When Classification Becomes Reality

A claim has circulated for years in Germany’s political debate: Islamist attacks, or antisemitic offences committed by Islamists, are allegedly recorded as right-wing crime. If true on any meaningful scale, this would fundamentally undermine the country’s statistics on political violence. It would inflate the measured prevalence of right-wing violence, conceal Islamist violence and distort democratic debate at its statistical foundation.

The general claim, however, is false.

Germany’s system for recording politically motivated crime explicitly distinguishes between right-wing, left-wing, foreign-ideological and religious-ideological motivations. Islamist offences belong, in principle, to the category of religious ideology. Political violence rooted in a non-religious ideology originating abroad belongs to foreign ideology. The nationality, ethnicity or immigration status of the offender does not determine this classification. What matters is the motive attributed to the offence.

But the false general claim points toward a narrower, real and more interesting problem.

For at least part of the period examined here, German classification rules allowed certain antisemitic offences to be attributed to right-wing political crime when no contrary indications concerning the offender’s motivation were available. That rule did not mean that every antisemitic offence was classified as right-wing. Nor did it mean that known Islamist attacks were routinely moved into the right-wing category. It did, however, allow the absence of contradictory evidence to function as a substitute for positive evidence of a right-wing motive.

That distinction is the subject of this essay.

The question is not whether German crime statistics are fabricated. There is no evidence for that proposition. The question is how administrative classifications acquire the appearance of objective reality, how much uncertainty survives the publication process and what conclusions the resulting numbers can reasonably support.

The broader lesson reaches beyond crime:

Public statistics do not merely count events. They encode institutional decisions about what an event is.

The finding in plain language

Five conclusions can be defended from the available evidence.

First, Islamist attacks are not systematically recorded as right-wing violence. Germany maintains a separate category for religiously motivated political crime, and contemporary data show substantial numbers of Islamist and Middle East-related offences being placed in either the religious-ideology or foreign-ideology categories.

Second, a genuine classification problem existed in relation to antisemitic offences for which the motive was insufficiently established. The problem was not that authorities knowingly took proven Islamist offences and called them right-wing. It was that a default presumption could turn uncertainty into a positive ideological classification.

Third, politically motivated violence and general violent crime are different statistical universes. One is organised by political motive. The other is organised primarily by criminal offence. Nationality statistics from the general Police Crime Statistics cannot be inserted into the political-crime statistics without committing a category error.

Fourth, right-wing and left-wing political violence were recorded at remarkably similar aggregate levels across 2016–2025, although their trajectories were very different. Left-wing violence was higher during much of the first half of the decade. Right-wing violence was considerably higher during the second half.

Fifth, non-German nationals are substantially overrepresented among police-identified suspects in general violent crime, including on a resident population basis. That is a real descriptive finding. It is not, by itself, a causal explanation of violence.

These conclusions can coexist. None requires the suppression of another.

1. What Germany’s PMK system actually measures

Germany records politically motivated crime through the Kriminalpolizeilicher Meldedienst in Fällen Politisch motivierter Kriminalität, abbreviated KPMD-PMK.

The system is operated jointly by the federal government and the Länder. Local police and, ordinarily, the relevant State Criminal Police Office assess an offence and transmit it to the Federal Criminal Police Office. The Länder retain primary responsibility for the classification because they possess the detailed case information.

An offence may enter the PMK system when its circumstances or the apparent attitude of the offender indicate that it was intended, among other possibilities, to influence political decision-making, oppose political objectives, attack the constitutional order, endanger Germany’s foreign interests through violence or target a person because of a politically relevant attributed characteristic.

The system then records several dimensions.

These may include:

• the criminal offence;
• the attributed political motive;
• the ideological category;
• thematic fields such as antisemitism, racism or hostility toward foreigners;
• the target of the offence;
• whether the offence is considered extremist;
• whether it meets the definition of political violence.

