Case Study 1: The Truncated Axis on Cable News

How a single design choice — where to start the y-axis — turned a routine employment update into one of the most studied examples of visual dishonesty in modern media.

The Situation

On the morning of December 7, 2012, the U.S. Bureau of Labor Statistics released its monthly Employment Situation report. The headline figure was the unemployment rate, which had fallen to 7.8 percent — down from 7.9 percent the previous month and continuing a slow, three-year decline from the post-2008 recession peak of 10.0 percent. It was a routine release. The change was small. The story was boring.

Cable news programs do not do boring.

Later that same day, Fox News's America's Newsroom broadcast a bar chart accompanying a segment on the jobs report. The chart was on screen for roughly eleven seconds — long enough for a viewer to form an impression, too short for anyone to study the axis labels carefully. The chart showed three bars representing unemployment rates at three different moments: 8.6 percent, 8.8 percent, and 7.8 percent. Three simple numbers, three blue bars, a white background, a red banner across the top reading "CHANGES IN UNEMPLOYMENT RATE."

To a casual viewer, the chart suggested something dramatic: unemployment had been climbing, climbing, and then suddenly fallen off a cliff. The final bar was visually about one-third the height of the first two. A 1.0 percentage point decline — a real but modest improvement — had been rendered as a visual plunge of roughly 70 percent.

The visual impression was possible because of one editorial choice: the y-axis of the chart started at 7.0 percent, not at 0. Everything above that baseline was blown up. Everything below was invisible. The chart was technically accurate — all three values were correctly labeled — but the visual encoding made the small decline look enormous.

Within hours, screenshots of the chart were circulating on the internet. Media Matters for America published a side-by-side comparison showing what the chart would have looked like with a zero-based y-axis: three bars of nearly identical height, the decline visible but proportionate. By the end of the day, the Fox News chart had become a canonical example in dozens of blog posts, academic lectures, and undergraduate visualization courses. It has since been cited in Alberto Cairo's The Truthful Art and How Charts Lie, in Nathan Yau's Data Points, and in countless classroom presentations as the textbook case of the truncated-axis distortion.

The Data

The underlying numbers are publicly available from the Bureau of Labor Statistics, which has published monthly unemployment rates without interruption since 1948. For the three data points shown on the Fox News chart:

• September 2012: 7.8 percent unemployment
• October 2012: 7.8 percent unemployment (revised)
• November 2012: 7.7 percent unemployment (later revised to 7.8 percent)

Some contemporary reports noted that the exact values shown in the chart — 8.6, 8.8, and 7.8 — did not match the actual BLS numbers for any obvious three-month window in 2012. The 8.8 figure had last been reported in March 2011. This compounded the problem: the chart was not only visually distorted, it was working from data points that appeared to have been selected rather than pulled from the most recent release. But the central distortion — the one that made the chart famous — was not the specific numbers. It was the y-axis.

The actual change the chart was trying to depict was modest. Taking the extreme case of the chart's own numbers, the drop from 8.8 to 7.8 is a 1.0 percentage point decline, which in relative terms is a decline of about 11.4 percent (1.0 divided by 8.8). That is the effect size in the data.

The visual effect size is easy to measure. On the broadcast, the tallest bar (8.8 percent) occupied roughly 245 pixels of vertical space. The shortest bar (7.8 percent) occupied roughly 70 pixels. The visual change from 245 pixels to 70 pixels is a decline of about 71 percent.

Apply Tufte's lie factor formula from Section 4.1:

$$\text{Lie Factor} = \frac{\text{Size of effect shown in graphic}}{\text{Size of effect in data}} = \frac{71\%}{11.4\%} \approx 6.2$$

A lie factor above 6. The visual representation exaggerates the data by more than a factor of six. By Tufte's own criterion — which calls anything outside the 0.95 to 1.05 range misleading — this chart is approximately a hundred and twenty times past the threshold of honesty. And this is computing the lie factor in percentage terms. In absolute terms, comparing pixel heights to percentage points, the distortion is even more severe.

The Visualization

It is worth being precise about what the chart was and was not. It was not a chart that lied about the data. Every number on every bar was correct (relative to the figures Fox News chose to show). There were no digital alterations. The colors were professional. The typography was clean. Anyone examining the chart carefully could read the axis labels and see that the y-axis started at 7.0.

The chart lied through a structural choice that is invisible to a casual viewer but dictates the visual impression: the baseline.

Pre-attentive processing, the mechanism introduced in Chapter 2, meant that viewers' visual systems registered the relative heights of the bars in a fraction of a second — before they had any opportunity to read the axis labels. The brain's heuristic for comparing bars in a bar chart is simple and nearly automatic: taller bars represent bigger values, shorter bars represent smaller values, and the ratio of heights corresponds to the ratio of values. This heuristic is right when the y-axis starts at zero. It is badly wrong when the y-axis is truncated.

