We think personality is far more permanent than it really is. A 2026 study comparing public beliefs with eight longitudinal datasets covering 166,971 people found that people underestimated average Big Five changes in 90% of year-to-year comparisons—and every comparison spanning ages 15 to 90.

Personality can be stable without being permanent. That distinction sits at the centre of a 2026 study comparing what Americans believed would change between ages 15 and 90 with modelled trajectories from eight long-running panel studies.

The peer-reviewed paper in Communications Psychology combined two studies. First, 887 US adults estimated how much the Big Five personality traits and 20 other characteristics change across the lifespan. The researchers then analysed longitudinal data from 166,971 people in Germany, Australia, the United States and the Netherlands.

One statistical clarification belongs near the top. The headline’s 90% figure is the median rate at which net Big Five change was underestimated across the paper’s different analysis variants. Cumulative change was underestimated in every comparison. Respondents did not make a new prediction for each transition from age 15 to 16, 16 to 17 and onward. They made one whole-lifespan judgement for each characteristic.

The result is therefore not that the public guessed 90% of individual birthdays incorrectly. It is that a persistent personality-specific bias survived different definitions of change, comparison methods and model checks.

The Big Five describe dimensions, not five kinds of person

The Big Five are extraversion, agreeableness, conscientiousness, neuroticism and openness to experience. Each is a continuous dimension. Someone can be more or less extraverted, for example, without belonging to a permanent extravert or introvert category.

That makes the framework different from personality systems that sort people into a small set of types. Silicon Canals has previously examined why a four-letter Myers-Briggs result can look more fixed than the scores beneath it. Cross a category boundary by a small amount and the label changes, even if the underlying movement was modest.

The Big Five avoid that particular problem, but people may still treat the dimensions as parts of a permanent inner essence. “I am not organised.” “She is just an anxious person.” “He has never been open to new things.” A trait description quietly becomes a forecast.

What the public survey actually asked

Amanda Wright of the University of Zurich and four colleagues recruited English-speaking adults through Prolific. The sample was designed to represent the US population by sex, age and race. After 116 people failed an attention check and one gave the same rating throughout, 887 remained. Their average age was 45.36.

Participants were told to consider how much characteristics change in the population, on average, from ages 15 to 90. They used a scale from zero, meaning no change, to 11, meaning complete change.

The list included the Big Five and 20 other variables spanning identity, health, relationships, work, money, leisure and social life. Among them were self-esteem, life satisfaction, optimism, loneliness, physical health, sleep, exercise, personal income, relationship satisfaction, socialising and volunteering.

Extraversion received the lowest average change rating among the Big Five at 4.58. Conscientiousness received the highest at 5.87. The five-trait average was 5.37, and the Big Five occupied five of the seven positions the public treated as most stable.

Respondents were not saying personality never moves. Their judgement was comparative: personality changes less than almost everything else.

Eight panels supplied the empirical comparison

Study 2 used the German Socio-Economic Panel, Australia’s HILDA study, the US Health and Retirement Study, the Dutch LISS panel, Midlife in the United States, the US National Longitudinal Survey of Youth, the German Family Panel and the Wisconsin Longitudinal Study.

Together, they contributed 166,971 people with at least one wave of Big Five data. The typical participant had about 9.5 waves, and some people had as many as 40. Data collection across the eight projects stretched from 1984 to 2024. The combined age range ran from 14 to 111.

That large number needs a boundary. The study did not follow 166,971 individuals continuously from adolescence to age 90. The panels covered overlapping portions of life, used different schedules and measured some variables in different years.

The researchers used hierarchical age-period-cohort models to estimate average age trajectories while accounting for survey year and birth cohort. They then combined estimates across datasets. Because the public survey had specified ages 15 to 90, that was the range used for the main belief-versus-evidence comparison.

Outcomes were standardised, allowing movement in different units to be compared on the same effect-size scale. That does not make income, sleep and personality conceptually identical. It makes the size of their average changes comparable.

Net movement and total movement are not the same

The paper calculated two forms of change. Net change compared a variable’s modelled average at age 90 with its average at 15. Cumulative change added the absolute size of every year-to-year movement along the modelled trajectory.

The difference matters. Suppose life satisfaction falls through midlife and rises later. Its starting and ending levels could be similar, producing little net change, while the path between them contains substantial movement. Cumulative change retains that movement instead of allowing one direction to cancel the other.

