Link Search Menu Expand Document
Data visualisation toolbox

4.1 Bar charts

When to use: Bar charts are good for comparing different parts of a bigger set of data by showing different categories. Alternatively, then can be used longitudinally, i.e., if the same variable has been collected overall an extended period of time.

Bar charts are very useful for showing relationships between categorical (qualitative) variables, as is often used in development or humanitarian projects. For example, survey questions are often asked in the form of ‘select one’ or ‘select multiple’ in relation to a set of potential responses. Each of the responses is a narrative explanation. The distribution of the different responses among the sample can give us an indication of which factors are most important for the target, beneficiary population.

In the case study, we have asked survey respondents on the principal source of drinking water among the households, which can be used to determine access to improved drinking water sources.

The survey question was formed as follows: “What is the principal source of drinking water for members of your household?” Respondents were required to select one from a list of options (shown on the graph below, with the mostly commonly selected answer option in red).

image info

From graph, we can see that there is a significant variation among the households relating to their principal drinking water sources. However, handpumps/boreholes was selected far more than other sources, including 9% more frequently than the second most common water source (24% compared to 15% for piped connections to the house).

Best practices: The y-axis should start at 0, because this provides context in terms of the relative size of the different charts. Ensure that the grid-lines reduced (or removed), borders are removed, and directly label bars (if possible).

When to avoid: Avoid using bar charts if you’re using multiple data points, as bar graphs should be used only for categories. Also, avoid using bar charts if there are too many categories, as the visual can quickly become too busy and confusing for the reader (note: this is not applicable to the same variable overtime).