Research essay
When the chart headline overstates the chart
Published
Executive summary
Before believing a chart’s story, establish what it measures. A change is not a level, a forecast is not an observation, and two lines moving together do not establish a cause.
Lessons from reviewing financial charts. All numerical examples below are invented illustrations, not market data or reproduced research charts.
A chart can be accurate while the sentence attached to it is wrong. One common slip is moving from “these firms account for most of the increase” to “these firms account for most of the total”. Another is reading the amount of interest-rate tightening as the final interest rate.
Write the measured quantity as a plain sentence before interpreting the picture. Include the unit, starting point, end point and whether the number is a level, change, growth rate or share. If that sentence does not support the headline, the headline needs another source.
Consider these hypothetical capital-spending figures:
| Group | Earlier spending | Later spending | Increase |
|---|---|---|---|
| A | 20 | 30 | 10 |
| B | 100 | 105 | 5 |
Group A contributes twice as much to the increase. Group B still spends three and a half times as much at the end. Both statements are true; substituting one for the other reverses the impression.
A comparison of trailing twelve-month spending also differs from spending within a single quarter. The time window is part of the quantity, not a caption detail.
A hypothetical policy rate rising from 2% to 4% increases by 2 percentage points, or 200 basis points. Its final level is 4%.
A chart relating starting inflation to the amount of tightening cannot directly tell you the final policy rate without the starting rate and further assumptions. Nor can a fitted historical relationship automatically be reversed into a forecasting rule.
A bar for expected full-year issuance is not directly comparable with an unadjusted year-to-date total. A blended earnings number containing estimates is not a fully reported outcome. A claim about a long historical record cannot be established by a plot showing only recent years.
These distinctions often appear in small labels. They belong beside the conclusion whenever they change what a reader should believe.
Two lines on separate axes can look closely aligned because of the chosen scales. If one axis is inverted, even their apparent direction may be misleading. The vertical distance between them has no common numerical meaning unless their units and scales make it meaningful.
Check each axis independently. Re-express the proposed relationship in words or in a calculation using the actual series. An annotation marking an intervention or policy event is a lead to verify, not proof of the event’s amount or effect.
A few large firms can dominate a change measured over one interval without representing a stable share of the economy. A sample chosen from unusually turbulent episodes need not describe ordinary periods.
Identify the population and the selection rule. Check whether the series is nominal or inflation-adjusted, whether it includes acquisitions or disposals, and whether later revisions changed the historical values. Pursue the checks that matter to the claim rather than applying a ritual checklist to every graphic.
A close fit describes the selected observations. It does not establish which variable caused the other, whether both respond to a third force, or whether the relationship will survive a different regime.
When the title, labels and plotted values conflict, leave the interpretation unresolved until the source is clarified. Guessing which label was intended creates an apparently clean answer by discarding the uncertainty.
Before taking a chart into a thesis, finish four sentences:
- This chart measures …
- The comparison is between … over …
- It supports the narrower statement that …
- It does not yet establish …
Then ask what extra evidence would justify the final step from measurement to explanation. Sometimes the answer is a simple unit correction. Sometimes it is underlying data or an entirely different study.
The point is to preserve what the chart genuinely contributes. Precision makes the argument stronger, even when it makes the headline smaller.