Subtle distortion: Whether infection numbers or economic data – how we assess a statistical trend depends surprisingly heavily on the shape of the chart, as an experiment reveals. Thus, we are more pessimistic about a trend when presented to us as a bar chart than when we see it as a line or point cloud. This psychological distortion effect also works regardless of the direction of the trend and the coloring of the bars.
Diagrams play an important role in science and business, but also in everyday life. They enable us to understand complex statistical relationships and developments and to derive decisions and forecasts from them. But as helpful as such visual representations are, they can also be misleading. It has long been clear that features such as scale, axis sections and also coloring influence our assessment of the data shown in the diagram.
Bars, lines or points
Stian Reimers and Nigel Harvey from University College London have now uncovered another distortion effect. They wanted to know to what extent the forecasts made on the basis of a depicted trend are influenced by the form of presentation. To do this, they showed their more than 4,000 test subjects diagrams of 50 data points each, which were displayed as bar charts, lines or point clouds. Depending on the test run, the trends were increasing or decreasing to varying degrees.
The test persons received the information that the weekly sales figures of a company were involved. Their task was to use the data presented to predict sales success for the next eight weeks. “If the same data series leads to different forecasts as a bar, line or point chart, then it is important to know how and why,” the researchers state.
More negative forecasts on bar charts
And indeed: When assessing the trends and the forecasts based on them, there were clear format-dependent differences. “Participants made more pessimistic predictions when viewing bar charts than when viewing line or scatter charts,” Reimers and Harvey report. Although the charts were based on the exact same data, their predictions for bar charts were more negative. “This effect was evident in both ascending and descending trends,” the researchers said.
But why do the bars of all things make us more pessimistic? What feature leads us to this distorted assessment? To find out, Reimers and Harvey conducted further tests. In these, they used various light and dark colored bar chart versions and in one go reversed the bars so that they were hanging down from above. In a test variant, they also modified the line chart in such a way that the line ascended or descended in steps – in principle similar to the top edges of the bars.
Color doesn’t matter, but orientation does
These tests showed that the coloring of the bars does not change the distortion effect, both light and dark bars lead to more pessimistic forecasts than the other chart types. The comparison with the stepped lines showed a similar result: Even then, the test subjects’ predictions for the bar trends remained more negative. According to the scientists, the stepped edges or the more strongly emphasized area of the bars are therefore not responsible for this effect.
Interestingly, however, when the researchers reversed the direction of the bars so that they hung down from above, the distortion still occurred – but the direction was reversed. “The forecasts based on the hanging bars were now more positive than on the line charts,” report Reimers and Harvey. They attribute this to an optical effect that causes participants to underestimate the length of the bars and their development when they are aligned normally, but overestimate them when they are hanging down.
All of this confirms that visual representation has a significant impact on how we interpret data and trends. “Our predictions about what will happen next are not only influenced by the data, but also by how it is formatted,” says Reimers. “It is obvious that this has considerable importance – also in everyday life. Because on this basis we consider, for example, whether we can meet relatives during a pandemic or whether we can afford a mortgage.”
According to the researchers, this places a significant responsibility on those who create such diagrams. Because they can manipulate our assessments. “When a fund manager shows past performance as a line chart, he can lead potential investors to have more optimistic expectations about future profits,” the scientists explain. “Conversely, a crime statistic can have a more reassuring effect when presented as a bar graph.”
At the same time, knowledge of these effects also offers the possibility of compensating for other distortions that are less easily influenced: “Many of these distortions are deeply anchored in our worldview and are therefore difficult to change,” explains Reimers. “By contrast, we can easily control the chart format – and perhaps use it deliberately to offset other distorting effects.” (International Journal of Forecasting, 2023; doi:10.1016/j.ijforecast.2022.11.003 )
Source: City University London