Census variable selection?

[ericmanning]

Why are race totals assigned by the following variables

r_whi = "P12I_001N"
r_bla = "P12B_001N"
r_his = "P12H_001N"
r_asi = c("P12D_001N", "P12E_001N")
r_oth = c("P12C_001N", "P12F_001N", "P12G_001N")

which correspond to the following Census tables

2020 DHC table	2010 DHC table	Title
P12B	P12B	SEX BY AGE FOR SELECTED AGE CATEGORIES (BLACK OR AFRICAN AMERICAN ALONE)
P12C	P12C	SEX BY AGE FOR SELECTED AGE CATEGORIES (AMERICAN INDIAN AND ALASKA NATIVE ALONE)
P12D	P12D	SEX BY AGE FOR SELECTED AGE CATEGORIES (ASIAN ALONE)
P12E	P12E	SEX BY AGE FOR SELECTED AGE CATEGORIES (NATIVE HAWAIIAN AND OTHER PACIFIC ISLANDER ALONE)
P12F	P12F	SEX BY AGE FOR SELECTED AGE CATEGORIES (SOME OTHER RACE ALONE)
P12G	P12G	SEX BY AGE FOR SELECTED AGE CATEGORIES (TWO OR MORE RACES)
P12H	P12H	SEX BY AGE FOR SELECTED AGE CATEGORIES (HISPANIC OR LATINO)
P12I	P12I	SEX BY AGE FOR SELECTED AGE CATEGORIES (WHITE ALONE, NOT HISPANIC OR LATINO)

and not the following tables' variables instead?

2020 DHC table	2010 DHC table	Title
P12H	P12H	SEX BY AGE FOR SELECTED AGE CATEGORIES (HISPANIC OR LATINO)
P12I	P12I	SEX BY AGE FOR SELECTED AGE CATEGORIES (WHITE ALONE, NOT HISPANIC OR LATINO)
P12J	N/A	SEX BY AGE FOR SELECTED AGE CATEGORIES (BLACK OR AFRICAN AMERICAN ALONE, NOT HISPANIC OR LATINO)
P12K	N/A	SEX BY AGE FOR SELECTED AGE CATEGORIES (AMERICAN INDIAN AND ALASKA NATIVE ALONE, NOT HISPANIC OR LATINO)
P12L	N/A	SEX BY AGE FOR SELECTED AGE CATEGORIES (ASIAN ALONE, NOT HISPANIC OR LATINO)
P12M	N/A	SEX BY AGE FOR SELECTED AGE CATEGORIES (NATIVE HAWAIIAN AND OTHER PACIFIC ISLANDER ALONE, NOT HISPANIC OR LATINO)
P12N	N/A	SEX BY AGE FOR SELECTED AGE CATEGORIES (SOME OTHER RACE ALONE, NOT HISPANIC OR LATINO)
P12O	N/A	SEX BY AGE FOR SELECTED AGE CATEGORIES (TWO OR MORE RACES, NOT HISPANIC OR LATINO)

Using the former yields aggregate population counts that exceed the population total for each geography because it ought to double-count non-white Hispanic or Latino individuals. The latter yields matching counts.

[ericmanning]

Might be related to #138

[1beb]

Thank you, getting these tables right is a challenge sometimes. I'm checking in with the team on this one.

[ericmanning]

The Census Bureau did not publish the P12J through P12O information in any summary file for the 2010 census. So if my suggestion is correct, then you can't actually tabulate age and sex by block and H/L and race for 2010.

[1beb]

Correct, we need to setup warnings so that people use an older version of the package as it's not backwards compatible with pre-2020.

[ericmanning]

FWIW, the package has always used the current set of variables for sex and age, which are incorrect -- so (correct me if I'm wrong) any version will produce inaccurate estimates for 2010 if age OR sex is TRUE

From wru-0.1-12/R/census_geo_api.R,

if (age == F & sex == F) {
    num <- ifelse(3:10 != 10, paste("0", 3:10, sep = ""), "10")
    vars <- paste("P0050", num, sep = "")
  }

  if (age == F & sex == T) {
    eth.let <- c("I", "B", "H", "D", "E", "F", "C")
    num <- as.character(c("01", "02", "26"))
    vars <- NULL
    for (e in 1:length(eth.let)) {
      vars <- c(vars, paste("P012", eth.let[e], "0", num, sep = ""))
    }
  }

  if (age == T & sex == F) {
    eth.let <- c("I", "B", "H", "D", "E", "F", "C")
    num <- as.character(c(c("01", "03", "04", "05", "06", "07", "08", "09"), seq(10, 25), seq(27, 49)))
    vars <- NULL
    for (e in 1:length(eth.let)) {
      vars <- c(vars, paste("P012", eth.let[e], "0", num, sep = ""))
    }
  }

  if (age == T & sex == T) {
    eth.let <- c("I", "B", "H", "D", "E", "F", "C")
    num <- as.character(c(c("01", "03", "04", "05", "06", "07", "08", "09"), seq(10, 25), seq(27, 49)))
    vars <- NULL
    for (e in 1:length(eth.let)) {
      vars <- c(vars, paste("P012", eth.let[e], "0", num, sep = ""))
    }
  }