Descriptive variable / column names?

[arilamstein]

I'm using tidycensus for the first time in a year. I'm trying to get output from a table, and am surprised that the variable names are codes, and not English words. For example:

> get_acs("congressional district", table="B03002", output="wide", cache_table=TRUE)
>  A tibble: 440 x 44
   GEOID NAME  B03002_001E B03002_001M B03002_002E B03002_002M B03002_003E B03002_003M
   <chr> <chr>       <dbl>       <dbl>       <dbl>       <dbl>       <dbl>       <dbl>
 1 2701  Cong…      672649         322      629764         419      576824         542
 2 2702  Cong…      698522         399      656109         615      568269         966
 3 2703  Cong…      710378         517      678509        1781      551545        2888
 4 2704  Cong…      706259         337      658885         583      472459         985
 5 2705  Cong…      707352         449      641898        1748      444397        2775
 6 2706  Cong…      702583         541      683042        1071      626547        1323
 7 2707  Cong…      664941         540      633419         682      588501         706
 8 2708  Cong…      664674         326      652968         351      609529         431
 9 2801  Cong…      760715         674      735566         664      508360         623
10 2802  Cong…      711716        1621      695686        1676      216997        1066

I recall dealing with the issue when I last used Tidycensus. But I can't seem to remember how I solved it.

When I type the table name directly into Google I wind up here: https://censusreporter.org/tables/B03002/. My assumption is that what Tidycensus calls B03002_001E is the first row in that table ("total"). But is _002E "Not Hispanic or Latino:", or is that a placeholder with no data? In that case _002E is actually "White alone". Or maybe this website is from a different vintage of the survey, in which case the variable codes might map to totally different data.

I guess my question is this: does tidycensus have support for using English variable names instead of codes? If not, what is the best way to do this myself? I'd really like to avoid winging it with something this important.

Thanksl

[walkerke]

Hi @arilamstein! Correct, all variable names are codes, and I don't plan to support English words for variable names. None of the variables in the ACS are placeholders - they all represent some level of aggregation. So B03002_002E in your table above is the number of residents of any race by congressional district who responded Not Hispanic or Latino.

The way that Census organizes these variables within tables like the one you've acquired is within varying nested hierarchies. When we grab the linked JSON file in tidycensus with load_variables(), the levels of these hierarchies are separated by !! in the label field:

library(tidycensus)
library(tidyverse)

acs18 <- load_variables(2018, "acs5", cache = TRUE)

acs18 %>% 
  filter(str_detect(name, "B03002")) %>% 
  View()

B03002_001 represents the total for the table; B03002_002 and B03002_012 are the next level of the hierarchy (Not Hispanic or Latino and Hispanic or Latino, respectively); and the remaining variables represent racial identification with respect to the identified Hispanic or Latino ethnicity in the ACS survey.

The reason why I don't plan to support English variable names in tidycensus is because using English variable names gets messy quite quickly. For example, B03002_005 is much shorter than typing "Not Hispanic Or Latino!!American Indian and Alaska Native alone". Plus - these variable labels are themselves organized within concepts (in this case, Hispanic or Latino Origin by Race" which get duplicated across concepts. For example, let's say we just want the total Male population, for which we'd usually use B01001_002:

acs18 %>% 
  filter(label == "Estimate!!Total!!Male") %>%
  View()

We get back 149 different possibilities corresponding to a wide variety of different tabulations (shown in the concept column), many of which are quite verbose to express in English.

My general recommendation is that users call load_variables() and carefully browse the returned object to identify the variable codes they need (or use an external website like Census Reporter which has a very nice interface). When they have collected these variable codes, they can then pass a named vector to get_acs() to create their own simpler variable names for their own work. For example,

get_acs(
  geography = "congressional district",
  variables = c(not_hispanic = "B03002_002",
                hispanic = "B03002_012"),
  output = "wide")

Getting data from the 2014-2018 5-year ACS
# A tibble: 440 x 6
   GEOID NAME                            not_hispanicE not_hispanicM hispanicE hispanicM
   <chr> <chr>                                   <dbl>         <dbl>     <dbl>     <dbl>
 1 2701  Congressional District 1 (116t…        629764           419     42885       380
 2 2702  Congressional District 2 (116t…        656109           615     42413       523
 3 2703  Congressional District 3 (116t…        678509          1781     31869      1694
 4 2704  Congressional District 4 (116t…        658885           583     47374       498
 5 2705  Congressional District 5 (116t…        641898          1748     65454      1651
 6 2706  Congressional District 6 (116t…        683042          1071     19541      1030
 7 2707  Congressional District 7 (116t…        633419           682     31522       415
 8 2708  Congressional District 8 (116t…        652968           351     11706        95
 9 2801  Congressional District 1 (116t…        735566           664     25149       146
10 2802  Congressional District 2 (116t…        695686          1676     16030       658
# … with 430 more rows

[mfherman]

These column names are very much not pretty, but here's one approach you could use to take the label info from the data frame returned by load_variables() and join them with the variable codes returned by get_acs().

library(tidycensus)
library(tidyverse)

var_list <- load_variables(2018, "acs5", cache = TRUE) %>% 
  select(-concept) %>% 
  mutate(label = str_remove(label, "Estimate"))

df <- get_acs(
  geography = "congressional district",
  table = "B03002"
  )
#> Getting data from the 2014-2018 5-year ACS

df %>% 
  left_join(var_list, by = c("variable" = "name")) %>%
  select(-variable) %>% 
  pivot_wider(names_from = label, values_from = c(estimate, moe)) %>% 
  janitor::clean_names()
#> # A tibble: 440 x 44
#>    geoid name  estimate_total estimate_total_… estimate_total_… estimate_total_…
#>    <chr> <chr>          <dbl>            <dbl>            <dbl>            <dbl>
#>  1 0101  Cong…         706503           683855           459898           193875
#>  2 0102  Cong…         680575           656410           420590           211747
#>  3 0103  Cong…         706705           684817           478120           179170
#>  4 0104  Cong…         683391           639036           571190            47200
#>  5 0105  Cong…         714145           677764           518936           123821
#>  6 0106  Cong…         703715           669725           535082           107947
#>  7 0107  Cong…         669646           649927           212914           421977
#>  8 0200  Cong…         738516           687330           450754            22817
#>  9 0401  Cong…         753580           583124           366818            16209
#> 10 0402  Cong…         719924           515278           438332            26199
#> # … with 430 more rows, and 38 more variables:
#> #   estimate_total_not_hispanic_or_latino_american_indian_and_alaska_native_alone <dbl>,
#> #   estimate_total_not_hispanic_or_latino_asian_alone <dbl>,
#> #   estimate_total_not_hispanic_or_latino_native_hawaiian_and_other_pacific_islander_alone <dbl>,
#> #   estimate_total_not_hispanic_or_latino_some_other_race_alone <dbl>,
#> #   estimate_total_not_hispanic_or_latino_two_or_more_races <dbl>,
#> #   estimate_total_not_hispanic_or_latino_two_or_more_races_two_races_including_some_other_race <dbl>,
#> #   estimate_total_not_hispanic_or_latino_two_or_more_races_two_races_excluding_some_other_race_and_three_or_more_races <dbl>,
#> #   estimate_total_hispanic_or_latino <dbl>,
#> #   estimate_total_hispanic_or_latino_white_alone <dbl>,
#> #   estimate_total_hispanic_or_latino_black_or_african_american_alone <dbl>,
#> #   estimate_total_hispanic_or_latino_american_indian_and_alaska_native_alone <dbl>,
#> #   estimate_total_hispanic_or_latino_asian_alone <dbl>,
#> #   estimate_total_hispanic_or_latino_native_hawaiian_and_other_pacific_islander_alone <dbl>,
#> #   estimate_total_hispanic_or_latino_some_other_race_alone <dbl>,
#> #   estimate_total_hispanic_or_latino_two_or_more_races <dbl>,
#> #   estimate_total_hispanic_or_latino_two_or_more_races_two_races_including_some_other_race <dbl>,
#> #   estimate_total_hispanic_or_latino_two_or_more_races_two_races_excluding_some_other_race_and_three_or_more_races <dbl>,
#> #   moe_total <dbl>, moe_total_not_hispanic_or_latino <dbl>,
#> #   moe_total_not_hispanic_or_latino_white_alone <dbl>,
#> #   moe_total_not_hispanic_or_latino_black_or_african_american_alone <dbl>,
#> #   moe_total_not_hispanic_or_latino_american_indian_and_alaska_native_alone <dbl>,
#> #   moe_total_not_hispanic_or_latino_asian_alone <dbl>,
#> #   moe_total_not_hispanic_or_latino_native_hawaiian_and_other_pacific_islander_alone <dbl>,
#> #   moe_total_not_hispanic_or_latino_some_other_race_alone <dbl>,
#> #   moe_total_not_hispanic_or_latino_two_or_more_races <dbl>,
#> #   moe_total_not_hispanic_or_latino_two_or_more_races_two_races_including_some_other_race <dbl>,
#> #   moe_total_not_hispanic_or_latino_two_or_more_races_two_races_excluding_some_other_race_and_three_or_more_races <dbl>,
#> #   moe_total_hispanic_or_latino <dbl>,
#> #   moe_total_hispanic_or_latino_white_alone <dbl>,
#> #   moe_total_hispanic_or_latino_black_or_african_american_alone <dbl>,
#> #   moe_total_hispanic_or_latino_american_indian_and_alaska_native_alone <dbl>,
#> #   moe_total_hispanic_or_latino_asian_alone <dbl>,
#> #   moe_total_hispanic_or_latino_native_hawaiian_and_other_pacific_islander_alone <dbl>,
#> #   moe_total_hispanic_or_latino_some_other_race_alone <dbl>,
#> #   moe_total_hispanic_or_latino_two_or_more_races <dbl>,
#> #   moe_total_hispanic_or_latino_two_or_more_races_two_races_including_some_other_race <dbl>,
#> #   moe_total_hispanic_or_latino_two_or_more_races_two_races_excluding_some_other_race_and_three_or_more_races <dbl>

^{Created on 2020-07-27 by the reprex package (v0.3.0)}

[arilamstein]

Guys, thank you so much for the thorough answer!