The Dark Figure SOC 3333 · Crime Lab · Midwestern State University
Last week you measured the gap between belief and data. This week you find out where the data itself comes from, and what it cannot see. Every crime number you will ever read was produced by a counting system with rules, blind spots, and a definition of what counts. This page walks you through the two biggest systems in American criminology, the reasons they disagree, and a counting rule that quietly erased offenses for ninety years. Guess before each reveal. Nothing here is recorded, so write your numbers down. You will use them in Forum 2.
Police data measure reported crime, not crime
Every year the Bureau of Justice Statistics runs the National Crime Victimization Survey, asking a large sample of Americans age 12 and older whether they were victimized, and then whether they told the police. The distance between those two answers is the dark figure of crime. Start with the headline number.
In 2024, what percent of violent victimizations were reported to police?
Fewer than half of violent victimizations reached the police in 2024, and fewer than a third of property victimizations did. Look at the top and the bottom of that list. Motor vehicle theft is reported at 75 percent because insurance claims require a police report, which means the incentive to report is built into the system rather than into the crime. Rape and sexual assault sit at 23.6 percent, the least reported serious offense in the survey. Both facts are about reporting behavior, not about how often the offenses happen, and any chart built only from police records inherits both distortions silently.
Source: Bureau of Justice Statistics, Criminal Victimization, 2024 (published 2025), Table 4, percent of victimizations reported to police. The NCVS covers persons age 12 or older living in U.S. households and excludes homicide, crimes against businesses, and children under 12.
The same year, counted twice
Two federal systems measured American crime in 2024. Police agencies reported offenses known to them, compiled by the FBI and published by BJS. The victimization survey asked households directly. Same country, same year, two very different numbers. Before you look, one question.
For most offenses the victim survey finds far more crime than police records do. For exactly one offense below, police recorded MORE than the survey estimated. Which one?
| Offense | Victim survey (NCVS) | Known to police | Survey ÷ police |
|---|---|---|---|
| Rape or sexual assault | 560,890 | 134,300 | 4.2× |
| Robbery | 642,150 | 212,670 | 3.0× |
| Theft (larceny) | 10,618,790 | 4,538,650 | 2.3× |
| Burglary or trespassing | 1,609,650 | 782,100 | 2.1× |
| Aggravated assault | 1,341,950 | 896,820 | 1.5× |
| Motor vehicle theft | 841,120 | 920,300 | 0.9× |
| Murder | Cannot be measured | 17,420 | n/a |
Motor vehicle theft is the exception, and the reason is a blind spot rather than a contradiction. The victim survey interviews people in households, so a stolen delivery van, rental car, or fleet vehicle never appears in it, while the police record it like any other theft. Add the high reporting rate you saw on the first tab and the survey's advantage disappears entirely. Murder is the other instructive row: the survey cannot measure it at all, for the obvious reason, which is exactly why criminologists lean on police homicide data when they want the most trustworthy crime trend.
Before you use this table as proof of anything, read the fine print. The two systems do not define offenses identically. The survey counts burglary and trespassing together while the police category is burglary alone. The survey excludes crimes against businesses, which the police data include. The survey counts victimizations while police data count offenses known. These are not rounding differences, they are different questions, and the honest conclusion is not that one system lies but that neither one alone is "the crime rate."
Sources: BJS, Criminal Victimization, 2024, Tables 1 and 2 (survey estimates, persons age 12 or older). BJS, Crime Known to Law Enforcement, 2024 (police-recorded offenses: rape 134,300, robbery 212,670, aggravated assault 896,820, burglary 782,100, larceny-theft 4,538,650, motor vehicle theft 920,300, homicide 17,420). Ratios in the last column are arithmetic from the two published counts and are shown to one decimal.
Why people stay silent
The dark figure is not an accident of paperwork. It is the aggregate of millions of individual decisions not to call. When BJS asked victims of unreported violence for their single most important reason, the answers were not what most people assume.
What was the most common reason victims gave for not reporting a violent victimization?
The single largest reason is that people handled it themselves. A third of unreported violence was treated as a private matter, settled inside a family, a friendship, a school, or a workplace. Another 16 percent came from a judgment about the police, that calling would not help or would not be worth it. Put those together and you have most of the dark figure: not fear, but two quiet conclusions that the criminal justice system was either the wrong tool or an unreliable one. Bar lengths above are drawn relative to the largest category.
