One of the great economists of the last century, John Kenneth Galbraith, once quipped that “the only function of economic forecasting is to make astrology look respectable.” The same can be said for pretty much anything that we foolishly insist on forecasting – weather, stock prices, the next recession, and if we are honest, demographics. This is especially true of small area demographics, largely because of the rather annoying human tendencies to wander about.

Superficially, projecting population is a simple process. Take the current population, add in the births, subtract the deaths, and track those who didn’t stay put. At a national scale, this seems like it should work well over the short run – say a decade since fertility and mortality rates by age and sex generally change slowly. Since our ‘known’ data is always in fact historical data, even current estimates are short-term projections. We don’t call them that, because the term is sufficiently short that we don’t think much has changed and that the change which has occurred we have accounted for. The major source of error in the short run is net migration, even at a national level – evidenced by the substantial discrepancy between the population estimates released by the census bureau in the summer of 2020 and the actual census counts.

At its core, the issue with projecting pretty much anything is that change tends to have two distinct components: one continuous, the other (in mathematical terms) catastrophic. Continuous change is reasonably predictable – so given the current age/sex distribution of the population, we can be reasonably confident in predicting how many deaths will occur over the next five years.  But what if there is something unexpected which disrupts that trend, such as a pandemic or a medical innovation? The overall trend may continue, but on a completely different plane. Note that ‘catastrophic’ in this context doesn’t mean bad, it just means an event that shifts the data to a new plane.

On a national scale, natural increase (births minus deaths) most often is the dominant source of change, and the error is in the net migration term. But at the local scale, the comfortable predictability of natural increase is swamped by net migration which has a significant catastrophic component. While the closing of a factory in Muncie, Indiana would have negligible impact on national projections, it would have tremendous impacts locally. The landscape in five years will be the result of an unstable marriage of long-term trends, a huge number of locally significant catastrophic shifts, and likely some unforeseen major events just to gum up the works.

A good projection series is in effect a ‘ceterus paribus’ projection – given what has been happening in any area recently – expect it to continue if you know nothing else. There are things that we do know that should be utilized in small area projections:

• A projection system must utilize spatial context. Using the demographics of a block group to predict five years down the road is virtually pointless, since migration dominates the equations at the local level. As in retail site selection, locational context is everything.
• It should include projections from a range of geographic levels from national to metropolitan but should likely not include local government projections which are often simply marketing statements; your local economic development department is always going to be optimistic. What is a black swan event for a block group may be a known trend which is accounted for in regional projections.
• Projections must take account of local conditions. There may be growth pressures, but is there any developable land left? What impact on the pattern of change within a city can be expected with the construction of a new beltway freeway? What is the industrial base of an area, and what overall trends are occurring in that industry? A projection which simply uses a straight line from past patterns will simply not work.

An intelligent projection model, though, will still only ever answer the question “what will this trade area most likely look like in five years, if there are no major local shifts?” In 1990, would any reasonable forecaster have predicted the rise of the banking sector in North Dakota? Or the dramatic shift in working at home that occurred with COVID-19? Or even that the population of California would decline? By their very nature, forecasts are conservative – we assume that nothing major will happen to disrupt the patterns.

Most in the commercial real estate world understand the limitations of forecasting and tend to treat them as a baseline expectation of change, to be modified by local knowledge and a general awareness of the economic, social, and political winds affecting any particular area. Valuable to consider, but always taken with a grain of salt and with the expectation that black swans not only exist but come in local, regional, and national flavors.