This Week In Public Transit

Transcript 43 lines

Cold Open Stats Overview Paper Walkthrough free_promo

Cold Open

Jenny When a city says it’s “improving transit,” what change would you actually notice first in your daily life?

Davis Honestly, I’d notice the wait before I notice anything else—if the bus shows up every 8 minutes instead of every 20, my whole day loosens up.

Jenny Okay but is that a feelings story or a measurable story, because cities love to say “better service” and then you’re still late to work and nothing else changes…

Davis I mean, if frequency really jumps, you’d expect it to show up in stuff that’s hard to fake—like more people actually getting to more jobs without moving, and neighborhoods feeling less cut off.

Jenny And that’s the tension this week: what counts as real change versus a tidy metric—because sometimes the number moves a little and the stakes are huge…welcome to This Week In Public Transit on paperboy.fm.

Stats Overview

Davis Stats check for the week: we’ve got 5 papers, and all 5 made it through—so the whole pool is the show. There are 23 unique authors across 3 countries, which is a tight set, but it’s not just one place talking to itself.

Jenny And with only five, every pick becomes a feature, not a sample. So I’m gonna keep asking the annoying question: are we seeing a pattern, or just five one-offs that happen to rhyme?

Davis Geography’s doing some work here. The papers span the U.S., Uzbekistan, and India, so the through-line isn’t “one city’s transit agency,” it’s whether policy and operations move jobs, health, and mobility in really different systems.

Jenny Methods skew applied and causal-ish, not just vibes. We’ve got one difference-in-differences study—that’s basically comparing a before-and-after change in a treated group against a similar group that didn’t get the change—plus quasi-experimental design, GIS-based modeling, and a couple simulations like traffic microsimulation. But with one of each, I want to know: are these methods here because the questions demand them, or because these were the only papers that cleared the bar this week?

Davis Author mix is interesting too: about two-thirds are emerging researchers—15 out of 23—about 7 are experienced, and there’s 1 first-time author, meaning first-ever paper. That’s a lot of early-career energy, which can mean fresher datasets and bolder designs, but also more one-study claims we’ll want to see replicated.

Jenny Theme sweep: “public transit” shows up twice, and then we’ve got public health and health equity sitting right next to employment outcomes and political advocacy. That fits the episode’s promise—measurable change without overclaiming—so the question for each paper is simple: what’s the outcome, what’s the counterfactual, and who actually benefits?

Paper Walkthrough

Paper 1 Estimating the Impact of High-Frequency Public Transit on Employment Outcomes in Chicago Neighborhoods

Jenny Alright, let’s get into the papers, and paper one is called Estimating the Impact of High-Frequency Public Transit on Employment Outcomes in Chicago Neighborhoods.

Jenny Two researchers, Fatemeh Noorizadehsalout and Amirhossein Vaziri, use the August twenty-nineteen launch of Pace’s Pulse Milwaukee Line to ask a simple question in a careful way: if you make buses come a lot more often, do jobs and work outcomes around the corridor actually change?

Jenny In plain terms, they don’t see residents suddenly getting employed or earning more right away, but they do see more jobs showing up near the upgraded service.

Jenny Their headline number is an increase of zero point zero six six workplace jobs per resident, which they say is about fourteen percent of the pre-upgrade average in those neighborhoods.

Davis If resident employment and incomes don’t move, what exactly is “jobs per resident” capturing here—are we talking about jobs that physically relocated into the corridor, or just better measurement, or maybe the control areas losing jobs?

Jenny Good catch, because “workplace jobs” here is literally where the job is located, not where the worker lives, using LEHD workplace counts, and then they scale it by the tract’s resident population.

Jenny Method-wise they run a difference-in-differences, which is just “compare the before-after change in the treated area to the before-after change in a similar nearby area,” with treated tracts within a half-mile of new Pulse stops and a control ring from a half-mile to two miles out.

Jenny They do the usual credibility checks too: an event study with flat pre-trends, placebo corridors that come up null, and buffer tests where the effect gets stronger the closer you are to the stops.

Jenny But the big limitation is right in their abstract: with tract-clustered stats, the p-value is zero point zero seven three, so it’s suggestive, not a slam dunk.

Davis I like how this forces you to separate “jobs moved closer to the bus line” from “people’s lives improved,” because those are different timelines and maybe different policies.

Davis And it’s such a clean example of our “what counts as evidence” thread: even with a tidy design and a bunch of robustness checks, they’re still honest that they’re just shy of the usual cutoff, so you’d track corridor job shifts now and plan the land-use or workforce pieces if you want household income to follow.

