Todd W. Schneider

Chicago’s Public Taxi Data

Trends across neighborhoods, the Cubs World Series, and a good-faith effort to protect privacy

The City of Chicago has released a public dataset containing over 100 million taxi rides since 2013. I adapted my analysis of the similar New York City dataset to process the Chicago data, and created a GitHub repository with the relevant code.

The Chicago dataset does not include data from ridesharing companies like Uber and Lyft, but the data makes clear that taxi usage in Chicago has declined dramatically since 2014. As of November 2016, Chicago taxi usage was declining at a 35% annual rate, and had fallen a cumulative 55% since peaking in June 2014.

monthly chicago taxi trips

Again, the public dataset does not include any data from ridesharing services like Uber and Lyft, but the Chicago taxi industry claims that ridesharing services caused cabs to lose 30–40% of their business in the summer of 2015.

Chicago’s taxi industry is shrinking faster than NYC’s

New York taxis have also been losing market share to ridesharing companies—NYC releases data that confirms this—but in fact Chicago taxis are losing market share even faster than their NYC counterparts. While NYC taxi usage has been declining at around 10% per year, Chicago’s declines have reached 35% year-over-year.

chicago vs nyc taxi growth

New York taxis make about 8 times more trips per month than Chicago taxis do, but a rescaled monthly trips index shows that Chicago has a larger cumulative decline on a percentage basis.

Areas closest to downtown show smaller taxi declines

Chicago’s taxi pickup declines are not evenly distributed among the city’s 77 community areas. For example, the Loop, Chicago’s central business district, shows a 23% annual decline, while Logan Square on the northwest side shows a 50% annual decline. In general, the areas located closest to the central business district show smaller declines in taxi activity.

I defined 5 particular community areas—the Loop, Near North Side, Near West Side, Near South Side, and O’Hare Airport—as the “core”, then compared pickups inside and outside of the core. As of November 2016, pickups inside the core show a 27% annual decline compared to a 42% annual decline outside of the core. On a cumulative basis, core pickups have declined 39% since June 2014, while non-core pickups have declined a whopping 65%. The smaller taxi decline near the central business district is consistent with NYC’s taxi and Uber data, where taxi share has fallen less in Manhattan than in the outer boroughs.

Data by community area is available here in spreadsheet form.


A map of the official community area definitions is available here, and you can select community areas in the menu below to view taxi pickups since 2013.

community areas

Anonymized medallion numbers

Chicago’s public taxi data, unlike New York’s, includes anonymized taxi medallion numbers for each trip. This makes it possible to do things like:

  • Count the number of unique taxis in service each month
  • Measure the distribution of trips per day for active taxis
  • Observe the sequence of trips made by individual taxis

The Chicago dataset is also missing some of the details provided by New York, though this is explicitly for the purpose of privacy, and is probably on the whole a good thing.

The number of taxis that make at least one pickup per month has declined nearly 30%, from a peak of over 5,000 to 3,600 more recently.

taxis in use

Since taxi trips have declined by 55% over a time period when unique taxis have declined by 29%, that means fewer trips per day for each active taxi. Active taxis used to average 20 trips per day, but more recently have averaged 13 trips per day.

trips per taxi per day

A histogram of daily trips per taxi shows a bit of a right skew, with a mean of 18 and median of 16 trips per day over the entire dataset. On the plus side for taxis, average fares have increased over time, at least partially due to a 15% fare increase in early 2016, and so the decline in total fares collected per taxi per day is not as large.

The best and worst places for a taxi to make a drop off

With anonymized medallion numbers, we can see when and where a taxi picked up its next fare after making a drop off. For each drop off, I looked at the time of the next pickup, and calculated the percentage of drop offs in each area that were followed by a new pickup within 30 minutes. For privacy reasons, trip timestamps are all rounded to 15-minute intervals, so this calculation is not exact, but it should be close enough.

