Do your core values stay the same? The stability of personal values is hard to observe directly, so this study looks at something narrower: how much a ranking repeats. We looked at 1,256 people who took the personal values assessment more than once, to see how much of their ranking repeated.
By Arkadiusz Zalewski, founder of personalvalu.esData updated
This study measures how much people repeat their choices and rankings when they take the Personalvalu.es assessment again. It does not establish whether their underlying values changed.
Some figures are set against two different people: one person's first list compared with the second list of another person in the same group, paired at random 200 times and averaged. The pairing is random and says nothing about whether the people know each other; the study does not look at relationships between participants.
Chapter 02, Key findings
People came back on their own, anywhere from minutes to more than three years later, rather than on a set date. This shows how rankings repeated among people who chose to return to the assessment. It also means each time group is made up of different people.
51.7% of the values in a first Top 10 appeared in the second, among the 516 people who returned at least a day later. That is about 5 of 10.
The lists of two different people overlapped by 22.7%, partly because some values are popular with many people.
See how much appeared again →30.6% of the 516 people who returned at least a day later put the same value first again. Between the lists of two different people it is 4.9%.
The other two thirds put a different value first.
See it by time apart →31.5% of people who came back within an hour chose the same number one. Among people who came back after at least 365 days it was 31.3%.
These are different groups of people who returned after different intervals, not the same people measured twice. Leaving out everyone who came back within 30 days, in case quick retakes were people trying out the tool, gives 32.6%. This pattern is consistent with ranking variation between sessions, but this study cannot determine why a person's answer changed.
See it for every wait →0.40 is the Spearman rank correlation, where 1 is the same order and 0 no link. Kendall's is 0.34.
Both are calculated only for the 459 of 516 people with at least three values in common. People with fewer tended to have less stable lists, so these figures may overstate how similar repeat rankings were. Between the lists of two different people they are 0.09 and 0.08.
See what these two measures mean →73% of number ones appeared somewhere in the second Top 10, while 66% of the values in tenth place did not.
The lower a value sat on the first list, the more likely it was to drop out. A value near the cut-off needs only a small change to fall out, so this does not show that lower places matter less.
See where values ended up →49.9% of a Top 10 appeared again among people who returned after more than 180 days, against 51.6% at 1-7 days and 56.9% within an hour.
These are comparisons between different groups of people who returned after different intervals, not the same people measured again. The study did not assign people to a retest schedule, so the results do not show what would happen to the same person over time. The groups also differ in size, from 45 to 508 people, and some are small.
See every time gap →Chapter 03, The Top 10
People's repeat rankings shared more values than rankings from two different people did.
Among the 516 people who took the assessment again at least a day later, 51.7% of the values in their first Top 10 appeared in their second. For comparison, the first and second rankings of two different people shared 22.7%.
How it is measured. Some values are popular, so any two people will have a few in common. For the comparison, each person's first ranking is set against the second ranking of someone else in the same group, paired at random 200 times and averaged. Keeping all ten was rare: only 3 of 516 people did. The second chart shows the 740 people who came back within 24 hours.
516 people, 52% on average
Values in common, out of 10
740 people, 55% on average
Values in common, out of 10
Same person, % of peopleTwo different people
Chapter 04, The order
The higher a value sat on the first list, the more likely it was to appear again.
On average, 5 of the 10 values are swapped for new ones. The ones that stay usually shift by about 2 places, and 18% stay exactly where they were.
Read across a row. Of the values people put first, 31% were first again, 13% came second, and 27% left the Top 10 altogether. The darker the square, the more common the move.
| Place | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | Gone |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 31 | 13 | 8 | 5 | 5 | 4 | 3 | 2 | 2 | 1 | 27 |
| 2 | 12 | 17 | 8 | 8 | 5 | 4 | 3 | 2 | 4 | 2 | 35 |
| 3 | 9 | 11 | 8 | 8 | 7 | 7 | 4 | 3 | 3 | 3 | 38 |
| 4 | 5 | 7 | 10 | 7 | 9 | 7 | 5 | 2 | 3 | 4 | 42 |
| 5 | 4 | 5 | 7 | 8 | 5 | 5 | 6 | 6 | 4 | 3 | 46 |
| 6 | 4 | 3 | 5 | 7 | 5 | 4 | 6 | 5 | 4 | 3 | 54 |
| 7 | 2 | 3 | 5 | 5 | 6 | 7 | 5 | 6 | 5 | 3 | 52 |
| 8 | 1 | 3 | 3 | 3 | 3 | 6 | 4 | 5 | 4 | 4 | 62 |
| 9 | 2 | 3 | 3 | 4 | 3 | 5 | 5 | 5 | 6 | 5 | 60 |
| 10 | 2 | 2 | 2 | 3 | 3 | 3 | 4 | 4 | 6 | 5 | 66 |
No value tends to gain or lose ground on the second try. Rankings changed in many directions, not toward any particular value.
Spearman and Kendall are two rank correlation coefficients. Here they measure whether the values that appear in both rankings come in the same order. Both run from 1, the same order, through 0, no link, to −1, the exact reverse. A value outside the Top 10 has no known place, so only values in both rankings are used.
Kendall usually reads lower than Spearman, so the two figures are not compared with each other. These correlations are calculated only for people with at least three values in common across their two Top 10s, 459 of 516. People with fewer shared values are excluded, and they tended to have less stable lists. The correlations may therefore overstate how similar repeat rankings were across all returners. The correlation also rises with the number of shared values, as the table shows.
| Shared values | Pairs | Spearman | Kendall |
|---|---|---|---|
| 3 | 53 | 0.23 | 0.21 |
| 4 | 70 | 0.31 | 0.27 |
| 5 | 101 | 0.35 | 0.30 |
| 6 | 86 | 0.49 | 0.41 |
| 7 | 87 | 0.48 | 0.39 |
| 8 | 46 | 0.46 | 0.37 |
| 9 | 13 | 0.66 | 0.52 |
| 10 | 3 | 0.65 | 0.50 |
Chapter 05, Value by value
Some values appeared again more often than their popularity alone would explain.
A value that stays is not always a sign of something deep. Health came back for 71% of the people who had it, but it is so popular that 26% of those who did not have it picked it the second time. We call that second figure the new-pick rate. So each value also gets a stickiness score: Cohen's kappa, which compares how often people's two lists agreed on a value, in or out, with how often they would agree by chance given how often it is picked. It measures repeat agreement in this sample, not how deep, important or lasting a value is. There are hints that Love and Financial stability scored above average, and Trust and Contribution less, but the data is not firm enough to be sure. Most other values are too close together to separate.
0.68 against an average of 0.37 across the 54 values shown in the table. A score of 0 means people's two lists agreed on the value no more often than chance would predict, and 1 means they agreed completely. Religion had the highest observed score in this sample. Only 38 people had ranked it in their first Top 10, so this estimate is uncertain. Religion and Family were the only values whose scores differed by more than chance would explain from the average of all 62 values, 0.35, in a check that allows for testing many values at once. Both were above it. That average is a little lower than 0.37 because it also counts the 8 values too few people had to show in the table. The data should not be read as establishing that either value is inherently more stable.
