Google’s Public Data directory has visualizations of data from many sources, and this post examines a couple of them, thinking about their design and possible ways the visualizations may be misleading.
The first visualization I’m going to consider has fertility rate, or rather the average number of children per woman, on the Y-axis and life expectancy on the X-axis. The data comes from World Bank.
One shortcoming of this graph is in the nature of the bubble chart, namely that the size of the bubble is supposedly proportionate to the population of the country it represents, which is difficult to do accurately with circles. This problem is exaggerated with this data, as the degree to which bubbles are magnified can be altered by the viewer.
A second shortcoming is in the dimensions of the graph, which change with the size of the browser window. The scales seem reasonable for the data being presented, but depending on how narrow or elongated the X-axis is, it can appear that the difference between countries is more or less drastic.
This second visualization considered here is a line graph showing the broadband penetration rate, i.e. “Number of high-speed internet connections (capacity equal or higher than 144 kbit/s) per 100 inhabitants,” on the Y-axis and the year, ranging from 2003 to 2010, on the X-axis.
A line graph in this case is appropriate to show continual change of each country.
This graph is based on data from Eurostat, which importantly shows data only from countries in the European Union, and so does not include all European countries, much less allows comparisons with those of other continents. What countries are included in the graph can be specified from the checklist to the left of the chart. Selecting too many makes for a cluttered graph, so I have a more reasonable number checked.
This graph also has the issue of being able to tell a story of rapid implementation of broadband or one of slower, gradual adoption depending on the size of the browser window.
These visualizations are good at communicating information, but it must be recognized that there is malleability and distortions that shape how data is interpreted.