Sure enough, nearly 80% of drop offs in central business districts are followed by a pickup within 30 minutes, while as little as 20% of drop offs in more remote areas, e.g. airports, are followed by pickups within 30 minutes.

Likelihood of a taxi finding a new fare within 30 minutes by drop off area

This basic analysis doesn’t necessarily imply that it’s a bad thing for a taxi to make a trip from the Loop to O’Hare. It’s true that it’s less likely for a taxi to get a new fare after dropping off at the airport, but a more thorough analysis would have to take into account that fares to the airport are higher than average, and so the question becomes whether that higher fare is enough to offset the longer wait time after drop off. Time of day and day of week might also be relevant, and should be considered in a more complete analysis.

Wrigley Field and the 2016 Cubs

I’m not a native Chicagoan, but you don’t have to be one to know that the Cubs winning the 2016 World Series was a big deal. I grabbed the 2013–2016 Cubs home game schedules from Baseball Reference and compared taxi drop offs near Wrigley Field on game days to non-game days.

wrigley field

wrigley field drop offs

Not surprisingly, taxis do more business around Wrigley Field on game days. Total drop offs have declined since 2013—remember taxis have lost market share everywhere—but more interesting is to look at the patterns within each season. In particular the 2016 championship team generated the most taxi activity during the World Series games in October, when in previous seasons peak taxi activity had been during the mid-summer months.

Privacy measures

Chicago’s dataset is missing some of the details provided by New York, most notably:

  • Precise timestamps
  • Precise latitude/longitude coordinates

All timestamps are rounded to the nearest 15-minute interval, and instead of latitude/longitude, the data includes census tract and community area identifiers. Furthermore, census tracts are only included when there are multiple trips within the same tract over the same 15-minute interval.

The press release announcing the dataset’s publication specifically points out that these measures were taken to protect privacy, presumably of both drivers and riders. I think on the whole it’s a good thing, even if it means that there won’t be any fancy maps of the Chicago trips, frankly that’s a small price to pay.

Still, anonymizing data is a very hard problem, and it seems like the Chicago dataset has not completely eliminated the risk. If we define a “uniquely identifiable” trip as one where there was exactly one pickup or drop off in a community area over the course of an hour, then 66% of all taxis in the dataset made at least one uniquely identifiable trip.

That means, for example, if you got into a taxi in some area at some time, recorded its medallion number, then later checked the data and there was only one pick up in that area during that hour, then you could map that particular “anonymized” medallion number to the actual medallion number. It might be impractical to find the real medallion numbers for these uniquely identifiable trips—you wouldn’t know the trip was uniquely identifiable until well after the fact—but with the proliferation of cameras and computer vision technology, it’s not that far-fetched either.

Even though only 0.7% of the trips in the dataset are uniquely identifiable by my definition, taxis that made at least one uniquely identifiable trip account for nearly 98% of the total trips. Again, this isn’t to say that I or anyone else has managed to de-anonymize the data, but it’s a reminder that even when good-faith efforts are made to anonymize data, it’s extremely difficult to do it well.

Uber and New York are currently fighting over data disclosure, with the city asking for more data from Uber for planning and regulatory purposes, and Uber refusing to provide it because NYC has done a bad job protecting privacy in the past. Chicago’s privacy measures are not perfect: there might still be ways to de-anonymize the data, and just the fact that they have more detailed data means there’s a risk of accidental or malicious release. But in my mind the Chicago data strikes an appropriate balance, on the one hand enabling analysis that could lead to real insights and quality of life improvements, while simultaneously protecting the privacy of those involved. New York could do worse than adopt a similar approach.

Code on GitHub

All code used in this post is available on GitHub.

“If they can dye the river green today, why can’t they dye it blue the other 364 days of the year?”

