Research Question: Importance of GTFS-Flex

[e-lo]

Journey-planning applications have asked for some data points to help prioritize GTFS-Flex capabilities. Some data points which would help demonstrate the importance of demand-responsive transit:

1. Number of transit services which have demand-responsive transit
2. % of California (by area, pop, employment) not served by fixed-route transit service
3. % of California served by demand-responsive transit
4. Total ridership of demand-responsive transit services

Priority

Med-high: This will feed into [unnamed journey planning application]'s prioritization process. There isn't a specific deadline so much as "as soon as we provide the data, they can start the process that will then start a process.

Data Sources

• Population: [census]
• Employment: LODES
  • Data dictionary
  • total jobs is S000_JT00 but could want to split up later and look at minority jobs
  • Metadata to reference in analysis
  • dataset for workplace location for all jobs in 2018
  • geography crosswalk
• demand-responsive and fixed-route operators:
  • Transit Provider db with NTD data linked
  • Hunter's previous research?
• Geographic areas: combo of stops.txt from all GTFS files with 1/4 or 1/3 mile buffers
• Ridership:
  • Transit Provider db with NTD data linked
  • Hunter's previous research?

[hunterowens]

previous report and research

[machow]

@hunterowens and @Nkdiaz to pair in Sep. Key questions:

• where does the data for this question live right now?
• where will we put the results?
• can we produce snippets on the following...
  • calculating % of california
  • calculating whether a service has demand-responsive transit
  • e.g. X from this data, Y from that data, combined in this way

[hunterowens]

to do this, we'll need a final dataset that looks something like this

census_tract_geoid	has_demand_responsive	has_fixed_route	population	jobs	area
xxx	True	False	10	1	10M2

this will require 3ish datasets.

1. Census Tract Population / Jobs Numbers for all of CA
2. A list of tract with fixed route service - join the stops dataset to tract geometeries, assume all that have a stop have fixed route service
3. A list of tracts with DR service. TAke the City / County fields from the Transitstacks / NTD HQ data, assume the DR provider services the entirity of that geog. join to the state geoportal dataset, simplify into a single geometry, take the centroid of the tract, point in polygon insersect that to the tracts dataset to get "has_DR"

[hunterowens]

@hunterowens to link census datasets here for @Nkdiaz

[hunterowens]

data - you'll want 3 year ACS tract level population estimates.

@e-lo linked to the LODES employment statistics by tract in the issue description

[machow]

For data team meeting (from pairing w/ @Nkdiaz):

• Is this the data we want? (ACS 5 year demographic and housing info): https://data.census.gov/cedsci/table?q=United%20States&tid=ACSDP5Y2019.DP05
• here's an api link for that data

Intermediate next step:

• calculate population per geoid using census data
• fetch lodes data linked up top, calculate jobs per geoid

Final piece:

• Join census with stops (so we have a geoid associated per stop)
• Sync w/ Michael to check
• (Natalya TODO): Layout how to finish has_demand_responsiveness, has_fixed_route

[machow]

From pairing w/ @Nkdiaz, I just remembered that R has a very nice census library called tidycensus (as I recall, python's census lib is a bit more low level). Here is an R script for pulling out the data in the format @hunterowens listed above.

May need to tweak the variables we grab (I just grabbed a couple to fill in)

library(tidycensus)
library(tidyverse)

census_api_key("<API_KEY_HERE>")

# uncomment this to see possible variables to load
load_variables(2019, "acs5", cache = TRUE) %>% View()

# get ACS data. modify variables argument to get additional measures
# we'll transform them to wide format so each variable will be its own column
tracts <- get_acs(
  geography="tract",
  state = "CA",
  year = 2019,
  variables = c(ttl_population = "B01003_001", ttl_jobs = "B24011_001")
)

wide_tracts <- tracts %>% select(-moe) %>% spread(variable, estimate)

wide_tracts %>% write_csv("calitp_acs5_tracts.csv")

[machow]

In order to calculate has_demand_responsive that Hunter lists above. We need to do the following...

• get data from this google sheet https://docs.google.com/spreadsheets/d/105oar4q_Z3yihDeUlP-VnYpJ0N9Mfs7-q4TnribqYLU/edit#gid=471807468
• Use the Modes column to figure out if they have "DR" (for now, just check that DR is one of the modes listed)
• Use the City and County columns to map back to census data.

[machow]

@e-lo for this piece...

Total ridership of demand-responsive transit services

Where do total ridership numbers come from? Is it somewhere in transitstacks?

edit: nvm we found it in the NTD Stats has UPT_DR which @Nkdiaz is planning to use. Definitely let us know if there's a better source

[machow]

Blocked until the data source for demand responsive provider info is provided. See this slack thread.

[machow]

Here is the notebook Natalya worked on for this: https://github.com/cal-itp/data-analyses/blob/main/gtfs_flex_research_questions/calitp_flex_research_demand_responsive.ipynb

[machow]

The draft of this analysis is completed--let's wait until other needs to re-open

see this related issue: cal-itp/data-analyses#171

[tiffanychu90]

@Nkdiaz: With the notebook, I see a couple of datasets that could be documented in intake:

• [ ] CA census tracts (we have some versions of this with population), so let's decide on a version and stick with it
• [ ] CA cities
• [ ] CA counties
• [ ] LODES/LEHD (with this version and one from the bus_service_increase analysis, let's pick a version and document it)

The demand responsive data is fairly analysis-specific....but it comes from a Google Sheet and not our warehouse? If It's not elsewhere in the warehouse and this is a canonical source, then maybe it should either live in the warehouse or a GCS bucket and documented in intake