| Value | People | Came back | New pick | Kappa, likely range | Cohen's kappa |
|---|---|---|---|---|---|
| Religion | 38 | 66% | 2% | 0.68 | |
| Intelligence | 64 | 55% | 6% | 0.51 | |
| Job security | 60 | 55% | 6% | 0.48 | |
| Family | 171 | 64% | 17% | 0.48 | |
| Love | 137 | 69% | 19% | 0.47 | |
| Financial stability | 190 | 67% | 20% | 0.47 | |
| Variety | 44 | 50% | 4% | 0.46 | |
| Health | 264 | 71% | 26% | 0.45 | |
| Peace | 188 | 63% | 20% | 0.44 | |
| Authenticity | 172 | 63% | 20% | 0.43 | |
| Compassion | 101 | 54% | 11% | 0.43 | |
| Meaningful work | 116 | 59% | 14% | 0.43 | |
| Wisdom | 123 | 52% | 11% | 0.43 | |
| Curiosity | 163 | 60% | 18% | 0.42 | |
| Creativity | 134 | 57% | 15% | 0.42 | |
| Certainty | 99 | 55% | 12% | 0.42 | |
| Inner harmony | 132 | 56% | 14% | 0.42 | |
| Equality | 53 | 53% | 8% | 0.42 | |
| Courage | 65 | 45% | 6% | 0.42 | |
| Pleasure | 144 | 56% | 15% | 0.41 | |
| Adventure | 57 | 47% | 6% | 0.41 | |
| Spirituality | 67 | 55% | 10% | 0.41 | |
| Wealth | 66 | 52% | 9% | 0.40 | |
| Discipline | 84 | 52% | 12% | 0.39 | |
| Determination | 48 | 42% | 5% | 0.39 | |
| Honesty | 82 | 54% | 13% | 0.38 | |
| Control | 76 | 45% | 9% | 0.37 | |
| Beauty | 28 | 46% | 5% | 0.37 | |
| Growth | 206 | 60% | 24% | 0.37 | |
| Security | 118 | 49% | 14% | 0.36 | |
| Success | 69 | 45% | 9% | 0.36 | |
| Independence | 91 | 44% | 10% | 0.36 | |
| Ethics | 90 | 49% | 12% | 0.36 | |
| Freedom | 167 | 53% | 18% | 0.36 | |
| Excitement | 41 | 34% | 4% | 0.34 | |
| Friendship | 74 | 43% | 10% | 0.34 | |
| Teamwork | 43 | 37% | 5% | 0.33 | |
| Respect | 91 | 47% | 13% | 0.33 | |
| Challenge | 52 | 42% | 8% | 0.33 | |
| Loyalty | 69 | 42% | 9% | 0.33 | |
| Uniqueness | 31 | 45% | 6% | 0.32 | |
| Forgiveness | 34 | 41% | 6% | 0.32 | |
| Competence | 47 | 36% | 6% | 0.31 | |
| Justice | 67 | 40% | 9% | 0.31 | |
| Acceptance | 50 | 32% | 5% | 0.31 | |
| Ambition | 43 | 33% | 5% | 0.30 | |
| Helpfulness | 82 | 40% | 11% | 0.29 | |
| Passion | 65 | 45% | 12% | 0.28 | |
| Accountability | 97 | 38% | 11% | 0.28 | |
| Privacy | 43 | 35% | 7% | 0.27 | |
| Tolerance | 79 | 35% | 10% | 0.26 | |
| Trust | 148 | 45% | 23% | 0.23 | |
| Appreciation | 74 | 34% | 12% | 0.22 | |
| Contribution | 52 | 23% | 7% | 0.17 | |
| Authority | 13 | Too few people to show | |||
| Commitment | 46 | Too few people to show | |||
| Excellence | 34 | Too few people to show | |||
| Fame | 10 | Too few people to show | |||
| Influence | 24 | Too few people to show | |||
| Popularity | 9 | Too few people to show | |||
| Reputation | 19 | Too few people to show | |||
| Tradition | 16 | Too few people to show |
"People" had the value in their first Top 10. "Came back" is the share of them who had it in their second. "New pick" is the share of people without it in their first Top 10 who had it in their second. Cohen's kappa is 0 when people's two lists agree on a value no more often than chance would predict from how often it is picked, and 1 when they agree completely: everyone who had it kept it, and no one else picked it. It can fall below 0. The dot is the kappa, the line its likely range, and the pale tick the average. A hollow dot marks a value fewer than 30 people had, so treat its score with extra caution.
Chapter 06, Over time
Top 10 overlap was similar across several observed time-gap groups.
These are comparisons between different groups of people who returned after different intervals. The study did not assign people to a retest schedule, so the results do not show what would happen to the same person over time. The groups differ in size, from 45 to 508 people, and some are small. People who came back within an hour shared the most of their Top 10, and may still have remembered their answers. Their number one was no steadier. After that, Top 10 overlap stayed nearly flat. The lists of two different people in the longer-gap groups also overlapped a little more, so the lead over them narrows from 30 percentage points at 1-7 days to 25 after more than 180 days. People who return after a long time may differ from those who do not, and they first saw the cards in the old fixed order.