An offence can therefore be both antisemitic and right-wing, antisemitic and Islamist, or antisemitic and connected to a foreign political ideology. “Antisemitic” is a thematic description. “Right-wing” or “religious ideology” is an attributed ideological background. They are not interchangeable labels.

The five principal ideological categories

For the period examined here, the principal categories are:

1. PMK-right
2. PMK-left
3. PMK-foreign ideology
4. PMK-religious ideology
5. PMK-other assignment

Until the end of 2022, the last category was described as not assignable. For offences committed from 2023 onward, it is called other assignment. It is used when a politically motivated offence cannot reasonably be subsumed under the four principal ideological categories.

That category is not statistical waste. It is an essential admission that political motivation does not always fit a clean left–right–religious–foreign taxonomy.

What counts as political violence

Political violence is a subset of all politically motivated crime. It includes offences such as:

• homicide;
• bodily injury;
• arson and explosives offences;
• breach of the peace;
• dangerous interference with transport;
• deprivation of liberty;
• robbery;
• extortion;
• resistance offences;
• certain sexual offences.

This breadth matters. “A politically motivated violent offence” is not synonymous with “a terrorist attack.” It can describe a homicide, but also a physical altercation, resistance against police or violence committed during a demonstration. Terrorism is a narrower and more severe concept, although terrorist offences are also included within the broader PMK system.

A count of violent offences therefore measures frequency within a defined catalogue. It does not measure severity, casualties, social impact or terrorist capability.

2. Methodology for the ten-year comparison

This analysis covers the complete reporting years from 2016 through 2025. The incomplete year 2026 is excluded from the longitudinal totals.

The comparison follows five rules.

Rule one: use violence, not total PMK offences

Total PMK figures include large numbers of propaganda offences, incitement, insults, property damage, offences involving prohibited symbols and violations connected to demonstrations or elections.

Those figures are important, but they are not measures of physical violence.

In 2025, for example, Germany recorded 85,837 politically motivated offences, of which 4,156 were classified as violent. The two figures describe different phenomena.

Rule two: retain the official categories

The table below does not independently reclassify individual cases. An official right-wing case remains right-wing unless public case-level evidence provides a defensible basis for a different conclusion.

It would be methodologically improper to subtract all offences involving unknown perpetrators, all antisemitic offences or all offences committed by people with a migration background from the right-wing category.

Rule three: recognise the 2017 structural break

In 2016, the statistics still used the broader category of politically motivated foreigner crime. From 2017 onward, religious ideology and foreign ideology were reported separately.

The 2016 figure in the foreign-related column is therefore not directly equivalent to the later foreign-ideology figures. Any ten-year total involving these categories must be interpreted with this discontinuity in mind.

Rule four: treat the figures as administrative observations

The PMK figures describe offences known to and classified by the authorities. They are not a census of every politically motivated act committed in Germany.

Changes can reflect:

• actual changes in offending;
• changes in reporting;
• changes in police detection;
• legal changes;
• revisions to definitions;
• major political events;
• altered classification practice;
• cases reported or reclassified after the initial reporting date.

Rule five: do not infer causation from descriptive categories

A category can show that authorities attributed a motive to an offence. It does not, without further evidence, explain why the offender acted, why a trend changed or why one population has a higher recorded rate than another.

3. Politically motivated violence, 2016–2025

The following table compiles the final annual counts of politically motivated violent offences by official category.

¹ The 2016 value belongs to the former umbrella category of politically motivated foreigner crime and is not fully comparable with the foreign-ideology category used from 2017.

² Religious ideology is reported separately only from 2017 in this series.

Source note: author’s calculation from the BKA’s annual PMK reporting series. The 2023 annual report, the official 2024 reporting and the 2025 release were used to verify the recent series; the historical values correspond to the BKA-based long-run table.

What the table shows

Across the full decade, the official system recorded:

• 12,630 right-wing violent offences;
• 12,397 left-wing violent offences;
• 4,401 offences in the foreign-related column;
• 627 religiously motivated violent offences from 2017 onward;
• 7,319 offences in the other or unassignable category.

