WP74 · The Price of a Hearing · Fee Structure and Language Access in New Jersey's Municipal Courts

1 · A notice for a hearing that was not scheduled

The specimen that prompted this paper is ordinary. A $39 citation for an expired meter. A subsequent notice asserting failure to appear at a hearing the recipient was never scheduled for, adding $10, and stating that appearing in court could result in a larger sum than paying now.

Two of those three elements are statutory and statewide. One is not, and the separation is the whole of the analysis.

element $10 added after a missed appearance
          22A:3-4, per supplemental notice — statewide
element appearing may cost more than paying
          22A:3-4, up to $33 discretionary costs on conviction — statewide
element the hearing was never scheduled
          not statutory — a local noticing practice, and the only measurable local variable

The first instinct on receiving such a notice is to read it as a municipal strategy. That reading is available, it may even be correct, and this paper does not adopt it — not from charity but because it is the weaker claim. A structure that requires intent can be dissolved by denying intent, and denial is cheap. The argument below holds with every actor in the system behaving in good faith.

2 · The schedule

Under N.J.S.A. 22A:3-4 a municipal court may assess:

item	amount	trigger	suspendable
Discretionary court costs	≤ $33	conviction	yes, at discretion
Supplemental failure-to-appear notice	$10 ea.	notice sent, unless the defendant proves non-receipt	on proof of non-receipt
Automated Traffic System	$2.00	any violation	no
ATS modernization	$3.00	any violation	no
EMT Training Fund	$0.50	motor vehicle / traffic fine	no

Note on the escape hatch

The $10 supplemental-notice assessment carries its own defeasance in the statutory text: it does not attach where the defendant proves non-receipt. A statute that anticipates the failure mode is not the same as a statute that prevents it. What the defendant must do is prove a negative about the postal history of a document they never saw, in a forum whose fee schedule discourages them from appearing to try.

3 · The dominance argument

Let $F$ be the base fine, $n$ the number of supplemental notices already assessed, and $c \le 33$ the discretionary costs imposed on conviction. A defendant chooses between settling through the violations bureau and appearing to contest:

C_pay = F + 10n
C_win ≈ 0
C_lose = F + 10n + c

Contesting is weakly preferred on money alone when $p\,C_{\text{win}} + (1-p)\,C_{\text{lose}} \le C_{\text{pay}}$, giving the indifference probability

p* = 1 − C_pay / C_lose

For the specimen — $F = 39$, $n = 1$, $c = 33$ — this is $p^* = 1 - 49/82 \approx 0.40$.

Model

A defendant must believe they hold a two-in-five chance of prevailing merely to break even on the money. Now introduce a time cost $\tau$ for the half-day an appearance consumes. The half-day is spent whether the defendant wins or loses, so $\tau$ enters both branches: $$\tau + (1-p)\,C_{\text{lose}} \le C_{\text{pay}} \quad\Longrightarrow\quad p^* = 1 - \frac{C_{\text{pay}} - \tau}{C_{\text{lose}}}$$ At any positive wage the required confidence climbs. For a defendant paid hourly and not paid for absence, $\tau$ alone can exceed $F$ — and once $\tau > C_{\text{pay}}$ the threshold exceeds unity, which is to say no probability of prevailing is high enough.

Consequence

Contesting is expected-cost-inferior for any defendant whose subjective probability of prevailing falls below $p^*$. And once the time cost exceeds the amount at stake — half a day's unpaid absence set against a $49 citation — contesting costs more than paying in every outcome, winning included. There it is strictly dominated in the exact sense of the term: no probability of prevailing makes appearing worthwhile. How many defendants stand on which side of that line is not known, and §8 records it as open. The defences are not weighed and rejected; they are never presented. What the docket records is not a population that was mostly guilty but a population that was mostly priced out, and the two are indistinguishable in every statistic the court publishes.

□

The suppression is a property of the schedule, not of anyone's conduct. That is what makes it durable, and it is why intent is the wrong thing to litigate.