Separate groups: each person is in one row.
Same person, with likely rangeTwo different people
Separate groups: each person is in one row.
Same person, with likely rangeTwo different people
Someone who ranks their values twice within minutes may just be trying out how it works. If that were most of what happens, people who wait longer should have kept their number one more often. They did not. Leaving out everyone who came back within a month leaves 258 people, and 32.6% of them chose the same number one.
Added after the analysis plan was fixed, to check this concern. Unlike the charts above, these groups overlap: "Under 1 hour" includes everyone under 10 minutes, and "At least 30 days" includes everyone at 90, 180 and 365 days.
| Time apart | People | Same number one | Likely range | Two different people | Still in Top 3 | Left the Top 10 |
|---|---|---|---|---|---|---|
| Under 10 minutes | 239 | 29.3% | 23.9 to 35.4 | 4.1% | 48.1% | 31.0% |
| Under 1 hour | 508 | 31.5% | 27.6 to 35.7 | 3.5% | 49.6% | 29.1% |
| At least 1 day | 516 | 30.6% | 26.8 to 34.7 | 4.8% | 51.0% | 26.7% |
| At least 30 days | 258 | 32.6% | 27.1 to 38.5 | 5.6% | 52.3% | 24.0% |
| At least 90 days | 207 | 30.9% | 25.0 to 37.5 | 6.0% | 51.2% | 25.6% |
| At least 180 days | 162 | 31.5% | 24.8 to 39.0 | 6.4% | 53.1% | 25.9% |
| At least 365 days | 96 | 31.3% | 22.9 to 41.1 | 7.2% | 51.0% | 27.1% |
Each person is in one row, except "0-7 days, combined", which adds up the three rows above it. Top 10, Top 5 and Top 1 are the share of values or number ones that appeared again. Spearman and Kendall run from 1 to −1 and only cover people with at least three values in common, so they may overstate how similar rankings were.
| Time apart | Pairs | Top 10 | Top 5 | Spearman | Kendall | Top 1 |
|---|---|---|---|---|---|---|
| Under 1 hour | 508 | 56.9% | 47.1% | 0.45 | 0.39 | 31.5% |
| 1-24 hours | 232 | 51.5% | 41.8% | 0.42 | 0.36 | 35.3% |
| 1-7 days | 196 | 51.6% | 42.5% | 0.40 | 0.34 | 28.6% |
| 0-7 days, combined | 936 | 54.5% | 44.9% | 0.43 | 0.37 | 31.8% |
| 8-30 days | 62 | 52.3% | 46.5% | 0.49 | 0.42 | 29.0% |
| 31-90 days | 51 | 55.3% | 43.5% | 0.30 | 0.23 | 39.2% |
| 91-180 days | 45 | 53.6% | 47.6% | 0.33 | 0.26 | 28.9% |
| 181+ days | 162 | 49.9% | 43.8% | 0.42 | 0.35 | 31.5% |
Chapter 07, How it was checked
From more than 2,100,000 results, we kept the pairs most likely to belong to one person.
The main plan came first. The primary rules and measures were documented before the main results were calculated. Some additional comparisons were added later and are identified in the technical details.
We compared against two different people. Every main figure sits beside the same figure for two different people, so popular values do not get mistaken for consistency.
We removed what did not belong. Results saved from clients, shared results, people clicking the same side every time, test accounts and accounts whose answers pointed to two different people, all without looking at how similar the rankings were. 1,256 people remain.
We tried it other ways. Under seven different versions of the rules, the share of a Top 10 that comes back stays between 50.9% and 52.1%.
Chapter 08, How to read this
Four things worth keeping in mind.
