Using Causal Inference to Estimate the Impact of Tube Strikes on Cycling Usage in London

2026-04-22 · Source: Towards Data Science · Field: Technology & Digital — Data Science & Analytics, Public Transportation & Urban Mobility · Depth: Intermediate, long

Summary

Transport for London (TFL) shares its Santander Cycle usage data, encompassing 9.2 million station-hours across 800 bike stations from 2015 to 2025, provided in 144 weekly CSV files. This data, aggregated to H3 cell-day level, is used to causally analyze the impact of major London Underground strikes on cycle usage. The analysis involves extensive data wrangling, including converting CSVs to Parquet, grouping by bike station and hour, and joining with H3 cell coordinates and strike data from a Freedom of Information request (FOI-2596-1819). The study defines strike exposure based on proximity to striking tube lines and uses a two-way fixed effects (TWFE) model to estimate the causal effect, controlling for confounders like weather and seasonality. The most refined model, focusing on cells near 42 central interchange stations and within a 45-day window of strikes, estimates a 3.95% increase in Santander bike usage on strike days.

Key takeaway

For data scientists or urban planners analyzing public transport patterns, this analysis demonstrates how to quantify the causal impact of disruptions using open data and robust econometric methods. You should consider panel data approaches like two-way fixed effects when assessing interventions with recurring binary treatments, as they effectively mitigate selection bias and time-invariant confounding. This framework can inform infrastructure planning and emergency response strategies during transit strikes.

Key insights

Tube strikes in London cause a measurable increase in Santander Cycle usage due to commuter substitution.

Principles

• Panel data is effective for isolating time-invariant characteristics.
• Causal mechanisms must credibly link treatment to outcome.
• SUTVA violations can attenuate causal effect estimates.

Method

A two-way fixed effects model, clustering errors at the cell level, estimates causal treatment effects by controlling for both time-invariant cell characteristics and cell-invariant day-specific variations.

In practice

• Use H3 cells for spatial aggregation in urban mobility analysis.
• Restrict analysis to relevant geographic and temporal windows.
• Cluster standard errors at the unit level in panel data regressions.

Topics

• Causal Inference
• London Tube Strikes
• Santander Cycles
• Panel Data Analysis
• Two-Way Fixed Effects Model

Code references

• stucsk99/tfl_bike_casual

Best for: Data Scientist, Research Scientist, AI Student

Related on AIssential

• Understanding Long-Term Dynamics of Individual Metro Usage: A Hidden Semi-Markov State Framework with Survival Analysis — The framework models individual metro usage lifecycles by integrating latent mobility states with exit and re-entry hazards.
• é — Causal Inference Methods Every Data Scientist Should Know — Causal inference distinguishes correlation from causation, enabling robust decision-making in complex systems.
• CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting — CausalPOI uses spatio-temporal causal modeling to predict check-ins for new POIs, outperforming correlation-based methods.
• Epidemiologist Donna Spiegelman sez: SUTVA is “mostly not necessary for valid causal estimation and inference most of the time” — SUTVA is often unnecessary for valid causal inference, as interference and stochastic outcomes can be directly modeled.
• Balanced Twins: Causal Inference on Time Series with Hidden Confounding — B-Twin robustly estimates treatment effects in time series by matching latent representations and propensity scores under hidden confounding.
• Beyond Means: Topological Causal Effects under Persistent-Homology Ignorability — Persistent homology enables identifying causal effects on outcome distribution shapes, addressing limitations of mean-based estimands.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Towards Data Science.