§3

4 · The language layer

New Jersey's Judiciary Language Access Plan, issued under Directive #10-22 and updated in 2023, provides that an interpreter shall be furnished to any court user when that user — or that user's attorney — represents an inability to understand or communicate proficiently in English. The plan does not require the court to offer proactively. Separately, and in the same document: costs for municipal court interpretation are the responsibility of the municipality.

Who must act

The defendant, by representing a need — a representation that must be formulated and delivered in the institutional setting whose language they do not command, at a counter, under time pressure, usually without counsel.

Who bears the cost

The municipality. Which also employs the counter staff and the court personnel whose bearing, patience and phrasing determine whether the representation is invited, tolerated, or quietly discouraged.

Definition

Discretionary trigger on an unfunded mandate. An entitlement whose activation depends on an act by the beneficiary, where the cost of honouring it falls on a party positioned to influence whether that act occurs. No misconduct is required for the arrangement to bias outcomes; the incentive is structural and operates through ordinary discretion.

For a limited-English-proficient defendant, the decision problem of §3 acquires further cost terms: comprehension of the notice itself, the act of requesting an interpreter, and a longer expected appearance. Each raises $C_{\text{contest}}$ without touching $C_{\text{pay}}$. The dominance result therefore does not merely persist for this population — it deepens.

5 · What the 2018 committee did not look at

Chief Justice Rabner convened the Supreme Court Committee on Municipal Court Operations, Fines, and Fees in March 2017. It reported in 2018 with 49 recommendations, having found roughly 2.5 million outstanding bench warrants and some $22 million in discretionary contempt assessments across 2015–2017. Its guiding principle is unambiguous: financial obligations shall be imposed on the basis of the fair administration of justice, and not the generation of revenue.

Finding by absence

The report contains no recommendations concerning language access. A fairness review of the municipal courts, conducted at the level of the Supreme Court and running to 49 recommendations, did not address the defendants least equipped to navigate them. The Language Access Plan and the fines-and-fees reform effort are contemporaneous, both are Judiciary instruments, both govern the same defendants in the same rooms on the same afternoons, and neither references the other.

Implementation is partial in a pattern worth recording, because the pattern is older than the report.

recommendation area	status	evidence
Bench warrant restraint	directive	Directive #04-22 (16 May 2022), immediate release on recognizance of certain defendants
Ability-to-pay hearings	practised, uncodified	assessment required in practice; no codified standard, no presumption of indigence
Willfulness finding before incarceration	not required	no obligation to establish willful nonpayment
Court consolidation	not enacted	proposed 1958, 1969, 1971, 1979, 1983, 2018
Judicial tenure and appointment	not enacted	carried unimplemented from the 1983 task force
Revenue conflict-of-interest protection	absent	nothing bars fine revenue from funding the agencies generating it
Data collection on fines and demographics	absent	no statutory reporting obligation of any kind

The last row is not a complaint. It is the methodological opening, and §7 depends on it.

6 · Hypotheses and identification

H1. Contest rates decline in the ratio of exposed costs to base fine, across municipalities.
H2. Contest rates decline further in limited-English-proficient population share, conditional on fine magnitude.
H3. The ratio of supplemental failure-to-appear assessments to actually-scheduled events varies across municipalities beyond what caseload composition explains.

Which one is the paper

H1 alone is confounded by everything correlated with municipal wealth, and a critic disposes of it in a sentence. H2 is the claim worth defending: an interaction between fine magnitude and language composition is far harder to attribute to a common cause than a level difference, and it is the point at which §3 and §4 stop being two observations and become one. H3 isolates the single quantity that can vary locally, since the schedule in §2 is statewide — a municipality cannot be unusual in what it charges, only in what it sends.

New Jersey operates on the order of five hundred municipal courts under one fee schedule, with wide variation in fiscal capacity and language composition. The institutional multiplicity that the 1958, 1969, 1971, 1979, 1983 and 2018 consolidation proposals all treated as the defect is, for identification purposes, the resource.

7 · The instrument, and the trap inside it

No aggregate compliance record exists, which follows directly from the final row of §5. The underlying records are therefore the only evidence: docket-level dispositions by municipality, municipal demographics and budgets, and the proceedings themselves. Municipal sessions are captured on digital audio under Directive #07-10 — a corpus that substantially exceeds what has ever been transcribed from it, since transcripts are prepared on request while the recording runs regardless.

