Today in one Croatian newspaper I stumbled upon this graph. It is showing a distribution and comparison of gasoline prices over regions for two types of gasoline. On first look, there is nothing wrong at all with this graph:
Then we examine further, let’s say upper graph for ‘Eurodizel’ gasoline.
Graphs encode data as visual objects. Therefore, these objects have to be accurate and clear. At least, on this particular graph, objects are clear and overall graph design is pretty good. But, it is not accurate!
Lets write this graph on computer:
This graph is using line length attribute of visual perception to encode quantitative information, in form of bars.
Problem is, a bar that is twice as long as some another is immediately perceived as having twice the quantitative value.
So, what the readers get from this graph is that price in Serbia (Srb) is more than two times the price in Bosnia (Bih)! And that is simply not true. Sometimes graphs are even intentionally manipulated in this way to conceal the truth. I hope this was out of ignorance.
The problem with the graph is with the scale that is not starting at 0 EUR. It is showing just a portion of quantitative value.
The right-scaled graph would look like this:
Now the difference is obvious and graph is telling us the truth. There are only sight changes in gasoline price over countries.
Only when this change is too subtle to show it right, graph designer is allowed to set the bottom of quantitative scale to a value greater then zero. But, he has to make sure that his readers are alerted that graph does not give an accurate visual representation of the values in form of text alert or short message.