When “Overrepresentation” Is a Statistical Illusion

A recent media analysis claims that leading German print and TV outlets report violent crimes involving foreign suspects far more often than would be expected based on police crime statistics — with roughly 90–95 % of reports naming foreign suspects despite foreigners accounting for about 34 % of violent crime suspects. The study also asserts that suspects from predominantly Muslim countries are especially overrepresented in media reporting compared to their share in the official data.

Why comparing crime statistics to media coverage leads us astray

Recent discussions claim that German media report disproportionately often on crimes committed by foreigners, pointing to a mismatch between media coverage and police crime statistics. At first glance, the numbers look striking. But the conclusion rests on a basic statistical mistake: it assumes that media coverage should be proportional to crime frequency or other rational/dry crime facts (such as number of victims).

That assumption is wrong.

Media do not report “average” events

Police crime statistics record all known crimes: from bar fights and domestic violence to extreme acts of public violence. Media outlets, by contrast, report on a highly selected subset of cases — those that are rare, shocking, and likely to provoke public fear.

This difference matters enormously.

National media attention is driven by factors such as:

severity and brutality,
randomness and public exposure,
symbolic or ideological meaning,
perceived threat to social norms.

These factors have little to do with how often a type of crime occurs.

Why “overrepresentation” appears automatically

Once we accept that the media focus on extreme cases, the rest follows almost mechanically.

Certain types of crimes — terrorist attacks, ideological violence, attacks on children or random victims in public spaces — almost guarantee intense media coverage. These crimes are rare, but when they occur, they are reported extensively.

If some of these extreme crimes are more likely to involve foreign perpetrators, then foreigners will appear frequently in media reports even if journalists apply the same standards to everyone.

This is not evidence of unfair reporting. It is a consequence of non-random selection.

A hidden statistical trap: looking only at cases where nationality is mentioned

Many analyses focus only on reports where the suspect’s nationality is explicitly mentioned. This creates a further distortion.

Nationality is more likely to be mentioned when it is seen as relevant to the story — for example in cases involving terrorism, migration policy, or international security. In routine crimes, nationality is often omitted altogether.

By restricting attention to cases where nationality appears, the analysis builds the conclusion into the data. The resulting “overrepresentation” is partly a product of how the sample was constructed.

What would a fair comparison look like?

A meaningful analysis would not compare media reports to all crimes. It would compare similar crimes:

equally severe cases,
similar numbers of victims,
similar levels of randomness and public exposure,
similar ideological context.

Only after accounting for these factors could one ask whether nationality still plays an independent role in media attention.

Without this step, claims about bias rest on shaky ground.

Why this distinction matters

None of this denies that media coverage shapes public perception, or that selective reporting can have social consequences. But diagnosing a problem requires getting the logic right.

When we mistake selection effects for unfairness, we risk drawing strong moral conclusions from weak statistical foundations.

The core takeaway

The fact that foreigners appear more often in crime reporting does not imply that media reporting is unfair. It mainly shows that media coverage is driven by rarity, fear, and symbolism — not by proportional representation.

Or put more simply:

Media reports are about what shocks us, not about what happens most often.

Anecdotal Cherry Picking is not science

The “two car attacks” example

A central illustration in Hestermann’s analysis compares two car attacks that occurred within a few weeks of each other: one committed by an Islamist extremist, the other by a German offender described as mentally unstable. Because the attacks appear superficially similar, the difference in media coverage is attributed to nationality. However, this is not a valid inference. Media attention is highly non-linear and sensitive to symbolic and ideological context. An attack framed as ideologically motivated signals broader risks — future attacks, organized networks, political consequences — and therefore almost automatically generates sustained coverage. A single carefully chosen pair of events, no matter how vivid, cannot establish a general pattern. At best, it illustrates that differences in coverage can occur; it does not show that they systematically occur because of nationality rather than narrative relevance.

A note on scientific standards

There is a further issue that should not be ignored. The analyses cited in support of these claims have not undergone peer review in established academic journals, and the underlying data are not publicly available. This means that independent researchers cannot replicate the results, test alternative specifications, or verify the coding decisions that are crucial in content analysis.

In empirical social science, transparency and reproducibility are not optional luxuries; they are basic safeguards against error and overinterpretation. When data and methods are not openly accessible, strong conclusions about systemic bias should be treated with appropriate caution.

This does not imply bad faith. But it does mean that the findings have not been subjected to the level of scrutiny normally expected before drawing far-reaching claims about media behavior.

When Uncritical Reporting Becomes Part of the Problem

When a publicly funded broadcaster presents contested research as settled fact, scrutiny is not optional. Yet this is precisely what happened when Deutsche Welle uncritically amplified claims that German media systematically overreport crimes committed by foreigners. Framed with authoritative phrases such as “Expertinnen und Experten warnen: So entsteht ein verzerrtes Bild,” the piece leaves little room for methodological doubt. More troubling still, the message appears in DW’s Learn German program — an educational format aimed at language learners who are unlikely to question the statistical foundations of what they are being taught. Before accepting such claims as established truth, it is worth examining whether the underlying evidence can actually support them.