Right-wing and left-wing violence were therefore almost equal in aggregate over the complete ten-year period: approximately 33.8 percent and 33.2 percent of the total respectively.

That numerical proximity can be misleading, however, because it conceals a pronounced shift over time.

From 2016 through 2020, the system recorded:

• 7,587 left-wing violent offences
• 6,062 right-wing violent offences

From 2021 through 2025, it recorded:

• 6,568 right-wing violent offences
• 4,810 left-wing violent offences

The decade did not contain one stable hierarchy. Left-wing violence dominated much of the earlier period, while right-wing violence became the largest individual category in each year from 2022 onward.

In 2025, right-wing violent offences rose to 1,598. Left-wing violence also increased sharply, from 762 to 1,087. Foreign-ideological violence declined from the unusually high 2024 level of 975 to 704, while religiously motivated violence increased from 87 to 98.

The 2025 total of 4,156 was the highest level recorded since 2016, but it remained below the 4,311 violent offences shown for 2016. The official description “highest since 2016” is therefore the accurate formulation.

What the table does not show

It does not show:

• the number of distinct perpetrators;
• the number of victims;
• the severity of injuries;
• whether a case involved organised terrorism;
• whether multiple offences arose from one event;
• whether the suspected motive was later confirmed in court;
• whether an offender held additional or conflicting ideological beliefs;
• the number of offences never reported or never identified as political.

It also cannot establish moral equivalence between categories. One thousand cases in two categories need not involve equivalent conduct, equivalent targets or equivalent consequences.

A count is not a complete threat assessment.

4. Why total PMK numbers are frequently misused

The most common mistake in public debate is to compare total PMK offences as though they were equivalent to violent attacks.

In 2025, PMK-right accounted for 42,544 recorded offences and 1,598 violent offences. PMK-left accounted for 13,490 offences and 1,087 violent offences. The right-wing total was more than three times the left-wing total, while the violent-offence count was approximately one and a half times as high.

The difference is partly explained by the composition of the categories. Right-wing statistics contain very large numbers of propaganda offences and offences involving prohibited National Socialist symbols. Left-wing totals are strongly influenced by demonstrations, elections, political summits, damage to property and conflict with political opponents or police. The BKA itself notes that major events can materially alter the annual volume and composition of PMK-left.

Neither statistic is illegitimate. They simply answer different questions.

The total number of right-wing offences is relevant to the prevalence of illegal propaganda, incitement and organised ideological activity. The violent subset is more relevant when the question concerns physical aggression. Terrorism statistics, homicide figures, victim numbers and intelligence assessments answer still other questions.

Selecting whichever number best supports a preferred narrative is not analysis. It is metric shopping.

5. Were Islamist offences recorded as right-wing?

The general allegation

The proposition that German authorities generally count Islamist attacks as right-wing violence is not supported by the classification system or by the available data.

Islamist motivation falls under religious ideology. A non-religious ideology originating abroad falls under foreign ideology. Authorities also retain an other-assignment category when the available facts do not support one of the principal classifications.

The distribution of offences connected to the Middle East conflict provides a practical test.

During the first nine months of 2024, authorities recorded 3,931 politically motivated offences connected to that conflict. Of these:

• 2,816 were attributed to foreign ideology;
• 383 to religious ideology;
• 249 to left-wing motivation;
• 166 to right-wing motivation;
• 317 to other assignment.

Those figures are incompatible with a system that automatically converts Islamist or Middle East-related offending into right-wing crime.

The narrower and valid criticism

The serious criticism concerns the historical treatment of antisemitic offences when the motivation could not be positively established.

The federal government stated in 2021 that antisemitic offences were assigned to PMK-right when neither the circumstances of the offence nor the apparent attitude of the offender provided contrary indications. Such contrary indications could include characteristics of the offender, the language used or the symbols involved. Where corresponding evidence existed, an offence was to be assigned to left, foreign ideology or religious ideology instead.

