The Census Bureau released its first set of Vintage 2021 population estimates today covering population changes over the year ending 30 June 2021. The estimated population grew by 392,665, or 0.1%, the "lowest rate since the nation's founding", according to the Census Bureau.

The slowdown in population growth is part of a continuing trend—estimated immigration and fertility rates have fallen in recent years. The slowdown is also related to the pandemic, immigration fell to almost nothing in the early months of the pandemic, deaths soared, and births dropped, although not nearly as much as some expected.

The natural increase estimate came in as expected—it almost perfectly aligns with CDC source estimates published in September. Births fell to 3.58 million from 3.77 million in 2019 and 3.75 million in 2020; deaths increased to 3.43 million from 2.85 million in 2019 and 3.07 million in 2020.

The Bureau of Labor Statistics (BLS) will incorporate Vintage 2021 population estimates into the household employment data in the January Employment Situation report. Once it does, the data will officially resolve a perplexing data anomaly. The BLS produces an adjusted household employment series similar to the payroll employment concept. The adjusted household series excludes self-employment and includes second jobs. In theory, the payroll and the adjusted household employment series should align over time. Since 2010, they have not. From the first quarter of 2010 to the fourth quarter of 2019, payroll employment grew by 21.9 million, while adjusted household employment increased by 20.7 million. The 1.3 million difference will likely shrink to a small number explainable as statistical noise in the January Employment Situation release. Since the BLS does not revise the household estimates, it will show up as a discrete jump of over one million in adjusted household employment in January 2021.

Not included in this update (and eagerly awaited) are the intercensal estimates for 2010-2019. These are expected to come out in early 2023. The challenge in putting together this data set will be allocated across ten years and fifty states the 2.1 million additional immigrants that the 2020 Census uncovered.

How good is the Vintage 2021 estimate for net international migration?

Because the quality of source data deteriorated during the pandemic, it's not as good as the estimates for 2010-2020, which undercounted immigration by 2.1 million over 2010-2020, according to decennial census data. The Census 1 April 2021 count was 2.1 million higher than expected. Since births and deaths are well measured, net international migration accounts entirely for the higher than expected number.

How accurate was the Census 2020 estimate for net migration over 2010-2020?

It was likely an undercount. New York state's experience illustrates this uncertainty: To avoid losing two congressional seats, New York state, according to news reports, spent $30-$40 million on census advertising, texts, events, and, crucially, finding addresses missing from the US postal databases. The investment partially paid off; instead of losing population, its count came in 818,876 higher than the previous estimate from the Vintage 2020 data set. New York lost one congressional seat and came up 89 short of losing none. That seat went to Minnesota, a state which also invested heavily to increase the census count. Arizona, on the other hand, which spent little to raise its numbers, saw its count come in 242,398 below expected; it fell short of gaining a seat by about 80,000. That New York and Minnesota were able to increase the population count by investing resources to locate people suggests that the 2020 Census undercounted. It is uncertain how large that undercount might be.

Bottom line

International migration is hard to measure. The Census estimates should be seen as approximations.

Posted 21 December 2021 by Patrick Newport, Executive Director, US Economics, S&P Global Market Intelligence