How Much Crime in South Africa Goes Unreported?
Reporting rates depend heavily on what crime happened. Stats SA's GPSJS 2024/25 survey found South Africans reported 43.4% of housebreaking incidents to police (the most common household crime, 1,493,000 incidences), and 31.0% of theft-of-personal-property incidents (the most common individual crime, 1,323,000 incidences). Across all crime types the survey measured, reporting rates ranged from 41.1% (theft of livestock/poultry) to 100% (murder).
How to read this page: SAPS crime statistics, used everywhere else on SafeSuburb, count crimes reported to and recorded by police. GPSJS is a separate Stats SA household survey that asks a nationally representative sample whether they experienced specific crimes in the past 12 months, whether or not they reported them. This page uses GPSJS to show the size of that gap, by crime type. It does not adjust, multiply, or "correct" any SAPS figure shown elsewhere on the site. See methodology for how SafeSuburb presents SAPS data.
Reporting rates by crime type
GPSJS asks about crime in two separate modules: household crimes (things that happen to a household's home, vehicle, or livestock) and individual crimes (things that happen to a person). Each module's reporting rate below is the share of households or individuals, among those who experienced that crime type in the 12 months before the survey, who reported all or some of the incidents to SAPS.
Household crimes
| Crime type | Incidences 2024/25 | Reported to SAPS | Sample (n) |
|---|---|---|---|
| Theft of livestock/poultry | 248,637 | 41.1% | 200 |
| Housebreaking | 1,493,000 | 43.4% | 1156 |
| Sexual offence (household) | 50,000 | 47.6% | 34 |
| Home robbery | 259,000 | 51.3% | 210 |
| Deliberate damage to dwellings | 123,000 | 53.3% | 112 |
| Assault (household) | 158,000 | 67.0% | 137 |
| Theft of motor vehicle | 77,000 | 82.5% | 58 |
| Murder | 14,000 | 100.0% | too small to reproduce |
"Reported to SAPS" is the weighted share of households that reported all or some incidents, among households that experienced that crime type. Sample (n) is the unweighted count of households in the GPSJS 2024/25 microdata who experienced that crime, before weighting. Murder is cited directly from Stats SA's published table (Table 17): the microdata export has no separate murder experience/report column, so it is not independently reproduced here like the other seven.
Individual crimes
| Crime type | Incidences 2024/25 | Reported to SAPS | Sample (n) |
|---|---|---|---|
| Theft of personal property | 1,323,000 | 31.0% | 528 |
| Consumer fraud | 811,000 | 34.9% | 232 |
| Street robbery | 529,000 | 36.4% | 220 |
| Assault (individual) | 561,000 | 54.4% | 165 |
| Sexual offence (individual) | 60,000 | 56.3% | 25 |
| Hijacking | 119,000 | 76.8% | 47 |
Hijacking has the highest individual-crime reporting rate (76.8%). Sexual offence and hijacking have the smallest samples in the 2024/25 survey (n=25 and n=47 respectively). Stats SA still publishes these nationally, and SafeSuburb follows that judgement, but smaller samples carry wider uncertainty than housebreaking or theft of personal property.
GPSJS and SAPS measure different things. GPSJS asks whether a household or individual experienced a crime in the 12 months before their interview, regardless of whether it was reported. SAPS counts crimes reported to and recorded by a police station in a given period. The two use different reference windows, different geographic units (GPSJS is representative at national, province, and metro/non-metro level; SAPS is recorded per precinct), and different crime definitions in places. They are not the same measurement with a gap that can be closed by multiplying one by the other, which is why SafeSuburb does not apply a reporting-rate multiplier to any SAPS figure.
Why people don't report crime
GPSJS asks victims who didn't report their most recent incident why not. The same two reasons dominate across crime types: doubt that police can or will act, and people resolving the matter themselves.
Housebreaking
- Police could do nothing / lack of proof25.1%
- Police would not do anything about it21.0%
- Solved it myself18.7%
- Inappropriate for police / police not necessary11.0%
- Reported to other authorities instead7.1%
- My family resolved it4.6%
Street robbery
- Police would not do anything about it32.6%
- Police could do nothing / lack of proof26.3%
- Solved it myself8.5%
- Inappropriate for police / police not necessary7.2%
- Did not dare (fear of reprisal)4.0%
- I do not trust the police3.6%
Theft of personal property
- Police would not do anything about it25.4%
- Police could do nothing / lack of proof24.8%
- Solved it myself16.7%
- Inappropriate for police / police not necessary13.5%
- My family resolved it4.9%
- I do not trust the police4.3%
Reasons are the weighted percentage distribution among households or individuals who did not report the stated crime type in 2024/25, as published in the GPSJS 2024/25 statistical release. Only the top reasons are shown here for space; the full distribution is in the published release.
What this means for reading SafeSuburb's crime pages
Recorded counts are a floor, not a ceiling. Every raw count on SafeSuburb is the number of crimes SAPS recorded, which GPSJS shows is reliably lower than the number that actually happened. A precinct's true crime volume for any given category is at least as high as the recorded count, and for categories like theft of personal property or consumer fraud, likely much higher.
Under-reporting is not uniform across crime types. Murder is reported close to 100% of the time nationally; theft of personal property is reported about 31% of the time. That means comparing raw counts across different crime categories (for example, "housebreaking is 10x more common than home robbery in this precinct") mixes categories with very different reporting behaviour. Comparing the same category across precincts, or the same category over time within a precinct, is not affected by this in the same way, since the national reporting-rate gap should apply similarly regardless of area.
This does not change any number shown elsewhere on SafeSuburb. Precinct, suburb, and province pages continue to show exactly what SAPS published, unadjusted. This page is context for interpreting those numbers, not a correction to them.
Methodology
| Survey | Governance, Public Safety and Justice Survey (GPSJS) 2024/25, Statistics South Africa, Statistical Release P0341. Successor to the Victims of Crime Survey (VOCS). Published 2025-08-26. |
| Fieldwork | April 2024 - March 2025. Respondents were asked about 12 months preceding interview (moving reference period), not a fixed calendar window, so the reference period differs slightly across respondents interviewed at different times in the fieldwork year. |
| Sample design | Stats SA's Master Sample: 3,324 primary sampling units, stratified two-stage design with probability-proportional-to-size sampling, targeting roughly 33,000 dwelling units. Representative at national, provincial, and metro/non-metro level. |
| Achieved sample | 20,788 households and 20,703 individuals aged 16 and older interviewed. National response rate 87.83%; lowest area response rate 73.34% (City of Tshwane). |
| Geographic scope | This page reports national figures only. The microdata supports province and metro breakdowns, but most crime types have too few respondents at metro level to publish reliably at the precision shown here (some national samples are already as small as n=25). SafeSuburb has not built province or metro-level reporting-rate pages from this survey. |
| Weighting | All percentages use Stats SA's calibrated household or individual survey weights, which expand the sample to represent the national population. Unweighted sample sizes (n) are shown alongside each figure for transparency. |
| Independent verification | 15 of 15 reporting-rate figures on this page were independently recomputed from the raw GPSJS 2024/25 microdata and matched the published Stats SA figure within ±0.05 percentage points. Murder is cited directly from the published table, not independently reproduced, since the microdata export has no dedicated murder experience/report column. |