Crime Can Fall Even as Fear Rises

At a press conference, a downward-sloping line appears on the screen. The police announce that they recorded fewer offences nationwide than the year before. A few hours later, at a district council meeting, a woman says she no longer uses the underpass by the station after dark. A convenience-store owner describes aggressive customers, parents circulate footage of a fight in a local group, and someone comments beneath the post: “And then they’ll tell us it’s safe.”

At first glance, there seem to be only two possibilities. Either people are exaggerating because the media and social networks have frightened them, or institutions are hiding reality behind tables. One side must be naïve and the other dishonest.

Yet the police chart, the woman’s experience in the underpass and the parents’ anxiety can all be true at the same time. They are not three answers to the same question. The first describes the number of recorded offences across a large territory. The second describes how a particular person makes decisions in a particular place and time. The third captures how information about danger spreads through a community. The dispute arises only when we call all these phenomena by a single word: crime.

Crime statistics are a map of what has passed through an institution. A sense of safety is a map of how people move through the world. Both maps can be accurate—and still use different legends.

1. The False Referendum: People or Numbers?

The question “Is crime rising?” sounds as though it calls for a single figure. In reality, it contains at least five further questions. Has conduct meeting the legal definition of an offence become more common? Are there more victims? More reports? More police records? Or has people’s sense that they can use a street, park or public transport without disproportionate fear changed?

These variables are related, but they are not interchangeable. The true incidence of crime cannot be observed directly. Some incidents are never recognised as crimes, some are not reported by victims, some reach the police by other routes, and some reported events do not appear in the final statistics in the same form in which the reporter described them. At the other end lies the sense of safety: it may respond not only to the probability of an attack, but also to the severity of the possible consequences, personal vulnerability, control over the situation, the condition of the environment and confidence that someone will intervene in an emergency.

Police-recorded crime is therefore not a false version of “real crime”. It is a distinct and important administrative indicator: it tells us which offences came to the police’s attention and which they recorded under the applicable rules. Nor is a victimisation survey simply a better police statistic. By asking a representative sample about their experiences, it can illuminate part of the dark figure; at the same time, it is affected by memory errors, reluctance to disclose sensitive events, differing interpretations of questions and the fact that some offences have no living individual victim who can be asked.

2. How an Event Becomes a Row in a Table

Between an offence and the annual total there is no single neutral sensor. There is a chain of human and institutional decisions. Someone must notice the event, judge it serious enough, decide to report it, describe it, substantiate it and classify it. The Police of the Czech Republic state that their statistics use recorded offences under tactical-statistical classifications for which criminal proceedings were initiated. The figures are therefore neither a list of final convictions nor a mechanical transcript of every call to emergency number 158.[1]

The size of the “loss” between experience and reporting varies dramatically by offence. A Czech victimisation study by the Institute of Criminology and Social Prevention, published in 2019, found that among the most recent incidents examined, 92 per cent of car thefts were reported to the police, compared with only 31 per cent of physical assaults, 21 per cent of sexual assaults, 17 per cent of domestic violence and 12 per cent of stalking. For online-shopping fraud involving financial loss, the figure was eight per cent.[5]

These figures are not a current estimate of today’s prevalence and cannot simply be transferred to every case. Their value lies elsewhere: they show that there is no single universal “under-reporting rate”. Car theft is often tied to insurance and an unmistakable loss. When violence occurs between people who are close, fear of the perpetrator, shame, financial dependence and doubt that institutions will help may all shape the decision. Two categories with the same actual incidence can therefore generate entirely different numbers of records.

This is where the first paradox arises. If victims begin to trust an institution more, the police improve the intake of reports, or a specialised unit actively uncovers more cases, recorded crime can rise without the actual incidence rising at the same pace. Conversely, declining trust, victims’ resignation or a shift into less visible settings can push the statistics down even though people’s everyday experience has not improved.

3. What the Czech Data Actually Say

In 2025, police in the Czech Republic recorded 170,051 offences, 1.9 per cent fewer than the year before. That statement is true. It is equally true that violent crime rose slightly to 13,752 offences and sexual and morality offences rose by 8.8 per cent to 3,695. The fall in the total was driven by other parts of the sum.[1][2]

The total number of offences is not an index of social harm either. One record of a less serious property offence and one record of a violent offence each count as one unit in the total, even though their impact on the victim and on public space may be incomparable. A declining total can therefore contain growth in a category to which people are more sensitive. Yet growth in a category does not automatically mean an equal rise in real-world conduct: for example, police linked part of the increase in recorded child pornography in 2025 to a greater volume of intelligence from international databases and targeted operations. The capacity to see the phenomenon had changed as well.[1]

The long Czech time series provides an even more important correction. Between 2002 and 2023, the number of recorded offences fell from roughly 372,000 to 181,000—about half. Over comparable CVVM surveys, the share of people who felt safe in the Czech Republic rose at the same time from 45 to 82 per cent.[2][3] The Czech data therefore do not support the cheap thesis that the public simply succumbs to ever-greater fear regardless of reality.

