An analysis of changes in CTA ridership between 2019 and 2022
This project looks at which Chicago Transit Authority (CTA) train stations recovered the most, and which recovered the least, of their pre-pandemic ridership levels. Because ridership fluctuates seasonally within each year, I focused my analysis on change in ridership between the last full year before the pandemic (2019) and the first full year after the peak of the pandemic (2022).
I also set the stage for future demographic analysis by aggregating ridership stats for the 143 stations up to Chicago's 77 community areas.
Here's my story, Who's Still Riding the CTA, as self-published in my portfolio.
My key findings are as follows:
- Citywide, 2022 train ridership was at 49% of pre-pandemic (2019) levels
- Stations ranged from 33% to 71% ridership retention
- The five stations with the best recovery (ranging from 66% to 71%) were all Pink Line stops serving South Lawndale and the Lower West Side
- The station with the weakest recovery is the suburban Oak Park Blue Line station, at 33% retention
- The next four weakest stations were all in the Loop, Chicago's downtown area
Interestingly, I have not seen the Pink Line trend reported in Chicago. Prior reporting by the Tribune and the Urban Institute highlighted the Pulaski and Garfield Conservatory Green Line stops as recovering the best by December 2021 by that time.
| Data Source | Description |
|---|---|
| CTA - Ridership - 'L' Station Entries - Daily Totals | Daily count of # of ridership on CTA by station of entry |
| CTA - System Information - List of 'L' Stops | Additional station info including GPS coordinates and community area IDs |
| CTA station additional info | Manually-entered info on city and train line color |
| Chicago Community Area Names | ArcGIS Community Area IDs and Names |
| Chicago Community Geographic Boundaries | shapefile I used for mapping train stations in Flourish within Chicago community areas |
My data analysis process required the following steps:
- Acquired ridership and station data via API, aggregated through Socrata's query engine. Socrata is an API platform used by the City of Chicago and many other cities. Its query language helped me group and sum results while extracting data, minimizing the amount of data transfered.
- Cleaned up data by removing a previously closed station, re-organizing data structures to get Lat and Lng coordinates as separate columns, dropping duplicate rows representing different entries to the same station
- Merged supplemental info about stations (lat/lng, station line color, city) into station data
- Created an aggregate community-level summary, merging in community names
- Calculated station-level and community-level summary metrics- primarily, percent ridership retention
% retention = ((2022 ridership) - (2019 ridership)) / (2019 ridership)
- Prepared graphics in DataWrapper and Flourish. I found the hex color codes in Chicago's train wikipedia page to be helpful in getting the precise colors.
In addition to some problems mentioned above, I addressed the following problems:
- Removed 4 red line stations from station-level (but not community level) summaries, because two stations closed for renovation in May 2021 while adjacent stations picked up their ridership
- Annoyingly, community area IDs in CTA data did not exactly match IDs in the community area boundary file. I had to match by community area name
- Using a point map, I found it difficult to visually discern the difference between 35% ridership retention and 70% retention. It's worth weighing whether mathematical precision or perception is more important in portraying differences. Though I ultimately opted against "perceptual scaling," here's an interesting explainer on the ethics of perceptual scaling.
Though I improved this repo in October 2023, I started this project in June as a Columbia University portfolio project. I credit this work shifting my habits away from Google Sheets and towards Python/Pandas within Jupyter Notebook. I'm convinced the latter approach is much more effective in integrating documentation within code, creating a reproducible and automated template, and making it easy for me to retrace my steps.
In June, I also learned:
- Socrata's API platform, commonly used in city data portals
- GitHub, a means of organizing and sharing my code
- Scoping a project for timely completion. I really wanted to delve into demographic patterns corresponding to ridership, but that will have to wait for a Part II.
- Layer demographic data about the community areas.
- Size images to look good across platforms, especially desktop vs. mobile. The default iframes in Flourish work reasonably well compared to anything else I've tried, but it's worth more development.
- Sandhya (LA Times) challenged me to provide geographical orientation- the lake, streets, etc. I could do this in DataWrapper using a basemap, but the tradeoff is that I'd be limited in the # of data points (stations) I could portray. Worth exploring more, perhaps for a neighborhood-specific zoom-in.
Most data for this project is collected directly in Python via API. You'll also find in this folder: results includes exported data analysis
- source_data- includes only my own manually-entered lookup table for CTA stations
- Jupyter Notebook- steps through the analysis
- results- results output from Jupyter Notebook for mapping/charting in DataWrapper and Flourish