Does Social Desirability Invalidate a Personality Assessment?

Does socially desirable responding automatically invalidate a personality assessment?

No. Socially desirable responding can weaken a particular score’s interpretation, but a favorable answer pattern or high indicator score alone does not prove someone lied or make every result unusable. Ask: what does the report claim, and is that interpretation supported for this assessment’s purpose?

A person may answer favorably to manage an impression, because they genuinely see themselves that way, or because the questions and setting invite it. A response-style indicator raises a question; it does not reveal motive. Treating it as proof of dishonesty turns an uncertain signal into a judgment about character.

The Standards for Educational and Psychological Testing frame validity as evidence supporting an interpretation and use, not a permanent label attached to a test. A warning may qualify one statement without erasing all information in the assessment. Self-reflection and consequential decisions also require different evidence, so a score should not automatically carry equal weight in both settings.

Do not decide that the whole report is trustworthy or worthless. Identify the statement you might act on, the intended use, and what the instrument’s evidence says about that interpretation. A flag deserves attention, but its meaning depends on the specific measure and use.

Sources: Standards for Educational and Psychological Testing

What does “invalid” mean for a report?

A score is not simply valid or invalid in the abstract. The useful question is whether evidence supports a particular interpretation for a defined purpose and population. A concern about socially desirable responding can qualify one inference without erasing every result in a report. A flag alone does not show that the person answered dishonestly or that the whole report is unusable. Socially desirable responding means answering in a way that presents oneself as socially approved, whether deliberate or not. It is not identical to deliberate faking. Answers may reflect impression management or an overly favorable self-view, without a clear plan to deceive. Careless responding is different: it concerns insufficient attention to questions, rather than the direction in which answers make someone look. An indicator can raise a question about how answers were produced, but it does not reveal motive by itself. The *Standards for Educational and Psychological Testing*, published jointly by the American Educational Research Association, American Psychological Association, and National Council on Measurement in Education, frame validity as evidence supporting interpretations and uses of scores. This shifts attention from the test as a whole to the claim being made with it. A report might describe a tendency, compare a score with a reference group, or support a consequential decision. Each interpretation needs evidence suited to that purpose and population. Evidence for one use does not automatically establish another. The Standards also place responsibility on users who extend an interpretation beyond the use supported by the test's evidence. This is separate from reliability, which concerns the consistency or precision of scores under specified conditions. A score can be consistent yet fail to support a particular interpretation; a response-style concern does not by itself establish that all scales are unreliable. They answer different questions: reliability concerns score consistency; validity concerns which conclusions scores support, and for whom. The same caution applies to a flag, cutoff, band, or correction. A number becomes meaningful through the instrument's documented scoring and interpretation rules, including evidence for the intended population and use. A threshold from one assessment cannot be carried over as a universal line separating truthful from deceptive respondents. Nor does “invalid” tell a reader which result is affected unless the report explains the rule and its scope. The 2014 Standards provide a general framework for evaluating claims, not a social-desirability cutoff or universal remedy. Suppose a report describes someone as exceptionally dependable because they endorsed consistently favorable statements. If a response-style indicator is elevated, the reader can pause over that description: does the manual say the indicator changes its interpretation, and is that rule supported for this assessment and setting? The answer might qualify the claim while leaving other, differently supported observations available for reflection. This is a way to examine a report statement, not evidence that a respondent answered falsely. A warning is best read as a boundary on what can safely be concluded until its meaning is established, not as a character judgment or a global verdict.

Sources: Standards for Educational and Psychological Testing

Why can a validity result look intact while a score is distorted?

