Skip to contents

psrcctpp basics

psrcctpp_basics.Rmd

Why CTPP data?

Although it is derived from the American Community Survey (ACS), CTPP includes data items that are not published in the ACS release, and at scales substantially smaller than ACS Public Use Microdata (PUMS). As with other ACS products, it is consistent across scales, supports longitudinal analysis, and includes margins of error. It is the most authoritative nationwide, Census Bureau source for transportation flow data in particular.

The primary drawback to CTPP data is the approximately 3-year lag to develop it; since it’s restricted to non-overlapping 5-year spans, it may involve a remove of nearly seven years from the conclusion of the first year in a reported span. Issues such as the transportation impacts accompanying the 2020 Covid pandemic have a long signature in the data (i.e. the next data without it would arrive around 2029).

Identify and retrieve CTPP data

Determine which table to request

In order to request data, as with the ACS, one must know the CTPP table code (or variable codes). To assist in identifying this, the package has a search function, ctpp_tblsearch(). It has two required parameters:

  • prefix - CTPP uses a pattern in table names: the first character “A” denotes originally reported data (with confidential values suppressed); “B” indicates data with noise infusion instead of suppression (“perturbed”), and “C” data with noise infusion and fewer categories (“condensed” to enable greater statistical reliability). Release of the 2017-21 dataset saw full retirement of “A” tables and introduction of “C” tables. The second character indicates either residence geography (“1”), workplace geography (“2”), or flows (“3”, i.e. both residence and workplace geography). For example, a condensed table reported by workplace geography uses the prefix “C2”. The ctpp_tblsearch() function can also take a question mark as wildcard in place of either character, i.e. “?3” would show all relevant flow tables, regardless of whether they are original or perturbed data.

  • “regex” - This is the regular expression search term to find in the table description. It is not case-sensitive.

  • (optional) year - i.e. specify which survey; otherwise defaults to latest (currently 2012-16)

In our example, we’re looking for a workplace-geography table showing mode, aka “means of transportation” in Census parlance.

shhh <- suppressPackageStartupMessages
shhh(library(psrcctpp))
shhh(library(magrittr))
shhh(library(dplyr))
shhh(library(stringr))

ctpp_tblsearch("?2", "means of transp", year=2021) %>%
  mutate(desc=str_replace(description, "by Means .*", "bmtw")) %>%
  select(c(name, desc)) %>% head()
##       name                                        desc
##     <char>                                      <char>
## 1: B202105        Means of Transportation to Work (18)
## 2: B202200 Race and Hispanic or Latino Origin (3) bmtw
## 3: B202202                      Age of Worker (8) bmtw
## 4: B202213               Travel Time to Work (15) bmtw
## 5: B202216             Time Arriving At Work (17) bmtw
## 6: B202222                   Class of Worker (10) bmtw

In our example, the first table (A202105) is the one we’re after. Notice, even if your attribute of interest is listed secondarily, there will typically be subtotals by that dimension alone within the table.

Get the data

Once we’ve identified the table code, we can now use the get_psrc_ctpp() function to retrieve the data. Its parameters are:

  • scale - for residence or workplace tables, either “county”, “place”, or “tract”
    for flow tables, either “county-county”, “place-place”, “place-county”, “county-place”, or “tract-tract”
    “block group” scale will be available starting with the 2017-21 survey.
  • table_code - identifier (string) for the table you want (you can also request individual variables)
  • dyear - aka data year, the last year of the CTPP span (i.e. 2016 for 2012-16)

By default, the resulting dataframe is restricted to geographies within the Central Puget Sound region (or flows with one end within the region). To specify a narrower set–or geographies outside the region–you can provide a character vector using the optional geoids argument containing the desired FIPS codes.

Notice that all results will include both res_geoid, res_label and work_geoid, work_label fields, but the work fields will be NA for residential tables and the residence fields will be NA for workplace tables. Categorical fields are left as character rather than factor datatype. Suppressed values in original data (“A”) tables are also coded NA.

x <- get_psrc_ctpp("tract", "B202105", 2021)            # get data

mutate(x, tract=str_extract(work_label, "[\\.\\d]+")) %>%  # fit in frame
  select(10, 7:8) %>% filter(tract=="82") %>% head()                   
##     tract                                    category estimate
##    <char>                                      <char>    <num>
## 1:     82                                    Bicycle       350
## 2:     82                                         Bus     7720
## 3:     82           Car, truck, or van -- Drove alone     8120
## 4:     82 Car, truck, or van -- In a 2-person carpool     1165
## 5:     82 Car, truck, or van -- In a 3-person carpool      120
## 6:     82 Car, truck, or van -- In a 4-person carpool       60

Aggregate CTPP data using custom variables

CTPP data tables include totals as well as category breakdowns, but if you wish to define either a custom category or a custom geography (as an aggregate of a smaller geography, e.g. tract or block group), you can create your own grouping variable, and then summarize using that variable in the psrc_ctpp_sum() function, as follows.

psrc_ctpp_sum() results include both estimates and corresponding margins of error. The incl_na=FALSE option can be used to remove irrelevant categories without having to create a separate filtered data object first (NA values for relevant categories are still preserved).

You can utilize the convenience function ctpp_shares() to append the share and share MOE to any CTPP dataset as long as it contains category totals (as is typically the case).

x %<>% mutate(
  custom_geo=case_when(
     str_sub(work_geoid,4L,11L) %in% paste0("3302380",3:4)         ~ "Downtown Bellevue",
     str_sub(work_geoid,4L,11L) %in% paste0("61040",c(4,7:8),"00") ~ "Downtown Everett",
     TRUE                                                          ~ NA_character_),
  category=factor(
    case_when(
      grepl(" carpool$",                category) ~"Carpool",
      grepl("Drove alone|Motorcycle",   category) ~"Drove Alone",
      grepl("Bus|Subway|Ferry|[rR]ail", category) ~"Transit",
      grepl("Bicycle|Walked",           category) ~"Bike/Ped",
      grepl("^Total, means",            category) ~"Total",
      grepl("Other|Taxi",               category) ~"Other",
      !is.na(category) ~category),
    levels=c("Bike/Ped","Transit","Carpool","Drove Alone","Other","Total")))
rs <- psrc_ctpp_sum(x, group_vars="custom_geo", incl_na=FALSE) %>% ctpp_shares()
head(rs[,2:7])
## # A tibble: 6 × 6
##   custom_geo       category    estimate estimate_moe   share share_moe
##   <chr>            <fct>          <dbl>        <dbl>   <dbl>     <dbl>
## 1 Downtown Everett Bike/Ped         760        287.  0.0479    0.0179 
## 2 Downtown Everett Transit          310        146.  0.0196    0.00915
## 3 Downtown Everett Carpool         1295        295.  0.0817    0.0179 
## 4 Downtown Everett Drove Alone    12630        865.  0.797     0.0240 
## 5 Downtown Everett Other             59         48.2 0.00372   0.00303
## 6 Downtown Everett Total          15855        976.  1         0.0870