What Has Changed with Detailed Asian Group Data for the 2020 Census?


By Howard Shih

On September 21, 2023, the Census Bureau released the population counts for detailed race and ethnic groups from the 2020 Census. This product — called the Detailed Demographic and Housing Characteristics File A (Detailed DHC-A) — is comparable to the population count, age, and sex data from the Summary File 2 product from the 2000 and 2010 Censuses. In the fall of 2024, the Census Bureau will release the Detailed Demographic and Housing Characteristics File B, which will contain detailed household counts and housing tenure (own or rent) data that is comparable to other parts of the Summary File 2. Due to delays in processing data from the 2020 Census, the Census Bureau chose to split up the population count and household count data into the A and B files so that population count data could be released earlier. For comparison, the 2010 Census Summary File 2 was released in July of 2012. The Detailed DHC File series provides researchers and community advocates with the most detail on Asian ethnic groups and is tied directly to the allocation of federal funding programs.

This blog post will discuss the changes being implemented for this 2020 Census data release that impact the accuracy and utility of the detailed race and ethnicity data. The post will conclude with several recommendations on how to best use the new data.

Changes to the 2020 Census that Impact Accuracy

The Census Bureau made major changes in how they approach privacy protection in the 2020 Census. Just like how someone can protect their privacy either by not going out in public or by putting on a disguise, the Census Bureau uses two ways to protect the privacy of personal data from the 2020 Census. First, the Detailed DHC series of files will contain fewer data categories than in the past. For example, median household size will not be reported in the Detailed DHC series as it was in the Summary File for the 2000 and 2010 Censuses. Limiting the amount of data released protects privacy by not letting the data out in public.

Second, the Census Bureau disguises the data for the Detailed DHC series of files using an algorithm for privacy protection that is distinct from the one used for prior 2020 Census releases such as the Redistricting Data (PL 94–171) Summary File and the Demographic and Housing Characteristics File (DHC).1 This algorithm, called SafeTab-P, randomly changes data released in the Detailed DHC-A to protect the privacy of individuals and households. The Census Bureau has previously changed decennial census data when released to prevent the disclosure of personal information by swapping or suppressing data in release files. However, the SafeTab-P algorithm makes the Detailed DHC files unique from other 2020 Census and Census Bureau products in the following ways:

  • First, summing up populations for smaller geographies (like counties) to form a larger geography (like a state) will not match the reported population for the larger geography, unlike other 2020 Census and Census Bureau products. For example, adding up the total Asian Indian population in all 21 counties of New Jersey results in 415,496 individuals, while the total Asian Indian population for New Jersey is reported as 415,342. For previous censuses, the total would be the same.
  • Second, adding up categories like age, sex, detail groups, or geographies (such as census tracts or counties), will result in less accurate data because of the way the SafeTab-P algorithm randomly changes the data. This is the opposite of our previous experiences with Census Bureau data products. For example, other 2020 Census products are internally consistent, so the random changes to protect privacy are somewhat linked together. The oversimplified explanation is that changes in one area are often balanced somewhere else so that the overall population totals remain the same in previous Census products. However, the random changes (variance) made to the Detailed DHC data are independent of each other and that variance will accumulate as one adds more data points together.
    As a practical example, for prior data products, the Census Bureau has recommended adding together data from smaller geographies for more accurate counts and estimates — like adding up several census tracts to describe a neighborhood more accurately. For the Detailed DHC series of data products, the Census Bureau states that adding together census tracts will not result in more accuracy and to proceed with caution. Instead, the Census Bureau recommends for the Detailed DHC files adding or subtracting the fewest number of data points to achieve the desired results.

In exchange for less accuracy, the new privacy protection methods have allowed the Census Bureau to publish count data for many more detailed Asian groups than before. In 2010, 22 distinct detailed Asian groups received nationwide population counts. By contrast, in 2020, there were 41 distinct Asian groups, (47 total if “other” group categories are included) and 5 new regional groups (Central Asian, East Asian, South Asian, Southeast Asian, and Other Asian) that received nationwide counts.

Recommendations on How to Use the Detailed Race and Ethnicity Data

With these new limitations in mind, there are some important steps to take when using the new detailed race and ethnicity data in the Detailed DHC-A.

  1. Recognize that the Census Bureau has changed how much data is available to protect privacy. For the 2010 Census and earlier, if a population group met the criteria for publishing data, then they received the same number of age-by-sex categories in data tables. For 2020, the amount of age-by-sex data available for each group depends on the population size of that group. The smallest groups will only have total population data, not age or sex data. For larger population groups, age-by-sex tables will be available, with the widest range of age categories available only for the largest population groups. This limits how comparable smaller population groups are to larger population groups.
    For example, if we compared the child populations of Korean Americans in Jersey City and Fort Lee, NJ, we would only have the population under 18 years of age for Jersey City. But we would have the preschool (under age 5) and the school-age populations for Korean Americans for Fort Lee. This is because Jersey City’s Korean American population of 2,577 received Table T02001 with 4 age categories, including only a category for under 18 years. Fort Lee’s Korean American population of 9,783 received Table T02002 with 9 age categories, including under 5 years and 5- to 17-year-old categories.
  2. The population data for smaller geographies will not total up to match the data for larger geographies. For example, adding up the population of Burmese Americans for each state and the District of Columbia will not exactly match the population of Burmese Americans for the whole nation.
  3. Because the accuracy of the Detailed DHC-A data decreases the more individual data points are added together, the Census Bureau recommends removing or adding together as few groups as possible. The Census Bureau provides some examples in their technical documentation:
    – “To create a count for the Pacific West states, add together the counts for Alaska, California, Hawaii, Oregon, and Washington rather than subtracting the 46 other states and state equivalents from the national counts.”
    – “To create a count for Arizona counties that are majority urban, remove Apache, Graham, Greenlee, Navajo, and Santa Cruz counties from the Arizona state total (which involve 6 geographic regions) rather than adding together the ten majority urban counties.” Note that there are 15 counties in Arizona in total.
  4. Keep in mind the new race and ethnicity coding for the 2020 Census when comparing race and ethnicity data from 2010 and 2020. For example, Sikh Americans were previously tabulated as Asian Indians in previous censuses but now are tabulated separately in 2020.
  5. Extremely small populations (around one thousand individuals or less) will more likely see extreme swings in population counts across various levels of geography if their data was not already suppressed. Use this data only if it matches the real experiences of the community in question. For example, the Kyrgyz American population in New York City was reported as 914, and the total population for the 5 boroughs was 940, for a differential count of 26 and a percentage difference of 2.8%. Similarly, for the Kazakh American population, the count was 1,634 for New York City as a whole, and the five boroughs total was 1,664, for a differential count of 30 and a percentage difference of 1.8%. Meanwhile, for larger groups, like Asian Indians and Chinese, the percentage difference was 0.01% or less. The population counts for these examples were for the population group alone (single reported race or ethnicity).


The new changes in privacy protection that the Census Bureau has implemented for the release of the detailed race and ethnicity data from the 2020 Census have resulted in many more detailed groups than before. There are 47 detailed Asian groups and 5 new Asian regional groups that have national counts.

For this expanded coverage of detailed race and ethnicity data, we have sacrificed internal consistency of the population totals across geographic levels. Also, due to the independent way that random changes to the data protect privacy, adding up geographic areas to form custom analysis areas will result in less accurate data, not more — as was the case in earlier censuses. The impacts are felt most for small populations. We strongly advise that people using this data work directly with the communities they are characterizing to ensure that the data matches the lived experiences of those communities.

Howard Shih is the Assistant Director of Quantitative Research at Asian Americans Advancing Justice | AAJC.

Asian Americans Advancing Justice | AAJC has a mission to advance the civil and human rights of Asian Americans and to build and promote a fair and equitable society for all. Visit our website at advancingjustice-aajc.org.



Advancing Justice – AAJC
Advancing Justice — AAJC

Fighting for civil rights for all and working to empower #AsianAmericans to participate in our democracy.