by Jan Vink
Last month 50 years ago, Congress made a change to 13 U.S. Code § 181 and established the obligation of the Census Bureau to produce data on total population and population characteristics during the intervals between each census of population. These estimates are used in federal funding allocations, as survey controls, as denominators for vital rates and per capita time series, and as indicators of recent demographic changes for planning and other use cases. The Population Estimates Program (PEP) within the Census Bureau produces these estimates with a variety of releases throughout the year.
Title 13 U.S. Code § 181 doesn’t define what Congress meant by “estimating the total population,” but the Census Bureau’s estimates of the total population should be interpreted as an annual average number of people living in an area. And a commonsense definition of living somewhere is that one spends most of their time there, which is very comparable with the definition of usual residence as defined for past decennial censuses. Congress also did not define “population characteristics,” but for these fifty years, population characteristics included age, sex, race, and more recently ethnicity.
I have been part of the FSCPE for almost 20 years, which makes the 2030 Census my third Census as a well-informed outside stakeholder. I have always appreciated the expertise and thoughtfulness that goes into preparing for the next Census, and it pains me to see that the Census Bureau is proposing to shoot itself in the foot this time around.
Annual population estimates start with a base population and estimate how that population has changed since the date of the base. A high-quality base population is essential for the population estimates and the decennial census has always provided such a quality base.
After each census, the base is reset to start fresh with a new decade of postcensal estimates. For example, the last estimate before the 2020 Census results were published estimated that New York City had 8.25 million residents, 1.24 million of them Non-Hispanic Black alone on July 1, 2020. Now that the base has been reset and the 2020 Census results are fully integrated in the base, this estimate is revised to an updated estimate of the total of 8.75 million residents (+6%) on July 1, 2020, including 1.40 million Non-Hispanic Black alone people (+13%). Many factors play a role in these differences, including changes in the accuracy of the count over time.
If the proposals laid out in the FRN are accepted as written, they will have huge impacts on the applicability of the 2030 Census as a base for the population estimates going into the next decade.
The proposed rules exclude some people from the count that live here based on the commonsense definition of living here. If those people are removed, they weaken the base population needed for PEP. To maintain an accurate total count of the population, PEP would need to find a way to add those people back into the estimates.
Defining usual place of residence by introducing a reference period (Jan 3 – April 1) and counting people where they spend most of the time during this period is very different from the commonsense definition of usual residence for many people. I happen to like the New York winters, but they are not for everyone, and some choose to spend the winter months in a warmer place. These people are still New Yorkers and should be counted in New York. Similarly, should a student spending their winter break with a friend’s family count that time towards a residence in that household? The rules as proposed are very confusing and should have been thoroughly tested, in the way the Census Bureau normally does in preparation for a Decennial Census.
This proposed definition of usual residence is very different from the commonsense definition used in the population estimates and would potentially disqualify the decennial census results as a base for the next decade’s estimates. Aligning the definition of usual residence from the population estimates with the proposed redefinition is also not an option as the estimates reflect an average population over a whole year and not just the winter months.
In addition, eliminating demographic questions like race and ethnicity would also negatively impact the population estimates. The reset after each decade is necessary to keep up with changes in definitions and counter imperfections that slip into the estimates over the decade. There are differences by race and ethnicity in health outcomes, labor force participation, poverty rates etc. and the population characteristics estimates are the best source for the denominators of many of those rates.
The Preliminary Regulatory Impact Analysis that accompanied the proposed rule notes that “Omitting race and ethnicity from the 2030 Census questionnaire would require the development of estimates from a base population derived from outdated and/or less accurate data and would deteriorate the accuracy of the estimation method itself” but there is no further analysis that tries to quantify this cost. However, as a demographer I know that the cost is substantial. Local governments and communities, especially already disadvantaged communities will bear the costs of bad data caused by an incomplete census that will change everything from funding allocations to public health tracking, etc.
Asking about citizenship and legal status on the questionnaire would likely cause an undercount of the population. Many households might not like or trust what the government is doing with their responses. A direct purpose of asking this question mentioned in the proposal is excluding people from the count. Why would someone answer when the result is that he/she will not be counted? Many families with some family members that are included in the count as proposed but also some that would be excluded might choose not to fill out a questionnaire at all. This will lead to undercounts. This is a problem for many use cases for Census data, including its use as base for the population estimates.
I am confident that PEP will find a way to keep producing estimates of total population and characteristics, just like they did when there were delays in the 2020 Census. But counting everyone, using a commonsense definition of usual residence and asking demographic questions while leaving out a question about citizenship and legal status would definitely result in higher quality estimates.
Jan Vink is Chair, Federal State Cooperative on Population Estimates (FSCPE) steering committee at the Cornell Program on Applied Demographics
Opinions expressed are the author’s