The problem in one sentence
Recorded crime statistics and victimisation surveys routinely produce different pictures of the same situation, and anyone using either without understanding why will eventually reach a conclusion the data does not support.
What each measure actually counts
Recorded crime counts incidents that came to the attention of an authority, were accepted as crimes, and were entered into a recording system. Each of those three steps is a filter. An incident that was never reported is not counted. An incident reported but not recorded is not counted. An incident recorded under a different category appears somewhere else.
Victimisation surveys ask a sample of the population whether they experienced particular things over a reference period, regardless of whether they told anyone. They bypass the reporting and recording filters, which is their purpose.
The difference between the two is sometimes called the dark figure of crime: offences that occurred but never entered official records. It is not a measurement error to be corrected. It is a structural feature of how the two instruments work.
Why the gap opens and closes
- Reporting propensity varies by offence. Offences requiring documentation for insurance or replacement are reported at high rates. Offences involving shame, fear of reprisal, or a relationship between the parties are reported at far lower rates. The gap is therefore widest exactly where accurate figures matter most.
- Confidence in institutions changes reporting. If people expect a report to produce nothing, fewer report. Rising confidence can increase recorded crime while actual offending is flat or falling.
- Recording rules change. A change in counting rules, classification or audit practice moves recorded figures without anything changing on the ground.
- Survey design changes estimates. Reference period, question wording, whether a screening question is used, and how repeat victimisation is counted all move survey estimates substantially.
- Coverage differs. Household surveys typically exclude people not living in households and may exclude some age groups. Recorded crime has no such boundary but is shaped by where enforcement attention falls.
A worked example
The figures below are synthetic and illustrate the arithmetic only.
Suppose recorded offences of a given type rise from 1,200 to 1,560 over three years, an increase of 30 per cent. Over the same period a victimisation survey estimates that the proportion of households experiencing that offence fell from 7.4 to 6.1 per cent.
| Measure | Year 1 | Year 3 | Direction |
|---|---|---|---|
| Recorded offences | 1,200 | 1,560 | Up 30 per cent |
| Households victimised, survey estimate | 7.4% | 6.1% | Down 1.3 points |
| Implied reporting rate | Lower | Higher | Consistent with both |
Both series can be correct. If victimisation fell while the share of victims who reported rose, recorded crime increases even as offending declines. A report claiming a crime wave from the first row alone would be wrong, and so would one claiming success from the second row alone.
The analytically useful statement is narrower and more defensible: recorded offences rose over the period, survey estimates of victimisation fell, and the two are consistent with an increase in reporting. Establishing which explanation holds requires evidence the two series by themselves do not contain.
Practical guidance
- Say which measure you are using in the sentence where you use it. "Recorded offences rose" and "crime rose" are different claims.
- Check whether counting rules, classification or survey design changed during the period before interpreting a change over it.
- Treat a divergence between the two as information rather than a problem. It usually points at reporting behaviour, which is often the thing worth knowing.
- Be most cautious with offence types where reporting is known to be low. Those series measure reporting at least as much as offending.
- Avoid converting either measure into a rate without stating the denominator and its source.
Sources and standards
Abbreviated for this sample. A real deliverable carries full citations with publication and access dates for every source.
- United Nations Office on Drugs and Crime, International Classification of Crime for Statistical Purposes, the standard framework for making crime statistics comparable between jurisdictions.
- United Nations Office on Drugs and Crime and United Nations Economic Commission for Europe, Manual on Victimization Surveys, which sets out survey design practice and the reasons survey and administrative measures diverge.
- Kenya National Bureau of Statistics, which publishes official statistics including crime data supplied by the relevant institutions.
- National Crime Research Centre, Kenya, which conducts and publishes crime research.