Blog: Digging Into Data Sources

by Jenna Laib

“Data” is not neutral.

When we look at a data visualization, it can feel cool and objective — maybe even authoritative. But every graph, chart, or infographic has been shaped by a series of human decisions. Someone decided what to measure and how to measure it. Someone then decided what data to include (or exclude), how to scale the axes, and what labels to use. These choices influence how we interpret the information and what conclusions we draw. Understanding that data is constructed — not just collected — is a critical part of data literacy.

When I first started posting slow reveal graphs to this site, I made certain to always include a link to the “source” for the graph. This felt like due diligence, as both a way to give credit and an entry point into digging deeper. A few years later, I decided that this was not sufficient. Data visualizations, even the most clever and well-intentioned, are the product of decisions about what to include and exclude, what definitions to use, and what assumptions to make. Often, we agree with these decisions, and we derive meaning from the way that the data has been represented. However, there are times when knowing about those decisions colors our interpretation.

So I started linking directly to the data source, when I could find it. (And I am increasingly hesitant to post graphs where I can’t find the data source.) I add questions to the slide deck that encourage students to critique the source of the data. That doesn’t mean that we inherently disagree with the data, but that we use a critical lens to understand it.

Data Source for “Days Between Billion-Dollar Natural Disasters in the US”

Yesterday, I posted a slow reveal graph called Days Between Billion-Dollar Natural Disasters in the US. I had seen the graph posted on social media, and then dug around a little on the internet to find the original source of the data visualization. (Turns out it was on a website called ‘Climate Central.’) The original data visualization had a lot of intense, dark colors, and so I decided to modify it for classroom use. I spent time removing the background, and darkening the bars along the axes. I changed the color of the text to make it more readable, and re-typed portions to increase visibility. Then, finally, I designed the slow reveal and the questions.

There were some questions I had, too, about the data. Usually graphs mention that whether the dollar amounts have been adjusted for inflation, or CPI-adjusted (using the consumer price index as a benchmark). This graph did not have any information, which led me to assume it wasn’t. But this was merely an assumption.

I looked at the clock. It was almost time to pick up my kids from camp, and so I started to rush. I put on an additional slide with a comparison of the dollar’s value. Roughly $250m in 1980 would have a similar purchasing power to $1b in 2025. Without looking up the data source — I listed it as NOAA/NCEI, as on the original graph — I posted it.

ARE those dollar amounts adjusted for inflation?

Scott Farrar had the exact same question I had. We exchanged a few messages via Bluesky — it felt like old Math Ed twitter in the best way! — and then embarked on a simultaneous google search. We both discovered pages on the NCEI website about billion-dollar events from natural disasters.

…and the pages were data dense!

The NOAA/NCEI page that Scott found listed out events individually, offering information about the type of event, the start and end dates, the CPI-adjusted estimated cost, and the number of deaths.

There are so many ways to visualize this data. The graph that I had been working with was all about the average number of days between billion dollar events in a given year. This can be calculated using the start and end dates, and offers a new insight into the frequency of billion-dollar events.

Because this data uses the date for each event, it is also possible to break the data down in different ways, e.g. calculating the number of events per month, or the total spent in a given year, by month, like the graph to the right. There is consistently a large increase in spending around July/August, which suggests that hurricanes are a major driving cost of natural disaster relief and repair.

Oh, and look at that: the data is CPI-adjusted. I am so glad that Scott pushed me to consider the source, once again.

Digging Into Data Sources

Looking at the original source of the data answered the question that Scott and I had about whether the figures were adjusted for inflation using the CPI. It can also prompt new questions, and new ways to look at the data. What other questions do we now have?

• Are events getting more expensive over time?
• Are there certain types of events that cause more damage?
• Has the human toll and number of deaths changed over time?
• Are certain parts of the US impacted more heavily?

etc.

These questions may have arisen by looking at the original graph. Maybe students would want to conduct the research to determine answers to those questions. Knowing the source of the data, not just the visualization, gives an entry point into that research.

Good data work is driven by good questions.

By bringing attention to the original data source, I encourage you to trace and critique the decisions behind a data visualization. This is how we become more than consumers of information. If we engage in and model these habits, we can also teach it to the next generation of data literate citizens of the world.