The government defended this practice by pointing to the historically central role of antisemitism in right-wing ideology and stated that multistage quality controls had not produced evidence of a statistically distorting effect.

Both parts of that statement deserve to be preserved.

It is historically and empirically reasonable to recognise antisemitism as a central element of German right-wing extremism.

It is also methodologically problematic to treat a lack of contrary evidence as though it were positive evidence of a right-wing motive.

The issue can be expressed formally:

“No evidence that the offence was not right-wing” is not equivalent to “evidence that the offence was right-wing.”

That does not prove that a large number of offences were misclassified. It establishes that the rule contained the possibility of misclassification.

The scale of that effect cannot be reliably calculated from aggregate public data.

Unknown offenders and ideological symbols

A related controversy concerns offences committed by unknown persons using explicit National Socialist propaganda or symbols.

The federal government confirmed in 2025 that offences involving prohibited symbols such as swastikas or SS runes are assigned to PMK-right when no contrary facts concerning the motivation are available.

This should not be described as the authorities claiming to know the personal ideology of an unidentified offender. The classification concerns the observable ideological content of the act.

That method remains open to error. A symbol could conceivably be used provocatively, ironically, fraudulently or by someone seeking to implicate another group. But classifying an explicit National Socialist symbol as right-wing is substantially different from classifying an ideologically ambiguous antisemitic act as right-wing merely because another motive has not yet been demonstrated.

The evidentiary quality is different and should be reported as such.

6. Why the disputed rule cannot simply be “corrected”

It may be tempting to reconstruct a supposedly clean statistic by removing a proportion of antisemitic offences from PMK-right and reallocating them to religious ideology.

That would not improve the data. It would replace one unsupported assumption with another.

A defensible reclassification requires case-level evidence such as:

• a confession or ideological statement;
• communication records;
• membership in an organisation;
• propaganda consumed or distributed by the offender;
• the target selection;
• the language and symbols used;
• the preparation of the offence;
• findings by prosecutors or courts;
• a later official reclassification.

The offender’s name, nationality, religion or immigration history is not sufficient.

A Syrian offender is not automatically an Islamist. A German offender is not automatically right-wing. A Muslim offender may act from a personal, criminal, nationalist, Islamist, antisemitic, psychiatric or mixed motive. A person with German nationality may act on behalf of a foreign or religious ideology.

Any serious audit must therefore preserve at least three separate fields:

1. official classification at the reporting date;
2. later established findings;
3. independent evidentiary assessment, including confidence.

Where the evidence remains insufficient, the proper result is not an ideologically convenient estimate.

It is unknown.

7. The “other assignment” category is not a defect

The large number of offences in the other or formerly unassignable category is sometimes treated as evidence that the PMK system has failed.

The opposite interpretation is often more appropriate.

Between 2016 and 2025, 7,319 violent offences—almost one fifth of the total—were recorded in that category. The number was particularly high in 2021 and 2022.

A substantial unknown or residual category reduces the apparent precision of the statistics. But that is preferable to forcing ambiguous cases into a politically recognisable box.

The purpose of statistical classification should not be to eliminate uncertainty from the final table. It should be to represent uncertainty honestly.

A reliable system must be allowed to say:

• politically motivated, ideology unresolved;
• mixed ideological motive;
• initial assessment only;
• later reclassified;
• insufficient evidence;
• non-political motive subsequently established.

A database that cannot represent uncertainty does not remove uncertainty.

It hides it.

8. General violent crime and nationality: a separate question

The debate often moves from political violence to “foreigner crime” without acknowledging that the dataset has changed.

The Police Crime Statistics, or PKS, record general criminal offences handled by the police. Unlike the PMK system, they are not primarily organised around political ideology.

In 2025, violent crime in the PKS declined by 2.3 percent. The number of suspects declined by 2.6 percent, and the number of suspects classified as immigrants under the specific PKS definition fell by 7.2 percent. Non-German nationals nevertheless represented 42.9 percent of identified violent-crime suspects.

