Public debate about criminal justice tends to focus on the end of the process: sentence lengths, prison populations, parole decisions. But those decisions apply to only a small slice of the offenses that actually occur. Before anyone is sentenced, a case has to pass through a long series of filters, and most cases do not make it.

Criminologists call this case attrition, and it is often pictured as a funnel. At the wide top sit all crimes committed. At the narrow bottom sit the comparatively few that end in a conviction. Each stage in between drops a substantial share of cases: reporting, recording, investigation, arrest, charging, and adjudication. Because each stage works on what the previous one let through, the losses multiply.

The funnel matters because it shapes what reform can realistically accomplish. A policy that doubles a penalty affects only cases that reach the bottom. A policy that improves detection affects every stage below it. Tracing where cases fall out, and why, is the first step toward judging which levers actually move public safety and which mostly move headlines.

The largest losses occur before police are involved at all. In the United States, the National Crime Victimization Survey asks households directly about their experiences of crime, whether or not they called the police. It consistently finds that fewer than half of violent victimizations are reported to police. For property crimes the figure is lower, typically around a third. Victimization surveys in other countries generally find a similar pattern: a large share of crime never enters official view.

Victims give reasons that are rational from their perspective. The loss seemed too small to bother. They assumed police could do nothing. The offender was a partner, relative, or acquaintance. They feared retaliation, or they distrusted the authorities. Sexual violence has some of the lowest reporting rates of any offense category. Motor vehicle theft has some of the highest, largely because insurance claims require a police report. Reporting follows incentives as much as harm.

A report is also not the same as a record. Officers exercise discretion over whether an incident is written up as a crime, downgraded to a lesser category, or classed as non-criminal. In 2014, the police inspectorate in England and Wales audited crime recording. It estimated that roughly one in five crimes reported to police was never recorded, and the shortfall was considerably higher for sexual offenses. Performance targets that reward falling crime figures can quietly encourage such gaps.

As a result, official crime statistics measure victim behavior and police practice as much as they measure offending. A rise in recorded domestic violence may reflect greater willingness to report, not more violence. A drop in recorded burglary may reflect changed recording rules. Analysts who rely on recorded crime alone to compare jurisdictions or track trends risk mistaking a change in the filter for a change in the underlying problem.

Takeaway

Recorded crime is not a measure of crime. It measures the crime that victims chose to report and police chose to record, so any trend in the data could reflect a change in either decision.

Once a crime is recorded, the next question is whether anyone is held responsible for it. Police measure this through the clearance rate. That is the share of offenses closed by an arrest or, less often, by exceptional means, such as when the suspect has died or the victim refuses to cooperate. In FBI data, US clearance rates for violent crime have hovered in the forties for decades. Property crime clearances generally sit below twenty percent.

Those averages conceal enormous variation by offense. Homicide historically has the highest clearance rate, although it has fallen from roughly 90 percent in the 1960s to around half or somewhat more in recent years. Aggravated assault clears about half the time and robbery roughly a third. Burglary and motor vehicle theft often clear in only around one case in seven, or fewer. The pattern follows evidence. Crimes with witnesses, injured victims, and offenders known to the victim are easier to solve than anonymous property crimes.

Jurisdiction matters too. Clearance rates differ widely between cities with similar crime levels. Detective caseloads, forensic capacity, and community cooperation all shape the outcome. Where residents distrust police, witnesses stay silent, and each unsolved case erodes trust further. Several studies have found that homicides of victims in marginalized neighborhoods clear at lower rates. That raises questions of equal protection, not just efficiency.

The definition of clearance also has a catch. A crime counts as cleared when someone is arrested and referred for prosecution, not when anyone is convicted. A single arrest can clear several offenses at once. Pressure to improve the statistic can encourage practices that close cases on paper without resolving them in fact. The metric tracks investigative output, not whether the right person was ultimately held to account.

Takeaway

Most recorded crimes are never solved, and which ones get solved depends heavily on evidence, resources, and community trust. The certainty of being caught is largely determined during investigation, long before any sentence is imposed.

An arrest opens the courtroom phase, but attrition continues. Prosecutors may decline to file charges for several reasons: the evidence is weak, witnesses are unavailable, the arrest was legally flawed, or the case is better handled through diversion. Declination rates vary sharply between offices. Those differences often reflect each office's charging standards and policy priorities rather than differences in the underlying cases.

Among charged felony cases, a substantial minority are dismissed. Studies of felony defendants in large US urban counties have found that roughly two-thirds were convicted and around a quarter had their cases dismissed. Of those convicted, the overwhelming majority, typically well over 90 percent, plead guilty rather than go to trial. Many pleas are to reduced charges, so the conviction on record often differs from the offense that started the process.

Incarceration narrows the funnel further. Many convictions lead to probation, fines, or time served rather than a prison term. Stacking the stages together produces striking arithmetic. Suppose roughly 40 percent of serious crimes are reported, perhaps a fifth of those lead to an arrest, and about two-thirds of arrests end in conviction. Under those illustrative assumptions, only around five percent of offenses produce a conviction, and fewer still produce a prison sentence.

This does not mean most offenders escape entirely. Offending is concentrated among a relatively small group of people. Those who commit many crimes face repeated chances of detection, so a meaningful share are eventually caught for something. But they are rarely caught for most of what they do, and the sanction that follows attaches to a sliver of their conduct. The system punishes samples, not totals.

Takeaway

Criminal courts see a small, filtered sample of offending, shaped by discretion and plea bargaining at every step. A conviction shows that a person was caught once, not how much crime they committed or how much went unanswered.

The attrition funnel is not simply a malfunction to be eliminated. Some filtering reflects sound judgment. Weak cases should be dropped, minor matters diverted, and civil liberties protected. A system that turned every offense into a conviction would require a level of surveillance few people would accept.

Still, the funnel's shape has policy implications. Deterrence research consistently suggests that the certainty of being caught matters more than the severity of the punishment. Yet reform debates disproportionately focus on the narrow end, adjusting penalties for the few cases that reach it.

A more effective agenda would ask where cases fall out and why: why victims stay silent, why investigations stall, and why charges collapse. Building community trust, investigative capacity, and evidence quality widens the funnel fairly. That is where justice and public safety tend to align.