What explains the weak correlation between service frequency and ridership?

[Hannah22]

I’ve been looking at the data from my community survey on public transit use, and I’m stuck on a specific methodological issue. My initial hypothesis was that frequency of service would be the strongest predictor of ridership, but the correlation is surprisingly weak, and I can’t figure out if my survey design failed to capture a key variable or if my analysis is just missing something.

[RichardS]

Honestly I tried something similar last year. I pulled usage data from the city counts and paired it with service frequency by route. The scatter was messy and the correlation was weak. I did one simple check: are you measuring headways or scheduled frequency? The two can diverge a lot, especially on routes with variable headways.

[Aaron86]

Reliability mattered more. I grabbed on time performance from the transit agency and added it as a variable, and when I included an interaction with frequency, ridership aligned a bit on some corridors. It suggested that frequency alone isn’t the whole story. Also I noticed access to stops and first mile issues could mute the effect of frequency.

[Harper.L]

Could the problem be the survey design itself? If you only sampled current riders, you might miss a latent group that would ride with small improvements.

[Steven31]

Drifted into a tangent once, thinking about weather or events pushing people off schedules. When I checked by time of day, mornings looked different from evenings and weekends, and that teased apart a little more nuance but also created more questions. I guess the data wants a more nuanced model, not a single predictor.