Politically Motivated Killers: 51 Years of Terrorist Murders on US Soil, 1975–2025

The 208 terrorist attackers who killed at least one person on US soil between January 1, 1975, and December 31, 2025, murdered 3,577 people. Murders in politically motivated attacks account for about 0.35 percent of all murders during that time, which translates to an annual chance of being killed in such an attack of about 1 in 4 million per year. The deadliest motivating ideology for those attacks is Islamism, which accounts for a total of 87 percent of all deaths and includes those who were murdered in the 9/11 attacks. Right-wing ideology is the second deadliest motivation at 10 percent, left-wing ideology at 2 percent, and unknown/​other, foreign nationalism, and separatism together accounting for less than 1 percent of all murders.

Two attacks are responsible for most of the terror deaths on US soil: 9/11 and the 1995 Oklahoma City bombing. Those two attacks caused 3,147 deaths, which is 88 percent of all terrorist murders over the 51-year period. Of the other 430 murders, right-wing terrorists account for 45 percent of people murdered, Islamists are responsible for 32 percent, left-wing terrorists are responsible for 16 percent, and attackers motivated by the other ideologies account for the remainder. The number of politically motivated killings is not increasing over time.

Of course, murders don’t capture the totality of politically motivated violence or the amorphous feelings that people have that American politics are out of control. However, they are the best measure of the actual violence and an excellent indicator that the United States is not suffering through a wave of politically motivated terrorist killings.

Politically motivated terrorist attacks are acts of violence committed by nonstate actors to achieve political, ideological, religious, or social goals. These types of attacks can include assassinations of prominent political figures such as conservative activist Charlie Kirk in 2025 or the attempted assassination of Donald Trump in 2024; mass shootings such as those at the San Bernardino Inland Regional Center in 2015; the Pittsburgh Tree of Life Synagogue in 2018; the El Paso Walmart in 2019; or other planned and coordinated terrorist attacks such as 9/11, the 1993 World Trade Center bombing, and the 1995 Oklahoma City bombing. In this policy analysis I examine all politically motivated terrorist attackers from January 1, 1975, to December 31, 2025, noting the dates of their attacks, how many people they murdered, their countries of origin, and specific political, ideological, religious, or social motivations. The data indicate that people often greatly overestimate the risk of death from terrorism and the number of terrorist murders on US soil. This is understandable given that these attacks are tremendous tragedies and horrific crimes whose felons deserve to be punished to the fullest extent of the law.

The 208 attackers who killed at least one person on US soil over the period of this study murdered 3,577 people. Those murders account for only about 0.35 percent of all murders committed in the United States over that time. This translates to Americans facing about a 1 in 4 million per year chance of being murdered in a terrorist attack. By comparison, the annual chance of being the victim of homicide that is unmotivated by political or other terroristic ideals was about 1 in 14,000 per year, or about 284 times as great. No matter how one views the number of politically motivated terrorist killings, whether in absolute numerical terms or as a share of all murders, they are a minuscule threat to Americans. Regardless, they are still tragic and vicious crimes that have produced too many victims. And they may have other negative effects that are not entirely expressed by the number of victims.

The 9/11 attacks were the deadliest in this dataset, killing 2,979 people and accounting for 83 percent of all deaths recorded. These attacks were nearly 18 times more lethal as the next deadliest attack, which was the 1995 Oklahoma City bombing that murdered 168 people and accounted for just under 5 percent of all those murdered terror victims. By far, the deadliest motivating ideology was Islamism, which inspired attacks that led to 3,118 murders, or 87 percent of all deaths in this dataset. Attackers inspired by right-wing ideology were the second deadliest and were responsible for 10 percent of all the murders, or 363. Attackers motivated by left-wing ideologies murdered 68 people and were responsible for under 2 percent of the killings. Lastly, unknown/​other, foreign nationalism, and separatism together account for 28 killings and less than 1 percent of all murders.

Focusing on some political motivations or excluding others, homing in on different periods, and other ways of slicing the data or emphasizing different points all show that the number of politically motivated terrorist killings represents a small threat in the United States. As tragic as these killings are for the victims, their loved ones, friends, and communities, they are not a growing threat that requires an expanded federal response.

The Trump administration disagrees. President Trump issued an executive order designating ANTIFA—a mostly nebulous left-wing political movement—as a domestic terrorist organization on September 22, 2025, exercising a presidential power and group designation that does not exist in US law.¹ He followed up with a National Security Presidential Memorandum (NSPM‑7) directing the Joint Terrorism Task Force and the rest of the domestic national security and counterterrorism state to focus on rooting out domestic terrorist organizations, with a thinly veiled focus on left-wing organizations that the administration accuses of supporting violence.² Consistent with both of those extraordinary actions, Trump attended a conference with Attorney General Pam Bondi, Department of Homeland Security Secretary Kristi Noem, Federal Bureau of Investigation Director Kash Patel, and a slew of private-sector individuals to discuss the threat of political violence posed by progressive and left-wing groups.³ Bondi has issued follow-up memos to focus anti-terrorism investigations on left-wing groups.⁴ As part of this effort, there is substantial convincing evidence that the administration is coordinating an across-the-board investigative effort to disrupt left-wing philanthropy and other groups that it claims are supporters of domestic terrorism.⁵ The data indicate there is no current danger large enough to justify such a significant shift in domestic security priorities, let alone the assault on political liberty.

Politically motivated terrorism is the threatened or actual use of illegal violence by nonstate actors to attain political, ideological, religious, or social goals through coercion, fear, or intimidation. The definition is identical to terrorism without the attributive adjective “politically motivated,” so the terms will be used interchangeably or in combination throughout this policy analysis. This definition is slightly broader than how federal law defines the term so as to conform with popular perceptions of terrorism in the press and political debate, but it is still very similar to federal law.⁶ Federal law divides terrorism into international and domestic terrorism, where the former primarily occurs outside the United States and the latter occurs on US soil without foreign direction.⁷ Federal law allows the US secretary of state to designate foreign organizations as terrorists if certain conditions are met, but there is no similar designation for domestic terrorist organizations.⁸

This analysis focuses on politically motivated terrorists who killed at least one person in an attack on US soil over the 51-year period from January 1, 1975, to December 31, 2025. The methodology below and the particular choices made are intended to identify a reasonable count of the number of people murdered in politically motivated terrorist attacks in the United States. The dataset is organized by attackers, but it is also easy to reorganize it by attack type. This analysis only includes attacks that resulted in at least one death because other measures of violence are extremely variable, often vague, and don’t easily allow comparisons between attackers. Additional justifications and explanations of the tradeoffs associated with this methodological choice are provided in the next subsection.

This is an analysis of people who kill to further their own political and other terrorist goals. It is not an analysis of killers with incidental political opinions, nor is it an analysis of the political opinions of the victims. For instance, Anthony Comello was a right-wing obsessive who believed in the QAnon theory that a liberal elite gang of pedophiles was secretly controlling the US government and thwarting President Trump’s agenda. Comello tried to carry out a citizen’s arrest of Gambino crime family boss Francesco Cali in 2019 and killed him in the process. Comello’s devotion to a once-prominent conservative conspiracy theory inspired him to act, but there was no wider political goal for his actions, so he is not included among the terrorists in this study.

