Truncated Axis: How Cropping the Y-Axis Turns a Ripple into a Tidal Wave

In 2012, Fox News aired a bar chart comparing unemployment rates across three months: 8.6%, 8.5%, and 9.0%. The bars looked dramatically different — the 9.0% bar towered over the others like a cliff face above a plain. The Y-axis started at 8.3%. Had the axis started at 0%, as standard practice requires, the three bars would have been visually nearly identical — because the differences between 8.3% and 9.0% are genuinely small. The chart was not fabricated. Every number was accurate. But the visual impression it created was the opposite of the underlying data's message.

What Is a Truncated Axis?

A truncated Y-axis (also called a cropped axis, suppressed zero, or manipulated baseline) is a chart in which the vertical axis does not start at zero, causing the visual space between bars or data points to be disproportionately large relative to the actual differences between values. The effect is to visually amplify variation that is numerically small.

The distortion works because human visual perception interprets bar height as representing quantity. When a bar chart shows unemployment at 8.5% and 9.0%, the visual expectation is that the bar representing 9.0% should be approximately 5.9% taller than the bar representing 8.5% (since 9.0 is 5.9% larger than 8.5). When the axis starts at 8.3%, however, the bar for 9.0% is visually six times taller than the bar for 8.5% — representing a 600% visual amplification of a 5.9% numerical difference. The eye sees a dramatic gap where the data contains a marginal one.

The Geometry of Misleading Charts

The core principle was formalised by statistician Edward Tufte as the "lie factor": the ratio of the visual effect size in a graph to the actual numerical effect size in the data. A lie factor of 1.0 means the chart accurately represents the data. A lie factor significantly above 1.0 means the chart visually exaggerates the effect; below 1.0 means it understates it. Truncated axes routinely produce lie factors of 5 to 20 or more.

For bar charts and histograms — where the bar's entire area is perceived as representing the value — starting the Y-axis at any value other than zero is almost always misleading. This is a different standard from line charts, where context matters more and where the zero baseline may be genuinely irrelevant. A line chart showing temperature variation between 18°C and 22°C can reasonably start its axis at 17°C without deception, because the viewer is tracking change over time, not comparing absolute quantities against a meaningful zero. The key distinction: bar charts use area to encode value (area must start from zero); line charts use position and slope to encode change (the relevant comparison is relative).

The Fox News Catalogue: Recurring Offences

Media Matters for America documented multiple instances of misleading charts from Fox News broadcasts over several years, not all involving truncated axes but many doing so. The 2012 unemployment chart mentioned above is among the most-cited examples in data visualisation pedagogy. A 2014 Fox News graphic on Affordable Care Act enrolment numbers similarly distorted a modest increase into an apparent dramatic spike by anchoring the axis well above zero and omitting clear axis labels.

The pattern is significant not because Fox News is uniquely dishonest but because it is unusually well-documented. Media monitoring organisations catalogued the instances. The same visual technique appears throughout political and corporate communications across the ideological spectrum — Fox News simply provided unusually clear and repeated examples that became pedagogically useful.

The 2012 US presidential campaign produced a widely circulated chart from a Republican campaign organisation showing job growth over time with a Y-axis that started near the data minimum, transforming a modest improvement in employment into what appeared to be a dramatic recovery followed by collapse. The same data, plotted with a zero baseline, showed steady but unremarkable improvement.

Corporate Earnings and the Deceptive Upswing

Financial communications make extensive use of truncated axes, sometimes for legitimate reasons and sometimes not. A company reporting earnings-per-share of $4.21 in Q3 versus $4.19 in Q2 might present this as a dramatic upward trend with a Y-axis starting at $4.15 — visually suggesting a steep ascent. Investors whose attention was engaged by the chart rather than the underlying numbers might retain the impression of strong momentum where the actual change is less than 0.5%.

Quarterly earnings presentations, investor relations materials, and annual reports are fertile ground for this technique. Analysts who work with raw data are typically unaffected, but the charts in these documents are often designed for stakeholders who will not read the tables — and for those audiences, the visual impression may be the primary takeaway.

