Data Journalism

Numbers as Narrative

Data journalism uses quantitative evidence — statistics, datasets, visualizations — to tell stories. At its best, it reveals patterns invisible to anecdotal reporting. The Panama Papers (2016) analyzed 11.5 million leaked documents using data techniques to expose offshore tax havens used by world leaders. ProPublica's 'Machine Bias' investigation used statistical analysis to show racial bias in criminal sentencing algorithms.

But data journalism can also mislead, sometimes without the journalist intending to. Numbers carry an aura of objectivity that text doesn't. When someone says 'studies show' or presents a chart, readers tend to lower their critical defenses. This is the authority bias applied to quantitative information.

The core skill of media-literate data reading is asking: what choices were made before I saw this number? Every statistic is the product of dozens of decisions — what to count, how to count it, what time period to use, what to compare it against, and how to visualize it. Each decision can change the story.

Six Ways Data Misleads in News

1. Truncated axes. A bar chart where the y-axis starts at 95% instead of 0% makes a difference between 96% and 98% look like a massive gap. Always check the axis.

2. Cherry-picked time frames. Crime went up 40% — but from when to when? Picking a low point as the start date can make any trend look dramatic. Ask: what happens if you extend the time frame?

3. Absolute vs. relative numbers. 'Doubled' sounds alarming until you learn it went from 2 to 4 cases in a population of 10 million. Relative risk (100% increase) sounds scary; absolute risk (0.00002% increase) does not.

4. Correlation presented as causation. 'Countries that eat more chocolate win more Nobel Prizes' was a real published study. The correlation is real; the causal claim is absurd. Both correlate with national wealth.

5. Survivorship bias in data. Measuring only successes. 'Our graduates earn $90K on average' — but what about the 40% who dropped out?

6. Missing denominators. '500 adverse events reported' means nothing without knowing how many total doses/uses occurred. The denominator is always the most important number that's most often missing.

Evaluating Data Sources

When a news article cites data, trace it back to the source. A story might say 'according to a study' — but was it published in a peer-reviewed journal, or posted as a white paper by a think tank with an agenda? Was the sample size 30 people or 30,000? Was the study funded by a party with a financial interest in the outcome?

Good data journalism shows its work. The best practitioners — FiveThirtyEight, The Economist's data team, The Guardian's datablog — publish their datasets and methodology alongside the article. If you can't see the underlying data, you're being asked to trust on faith.

One powerful technique: when a story presents a striking statistic, search for the same statistic in a different outlet. If no one else is reporting it, either one outlet has a scoop or someone made an error. If others report it differently, comparing versions reveals which framing choices were made.