Whether they come from Dr. Anthony Fauci, Warren Buffet, a politician, a bombastic sports commentator or even just a local weather forecaster, our natural inclination is to see how accurate predictions are.
But the people making the predictions themselves are often the most curious of all.
Such was the case for Qian Cai and Shonel Sen, demographers at the University of Virginia’s Weldon Cooper Center for Public Service, who had prepared population projections for the 50 states and the District of Columbia and were eagerly awaiting the 2020 Census results.
Their work marked the first time any organization other than the U.S. Census Bureau had produced projections for all 50 states.
Recently, Cai and Sen compared their projections to the actual census counts, and, let’s just say, came away quite pleased.
UVA Today caught up with Cai and Sen via email to learn more.
Q. Can you give UVA Today readers, who may not be that familiar with the work that you do, an idea of why these projections are so important, as well as a couple examples of how the projections were used over these last two or so years?
A. In the decade of the 2010s, the U.S. Census Bureau stopped producing state-level population projections, but many federal agencies, state governments, businesses and nonprofits rely on a consistent set of state-by-state projections for their planning and research purposes. In order to fill this void, our team decided to step in and develop these projections pro-bono, for each state and Washington, D.C., over 2020, 2030 and 2040.
The national 50-state projections are widely used and have practical applications in health, economics, politics and environmental research, to name a few. Our projections, for example, were used by the Centers for Disease Control and Prevention, the U.S. Environmental Protection Agency, the U.S. Government Accountability Office and the National Park Service.
Q. What were your key findings, when comparing your projections to the actual 2020 census count?
A. Nationally, we projected the 2020 U.S. population to be 332,527,548, which is a bit more than the Census count of 331,449,281, representing a difference of 0.33%. Our projection is slightly closer to the actual count than the U.S. Census Bureau’s [preliminary] projection of 332,639,000.
Some of the key findings when we compare our projected population numbers to the 2020 Census head counts are the following:
- Of all 50 states, the absolute difference between our projected population and the Census count ranges from 0.06% for Kansas to 3.4% for Idaho. All differences are under 3.5%.
- Thirty states are within plus-or-minus 1% difference from the Census count, including 16 states that are within plus-or-minus 0.5%.
- The average difference of the 50 states is 1.1%. This is very low. According to the Census Bureau’s empirical evaluations of its own projections in previous decades, the typical average difference is between 3% and 5%.
Q. How did it make you feel when you realized your projections were so accurate?
A. We are very pleased with the results. Since it was the first time any organization other than the U.S. Census Bureau has produced projections for all 50 states, this evaluation not only provides a concrete validation for our work, but more importantly, affords greater confidence for data users that our projections are of high quality.
Q. The average difference in your projections of the states was just 1.1%, while Washington, D.C., had a 6.2% difference. This isn’t the first time D.C. has been an outlier. What has made projecting D.C. so challenging?
A. Projecting the future population for Washington, D.C., is definitely a challenge, and it had the highest difference of 6.2% for our 2020 projected total. Even back in 2005, when the Census Bureau released its final set of 50 state projections for the 2010 population, D.C. was off by 12%, also higher than all 50 states.
The simplest explanation is that the District is not a state, and so it does not behave like one in our projection’s models. The methodology we applied is more appropriate for larger geographies. Even though D.C. has a population size bigger than a few states, it is not composed of counties, cities and towns, but rather is a jurisdiction within a larger metro area. Factors such as high density and quick population turnover also contribute to the more transient, fluid nature of the population in the nation’s capital.
Q. Which state did you project most accurately, and on the flip slide, which was your least accurate? What were your biggest surprises?
A. Projections for Kansas were the most accurate, lower by only 1,700 residents, or minus-0.06%. D.C., on the other hand, is our least accurate, higher by 43,000 residents, or 6.2%.
Of all 50 states, 21 were projected higher than the actual counts, and 29 were projected lower.
One interesting observation is that we over-projected most Southern and Western states and under-projected New England and the Midwestern states. The fast-growing Sun Belt region grew more slowly than anticipated, and the slower-growth states in the northeast and Midwest grew faster than expected.
Q. Your projections here in Virginia were only 0.27% higher than the Census count. Does living and working in the state give you any added insight, kind of like a home-field advantage, that you don’t have when you make projections in states out West? Or is this all 100% data-driven?
A. Our projection for Virginia was highly accurate, but it was not treated any differently; we applied the same methodology to all 50 states and D.C. However, our knowledge and insight of the commonwealth and its detailed demographics do provide us a home-field advantage when we develop the locality-level projections for Virginia’s individual cities, counties and large towns.
Q. What are some lessons learned from this process that you think you can use to your benefit the next time around?
A. Our experience projecting an inherently unknown future has taught us several valuable lessons. The biggest takeaway is that carefully constructed, simple methodologies can yield more reliable results than some of the elaborate and complicated modeling techniques. High-quality input data is critical to generate accurate projections. Finally, projections need to be updated every two or three years to better capture the latest population trends.
Q. Is there anything else you’d like to add?
A. Our latest national population projections were produced in 2018, prior to the pandemic. The pandemic, however, is likely to have a long-lasting impact on population change and distribution in this decade and beyond. Births may continue to drop further, and migration patterns may alter.
We look forward to capturing these, and more, demographic changes as we prepare to produce the next round of population projections out to 2050.