Visualizing the 2018 Kerala Flood Donations
The recent Kerala floods caused massive destruction that left its unprepared citizens in a state of despair and chaos. This was the worst flood after the Great Flood of 99 that occurred in 1924, with the Indian Government raising a level 3 calamity. Several individuals, private organizations, government bodies came to Kerala’s help by organizing donation drives, making private contributions and mobilizing resources for those in need.
As this happened to be around the same time as our data visualization class, we were asked by our professor to create an interactive data visualization on the same. This seemed like an interesting opportunity since there were many geographical, climatic, social, political and financial facets to look at. There were also prevalent speculations that we could use data to probe into. I happened to come across the Kerala Chief Minister’s Disaster Relief Fund (CMDRF) website which listed all the donations made to the Chief Minister’s Relief Fund, the sources for the same and how they were spent.
There were also a lot of circulating news articles on the contributions made by some famous entities. There seemed to be some interesting patterns in the data with the portal having collected around 1027 crores then, more than the 600 crores that the state was promised by the Prime Minister.
Hence I decided to base my visualization on the donations made by various entities. I could use them to further answer what are the possible factors that have an effect on people’s giving? Is it the level of disaster, media coverage, tax benefits, transparency, ease of payment, time of the year etc. that make people give?
With some of these in mind, I started with the data collection
Data Collection
Very soon I realized that the CMDRF does not have a collated list of all the donations and its sources, they only publish the list of donations made for that day (that is, of course, apart from the special excel sheet snapshot posted by the CM on his Facebook page for the first 200 donations).
The data was however available, scattered across the internet in the form of online articles, social media posts and Wikipedia listings. I made a schema of all the data types I needed in sheets and began on a rampage to find them on the world wide web.
My final sheet had the following datatypes:
Entities: Entities list the main sources who contributed.
Categorization: Categories were the broader groups that entities fell under. After a lot of organizing and reorganizing, I limited my category to 9 types. I further made subcategories to make sense of the data. Categorization was an important part of my visualization and I had to be careful with how I made them keeping my narrative in mind. For example, donations made by Asianet employees fell under ‘category: People, subcategory: Fundraising’ and not TV networks, whereas that made by the company itself fell under ‘category: Entertainment, subcategory: TV network’.
Amount: List the amount of donation in crore rupees.
Donations made to: I also was coming across data on where these donations were given, which I thought could add an interesting dimension to the analysis. I decided to add the same. Apart from CMDRF where the largest volume of donations was made, there were other several disparate sources. I clubbed some of them together into ‘NGOs and other organizations’ and ‘Relief Support’ to further clean the data.
Apart from the extensive human effort and time spent in data collection, I faced some other issues. Here a list of them and what I did to resolve them:
1. Inconsistency in the data: Sourcing some data from online articles invited a lot of inconsistencies along with it. Some articles stated different donation amounts for the same entity. In that case, I had to rely on the most trusted source for the data.
2. There was excessive media coverage on the donations made by famous celebrities and not so much for others. Unfortunately, I could not do much to help this, except try and collect as many data points as possible. I felt it was important to include a note in my visualization about the same.
3. To estimate the amount of donation contributed by common people, I subtracted the amount contributed by other entities from the total amount collected.
This is what my final data set looked like with 154 data points.
Data Visualization and encoding
I wanted my data visualization to be a medium that people can use to make their own interpretations. I decided to have a main visualization for the entities and donations and, smaller visualizations around the categories and where the donations went to. Tableau seemed like a good fit for my data visualization and was simple enough to learn, hence I used the same.
Donations by each entity: For my main visualization, I decided to go show all the entities using packed bubbles in Tableau, with colours, representing the category they belong to and size representing the amount contributed by them
Category wise donations
Category wise donations show the amount contributed by each category using a pie chart. The encoding is similar to the main visualization with colours representing categories and size of the pie representing the amount
Donations given to
I used a simple bar chart to plot the sources and the amounts they received. I had earlier used stacked bars with colours representing categories. However, while making the bar act a filter in the Tableau dashboard, it would filter all entities in that category and not those who contributed to that source. Hence due to this tableau bug, I decided to drop the colours and use grey for all the sources with the height representing the amount.
Interactivity
Tableau offers an excellent means to pull different aspects of your data together in the form of a dashboard. I put together all my visualizations together in the dashboard and further used actions to turn them into filters.
I added a ‘filter by amount’ and ‘search for entity’ filters and applied them across worksheets to make the visualization more interactive. This is what the final dashboard looks like. You can view the interactive version here. It is recommended to view it in full-screen mode.