Assigning ideological motivations to the individual killers requires looking at their statements and actions. The most useful are manifestos, notes, online posts, or other forms of communication produced by the attackers before their attacks that justify their crimes. Statements by the attackers after arrests, during their trials, or in subsequent interviews are also useful, but slightly less so, because they may be strategic legal reactions prompted by competent defense attorneys in attempts to avoid the most serious punishments. Other statements and actions prior to the attacks can also provide insight into motivation. However, judging the ideological motivations of the attackers is the most subjective component of this analysis. In borderline cases, the inference of political motivation may be partially endogenous to target selection. Such cases are rare and, importantly, do not affect the measurement of the scale of all political violence, only its distribution over different ideological motivations.

The political motivations of the terrorist killers are divided into the following categories: Islamism; foreign nationalism; right (anti-communism, white supremacy, anti-vaccine, incel beliefs, anti-abortion, pro-Christian, etc.); left (communism, black supremacy, anti-white, animal rights, environmentalism, etc.); separatism of various kinds; and unknown/​other. Violent attacks against the police are ideologically categorized based on available evidence, but most are committed by left-wing terrorists. Men with incel (“involuntarily celibate”) beliefs are included in the right ideological motivation category because those beliefs include social misogyny, anti-feminist, anti-women, anti–out-group, and pro-hierarchy.⁹ There are only a few incel killers, so readers who disagree with this ideological judgment can make the adjustment themselves and see that it barely affects the ideological distribution of the killers.

Most of the terrorists in the unknown/​other category have an “other” ideology, such as Thomas Jacob Sanford, who murdered four people in a 2025 attack on a Church of Jesus Christ of Latter-day Saints meetinghouse in Michigan. His premeditated attack was motivated by his anti-Mormon beliefs. Thomas Matthew Crooks’ motivation in shooting Trump on July 13, 2024, remains a mystery because he left behind no writing indicating a motive.¹⁰ However, his target was so political that the odds are that the assassination attempt was politically motivated. There are only a handful of killers in this dataset with “unknown” motivations that almost certainly were political; these are the exceptions that violate the general rule of judging entirely the motivations of killers based on their writings rather than the target.

Some of the killers were motivated by more than one ideology. In those cases, the ideological motivations are grouped into primary and secondary motivations. For instance, the four Cuban-born terrorists Valentin Hernandez, Alvin Ross Diaz, Guillermo Novo Sampol, and Eduardo Arocena were primarily motivated by a fierce anti-communism directed at the Castro regime that took over their country of birth, but their motivation could also reasonably be described as a foreign nationalist belief. Thus, their primary motivation was coded as right-wing, and their secondary was coded as foreign nationalism. Many of the specific ideological motivations for attacks are not mutually exclusive. Unless otherwise stated, the primary ideologies are reported herein with additional data available in the appendix.

Crimes committed for personal financial gain do not count as politically motivated attacks. The only types of crimes that could be confused as an exception to this are if the murders happen during a robbery that is intended to fund terrorist activities. Killings that occur in other crimes meant to support politically motivated terrorist attacks are included as terrorism, such as the 1981 robbery of a Brinks armored car by an offshoot of the Black Liberation Army that resulted in three deaths. Common-law rules about felony murder guide my inclusion of victims in terrorist attacks, except that terrorists who die in the attack are not counted as victims. Members of a terrorist organization who are intentionally killed by other members of the organization for any reason, such as internal security, are not counted as victims. If a victim in an attack is injured and later dies of their wounds, then they are included as a murder victim even if years pass between the initial injury and death. Innocent bystanders of an attack who are accidentally killed by the police or security forces in response to the terrorist attack are included as victims of the terrorist attack. Politically motivated killers who commit attacks in prison or as members of prison gangs are excluded. Although many prison gangs have an overt political agenda or ideological organizing principle, such as race or religion, the goal of those groups is to supply club goods to gang members and only incidentally to advance political or ideological goals unrelated to personal gain.

The country of origin of the terrorists refers to their country of citizenship at birth. If the politically motivated terrorists committed a series of attacks before and after the start date of the analysis, which is January 1, 1975, this analysis only includes those that were committed on or after that date. Thus, the dates recorded in the dataset are when the attack occurred. If an attacker committed several attacks over many years and each attack resulted in a different murder, then the first date is recorded in the dataset. It’s rare that this happens, but it most prominently occurred in the case of left-wing terrorist Theodore Kaczynski, also known as the Unabomber. Kaczynski murdered Hugh Scrutton, a computer store owner, on December 11, 1985; advertising executive Thomas Mosser on December 10, 1994; and president of the California Forestry Association Gilbert Brent Murray on April 24, 1995.¹¹ The method of recording Kaczynski’s victims slightly distorts the temporal distribution of deaths, but the paucity of such cases ensures that the effect is barely detectable. Still, the committed reader can adjust the dataset accordingly if they wish.

The hardest challenge in this analysis is distinguishing terrorism from so-called hate crimes, where an offender commits a violent or property crime targeting an individual or small group and is motivated by ethnic, racial, national, religious, or other forms of bigotry based on the offender’s personal malicious opinions without broader political, economic, religious, or social goals.¹² Terrorism is often motivated by personal hatred, but the attacker must be aiming at a broader political, economic, religious, or social goal for their violent actions in order for the attack to qualify as terrorism.

The difference between politically motivated terrorist attacks and hate crimes is often that the former are premeditated and the latter are spontaneous. That distinction doesn’t hold in every case, but it is a good starting point. For instance, Travis Ikeguchi noticed a gay pride flag hanging in front of Lauri Ann Carleton’s store in Cedar Glen, California, on August 18, 2023. Ikeguchi was opposed to homosexuality and started an argument with Carleton about her flag, and he eventually shot and killed her. Ikeguchi committed murder that was plausibly a hate crime, but it was not politically motivated terrorism because it resulted from an argument and had no wider goals.¹³ Another distinction is that rage and a sudden desire to violently vent it usually inspire hate crimes. Those features typically manifest as characteristics of the crime or attack, such as the opportunistic choice of a target rather than a planned violent act. Regardless, distinguishing hate crimes from politically motivated terrorist attacks remains a challenge and may be the source of errors in the dataset.

To count as a politically motivated terrorist attack, the killing must be independently premeditated and motivated by the ideological and/​or political beliefs of the attacker. The attack must also not stem from another dispute that then results in a killing, even if the killer has strong political opinions or is a member of a terrorist group. For example, a follower of an extremist ideology who is detained by law enforcement for an unrelated legal infraction and ends up in a shoot-out with the police did not commit a politically motivated terrorist attack.¹⁴ Right-wing self-proclaimed “sovereign citizens” Joe and Jerry Kane were pulled over by Memphis, Tennessee, police in May 2010. During the traffic stop, Joe fired an AK-47 rifle at the police, killing two officers. The duo later died in another shoot-out.¹⁵ They are not included as politically motivated killers.

The political ideology of the attackers was independently checked and coded by four researchers who acted without collaboration during the coding process, and without sharing their reasoning, evidence, or judgments with each other prior to completing the task. They were given similar research on the attackers, such as police statements, manifestos, and news stories, and allowed to gather additional evidence at their discretion. The researchers are the author, two former interns at the Cato Institute, and an independent undergraduate researcher from Dartmouth College.¹⁶

To test the reliability of those judgments, we ran several standard intercoder checks to measure whether the coding methodology is reliable and not dependent on any single coder’s potentially biased judgment. The coders agreed 95.7 percent of the time, but the raw percentage of agreement between coders doesn’t control for chance alignment. Thus, we use Cohen’s kappa as a chance-corrected measure. It estimates agreement according to

where P₀ is observed agreement and P_e is expected agreement under random coding.¹⁷ The estimated result here is K ≈ 0.939, which indicates near-perfect agreement and is consistent with Krippendorff’s alpha, another measure that is robust to missing data and uneven category distributions.¹⁸ Scott’s pi, another method that relies on pooled category proportions rather than coder-specific margins, returned an almost identical estimate (π ≈ 0.939).¹⁹ The convergence of these different statistics is evidence that coder agreement is not an artifact of the underlying distribution of ideological categories.

The remaining disagreements were mostly limited to boundary cases where the coders classified foreign nationalism as right or, in a few instances, as unknown/​other. No category was large enough to inflate agreement, and the coders were in perfect agreement for rare categories such as separatism (n = 3). Even if every disputed case were recoded against my judgment, the ideological distribution of the attackers would not meaningfully change.

This policy analysis is an intellectual descendant of several of my previous papers about foreign-born terrorism in the United States, another paper on mass shootings that was published in 2025, and several articles I wrote that same year.²⁰ The methodology sections of those papers are very similar, but there is a critical difference here that is intended to reduce the incidence of overinclusion. When studying foreign-born terrorism, failure to include terrorists was a major potential problem because the downside of underestimating the risk was greater than any other potential problem. Based on conversations with the Ohio State University political scientist and Cato colleague John Mueller, the foremost quantitative expert on terrorism, in my previous papers I erred on the side of overinclusion. If a foreign-born criminal was possibly a terrorist or there was excellent evidence to support it, that person was included to eliminate the possibility of underinclusion. In such a case, it was important to assign a high risk weighting to systematically bias the methods against my prior belief that foreign-born terrorism was not a serious problem that justified immigration restrictions or further security actions.

However, such a simple method of overinclusion is inappropriate when studying deaths in politically motivated terrorist attacks on US soil in recent years because it can lead to ideologically biased results. Most readers of Cato’s earlier papers about foreign-born terrorism were primarily interested in the annual chance of being murdered in an attack, but readers of analyses of terrorists’ political motivations are more concerned with the proportions of attacks that are credited to various ideological motivations. Thus, my goal in this policy analysis is to eliminate overinclusion and only count attackers who are definitely politically motivated. Earlier and rougher versions of this analysis implicitly used the overinclusion method, which led to some errors that were identified by internet commentators.²¹ Even though the magnitude of those errors ultimately accounted for less than a 1 percent difference in the final number of victims, a slight net reduction in the number of victims of right-wing terrorists, and a small increase in the number of people murdered by left-wing terrorists, the overinclusion principle is dropped in this analysis.

This analysis focuses on killings because they are the best measure of violence. The number of deaths in a politically motivated attack is affected by the intent of the attackers, the degree of their preparation and competence, the response from law enforcement and emergency services, and chance. Killing is difficult, and it stands to reason that attackers who intend to cause many deaths are more likely to do so. Thus, the number of deaths is a decent measure of the deadly intent of the attacks, even if it is not a perfect one.²² Each murder is the same as another in terms of the number of lives lost. It’s trivially easy to compare the severity of an attack by comparing the number of people killed, which is why journalists, historians, and ordinary people compare deaths rather than other measures when the number of deaths are available. The final reason for choosing the number of people murdered in attacks as the dependent variable is that it’s the easiest to uncover and the least likely to be misreported or underrecorded.

There are alternative measures of politically motivated violence that have some merit. One such measure is injuries in attacks, which can include wounds as varied as scratches and amputations. Another measure is simply using the number of attacks regardless of their extent. There are admirable attempts to count the number of incidents that could include successful attacks, foiled attacks, attacks that resulted in no deaths, and any amount of damage, but the ambiguity is high in all these cases. Compare two situations: In the first, the police intercept an attacker shortly before they commit a violent action. In the second, the attacker gets cold feet at the last moment, retreats, and is then apprehended. Should those count the same? The dollar value of property damage from an attack would also seem to be a useful measure until you realize that three terrorist attacks—the 1993 World Trade Center bombing, the 1995 Oklahoma City bombing, and the 2001 9/11 attacks—collectively account for well over 99 percent of the property damage in politically motivated violence.

Weighting the murders of prominent people such as Kirk or the attempted assassination of Trump more heavily than the victims of a mass shooting is intuitively appealing when judging the intensity of politically motivated violence. To put it simply, some people are more politically significant than others and their murder can cause serious problems, whereas a greater number of victims in politically motivated mass shootings has less of an effect. The 1960s assassinations of President John F. Kennedy, Rev. Martin Luther King Jr., and Sen. Robert F. Kennedy Jr. were just three murders in a turbulent decade, but they shaped political trends, fomented widespread riots, and prompted other forms of political violence that affected millions. If there were a consistent and transparent means to weight those lives more heavily when judging the political effects of violence, then it would be worth employing that method, but such a means is absent here.

Another possible measure would be to use the value of a statistical life, a measure employed by economists to gauge the efficiency of safety procedures that have costs. A value of a statistical life uses a population’s revealed preferences on safety spending to estimate how much it values a human life. However, for the purpose of this study, monetizing the loss of life would provide no additional insight: The distribution of costs from such a calculation would be the same as the number of lives taken, but expressed in a different unit. A related measure, the number of quality adjusted life years (QALYs) lost in an attack, is a theoretically attractive alternative measurement because it would take account of both the quality and quantity of life destroyed in an attack. QALYs are often used in government health care systems to evaluate the efficiency of policy changes, new medicines, and medical procedures.²³ The more high-quality years that are lost, the more damaging the attack. Such a calculation requires highly detailed health and age information about victims and statistical estimates of life expectancy, but it would be a handy and defensible measure for this analysis if the cost of calculating it were lower.

The Chicago Project on Security and Threats (CPOST) catalogs and analyzes threats and attacks against members of Congress that are severe enough to be prosecuted.²⁴ While useful, it’s a sample of a small number of victims who are less prominent than executive branch officials and other national figures. Threats against prominent individuals such as Kirk, for instance, wouldn’t be recorded in CPOST even though they are prominent targets of politically motivated attackers. CPOST includes murders, injuries, and threats—a combination that reduces the precision of comparisons. How many threats equal one physical assault? How many physical assaults equal one murder? There are ways to estimate that, but they are all subject to contestable assumptions. As anyone who has received many threats can tell you, the law enforcement decision to investigate and prosecute them is uneven. Thus, what is prosecuted is certainly not a random sample. Lastly, a threat followed by an attack that results in a murder indicates a more motivated killer than a threat alone, so a murder by itself matters more than a threat or even an attempted attack. Aggregating threats, assaults, and murders into a single metric blurs the distinction between rhetoric and violence. Counting them together introduces statistical noise while providing less underlying signal of political violence.