Even this long series is not laboratory-clean. The police note that raising the loss threshold in October 2020 reduced the number of recorded offences, especially in property and economic crime, and that the structure of tactical-statistical classifications underwent extensive revision from 2021. Part of the break therefore describes a change in reality and part a change in the measuring instrument; the share of each cannot be separated from a single total.[1]

The two curves do not move as mechanical mirror images. Between 2019 and 2021, recorded offences fell by more than 23 per cent, while the share of people who felt safe nationwide fell by two percentage points. Pandemic restrictions reduced opportunities for some offences and changed the number of recorded events, but they certainly did not create a general sense of calm and predictability. From 2021 to 2022, recorded crime then rose by almost 19 per cent and the sense of safety fell by five points. In the following year, the offence count barely changed while the sense of safety rose slightly again.

4. The National Average Is Not the Street You Walk Home On

A national average is essential for governing a country, but no one lives in a national average. People move along a handful of routes, at particular hours and among particular people. Their personal risk depends not only on the number of inhabitants, but on exposure: how often they enter a given environment, what kinds of situations occur there and whether they can escape them.

A resident can therefore be entirely justified in saying that conditions in their neighbourhood have worsened even as the nationwide figure falls. The particular offence they fear may have increased nearby. A few repeat offenders may have concentrated their conduct in a small area. Evening transport, lighting or the presence of people who once provided natural surveillance may have deteriorated. Some changes need not themselves meet the legal definition of a crime: loud conflicts, visible intoxication, harassment, disorder or abandoned premises can alter how legible and controllable a space feels before police statistics move significantly.

Research on fear of crime has long shown that people read the condition of their surroundings as information about informal control and community cohesion. In a large British analysis, fear of violence was associated not only with personal experience, but also with perceived disorder and weaker collective capacity to uphold ordinary rules.[8] Transferring the finding to the Czech context has limits, but the mechanism is readily understandable: a broken light is not an assault, yet it can signal that an assault would be harder to see and help might not be close at hand.

The difference between the whole country and one’s place of residence is visible in Czech surveys as well. In 2023, 91 per cent of respondents rated safety where they lived positively, compared with 82 per cent for the Czech Republic as a whole.[3] That does not prove that the nine-point gap was created by the media. It does show, however, that people assess nearby and distant environments under different information regimes: they know their street from everyday life, while they assemble their picture of the country indirectly.

5. Fear Is Neither a Single Emotion nor a Single Question

The phrase “fear of crime” is often used in debate as though it were a thermometer with a single scale. Academic work, however, distinguishes specific episodes of worry, broader anxiety, estimates of the probability of victimisation, and behaviour altered because of perceived risk. Gray, Jackson and Farrall also distinguish functional worry—which leads to proportionate caution—from dysfunctional worry that repeatedly restricts movement, relationships and quality of life.[7]

The wording of the question matters too. “How likely is it that you will be mugged?” measures an estimate of risk. “How many times in the past month did you actually feel worried?” looks for specific emotional episodes. “Do you feel safe walking alone after dark?” includes vulnerability, possibilities of escape, lighting, the presence of other people and expected assistance. A respondent may consider an attack unlikely and still change route because the possible consequence is unacceptable.

A subjective feeling is therefore not an inaccurate copy of objective risk. It measures another layer of safety: the degree of control and freedom. Two people facing the same statistical probability can bear different expected costs. A person with limited mobility, a woman returning from a night shift or someone who has previously been assaulted may not be “bad at percentages”. They may be realistically accounting for a reduced ability to escape, a higher cost if something happens, or vivid knowledge of how quickly an ordinary situation can change.

6. How One Event Can Fill the Entire Mental Field

The best-known explanation draws on the availability heuristic. When estimating frequency, we do not always consult a statistical set; often we rely on how easily an example comes to mind. A vivid, recent and emotionally powerful case is recalled faster than hundreds of days on which nothing happened. Tversky and Kahneman described this mechanism as early as 1973.[9] That does not make human judgement worthless. Availability is a useful shortcut in environments where personal experience often does reflect frequency. The problem begins when the selection of remembered events is not representative.

