This is a Medium Rare template. Explore our complete range of designs here.
top
Analysing survey results by generation is common practice, yet the arguments against doing so are rarely acknowledged or understood.
The sociologist Karl Mannheim set out the modern concept in 1927, in an essay later translated as "The Problem of Generations." His account was careful, and conditional above all. Being born around the same time does not make a generation; it creates only the potential for one. Mannheim called that bare starting point a generational location — a shared position in historical time, and nothing more.
A generation, for Mannheim, required more than proximity in birth years. It required a shared historical experience, encountered at a formative stage of life, close enough to mark people in common. Without that shared experience, a birth cohort is just a set of people of similar age.
Even then, Mannheim expected no uniformity. People who live through the same events respond to them in opposite ways. The youth of post-war Germany were his case — some turning to the revolutionary left, others to the nationalist right, out of one shared upheaval. He saw a generation as divided from within by class and circumstance, not as a single personality type.
Two conditions in his account decide everything that follows. The events that form a generation have to land in youth, while attitudes are still setting. And being alive in the same year is not the same as living through it alike: what reaches a twenty-year-old and a fifty-year-old in the same moment is not one shared experience.
A generational cut of survey data meets none of Mannheim's conditions. It assumes a shared formative experience without checking for one, treats a fifteen- or twenty-year span as a single type, and ignores life stage, which for Mannheim was inseparable from the idea.
Mannheim's account asked a lot: establish the shared experience, allow for division within the cohort, keep life stage in view. The popular version kept the labels and dropped the conditions.
The popular scheme most people know comes from William Strauss and Neil Howe, whose books, from Generations in 1991 onward, fixed the generational types onto a recurring cycle and gave each one a personality. Mannheim's conditions fell away. A birth-year range was now enough to tell you what someone valued at work and what they wanted from an employer. The events said to have formed each generation were often picked out long after the fact; Strauss and Howe at one point offer the 1987 rescue of a toddler from a Texas well as a formative event for Millennials. They do not explain why.
Blaming marketers for this lets the research off too lightly. Ravid, Costanza and Romero's 2025 meta-analysis did something most do not: alongside testing whether the differences exist, it examined how the academic literature itself writes about generations, and found studies framing their results through generational stereotypes the data underneath did not support. The idea was being reinforced from inside the field meant to be testing it.
So the concept that reaches an HR team is not a simplified Mannheim. It keeps his labels and discards the conditions that gave them meaning.
If generational types were real in the way HR uses them, the differences would show up in the data. They do not. Costanza and colleagues pooled the available studies in 2012 and found the gaps between generations in job satisfaction, organisational commitment and intention to quit to be small and inconsistent — in most cases indistinguishable from zero. Where differences appeared at all, they tracked age and career stage rather than birth cohort.
The picture has not changed with better data. Ravid and colleagues returned to the question in 2025 with a larger meta-analysis and reached the same place: few systematic differences across the outcomes that matter at work, and none large enough to justify managing people by their birth years.
The boundaries themselves are not even agreed. The birth years that define each generation shift from study to study — by anywhere from three to nine years, depending on the study — so two researchers can sort the same person into different generations. A grouping whose membership depends on who draws the line is not a stable thing to measure.
None of this is a fringe position. In 2020 the US National Academies of Sciences, Engineering and Medicine reviewed the field and concluded that generational labels are not supported by research and cannot adequately inform workforce decisions — about as firm a verdict as a consensus body issues.
A construct this weak should have faded. It persists because a real mechanism sits at its centre, and the mechanism is one everyone recognises from their own life. What happens to you at seventeen marks you in a way the same event at fifty does not.
That intuition has evidence behind it. Mannheim called it fresh contact — the way the young meet the world without an existing frame to absorb it into. Krosnick and Alwin gave it an empirical shape in 1989, at least for political attitudes: openness to change peaks in late adolescence and early adulthood, then falls sharply and stays low for the rest of life. The formative years are formative in a measurable sense.
Grant the mechanism in full, and it still cuts against the way HR uses generations. Its effects are specific, not global: coming of age in a financial crash might durably shape attitudes to job security without making someone a different kind of employee across everything a survey measures. It respects no fifteen-year boundary, either — the unit that matters is whoever was in their late teens or early twenties when a particular event hit, which means someone born in 1981 and someone born in 1996, both filed under Millennial, had nothing like the same formative world. And it produces a cohort effect, which, as the next section shows, a single survey cannot see in the first place.
So the part of generational thinking that feels truest is the part that, followed honestly, takes apart the categories it is used to defend.
The deeper problem is not that generational effects are hard to measure. It is that, in the data an employer actually holds, they cannot be separated from age at all.
Three things move together whenever we measure people over time: age, period, and cohort. Age is how long you have lived. Period is the moment of measurement — the conditions of the year you are surveyed in. Cohort is when you were born. The constraint is arithmetic: knowing any two fixes the third. Someone aged forty-five in 2026 was born in 1981, necessarily, with no other possibility available.
