Axis Shenanigans

Definition

Because the human eye reads visual patterns far more readily than tables of numbers, a graph's axes — their labels, starting point, continuity, and scale — do most of the persuasive work, independent of the underlying data. A graph can be numerically accurate and still visually lie, by choosing an axis treatment that exaggerates or minimizes the difference the data actually shows.

In the Book

Levitin catalogs the specific techniques: unlabeled axes (a conference poster comparing "SZ" and "HCs" on a y-axis with no units, so the reader cannot tell what is even being measured); a truncated vertical axis, illustrated by a 2012 Fox News graph of the Bush tax cuts expiring, where a bar six times the height of another visually implies a 600% tax hike when the real change is 35% to 39.6% — a 13% relative increase, made obvious once the axis is redrawn starting at zero; a discontinuous axis, where a steady 5%-per-year crime increase is made to look like a dramatic spike by compressing five years of data into the graphic space previously used for two; and an artificially extended axis on a home-price graph, which bends a constant 15% annual growth rate into a curve that visually implies runaway acceleration — a distortion Levitin resolves by noting that steady percentage growth should be plotted on a logarithmic scale, where it shows up correctly as a straight line.

Why It Matters

Once you know the specific repertoire of axis manipulations, you can check any chart in seconds — look at the axis labels, the starting value, and whether the scale is continuous — before accepting the visual impression it creates. This matters anywhere data gets summarized into a picture for a non-technical audience: news broadcasts, investor decks, dashboards, policy debates. The concept generalizes beyond graphs to any visual encoding of a quantity: the presentation format itself is a persuasive choice, separable from and sometimes contradicting the data it claims to represent.