The correct descriptive conclusion is:

Non-German nationals remain substantially overrepresented among police-identified suspects in violent crime, despite a modest decline in the latest reporting year.

That is a meaningful finding. It must neither be concealed nor made to carry more explanatory weight than it can bear.

Why population shares are not enough

A comparison between 42.9 percent of suspects and the overall foreign share of the population is informative, but incomplete.

Crime risk is not distributed evenly through a population. It varies strongly by:

• age;
• sex;
• income;
• family circumstances;
• urban environment;
• exposure to criminal networks;
• social exclusion;
• education;
• experiences of violence;
• residence status;
• length of residence;
• policing and reporting patterns.

Foreign populations in Germany do not have the same age and sex structure as the German-citizen population. A population containing a higher proportion of young men will tend to have a higher recorded rate of violent offending even when nationality has no independent causal effect.

This is not a reason to dismiss the observed difference. It is a reason to calculate it more carefully.

Tatverdächtigenbelastungszahl

Germany’s Tatverdächtigenbelastungszahl, or TVBZ, improves on raw suspect shares by calculating the number of resident suspects per 100,000 residents of the corresponding population, excluding children younger than eight.

For violent crime in 2025, the official special evaluation reported:

These figures concern the ten non-German nationalities with the largest absolute suspect counts for the relevant category. They are not a ranking of every nationality present in Germany.

The differences are too large to be dismissed as an artefact of using absolute numbers alone. The rates show a genuine overrepresentation of several groups among police-identified resident suspects.

But the TVBZ still does not answer the causal question.

It counts police suspects, not convictions. It uses citizenship rather than ethnicity, religion or migration biography. It does not automatically standardise all groups to the same age, sex, socioeconomic and geographic composition. Some population denominators are relatively small, and the government warns that low population counts can produce highly volatile rates. The PKS is also an exit statistic: cases are recorded when police processing is completed and the file is transferred, not necessarily in the year in which the offence occurred.

The defensible statement is therefore neither:

There is no meaningful difference.

Nor:

Nationality causes the difference.

It is:

Certain non-German citizenship groups are substantially overrepresented among recorded violent-crime suspects, including when resident population size is considered. The statistics alone do not identify how much of the difference is attributable to demographics, social conditions, selection effects, migration circumstances, cultural factors, individual behaviour or other mechanisms.

Observation must precede explanation.

It cannot replace it.

9. Four statistical category errors

Much of the public argument can be reduced to four recurring mistakes.

Ideology is not nationality

An offence is not Islamist because the suspect is Muslim, Arab, Afghan or Syrian. It is Islamist when evidence indicates that a religious-political ideology motivated the offence.

Conversely, an offender does not need to be ethnically German to commit a right-wing offence.

A suspect is not a convicted offender

The PKS records persons whom the police identify as suspects after completing their handling of a case. A suspect may later be acquitted, the case may be discontinued or the legal assessment may change.

Conviction statistics answer a different question.

Political violence is not general violence

PMK violence concerns an attributed political motive. PKS violence concerns defined criminal offences regardless of political motive.

A fight between two people may be general violent crime. The same conduct can enter PMK if the evidence indicates a political or hate-based motivation.

Frequency is not severity

A system counting offences gives one unit to each counted offence under its counting rules. It does not automatically weight an attempted mass-casualty terrorist attack more heavily than a bodily-injury offence at a demonstration.

A comprehensive threat assessment would need at least:

• offence counts;
• victims;
• fatalities;
• injuries;
• weapon type;
• target selection;
• organisational capacity;
• completed and attempted offences;
• terrorist preparation;
• repeat offending;
• severity-weighted measures.

No single number can represent all of these dimensions.

10. What a more trustworthy system would publish

Germany’s PMK system is more differentiated than its critics often acknowledge. It distinguishes themes from ideological categories, provides several ideological domains and retains a residual category.

It could nevertheless be made substantially more transparent.