From 15 to 90, extraversion fell by 0.784 standard deviations in the modelled population average. Openness fell by 0.540 and neuroticism by 0.400. Conscientiousness rose by 0.484, while agreeableness rose by 0.221.

These are differences between modelled age averages, not a promise about an individual. A particular person may move more, less or in the opposite direction.

The cumulative results made personality look even less static. Conscientiousness accumulated 1.91 standard deviations of modelled movement across the 75-year path. Among 32 standardised variables, only socialising frequency, personal income, health limitations and nightly sleep moved more. Conscientiousness moved more than self-esteem, loneliness and life satisfaction.

Across the five traits, average annual cumulative change ranged from 0.017 standard deviations for extraversion to 0.026 for conscientiousness. Those increments are small. Their relevance comes from repetition across decades, not from a dramatic annual reinvention.

The 90% figure comes from robustness comparisons

Wright and her colleagues did not rely on one calculation. They compared perceived and empirical change using standardised magnitudes and rankings. They examined net and cumulative change. They ran estimates with and without corrections for imperfect measurement reliability, then repeated key analyses using only the four US datasets to make the empirical sample more comparable with the American belief survey.

Across those variants, the median rate of underestimating Big Five net change was 90%. For cumulative change it was 100%. In the complete four-country analysis, every Big Five trait was underestimated regardless of change metric or comparison method. In the US-only checks, net underestimation occurred for 60% to 80% of the traits, while cumulative underestimation remained universal.

This is the precise origin of the headline number. It is a summary across model and comparison variants, not the proportion of survey respondents who gave a wrong answer and not the proportion of individual annual transitions they failed to predict.

The public was not generally clueless. Perceived change correlated moderately with empirical change across the full set of variables. For standardised values, the correlation was 0.59 with net change and 0.55 with cumulative change. People had a rough sense of what tends to move more and less.

Personality was the exception. The five traits were repeatedly judged too stable, while several identity-related variables, including life satisfaction and self-esteem, were more often judged to change more than the longitudinal estimates suggested.

Average change can coexist with a recognisable person

The study measured mean-level change. It did not ask whether people preserve their rank relative to peers. Those are different forms of stability.

Imagine an entire school year becoming more conscientious over several decades. The student who began as the most organised could remain near the top even while everybody’s average rose. Relative order can be stable at the same time as the population mean changes.

Nor does mean-level change mean a person becomes unrecognisable. Extraversion declining by a fraction of a standard deviation across decades is not the same as every highly social teenager becoming solitary at 90. It means the average expression of the trait differs across age.

This distinction fits an earlier Silicon Canals report on a 50-year American study of personality maturation. The same people retained meaningful continuity from 16 to 66, yet became more agreeable, conscientious and emotionally stable on average. Continuity and development were present together.

What the paper cannot turn into a promise

All eight panels came from Western democratic countries, and the public-belief sample came only from the United States. Ideas about a “true self” are shaped partly by culture, language and social expectations. The result should not be treated as a universal measure of what every society believes.

Most measures were self-reported. The authors adjusted multi-item scales for reliability and found similar relative rankings without those corrections, but comparable corrections were not available for every single-item measure.

The age-period-cohort method is also a model, not a 75-year film of one nationally representative group. It helps separate average ageing patterns from historical periods and generations, but it cannot remove every difference among datasets or every assumption behind the fitted trajectories.

Most importantly, the study describes population averages. It does not show that any chosen trait can be changed quickly, that all life events reshape personality, or that an intervention will reliably produce a preferred self. It did not randomly assign anyone to become more conscientious or less neurotic.

The study’s preregistrations, public survey data, analysis code and supplementary results make the analytical choices unusually inspectable. That transparency strengthens scrutiny; it does not remove the boundaries of an observational design.

A trait is a tendency, not a sentence

The fixed-personality belief survives partly because traits feel like explanations. If someone misses deadlines because they are “just disorganised”, the description appears to end the inquiry. Once the trait becomes an essence, change starts to sound false or performative.

Longitudinal evidence supports a more restrained position. Personality has enough continuity to be meaningful and enough movement to keep developing. The relative balance differs by trait, life stage, person and environment.

People readily accept that work, income, health, relationships and satisfaction can look different at 60 than at 20. The harder idea is that the person moving through those changes is not standing still either.