks — but prayer frequency had zero effect. In an experiment with Israeli Jewish settlers, priming synagogue attendance increased approval of Baruch Goldstein’s massacre from 15% to 23%, while priming prayer reduced it to 6%. Replicated across Indonesian Muslims, Mexican Catholics, British Protestants, Russian Orthodox, Israeli Jews, and Indian Hindus. The coalitional function of religion links to violence; the devotional function restrains it.
Ginges, Sheikh, Atran, and Argo (PNAS, 2016) showed that when 555 Palestinian Muslim adolescents evaluated a moral dilemma from Allah’s perspective rather than their own, they valued Israeli Jewish lives and Palestinian Muslim lives more equally. The Abu Dhabi Gallup Center’s 131-country study found Muslim Americans were the most opposed of any U.S. religious group to military targeting of civilians (78% “never justified” vs. ~58% for Catholics).
03 – The just war traditionThe just war tradition — from Augustine through Aquinas, Vitoria, and Suárez to Hugo Grotius — directly produced the principles of distinction (between combatants and civilians) and proportionality that form the backbone of modern international humanitarian law and the Geneva Conventions.
Evidence that the Ginges et al. attendance-vs-prayer finding does not replicate in pre-registered multi-site studies. Evidence that religious peacebuilders were no more effective than purely secular mediators in comparable conflict contexts.
Atheist-State Atrocities — Totalitarianism, Not Disbelief
Not one of the major historians of these regimes — Conquest, Applebaum, Wheatcroft, Snyder, Dikötter, Kiernan — identifies atheism as the primary causal factor. Tim O’Neill (atheist medieval historian) argues both the Christian apologist position (“atheism uniquely caused the bloodshed”) and the New Atheist position (“atheism had nothing to do with it”) are wrong. Atheism was a constitutive element of Marxist-Leninist ideology — churches were destroyed because of committed beliefs about religion as false consciousness. But atheism operated within a package including class warfare, personality cults, and elimination of dissent.
R.J. Rummel’s own research — paradoxically, given how his numbers are used in debate — found that the democratic-totalitarian spectrum, not the religious-atheist spectrum, was the key predictor of mass killing. His principle: “Power kills; absolute Power kills absolutely.”
Claiming “atheism caused the Gulag” commits the same error as claiming “religion caused the Crusades.” The common factor across history’s worst atrocities is not the presence or absence of God-belief but the combination of utopian ideology demanding total social transformation, totalitarian concentration of power, dehumanization of out-groups, and suppression of dissent. Highly secular liberal democracies (Scandinavia, Japan, Czech Republic) are among the least violent societies in history. Deeply religious societies have produced catastrophic violence (Rwandan genocide: 90% Christian country).
Evidence that secular liberal democracies produce mass violence at rates comparable to totalitarian states. Or: evidence that the anti-religious campaigns were the primary motivating force (rather than political consolidation) behind the Great Purge, Great Leap Forward, or Khmer Rouge killing fields.
Both Sides Steelmanned & What the Quantitative Models Show
The share of civil conflicts with religious dimensions has surged from 3% to 55%. Religious terrorism is disproportionately lethal. Religious civil wars last longer, recur more, and are harder to negotiate. The experimental evidence (Ginges et al.) shows that communal religious participation specifically predicts support for suicide attacks across six religions. The Crusades, Wars of Religion, and modern jihadist campaigns demonstrate that religious conviction can motivate extraordinary violence. These are not mere identity markers — fighters explicitly cite scripture, promise divine reward, and define enemies in theological terms.
Only 7% of historical wars are classified as religious. Fearon and Laitin’s landmark study (2003, APSR, 12,000+ citations) found that after controlling for per capita income, more religiously diverse countries were no more likely to experience civil war. Poverty, state weakness, and rough terrain predicted civil war regardless of religious composition. Bormann, Cederman, and Vogt (2017) found no support for the thesis that Muslim groups are particularly conflict-prone. Popular historical atrocity claims are inflated by 1,000×. The just war tradition produced international humanitarian law. Religion also provides powerful peacebuilding resources.
The most important and least-discussed findings come from quantitative conflict models:
1. Historically, religion was the primary driver of only a small minority of wars (≈7%), but the share of contemporary armed conflicts with religious dimensions has risen dramatically since the late 1970s. 2. Religiously motivated terrorism is disproportionately lethal, though concentrated in transnational jihadist groups; the vast majority of victims are Muslim. 3. Leading quantitative models consistently show that poverty, state weakness, and political exclusion explain far more variance than religious diversity. 4. Popular historical atrocity claims are inflated by orders of magnitude. 5. Religion also provides powerful peacebuilding resources. Genuinely contested: Whether religious civil wars are increasing because religion is more salient or because Cold War proxy conflicts have exited the denominator. Whether the lethality gap reflects religious ideology or organizational characteristics of specific groups.
