Divergent Opinions on Transatlantic Alliance — #MakeOverMonday
The #MakeOverMonday challenge of December 5th was about divergent opinions on the status of their partnership between Germany and the United-States.
A few observations on this original plot :
- It is composed of two images to answer basic questions about German respondents and US respondents
- Two images makes it impossible to answer all possible questions in a single instant of perception (we need to look back and forth to compare Germany and the USA, which defeats the entire purpose of the visualization)
- It is redondant because the sum of the percentages is always 100%
- There is a lot of non-data ink : the percentage labels are written every 10%, the parallel lines are drawn, there is no new information in plotting the Bad / Good instead of only the Bad and the illustration in the background is distracting
- The scale does not start at 0%
- Hides the true information of the data, which is to compare how the gap amplifies over time
Pfiou, that’s a lot for such a few data…
My analysis suggested a different approach. Using Jacques Bertin “Semiology of graphics” framework, we can define:
Invariant : Bad opinion about a partner
Components :
- Date
- Q per 100 respondents
- Partner
There are three components here, hence three types of questions : For a given date, for a given percentage and for a given partner. Each of these questions has three levels of reading : elementary (single correspondence), intermediate (group of elements in the component), overall (entire component). Since we have three components, all these questions (3 × 3 = 9) can be answered within the frame of a single image.
Next, we need to define the visual variables for each component beside the two dimensions of the plane which will define the Date and the Q per 100 respondents. I chose to use a different color as a retinal variable for the Partner, and to connect the ends with a bar to represent the gap between the opinions of Germans and Americans. To clarify the gap, I also use size as a second retinal variable to label the difference between the Bad opinion from Germany towards the USA and reciprocally.
To maximize the data-ink used I chose only to label the length of the bar representing the difference in respondents opinions.
I did not plot the individual opinions for each respondent because the real story is in the gap between Germany and the USA and its variation over time, and it would charge the graph uselessly.
Hence, the plot is improved because of the following:
- One image to answer all the questions
- Every piece of data is valuable : the data-ink is maximized
- The story behind the data stands out