Cleveland and McGill's encoding accuracy research (again, from Chapter 2) explained why the heuristic exists. Length on a common scale is the most accurately perceived visual encoding for quantitative comparisons. Viewers are good at judging bar heights. The perceptual system has been trained by thousands of bar charts to take those judgments at face value. The truncated axis exploits that training — it hijacks a visual skill that usually produces accurate readings and redirects it toward a distorted conclusion.

The result was a chart that was simultaneously true (in its labels) and false (in its impression). This is the definition of visual dishonesty: a gap between what the data says and what the chart implies.

The Impact

The chart became one of the most widely circulated examples of visual distortion in the modern era. Media Matters published its corrected version within hours, showing side-by-side what the same data would look like on a zero-based axis: three nearly identical bars, the 1-point decline barely visible, the story visually honest but politically deflated. The contrast was the point. Without the truncation, there was no dramatic "cliff" to show, and without the dramatic cliff, there was no compelling chyron to wrap around an eleven-second graphic.

Professional visualization journalists picked up the story almost immediately. Alberto Cairo, then at the University of Miami, wrote about it in his blog and later included it in multiple books. Nathan Yau of FlowingData used it as a teaching example. Academic courses in journalism schools, statistics departments, and data visualization programs adopted it as a standard exhibit. A generation of undergraduates now learns about the truncated-axis distortion by starting with this chart.

Fox News did not formally respond to the criticism, and the chart was not corrected on air. But the incident had a lasting effect on the broader conversation about visualization in news media. Major outlets — including The New York Times, The Washington Post, The Wall Street Journal, and The Financial Times — explicitly updated their visualization guidelines in the years that followed, adding explicit language about when bar chart axes must start at zero. The New York Times graphics desk had used zero-based baselines for bar charts long before 2012, but they made the principle more prominent in internal training materials after the Fox News incident drew attention to how much damage a single truncation could do.

Meanwhile, the techniques for spotting truncated axes became part of basic visualization literacy. Teachers now routinely ask students to find the axis, check whether it starts at zero, and compute the lie factor before interpreting any chart from any source. The Fox News chart is the canonical example precisely because it is so extreme and so easy to analyze: one chart, three bars, one choice, a lie factor of more than six.

Why It Worked: A Visualization Analysis

The Fox News chart was effective — in the cynical sense that it produced the intended visual impression — for reasons that connect directly to principles from Chapters 2, 3, and 4.

1. The distortion bypassed conscious scrutiny. Pre-attentive processing handles relative bar heights in under 250 milliseconds. A broadcast graphic that appears for eleven seconds gives the viewer roughly forty-four opportunities to form a pre-attentive impression — and approximately zero opportunities to read, interpret, and correct for the axis labels. The viewer's conscious mind never got a chance to ask, "Wait, does that axis start at zero?" The impression was formed and locked in before the question could arise.

2. The chart obeyed every surface convention. The design looked professional. The bars were cleanly drawn, the colors were appropriate for a news broadcast, the typography was legible, the axis was labeled. Every element of the chart signaled competence and authority — and all of those signals were correct, at the surface level. A viewer looking for obvious warning signs of manipulation would have found none. The distortion was structural, not cosmetic.

3. The encoding matched the viewer's expectations for bar charts. Viewers have been trained by thousands of textbook bar charts to read "taller bar = bigger number, ratio of heights = ratio of numbers." The Fox News chart exploited this training by violating the invisible assumption on which it rests — that the baseline is zero. The viewer's heuristic kept functioning, but it was applied to a chart where the heuristic was wrong.

4. The chart framed the editorial choice as a technical decision. The choice to start the y-axis at 7.0 was presented as a neutral production decision, the kind of thing a graphics intern might do to "make the chart look good." In fact, it was an editorial decision that carried a specific argument — that the drop in unemployment was dramatic rather than modest. By hiding the editorial choice inside a technical setting, the chart maker avoided having to defend the editorial claim directly.

5. The audience had no opportunity for comparison. A viewer watching live television does not have the option to pause, screenshot, and compare the chart to an alternative framing. The only version of the chart the viewer saw was the misleading one. The corrected version existed — Media Matters published it — but it circulated on the internet, for an audience that was already curious enough to seek out fact-checks. The original chart's audience and the corrected chart's audience were mostly different people.

Complications and Counter-Arguments

Intellectual honesty requires acknowledging the counter-arguments that have been offered for the Fox News chart — and explaining why they do not rescue it.

Counter-argument 1: "The chart shows the real numbers. Nothing was fabricated." This is true and irrelevant. The definition of a lie factor distortion is that the data itself is accurate but the visual representation is not. The lie lives in the encoding, not in the numbers. Section 4.1 of this chapter is built around exactly this insight.

Counter-argument 2: "Truncated axes are sometimes legitimate." This is true. As Section 4.2 discussed, line charts of variables that cannot meaningfully be zero (body temperature, stock prices, atmospheric pressure) often require truncated axes to show meaningful variation. But the Fox News chart was a bar chart, not a line chart. The visual encoding of a bar chart — length from a baseline — requires a zero baseline to be interpreted correctly. The exception for line charts does not apply.

Counter-argument 3: "The viewer should pay attention to the labels." This argument shifts the responsibility for honesty from the chart maker to the viewer. It is the visualization equivalent of "the fine print said so." Pre-attentive processing means the impression forms before the labels can be read. The viewer is not being careless; the viewer's visual system is doing what visual systems do. Blaming the viewer for being deceived by a designed-to-deceive chart is a failure of ethical reasoning.

Counter-argument 4: "Every chart has to start somewhere. Zero is arbitrary." This is a misunderstanding of visual encoding. Zero is not arbitrary for bar charts because bar length is the encoding, and length implies a baseline. If the baseline is not zero, the encoded quantity is not "the value" but "the value minus some offset." Communicating that offset to the viewer is the chart maker's job, and a truncated axis without extreme visual cues fails that job.

Lessons for Modern Practice

The Fox News chart is the most-cited example of truncated-axis distortion, but its real lessons apply to charts that will never be broadcast on national television.

Check your bar charts against the zero-baseline rule, automatically and every time. The rule is simple: if the chart type encodes data through length or area (bar charts, column charts, stacked bars, area charts), the axis starts at zero. Full stop. If this makes your data look too flat, the flatness is the honest message. The rule exists because the visual encoding demands it, not because convention requires it.

Do not trust defaults. Most modern plotting libraries — including matplotlib, seaborn, Excel, Tableau, and Power BI — auto-scale axes by default. For bar charts, auto-scaling can produce truncated axes that look reasonable to the software but violate the encoding principle. You must override the defaults. We will return to this repeatedly in Parts III and IV, but the principle starts here.

Compute the lie factor when you are unsure. If you suspect that your chart might be misleading but cannot tell by eye, measure it. Compare the visual change in bar height or slope to the numerical change in the data. If the ratio is outside 0.95 to 1.05, fix the chart. You do not need professional judgment — you need a ruler and a calculator.

Remember that most distortion is accidental. The Fox News chart became a public case study because it was broadcast by a partisan news organization during a politically sensitive news cycle. That framing suggests intentionality. But the same distortion appears hundreds of thousands of times a day in internal corporate slide decks, student term projects, and automated BI dashboards — almost always without any intent to deceive. The fix is the same whether the cause was partisanship or inattention: start the axis at zero for bar charts, and check your work.

Honest charts can be boring, and that is sometimes the honest message. A bar chart showing three values within a narrow range may indeed look flat. The reader will conclude: nothing much changed. If nothing much changed, that is the story, and the chart has done its job. The temptation to "make the chart more interesting" by truncating the axis is the temptation to substitute a different story — a more dramatic story — for the one the data actually tells. Resist it.

Discussion Questions

1. On intent versus impact. Do you think the Fox News graphics team intended to mislead, or is it more likely a case of production shortcuts and default software settings? Does the answer change how you evaluate the chart ethically? Where would you place this incident on the four-zone spectrum introduced in Section 4.6?

2. On the audience's responsibility. Some commentators argued that viewers should simply "read the axis labels" and correct for themselves. What does pre-attentive processing (Chapter 2) tell us about whether that is realistic? Is it ever fair to blame the viewer for being misled by a chart?

3. On repetition and cumulative effect. A single viewing of a misleading chart may not change anyone's mind. But what about repeated exposure to similar distortions across multiple broadcasts, news stories, and social media feeds? Does the cumulative effect of many small distortions matter more than any single example?

4. On industry-wide norms. Major newspaper graphics desks — The New York Times, The Washington Post, The Financial Times — consistently start bar chart axes at zero. Television news graphics departments have been more inconsistent. What explains the difference? What structural changes might encourage consistent visualization ethics in broadcast media?

5. On your own work. Think of the last chart you made with a truncated axis. Was it a line chart or a bar chart? Did you have a good reason for the truncation, or did you accept the software default? If you had to defend the choice to a skeptical reviewer, could you?

The Fox News chart is remembered not because it was uniquely dishonest but because it was so easy to analyze. Three bars, one axis, one choice. When you can point to a single design decision and compute exactly how much it distorted the message, you have a teaching example. The Fox News chart is a teaching example. Use it on yourself.