Recent publications in population projection and forecasting research have called on scholars to broaden the methods available to scholars in order to make more accurate and detailed estimates of future populations. While a wealth of research has been devoted to population projections for large entities such as nations, less focus has been placed on making projections for more localized areas, such as metropolitans, cities, and towns. For community leaders, city planners, and a number of other stakeholders these more localized projections are essential for decision making regarding the future direction of their communities. In addition, even at the local area level, needs are not equal among different population groups. For example, racial and ethnic groups may have different population needs, such as language needs, and making projections by these groups may improve local leaders ability to supply those needs in the future.
While the demand for detailed population projections, such as for local areas and by racial and ethnic groups, is high, the availability of methods which can produce sensible estimates of future populations in such detail are lacking due to the difficulty of dealing with small population numbers. Population projection methods tailored to much larger populations cannot simply be applied to much smaller populations who fluctuate in their size at much faster, more unpredictable rate than larger populations. While some attempts have been made to apply traditional population projection methods to these small populations, the results have been mixed at best.
In an attempt to address this issue we present a new approach to making projections for detailed population groups, which allows for accurate population projections even in the presence of rapid changes in population in the past. This method builds off of traditional demographic understanding of population change to make projections of populations for small population groups which resemble the greater populations they are a part of. By using statistical methods to capture general patterns of population change , we create population projections for very detailed populations which tend to be more accurate than projection methods designed for large populations. In the figure we demonstrate an application of this method, which we refer to as the multi-stage smoothing method, to make projections of the King County population. While projections are made at a very detailed level, by age, sex, race, and ethnicity for a particular census tract, the projections can also be compiled into a more aggregate level as seen in the figure. We demonstrate here that our new multi-stage smoothing approach produces more reasonable estimates of the population in King County when aggregated by race and ethnicity than more traditional methods such as the Hamilton-Perry (HP) method.
In addition to testing this method against other projection methods to demonstrate its effectiveness, we have also created an R library for other researchers and those interested in population projections to easily apply this method to their own use cases.
Computation & Reproducibility
All code necessary to implement the methods and reproduce the figures and results in Tract-level projections by age, sex, race and ethnicity for local policy makers has been archived as of publication on November 3, 2021 by the Population Dynamics Lab here: UW PopLab Github.
(2021, November 3). Tract-level projections by age, sex, race and ethnicity for local policy makers. The Download, Population Dynamics Lab. https://population-dynamics-lab.csde.washington.edu/the-download/2021/11/03/accurate-projections-for-local-policy-makers/ [Accessed January 21, 2022].