A different measure would use survey data on Americans’ attitude toward the acceptability of terrorist behavior. Results from several polls and surveys purport to show that Americans’ level of support for politically motivated violence is high and rising, particularly among Democrats, liberals, and left-wing respondents.²⁵ Vice President J. D. Vance argued that such polling results are evidence of a surge in left-wing violence when he hosted Kirk’s podcast after the advocate’s murder. “Political violence, it’s just a statistical fact that it’s a bigger problem on the left,” said Vance.²⁶ Contrary to the claims of the vice president, the polling and survey data are not clear, nor can they be relied upon even if they were.

There are strong general criticisms of surveys and polls that undermine their general reliability and reveal the limitations of their ability to discover useful knowledge.²⁷ Surveys and polls often produce data that offer no real insight into respondents’ thinking because the underlying assumptions about respondent truth-telling, comprehension, and shared meaning only occasionally hold in practice. Some of the problems with survey and poll data are satisficing (quickly choosing a survey response that is roughly correct instead of the one that is most fitting); lack of sincerity (known as the Lizardman Constant); dishonest motivation; culturally loaded wording such as “politically motivated violence” that can have radically different meanings across individuals and contexts; and inaccurate respondent perception. In addition, we should be generally skeptical of results that purport to measure things that do not actually exist in an independent measurable form, such as support for violence.²⁸ By contrast, asking people about the number of cars they own or whether they are actually lizardmen can be independently verified through administrative data or outside observation.

The specific poll and survey questions attempting to gauge support for political violence are especially poorly designed, even beyond the general criticisms of polls and surveys. Ohio State University professor Ryan Kennedy is the most prominent critic of them.²⁹ The central problem is that the researchers who ask the questions and the respondents who answer them don’t understand the questions in the same way, which means that both the researchers and readers of the results, like Vance, tend to misinterpret the results.

Kennedy tries to resolve the misinterpretation problem via a method called cognitive interviewing. He starts this by first asking respondents about their support for political violence by repeating questions from other surveys. He then asks the respondents what they were thinking when they answered those questions. Many supporters of political violence were thinking of using legal force or violence such as court orders, forcing a president out of office if they seized power in a coup, or violence as a last resort after legal methods had first been tried, such as in the American Revolution, or as violent resistance to a dictator, such as the French Resistance fighting against Nazi occupiers.³⁰ Cognitive interviewing reveals that both left-wing and right-wing support for politically motivated private violence is substantially lower than initially reported, and it also shows that right-wing support for using the military to crack down on protesters is also greatly exaggerated.³¹ These surveys and poll questions also exaggerate support for violence by ignoring the effects of partisan response bias that emerge from timing when the questions are asked shortly after political violence that is frequently partisan or perceived as such.³² Other experts share Kennedy’s skepticism of the results from polls and surveys of political violence.³³

Besides cognitive interviewing, other studies of support for political violence ask more refined questions about violence by prompting respondents to think about the partisan identities of violent actors and to separate murder from other forms of real violence, which is wise because reasonable people think murder is worse than assault.³⁴ The result of those superior lines of questions is less support for political violence, but those studies predate the current panic.³⁵ The relative difference in partisan responses also varies substantially based on nuances in question asking that should not produce such wildly different results if there were an actual underlying level of support for politically motivated violence that varied with partisanship.³⁶ Attempts to measure opinions qua opinions will reveal less satisfying results than analyzing the observable violent actions that people take.

In addition to the reasons above for using the number of murders to measure politically motivated violence, there is a practical justification that is almost as important: Counting murders is easier, more accurate, and less subject to bias or error than any other variable of violence. Murders are the crime least likely to be miscounted; they are well-reported because they are the most serious crime, there’s little dispute over the number of murder victims, and it’s easier to discover errors when the number of murder victims is miscounted.

The different measures mentioned above, such as injuries, incidents, property damage, QALYs lost, and chargeable threats and assaults against Congress, are all valuable indicators of politically motivated violence. Poll and survey results are not revelatory and are thus ignored in the rest of this analysis. The valuable indicators of politically motivated violence other than murder are variable, difficult to measure, hard to compare between attacks, challenging to calculate because they require several tendentious assumptions, or otherwise inferior to measuring the number of deaths. The homicide rate is the best measure of crime in a locality because it’s the most serious offense, and it’s correlated with other serious crimes. The same holds for gauging the frequency and severity of politically motivated violence.

The distribution of murders in politically motivated attacks limits the types of statistical analyses that are possible. Just two attacks account for almost all deaths: The 9/11 attacks resulted in 2,979 fatalities, and the Oklahoma City bombing caused 168 deaths, which together account for 88 percent of all the deaths in the entire dataset. Relatedly, there are many fatal attacks with very few deaths. About 56 percent of the politically motivated killers murdered just one person, and 73 percent murdered fewer than three. Just 10 percent of all the terrorists are responsible for 88 percent of deaths. The median is 1 murder in such attacks, and the mean is 17, which together statistically indicate how skewed the data are.

The full data are presented in the first few exhibits below, with the 9/11 and Oklahoma City attacks included, but those attacks are removed in subsequent tables and figures because they are such outliers that it’s impossible to analyze recent trends with them included. Every politically motivated instance of violence looks like nothing next to 9/11, and only a small blip next to the Oklahoma City bombing. The problem of how to remove extreme events arises because the skewed distribution of deaths in attacks makes the standard deviation a useless measure of data dispersion. The standard deviation treats large and small deviations as roughly symmetrical when the data are distributed normally in a bell curve, but that’s not the case in this dataset because the 9/11 attacks would expand the variance so much that every other attack would look typical by comparison. These data are in what analysts call a power law distribution.

Thus, this policy analysis uses a robust outlier detection method based on the median absolute deviation (MAD) to identify extreme values and avoid distortions. This is a robust measure of spread that is simple to calculate and reports a modified Z‑score for attacks rather than individual attackers. This analysis switches to attacks at this point, but the results are similar if individuals remained the unit of analysis. The modified Z‑score is calculated as follows:

Where y_i is the natural log of fatalities for attack i. The denominator MAD_y is the median of the absolute deviations: MAD_y = median(|y_i – median(y)|). The constant 0.6745 rescales the statistic so it’s roughly comparable to a standard deviation if the data were normally distributed. Attacks with |z_i| > 3.5 are classified as extreme tail events, meaning they’re much deadlier than nearly all other attacks. These extreme tail events, or just extreme events, would be called outliers if the data were normally distributed.

A Z‑score is simply a means to describe how far the number of fatalities in a specific attack sits from the middle of the distribution of attacks, which is measured in units of typical variation. For instance, a score of 0 means an attack was close to the median, a score of 1 means it was about one typical step more lethal, and a score 3.5 or higher means it was far beyond the ordinary range. It’s like a standardized distance from the normal number of deaths. Here, it translates a messy and skewed set of fatality numbers into a single comparable measure of how extreme each attack was.

The intent of MAD-based robust Z‑scoring is to systematically, transparently, and defensibly separate the rare extreme events from the common low-fatality cases that are so common in politically motivated terrorist attacks. The dataset covers the entire population and is not a sample, and the goal is not to estimate an unobserved universe of attacks, so there’s no need for the usual statistical corrections for sampling error. Only the 1995 Oklahoma City bombing (z = 4.31) and the 9/11 attack (z = 7.31) have Z‑scores above 3.5, so they are systematically excluded from later tables and findings below.

Given the rarity of attacks even after excluding extreme events, we use a negative binomial regression to assess whether the frequency of politically motivated attacks has changed over time. Attacks are defined as distinct calendar days on which politically motivated violence occurred, with multiple perpetrators treated as a single coordinated event. Attacks are aggregated into triennial periods beginning in 1975. The model does not speak to fatalities, severity, or the number of perpetrators. It is well suited to the data, as the number of attacks per period exhibits substantial overdispersion, a common feature of rare and episodic forms of political violence that violates the assumptions of simpler count models. Formally, the model estimates

Where A_t represents the number of attacks in period t, and t is a linear time index.

Under this specification, the estimated coefficient on time is positive in magnitude, corresponding to an incidence rate ratio of approximately 1.07 per three-year period, but it is not statistically significant. Accounting for overdispersion, there is no statistically significant evidence of a systematic increase in the frequency of politically motivated attack events over time. In summation, the estimated time trend is positive but not statistically distinguishable from zero.

The identities of the attackers, the number of people they murdered, their ideological motivations, and other information were gathered from about 20 different sources, including research by John Mueller;³⁷ US Department of Justice reports;³⁸ press releases, statements, or speeches issued by the Department of Justice;³⁹ Fordham University’s Center on National Security;⁴⁰ several Congressional Research Service reports;⁴¹ the RAND Corporation’s Database of Worldwide Terrorism Incidents (RDWTI) and other RAND publications on terrorism;⁴² the Global Terrorism Database (GTD) maintained by the National Consortium for the Study of Terrorism and Responses to Terrorism at the University of Maryland, College Park, and other research produced by the GTD;⁴³ numerous Freedom of Information Act requests by various organizations and individuals asking for all terrorism-related convictions since 9/11;⁴⁴ the New America Foundation; The Intercept; the Investigative Project on Terrorism; the research of University of North Carolina sociology professor Charles Kurzman; the George Washington University Program on Extremism; the National White Collar Crime Center; the Terrorism Research Center at the Fulbright College of Arts and Sciences at the University of Arkansas; a dissertation by Catlyn Kenna Keenan; numerous FBI reports from 1982 to 2005 on terrorist incidents in the United States; the Prosecution Project;⁴⁵ the Southern Poverty Law Center;⁴⁶ the Center for Immigration Studies;⁴⁷ the Center for Strategic and International Studies;⁴⁸ and news stories, articles, and email correspondence.

This section contains three iterative refinements of the data that progressively remove attacks where the number of deaths qualifies as an extreme event. The first iterative refinement includes all deaths from politically motivated terrorist attacks since 1975. The second iterative refinement removes deaths from the 9/11 attacks. The third iterative refinement removes both the 9/11 attacks and the Oklahoma City bombing. The third is the preferred analysis, although the first refinement is useful because it contains the entire population of people murdered in politically motivated terrorist attacks on US soil.

This dataset identified 208 terrorists who murdered 3,577 people in politically motivated attacks on US soil from the beginning of 1975 through the end of 2025. During this 51-year period, the annual chance of being murdered in such an attack was about one in four million per year, but with substantial variation by motivating ideology. The annual chance of being murdered in such an attack isn’t as useful a metric when it comes to rating the danger of politically motivated violence as it is for normal homicide or other terrorist attacks because a disproportionate number of the victims of politically motivated killings are prominent individuals such as Charlie Kirk. Still, it is a means of showing relative risks. Islamists were the deadliest terrorists by ideology, responsible for 3,118 killings, or 87 percent of the total; followed by right-wing terrorists with 363 killings, accounting for 10 percent; left-wing terrorists with 68 killings, accounting for 2 percent; and the other ideologies combined, accounting for about 1 percent (Table 1).

The largest attacks are 9/11 and the Oklahoma City bombing, with 2,979 and 168 victims, respectively. As mentioned above, they account for 88 percent of all murder victims and are the two most extreme attacks in this dataset. Their inclusion biases the results to such a degree that Figure 1 is almost worthless because there are only two clearly discernable blips, one for each extreme event. (Tables 2 and 3 and Figures 2 and 3 will gradually remove those extreme events in two subsequent iterations to show how deaths in politically motivated attacks have changed over time.)

Shorn of the most extreme values of the 9/11 attacks, Table 2 shows that 189 killers murdered 598 people during the entire 51-year period. Another effect of removing 9/11 is the reduction in the Islamist share of all deaths from 87 percent of the total to just 23 percent. Similarly, the share of victims from right-wing attacks grew from 10 percent to 61 percent, the highest of all murders. The share of left-wing attacks also increased more than fivefold, from 2 percent to 11 percent. Figure 2 is more useful than Figure 1 because the annual variation since 2009 really shows up when the deaths from the 9/11 attacks are removed, but the death toll from the Oklahoma City bombing still dominates.

Table 3 removes the extreme events of 9/11 and the Oklahoma City bombing, which allows readers to see how deadly typical attacks are, as well as their ideological motivations. This reduces the total number of killers to 186 and total number of deaths to 430. The only decline in ideological killings from this refinement is for those committed by right-wingers because that was the ideological motivation of the Oklahoma City bombing. Excluding its 168 victims reduces the number of right-wing victims by about 45 percent and means that right-wing terrorists are now responsible for 195 deaths and 45 percent of the total. Islamist terrorists are the second deadliest, followed by left-wingers. That excluding two extreme events so radically changes the number and ideological composition of victims in politically motivated terrorist attacks is evidence that this is a very small problem with relatively few attacks and victims. If politically motivated terrorism weren’t such a small problem, the exclusion of two events wouldn’t have such a radical effect.

Table 4 displays the number of killings in politically motivated terrorist attacks, the chance of being killed in such attacks, and the motivating ideology of the attackers by the number of victims in three-year periods since 1975, excluding the extreme events of the 9/11 attacks and the Oklahoma City bombing. The number of deaths in each three-year period and the triennial chance of being murdered are the best measures of the risk over time. In that light, the most recent period, from 2023 to 2025, is more dangerous than average, with a triennial chance of death in such an attack at around 1 in 19.1 million, compared to 1 in 33.1 million over the entire period. Notably, the 2014–2016 and the 2017–2019 periods were more dangerous, with more total killings and a higher annual chance of being murdered (Table 4). The results are about the same if the data are organized on a two-year, three-year, five-year, or any other yearly period.

Figure 3 graphically displays the deaths by ideology on a triennial basis as organized in Table 4, focusing on those who were murdered in politically motivated attacks committed by Islamists, right-wing terrorists, and left-wing terrorists. The period 2014–2016 had the greatest number of deaths because the two extreme events of 9/11 and the Oklahoma City bombing are excluded. Right-wing terrorists were the deadliest in eight of the periods; they tied with separatism and foreign nationalism in one period, Islamists in five periods, and left-wing terrorists in three periods.

Most readers will be interested in the share of attacks committed by left-wing and right-wing terrorists. Their interest is in contrast with the topline results that there are few politically motivated murders and that Islamists are responsible for most such killings since 1975, when the 9/11 attacks are included. With both the extreme events of 9/11 and the Oklahoma City bombing excluded, as in the preferred iteration, right-wing politically motivated terrorists murdered the most people, followed by Islamists, and then left-wing terrorists. The obsession with the share of people killed by right-wing or left-wing terrorists is due to the divisive current political atmosphere. Regardless, the numbers are small.

Perhaps the earlier and more dangerous periods felt less divisive because Islamists accounted for almost half of all deaths in attacks in 2014–2019 and only 30 percent in the most recent triennial period of 2023–2025 (Table 4). Islamism is a foreign ideology that is aesthetically and intellectually un-American in most ways, even if some of the deadliest attacks were committed by American-born adherents. But there is no such comforting thought when it comes to analyzing deaths in right-wing or left-wing terrorist attacks. When Utah Gov. Spencer Cox (R‑UT) announced the arrest of Tyler Robinson, the 22-year-old Utah resident accused of assassinating Charlie Kirk, he said: “For the last 33 hours, I had been praying that this person (who murdered Charlie Kirk) was from another country. That he was not one of us because we are not like that. But it was one of us.”⁴⁹ Cox’s statement was quite unusual, and many interpreted it as an anti-immigrant statement, but it was more likely just the surprised exclamation of a man who realizes he can’t blame this problem on outsiders and that his society must instead deal with its own deformities.

Right-wing terrorists were responsible for 195 murders over this period, or the plurality of all killings in the preferred analysis, where the extreme events of 9/11 and the Oklahoma City bombing are excluded. Seventy-five percent of those murders occurred after 2008, the year that Barack Obama first ran for and won the presidency. Left-wing terrorists murdered 68 people in their attacks during the entire period, 62 percent of them since 2008. The ratio of deaths in right-wing politically motivated terrorist attacks to left-wing politically motivated terrorist attacks is 2.9 to 1 over the 51-year period, and 3.5 to 1 since 2008. On a relative basis, deaths in right-wing terrorist attacks have increased relative to left-wing terrorist attacks.

The share and number of deaths in attacks committed by right or left terrorists is extremely unstable, fragile, and could be reversed by a single attack. There are just not that many people killed in terrorist attacks in the United States. As a result, political partisans should not use the above numbers to bash their political opponents. It would be dishonest and mathematically myopic to pretend to understand any profound truths about one’s partisan political opponents or one’s partisan allegiance based on a few hundred politically motivated murders over a period of 51 years. The paucity of deaths by right or left terrorists, together or separately, does not justify a domestic security crackdown or new war on terrorism. The first war on terrorism was a grossly disproportionate, destructive, and net-deadly overreaction in the United States to the minor threat of Islamist terrorism, to say nothing of the harm done overseas. A similar overreaction to the smaller amount of politically motivated domestic killings would add farce to tragedy.

There are several killers whose additions or omissions from this dataset are controversial. Below are detailed explanations of various inclusions, exclusions, specific classifications, and other coding choices. These coding decisions are subject to change and correction based on additional evidence.

The most prominent controversial inclusion is Colin Ferguson, a Jamaican immigrant who murdered six people in a shooting on the Long Island Rail Road on December 7, 1993. Noted—and perhaps notorious—defense attorneys William Kunstler and Ronald Kub defended Ferguson, at first by attempting to introduce a “black rage” defense on his behalf, arguing that anti-black racism had driven him insane so that he did not act willfully when he opened fire on strangers packed in a railroad car during the evening rush hour.⁵⁰ Ferguson disagreed with the black rage insanity defense and instead argued that he was framed, even though there were dozens of eyewitnesses who saw him open fire. His legal strategy was perhaps evidence that he was insane.

The black rage defense did not spring entirely from the minds of Kunstler and Kuby. The first source was a 1968 book called Black Rage by psychiatrists William H. Grier and Price M. Cobbs. The authors argued that supposed negative psychological damage from white racial oppression caused deep psychological scars on some black Americans that manifest in anger, alienation, and violence.⁵¹ Besides the intellectual basis of a floated defense in Ferguson’s case, the book’s other infamous impact was to provide an ideological basis for Mark Essex’s black nationalism–inspired murder of nine people in a sniper attack in New Orleans in 1973.⁵²

The second, and more important source, of the black rage defense was notes written by Ferguson that police discovered in his pockets after his arrest that expressed serious and deep-seated hatred of whites, Asians, and “Uncle Tom Negroes.”⁵³ Witnesses reported that Ferguson yelled about “black rage” and “how it will get you” in a separate incident prior to the shooting.⁵⁴ Black Rage was rereleased with a new introduction by the authors in 1992, a year before Ferguson committed his mass murder, which can possibly explain why he used the phrase when he did. His statements prior to the attack that seemingly reference the book Black Rage and the notes in his pockets during the attack are good evidence that it was politically motivated terrorism.

The most controversial inclusion is Roland James Smith Jr., who opened fire at Freddy’s Fashion Mart in Harlem and then set the building aflame on December 8, 1995, murdering seven people. The police treated Smith’s actions as normal crimes rather than a terrorist attack or hate crime, but he was a protester who was part of a crowd inflamed by black activist Al Sharpton. Sharpton had organized an anti-business protest and addressed the crowd, saying, “We will not stand by and allow them to move this brother so that some white interloper can expand his business.”⁵⁵ Smith was perhaps inspired by those words, his own long-held opinions about race, or the recent loss of his own street vending stall and home to arson.⁵⁶ Regardless, he yelled, “Brothers, get out! It’s going to go down!” when he entered the store, and that’s sufficient evidence, along with his history, to conclude that the violence was politically motivated.⁵⁷

Another controversial inclusion is Robert Aaron Long, who murdered eight people in attacks on three Atlanta-area spas on March 16, 2021. He committed the crimes because he was a Christian religious fundamentalist who wanted to eliminate temptation, in addition to ending his own sex addiction, by targeting public murders meant to intimidate others.⁵⁸ He was indicted in Fulton County on state charges of domestic terrorism and other charges such as murder.⁵⁹

The most recent controversial inclusion is Rahmanullah Lakanwal, an Afghan immigrant and former member of a CIA-funded Afghan Zero Unit who fought in Afghanistan. He opened fire on two West Virginia National Guard members in Washington, DC, on November 26, 2025. Prosecutors allege he yelled “Allahu Akbar” during the attack, which raises the high possibility of an Islamist motive.⁶⁰ However, unlike other attacks whose ideological commitments are documented in manifestos, notes, or prior statements, there is at present no publicly available evidence of ideological justification in any of Lakanwal’s writings, communications, or affiliations. Prosecutors have not filed terrorism charges yet, and the charging documents contain no discussion of ideology. Federal investigators have also indicated that Lakanwal suffered from depression, struggled with war trauma, and became increasingly withdrawn in the years before the attack.⁶¹ This lack of affirmative evidence distinguishes his case from clearer Islamist attacks in the dataset and produced an internal coding disagreement, the main issue being whether the shouted phrase alone is sufficient to place him in the Islamist category. As such, his classification here as motivated by Islamism is provisional and reflects a prediction of where the case may ultimately end up rather than a confirmed motive.

