We rely on many forms of data visualization in our daily lives, often without even realizing it. From weather maps to public transportation guides – they play a critical role in visually communicating data. This was the premise of the project undertaken by my teammate, Karla Cruz Sanchez ’26, and myself, Rachel Kim ’27, wherein we constructed graphs for Dr. Annette Nierobisz’s (from our own Sociology Department!) research article on the long term impacts of late-career job loss.
Our dataset consisted of survey responses from 25 individuals, mostly in their 60s and 70s, who were part of an original group of 60 interviewees recruited a decade prior. After thoroughly reviewing the article, Karla and I identified our focus: visualizing the long-term effects of late-career job loss, particularly among genders and different financial landing situations of the respondents. These effects were identified in the first round of interviews in 2013/2014, which looked at soft vs downward mobility. Here, the “soft” group refers to those who had sufficient financial resources to turn to following their job loss such as their spouses or parents, whereas the “downward mobility group” refers to those who suffered more financial hardship.
<Click images for full-size versions>
Using the programming language R, Karla tackled emotional responses (“As a result of my initial job loss, I am now experiencing: ______”) while I focused on satisfaction levels (“In retrospect, I am satisfied with how my life turned out after experiencing job loss”) Initially, I planned to create three graphs: satisfaction levels by gender, satisfaction levels by group type, and current employment statuses by group type. As I was dealing with categorical variables, I initially considered pie charts, but quickly learned that they are quite often considered as ineffective visualizations due to their inaccurate and often distorted representation of quantitative data using pie segments rather than height or length like line and bar charts.
Following a meeting with Paula Lackie, Academic Technologist for Data, we decided to consolidate the information into a comprehensive table, offering a clear snapshot of the key findings from Dr. Nierobisz’s article in one swoop. Here, we actually found that differing the relative sizes of the pie charts would be helpful in showing the different number of responses for each category. I was able to come close but encountered some technical difficulties with the pie charts that stubbornly wanted to be donut charts, which was a bit frustrating but nonetheless a great learning experience. Subsequently, after further discussion with Dr. Nierobisz, we opted to revert to our original plan of breaking up our graphs into separate ones for better audience comprehension. Not to mention, we also got to apply some statistical analysis techniques to perform a Welch’s t-test in order to compare retirement ages between individuals with different financial landing situations, which yielded statistically significant results!
This project provided me with invaluable experience in data visualization and analysis, allowing me to refine my R skills and explore new tools like GitHub for collaborative coding. Not to mention, I also learned a great deal about late-career job loss and its critical socioeconomic effects. Apart from being the first “real” task that I got to tackle as a new Data Squad member, I also really came to understand the importance of inclusivity in the graph colors we choose. This is an area where I definitely think I have to improve a lot, as achieving a good balance between hue, saturation and value for a visually appealing and harmonious color palette was actually one of the most challenging aspects for me in this project. Furthermore, I learned that it is not a good approach to represent aspects such as gender solely through colors, and alternative approaches such as using clear labeling or other visual cues such as icons or symbols are better to use. However, I still found various tools such as this Coblis simulator helpful in making sure my color choices were inclusive. For more information for your own data visualizations, try reading this article!