Republicans Refuse Medicaid for Poor: 5 Data Viz Lessons

The New York Times has an insightful article on the decision by states with Republican Governors to refuse the expansion of Medicaid, the medical insurance program for the poor. The Federal government will pay for this 100% through 2016 and no less than 90% of the cost in later years. This decision impacts eight million extremely needy Americans whose income limit is just under half of the federal poverty level - about $5,600 a year for an individual.

The NYT team has created an insightful visualization to go with the article. I felt it would be a great example to use to remind ourselves of five critical data visualization lessons we can apply in our jobs as we present data.

1. Build Clear Segmentation Options.

I've said that all data in aggregate is crap.

We can focus on all states (aggregate), or just as easily only the states refusing medicaid for the poor or those that are accepting it. Even at the highest level it lets you see the string of states in the south, Texas, Florida as an example, that are making this choice.

When you don't have segmentation, it is hard for your decision makers to quickly pick the first focusing area.

2. Build Micro Drill-Down Capability, with Annotations.

Once the first focusing area has been picked, you want your visual to include the capability to allow the audience to drill-down deep into your data-set.

In this case the zoom-in feature allows us to drill into Florida and hover our mouse on the reddest areas and identify exactly how many underprivileged people will be adversely impacted.

3. Build Multiple Data-Visualization Options.

People are picky, bosses are picky squared. Hence one of the hardest things to do is to pick the right visual layer that will appeal to everyone. We also face the same challenge with how to show the data.

The NYT team has included two small boxes on the top right that allows us to pick the data type (share or number) and switch the visualization type. This version allows us to hone in even more sharply on states we are interested in and see the number of eligible adults who will be left uninsured. Very nice.

4. Build Less Complexity, Yet Include Contextual Data.

There is a tendency amongst even the best analysts to try and be too clever, to go "omg, would it not be cool if I also..." Resist that tendency.

Yes, there is always data that could provide key context, one more little bit of data, something cool you found. Share it, just don't make it a part of your core visualization. In our case study today it was which racial types will suffer the most by this refusal. It is shown just below.

It would have been easy to add drop-downs in the map that show race, or to try to stuff in some clever color/shading thingy. The NYT made a very wise choice to have the data included as a simple graph below the main visualization.

Also notice this is not a filled out bar chart or some kind of weird stacked bar charts (standard excel stuff). Simple side-by-side bars work very well, they show that the impact of this decision by Republican governors will disproportionately harm African Americans.

5. Don't Let Data Speak For Itself!

This might seem super odd. We have all the data there, people are intelligent, they will be able to infer the most important point the data is going to make.

This is a gigantic flaw. Never assume unlimited time or intelligence at the other end.

You have spent the most amount of time with the data. You've done all the analysis. You've created a masterful visualization. Write down the biggest, most important point you've discovered.

Here is how simple and direct it can be: "A sweeping national effort to extend health coverage to millions of Americans will leave out two-thirds of the poor blacks and single mothers and more than half of the low-wage workers who do not have insurance, the very kinds of people that the program was intended to help."

Five simple lessons that, if you follow them, will awesomize your next data visualization.