Measuring armed conflicts and conflict deaths across the world helps us understand how people’s lives and livelihoods are affected by large-scale violence.

But this comes with many challenges. People do not always agree on what characteristics define an armed conflict. Even once defined, these characteristics — especially how many people died in them — are difficult to assess.

The people affected are not always asked who has died around them due to the conflict.

The conflict parties may underreport deaths to claim success, or overreport them to encourage intervention by third parties.

Independent observers may also struggle to be present in all places and document a conflict’s death toll.

So how do researchers address these challenges?

In our work on War and Peace, we provide data from six sources that identify armed conflicts and count their deaths:

• Uppsala Conflict Data Program (UCDP)¹
• Project Mars by Jason Lyall (2022)²
• Militarized Interstate Events by Douglas Gibler and Steven Miller³
• Correlates of War (CoW)⁴
• Peace Research Institute Oslo (PRIO)⁵
• Conflict Catalog by Peter Brecke (1999)⁶

These sources all measure armed conflicts and their deaths; they cover many countries and years, and researchers and policymakers frequently use them.

In our Conflict Data Explorers, you can delve into global and regional overviews of conflicts and their deaths.

Reassuringly, sources typically agree about big differences in the deaths from armed conflicts: they readily distinguish between deadly and peaceful years.

But they do not always agree. They often come to different assessments about how widespread and deadly some years were.

Why do these sources arrive at different numbers? And is there any best approach?

I summarize the key similarities and differences between the different approaches in this article and when to use which one.

What conflicts are covered?

In this table and the ones that follow, I summarize what armed conflicts and conflict deaths each data source covers.⁷

All sources characterize an armed conflict as a disagreement between organized groups that involves force. For most datasets, this means that force is used. But for the Militarized Interstate Events data and Correlates of War’s Militarized Interstate Disputes data, this can also mean that force is threatened or displayed.

Approaches also differ in what types of conflicts they cover and how many deaths they must cause to be included.

They all include armed conflicts where the primary combatants are states (interstate conflicts), and most cover those conflicts between a state and an armed group that is not a state inside the state’s territory (intrastate conflicts) or outside the state’s territory (extrastate conflicts).⁹ These conflicts are sometimes grouped together as “state-based” conflicts.

Some approaches go further and also include conflicts where neither armed group is a state (non-state conflicts). ¹⁰ And some even include violence by an armed group — state or non-state — against civilians (one-sided violence).

Most sources only include conflicts if they cause many deaths. Some set a threshold of at least 500 or 1,000 annual or total deaths, and then call these deadly conflicts wars. Others include more minor armed conflicts that cause a few dozen deaths during a year. And the Militarized Interstate Events data and Correlates of War’s Militarized Interstate Disputes data also include conflicts in which no one died.

Finally, the specific definitions of the sources mean that their conflicts differ slightly beyond their differences in conflict type and death toll. The biggest of these minor differences is that Project Mars only includes conventional wars, those fought between differentiated militaries along clear frontlines. Others do not make such a distinction.

Which deaths are included?

The data sources also differ in who they include in their death counts and how they were killed.

All sources record deaths, the people killed in the conflict. This is different from other sources which record casualties, which include both deaths and those wounded. Yet other sources even tally losses, including both casualties and those deserted.¹¹

But some approaches only include deaths of combatants — those doing the fighting. Others also include civilians — those who do not take up any weapons but may be killed accidentally or intentionally (in cases of one-sided violence).

These civilian deaths also include victims of terrorist attacks, as long as the attack was carried out by an armed group.¹²

And some approaches only include direct, or battle-related, deaths — those due to the fighting. Meanwhile, others include direct and indirect, or war-related, deaths — both those due to fighting, and those due to disease or starvation caused by the conflict.

What years and countries are covered?

All data sources provide historical data and cover conflicts worldwide. But they differ in what years they cover, and which countries and regions they consider sovereign states.

UCDP currently provides the most recent data, as it is updated each year. Other sources are updated less frequently and currently have a delay of about a decade on their most recent data (Project Mars, Militarized Interstate Events, and Correlates of War), or they are no longer maintained and will likely not be updated for recent decades (Conflict Catalog and PRIO).

In turn, UCDP only begins at the end of the Cold War. In contrast, the other sources start in the mid- or early 20th century (PRIO), the early 19th century (Project Mars, Militarized Interstate Events, and Correlates of War), or even reach as far back in time as the year 1400 (Conflict Catalog).

The data sources also differ in which countries and regions they consider sovereign states. Correlates of War and Militarized Interstate Events use the state system by Correlates of War, which until 1919 requires diplomatic recognition by France and the United Kingdom for a country to be included.

UCDP and PRIO meanwhile use the slightly more inclusive list of sovereign states by Gleditsch and Ward. Project Mars and the Conflict Catalog have even more expansive definitions of statehood, and are closer to the state system as identified by Butcher and Griffiths.

This affects how some conflicts are classified, as an extrastate or non-state war in Correlates of War’s war data can be classified as an interstate or civil war in Project Mars. And it affects how conflict rates are calculated, as the number of countries for each year can differ a lot, especially for Africa and Asia and Oceania.

How are conflicts and deaths identified?

Generally, the data sources are similar in how they go about assessing conflicts and the deaths in them.

All rely on researchers evaluating independent news reports, academic research, or both. Some also supplement these primary sources with available surveys of the affected populations, or official death statistics.

The primary sources they use can partially overlap. Some use the same sources, such as the encyclopedias by Micheal Clodfelter¹⁴ Lewis Fry Richardson¹⁵, and Pitirim Sorokin¹⁶. Or they outright rely on each other as the sources: PRIO and Project Mars use UCDP data, and Project Mars, Militarized Interstate Events, and the Conflict Catalog build on Correlates of War.¹⁷

How do sources work to measure what they want to measure?

The next tables summarize how the data sources address the specific challenges of measuring armed conflicts and conflict deaths. The first challenge is to make their assessments valid — to actually measure what they want to measure.

In general, the sources relying on their own researchers have the advantage that these researchers know the respective definitions of armed conflicts and related deaths well and are very familiar with the measurement procedures, which makes it more likely that they code as intended.

However, the sources differ in the deaths they include because they weigh the challenges of measurement differently. Some include civilian deaths because they often find it too difficult to distinguish them from combatant deaths.

Others exclude them because they find them too difficult to identify comprehensively or because they are only interested in the actions of militaries on the battlefield.

Some approaches include deaths due to disease and starvation because they consider them a fundamental part of war, and because primary sources often do not distinguish between them and deaths due to fighting. But others do not, as they find such indirect deaths too challenging to identify comprehensively.

Finally, Correlates of War’s war data sometimes estimates the number of deaths based on the number of conflict casualties, if only those are available. And its Militarized Interstate Disputes data mostly groups deaths into broad ranges, as it finds determining more specific numbers too uncertain.

How do sources work to make assessments precise?

The sources are also concerned with making their assessments precise and reliable.

They all use many primary sources, and most use different types of sources, such as news reports, academic books and articles, or NGO documents.

They also all avoid covering conflicts that happened even earlier — even though they differ in how far they think they can go back in time and still have data that is sufficiently reliable: UCDP begins in 1989, whereas the Conflict Catalog starts in the year 1400.¹⁸

Several sources also disaggregate their data to make it more precise: they split their estimates of deaths into total, combatant, and civilian deaths, when possible; disaggregate bigger conflicts into smaller ones; or each conflict into years, campaigns, or even specific events.

How do sources work to make assessments comparable?

The sources also face the challenge of making the data comparable across conflicts and time.

Some do so by providing their researchers with precise definitions of conflicts and deaths, making it easier to decide which to include.

And some also use primary sources in different languages or first identify conflicts and deaths with global sources and then use more local ones. Others, meanwhile, prefer global sources for their more comprehensive coverage of international affairs.

How are the remaining differences between primary sources dealt with?

The sources work to address any remaining differences between primary sources, even if they do so differently.

They all consider the quality of primary sources and prefer more reliable ones. Some also consider the context of the conflicts — such as army size and weapons used — to assess which accounts are more plausible.

Others also discuss any discrepancies between researchers on their team. This adds an additional step to make the assessments comparable across coders, conflicts, and time.

Correlates of War’s Militarized Interstate Disputes data partially reconciles differences between conflict parties by erring on the side of including accusations by a targeted state even if the state allegedly targeting it denies them.

Several approaches provide more than one estimate of conflict deaths and share a low, high, and sometimes also a best estimate. By doing so, they avoid forcing themselves to eliminate all uncertainty and possibly biasing their estimates.

UCDP combines these strategies and uses the lower estimate as the best one if no primary source is more reliable than others. It thereby reduces the risk that the estimate overcounts deaths.

How do sources work to make data accessible and transparent?

Finally, the data sources all make their data accessible and the underlying measurement transparent. They all publicly release their data, and most make it straightforward to download and use.

All release descriptions of how they characterize conflicts and deaths and share at least some of the coding procedures guiding the researchers.

Militarized Interstate Events and Correlates of War also provide additional helpful information by describing conflicts in accompanying reports. And UCDP shares the primary sources for each event underlying every conflict.¹⁹

The best source depends on your questions

There is no single ‘best’ approach to measuring armed conflicts and their deaths. All sources put a lot of effort into measuring conflicts and deaths in ways that are useful to researchers, policymakers, and interested citizens.

The most appropriate source depends on what questions you want to answer. It is the one that captures the conflicts, deaths, and time you are interested in.

If you are interested in both big and smaller conflicts in recent years, in combatant and civilian deaths, or also non-state conflicts and acts of one-sided violence, UCDP is best.

If you are less interested in the most recent years and more in big and smaller conflicts that involve at least one state in the decades since World War II, PRIO data is best. If you also want the most recent years, you can use our combined UCDP+PRIO data.²⁰

If you want to study long-term differences and changes over the last two hundred years and it is sufficient to focus on bigger conflicts, then Project Mars is best.

If you are especially interested in specific types of large conflicts over recent centuries, the Correlates of War data is best.

If, instead, you want to explore both big and small conflicts between states, the Militarized Interstate Events data is best.

If you want to explore the source that was researchers’ go-to for interstate conflicts for a long time and are fine with its less precise data, Correlates of War’s Militarized Interstate Disputes data is best.

And if you want to study bigger conflicts over more than the last half millennium and are willing to accept lower data quality, the Conflict Catalog is best.

Even if you have a preferred source, it can still be useful to see what other sources show and where they agree and differ.

This means that having several approaches to measuring conflict deaths is not a flaw but a strength: it gives us different tools to understand the past, current state, and possible future of armed conflict around the world.

If you want to explore and compare the data that each dataset produces, you can do so in our work on War and Peace and in our Conflict Data Explorers.

Keep reading at Our World in Data

How different measures capture how common and deadly conflicts are, and when to use which one

There are several ways to capture how common and deadly armed conflicts are. What are they? And when is which measure best?

Our Data Explorers on Armed Conflict and War

Browse our two detailed data explorers on armed conflicts and war, with almost 300 interactive visualizations across six data sources.

Bastian Herre (2023) - “How major sources collect data on conflicts and conflict deaths, and when to use which one” Published online at OurWorldinData.org. Retrieved from: 'https://ourworldindata.org/conflict-data-how-do-researchers-measure-armed-conflicts-and-their-deaths' [Online Resource]

@article{owid-conflict-data-how-do-researchers-measure-armed-conflicts-and-their-deaths,
    author = {Bastian Herre},
    title = {How major sources collect data on conflicts and conflict deaths, and when to use which one},
    journal = {Our World in Data},
    year = {2023},
    note = {https://ourworldindata.org/conflict-data-how-do-researchers-measure-armed-conflicts-and-their-deaths}
}

Reuse this work freely

All visualizations, data, and code produced by Our World in Data are completely open access under the Creative Commons BY license. You have the permission to use, distribute, and reproduce these in any medium, provided the source and authors are credited.

The data produced by third parties and made available by Our World in Data is subject to the license terms from the original third-party authors. We will always indicate the original source of the data in our documentation, so you should always check the license of any such third-party data before use and redistribution.

All of our charts can be embedded in any site.