The exclusion of other killers from the dataset may seem controversial but they are explained here.

Devin Patrick Kelley was an abusive ex- and estranged husband, a court-martialed enlistee in the Air Force, and a vocally anti-Christian online poster who murdered 26 people in a shooting at the First Baptist Church in Sutherland Springs, Texas, on November 5, 2017. Kelley was critical of Christianity but that did not motivate his church shooting. His estranged wife and her family often attended the First Baptist Church. Shortly before the shooting, Kelley sent threatening text messages to his mother-in-law and went to the church to murder her and others.⁶² His mother-in-law was not attending services that day, but her mother was, and Kelley killed her. Freeman Martin, a regional director for the Texas Department of Public Safety, said, “This was not racially motivated. It wasn’t over religious beliefs.… There was a domestic situation going on with the family and in-laws.”⁶³

Another prominent exclusion from this dataset is Jared Lee Loughner, who murdered six people and injured member of Congress Gabby Giffords (D‑AZ) in an attack in Tucson on January 8, 2011. Giffords was a target because she was an elected politician, but Loughner did not target her for political reasons; he did so because he believed in incomprehensible conspiracy theories and had paranoid schizophrenia.⁶⁴ The latter likely influenced the former. Since the motivations of the attacker and not the identity of the intended target are what matter, Loughner’s murders are not terrorism.

Richard Andrew Poplawski killed three police officers and wounded two others in Pittsburgh on April 4, 2009, after his mother called the police asking that he be removed from her property. Poplawski was a white nationalist and a fan of Nazi conspiracies, but those did not seem to have inspired him to attack the officers, and so he is excluded from the dataset.

Five members of a cult centered around the writings of an artificial-intelligence doomer named Jack “Ziz” LaSota, and that also held uncommon beliefs about animal rights, theories of the brain, and veganism, were involved in at least two murders and as many as four, as well as the deaths of several members of their own group.⁶⁵ Cult member Maximilian Snyder is alleged to have murdered Curtis Lind, the landlord of cult members Alexander Leatham and Suri Dao, who in turn are facing murder charges in the death of another Zizian cult member. Another member, Michelle Zajko, may have murdered her parents in late 2022, although she has not been charged with the crime. In early 2025, cult members Teresa “Milo” Consuelo Youngblut and Felix “Ophelia” Bauckholt started a gun battle with Border Patrol officer David Maland in Vermont during a traffic stop, killing him. Although the Zizians have extremely strong ideological opinions, especially about the science fiction scenario of large language models taking over the world and enslaving or killing humanity, they are excluded from the dataset because they did not commit their crimes to further their unusual ideology.

Andrew Joseph Stack III published a six-page manifesto before crashing his airplane into an Internal Revenue Service building on February 18, 2010. It was a poorly written and populist explanation for his crimes, inspired by both left and right ideas, with the latter dominating. Populism was overwhelmingly a right-wing opinion at the time of Stack’s attack during the heyday of the Tea Party movement. Many of his anti-IRS, anti-political, and anti-institutional statements were right-wing. In addition, he made many statements consistent with right-wing political opinions. Some of his statements were also left-wing, and they have been used by commentators to dismiss the claim that Stack had any right-wing political inspiration, but those comments are also mostly consistent with a right-wing populist perspective.⁶⁶ This study classifies Stack as being a right-wing terrorist, but readers can move his single victim from the column of right-wing victims in terrorist attacks into another column if they are convinced otherwise.

Thomas Matthew Crooks shot Donald J. Trump at a presidential campaign rally near Butler, Pennsylvania, on July 13, 2024. Crooks wounded Trump, several bystanders, and killed rally attendee Corey Comperatore before being shot and killed by the Secret Service. The mystery is why Crooks attempted to assassinate Trump. He did not leave behind a manifesto, political justification, or other clear political reason for his actions.⁶⁷ Since he tried to assassinate Trump, it would seem clear that he disagreed with Trump’s politics, but Crooks also searched for nearby political appearances of President Joe Biden, FBI Director Christopher Wray, and Democratic Attorney General Merrick Garland.⁶⁸ He was a registered Republican, contributed $15 to the left-wing Progressive Turnout Project in 2021, and didn’t appear to have strong ideological or political opinions.⁶⁹ It’s tempting to classify Crooks as a shooter inspired by left-wing political ideology because he tried to assassinate a conservative Republican candidate, but there’s no good direct evidence that his shooting was politically motivated. His search for any prominent politician of either political party to assassinate may even be evidence that he was afflicted by Herostratus syndrome—the desire to become famous by committing an infamous act—rather than be driven by ideology.⁷⁰ As a reminder, it is the killer’s motivation that is the subject of this study, and though that is often related to characteristics of the target, it isn’t always.

Nobody disagrees that Islamist terrorists have murdered more people on US soil than killers motivated by all other ideologies combined when all attacks are included. Similarly, nobody disagreed with my overinclusive terrorist-identification methods in earlier research on foreign-born terrorism. However, some readers may object to the finding that right-wing politically motivated terrorists have murdered more people in attacks than left-wing terrorists in every iteration above.⁷¹ Few critics take issue with the specific methods, but instead disagree with how a few of the terrorists above are ideologically classified or whether they should have been included at all.

Reasonable people can always disagree over the inclusion and ideological classification of certain individuals. Since the publication of my first draft of this report on Cato’s blog, I’ve removed and added attackers who account for 21 murders based on feedback from readers, comments from other scholars, an examination of other sources, and changes to methods to reduce overinclusion.⁷² Those changes have not meaningfully shifted the relative shares of people killed by terrorists among the different ideological motivations, but the total number is smaller. I also included the intercoder reliability test mentioned in the methodology section to guarantee that coder bias isn’t driving results.