Classification provenance

For each anonymised case, the system should preserve:

• the evidence supporting the classification;
• whether the motive was admitted, inferred or unresolved;
• the classification authority;
• the date of the assessment;
• the confidence level;
• later changes.

Version history

Annual statistics are often presented as final snapshots. A better system would disclose:

• the initial classification;
• subsequent corrections;
• the reason for each correction;
• the statistical effect of revisions.

This would make it possible to distinguish real changes in offending from changes in administrative knowledge.

Separate evidence states

The published data should distinguish:

• confirmed motive;
• strongly supported motive;
• probable motive;
• preliminary attribution;
• mixed motive;
• unknown motive.

An ideological label without an evidence state creates more certainty than the underlying case may justify.

Event and offence levels

One event can generate several criminal offences. Public reporting should allow analysts to distinguish:

• number of events;
• number of offences;
• number of suspects;
• number of victims;
• number of injured people;
• number of fatalities.

Severity measures

Violence totals should be accompanied by separate series for:

• homicide;
• attempted homicide;
• serious bodily injury;
• arson and explosives;
• terrorism;
• attacks on people;
• attacks on property;
• resistance and public-order offences.

This would prevent a broad violence total from being mistaken for a count of terrorist or life-threatening attacks.

Reproducible historical tables

Whenever definitions change, the authorities should publish:

• the former definition;
• the new definition;
• the effective date;
• the affected categories;
• a back-cast series where possible;
• an explicit warning where comparison is not possible.

The 2017 separation of foreign ideology and religious ideology is precisely the kind of structural break that should be visually marked in every longitudinal chart.

11. The responsibility of interpretation

A public statistic passes through several stages before it appears in a headline:

1. an event occurs;
2. someone reports or detects it;
3. police investigate it;
4. an offence is legally coded;
5. a motive is attributed;
6. a thematic and ideological classification is assigned;
7. the case is transmitted and quality-controlled;
8. aggregate figures are published;
9. politicians and media select a comparison;
10. the public encounters a simplified claim.

At every stage, information can be added, lost or compressed.

That does not make official statistics arbitrary. It makes them constructed observations: structured attempts to describe reality through legal and administrative rules.

The appropriate response is neither blind faith nor conspiratorial rejection.

It is methodological scrutiny.

Conclusion

The claim that Germany systematically records Islamist attacks as right-wing violence does not withstand examination.

Islamist and other religiously motivated offences have their own category. Foreign ideological violence has another. Contemporary Middle East-related offences are visibly distributed across religious, foreign, left-wing, right-wing and other classifications.

A narrower criticism remains justified.

Where an antisemitic offence was assigned to the right-wing category because no contrary motivation was evident, the classification system risked converting missing knowledge into ideological certainty. The publicly available data do not show how many cases, if any, were materially misclassified as a result. They therefore do not support a wholesale statistical correction.

The ten-year violence series also resists simple political storytelling.

Right-wing and left-wing violence were almost equal in aggregate between 2016 and 2025, but followed different trajectories. Left-wing violence was higher through much of the earlier period. Right-wing violence became dominant in the later period. Foreign-ideological violence fluctuated strongly with international conflicts. Religiously motivated violence was smaller in numerical frequency, while the large residual category demonstrated how much politically motivated violence does not fit a clean ideological taxonomy.

General violent crime introduces another uncomfortable finding: certain non-German citizenship groups are heavily overrepresented among police-identified suspects, including on a resident population basis. That observation should be reported honestly. It should not be mistaken for proof that nationality itself causes violence.

The standard required is symmetrical.

We should not minimise an observed difference because it is politically inconvenient.

We should not invent an explanation because it is politically useful.

We should not classify uncertainty as certainty merely to complete a table.

The deepest problem with public statistics is rarely that every number is false. It is that a number can be correct within its administrative definition and still be used to claim something it never measured.

Trustworthy statistics therefore require more than accurate counting.

They require visible definitions, preserved uncertainty, documented revisions and restraint about what the evidence can support.

Because once a classification enters the public record, it does more than describe reality.

It begins to govern how reality is understood.