Does the BFI-2 Have Official Norms for Individual Percentiles?

Does the developer publish official BFI-2 norms?

No. On its “Big Five Inventory” page, the Berkeley Personality Lab says that it does not have an official BFI-2 manual with published norms. The page offers BFI-2 scoring instructions separately. Those instructions explain how to score responses; their availability is not itself evidence of a norm manual or a published conversion from a score to an individual percentile. The two resources have different jobs: scoring specifies how responses are combined into scale scores, while a norm publication would document a reference basis for interpreting relative standing. A scoring key should therefore not be described as a norm table simply because it is an official resource.

The distinction is about what the developer publicly supplies. The page’s direct statement supports saying that official developer norms are not published there. It does not establish that no independent researcher or provider has collected reference data, nor does it prove that a percentile could never be calculated from a suitable dataset. It does mean that a report cannot treat “the BFI-2” alone as identifying an official developer percentile scale. The comparison group and basis for any rank have to come from somewhere more specific than the instrument’s name.

This answer is deliberately narrow: it concerns an official BFI-2 norm manual from the developer, not every possible use of BFI-2 scores or every piece of research associated with the inventory. Berkeley’s page also points to material about the earlier BFI; that reference does not change what the page says about an official BFI-2 manual. A reader should therefore distinguish the status of the developer’s documentation from the separate question of whether an outside comparison dataset is appropriate for a particular report. The absence of the former settles only the official-publication claim; it does not settle the quality or availability of the latter.

Sources: Big Five Inventory — Berkeley Personality Lab

What does the BFI-2 development research establish?

Validation evidence concerns whether particular interpretations of scores are supported for defined purposes and populations. It is not one all-purpose stamp: evidence about how items group into scales, how consistently a scale measures, or how scores relate to other measures addresses different questions from where an individual falls in a reference distribution. Each result has a target: a score meaning, a relationship, a population, or a use. A finding can be relevant to more than one target, but it does not automatically answer questions the study was not designed to address.

The 2017 paper “The Next Big Five Inventory (BFI-2): Developing and Assessing a Hierarchical Model With 15 Facets to Enhance Bandwidth, Fidelity, and Predictive Power” reports development and evaluation of the full 60-item BFI-2. Its hierarchical structure has five broad domains, each represented by three narrower facets, for 15 facets in total. The paper evaluates the inventory in validation samples and examines whether the domain and facet scales capture personality content with useful breadth and fidelity, including their structure and relationships relevant to the intended measurement model. These are substantive measurement aims: the study is evidence about what the scale scores can represent and how the proposed hierarchy performs in the samples studied.

That work can support interpretations of BFI-2 domain and facet scores without answering a different question: what proportion of a specified population scores below one person? A validation sample is not automatically a norm group merely because it contains participants and produces score distributions. A percentile requires locating a score within a chosen reference distribution; the study must provide or be paired with a distribution intended for that comparison and a procedure for making the conversion. The development paper’s description of scale construction and validation, as recorded in the study, does not supply a general-purpose individual percentile table.

The distinction is therefore about evidence type and inference, not a judgment that the instrument is poor. Reliability evidence concerns consistency under specified conditions; evidence for a factor structure concerns whether a proposed organization of scores fits the observations; relations with other variables can show whether scores behave as expected in the studied context. Such findings may strengthen some interpretations and may be valuable to researchers and users. None alone identifies the distribution in which a particular person’s score should be ranked, or establishes that a distribution represents the population a report implies.

A practical reading rule follows: treat a development or validation result as support for the score interpretation it actually examines. Do not convert that support into a percentile claim unless the relevant reference distribution and conversion are also documented. The kinds of evidence can complement one another, but one cannot silently substitute for another. The 2017 BFI-2 paper establishes an investigated domain-and-facet measurement model; that contribution remains meaningful even though the paper does not, by that fact alone, tell a reader an individual’s relative standing in a chosen population.

Sources: The Next Big Five Inventory (BFI-2): Developing and Assessing a Hierarchical Model With 15 Facets to Enhance Bandwidth, Fidelity, and Predictive Power

Why does the exact BFI-2 form matter?

