The U.S. Census Bureau today released the first of its 2010-2020 intercensal population and housing unit estimates products. These estimates, which are produced once a decade, adjust the Vintage 2020 postcensal population and housing unit estimates to make them consistent with the April 1, 2020, counts from the 2020 Decennial Census. The 2010-2020 intercensal estimates are the recommended data source for population estimates in these years and should be used in lieu of the Vintage 2020 estimates whenever possible. One of the main benefits of the intercensal estimates is that several decades of data, including the latest vintage of postcensal estimates, can be linked together into one consistent time series.

Today’s release includes total population estimates for the nation, states, counties, metropolitan/micropolitan statistical areas, incorporated places, minor civil divisions, Puerto Rico and its municipios along with housing unit estimates for the nation, states and counties. Population estimates by demographic characteristics will be released in 2025.

Background

There are two inputs to intercensal estimates: Vintage 2020 population and housing unit estimates, also known as postcensal estimates, and 2020 Census counts. The Vintage 2020 estimates begin with the 2010 Census count and draw on information about components of change to produce estimates on July 1 of each year of the decade (2010 to 2020).  Components of population change include births, deaths, and domestic and international migration. Components of housing unit change, meanwhile, include new residential construction, new mobile homes and housing unit loss.

The 2020 Census values used as the end point for the intercensal population estimates come from a privacy-protected file of subcounty counts produced for the Population Estimates Program (PEP) to use as the base for their annual vintage estimates production. This file does not contain demographic detail on the population, which limits the level of detail included in the current release of the 2010-2020 intercensal estimates to population totals. The “modified race file,” which redistributes responses of Some Other Race from the census into the 31 race alone or in combination categories specified by the Office of Management and Budget, will be used as the end point for intercensal estimates by demographic characteristics.

Postcensal estimates are not always consistent with census counts. Differences between the two can come from several sources: errors in the censuses, errors in the data and methods used to estimate the components of change, geographic boundary changes, and changes in the measurement and coding of variables across censuses.    

Intercensal Estimates Methodology

The gap between the 2020 Census and the Vintage 2020 Estimates on April 1, 2020, is referred to as the error of closure (EOC). The method for producing intercensal estimates distributes the EOC across the decade, modifying the vintage estimates so they are consistent with both the April 1, 2010, Census count at the start and the April 1, 2020, Census count at the end. Importantly, intercensal estimates do not adjust the components of change in the vintage estimates, just the population and housing unit values.

For the 2010-2020 intercensal estimates, we are changing the model we use to distribute the EOC. In prior decades, the assumption was that the EOC should be distributed exponentially across the vintage time series. This implies that the cumulative error of our inputs is smaller at the beginning of the time series and larger at the end, and the annual error is always positive or negative. This assumption has little importance when the EOC is relatively small compared to the population, but as the EOC increases, the time series becomes more distorted. The time series sags in the middle when there is a large positive EOC and bows in the middle when there is a large negative EOC. This decade, we are using a linear distribution approach that preserves the pattern of the original estimates.

There are several assumptions built into our method for producing intercensal estimates that data users should keep in mind. Our approach assumes that the enumerated values from the 2020 Census are the correct end point for the time series on April 1, 2020. Because of the disclosure avoidance protections used in the census this decade, values in the intercensal file may differ from with those published in other 2020 Census data products except for housing units, whose counts were invariant in the census.

Further, evaluations of the 2020 Census have revealed undercounts and overcounts for different populations. Our estimates do not adjust the census count and, thus, treat it equally for all geographies. That is to say, the 2020 Census is regarded as if it is a more accurate representation of the actual population than the estimated value for that date. Similarly, our method assumes that the trajectory of change in estimates across the decade as represented in the vintage estimates is correct and shifts it up or down to align with the new end point. Inherent in this approach is a risk that we are maintaining a pattern of change across the decade that does not reflect reality for a particular area or subset of the population. An exception to this occurs in some places where we are unable to preserve the time series pattern from the decade because doing so produces negative values. In these cases, linear interpolation is used to produce intercensal estimates for the decade. For these places, data users interested in year-to-year changes occurring throughout the decade might be better served using the Vintage 2020 population estimates.

There are also some areas that experienced boundary changes throughout the decade and although we were able to account for many of these, there are likely some places that are reflected in one of our input data sources but not in the other. For cases where this occurs, some of the difference in the error of closure that we have distributed across the decade will be due to the boundary changes. Data users should exercise caution when using intercensal estimates for places where they are aware this has happened.

Conclusion

The intercensal estimates for 2010-2020 released today represent the final estimates the Census Bureau will produce for this decade. Postcensal estimates for a particular area can change from year to year within a decade as inputs are revised, but the development of the intercensal adjustment following the release of a decennial census marks the final change to those numbers.  Intercensal estimates create a consistent time series of data that are an essential tool for understanding population dynamics over time. Although there are some limitations to the data, these estimates provide useful insights for planning and policy decisions and serve as valuable inputs to demographic studies.

To access the intercensal estimates, go to Intercensal Estimates of Total Population and Housing Units: 2010-2020.