Why this matters for every crime chart you will ever read. If reporting behavior changes, police-recorded crime changes even when actual crime does not. A community that begins to trust its police can look like a community with rising crime. A victim advocacy campaign that encourages reporting can look like an epidemic. When you see a jump in police-recorded offenses, "more crime" is one explanation, and "more reporting" is another, and you cannot tell them apart from police data alone.
Source: BJS, Victimizations Not Reported to the Police, 2006 to 2010 (special report), most important reason given for not reporting violent victimizations. This remains the standard BJS breakdown of reasons, and it pools five years of survey responses. Percentages sum to 99 due to rounding in the original.
The rule that erased offenses for ninety years
From 1930 until the FBI's switch to incident-based reporting, the Uniform Crime Reports used the hierarchy rule. When one incident contained several offenses, the agency counted only the most serious one and discarded the rest. The order, from most to least serious, was homicide, rape, robbery, aggravated assault, burglary, larceny-theft, motor vehicle theft, arson. Two offenses broke the rule and were always counted in addition: motor vehicle theft and arson.
Below are four incidents. For each one, check the offenses you think the old summary system would have counted, then reveal what the rule actually did. NIBRS, the system the FBI now uses, records every offense in the incident.
Keep the size of this effect honest. Roughly nine in ten incidents involve only one offense, so the hierarchy rule never erased anything in most cases, and research puts the resulting undercount at well under one percent for violent crime and a few percent for property crime. The rule did not invent the crime decline or hide a crime wave. What it did do is systematically flatten the incidents that were most complicated, which are often the most serious ones, and make it impossible to ask questions like how often burglaries turn violent. NIBRS was built to answer exactly those questions.
The larger lesson is the one to take to Forum 2. A counting rule is a measurement decision, not a fact about the world. Somebody chose an ordering of seriousness when the program was built, and that choice was written into ninety years of American crime statistics. Then came the switch itself. When the FBI retired summary reporting in 2021, roughly 7,300 of the nation's 18,000 agencies failed to submit data at all, including New York City and Los Angeles, and the 2021 national numbers became close to unusable for exactly that reason. A change in how crime is counted showed up in the data as a change in crime. Every number has a bureaucratic history.
Sources: FBI, Uniform Crime Reporting Handbook (hierarchy rule, ordered offense list, and the justifiable homicide, motor vehicle theft, and arson exceptions, including the rule that a vehicle stolen during the unlawful entry of a building is counted as burglary). FBI, National Incident-Based Reporting System documentation. Magnitude estimates from published analyses of NIBRS data comparing incident-based and hierarchy-rule counts. 2021 transition figures: The Marshall Project analysis of FBI reporting (about 7,287 of roughly 18,000 agencies did not submit 2021 data, including New York City and Los Angeles).
Pick your source
No source is "the truth," so the practical skill is matching the question to the system that can actually answer it. Five research questions. For each, choose the source you would reach for first. There is a defensible answer in every case, and the reasoning matters more than the label.
Triangulation is the whole method. When police data and victim surveys move in the same direction, criminologists get confident. When they diverge, that divergence is itself a finding worth explaining, and the explanation is usually about reporting, definitions, or coverage rather than about a sudden change in human behavior. Carry one habit out of this page: whenever you meet a crime statistic, ask which system produced it and what that system cannot see.
Forum 2: your post
Write roughly 250 to 300 words for your initial post, then reply to two classmates. Check that week’s module in the LMS for the posting and reply deadlines. This week you go to the databases yourself, and both are linked in the module.
- Pick one offense that interests you, ideally one connected to the offense type you are considering for your term paper.
- Look it up in the FBI Crime Data Explorer (police-recorded data) and in N-DASH, the BJS victimization dashboard (survey data). Report both numbers, with the year and the exact measure you used.
- Explain why they differ. Use the vocabulary from this week: the dark figure, reporting rates, definitional differences, coverage, the hierarchy rule and the shift to NIBRS. Point at a specific mechanism rather than saying the numbers are just different.
- Take a position. Which source would you trust more for that specific offense, and why? Notice from this page that the right answer changes by offense, so defend yours.
For your replies: ask about the mechanism. If a classmate picked the survey, ask what it cannot see for their offense. If they picked police data, ask how reporting behavior could be shaping their number. If you both picked the same offense and reached different conclusions, say so plainly and work out where you diverged.
One more thing while you are choosing an offense. Term Paper Part 1 is introduced the same week as this forum, and the offense you pick here can be the one you live with all semester, so treat this post as a test drive. Your first required office hours meeting, virtual or in person, comes due before Part 1 does. Both dates are in the LMS.