In 2025, CTA rolled out its Frequent Network—20 bus routes promised 10-minute headways. We analyzed ridership data and real-time vehicle positions to find out if it's delivering.
Using phase-level difference-in-differences with bootstrap confidence intervals, we compared Frequent Network routes against similar non-FN routes. When each rollout phase is analyzed separately, all three with sufficient post-launch data show statistically significant ridership gains. These results are corroborated by a Callaway-Sant’Anna staggered estimator and regression with fixed effects.
CTA's Frequent Network rolled out in four phases from March to December 2025. Click any route for details.
Phase-level difference-in-differences estimates with cluster-bootstrap confidence intervals, comparing Frequent Network routes to similar non-FN routes. Each phase is analyzed separately with its own control group. Why this approach?
| Phase | Routes | Effect (rides/day) | Effect (%) | 95% CI |
|---|---|---|---|---|
| Phase 1 | J14, 34, 47, 54, 60, 63, 79, 95 | +428 | +5.9% | [+42, +799] |
| Phase 2 | 4, 20, 49, 66 | +514 | +4.2% | [+163, +832] |
| Phase 3 | 53, 55, 77, 82 | +1,142 | +10.4% | [+213, +2,492] |
| Phase 4 | 9, 12, 72, 81 | — | — | No post-launch data |
Comparing matched calendar periods before and after each phase launched. 13 of 16 Phase 1–3 routes gained riders. J14 Jeffery Jump leads at +26.1%.
Frequent Network routes are capturing a growing share of CTA's total bus ridership after each phase launch. Importantly, total CTA bus ridership also grew during this period—the rising FN share reflects genuine ridership growth on frequent routes, not just riders switching from other buses.
We also tracked real-time bus positions using CTA's Bus Tracker API to measure whether routes actually achieve the promised 10-minute headways. Early results suggest a gap between schedule and reality.
See the headway analysis →This analysis uses difference-in-differences to estimate the causal impact of the Frequent Network on ridership, with phase-level estimation to avoid bias from staggered treatment timing. We ran eight separate analyses to stress-test the findings from different angles—pooled, phase-level, staggered (Callaway-Sant'Anna), and regression with fixed effects.
Read the full methodology →