Paper 2 GIS-based public transport network optimization in UNESCO World Heritage cities in the example of Bukhara, Uzbekistan

Davis Okay, we were just nitpicking a p-value of zero point zero seven three in Chicago, and it made me want a paper that just swims in measurement.

Davis So here’s "GIS-based public transport network optimization in UNESCO World Heritage cities in the example of Bukhara, Uzbekistan"—and the whole premise is, can you seriously improve mobility in a protected historic city without bulldozing anything.

Davis Their answer is basically yes: with mostly operational tweaks, their with-plan scenario doubles transit mode share from fourteen percent to thirty percent, with p less than zero point zero zero one.

Davis And the average trip time drops forty percent, from thirty-two point three minutes to twenty-three point four, while thirty-minute access—meaning the share of people who can reach key places within half an hour—rises from sixty-six point three percent to eighty-one point two.

Jenny How much of that is real-world observed change versus a model telling a nice story, and what would you need to see to trust this in, say, a different UNESCO city with different street patterns and tourist peaks?

Davis It’s mostly a projection, but it’s a projection fed by a lot of primary data: a stratified resident survey of three thousand one hundred seventy-nine people with about a plus-or-minus one point seven percent margin of error, plus video at sixty-two intersections logging two million eight hundred fifteen thousand eight hundred twenty-seven vehicle movements using YOLOv8—basically an AI classifier that labels vehicles in footage—and manual passenger counts at thirty bus terminals with forty-two thousand four hundred forty-eight observation events.

Davis Then they compare three scenarios—baseline twenty-twenty-four, do-nothing twenty-twenty-six, and with-plan twenty-twenty-six—running GIS accessibility like three-hundred-meter stop buffers and thirty-minute isochrones, and they even microsimulate corridors in Vissim.

Davis The limitation is the same reason it’s compelling: it’s deep for Bukhara, but it may not transfer cleanly to other heritage cities where travel demand, enforcement, and street geometry are totally different.

Jenny I love the practical moral, though: before you go hunting for a mega-project, try the boring stuff—route restructuring, signal coordination, station upgrades—because here it’s tied to big, countable changes like fourteen to thirty percent mode share and even about fifteen percent emissions cuts in the simulation.

Jenny And it’s such a "what counts as evidence" moment: not a randomized trial, but the data backbone is so thick that you can at least argue about assumptions in public, instead of waving your hands about “heritage character” and calling it a day.

Paper 3 SARS-CoV-2 Spread and Infection Risk in Public Transit Scenes: Simulation Study Featuring a Hybrid Crowd Dynamics and Disease Spreading Modek

Jenny You just said “try the boring stuff,” and I’m like, okay, what’s the boring stuff for disease risk on transit when the station’s packed.

Jenny This one’s called SARS-CoV-2 Spread and Infection Risk in Public Transit Scenes: Simulation Study Featuring a Hybrid Crowd Dynamics and Disease Spreading Modek, and it’s basically a virtual stress test of five everyday moments: corridor walking, buying a ticket, going through gates, waiting on a platform, and riding a train.

Jenny Plain-language takeaway first: if you’re stuck in a crowd, the biggest levers are masks, better air, and less time standing around.

Jenny They build this hybrid simulator called PeDViS that stitches together a people-movement model and a virus-spread model, then they vary four knobs—demand, waiting time, facial masks, and ventilation—and the direction is consistent across all five scenes: masks and ventilation cut the chance you infect others, while higher demand and longer waits push it up.

Davis When a model says “ventilation matters most,” what assumptions are doing the heavy lifting there—like, are they assuming the virus is mostly airborne, or that people mix evenly in the space, or a certain mask quality?

Jenny Yeah, the whole engine is assumptions, and they’re explicit about it: PeDViS links NOMAD for crowd dynamics—so who stands where, who passes whom—with QVEmod for virus spread, which is basically a mechanistic recipe for how much infectious stuff accumulates and gets inhaled.

Jenny Then they run the same five transit scenes while swapping demand levels and waiting times, and toggling masks and ventilation, and they look at relative differences in infection probability rather than pretending they’ve measured your exact risk on Platform Two at 8:12.

Jenny The limitation is the ranking depends on those scenario choices and parameters—so it’s robust inside their simulated world, but you shouldn’t treat it like a universal scoreboard for every city, every train car, every variant.

Davis Still, it lands for operations: if delays plus crowding are the “dangerous” combo, then the health move is the same move as the reliability move—keep queues moving, stop platform dwell from ballooning, and spend real money on ventilation where the crush loads happen.

Davis And it fits that transit-as-health-policy thread, but with the honesty label on it: it’s not a randomized trial, it’s a model you can argue with, and that’s way better than making mask-or-air decisions off vibes.