Emoji Use by Age and Gender: Top 50 Mapped by Bias

About This Chart

Data Visualization Method

• Data: For each participant, the share of every emoji among all their emoji was measured, then reduced to two association values — an age correlation (Pearson r) and a gender effect size (Cohen's d). These are computed statistics, not raw usage counts.
• Sample and period: 226 people who donated German-language WhatsApp chats (about 309,000 messages), with message timestamps running from 2012 to 2015.
• Selection: The 50 most frequently used Unicode emoji, ranked across 694 distinct emoji and together covering roughly 84% of all emoji occurrences. Text emoticons were excluded, and emoji were chosen by frequency rather than by how strong their bias is.
• Processing: Shares were normalized per participant, and users with no emoji were treated as zero. Two tokens were recomputed to fix a matching error in the original code, and two overlapping markers were nudged apart using display-only coordinates that leave the underlying values unchanged.
• Chart form: A scatter plot with a crosshair at the zero point: the horizontal axis carries the age association, the vertical axis the gender association. Every point is drawn from one consistent emoji image set (Twemoji) at equal size, and numeric ticks are replaced by directional axis titles.

Creator's Note

The idea was to avoid making two separate charts — one for age, one for gender — and instead place both demographic dimensions on a single coordinate plane, so each emoji's lean can be read in one glance rather than cross-referenced between two figures.

The spread is widest along the gender axis: the heart-eyes face sits at the female-biased extreme (Cohen's d ≈ 0.51) while the plain grinning face sits at the male-biased extreme (≈ −0.30). On a map like this, sitting near an edge means "leans that way," not "belongs there" — the more interesting object is the overall landscape, not any single ranking.

Data Sources

• Koch, Romero & Stachl (2022) — the study defining the sample, the emoji measures, and the intended direction of each effect size.
• OSF project archive — the analysis code and result tables used to recompute the per-emoji age and gender values shown here.

Limitations

These values describe one specific sample, not emoji users in general. The participants were predominantly German WhatsApp users, the gender split was uneven (162 women to 64 men), and ages skewed young, with few people above 40. The messages date from 2012 to 2015, so the chart reflects that moment rather than how emoji are used today.

A few points are the author's reanalysis rather than figures lifted verbatim from the study. Two emoji were recomputed to correct a substring-matching issue in the original code, and one heart symbol was rebuilt from a small subset of nine participants that fell below the study's reporting threshold. They are flagged here so the derived points are not read as published results.

Frequency was used only to decide which 50 emoji appear; it is not encoded in position or size. Every marker is the same size, and the two nudged markers moved for legibility alone — their age and gender values are unchanged.

What This Chart Doesn't Show

The horizontal axis is not average age, and the vertical axis is not the share of women or men who used an emoji. Both encode the direction and strength of an association — a correlation on one axis, an effect size on the other — so a marker near the "older" or "female" edge signals a stronger lean, not a literal age or percentage.

Distance from the center and marker size say nothing about how often an emoji is used. The emoji farthest from the middle is not the most-used one, and no point is larger for being popular; usage decided only which emoji made the Top 50.