Other reports about politically motivated attacks come to similar conclusions, although they often use different data and methods. A 2024 report by the National Institute of Justice, a research agency within the US Department of Justice, found that right-wing extremists have murdered 520 people in attacks since 1990, far more than the 78 people murdered by left-wing extremists.⁷³ The National Institute of Justice report covers a shorter period but, if anything, argues that my research undercounts the number of victims of right-wing terrorism far more than left-wing terrorism. It derives a ratio of 6.7 right-wing murders for every left-wing murder—much higher than the 5.3 to 1 ratio this paper calculates when all attacks are included. Another paper relying on data from the US Extreme Crime Database finds that right-wing terrorists commit about 5.3 homicides in attacks for every killing by left-wing terrorists, the same ratio as found here when the Oklahoma City bombing is included.⁷⁴ The number and ideological share of victims of right-wing attacks recorded elsewhere is generally greater than here.⁷⁵

Another criticism is that this research does not include people who were killed in riots with a plausible political motive, such as the Black Lives Matters (BLM) riots that followed the murder of George Floyd in 2020. The 19 most prominent homicides in the BLM riots were mainly looters shot by the police, shot by property owners, or victims of normal crimes.⁷⁶ It would be nonsensical to count the lawful killing of a looter as a politically motivated left-wing murder, especially if the looter were a politically motivated left-wing looter. In such a case, the looter is the attacker, not the victim. Some of the critics who want looters in the BLM riots to be counted as victims of left politically motivated violence also want me to include Ashli Babbitt as a victim of left-wing violence, even though a police officer shot her while rioting at the Capitol on January 6, 2021. Including neither makes sense.⁷⁷

Furthermore, politically motivated terrorist killings cannot be committed by state actors. Thus, the police cannot commit politically motivated terrorist killings if they’re doing their jobs, as they were during the BLM riots and while defending the Capitol on January 6, 2021. If a police officer were operating independently as part of a paramilitary organization, a member of a terroristic conspiracy, or murdered somebody for political reasons while off duty, then the killing could be politically motivated terrorism. If a police officer assassinated somebody on orders from their bosses or the president, it would not be a politically motivated terrorist killing because the government did it and not a private actor. President Trump ordered the US military to drop bombs on boats in the Caribbean, claiming that they were drug smugglers and terrorists. These killings are probably a war crime, illegal, unethical, and bad for many other reasons, but they are not terrorism nor were they committed by a politically motivated killer because President Trump and the US military are state actors who issued orders and then followed them, respectively.⁷⁸ It bears repeating that ultimate judgment of the ethics or efficiency of any action does not depend on whether it is a state or nonstate actor, but it does matter when identifying the action as politically motivated terrorism or not.

Critics who level the above objections want me to include many deaths that are downstream from political decisions. Such a methodological adjustment would amount to a statistical analysis of the effects of public policy on death rates more broadly, rather than a count of politically motivated killings. Such a macrosocial analysis is likely impossible, and any results would be uncertain because consideration of other tendentious social science theories would depend heavily on the assumptions. The critics are very confused here and didn’t think through the issue. Many of them want to judge whether a killing was politically motivated based on the politics of the victim or their partisan view of different rioters. On the contrary, the only way to understand the motivations of the killers is to investigate the motivations of the killers.⁷⁹ Sometimes those are unknown even though there was likely a clear political motivation, as mentioned above in the case of Thomas Matthew Crooks who tried to assassinate Trump in 2024, but cases like those are rare.

This policy analysis starts in 1975, partway through a spike in left-wing terrorism that began in the 1960s. However, that late start date did not bias the results in favor of left-wing terrorists because many of those 1960s–1970s terrorist groups weren’t very deadly and many of them intended to destroy property and sow chaos rather than kill or maim, much like left-wing terrorists in more recent times.⁸⁰ For instance, the left-wing terrorist group the Weathermen (also known as the Weather Underground) didn’t kill anybody in their bombings. Three Weathermen terrorists died when a bomb they were constructing exploded in Greenwich Village in 1970, but they weren’t counted as victims because they died in their own attack. The Black Liberation Army was much deadlier, killing 16 people at most, and those people murdered after 1974 are included here. For instance, the May 19th Communist Organization, an offshoot of the Black Liberation Army whose members included a few former Weathermen, murdered a Brink’s security guard and two police officers in 1981, and they are included in my data. The combined fatalities by Black Liberation Army and Weathermen terrorists were fewer than the 23 murdered by the right-wing nativist attacker Patrick Wood Crusius on August 3, 2019, at a Walmart in El Paso, Texas. Sometimes, when people mention the Weathermen, the thousands of bombings, and other terrorists of that period, they append the adjective “forgotten” to describe those campaigns of terror. Those attacks were forgotten mainly because there were so few killings.

One counterargument to my taxonomy is that right-wingers really don’t have an identifiable ideology like libertarianism or socialism. The closest is conservatism, but it is more of a disposition than an ideology with dogmas, according to the conservative philosopher Russell Kirk.⁸¹ Social psychologist Jonathan Haidt claims that people have slightly different innate moral foundations that make them unconsciously form their political intuitions, thus delineating them into left or right based on those differences.⁸² Economist Bryan Caplan’s simplistic theory of left and right holds that left-wingers dislike markets and that right-wingers dislike left-wingers.⁸³ Political scientist Verlan Lewis and his historian brother Hyrum Lewis argue that a political spectrum from left-wing to right-wing is an incoherent concept.⁸⁴ They believe that partisanship drives beliefs, not the reverse, and that most ideology is just post hoc justification, which is why the self-described right-wing and left-wing parties, politicians, and voters frequently switch positions on major policy issues. If the validity of a theory rests on its ability to correctly predict the beliefs and actions of different self-described political movements, then Caplan’s is the most tempting for conservatism, while the Lewises’ theory has much additional appeal. Thus, if we can’t reliably and consistently define what right-wing or left-wing ideologies are for the purposes of classifying the motivations of politically motivated killers because the concepts are so incoherent, much of this project’s results would be unreliable. No critics have made such an argument.

A total of 3,577 people were murdered in politically motivated terrorist attacks between the start of 1975 and the end of 2025. Under the preferred iteration of accounting that excludes the victims of the 9/11 attack and the Oklahoma City bombing, there were 430 victims. Of those, 45 percent were the murder victims of right-wing terrorist attacks, 32 percent by Islamist attacks, and 16 percent in left-wing attacks. The 2023–2025 period was the third-deadliest triennial period since 1975, with 53 murder victims. The other deadliest triennial periods were 2014–2016, with 110 victims, and 2017–2019, with 71. The annual chance of being killed in such an attack is about 1 in 33.1 million per year over the entire period, and 1 in 19.1 million during the last triennial period of 2023–2025. Those risks are dramatically smaller than the chance of being murdered in a nonterrorism homicide during that period, which is about 1 in 14,000 per year, and falls to about 1 in 20,000 per year in the latest triennial period. As a reminder, the above periods exclude the 9/11 attacks and the Oklahoma City bombing because they are extreme events or, under different statistical circumstances, known as outliers.

Every murder is a tragedy, but the victims of politically motivated terrorist attacks are socially tragic in a different way. They terrify Americans and have the potential to cause widespread civil disturbance, social disorder, or prompt authoritarian political and policy responses that would cause far more damage to the life, liberty, and private property of Americans. The Trump administration reacted to the murder of Charlie Kirk by reenergizing and refocusing the war on terrorism against domestic political opponents. Kirk and the other victims of politically motivated terrorism deserve justice, but Americans do not deserve to suffer under a new domestic war on terrorism in response to this small threat.

Nowrasteh, Alex. “Politically Motivated Killers: 51 Years of Terrorist Murders on US Soil, 1975–2025,” Policy Analysis no. 1012, Cato Institute, Washington, DC, February 19, 2026.