"Millennials are like this, Gen Z is like that" is one of the most confident stories in modern marketing and management. It is also, mostly, a mirage. The differences are real enough to see, but pinning them on the generation is usually a mistake, because three different forces are tangled together and only one of them is about who you are.
Let's begin with two words that get used interchangeably and should not be. A cohort is simply a set of people who share a starting point in time, usually a band of birth years, so they pass through history in step, meeting the same events at the same age. A generation is a stronger and shakier claim: a cohort said to share a distinctive consciousness, a common outlook stamped by the events it lived through while young.
Mannheim, who founded the serious study of this, was careful about the difference. Sharing a moment in history, he argued, only sometimes produces a genuine generation. A cohort becomes a "generation as actuality" only when its formative experiences crystallise into a shared identity and response, as arguably happened for those who came of age in a war or a depression. Most birth-cohorts never cohere into a generation in that strong sense; they are just people born around the same time (Mannheim, 1928). Ryder added the rigorous demographic version, treating the cohort as a precise analytic unit for tracking how societies change as one age group replaces another (Ryder, 1965).
Here is the problem that undoes most popular generation talk. When you notice that, say, younger people today behave differently from older people, that single observation is consistent with three completely different explanations, and they look identical in a snapshot:
Only the third is really about "generation". And you cannot tell which one you are looking at from a single snapshot; you need to follow cohorts across time and use what demographers call age-period-cohort modelling to separate the strands. Almost no popular generational claim does this.
Predictably, the popular industry raced ahead of the evidence. Strauss and Howe's influential books proposed grand cyclical archetypes repeating on roughly eighty-year cycles, a framework that sells consulting decks but has essentially no empirical support (Strauss and Howe, 1991). Twenge's widely read work documented apparent generational shifts in traits like narcissism, anxiety, and screen time, much of it drawing on cross-sectional data interpreted as cohort change, which is exactly the inference the APC problem warns against (Twenge, 2006; 2017).
The corrective came from researchers who took the confound seriously. Rudolph and Zacher argued that most workplace "generational differences" are age and period effects in disguise, and that treating generations as fixed types is conceptually broken (Rudolph and Zacher, 2017). The meta-analytic evidence backs them up: pooling many studies of generational differences in work attitudes, Costanza and colleagues found the differences were small to near-zero and did not follow the tidy generational story at all (Costanza et al., 2012).
The honest picture is not that generations are pure myth, but that the real signal is small and badly labelled, while the popular version is mostly confound.
The sharpest point is the identification problem above: age, period, and cohort are mathematically entangled, and without the right longitudinal design you simply cannot attribute a difference to "generation" rather than to getting older or to living in a particular year. A great deal of confident generational commentary is, technically, an unsolved equation presented as an answer.
That does not mean no cohort effects exist. Some are real and lasting: people who came of age in the Great Depression carried distinctive habits of thrift for life, a genuine formative imprint. The most debated current case is the rise in adolescent anxiety and depression documented across several datasets from around the early 2010s, which is a real cohort-linked trend. But the leap from "this trend is real" to "smartphones and social media caused it" is exactly where the evidence thins: careful reviews find the correlations between screen use and teen wellbeing are small, inconsistent, and nowhere near strong enough to carry the confident causal story, and warn that the topic has the hallmarks of a moral panic (Odgers and Jensen, 2020). A real cohort trend and a contested cause are being sold as one settled fact.
There is also a much older tell. The complaint that the young are vain, lazy, and disrespectful recurs in almost every era, so reliably that psychologists have studied it as an illusion in its own right: in every generation, adults rate today's youth as worse than the youth of their own day, largely because they compare real teenagers to a flattering, edited memory of themselves (Protzko and Schooler, 2019). If every young cohort in history has looked like it was in decline, the pattern is telling us about the observers and the life stage, not about any particular generation. Today's "Gen Z is fragile" is the newest verse of a very old song.
Finally, the labels themselves. "Gen X", "Millennial", and "Gen Z" have no principled scientific definition, their boundaries are drawn differently by different people, and they function as marketing and media categories rather than measured groups. The signal of how little rigour underpins them came when the Pew Research Center, long a major producer of generational analysis, announced it would stop using the labels unless it had the longitudinal evidence to justify them (Parker, 2023). They are general approximations at best, which seem to satisfy leadership when looking for "simple" solutions to the complex problem of targeting.
Because the ideal design, long-running panel data tracking the same people, is expensive and rare, most of what exists is cross-sectional and cannot in principle separate the three effects. The strong cohort effects that are real tend to be tied to genuinely major shared shocks (wars, depressions), which are hard to generalise from. And the research is heavily Western, while the events that might forge a cohort differ enormously across societies.
Which cohort effects survive proper age-period-cohort analysis, and how large are they really? Is the early-2010s adolescent mental-health decline a genuine cohort effect, a period effect hitting everyone, or an artefact of changing measurement, and what actually caused it? And is there any defensible, evidence-based way to bound a "generation", or is the whole labelling exercise irredeemably arbitrary?
The usable core: a difference between age groups is real, but before you call it generational, rule out that it is just people being young, or everyone reacting to the present. The labels are marketing, and the honest question is which of three forces you are actually seeing.
This is a direct warning about one of the most popular segmentation habits. Generational personas, the "Millennials value experiences, Gen Z wants authenticity" slide, are usually mislabelled age or period effects dressed as fixed traits, and building strategy on them means chasing a moving target while stereotyping millions of people by birth year (Rudolph and Zacher, 2017; Costanza et al., 2012). The disciplined replacement is a single diagnostic question: does this pattern persist within an age band as that group gets older (a true cohort effect worth planning around), or does it move with age (a life-stage effect that every cohort passes through)? The two imply opposite strategies, one is about a durable group, the other about a life stage you can expect to recur. The dual-use line is the consulting economy built on generational archetypes: much of it sells confident birth-year stereotypes that the evidence does not support, and while it is easy money, it produces bad targeting and ages badly. Segment on actual behaviour, needs, and values, which you can measure, rather than on a generational label, which you cannot really define.
Generational framing is seductive in politics ("young voters want...", "the boomer electorate...") but it routinely confuses a life-stage pattern, people often do shift as they age and acquire assets, with a permanent generational character. Treating a cohort as a fixed bloc misreads how its members will change, and pits age groups against each other on the basis of stereotypes that the data mostly does not support. Where a genuine cohort effect exists, a formative shared shock, it is worth understanding; the lazy label is worth dropping.
Policy and forecasting should lean on the demographic cohort concept done properly (Ryder, 1965), tracking how real birth-cohorts move through the population, rather than on media generation labels. It matters for the substance: whether a social trend is an age effect, a period effect, or a cohort effect changes what, if anything, to do about it. A period effect may pass on its own; an age effect will recur with every cohort; a true cohort effect will travel with a group for decades. Getting that attribution right is the difference between a sensible response and an expensive overreaction.
Three habits. When you see a "generational" difference, run the three-way test before believing the label: is it age (they are young), period (it is this year), or cohort (it is when they grew up)? Treat Gen X, Millennial, and Gen Z as marketing shorthand, useful for casual talk, useless as a basis for strategy, and segment on things you can actually measure instead. And when a striking generational trend comes with a confident cause, especially a technological villain, check whether the cause is established or just plausible, because a real trend and a contested explanation are often sold as a single fact.
.png)