Reporting a p value correctly is a mechanical formatting skill layered on top of a substantive methodological question: a p value on its own tells a reader almost nothing about whether a finding matters. This guide covers both. It walks through the APA-style formatting conventions for writing p values in a manuscript — exact-value reporting versus threshold reporting, decimal precision, and the specific rule against ever writing “p = .000” — and then covers the broader, now well-established methodological expectation that a p value should not appear in a manuscript alone, unaccompanied by an effect size and confidence interval.
Exact-value reporting vs. threshold reporting
There are two ways researchers report statistical significance, and current style guidance treats them as not equally acceptable defaults:
- Threshold reporting states only whether a result crossed a pre-set cutoff: “the difference was significant, p < .05.” This tells the reader a boundary was crossed and nothing else — a result with p = .049 and a result with p = .0001 are reported identically.
- Exact-value reporting states the actual computed p value: “the difference was significant, p = .032.” This preserves the actual strength of evidence against the null and lets a reader, or a later meta-analysis, use the real number rather than a censored range.
Why exact values are now generally preferred
APA’s Publication Manual (7th edition, 2020) and its companion Journal Article Reporting Standards (JARS) direct authors to report exact p values for all values greater than .001, reserving threshold notation only for the tail below that point (see the next section). The rationale is straightforward: threshold reporting was a legacy of an era when p values were looked up in printed tables at fixed cutoffs (.05, .01, .001) rather than computed exactly by statistical software. Once software reports an exact value as a matter of course, rounding it down to “< .05” discards information the analysis already produced for free, and it makes it impossible for a reader to judge how close a result sat to a threshold or to recompute the finding in a later meta-analysis. Modern statistical and editorial guidance — and most contemporary journal Instructions for Authors in psychology, education, and the biomedical and social sciences — now expect exact values as the default, with a fixed threshold cutoff (p < .05) reserved mainly for describing a general decision rule (e.g., “alpha was set at .05”) rather than for reporting an individual result.
Formatting mechanics
- Report exact p values to two or three decimal places (e.g., p = .032 or p = .008), consistent with the precision used elsewhere in the results.
- Drop the leading zero: write p = .032, not p = 0.032 — because a p value can never be less than 0 or greater than 1, APA style omits the leading zero on any statistic that is bounded to that range (the same rule applies to correlation coefficients and proportions).
- Use a space around the equals or inequality sign: p = .032, p < .001 — not p=.032.
- Italicize the p (as with other statistical symbols such as t, F, d, and r) when the manuscript’s formatting supports italics.
Never report “p = .000”
Some statistical software packages (SPSS is the most commonly cited example) round very small p values down to three decimal places and display them as “.000” in the output table. Copying that value directly into a manuscript as “p = .000” is a reporting error, not a correct exact value — a p value is a probability computed from a continuous test statistic and is mathematically never exactly zero, even when it is extremely small. What the software is actually showing is a value smaller than .0005 that its default display rounds to zero. The correct APA-style report for any p value below .001 is the threshold form p < .001, not the exact (and false) value the software happened to truncate to. This is one of the most common statistical-reporting errors editorial offices and peer reviewers flag, and it is easy to avoid: check the software’s underlying precision setting (most packages can display more decimal places, or scientific notation, on request) before assuming .000 is the real value, and default to “p < .001” whenever the exact figure is unavailable or genuinely below that cutoff.
The broader shift: p values alone are no longer sufficient
Correct p-value formatting solves a presentation problem; it does not solve the more consequential problem that a p value, reported by itself, does not tell a reader how large or practically meaningful an effect is, or how precisely it was estimated. A very small effect can produce a very small p value given a large enough sample, and a real, meaningful effect can fail to reach conventional significance in a small or underpowered study — statistical significance and practical importance are separate questions that a p value alone cannot distinguish.