The apparent contradiction disappears once two questions are separated. Criterion prediction asks whether people’s scores, across a group, are associated with an outside outcome or rating. Individual interpretation asks whether one person’s score represents the trait it is supposed to describe. A measure can retain a similar group-level association while some individual scores are distorted. The first result does not certify every profile; the second possibility does not show that every profile is unusable. A study can test a particular inference, such as whether self-ratings correspond with informant ratings, or whether scores relate to an outcome. It cannot automatically settle every other inference made from the same score. Here, the studies differ in what they measured and what their designs can support. In “Socially desirable responding in personality assessment: Still more substance than style,” researchers examined 222 college students who completed the NEO Five Factor Inventory and the Marlowe–Crowne Social Desirability scale. Informant ratings came from 128 randomly assigned roommates and 183 biological parents. Across 15 tests, higher Marlowe–Crowne scores did not predict lower agreement between a student’s self-rating and informant ratings. The scale also related positively to extraversion, agreeableness, and conscientiousness as rated by students and informants. In this sample, the indicator did not behave as a simple marker of inaccurate self-description. That finding is bounded. It concerns one social-desirability measure, a college sample, the NEO-FFI, and agreement with roommate or parent reports. Informant agreement is an external comparison, not direct access to a person’s “true” trait. Nor does a lack of moderation establish that every individual score was accurate. It says the study did not find the expected general pattern: people with higher indicator scores did not, on average, show lower self–informant agreement under those conditions. “Socially desirable responding in personality assessment: Not necessarily faking and not necessarily substance” broadened the test in another low-stakes student setting. It studied 602 undergraduates in 301 roommate pairs, using eight indices of socially desirable responding and a Big Five self-report measure. The indices did not show linear or quadratic moderation of the relationship between self-ratings and peer ratings. Some related to self-reported traits but were generally independent of peer-rated personality. The authors argued that socially desirable responding can reflect method variance without necessarily indicating faking or substantive personality. This weighs against a simple rule that a high social-desirability score means an assessment has become inaccurate. It does not show that response distortion never matters. The participants were undergraduates in a low-stakes setting, and the outcome was peer-report correspondence. The result does not establish what happens when applicants have a strong incentive to present themselves favorably, with a different instrument, or for a person whose profile raises concern. A statistical moderator test asks whether an association changes as an indicator varies across the sample; it is not a case-by-case audit. The other side of the puzzle appears in “Socially desirable responding and its elusive effects on the validity of personality assessments.” This Monte Carlo study created simulated trait scores, added specified amounts of desirability bias, and calculated score-to-criterion correlations. Those group-level predictive correlations could show relatively minor declines even when some simulated individuals’ obtained scores differed greatly from their underlying trait scores. Under the modeled conditions, individual interpretation could be badly compromised while aggregate tests gave little warning. This is not an observation that a known share of real respondents were distorted. It depends on assumptions about how bias entered scores and how traits related to criteria. Its value is methodological: it demonstrates why a modest change in a group correlation cannot rule out large errors for particular people. A correlation summarizes how scores and outcomes move together across cases; it does not reveal whether each person’s score is close to the intended trait. Put side by side, the findings answer different questions. The 2008 and 2010 studies tested whether measured desirability indices were associated with weaker average self–informant or self–peer correspondence in low-stakes student samples. They generally found no such moderation. The simulation asked whether aggregate criterion validity might remain relatively stable despite substantial individual score shifts. It showed that this pattern is possible under specified models. The findings are not mutually exclusive: group-level relationships can hold while some individual interpretations are wrong. For a reader, the practical consequence is restraint in both directions. A favorable response-style indicator alone does not justify rejecting an entire report or accusing its subject of deliberate deception. Yet intact average prediction does not establish that a flagged person’s profile is accurate. Ask which interpretation is being made, for whom, and for what purpose. The group result describes a defined sample; confidence in one person’s score requires evidence about that score’s interpretation in its instrument and context.

Sources: Social Desirability and Self- and Informant-Reported Personality; Socially desirable responding in personality assessment: Not necessarily faking and not necessarily substance; Socially desirable responding and its elusive effects on the validity of personality assessments

Can a social-desirability flag tell whether someone faked?

