Don’t Bet the Farm on Population Projections

Source: StatChat blog

The Demographics Research Group at the University of Virginia is the entity tasked with making official population projections for the Commonwealth of Virginia and its localities. Their projections feed into all manner of planning documents across the state. If the projections are off, so are the forecasts for school attendance and transportation demand. Getting the numbers right is a big responsibility.

Hamilton Lombard, a research specialist for the group, assumes an appropriate air of humility regarding long-range projections.

Forecasting population change, like forecasting the weather, is complex, requires one to make assumptions about the future, often based on past trends, and is rarely spot on,” he writes in the StatChat blog. “Because population projections are less familiar to the public, projections are often treated as something closer to a fact, rather than a forecast that can and likely will change. Unfortunately, not understanding population projections can lead to much larger problems than a rained out barbecue.

In the chart above, Lombard traces the history of state population projections for the year 2000 beginning in 1975. The 25-year projection was off by a significant margin. But, as a rule, shorter-term projections are more accurate, and the 10-year projection hit very close to the mark.

Numbers tend to be less accurate for localities because demographic trends tend to be more volatile. As an extreme example, Lombard cites, projections made of Bath County’s population jumped around 1980 when the lightly populated county experienced an influx of construction workers to build the Bath County pump storage facility. “Because of the temporary rise in Bath County’s population,” Lombard writes, “the projections expected the county’s population to keep growing, even after the 1990 census showed that most of the power plant construction workers had left.”

Bacon’s bottom line: Forecasting increased population for Bath County by projecting a trend line based on a temporary influx of construction workers was utterly foolish. Someone should have been strung up by the thumbs. Fortunately, not much was at stake (well, not much for anyone except, perhaps, the residents of Bath County). But sound planning for billions of dollars of transportation and infrastructure investments depends upon reliable population estimates.

For the 50-year reign of suburban sprawl, forecasters could reliably predict a shrinking of Virginia city populations and growth in surrounding suburban counties. Then an inflection point occurred in the mid-2000s when population and business began reversing the trend — moving from the burbs into core urban areas. Straight-line projects based on 2000 population trends would have gotten the numbers very wrong. I would urge Lombard to reconstruct the history of population projections for the year 2020 projections going back 25 years. I suspect he would find a much wider gulf between forecast and reality than in the graph shown above.

As long as the economy is in a steady-state condition, predictions tend to be reasonably accurate. When inflection points occur, forecasts go widely astray. Today demographers must ask, how long will the urban revitalization movement last? Will cities continue to gain population? Will the growth rate of counties continue to slow? Answers to those questions are beyond the ability of demographers to predict, for they depend upon the willingness of cities and counties alike to adopt policies that promote the kind of denser, mixed-used development that can accommodate growing populations.

So, as Lombard counsels, understand the limitations of long-term demographic projections. If demographers could predict the future with 100% accuracy, they wouldn’t be demographers — they’d be making a killing on Wall Street.