How to make a generational digital culture comparison

Comparing age groups by their online habits produces the same article every few years and settles nothing. The comparison is badly specified: the groups are defined by birth year, which is not a cause of anything, and the differences are attributed to the group rather than to what the group walked into.

A better comparison is available and rarely made. Compare the entry conditions instead: what it cost to get online, who was already there, what the default settings were, and what the first thing you were asked to do was. Those are the things that differ, and they explain the habits without appealing to anything about the people.

What to take away

The four entry conditions

None of those rows mentions age. All of them varied enormously over time, which is why they correlate with age, and that correlation is what generation talk is picking up. The correlation is loose: access cost varied by household as much as by decade, and the existing population varied by which country and which language you arrived in.

In the United States, dial-up plans in the early 1990s often charged by the hour, and phone time could be metered too.

AOL moved to a flat monthly fee for unlimited time in the late 1990s. As home broadband became more common, an always-on connection became ordinary rather than a choice.

The general mechanism here is the ordinary one described under diffusion of innovations, where early and late adopters differ systematically in ways that have nothing to do with when they were born. The same entry conditions show up in generational statistics, where age, cohort and period blur together.

Working an example without naming a generation

Take two people. One arrives where access is metered, the population is small and self-selected, obscurity is the default, and the first prompt asks for a handle. The other arrives where access is continuous, everyone they know is already there, discoverability is the default, and the first prompt asks for a real name and a contact list.

The first is a metered dial-up account in the early 1990s, used for Usenet and local bulletin boards, where a handle was normal. The second is a Facebook account in the late 2000s, opened on an always-on home connection, where a real name and a contact list came with signing up.

The first person will treat online space as somewhere separate from their offline life, will assume a handle is not a name, and will be surprised when something they wrote is found. The second will treat online space as continuous with everything else, will find handles slightly evasive, and will assume permanence from the start.

Those are two different sets of habits with a clean explanation. Neither requires a claim about attention spans, brains or values.

It also predicts what the generation account does not: a late arrival into the first set of conditions gets the first set of habits, whatever their age. What the labels get wrong more generally is under generational digital culture.

Why the label version keeps winning

Because it is easier to write and easier to sell. A label is short, a category is memorable, and the reader can place themselves in it immediately. Entry conditions require a paragraph and do not produce a headline.

The labels in circulation, Gen X, Millennials and Gen Z, each cover people who arrived under very different conditions.

A measurement reason also applies, since every dataset records birth year but none record entry conditions. So anyone with data has age and lacks the thing that age stands for. The finding follows the available variable, a general research problem rather than a fault of any one study.

The same pull toward what is measurable shows up in how benefits of social media get counted.

The result is a literature full of age differences, which are real, attached to explanations about age, which are usually not supported. Distinguishing the effect of an era from the effect of an age is the standard difficulty of a cohort study, and it does not go away by having more people in the sample.

Comparisons that do work

Each of those has a fact of the matter. "Which generation is better at this" does not, because the categories were not built for the question.

Using this when you read something

Ask what the writer thinks the mechanism is. If the answer is that a group grew up with a technology, ask which part of growing up with it is doing the work, and you will usually find one of the four conditions above, stated vaguely.

Ask whether the finding would still hold for someone who arrived under those conditions at another age. If yes, the finding is about conditions; the generation framing is decoration. If the writer has not considered it, the finding remains untested against its most obvious rival explanation.

The same test applied to claims about how formats spread is described under memes and virality.

Asking the entry questions

Three questions cover most of the ground.

Cost and company come back from memory reasonably well. Defaults do not, so ask what someone's first profile showed and whether they expected anyone to find it.

Common questions

Are there no real age differences at all?

There certainly are, in the ordinary ways that people at different life stages differ. The question is whether a given difference is one of those or an artifact of when someone arrived, and the two are very hard to separate.

Is this just saying labels are unhelpful?

It is saying something more specific: that a better variable exists, that it is observable by asking two or three questions, and that using it changes the conclusions rather than only the wording.

Can entry conditions be recovered retrospectively?

Partly, by asking. Recall about cost and about who was already there is reasonably good. Recall about default settings is poor, because nobody attended to them at the time.

Do entry conditions still vary now?

Considerably, by country, by household and by which device someone starts on. The idea that conditions have converged is itself a claim from one part of the world, and the differences behind it are covered under digital divide research.