No. A social-desirability flag is a result produced by a particular measure; it does not reveal, by itself, why someone answered as they did. People may describe valued qualities, hold favorable self-views, respond to approved wording, or deliberately manage impressions. These can look similar in a score. Calling one respondent dishonest requires evidence about that respondent’s motive that a scale score alone does not provide. The distinction matters because “social desirability” names more than one possible process. In high-stakes settings, an incentive to present oneself favorably can make deliberate response distortion plausible. The review The Lazy or Dishonest Respondent: Detection and Prevention distinguishes such distortion from careless responding, which can occur when a person is not attending carefully to items. But a theoretical reason to suspect impression management is not proof that a given person engaged in it. Nor does a social-desirability indicator necessarily isolate that motive from ordinary self-description or the content of the traits being measured. The scale label is a hypothesis, not a direct reading of intent. Evidence on social-desirability scales illustrates this measurement problem. The meta-analysis Are Social Desirability Scales Desirable? examined 41 studies, with 8,980 participants, relating scale scores to prosocial behavior in economic games. If the scales were measuring bias toward socially approved self-presentation, the authors reasoned, higher scores should relate negatively to prosocial behavior; if they measured prosocial traits, a positive relation might be expected. Instead, the pooled association was close to zero. That result challenges a simple interpretation in this context. It does not show that all response-style measures are useless, that respondents never manage impressions, or that any flagged profile is invalid. Its criterion was economic-game behavior, not every trait or use. Other findings also resist treating social-desirability scores as a clean lie detector. In Social Desirability and Self- and Informant-Reported Personality, an NEO-FFI study of college participants found that Marlowe–Crowne social-desirability scores did not reduce agreement between self-ratings and ratings from informants across 15 moderator tests. Higher scores were associated with some traits as rated by the respondents and informants. This could mean the scale captures personality or self-view as well as distortion. Yet the study involved a college convenience sample, one inventory and one indicator. Self–informant agreement cannot prove every item was answered candidly or that an individual score is accurate. It complicates a universal claim without settling every case. Three questions should therefore stay separate. Detection asks whether a procedure can identify a response pattern associated with distortion under specified conditions. A response-validity rule asks whether the instrument’s manual says a protocol should be interpreted, qualified or treated as uninterpretable. Score correction asks whether changing a score after administration improves the intended interpretation. Success at one does not guarantee the others: evidence that a pattern can be detected does not show that a particular person had a deceptive motive, and a plausible correction is not justified unless it has evidence for that measure and use. The review surveys methods across settings, but cannot validate a specific test’s flag, threshold or correction. For a reader, the defensible wording is correspondingly narrow: “This result may need qualification under the instrument’s guidance.” Ask what the flag measures, which rule applies, and which statement it affects. Avoid translating it into “the person lied,” “the profile is fake,” or the opposite claim that no distortion occurred. The studies do not provide a universal cutoff that turns a score into proof of motive. Only instrument-specific evidence can support a stronger conclusion about how a flagged profile should be used; even then, the conclusion concerns interpretability under defined conditions, not a moral judgment about the respondent.

Sources: Are social desirability scales desirable? A meta-analytic test of their validity; Social Desirability and Self- and Informant-Reported Personality; The Lazy or Dishonest Respondent: Detection and Prevention

What changes when the assessment has consequences?

Higher stakes change the conditions under which people answer. If a personality assessment is part of an application, a result may affect access to a job, so there is a reason to present oneself favorably. That makes response distortion more plausible than in private reflection, but it does not establish what any one person intended. Nor does a shift in answers, by itself, show that every interpretation drawn from the assessment has failed. The 2021 meta-analysis “Faking by Actual Applicants on Personality Tests” compared the same individuals’ responses in high-stakes applicant settings and lower-stakes settings across 20 within-subject studies. On average, applicant responses had more favorable means, slightly less variability, and stronger rank-order consistency. The authors also found that order mattered: studies that introduced the high-stakes condition first showed a stronger mean shift and weaker rank-order consistency than studies that began with the low-stakes condition. These are comparisons of distributions and relative ordering across administrations. They establish that context can move scores on average; they do not identify which applicant deliberately altered an answer, how much a particular person’s underlying tendency changed, or whether a specific flagged profile is unusable. These group statistics are not a lie detector. A favorable response could reflect deliberate impression management, a sincere positive self-view, a different reading of an item, or the situation. The meta-analysis compares answers across conditions, not private motives. A higher average and tighter spread can coexist with relative ordering across respondents; neither statistic says how far an individual moved. The broader review “The Lazy or Dishonest Respondent: Detection and Prevention” explains why the setting is relevant: in high-stakes assessment, the consequences can motivate favorable self-presentation, whereas careless responding is a distinct concern often discussed in lower-stakes settings. The review frames plausible mechanisms but does not validate a particular test’s scale, define a universal cutoff, or establish that a flagged person was dishonest. An indicator needs evidence for its interpretation in that instrument and use. Private self-reflection and applicant selection also ask different questions. In private reflection, a person may use a broad tendency statement as a prompt: “I often prefer clear plans; when did that help, and when did it limit me?” If the answers were unusually favorable, the reader can hold the statement lightly and compare it with concrete examples from more than one setting. The purpose is to generate questions, not to certify a trait or decide what someone deserves. In selection, the decision can affect a person’s opportunity, so evidence must reach further. The *Standards for Educational and Psychological Testing* define validity in relation to the evidence supporting a particular score interpretation for a specified use. They call for the intended population, construct, context, administration and scoring to be made clear, and say that users must assess whether the available evidence fits their local situation. Thus, evidence that an instrument works for self-reflection, or that it predicts some outcome in another sample, does not automatically support using it to rank applicants for a different role. The relevant question is whether the proposed interpretation and decision have support for the instrument, population and context at hand. This also separates two concerns that can be confused. Favorable responding may affect fairness if a process rewards presentation in a way unrelated to the intended qualities; it may affect prediction if score meaning or outcome relationships change. A warning flag does not answer either question by itself. Seek documented evidence about the assessment and decision, including how the indicator was validated and what action its manual recommends. Without it, treat a flag as reason to narrow confidence, not proof of misconduct or a substitute for job-relevant evidence. The practical conclusion is proportional. High stakes raise a credible concern, and applicant research finds average shifts, but neither warrants a blanket verdict about every person or scale. As consequences grow, interpretation should match evidence from the relevant population and use. For reflection, hold a favorable profile as a hypothesis and compare it with behavior. For consequential decisions, require purpose-specific support. Context warrants caution, not accusation from a score.

Sources: Faking by Actual Applicants on Personality Tests: A Meta-Analysis of Within-Subjects Studies; The Lazy or Dishonest Respondent: Detection and Prevention; Standards for Educational and Psychological Testing

When should a reader retain, qualify, or set aside a result?

A reader can retain, qualify, or set aside a report statement. The choice should depend on what the statement says, what the assessment's own rules say about the response pattern, and what someone plans to do with the result. This is a reading rule, not a universal scoring formula. Retain a result as a tentative reflection prompt when the use is low stakes, the report describes a bounded tendency, and no instrument-specific warning changes that interpretation. “You may prefer to plan before acting” leaves room to compare the idea with experience. It does not say that a person always behaves this way, nor that a score proves the cause of a work problem. The Standards for Educational and Psychological Testing frame validity around evidence for a proposed interpretation and use. That principle supports asking whether this particular statement is warranted, rather than assigning a blanket verdict to every page of a report. Qualify a statement when unusually favorable self-description or the setting makes its meaning uncertain, especially if the statement sounds broader than the available evidence. Suppose, as an illustration rather than a real report, a result describes someone as consistently calm in conflict. A careful reader might restate it as: “This report suggests you may see yourself as steady in disagreements; compare that with specific recent situations.” Then compare the claim with specific recent exchanges. These observations do not validate a score, but keep reflection tied to behavior rather than a fixed identity. Qualification is also appropriate when someone wants to use a result for a consequential decision. A profile designed or supported for reflection does not thereby become evidence for hiring, promotion, or another high-impact judgment. The review “The Lazy or Dishonest Respondent: Detection and Prevention” distinguishes response problems and discusses how stakes can shape responding; it does not supply a universal rule for interpreting every instrument’s indicator. Ask what the measure’s documentation says the flag means, which population and purpose its interpretation covers, and what action the manual specifies. If that information is unavailable, say that the conclusion is uncertain rather than supplying a threshold from another test. Set aside a statement, or seek an appropriately qualified interpretation, when the instrument’s documented protocol says that the profile cannot support that use, or when the evidence does not support the claim being made. This is narrower than discarding an entire assessment because one response-style indicator is elevated. A report may contain separate scales and claims, and the evidence for one does not automatically determine the others. Conversely, retaining unaffected-looking sections without checking the protocol can also overreach. In practice, write down the exact sentence you are unsure about, the decision attached to it, and what the manual says about the response pattern. Keep ordinary tendency language as a question to examine; narrow claims whose meaning may have shifted; and pause where the documented rules or intended-use evidence do not support reliance. This keeps a possible prompt from becoming a diagnosis or verdict on honesty.

Sources: Standards for Educational and Psychological Testing; The Lazy or Dishonest Respondent: Detection and Prevention

What should the reader do with a flagged result?

Start with the instrument’s own explanation of the flag. Ask the provider what pattern it represents, whether it signals a concern about interpretation, and what action the manual recommends. “Social desirability” does not establish motive, deliberate deception, or the effect on a trait score. The Standards for Educational and Psychological Testing frame validity around evidence for a particular interpretation and use. So identify which report statement depends on the flagged responses instead of treating the whole profile as cleared or condemned. For self-reflection or coaching, mark affected wording as a question to examine. If a report describes someone as dependable, ask what that means in ordinary situations: meeting an agreed deadline, warning a colleague when plans change, or following through after a difficult week. These prompts do not prove a score right or wrong. Compare several relevant instances and allow for context, such as unclear priorities or limited resources. The aim is a specific conversation about behavior, not a character verdict. When an assessment could affect an important decision, ask for evidence supporting that interpretation in the intended population and setting, and how flagged profiles should be handled. If that information is unavailable, do not treat the result as decisive. Do not retake the assessment just to obtain a more favorable result, or infer dishonesty from the flag. For low-stakes exploration, the publication’s Work Pattern Report offers ten continuums; it has no norms, cutoff, type, or selection score and is not validated for hiring or career matching. Discuss one uncertain statement alongside concrete examples, and ask what evidence would justify relying on it.

Sources: Standards for Educational and Psychological Testing

The verdict: a warning can narrow a conclusion without erasing the report

Socially desirable responding can make an assessment less interpretable, especially when a person has reason to present favorably. It does not automatically invalidate the whole report. Ask whether evidence supports this score interpretation for this population and purpose. A warning may qualify one claim while leaving observations as hypotheses. That is a judgment about evidence, not honesty.

The findings can coexist. The 2008 and 2010 self–informant studies found no consistent reduction in agreement tied to social-desirability indices in student samples. The Monte Carlo study showed how modeled individual scores could depart from intended traits even while aggregate prediction changed little. It demonstrates a possibility under its assumptions, not its frequency in practice. The 2021 applicant meta-analysis found average shifts toward favorable responses in high-stakes settings; it did not show that every applicant distorted answers or every profile became unusable. A flag is not a lie detector or universal invalidation rule.

Change the conclusion when instrument-specific evidence does: when an indicator is validated for the relevant population and use, distortion is documented, or the proposed interpretation no longer holds. At low stakes, treat results as tentative tendencies and compare them with experience. As consequences rise, require stronger evidence for the intended use and a rule for warnings. Until then, narrow the claim rather than turn uncertainty into certainty or blanket rejection.

Sources: Standards for Educational and Psychological Testing; Socially desirable responding and its elusive effects on the validity of personality assessments; Faking by Actual Applicants on Personality Tests: A Meta-Analysis of Within-Subjects Studies

Questions readers ask