It turns out that the annual St. Patrick’s Day Parade, made famous (at least in my adolescent mind) by The Fugitive, is the day with the most taxi trips in Chicago every year since 2013. Per IMDb, director and Chicago native Andrew Davis specifically wanted to capture the parade, though part of me now thinks that Dr. Richard Kimble should have ducked out by way of taxi…

daily trips

The Simpsons by the Data

Analysis of 27 seasons of Simpsons data reveals the show’s most significant side characters, a pattern of patriarchy, declining TV ratings, and more

The Simpsons needs no introduction. At 27 seasons and counting, it’s the longest-running scripted series in the history of American primetime television.

The show’s longevity, and the fact that it’s animated, provides a vast and relatively unchanging universe of characters to study. It’s easier for an animated show to scale to hundreds of recurring characters; without live-action actors to grow old or move on to other projects, the denizens of Springfield remain mostly unchanged from year to year.

As a fan of the show, I present a few short analyses about Springfield, from the show’s dialogue to its TV ratings. All code used for this post is available on GitHub.

The Simpsons characters who have spoken the most words

Simpsons World provides a delightful trove of content for fans. In addition to streaming every episode, the site includes episode guides, scripts, and audio commentary. I wrote code to parse the available episode scripts and attribute every word of dialogue to a character, then ranked the characters by number of words spoken in the history of the show.

The top four are, not surprisingly, the Simpson nuclear family.

If you want to quiz yourself, pause here and try to name the next 5 biggest characters in order before looking at the answers…


Of course Homer ranks first: he’s the undisputed most iconic character, and he accounts for 21% of the show’s 1.3 million words spoken through season 26. Marge, Bart, and Lisa—in that order—combine for another 26%, giving the Simpson family a 47% share of the show’s dialogue.

If we exclude the Simpson nuclear family and focus on the top 50 supporting characters, the results become a bit less predictable, if not exactly surprising.

supporting cast

Mr. Burns speaks the most words among supporting cast members, followed by Moe, Principal Skinner, Ned Flanders, and Krusty rounding out the top 5.

Gender imbalance on The Simpsons

The colors of the bars in the above graphs represent gender: blue for male characters, red for female. If we look at the supporting cast, the 14 most prominent characters are all male before we get to the first woman, Mrs. Krabappel, and only 5 of the top 50 supporting cast members are women.

Women account for 25% of the dialogue on The Simpsons, including Marge and Lisa, two of the show’s main characters. If we remove the Simpson nuclear family, things look even more lopsided: women account for less than 10% of the supporting cast’s dialogue.

A look at the show’s list of writers reveals that 9 of the top 10 writers are male. I did not collect data on which writers wrote which episodes, but it would make for an interesting follow-up to see if the episodes written by women have a more equal distribution of dialogue between male and female characters.

Eye on Springfield

The scripts also include each scene’s setting, which I used to compute the locations with the most dialogue.


The location data is a bit messy to work with—should “Simpson Living Room” really be treated differently than “Simpson Home”—but nevertheless it paints a picture of where people spend time in Springfield: at home, school, work, and the local bar.

The Bart-to-Homer transition?

Per Wikipedia:

While later seasons would focus on Homer, Bart was the lead character in most of the first three seasons

I’ve heard this argument before, that the show was originally about Bart before switching its focus to Homer, but the actual scripts only seem to partially support it.


Bart accounted for a significantly larger share of the show’s dialogue in season 1 than in any future season, but Homer’s share has always been higher than Bart’s. Dialogue share might not tell the whole story about a character’s prominence, but the fact is that Homer has always been the most talkative character on the show.

The Simpsons TV ratings are in decline

Historical Nielsen ratings data is hard to come by, so I relied on Wikipedia for Simpsons episode-level television viewership data.


Viewership appears to jump in 2000, between seasons 11 and 12, but closer inspection reveals that’s when the Wikipedia data switches from reporting households to individuals. I don’t know the reason for the switch—it might have something to do with Nielsen’s measurement or reporting—but without any other data sources it’s difficult to confirm.

Aside from that bump, which is most likely a data artifact, not a real trend, it’s clear that the show’s ratings are trending lower. The early seasons averaged over 20 million viewers per episode, including Bart Gets an “F”, the first episode of season 2, which is still the most-watched episode in the show’s history with an estimated 33.6 million viewers. The more recent seasons have averaged less than 5 million viewers per episode, more than an 80% decline since the show’s beginnings.



TV ratings have declined everywhere, not just on The Simpsons

Although the ratings data looks bad for The Simpsons, it doesn’t tell the whole story: TV ratings for individual shows have been broadly declining for over 60 years.

When The Simpsons came out in 1989, the highest 30 rated shows on TV averaged a 17.7 Nielsen rating, meaning that 17.7% of television-equipped households tuned in to the average top 30 show. In 2014–15, the highest 30 rated shows managed an 8.7 average rating, a decline of 50% over that 25 year span.

If we go all the way back to the 1951, the top 30 shows averaged a 38.2 rating, which is more than triple the single highest-rated program of 2014–15 (NBC’s Sunday Night Football, which averaged a 12.3 rating).


Full data for the top 30 shows by season is available here on GitHub

I have no proof for the cause of this decline in the average Nielsen rating of a top 30 show, but intuitively it must be related to the proliferation of channels. TV viewers in the 1950s had a small handful of channels to choose from, while modern viewers have hundreds if not thousands of choices, not to mention streaming options, which present their own ratings measurement challenges.



We could normalize Simpsons episode ratings by the declining top 30 curve to adjust for the fact that it’s more difficult for any one show to capture as large a share of the TV audience over time. But as mentioned earlier, the normalization would only account for about a 50% decline in ratings since 1989, while The Simpsons ratings have declined more like 80-85% over that horizon.

Alas, I must confess, I stopped watching the show around season 12, and Simpsons World’s episode view counts suggest that modern streaming viewers are more interested in the early seasons too, so it could just be that people are losing interest.

As I write this, The Simpsons is under contract to be produced for one more season, though it’s entirely possible it will be renewed. But ultimately Troy McClure said it best at the conclusion of the The Simpsons 138th Episode Spectacular, which, it’s hard to believe, now covers less than 25% of the show’s history:

troy mcclure


Automated episode summaries using tf–idf

Term frequency–inverse document frequency is a popular technique used to determine which words are most significant to a document that is itself part of a larger corpus. In our case, the documents are individual episode scripts, and the corpus is the collection of all scripts.

The idea behind tf–idf is to find words or phrases that occur frequently within a single document, but rarely within the overall corpus. To use a specific example from The Simpsons, the phrase “dental plan” appears 19 times in Last Exit to Springfield, but only once throughout the rest of the show, and sure enough the tf–idf algorithm identifies “dental plan” as the most relevant phrase from that episode.

I used R’s tidytext package to pull out the single word or phrase with the highest tf–idf rank for each episode; here’s the relevant section of code.

The results are pretty good, and should be at least slightly entertaining to fans of the show. Beyond “dental plan”, there are fan-favorites including “kwyjibo”, “down the well”, “monorail”, “I didn’t do it”, and “Dr. Zaius”, though to be fair, there are also some less iconic results.

You can see the full list of episodes and “most relevant phrases” here.

episode summaries

Another interesting follow-up could be to use more sophisticated techniques to write more complete episode summaries based on the scripts, but I was pleasantly surprised by the relevance of the comparatively simple tf–idf approach.



Code on GitHub

All code used in this post is available on GitHub, and the screencaps come from the amazing Frinkiac

Taxi, Uber, and Lyft Usage in New York City

Open TLC data reveals the taxi industry’s contraction, Uber’s growth, and the scramble for market share

The New York City Taxi & Limousine Commission publishes summary reports that include aggregate statistics about taxi, Uber, and Lyft usage. These are in addition to the trip-level data that I wrote about previously; although the summary reports contain much less detail, they’re updated more frequently, which provides a more current glimpse into the state of the cutthroat NYC taxi market.

I’ve updated the nyc-taxi-data GitHub repository with code to fetch and process the summary reports, and you can return here for updates in the future: the graphs on this page will update every month as the TLC releases more data.

Trips Per Day in NYC: Taxi vs. Uber vs. Lyft

The summary data includes the number of trips taken by taxis and for-hire vehicles:

This graph will continue to update as the TLC releases additional data, but at the time I wrote this in April 2016, the most recent data shows yellow taxis provided 60,000 fewer trips per day in January 2016 compared to one year earlier, while Uber provided 70,000 more trips per day over the same time horizon.

Although the Uber data only begins in 2015, if we zoom out to 2010, it’s even more apparent that yellow taxis are losing market share.

Total Vehicles on the Road

The summary reports also include the total number of vehicles dispatched by each service:

Again this graph will update in the future when more data is available, but as of January 2016 there are just over 13,000 yellow taxis in New York, a number that is strictly regulated by the taxi medallion system. Green boro taxis account for another 6,000 vehicles. Uber has grown from 12,000 vehicles dispatched per month at the beginning of 2015 to 30,000 in January 2016, while Lyft accounts for another 10,000.

However, the Uber/Lyft numbers might not be as dramatic as they seem: the TLC’s data does not indicate how many days per week Uber/Lyft vehicles work, only the total number of trips per month and the total number of vehicles that made at least one trip.

A study by Jonathan Hall and Alan Krueger reported that 42% of UberX drivers in New York work fewer than 15 hours per week, while another 35% work 16–34 hours per week. If those numbers are true, then a very rough guess might be that about half of those 25,000 vehicles make at least one pickup on any given day.

Yellow taxi utilization rates are much higher: the TLC statistics report that the average medallion is active 29 days per month, 14 hours per day (note that multiple drivers can share a medallion).

The controversial question is whether the influx of Uber, Lyft, and other for-hire vehicles has worsened congestion problems in NYC. I’ll stay out of that kerfuffle for now, but at least the popular narrative is that the city’s study did not blame Uber for increased congestion in Manhattan.

It would be interesting to look at the trip-level taxi data to see if taxi rides from point A to point B have gotten slower over the years in various parts of the city. But even if they have, it would be difficult if not impossible to blame it on for-hire vehicles—or any other single factor—using only the trip-level taxi data.

Lyft, Via, Juno, and Gett

Lyft is probably the most well-known Uber competitor, but there are others. Via, Juno, and Gett are among the newer ridesharing services to operate in NYC, and they report data to the TLC too.

Update 4/26/16: apparently there was a data reporting error between Lyft and the TLC in January 2016, which has now been corrected. When I originally wrote this post, the Lyft graph looked like this. Based on the revised data, it does not appear that Lyft usage declined in early 2016.

Uber’s Revenue in NYC

Uber’s revenue numbers are not publicly disclosed, but we can piece together different bits of information to arrive at a very rough estimate for Uber’s New York revenue in 2015:

  • The TLC data reports 36.3 million Uber trips in NYC in 2015
  • Uber published average NYC UberX fares for 2012–2014. The average fare was $27 in September 2014, but fares have been decreasing since 2012. Let’s guess a $25 average fare for Uber’s NYC trips in 2015, including UberX and UberBlack
    • By comparison, the average yellow taxi fare was a bit over $14 in 2015
  • Uber takes a 20–25% commission, call it 22% on average

That gives us (36.3 * $25 * 0.22) = $200 million estimated revenue for Uber in NYC in 2015.

2016 Outlook

UberX’s recent NYC fare cut will probably increase demand for rides while lowering the average fare. Simultaneously Uber might charge higher commissions, and who knows how surge pricing trends might evolve. I doubt we’ll see too many public data points surrounding revenue, but maybe there will be enough to continue the “rough estimate” game.

It will be interesting to see what happens in 2016. Like many New Yorkers, I’ll be curious to see if Uber continues to gain market share, if yellow taxis do anything to stanch their wounds, and if Lyft—or any other newcomers—can muscle their way into the ranks of the major players.