Speech recognition with speaker diarization can measure, at a scale no human coder reaches, whether an interpreter was present, whether a comprehension colloquy occurred, how long the defendant spoke, and how often they were interrupted. This is the point at which machine methods contribute something that is otherwise simply unavailable rather than merely faster.

The defect this instrument creates

Automatic speech recognition degrades on accented English and on non-English speech. Its error is therefore correlated with the treatment variable rather than independent of it. This is differential misclassification: unlike classical measurement error it does not attenuate estimates toward the null, and it can manufacture an effect of either sign. An instrument whose error tracks the variable of interest will produce exactly the finding the study hopes for, with a confidence interval attached. Mitigation: stratified human double-coding of a subsample, inter-coder agreement reported, and word-error rates estimated separately within LEP and non-LEP strata — published whether or not they are flattering.

Ordering

Any within-session analysis must first characterise how the docket was built. The parole-decision literature settled this: apparent decision-maker effects in court data frequently originate with the clerk who ordered the list rather than the adjudicator who worked through it. The confound lives in the scheduler.

8 · What this paper does not do

It does not establish that any municipality schedules hearings it does not notice, or notices hearings it does not schedule. It records one such account and specifies the measurement — H3 — that would settle it. A single receipt is a hypothesis with a date on it.

It does not claim that the fee schedule was designed to suppress contestation. The dominance result in §3 is indifferent to that question, which is its principal virtue and also its limit: an argument that does not need intent also cannot demonstrate it.

It does not know how many defendants stand on either side of the dominance threshold. That requires the distribution of subjective win probabilities among defendants holding meritorious defences, and the statewide contest rate on parking citations — neither of which is measured, and the second of which New Jersey does not publish. The threshold in §3 is computed exactly; the population on each side of it is not.

It offers no legal advice, and its arithmetic uses statutory maxima. The $33 is a ceiling rather than a fixed charge, and realised assessments vary by court — a variance that is itself a measurand and is nowhere tabulated.

Access constraint, unresolved

Judiciary records are governed by Rule 1:38 and the Judiciary's own process, not by OPRA — a distinction researchers routinely get wrong, and one that binds §7 harder than any technical question. Whether bulk audio access is obtainable at all, by request or per municipality, is unscoped as of this draft. If it is not, §7 reduces to a docket study and the interpreter-presence variable is lost.

9 · Provenance of this paper

Following the practice of WP73, the drafting errors are recorded rather than quietly repaired.

The paper began as an observation from a public gallery — court staff joking about defendants requiring interpreters — and the first framing built on that observation. It was the wrong object. Staff demeanour is an anecdote, contested by anyone present, and requires attributing motive. The structural reading in §4 was substituted because it survives conceding good faith entirely, and the same substitution was then applied to §3.
An early draft treated the warning that appearing may cost more as a local invention. It is the statewide discretionary-costs provision of 22A:3-4, stated accurately. The correction removed the paper's most inflammatory claim and improved it, since the dominance argument neither needs nor benefits from local malice.
Directive #07-23 was retrieved as an implementing instrument on warrants and ability-to-pay. It governs the Chancery Division, Family Part, on child support enforcement, is not municipal, and implements nothing in the 2018 report. It surfaces high in every relevant search and is recorded here as a false lead so the next reader does not spend the same hour on it.
The recommendation categories in §5 were read from the Judiciary's posted report; the primary PDF refused automated retrieval, and a verbatim enumeration of all 49 has not been obtained. Implementation statuses in that table combine directive text with a third-party policy index, which is outcome-level evidence rather than directive-level proof. The table is a research lead and should not be cited as a finding.
This chapter was produced across sessions of an automated assistant that retains no memory between them — the same condition that produced the defect audited in WP73, and the reason its own errors are listed here rather than in a private note.

Status

Draft. No data have been collected. H1–H3 are unfitted, no records request has been filed, and the ethics position — public-proceeding observation is ordinarily outside human-subjects review, but recording and identification of named personnel are not — is unresolved. Every statutory figure quoted is a maximum under 22A:3-4 as of August 2026 and will be false the day the schedule moves.