Standards of Proof

Two standards, and only two

At its foundation, the difference between the criminal and the civil legal tests is a difference in the standard of proof: the level of certainty a tribunal must reach before it will treat a fact as established. New Zealand law knows two such standards. The criminal standard is proof beyond reasonable doubt. The civil standard is proof on the balance of probabilities. Everything that follows in this area is a question of how those two standards are applied, not of how many standards exist.

The distinction is not a technicality. It reflects a deliberate choice about what each branch of the law is for. The criminal law sets the bar high because the Crown is seeking to condemn and punish an individual, and the law would rather let a guilty person go free than convict an innocent one. The civil law sets it lower because civil litigation exists to resolve disputes and allocate loss between parties, not to brand anyone a wrongdoer in the eyes of the community.

The criminal standard: beyond reasonable doubt

The criminal standard is the most demanding the law imposes. It does not require certainty, and it does not require the elimination of every fanciful or theoretical possibility. It requires that, on the whole of the evidence, the tribunal is sure — that any remaining possibility in the accused's favour is so remote that it can fairly be set aside. The classic statement is that of Denning J in Miller v Minister of Pensions [1947] 2 All ER 372, a formulation repeatedly adopted in New Zealand courts.

The standard does more than govern the verdict; it shapes the decision to prosecute at all. Under the Solicitor-General's Prosecution Guidelines, no prosecution proceeds unless the evidence is sufficient to provide a reasonable prospect of proof beyond reasonable doubt, and unless a prosecution is required in the public interest. The threshold is therefore engaged long before any trial begins.

The civil standard: the balance of probabilities

The civil standard asks a single, comparatively modest question: is the party's version of events more probable than not? If the tribunal can say that it is more likely than not that something occurred, the burden is discharged. If the probabilities are evenly balanced, it is not, and the party carrying the burden fails. There is no halfway house and no intermediate standard sitting between the civil and the criminal.

This is the standard that governs personal grievances, wage and holiday-pay claims, breach-of-contract disputes, and the great bulk of what is litigated before the Employment Relations Authority and the Employment Court.

One standard, applied flexibly

The real subtlety lies not in the definition of the civil standard but in its application. The balance of probabilities is not a mechanical exercise. The Supreme Court confirmed in Z v Dental Complaints Assessment Committee [2008] NZSC 41 that, although there is only one civil standard, it is applied flexibly according to the seriousness of the allegation and the gravity of the consequences. The standard itself never changes — the question remains whether the fact is more probable than not — but the cogency of the evidence required to satisfy a tribunal of that probability rises with what is at stake.

Two ideas drive that flexibility. The first is inherent improbability: serious misconduct is, in the ordinary course of things, less likely to have occurred, so stronger evidence is needed to displace that improbability. The second is consequence: where a finding would cost a person their livelihood or their reputation, a tribunal will look for evidence of corresponding weight before it is persuaded. The High Court of Australia captured the underlying idea in Briginshaw v Briginshaw (1938) 60 CLR 336, requiring an actual persuasion of the existence of the fact rather than a mechanical weighing of bare probabilities. It follows that an acquittal on the criminal standard does not preclude a finding of liability on the same facts in later civil proceedings: the two questions are different questions, asked for different purposes, against different thresholds.

The standard in the employment jurisdiction

In the employment jurisdiction the flexible civil standard has been applied directly to serious allegations.

There is, however, a distinction that must be kept firmly in view, because it is the point at which generic statements of the civil standard mislead. Where the matter in issue is the justification of a dismissal or other action, the Employment Relations Authority and the Employment Court do not themselves decide, on the balance of probabilities, whether the alleged misconduct occurred. They apply the statutory test of justification.

That is a reasonableness inquiry into the employer's conduct, not a burden of proof resting on the employer to establish guilt to the tribunal's own satisfaction. The two must not be collapsed. The balance of probabilities governs the tribunal's own findings of primary fact; the reasonableness standard in section 103A governs whether the employer's decision and process can stand.

Where the employment jurisdiction comes closest to the criminal in feel, it is in the penalty regime. Pecuniary penalties and declarations of breach are quasi-punitive, yet they too are determined on the civil standard. Section 142S of the Employment Relations Act 2000 is headed "Standard of proof" and fixes that standard as the balance of probabilities. The quasi-punitive character does not raise the standard; through the flexible approach, it raises the cogency of the evidence a tribunal will expect, as the penalty proceeding in Carr, Labour Inspector v Sharma [2016] NZEmpC 151 illustrates. The same logic governs grave allegations raised before the Authority, as in JXC v VGM [2023] NZERA 554.

Proving the facts: admissibility and the employment exception

A standard of proof tells a tribunal how firmly it must be persuaded. It does not tell it what material it may use to get there. That second question — admissibility — is governed in the ordinary courts by the Evidence Act 2006, which replaced a tangle of common law rules with a single principled code. Its fundamental principle, in section 7 of the Evidence Act 2006, is that all relevant evidence is admissible unless a statute excludes it; and its central safeguard, in section 8 of the Evidence Act 2006, requires a Judge to exclude relevant evidence where its probative value is outweighed by the risk of an unfairly prejudicial effect, or of needlessly prolonging the proceeding.

That code is weighted more heavily in favour of the accused in criminal cases, which is where the higher standard bites. The general exclusion is applied with the defendant's right to an effective defence expressly in mind; propensity evidence offered by the prosecution faces a stricter test than the ordinary balancing exercise; and improperly obtained evidence is presumptively excluded. These protections are the procedural counterpart of proof beyond reasonable doubt and the presumption of innocence.