The full BFI-2 contains 60 items. The shorter instruments are distinct forms: the BFI-2-S contains 30 items, and the BFI-2-XS contains 15. In “Short and extra-short forms of the Big Five Inventory–2: The BFI-2-S and BFI-2-XS,” the researchers describe developing and evaluating these abbreviated versions to reduce respondent burden while retaining useful measurement. Shortness alone does not make a form invalid; it changes which form’s evidence a reader needs to inspect.

A shared family name is not a score-conversion rule. The BFI-2-S and BFI-2-XS draw on the same broad inventory framework, but their item sets differ from each other and from the 60-item form. A derived percentile depends on the scores produced by the particular form and on the comparison data and conversion used for those scores. Related content may make results scientifically comparable for some questions, but that possibility does not establish that the forms produce identical score distributions or interchangeable individual ranks. The source record for the short-form study describes measurement evaluation and tradeoffs; it does not document a conversion that turns a short-form result into a full-form percentile.

This distinction matters when a report or cited study names only “BFI-2.” Before treating its percentile as relevant, identify whether the respondent completed 60, 30, or 15 items, and whether the comparison distribution and scoring procedure were built for that same form. If the form is omitted, the rank’s basis is underspecified. If the study used a short form, its findings may inform claims about that short form and its studied samples; they cannot be carried over as full-form individual norms without additional evidence establishing that specific link. That is a limit on the inference, not a reason to dismiss the abbreviated instruments or every result obtained with them.

A 15-item study can show what researchers evaluated that version, but item count and a shared framework do not show how a 60-item score ranks in a population. A short-form report likewise needs reference data for that form; using the family name does not supply it. Check that documentation links the administered form to the distribution and percentile procedure. Without that link, do not assume the full and abbreviated forms have interchangeable individual ranks.

Sources: Short and extra-short forms of the Big Five Inventory–2: The BFI-2-S and BFI-2-XS

What do large survey findings contribute—and leave open?

The OECD report “Psychometric properties of the BFI-2 inventory in the 2023 Survey of Adult Skills” is a useful boundary case. It examines the 30-item BFI-2-S and 15-item BFI-2-XS in an international adult-skills survey covering adults in 29 participating countries. This is substantial cross-country measurement research: it asks how short-form personality measures function in a defined survey setting, rather than offering a full-form manual or an individual score lookup table. Its size and breadth make its findings relevant to carefully bounded comparisons, but the study’s stated forms and survey context remain part of what those findings mean.

The OECD analysis considers translated survey deployment and cross-country comparability, including whether the measurement structure and items behave similarly across participating settings. The report describes a mixed pattern: many domains support useful comparability, while some items show noninvariance, and some model-based comparisons face limits. In practical terms, this is neither a blanket endorsement of every country-to-country comparison nor evidence that the short forms are unusable. The analysis identifies where comparisons may be informative and where differences in item behavior or model assumptions constrain interpretation. The participating countries, translated forms, and survey population are the conditions under which those conclusions were examined.

That contribution answers a different question from whether a report can assign one person an official, general-purpose percentile on the full BFI-2. The OECD report studies two abbreviated forms in a particular survey program; it does not supply a conversion table for the 60-item form. Nor does cross-country comparability analysis by itself define the reference distribution for a particular respondent’s rank. Comparability asks whether measured constructs or scores can be meaningfully related across groups under an analysis. A percentile requires an identified score distribution and a conversion applicable to the respondent’s form and comparison group. The first kind of evidence may inform the design of a careful comparative study; it does not automatically provide the second.

A reader can therefore take the OECD findings seriously without stretching them. They show that researchers have investigated short-form BFI-2 measurement across a large, multilingual adult survey and that the resulting evidence supports some bounded comparisons while retaining item-level and model-related cautions. They do not establish that every national sample is interchangeable, that one country’s results represent adults generally, or that a full-form report’s percentile is official. To use the report as support for an individual rank, a provider would still need to identify the exact form, relevant reference data, and a documented scoring-to-rank procedure. Those requirements follow from the gap between the OECD study’s comparative purpose and the report’s individual-percentile claim; they are not additional findings attributed to OECD.

A large survey may be designed to study measurement across participating settings, not to establish a nationally representative norm for every reader or form. Sample size can improve precision within that design, but cannot change its target population or make a comparison study a general-purpose individual conversion resource. OECD’s findings concern short-form deployment and cross-country psychometric analysis in the Survey of Adult Skills. They support bounded conclusions about that work, not a separate individual-norm claim.

