Data is not aligning properly

Hi again,

I used the IPUMS raceeth data to collect information about racial and ethnic minorities as a proportion of individuals under 125% of poverty in Iowa. However, when I did this, the data did not align well with the data my organization has collected. For example, my organization calculated that approximately 30.5% of the population was African American, whereas IPUMS calculated this value to be 44.1%. I used the online SDA analysis and used the 5 year 2017-2021 ACS data. Does anyone have suggestion on how to get data that is more reflective to the results my organization is collecting? Would there be a significant difference in using the 5 year vs. 1 year ACS? Any insight would be greatly appreciated!

raceeth is a variable created by another user in the online analysis tool that uses responses from RACE and HISPAN; it codes respondents of any race who reported being Hispanic into a Latino category. For example, someone who identified as Black and Hispanic will be coded as Latino and will not be included in the non-Hispanic Black category. Using this variable in combination with POVERTY and STATEFIP in the online analysis tool (see screenshot below), I’m finding that non-hispanic Blacks form 9.4% of the proportion of individuals under 125% of poverty in Iowa. This corresponds to 43,548 individuals from a total of 114,523, or around 38% of the total non-hispanic Black population of Iowa. This is within the margin of error of the official Census ACS data table which reports that 38.4% (with a 2.3% margin of error) of the state population that identifies as only Black or African American earns less than 125% of the poverty line. I recommend reviewing the variable description for POVERTY to note how it is constructed to determine if it’s comparable to the data your organization collected.

1 Like

Using the IPUMS raw data extract for the 2021 ACS 5yr sample, I am getting essentially-identical estimates for the population of Black, non-Hispanic people under 125% of the poverty level in Iowa that @Ivan_Strahof obtains using SDA*. Black non-Hispanic people being approximately 9.4% of the under-125% poverty line population, and Black non-Hispanic people under-125% of the poverty line being approx. 38% of the Black non-Hispanic population in the state.

Since you asked about whether the ACS 1-year sample might obtain different estimates, the answer is the percentage does not change significantly. I get Black, non-Hispanic people under 125% of the poverty line (39,784) as 8.6% of that poverty-line population (462,964), and 35% of that race/ethnicity population (113, 459).

Based on this information, I would wonder if your organization obtains different numbers from the Census because the sample year is different than the ones we use; 2021 ACS 1-year estimates are slightly less than 5-year, perhaps it was from an earlier year? Another explanation would be your organization looked at a sub-sample the Black, non-Hispanic population.

*I get very slightly different numbers for the sub-125% of the poverty line population-- 43,072 Black non-Hispanic individuals from 455,634 people under 125% of the poverty line. The same percentages as Ivan, but I wonder why this population doesn’t perfectly line-up like the estimate of the Black non-Hispanic population. I imagine if the 1-year ACS with that SDA variable were used, the numbers I gave for that sample similarly might not perfectly agree but occupy the same ball-park.