This concern is not new, but it has moved from methodological commentary into formal reporting requirements over the past decade. The American Statistical Association’s 2016 Statement on Statistical Significance and P-Values (published in The American Statistician) was an unusually direct professional-society statement warning against treating a p value crossing .05 as, by itself, evidence of the presence or absence of an effect, or as a substitute for reporting effect size. Consistent with that shift, APA’s Publication Manual (7th ed.) and JARS now direct authors to report an effect-size estimate — and, wherever possible, a confidence interval for that estimate — alongside every inferential test, not as an optional supplement. A results sentence built to current standards therefore reports three things together, not one: the test statistic and exact p value, an effect-size estimate appropriate to the test (Cohen’s d, r, η², an odds ratio, etc.), and a confidence interval (conventionally 95%) around either the effect size or the relevant parameter estimate. A representative example in APA format: “there was a significant difference, t(48) = 2.74, p = .009, d = 0.77, 95% CI [0.19, 1.35].”
The practical implication for anyone drafting a Results section: treat “did I report the p value correctly” and “did I report an effect size and confidence interval next to it” as a single checklist item, not two separate ones. A manuscript that formats its p values flawlessly but omits effect sizes throughout is still incomplete by current reporting standards, and increasingly likely to draw a reviewer comment asking for the missing effect-size estimates before the manuscript can proceed. See CASRAI’s Confidence Interval dictionary entry for the operational definition of the CI half of that pairing.
Where p-value misuse becomes a research-integrity issue
Formatting and completeness are one layer of this topic; how a p value is arrived at is a separate, more serious layer. Practices like running many statistical tests and reporting only the one that crossed .05, or continuing to collect data specifically until a result becomes significant, do not show up as a formatting error — the reported p value can look perfectly correctly formatted while still misrepresenting the actual strength of evidence. CASRAI’s P-hacking vs. HARKing comparison covers the two most commonly cited questionable research practices that inflate false-positive findings through exactly this kind of undisclosed flexibility in analysis and hypothesis-framing, and is the natural next read for anyone drafting or reviewing a Results section where the reported statistics need to be trusted, not just correctly typeset.
Frequently asked questions
Is p < .05 or p = .032 correct?
Both are grammatically valid APA notation, but current guidance treats exact-value reporting (p = .032) as the expected default for any p value above .001. Reserve the threshold form (p < .05) for describing a general significance criterion set in advance (e.g., in a Method section: “alpha was set at .05”), not for reporting the result of an individual test where the software has already computed an exact value.
Why is p = .000 wrong?
Because a p value is a continuous probability and is mathematically never exactly zero. “.000” is an artifact of a software package’s default decimal-rounding display for any very small value, not the true computed value. Report it as p < .001 instead, and check whether the software can display more decimal places if a more precise figure is genuinely needed.
How many decimal places should a p value have?
APA style generally calls for two or three decimal places for exact values (e.g., .03 or .032), with consistency maintained across a manuscript’s results. Values below .001 switch to the threshold form (p < .001) rather than being reported to additional decimal places.
Do I need to report an effect size every time I report a p value?
Per APA’s Publication Manual (7th ed.) and JARS, yes — an effect-size estimate, and a confidence interval around it wherever possible, is expected alongside every inferential test result, not treated as optional supplementary detail.
What’s the difference between statistical significance and practical significance?
Statistical significance (a p value below a chosen threshold) only indicates that an observed result is unlikely under the null hypothesis; it says nothing on its own about the size or real-world importance of an effect. A large enough sample can make a trivially small effect statistically significant, while a genuinely meaningful effect can fail to reach significance in an underpowered study. Effect size, not the p value, is the measure that speaks to practical significance.
Related CASRAI resources
- How to Write a Research Paper in APA Style — the fuller APA 7th edition manuscript-formatting guide this page’s rules sit inside.
- P-hacking vs. HARKing: what is the difference? — how undisclosed analytic flexibility and post-hoc hypothesis framing can misrepresent an otherwise correctly formatted p value.
- Confidence Interval — the operational definition of the interval estimate APA expects alongside effect sizes.
- Scholarly Writing hub — CASRAI’s full cluster on manuscript preparation and academic writing craft.
