While many graph types geared toward comparisons ask the viewer to subtract the difference between the heights of two bars or the space between two points on a line, a deviation bar graph simply graphs the difference. A deviation bar graph encodes the change between two points. The two points can be points in time, the difference between two groups, the change from pre to posttest, or any other dataset suitable for comparison.
Deviation bar graphs lose some detail. The exact percentages of the two points in comparison are not reported. Rather, the difference is reported. Thus a deviation bar graph emphasises the change and asks the viewer to focus on that aspect alone.
Change in ethnicity among students
Source: Stephanie Evergreen
Advice for CHOOSING this option (tips and traps)
When you want the audience to focus on how much change has occurred, use the deviation bar graph. If the audience is going to want to know the exact figures from the two points of comparison, this chart type will not be right for you.
Advice for USING this option (tips and traps)
To create this graph, simply make a regular bar graph, pulling from a data table where you have calculated the change. The negative numbers will go to the left and the positive will go to the right automatically. This could easily become a column graph as well. The only trick is that, by default, most software programs will put the label for the bar in the center, at the axis, partially masking it with the bar. In the example shown above, the labels were retyped into text boxes and positioned on top of the bars so they can be read easily.
Displaying change between two points at a time: Stephen Few articulates options for comparing two points in time, including the split axis bar graph, which he calls a deviation bar graph
Other ways to visualise changes over time
Displaying information as a series of data points connected by straight line segments, on two axes.
Comparing change between two points in time with a line.
Visualising how a group of quantities changes over time. Items are "stacked" in this type of graph allowing the user to add up the underlying data points.