Chapter 09, Questions
From a snapshot of 2,143,199 results taken on 30 October 2026, one pair per account: its two earliest results. Of 1,368 accounts with two or more results, we set aside those without two results of their own (results held by several accounts, or from the account's own invitations, 72), unlimited-invitation accounts (13), rankings made by clicking one side of every comparison (14), pairs with conflicting gender answers (9), and a few maintainer, test or very heavy accounts. A shared account we could not detect would lower similarity.
Overlap is the number of shared values divided by 10 or 5. Spearman and Kendall re-rank the shared values, and exist only where at least three are shared, because the 52 values outside a Top 10 have no known rank. The pairs left out are the least stable, so these correlations read high for everyone. Rank-biased overlap (p = 0.9) weights the top and covers every pair. "Two different people" is each person's first ranking against someone else's second, over 200 random re-pairings. Intervals are 95%, from a bootstrap over people, or Wilson for proportions. The stickiness score is Cohen's kappa.
Three sets appear on this page: the 516 people who came back at least a day later, the main group; the 740 who came back within 24 hours; and all 1,256 together.
Fixed in the analysis plan before the main results were calculated: how pairs were chosen and excluded, the main group of at least one day apart, the measures on this page, the time groups and the comparison with two different people. The one-day line was chosen after seeing how long people waited, but before any similarity was calculated. Added later: comparing the lead over two different people across time gaps, and number one by how long people waited at least. The stickiness check also used 10,000 resamples rather than the planned 2,000, and was repeated with six random seeds, because with 2,000 the list of values that differ from the average changed from run to run.
| Measure | Same person | 95% interval | Two different people |
|---|---|---|---|
| A day or more apart, 516 pairs | |||
| Top 10 overlap | 51.7% | 49.9 to 53.4 | 22.7% |
| Top 5 overlap | 44.0% | 41.7 to 46.1 | 14.4% |
| Top 1 consistency | 30.6% | 26.8 to 34.7 | 4.9% |
| Rank-biased overlap | 0.45 | 0.43 to 0.47 | 0.16 |
| Spearman | 0.40 | 0.36 to 0.45 | 0.09 |
| Kendall | 0.34 | 0.30 to 0.37 | 0.08 |
| Under 24 hours apart, 740 pairs | |||
| Top 10 overlap | 55.2% | 53.4 to 57.0 | 21.4% |
| Top 5 overlap | 45.5% | 43.4 to 47.4 | 13.3% |
| Top 1 consistency | 32.7% | 29.4 to 36.2 | 4.0% |
| Rank-biased overlap | 0.48 | 0.46 to 0.49 | 0.15 |
| Spearman | 0.44 | 0.40 to 0.48 | 0.07 |
| Kendall | 0.38 | 0.35 to 0.41 | 0.06 |
| All pairs, 1,256 pairs | |||
| Top 10 overlap | 53.8% | 52.5 to 55.0 | 21.8% |
| Top 5 overlap | 44.8% | 43.4 to 46.3 | 13.6% |
| Top 1 consistency | 31.9% | 29.3 to 34.5 | 4.3% |
| Rank-biased overlap | 0.47 | 0.45 to 0.48 | 0.16 |
| Spearman | 0.43 | 0.40 to 0.46 | 0.08 |
| Kendall | 0.36 | 0.34 to 0.39 | 0.07 |
| Rule | Pairs | Top 10 | Top 5 | Top 1 | Spearman |
|---|---|---|---|---|---|
| At least 1 day apart, first and second result | 516 | 51.7% | 44.0% | 30.6% | 0.40 |
| First and last result | 602 | 50.9% | 42.0% | 30.1% | 0.39 |
| Without accounts that sent invitations | 479 | 52.1% | 44.3% | 31.5% | 0.40 |
| Without pairs that include an invited result | 512 | 51.8% | 44.1% | 30.7% | 0.40 |
| Without second results from 14 October 2026 | 506 | 51.7% | 44.1% | 30.8% | 0.40 |
| Without the gender check | 521 | 51.5% | 43.7% | 30.3% | 0.40 |
| Without the one-sided check | 517 | 51.6% | 43.9% | 30.6% | 0.40 |
What psychology says about values changing over a life, and what tends to move them, is in Can my core values change over time?. Read how the wider research is built in the methodology.