Statistical Good News

principle

Important progress often arrives as small changes in percentages and rates, making it almost invisible as news. To detect it, compare consistent measures across long time horizons instead of treating vivid daily events as evidence of the historical direction.

A fall from 37% to 36% can represent thousands or millions of lives improving—and still sound too boring to report. If the rate later reaches 34%, the change remains uneventful even as the world quietly becomes different.

E1

Why progress disappears into the denominator

News favors discrete events: a disaster happened, a government fell, a company failed. Statistical improvement has a different shape. It is distributed across many people, arrives incrementally, and becomes visible only when the same measure is compared over time. Each percentage-point change feels negligible in isolation; the accumulated shift can be enormous.

That mismatch makes attention a poor instrument for measuring progress. Negativity Bias may amplify the problem, but the deeper issue is one of measurement: incidents tell you what happened today, while rates and baselines tell you whether the underlying condition is changing.

E1

Where it shows up

The unremarkable falling rate

The sequence from 37% to 36% and then 34% shows the characteristic rhythm of statistical good news: no single step looks transformative, yet the direction becomes consequential when the observations are placed on one timeline.

E1

A better average can conceal a worse reality

A falling overall rate does not prove that every group or place is improving, nor does a long-run trend make today’s suffering unimportant. Statistical good news is a correction to incident-driven perception, not permission to ignore distribution, reversals, or the people still inside the remaining percentage.

Replace the headline with a time series

Tomorrow, take one condition you believe is worsening and find a consistent rate for it at several widely separated dates. Record the baseline, the latest value, and any subgroup differences before deciding whether the latest incident represents the trend—or merely interrupts it.

Episodes that teach this