Sources: Psychometric properties of the BFI-2 inventory in the 2023 Survey of Adult Skills

What does an individual percentile actually compare?

An individual percentile compares a score with a specified distribution of scores. The National Council on Measurement in Education’s “Assessment Glossary” defines percentile rank as the percentage of scores in a specified distribution that fall below a given score; its definition of norm-referenced interpretation likewise makes the reference population part of the comparison. So a percentile is not a property carried by a BFI-2 scale score on its own. To say “the 70th percentile” meaningfully, a report must identify the set of scores and the method that locates the person’s score within it. Without that reference, the number leaves its central question unanswered: seventieth among whom?

A report may show several different quantities. A raw or scale score results from scoring responses under a specified rule. A standardized score transforms a score, often relative to a mean and spread, and depends on a reference basis. A mean summarizes a group’s center; it does not show an individual’s rank. A qualitative band labels a range under a defined scheme, but does not by itself identify the represented group or how boundaries were set. These numbers are not interchangeable just because they appear together.

For illustration only, imagine a researcher calculating a person’s rank against scores from one named BFI-2 study sample. If the form, scoring, sample, distribution, and rank procedure were specified, the resulting percentile would describe that person’s position relative to that sample under that procedure. It would not, without further evidence, mean the same position among adults generally, people of the same age in a country, or applicants for a particular job. Those are different comparison populations. The useful reading is local to the comparison actually made: it tells the reader about standing in the stated distribution, not a universal location on a personality scale.

A sample mean cannot provide that rank by itself. Nor can a POMP value: a percentage-of-maximum-possible transformation rescales a score against the instrument’s possible score range; it does not report the percentage of people scoring below an individual. A mean and a score’s transformed position can be informative summaries, but many different score distributions can share the same mean or range while placing a given score at different ranks. Recovering a percentile therefore requires an appropriate distribution and a procedure for comparing the individual score with it, not arithmetic applied to a mean alone.

That dependence does not make every nonofficial percentile meaningless. If the distribution is real, relevant to the reader’s question, and named clearly enough to understand the comparison, a sample-specific rank can communicate relative standing within that group. The boundary is the claim attached to it: a result based on one study sample may be useful as a comparison with that sample, while a broader population claim needs evidence that supports that broader reference. NCME’s definition clarifies what the number means; it does not validate any particular BFI-2 dataset or make a limited sample representative by labeling its output a percentile.

Sources: Assessment Glossary — National Council on Measurement in Education

Why do the older BFI age curves not answer the percentile question?

The age curves linked from the Berkeley Personality Lab page come from the earlier Big Five Inventory, or BFI, rather than the later BFI-2. “Development of Personality in Early and Middle Adulthood: Set Like Plaster or Persistent Change?” (2003) analyzes responses collected through an Internet volunteer sample and examines cross-sectional age patterns. Berkeley describes the linked figures as American sample means shown by gender and age. That is potentially useful descriptive material about the earlier inventory and its respondents; the link does not turn the figures into a BFI-2 norm table or individual percentile conversion.

The study reports group averages after transforming scores to POMP, short for percentage of maximum possible. POMP rescales a score relative to the lowest and highest possible values on the measure, expressing where a score or mean sits on that possible range. It is not a percentile rank: it does not state what percentage of respondents scored below one person. A group’s average POMP value describes its average transformed score, not the distribution of individual scores around that average. Even when plotted across ages, a line of group means is still a summary of groups, not a table that assigns individual ranks.

The distinction matters because the visual form can invite a stronger reading than the statistic supports. A curve may rise or fall across age categories and thus convey a historical descriptive pattern in this volunteer sample. But it does not reveal, for a particular BFI-2 respondent, the proportion of a defined reference group scoring below them. To make that individual comparison, a report would need an appropriate distribution of individual scores and a conversion procedure for the score at issue. The older study’s means do not provide that procedure simply because they are organized by age or displayed as reference-like curves.

Recruitment and design also bound what the curves can represent. Internet volunteers are not automatically a representative sample of adults, and cross-sectional age differences compare people observed at different ages rather than following the same people as they age. Thus the paper can show the pattern it analyzed in its recruited sample, while population-wide age claims require additional support. These limits do not erase its contribution; they specify the population and design behind it. For the present percentile question, the decisive mismatch is more basic: an earlier instrument, volunteer-sample group means, and a POMP scale are not an individual BFI-2 reference distribution.

To use those curves carefully, a reader can describe them as age-patterned averages in the earlier BFI Internet volunteer sample, with the POMP transformation stated. Calling them “norms” without qualification would conceal both the recruitment boundary and the statistic being shown. Neither qualification is a technicality: changing the instrument changes the score being summarized, while changing from an average to a rank changes the question answered.

Sources: Development of Personality in Early and Middle Adulthood: Set Like Plaster or Persistent Change?

Can a third party calculate a bounded BFI-2 percentile?

Yes, in principle: a third party could calculate a percentile that is interpretable as a rank within its own clearly described BFI-2 comparison dataset. The National Council on Measurement in Education’s “Assessment Glossary” defines percentile rank relative to a specified score distribution. That definition does not make the dataset official or representative; it makes the distribution the reference for the claim. If the provider says a score is at the 70th percentile in its dataset, the defensible meaning is that it falls at that position under the provider’s stated scoring and conversion, among the specified scores—not automatically among adults generally.

Imagine, for illustration only, a provider assembling BFI-2 responses and publishing its own conversion. A reader would first need to know which form respondents completed and how responses became the reported scale score. Otherwise, scores from different forms or scoring rules might be mixed or compared as though they were the same quantity. The provider would also need to describe where and when the comparison records came from, what population they represent, and how records entered the dataset. These are questions derived from what a specified distribution means, not a universal checklist prescribed by NCME or evidence that any actual provider has done this work.

The distribution itself must be sufficiently visible to understand the rank. A provider might show score frequencies or another account of the distribution and explain the conversion from an individual score to its reported percentile, including how ties or score boundaries are handled if those affect the result. The point is traceability: a reader should be able to see what was compared and how the number followed from that comparison. A displayed percentile without its reference and conversion conceals the defining operation. Conversely, a documented calculation can be locally meaningful even when it is not a nationally representative norm; its meaning remains limited to the comparison it actually describes.

Suppose the illustrative provider recruited volunteers through its own website. It could report a percentile relative to those respondents if it labels the group and explains the recruitment basis. The same result would not, by assertion alone, become a rank among all working adults, adults in a particular country, or people of a given age. A convenience sample can describe itself; extending its reach requires evidence that its recruitment, composition, and any adjustments support the broader comparison. The provider’s choice of reference is part of the interpretation, rather than a footnote that a reader can safely ignore.

The distinction is between source status and claim reach. An independent dataset is not invalid simply because the instrument developer did not issue it. A bounded rank can answer a bounded question, such as how a score compares with members of a disclosed sample under a stated procedure. Whether it answers a reader’s broader question depends on the design and limits of that sample, the match between the administered form and the data, and the clarity of the conversion. Calling the number a percentile is not enough; identifying its reference makes the comparison interpretable, while careful labeling prevents a local result from implying a wider population than its evidence supports.

Sources: Assessment Glossary — National Council on Measurement in Education

What evidence would justify using that comparison for a decision?

A well-described percentile supports a decision only when evidence supports the interpretation and use attached to it. The joint AERA, APA, and NCME “Standards for Educational and Psychological Testing (2014)” treat validity as a matter of evidence and rationale for proposed interpretations and uses, with attention to the population and context. A calculation may accurately rank a person in a disclosed sample yet still be a poor basis for a conclusion about a different group or a consequential action. Clear arithmetic answers how the rank was produced; it does not alone answer whether that rank is relevant to the decision.

For private reflection, a person might use a bounded comparison as one prompt for considering how their responses relate to a reference group. In coaching, it might help frame a question to explore alongside examples from the person’s own work and experience. Even here, the report should not turn relative standing into a fixed identity or prediction: a percentile describes position in a distribution, not why the person answered as they did or what they will do in every situation. The Standards’ general focus on the proposed interpretation means the user should ask whether this reference group and score meaning fit the reflective question, rather than treating any number as self-explanatory.

The evidentiary demand changes when a report is used to select, promote, set compensation, or manage performance. Those decisions concern a role, setting, and consequences that a general personality rank does not establish on its own. A percentile cannot by itself show that a person fits a job, will perform well, or should receive a particular employment outcome. Applying the Standards, the organization would need evidence and a rationale connecting the score interpretation to the specific population, context, and proposed use; relevant information beyond the rank would also be needed to support an individual decision. This is an application of general testing standards, not a claim that the BFI-2 has been validated for any such employment use.

A practical rule is to make the conclusion no broader than the evidence and to demand more relevant support as consequences rise. Ask whether the reference population resembles the people to whom the decision applies, whether the score is precise enough for the distinction being made, and whether evidence connects this interpretation to this use. An unexplained rank should carry little weight in a high-consequence decision; a readable percentile is not a substitute for evidence about the decision’s actual requirements. These questions organize the review, but they are not a mechanical checklist that certifies a test or use as acceptable—the Standards caution that judgment depends on the full evidence and context.

National norms are not the only possible basis for a useful comparison. Local evidence may fit a narrowly defined reflective or organizational question better than a broad reference would, provided the group and inference are described honestly and the evidence supports the intended use. The controlling issue is not whether a comparison has national reach, but whether its population, precision, and evidentiary basis justify the conclusion someone proposes to draw from it. The higher the stakes, the less a rank alone can reasonably carry. For the BFI-2 specifically, the general Standards guide how to evaluate a proposed claim; they do not themselves supply BFI-2-specific validity evidence or certify a percentile for a particular decision.

Sources: Standards for Educational and Psychological Testing (2014)

How should the reader interpret a report with missing norm details?

The right conclusion depends on what the report actually documents. Missing detail makes the comparison uncertain; it does not by itself show that the person has a particular trait, or that the underlying responses have no reflective value. Treat these as three different reporting cases:

| What the report provides | What the percentile can mean | What to do with it | | --- | --- | --- | | A percentile, but no identifiable reference group or conversion | No interpretable rank: the reader cannot tell whose scores were compared or how the number was produced. | Do not use the percentile as evidence of standing. Ask whether technical documentation is available. | | A named but narrow group, such as respondents in a particular study or provider dataset | At most, a comparison with that described sample under its stated scoring method. It does not automatically rank the person among adults generally or another population. | Keep the claim sample-specific; decide whether that group is relevant to the question. | | A described distribution that matches the administered form and the intended comparison | A bounded rank within that distribution, to the extent its method and limits support it. | Read it only against the named population and for the stated purpose. |

This distinction keeps a weak percentile claim from becoming a verdict about the respondent. If the group or conversion is unknown, the uncertainty is about the comparison, not proof that a score is unusually high, low, accurate, or inaccurate as a description of the person. A scale score may still prompt private reflection when read with its item and scoring context; without a reference distribution, however, it cannot carry the additional claim that the person ranks at a particular percentile.

A label such as “above average” needs the same scrutiny when it is derived from a percentile or band. Without the comparison basis, the label does not tell the reader whether the report describes a sample position, a score threshold, or an informal category. Keep the respondent-level observation separate from that missing reference: a person may recognize a pattern in their answers, but the unsupported rank adds no reliable information about how common that pattern is in a population.

A report may leave technical details out for space while making them available in a manual, methods page, or provider response. That possibility is a reason to request the documentation before deciding the comparison is unsupported. If the details identify a real group and method, interpret the result at that scope. If they do not, the number remains unresolved as a rank, regardless of how precise its presentation looks.

What should I ask the provider next?

Send one focused request: “Which BFI-2 form did I complete, what reference distribution and population does this percentile use, where and when were those data collected, how were scores converted to percentiles, and what use is this comparison intended to support?” If the provider answers, read the number only against that named group and purpose. If it cannot identify the basis, set the percentile aside; a documented independent dataset could support a sample-specific rank without becoming official general-purpose BFI-2 norms. A useful reply should let you connect the reported score to that distribution rather than merely naming a database: check whether the respondent group matches the stated population and whether the conversion applies to the form you took. If the provider supplies only a broad label such as “validated norms” without those particulars, ask for the underlying technical note or manual. The scope of the eventual answer should stay no broader than the documentation.

If your separate need is low-stakes reflection on recurring collaboration patterns, the live [Work Pattern Report](/assessment) is a separate, unnormed exercise. It is not validated for employment decisions and makes no job recommendation. [Explore the report topics](/topics).

Questions readers ask