The modern information environment radically changes that selection. One case does not appear once. There is the first report, a video, a police statement, a politician’s reaction, a profile of the offender, an interview with a witness, the trial, commentary and an anniversary. It is still one event, but a person experiences it as many separate encounters with danger. A study of US television news found an association between viewing crime-saturated local news and greater fear even after local crime levels were taken into account.[10] That is not proof that every Czech programme has the same effect; it is evidence of a mechanism that must be tested, not assumed.

Social networks add personalisation, repeated sharing and competition for attention. Research on tweets from eighteen Latin American countries found a strong overrepresentation of violent and sexual offences and only a weak correlation between the volume of posts and crime itself; the posts reflected levels of concern more closely.[10] Another study, based mainly on young adults in the United States, found a relationship between total social-media consumption and fear of crime, while explicitly noting the limitations of its sample and the preliminary nature of its conclusions.[10]

None of this warrants the cheap conclusion that “the media frighten us”. Media can expose a problem that long remained hidden, give victims a voice and increase willingness to report. A social network can exaggerate frequency yet accurately show that a particular group has lost trust in an institution or that harassment below the threshold of criminal law is recurring in a specific place. Visibility is a filter, not automatically a lie.

7. When Distrust of Statistics Is Justified

Official statistics are not automatically propaganda. Without them, we would be left with the loudest stories, political campaigns and our own limited surroundings. Police data make long time series, geographical breakdowns and analysis by offence type possible. But it is a mistake to attribute a quality they do not possess: a complete and consistently measured picture of every offence.

Eurostat explicitly warns that police-recorded counts are shaped by legal definitions, people’s willingness to report and police recording practices, and that methodological changes can create breaks in a time series.[6] A new offence, a broader definition, electronic reporting, a specialised campaign or better training can increase the number of records. A stricter threshold, administrative overload or a loss of trust can reduce it. For offences discovered mainly through proactive state activity, the statistics also partly measure the intensity of enforcement.

Public scepticism is therefore legitimate when an institution presents only an overall total and conceals changes in categories, methodology, reporting or local distribution. Scepticism is equally legitimate toward a campaign that turns a single increase in one category into the collapse of safety across an entire country. In neither case need the number be falsified; what is falsified is the scope of the claim the number is made to carry.

8. When Rising Fear Is a Relevant Signal

A sense of threat is not suitable for estimating the number of crimes. It can, however, be an important indicator of the quality of public space—especially when the same change repeatedly appears in the same place, among the same group and under the same question; when it is accompanied by route changes, fewer evening passengers, business closures, school absences, more requests for accompaniment or consistent experiences captured in victimisation research.

The most revealing case is the exposure paradox. A park may record fewer assaults because it has become safer. But it may also record fewer because people avoid it out of fear. The number of potential victims falls, opportunities for offences diminish and the police statistics improve—at the cost of the public space no longer serving its purpose. Statistically, the place is quieter; socially, it has been lost.

The opposite paradox is just as important. Recorded cases of domestic or sexual violence can rise at a time when victims are less afraid to report, institutions recognise the conduct more effectively and society stops tolerating certain behaviour. The statistics worsen, while the institution’s capacity to see the problem may improve. The direction of the number and the direction of reality are therefore not synonyms.

Public feeling is also a relevant signal when it responds not to frequency but to predictability. People may tolerate a stable level of risk better than an environment in which they do not know whether rules apply, whether help will answer and whether an institution will even acknowledge the event. In such a case, fear speaks not only about crime. It speaks about trust, control and the social contract.

9. Is It Safer Today?

Without further specification, the question is not precise enough. Safer where —in the country as a whole, in one municipality, at the station, at home or online? Safer for whom —for the average resident, a child, an older person, a woman working nights, a shopkeeper or a repeatedly victimised household? Safer from what —from theft, physical assault, fraud, domestic violence or public harassment? And by what measure—police records, victimisation surveys, final convictions, harm, sense of safety, or freedom to use public space?

An honest answer therefore does not emerge from one curve defeating another. It emerges from triangulation. Police statistics show offences captured by institutions. Victimisation surveys illuminate experiences that never entered the records. Local data reveal concentration. Surveys of safety measure human costs and trust. Changes in law and recording show whether the measuring instrument itself has changed.

And this is where the simple dispute turns inside out. Crime can fall even as fear rises. Fear can fall even while a particular form of crime rises. Recorded crime can rise because an institution has begun to see reality more clearly. And it can fall because people no longer believe reporting it is worthwhile.

The most precise question, then, is not merely: “Is it safer today?” It is: Safer where, for whom, from what, by which measure—and what changed on the path from the actual event to the number used to describe it?