Over decades of repeated measurement you can begin to prise the three apart, though never cleanly, and only on assumptions the data cannot test. In a single employee survey you cannot begin at all. Everyone answers in the same year, so period is constant across the whole sample. With period held fixed, cohort and age carry exactly the same information. A breakdown by generation is then a breakdown by age, redrawn with different labels on the axis.
Slice your engagement data by generation and you learn nothing a slice by age would not have told you — except that the labels tempt you to explain the gap with a stereotype the age bands never would.
This does not make age uninformative. Older and younger staff often do answer differently, and the difference can matter for how you act. The error is not in noticing it but in calling it generational, because the label carries explanations the data does not — that Gen Z wants purpose, that Boomers resist change, that each cohort arrives with a fixed profile of needs. Report the pattern as age, and you stand on firm ground. Report it as generation, and you have smuggled in a theory.
There is a more sophisticated version of the claim, and it deserves a more careful answer. Sometimes the assertion is not that cohorts differ in their scores but that they differ in their drivers — that what builds engagement is not the same for younger and older staff. That is a claim about relationships, not levels, and testing it properly is demanding. Before comparing drivers across age groups, you would first have to show that the questions mean the same thing to each group — a property called measurement invariance. If "opportunity to progress" carries one meaning at twenty-five and another at sixty, a difference in its relationship with engagement may reflect the wording of the question rather than a true difference in motivation. Only then could the strength of each driver be compared across groups. In practice neither step is taken. Separate driver analyses are run for each generation and the differing rankings are presented as if they settle the matter. They do not.
Bobby Duffy's book The Generation Myth is the most serious treatment of cohort analysis in general circulation, and it is routinely misread as a defence of the thing it dismantles. Duffy spends the book taking the stereotypes apart — the entitled Millennial, the lazy Gen Z-er — and is explicit that they have no business shaping how people are managed.
What he defends is narrower and harder: cohort analysis built on decades of repeated surveys, across many countries and millions of respondents, with age and period reasoned through case by case. That work can surface real generational imprints in a population's attitudes. But it rests entirely on the one thing an employee survey lacks — the long run of time that lets the three effects be teased apart. Duffy's method does not rescue the dashboard. It shows precisely what the dashboard would need, and does not have.
Generational labels are a vulgarised form of a careful idea, and they fail twice over. The differences they promise do not appear in the data. And the single annual survey most employers run could not detect them if they did. Read these patterns as what they are: differences due to age, and to the life and career stage that age usually reflects — not generational types. A single survey cannot tell a generational difference from an age one, so the label is best dropped rather than tested.
If your survey reporting still breaks results down by generation, it is worth asking what that cut adds that an age band does not.
Reviewing how your survey reports on age and demographics? We help organisations design analysis that stands up to scrutiny. Contact us for a conversation.
The weight of evidence says largely not. Meta-analyses by Costanza and colleagues in 2012 and Ravid and colleagues in 2025 find few systematic differences between generations in the attitudes that matter at work, and the differences that do appear track age and career stage rather than birth cohort.
In a single survey there is little point. Because everyone answers in the same year, a generational breakdown is mathematically identical to an age breakdown, and the label adds nothing except an invitation to read stereotypes into the gap. Analyse by age and interpret it as age.
Age, period (the year of measurement) and birth cohort are arithmetically linked: knowing any two fixes the third. They cannot be separated without strong assumptions, and in a one-off survey, where period is constant, cohort and age become the same variable.
Age and life stage directly, reported as such. If you want to claim that what drives engagement differs across age groups, that requires testing measurement invariance first and is rarely done, so treat such claims as hypotheses to test rather than findings.
Costanza, D. P., Badger, J. M., Fraser, R. L., Severt, J. B., & Gade, P. A. (2012). Generational differences in work-related attitudes: A meta-analysis. Journal of Business and Psychology, 27(4), 375–394.
Duffy, B. (2021). Generations: Does When You're Born Shape Who You Are? Atlantic Books. (Published in the US as The Generation Myth.)
Krosnick, J. A., & Alwin, D. F. (1989). Aging and susceptibility to attitude change. Journal of Personality and Social Psychology, 57(3), 416–425.
Mannheim, K. (1952). The problem of generations. In Essays on the Sociology of Knowledge (pp. 276–322). Routledge & Kegan Paul. (Original work published 1927.)
National Academies of Sciences, Engineering, and Medicine (2020). Are Generational Categories Meaningful Distinctions for Workforce Management? The National Academies Press.
Ravid, D. M., Costanza, D. P., & Romero, M. R. (2025). Generational differences at work? A meta-analysis and qualitative investigation. Journal of Organizational Behavior, 46(1), 43–65.
Rudolph, C. W., Rauvola, R. S., Costanza, D. P., & Zacher, H. (2021). Generations and generational differences: Debunking myths in organizational science and practice and paving new paths forward. Journal of Business and Psychology, 36(6), 945–967.
Strauss, W., & Howe, N. (1991). Generations: The History of America's Future, 1584 to 2069. William Morrow.
Let’s start a conversation about how employee surveys can help you develop a workplace where people and performance grow together.