In pharmaceutical advertising, the same technique is used to exaggerate the apparent benefit of drugs. A clinical trial showing a disease recurrence rate of 12% versus 15% — a modest absolute reduction — can be presented as a chart where the bar for 12% appears less than half the height of the bar for 15%, simply by starting the axis at 10%. The effect looks dramatic. The actual risk reduction is three percentage points.

When Truncated Axes Are Legitimate

Context genuinely matters. There are situations where starting the Y-axis at zero would itself be misleading or obscure meaningful information:

• Temperature data. A chart of average monthly temperatures in a European city might reasonably run from 0°C to 20°C rather than from -273°C (absolute zero) to 20°C. Zero degrees Celsius is not a meaningful baseline for temperature variation in this context.
• Financial time-series comparisons. When comparing the relative performance of two assets, the percentage change from a common starting point often matters more than the absolute value. A chart indexed to 100 at a start date communicates relative performance more clearly than one showing raw prices.
• Scientific measurements. Many physical measurements have no meaningful zero baseline in the context being discussed (pH, blood pressure, standardised test scores). Here, starting near the data range is standard practice and not deceptive.

The critical question is always: what is the chart trying to communicate, and does the visual representation accurately convey the magnitude of the phenomenon? If a chart is showing comparison between groups or conditions — especially with bar charts — starting the Y-axis at zero is the appropriate baseline. Deviations require explicit justification and clear labelling.

Why It Works: Visual Cognition and the Baseline Assumption

The truncated axis exploits a well-documented feature of human visual cognition: we compare the relative lengths of bars rather than reading their absolute values. When we see a bar chart, we are, roughly speaking, encoding "bar B is twice as tall as bar A" and translating that into "variable B is twice variable A." This is the encoding assumption that makes bar charts interpretable at a glance. Truncated axes break the encoding without breaking the appearance.

Research on graphical literacy has consistently found that most viewers — including educated professionals — do not habitually check whether bar chart axes start at zero. The visual impression is formed before the axes are read. Studies show that charts with truncated axes produce more extreme attitude change and more confident conclusions in viewers than the same data presented with zero baselines, even when both versions are accurate and the viewers understand that the difference exists.

This is related to, but distinct from, the availability heuristic: the vivid visual impression of a dramatic chart becomes the mentally available representation of the data, overriding the more effortful analysis of the actual numbers. The chart that looks dramatic is the one people remember and cite.

The Mcnamara Problem: When Truncated Axes Reinforce Metric Gaming

Truncated axes don't just mislead audiences — they can mislead the organisations that produce them. When management dashboards use truncated Y-axes to display KPIs, small improvements are visually amplified into apparent breakthroughs. Teams see a soaring chart and believe they are succeeding. The visual momentum can suppress the question of whether the underlying improvement is meaningful. This creates a feedback loop between misleading visualisation and the kind of metric-gaming described in Goodhart's Law: the chart becomes the target, and the target produces the chart that looks good rather than the outcome that is good.

Reading Charts Defensively

The practical defence is simple but requires a habit: before interpreting any bar chart, check where the Y-axis starts. If it does not start at zero, ask why, and mentally recalibrate the visual impression. For bar charts presented by parties with an interest in a particular interpretation — political organisations, corporations presenting financial results, pharmaceutical companies publishing clinical data — this check should be automatic.

Supplementary questions: Does the chart show absolute values or percentage changes? Are the axis labels clearly visible? Is there a contextual reason for the truncation, and is that reason stated? Would the visual impression change substantially if the baseline were moved to zero?

Data visualisation is not a neutral technology. The choices made in constructing a chart — scale, baseline, colour, labelling — are rhetorical choices that shape interpretation. The truncated axis is the most well-documented of these choices, but it sits within a broader repertoire of McNamara-style quantitative reductionism — the tendency to let visual simplifications of data substitute for engagement with what the numbers actually mean.

Sources & Further Reading

• Tufte, E.R. (1983). The Visual Display of Quantitative Information. Graphics Press.
• Cairo, A. (2019). How Charts Lie: Getting Smarter About Visual Information. W. W. Norton & Company.
• Huff, D. (1954). How to Lie with Statistics. W. W. Norton & Company.
• Media Matters for America. (2012–2014). A History of Dishonest Fox Charts. mediamatters.org
• Pandey, A.V. et al. (2015). How deceptive are deceptive visualizations? Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems.