1980 Census Age Data

NHGIS
1

I’m doing an analysis involving tract level data from the 1980 census, and I noticed a weird discrepancy between table NT10A (age of all persons at tract level in 1980) and tables NT7, NT8, and NT1A (respectively race, Spanish origin, and total persons at tract level in 1980). The sum of persons in all age brackets in NT10A doesn’t seem to match the count of persons listed in NT1A, or the sum of persons across all categories in NT7 and NT8.

image
image1397×860 9.96 KB

The errors aren’t usually so large, but I haven’t been able to find any reason that this should be the case at all, especially since it seems all of the tables are based on the same summary tape file. It’s admittedly entirely likely that I’m overlooking something, though. Does anyone have any insights into what’s going on here?

2

Dear Ashlynn,

In 1980, the Census Bureau used suppression to protect the confidentiality of respondents. As a result, you will observe situations in the data like you illustrated in your post.

For 1980, the following population information was never suppressed (more details about suppression are available on pages 19-29 of this technical document:

  • Total Population
  • Race
  • Hispanic Ethnicity
  • Total Housing Units
  • Year-round Housing Units
  • Occupied Units
  • Vacant Year-Round Housing Units

Suppression for other population characteristics, such as sex and age, was applied in various circumstances. For STF1, suppression for these characteristics was applied when a geographic unit had fewer than 15 people. In your example, suppression was applied to the age counts but not the total population, race or Spanish Origin tables.

It’s even more complicated than that, however. When the Census Bureau published their digital data files for 1980, data storage was expensive. In order to minimize space, the Bureau published data for “split census tracts” - these are census tracts that are split by city boundaries. The same suppression rules applied to split tract data.

When we started NHGIS, we realized that we needed to create 1980 data for “whole census tracts”. This required us to aggregate data for split tracts to create whole ones. We used the suppression flags on the split tracts when we created the whole tract data. Thus, if one part of a whole tract was suppressed, we then suppressed data for the whole tract. Thus, when you see a suppression flag for a census tract with a large population, that particular data record was probably created by aggregating split tracts AND one split part had suppression applied. Essentially, one part of the split tract had a very low population (under 15) and its characteristics were suppressed.

We actually provide the “split tract” data that we used to build the “whole tract” data that you’ve been looking at. When you’re on the Data Options page of NHGIS, you can click the hyperlink under GEOGRAPHIC LEVELS. The following popup will open:

Screenshot 2023-06-28 at 3.24.25 PM
Screenshot 2023-06-28 at 3.24.25 PM966×556 26.9 KB

If you click the SHOW COMPOUND GEOGRAPHIC LEVELS hyperlink, you will see the list of all “split” geographic levels for the dataset (e.g., 1980 STF1). To get the split tract data, you’ll want to click the button as shown on this screen shot:

Screenshot 2023-06-28 at 3.24.50 PM
Screenshot 2023-06-28 at 3.24.50 PM967×683 47.8 KB

When you get the data for this level, you will be able to isolate the “split tracts” and compare their statistics. Many of the parts with large populations will not have suppression applied. If the small part is quite small (e.g., 10 people), then you may be able to just use the larger part for your project.

Sincerely,
Dave Van Riper
IPUMS NHGIS research scientist

3

Hello Dave,

Thank you for your explanation of how suppression shows up in the 1980 tables. I wanted to ask some follow-up questions based on my own experience working with 1980 tables (I’m trying to do an analysis of children 0-17 by race and ethnicity at the tract level). My problem is that I am finding tracts where crosstabbed population totals for Persons of Spanish Origin are greater than the same totals for Persons.

I have figured out that part of my problem has to do with “complementary suppression” in 1980. For example, in tracts where the American Indian population is suppressed for a particular age group, the Other Race population for that age group is also suppressed in the Persons universe (table NT12B). The Persons of Spanish Origin universe (table NT13B), however, only has three racial categories (white, black, and other), so its Other category does not get suppressed, since it’s not complementary to a suppressed cell. This results in tracts where Other Race for a given age group is suppressed in Persons but is present in Persons of Spanish Origin for that age group.

However, I am also finding census tracts in 1980 where the total population in a given age group within Persons (table NT12A) is suppressed, but the total population in the same age group for Persons of Spanish Origin (table NT13A) is not suppressed. In some cases, the non-suppressed cells in Persons of Spanish Origin have population values greater than 15, which added another layer of head-scratching for me. Why would suppression be applied at a more general universe but not in a subset universe for a single tract, especially if the value for the subset universe is greater than 15? Is it possible that the total population for a given age group in NT12A could be considered a “complementary” cell to a suppressed race-specific cell in a more specific table like NT12B?

Another more general question I have is about the suppression standard (<15 for a given cell). I am seeing many cells in both the Age and Race by Age tables with tract population values less than 15, but where suppression hasn’t been applied.

Any insights you can provide would be much appreciated.

Ethan McIntosh
BPDA Research Division

4

Hi Ethan,

Could you provide me with a few example tracts where you are observing the counts you describe in your post? Having those examples will help me with my research and answer. It may take a couple of days because it’s a complicated topic!

Yours,
Dave

5

However, I am also finding census tracts in 1980 where the total population in a given age group within Persons (table NT12A) is suppressed, but the total population in the same age group for Persons of Spanish Origin (table NT13A) is not suppressed. In some cases, the non-suppressed cells in Persons of Spanish Origin have population values greater than 15, which added another layer of head-scratching for me. Why would suppression be applied at a more general universe but not in a subset universe for a single tract, especially if the value for the subset universe is greater than 15? Is it possible that the total population for a given age group in NT12A could be considered a “complementary” cell to a suppressed race-specific cell in a more specific table like NT12B?

What you’re observing is the result of how we at NHGIS rolled up data from census tract parts into whole census tracts. An example will hopefully help explain what you’re seeing.

I created an NHGIS extract from 1980 for census tracts and census tract parts (tract_080 geographic level, which is what is shown in the screen shot above). This first table below shows you the census tract part records for tract 0201 in Autauga, Alabama. The total suppression flag applies to tables such as Age (NT12A). The Spanish origin flag applies to tables with the Spanish origin universe such as Age (NT13A). The <5, 5-17 year, 18-64 years, and >=65 years columns contain counts from NT12A. The Sp. <5, Sp. 5-17 year, Sp. 18-64 years, and _Sp. >=65 years columns contain counts from NT13A.

State County County Subdivision Place Tract Total Suppression Flag Spanish Origin Flag Total Pop < 5 5-17 year 18-64 years >= 65 year Spanish Pop Sp. < 5 Sp. 5-17 year Sp. 18-64 years Sp. >= 65 year
01 001 020 1415 0201 0 0 1579 127 387 900 165 19 3 4 12 0
01 001 020 9999 0201 1 0 5 0 0 0 0 0 0 0 0 0

Suppression doesn’t apply to the first row in the table because the Total Population and the Spanish Origin Population both exceed 15. Thus, we observe Age counts for Total and Spanish Origin. Total Population suppression applies to the second row because its total population is only 5. The Spanish Origin suppression does not apply the the Age counts to the second row because there are zero persons of Spanish Origin in that tract part.

When NHGIS created the Age counts for the whole census tract 0201, we did our best to respect the suppression flags. The result of our processing yields the subsequent table.

State County Tract Total Suppression Flag Spanish Origin Flag Total Pop < 5 5-17 year 18-64 years >= 65 year Spanish Pop Sp. < 5 Sp. 5-17 year Sp. 18-64 years Sp. >= 65 year
01 001 020 1 0 1584 0 0 0 0 19 3 4 12 0

Since the second row has a suppression flag for Total Population (which applies to the Age counts), we put zeroes in the Age columns. Since neither tract part has a Spanish origin suppression flag, we just roll up the counts and place to sums in the Spanish Age columns. Thus, in this particular example, you observe age counts for the Spanish Origin group but no age counts for the Total Population.

I highly recommend downloading your tables of interest for census tract parts and inspecting your tracts of interest. You may decide that you disagree with our decision to apply the suppression flags during our roll-up process! Having the tract part dataset allows you to create your own census tract counts.

Another more general question I have is about the suppression standard (<15 for a given cell). I am seeing many cells in both the Age and Race by Age tables with tract population values less than 15, but where suppression hasn’t been applied.

The suppression standard does not apply to specific cells in tables; instead, it applies to the total number of people (or households or housing units) in a particular geographic area. For example, if a census tract contains 19 individuals of Spanish Origin, then the Bureau will tabulate and publish the Sex by Age table (NT13A) for persons of Spanish Origin. Many of the cells in NT13A will be smaller than 15, but suppression will not be applied to those cells because the tract contained more than 15 people of Spanish Origin. If a census tract contains 14 individuals of Spanish Origin, then the Sex by Age table will be suppressed.

– David Van Riper
IPUMS Research Scientist

6

Thank you for these clarifications!