The employment jurisdiction is different, and deliberately so. The Evidence Act 2006 does not formally bind the Employment Relations Authority or the Employment Court. In the Court, admissibility is governed instead by section 189 of the Employment Relations Act 2000.

That latitude is not licence for an arbitrary approach. The Employment Court has consistently held that, although it is not bound by the Evidence Act 2006, the Act's principles and contents remain an important source of reference whenever admissibility is in question, so that the broad discretion is exercised fairly and logically rather than at large. The relevance principle in section 7 of the Evidence Act 2006 and the general exclusion in section 8 of the Evidence Act 2006 are drawn on by analogy, as Lyttelton Port Company Limited v Pender [2019] NZEmpC 86 and GF v Comptroller of the New Zealand Customs Service [2023] NZEmpC 101 illustrate. The result is a jurisdiction that gathers the facts more fully than the strict rules of evidence would allow, while still applying the civil standard — the balance of probabilities — to what it finds.

A common example in practice: the manufactured screenshot

The flexibility of the employment jurisdiction has a practical edge that we see regularly. Because a screenshot or a photograph of a phone screen is so easily admitted, it is also easily manufactured. A common trick is to rename a contact in a phone, take a screenshot of an ordinary conversation, and present it as a message from someone else entirely — an employer, a manager, a colleague. The image looks convincing. On its own, it proves very little.

Consider the version that turns up most often. A photograph shows a phone held in the hand, a messaging app open, and at the top of the screen nothing but a saved contact name — a colleague's, or a manager's — beside a generic coloured initial, then a short personal message of the kind that might ground a complaint. What the image does not show, anywhere, is a phone number. That absence is decisive. The contact name is simply text the owner of the phone typed into their own address book; it can be changed in seconds, and on the manufactured version it is changed before the screenshot is taken. Nothing on the screen ties those words to the person named rather than to anyone at all whose contact was relabelled beforehand. A saved label is not proof of a sender. Strip the label away and what is left is an unattributed message from a number that is not even displayed — which is to say, almost nothing.

A photograph of a text message is secondary evidence, and a bare screenshot with no sender details has little probative value. Under the equity and good conscience approach the Authority and the Court will admit it, but admission is not weight. To carry any weight, the image must be supported by corroborating circumstances — independent evidence connecting the message to its alleged author. That is exactly where a manufactured screenshot tends to fall apart.

The most decisive check is the independent telecommunications record. In Chiu v New Deli & Café Ltd [2008] NZERA 260, faced with an allegation that call logs had been faked, the Authority used its powers to call for the carrier's records directly, which settled the question. The same logic exposes the renamed-contact trick: a discovery order for the phone records will show messages exchanged with the friend's real number, not the number of the person the screenshots purport to be from. Attribution can also be drawn by inference from independent documents, as in Brown v Adams [2014] NZEmpC 139, where a number was tied to its user because it appeared on documents the user had plainly authored.

Discovery of the original device, cross-examination on the maker's veracity, and analysis of metadata all do similar work. The Authority engages with this directly: in Labour Inspector v Basra & Khella Limited [2020] NZERA 534 the device was supplied at the investigation meeting and the photographs of a screen were examined alongside their metadata against an allegation that the computer's clock had been altered to change the date and time shown. Metadata is not foolproof — the time stamp on a photograph of a screen is only as reliable as the device clock behind it, which can be changed — but inconsistencies of exactly that kind are what give a fabrication away. The point for any worker tempted by a shortcut is blunt: the system is built to find this out.

Why the two standards differ

The differing standards mirror the differing functions of the two branches of law. The criminal law condemns and punishes: its consequences — loss of liberty, fines paid to the Crown, and the lasting stigma of a record — run between the Crown and the individual, and a conviction is a formal moral pronouncement by the community. The civil law compensates and resolves: its remedies run between the parties, and an adverse finding, however serious in its practical effect, does not carry public censure of the same kind.

For the worker and the advocate, the practical message is that the seriousness of an allegation cuts both ways. It does not lift the bar to the criminal standard, but it does mean that grave allegations — of theft, dishonesty, or violence in the workplace — must be supported by evidence of real cogency. A bare assertion will not carry a serious charge, even on the balance of probabilities.

Challenging how the standard was applied

Where a party considers that the Authority weighed the evidence wrongly, or misapplied the standard of proof, the route of challenge is not judicial review. It is the challenge to the Employment Court under section 179 of the Employment Relations Act 2000 which, where a challenge by way of hearing de novo is elected, gives a full rehearing on the merits in which the Court reconsiders the evidence and the findings for itself.

This is not merely a matter of practice. The general judicial review jurisdiction expressly yields to the employment regime: section 7 of the Judicial Review Procedure Act 2016 provides that that Act is subject to the provisions of the Employment Relations Act 2000 governing the jurisdiction of the Employment Court. A dissatisfied party therefore corrects an erroneous application of the standard of proof by way of the de novo challenge, not by seeking review in the High Court.

Legislation cited is linked to the current consolidated text on the New Zealand Parliamentary Counsel Office website at legislation.govt.nz. Case authorities are linked to the Working for Workers case law and reference repository at caselawandreference.workingforworkers.nz, except Miller v Minister of Pensions, Z v Dental Complaints Assessment Committee, and Briginshaw v Briginshaw, which are not held in the repository — being English, New Zealand Supreme Court, and Australian authorities respectively — and are cited in full by name so the reader may locate them in the official reports.