Victimisation and Self-Report Surveys

Police-recorded crime figures pass through a series of filters before they become a statistic. A crime must occur. The victim must recognise it as a crime. The victim must decide to report it. The police must record it according to their counting rules. Each filter removes incidents. In England and Wales, the CSEW consistently finds that only around 40 percent of the crime it records was reported to police, and not all of that was then recorded. In the United States, the NCVS shows that in recent years fewer than half of violent victimisations and fewer than one-third of property crimes were reported to police.

Reasons for non-reporting vary by offence type and by the victim's relationship to the offender. Victims of domestic abuse frequently do not report because of fear of retaliation, emotional ties to the offender, or distrust of the criminal justice system. Victims of sexual assault face additional barriers including stigma and anticipated disbelief. Minor property offences are often not reported because victims consider them too trivial or think the police cannot help. Fraud and cybercrime are frequently unreported because victims do not recognise what happened as a crime, or do not know which agency to contact.

Police recording practices introduce a further distortion. In England and Wales, the introduction of the National Crime Recording Standard in 2002 produced a sudden apparent increase in recorded crime that reflected changed recording rules, not a genuine increase in offending. Similar administrative discontinuities appear in recorded crime data in other countries when counting rules change. Victimisation surveys are not immune to methodological changes, but because their design is explicit and documented, changes in method can be tracked and their effects estimated.

All major victimisation surveys share a common structure. They draw a probability sample of households or individuals from the general population. Trained interviewers ask respondents about a defined set of offence types using standardised screen questions. Positive responses trigger detailed incident reports that collect information about the offence, the offender (where known), the injury or loss sustained, and whether the incident was reported to police. Responses are then weighted to represent the national population.

The NCVS uses a rotating panel design: sampled households are interviewed every six months for three and a half years, then replaced. The bounding interview technique addresses telescoping by using the previous interview as a fixed reference point, so respondents can be asked about incidents since they last spoke to the interviewer rather than relying on memory of a calendar date. The CSEW uses a cross-sectional design with a fresh sample each year, relying on a twelve-month recall window and careful question wording to manage memory effects.

Self-report studies ask respondents to report their own offending, typically by presenting a checklist of behaviours and asking whether the respondent has engaged in each one, either ever or within a defined period. Early work by Short and Nye (1957) used anonymous questionnaires with samples of school students and found that delinquent behaviour was far more widespread than arrest records indicated, and that the class-based concentration of delinquency visible in official data largely disappeared when self-report data was used. This finding had significant implications for criminological theory, calling into question explanations of crime that focused on poverty and lower-class culture.

Travis Hirschi's 1969 study Causes of Delinquency used self-report data from a sample of California schoolboys to test competing theories of delinquency and provided foundational support for social bond theory. Later large-scale efforts, including the Cambridge Study in Delinquent Development in Britain and the Dunedin Multidisciplinary Health and Development Study in New Zealand, combined self-report data with official records and longitudinal follow-up to trace offending careers over decades. These studies revealed that a small proportion of offenders account for a disproportionately large share of total offending, a finding replicated across multiple countries and cohorts.

Contemporary self-report instruments vary in scope. Some focus on specific offence categories such as drug use: the US Monitoring the Future survey has tracked self-reported drug use among American secondary school students annually since 1975. Others are broader: the Edinburgh Study of Youth Transitions and Crime followed a cohort of Edinburgh schoolchildren, collecting annual self-report data on offending alongside school and police records. The International Self-Report Delinquency Study (ISRD) collects comparable data from young people across thirty or more countries, enabling cross-national comparisons of offending that police data cannot provide.

Both victimisation surveys and self-report studies depend on respondents giving accurate, honest answers. The main validity threats differ somewhat between the two types but overlap in important ways. For victim surveys, the primary concerns are: under-reporting (not disclosing incidents due to embarrassment, distress, or because the respondent has normalised the experience), over-reporting (treating non-criminal events as crimes), telescoping, and the reference period boundary effect. For self-report studies, the primary concerns are: under-reporting of serious offending (concealment driven by social desirability or fear of consequences), over-reporting by those who exaggerate or misunderstand questions, and coverage gaps.

One method for assessing the validity of self-report data is to compare self-reported offending with official records for the same individuals where both are available. Studies using this cross-validation method generally find moderate to good correspondence for common offences but poor correspondence for the most serious crimes, which are both rare and most likely to be concealed. Randomised response techniques, in which respondents answer sensitive questions using a random device that gives them plausible deniability, have been used to improve disclosure rates for particularly sensitive items such as sexual offending.

Victimisation surveys have produced a number of durable findings that official statistics obscure or contradict. Repeat victimisation is more concentrated than official data suggests: a small share of survey respondents account for a disproportionately large share of all reported incidents, mirroring the concentration of offending found in self-report studies. Domestic abuse and sexual assault appear far more frequently in victim surveys than in police records. Victimisation is unevenly distributed across age, sex, area type, and socioeconomic group in broadly consistent patterns across countries, though the exact gradients vary.

The CSEW has been particularly useful for tracking trends in England and Wales independently of police recording practices. Between 1995 and 2020, the CSEW showed a sustained decline in household and personal crime of around 70 percent, a trend broadly consistent with similar declines shown by the NCVS in the United States and by victimisation surveys in Australia, Canada, and most European countries. This convergence across independent national surveys with different designs strengthens confidence that the decline reflects a real change in victimisation rates rather than a measurement artefact.

Self-report research has consistently found that the concentration of officially recorded crime in lower-income and minority populations is partly an artefact of differential enforcement and detection rather than a true difference in offending rates for many offence types. For drug use in particular, self-report studies in the United States, United Kingdom, and Australia have found similar prevalence rates across income and racial groups, while arrest and conviction data show very large disparities. This has direct implications for debates about racial disproportionality in criminal justice processing. Research connected to these themes draws on criminological frameworks covered in Labelling Theory and Social Reaction.

Governments and researchers use victimisation survey data for several purposes. They track crime trends independently of changes in police recording practice. They identify under-reported crime types as priorities for victim services or enforcement. They measure the impact of policy changes: the NCVS provided evidence on the effects of mandatory arrest policies for domestic violence in the 1980s and 1990s. They supply the denominators for risk calculations, such as the probability that a household will be burgled in a given year, which inform crime prevention advice and insurance pricing.

Self-report data informs debates about the distribution of criminal justice resources and the fairness of enforcement. It has been used in several countries to argue for changes in drug enforcement policy by demonstrating that arrest and prosecution are concentrated in particular groups despite broadly similar rates of drug use across groups. Longitudinal self-report studies have identified risk factors for persistent offending that inform early intervention programmes: in the United Kingdom, research drawing on the Cambridge Study influenced the development of youth justice policy from the 1980s onward.