Two departments, the same complaint, very different odds

Across New Jersey, 17.4% of internal affairs investigations opened between 2021 and 2025 ended in a sustained finding: 14,345 of 82,507.^[1][2] Every figure below uses that same investigation-level field, which is not the only sustained count the Attorney General publishes and is the lower of the two.

That statewide figure conceals an enormous amount. Narrow to the 74 agencies that opened at least 200 investigations across the period, so that small numbers cannot drive the result, and the sustained rate runs from 0.3% at one end to 41.4% at the other.

A complaint filed at one New Jersey department is on the order of a hundred times more likely to be sustained than the same complaint filed at another.

The range

At the high end:

Agency	Sustained	Investigations	Rate
Linden Police Department	182	440	41.4%
Lakewood Police Department	88	215	40.9%
Union County Sheriff's Office	85	212	40.1%
Essex County Sheriff's Office	284	769	36.9%

At the low end:

Agency	Sustained	Investigations	Rate
Morris County Corrections	4	219	1.8%
Morris County Prosecutor's Office	2	270	0.7%
Department of Corrections	42	8,758	0.5%
Plainfield Police Department	1	305	0.3%

The Department of Corrections figure is the one that stops you. Nearly nine thousand internal affairs investigations across five years, of which 42 were sustained. It is by a wide margin the largest single internal affairs operation in the state, and its sustained rate is a fortieth of the statewide average.

What a sustained rate is not

The temptation is to read a high rate as a department policing itself seriously and a low rate as a department covering for its officers. Resist it, in both directions.

Sustaining a complaint is an act of classification before it is an act of judgement. Departments differ in how they count. One agency records a single incident as one investigation; another records it as four allegations against three officers and closes three of them. One agency opens a formal investigation on every complaint that arrives, including the frivolous and the anonymous; another screens at intake and only opens files on complaints that survive a first look. The first will have a lower sustained rate than the second while doing more work, not less.

The population being policed matters too. The Department of Corrections investigates complaints arising inside prisons, where the complainant is incarcerated, the incident is often unwitnessed by anyone neutral, and the evidentiary picture is different from a traffic stop. A low sustained rate in that setting is not self-explaining in either direction.

None of that makes the spread uninteresting. It makes it a question rather than an answer.

What the spread does establish

One thing it does establish is that the sustained rate cannot be used as a comparative measure of misconduct without a great deal more information than this data contains. That matters because it is exactly the use such a number invites.

It also establishes that complaint outcomes in New Jersey are not produced by a uniform process. Whatever else is going on, two agencies handling the same complaint are not running it through the same machine. For a system built on agencies investigating themselves, that variation is the finding, independent of which end of the range is closer to correct.

Why this site does not rank on it

This site publishes rankings on several metrics. Sustained rate is not one of them, and this is the reason.

A ranking implies the metric orders agencies from better to worse. For internal affairs volume the site publishes a ranking while stating plainly that a high number is ambiguous, because volume at least measures a real quantity of recorded activity. A sustained rate measures the interaction of a classification practice, a screening practice and an evidentiary standard, and the same number can be produced by opposite underlying realities.

The figures are on each department page, alongside the investigation counts they come from, so anyone can compute the rate and read it with the caveats attached. What the site will not do is put them in a league table with a header implying the top is good.

What would make this legible

Two changes to what agencies report would resolve most of the ambiguity.

The first is a count of complaints received alongside investigations opened. The gap between those two numbers is the screening step, and it is currently invisible, which is what makes a low sustained rate impossible to interpret.

The second is a consistent definition of what constitutes one investigation. Whether a single incident involving three officers is one file or three changes every rate computed from this data, and nothing published says which convention an agency used.

Both are counting questions rather than policy questions, and neither would require an agency to disclose anything about an individual case.