This Week In Public Transit

Transcript 28 lines

Cold Open Stats Overview Paper Walkthrough free_promo

Cold Open

Jenny If the bus were free tomorrow, would you actually ride it more?

Davis Some days, yes, because free feels like permission, but if the bus still comes late or drops me four blocks from the errand, price isn't the only thing stopping me.

Jenny My gut says free should matter, especially for someone counting every trip, and that's why the King County result is such a useful gut check.

Davis Right, because when low-income riders got fare-free transit, the big promised changes in movement and well-being were surprisingly small, so this week we're asking what has to change besides the fare...welcome to This Week In Public Transit on paperboy.fm.

Stats Overview

Jenny This week we screened 12 transit hits and kept 5 qualified papers, with 15 unique authors across 4 countries: the U.S., Mexico, Russia, and China.

Davis That qualified pile is down from 6 to 5, about 17 percent, so the feed got a little narrower. The methods are scattered too: one experimental, one quasi-experimental, one survey, one regression, and one case study, which makes this feel less like a wave and more like five separate field notes.

Jenny But the search pool went the other way: 12 total hits, up from 11, about 9 percent. So more papers matched the query while fewer survived review — what's driving that, broader language around public transportation and mobility, or papers that mention transit without really showing what changes for riders?

Davis The author count dropped harder, from 24 to 15, down about 38 percent, while countries stayed flat at 4. That sounds less like geographic contraction and more like smaller teams, especially in a week where each top country contributes one paper.

Jenny The author mix is pretty balanced. Out of 15 authors, 4 are first-time, meaning their first-ever paper in the metadata, 4 are emerging, and 7 are experienced; that's about 27 percent, 27 percent, and 47 percent.

Davis The theme sweep fits the through-line: public transportation, mobility, and income subsidies each show up once, with health outcomes, COVID-19, ETA prediction, and nodal segmentation right behind. So the practical question stays steady: when policy, prediction tools, and urban design hit real places, who actually gets a better trip?

Paper Walkthrough

Paper 1 Means‐Tested Public Transportation Subsidies: Causal Evidence and Implications

Jenny Alright, let's get into the papers with Means-Tested Public Transportation Subsidies: Causal Evidence and Implications, a twenty twenty-six study of King County, Washington, where low-income riders were offered fare-free public transportation.

Jenny The surprising result is that making transit free had very limited effects on how much people moved around, and also on downstream health and well-being outcomes.

Davis If free transit didn't change mobility much, what exactly did they measure, and how did they know this wasn't just noise in King County's data?

Jenny They used both an experiment and a quasi-experiment, meaning they combined a direct test with a real-world comparison group, and they matched that to administrative records and surveys.

Jenny The big caution is that this happened after COVID, when travel habits and fare-payment norms had shifted, including more tolerated fare evasion, so the price elasticity of demand, meaning how much riding changes when price changes, may have been unusually low.

Davis So the practical lesson isn't that subsidies don't matter; it's that price is only one piece of the rider experience, and in King County the evidence is solid enough to take seriously but too local to turn into a universal rule.

Paper 2 Adaptive Time of Arrival (ETA) Prediction in Light Rail Transport Systems Using Regression and Nodal Segmentation: CDMX Light Rail Case Study

Davis That King County point about price being only one piece sets up this next paper really cleanly. Ricardo López-León and colleagues look at the other daily pain point in Adaptive Time of Arrival Prediction in Light Rail Transport Systems Using Regression and Nodal Segmentation: CDMX Light Rail Case Study.

Davis The plain version is that they try to make the wait feel less unknowable on Mexico City Light Rail Line one. They divide the thirteen kilometer route into seventeen segments, and their model predicts arrivals with eighty-three percent accuracy, better than simpler linear regression, which is just a straight-line prediction from past travel times.

Jenny Is eighty-three percent accuracy enough to change passenger trust, though, or is that mainly a model score that looks good in a paper?

Davis The evidence is operational, not psychological, so they don't measure trust directly. What they do is use nodal segmentation, meaning they break the line into operating chunks, then add time-series patterns, station dwell time, and the effect of at-grade road intersections, but the authors are careful that this is one Mexico City light rail line, not the Metro and not the whole city rail network.

Jenny That feels like the right-sized claim. It also fits the Price Versus Experience thread from the subsidy paper: if the fare is cheap but the sign lies, riders still pay with uncertainty, and for at-grade light rail the fix may be a model that knows where the street actually gets in the way.

Paper 3 Using GeoXAI to reveal gender differences in the spatially nonstationary influence of built environment on area-level happiness

Jenny That line about the street getting in the way is a nice bridge, because this next paper asks whether the city gets in the way differently for different people. It's called Using GeoXAI to reveal gender differences in the spatially nonstationary influence of built environment on area-level happiness, and in plain English, Ye and Ding use mapped machine learning to ask how transit distance, greenness, density, and street design line up with happiness across Shanghai.

Jenny The scale is the first thing that grabbed me: six hundred seventy-eight thousand one hundred twenty geotagged Weibo posts, spread across thirty-three thousand five hundred ninety-one five hundred meter grids. After filtering, they model three hundred twenty thousand six hundred eighteen female-labelled posts and one hundred eighty-seven thousand eighty-nine male-labelled posts, and the relationships aren't citywide averages; spatially nonstationary just means the pattern changes from neighborhood to neighborhood.

Jenny Their strongest equity result is that the same built environment feature did not dominate in both gender-labelled surfaces. Greenness, measured partly through NDVI, which is a satellite-based measure of vegetation, ranked first in the female-labelled model, while population count ranked first in the male-labelled model. And Distance to Transit contributed more to female-labelled happiness in eighty-four percent of paired grids.

Davis How do we know this is measuring happiness in the city, and not just who posts on Weibo, who geotags, and where they happen to be when they're bored or annoyed?

Jenny That's the right pressure point. They classify sentiment with an LSTM-CNN, basically a neural network that reads text patterns, and report zero point nine four two accuracy; then they use XGBoost, a tree-based prediction model, plus SHAP, which estimates which inputs pushed the model up or down, and they check that performance stays positive under spatial separation and a twenty-user grid threshold. But the limitation is real: this is geotagged Weibo and gender-labelled data in Shanghai, not a survey of the whole population.

Davis So the practical takeaway isn't, plant trees and everyone gets happier in the same way. It's more like the Place Shapes Access thread: if a transit agency or city planner only looks at one citywide average, they can miss that transit proximity, destination access, greenness, and density may matter differently across actual neighborhoods and gendered travel patterns.

This Week In Public Transit

Transit beyond the farebox

Episode

Cold Open

Stats Overview

Paper Walkthrough

Paper 1 Means‐Tested Public Transportation Subsidies: Causal Evidence and Implications

Paper 2 Adaptive Time of Arrival (ETA) Prediction in Light Rail Transport Systems Using Regression and Nodal Segmentation: CDMX Light Rail Case Study

Paper 3 Using GeoXAI to reveal gender differences in the spatially nonstationary influence of built environment on area-level happiness

free_promo

Other Episodes

This Week In Public Transit

Transit beyond the farebox

Episode

Cold Open

Stats Overview

Paper Walkthrough

Paper 1 Means‐Tested Public Transportation Subsidies: Causal Evidence and Implications

Paper 2 Adaptive Time of Arrival (ETA) Prediction in Light Rail Transport Systems Using Regression and Nodal Segmentation: CDMX Light Rail Case Study

Paper 3 Using GeoXAI to reveal gender differences in the spatially nonstationary influence of built environment on area-level happiness

free_promo

Other Episodes

Get This Week In Public Transit by email