Source: https://www.census.gov/prod/2013pubs/p20-570.pdf
I found this article while searching for some data in census.gov website. This visualization is from a paper published on Families and Living Arrangement trends in the United States in the year 2012. This is the 1st graph in the article, which includes a number of graphs depicting various trends in the American Family and Living Arrangements. The graph conveys the changing trends in different Household types from the years 1970 to 2012. The graph is a Stacked Bar Graph where each stacked bar for a given year, depicts the percentage share of that particular household with the total of the different stacks in a bar adding to 100%. We have stacked bars of household types for the years 1970, 1980, 1990, 2000, 2005, 2010 and 2012
The things I liked about the graph:
- The graph is extremely easy to understand. The title of the graph is Household types, 1970 to 2012 and the graph shows exactly that. There is no confusion as to what is in the graph. It is a fairly simple graph, conveying what it is supposed to convey.
- Each stack in the individual bars are labeled with the percentage number. Hence it is not very difficult to figure out the exact value of each household type share by looking at the graph.
- The X-axis and Y-axis are both labeled clearly and there is no missing values or confusion regarding the scales.
- The different colors used to identify the different household types helps in understanding the share of that household type in the whole bar.
Things I did not like about the graph:
- In the paper the first sentence below this graph, marked in red says “The share of households that married couples maintained has fallen since 1970, while the share of non family households has increased”. Although this statement does appear to look true by looking at the graph, but the change does not look so drastic especially if you consider the years from 1990 to 2012. The change in trend in these years does not look too drastic but rather gradual. I feel if this statement was intended to be conveyed by the visualization, then it should be obviously evident and should not take multiple looks to understand.
- The gap between the years for the consecutive bars, is not consistent. The gap between each of the first three bars is 10 years, then the gap between the years becomes 5 years for the next 4 bars and then ends with a 2-year gap between the last and second to the last bar. This inconsistency in the years may convey the wrong trends if the household type share for the missing years is considerably different from the depicted trend.
- Some bars do not add up to a perfect 100. As the graph is about the percentage share of each household type for each year, it is necessary that individual shares of each household type for a year add up to a 100%. For the years, 1980 and 1995, the total adds up to 99.9% and for the years 1990,2005 and 2012 it adds up to a 100.1%
Critical Analysis of the visualization:
- Beautiful: The visualization is clear and easy to understand. But I believe the use of stacked bar graph is not appropriate for this particular visualization. The aim of the visualization is to portray the changing trends in household types over the years. We know and Visualization Best Practices suggest that, line charts track changes or trends over time and show relationship between two or more variables. Thus, a line graph, with each household type depicted separately and differentiated by color would give a much clear view of the changing trends over the years.
- Enlightening: According to me, the visualization by itself is not very enlightening. An enlightening visualization is one which initiates a change in the audience. This visualization on household types is definitely informative. It gives us an idea of the changing trends over the years. But it does not make the audience take any specific action. Are there any relevant impacts due to changing household trends? This is not clear and hence there is no potential changes that one can take based on this information.
- I am also not sure as to why the start year is 1970, the visualization nor the article tries to explain the significance for the chosen time period. As we discussed in class on the validation of visualization, people can cherry pick the data to make the data look the way you want. Hence it is important that there is no question raised on the validity of data. May be if a longer period was chosen would have made the changing trends look different than what it shows now. There should be no question on the validity of data.
Redesign:
As I discussed, stacked bar graph is not the most ideal graph to design time changing trends. The use of a line graph would be a better choice to design the graph. The redesigned graph can be viewed at:
References:
1)Choosing the right visualization for your purpose:
https://www.gooddata.com/blog/5-data-visualization-best-practices
2)Scaling an axis properly:
https://blog.graphiq.com/data-visualization-best-practices-91a35f1b29fa
3) When are 100% Stacked Bar Graphs useful: