Why a forecast is not a prediction
When a city says its population will reach a certain number by a certain year, that figure is the result of assumptions, not a fact about the future. Planners use population forecasts to estimate how many homes, classrooms, buses or water connections may be needed. They are useful for that. But a forecast really answers the question “what happens if these assumptions hold?” and the assumptions often do not.
Reading a forecast well means asking what it assumed, how far ahead it looks, and how much room it leaves for being wrong.
How a basic forecast is built
Most methods are variations on a few ideas.
Trend extrapolation. The simplest approach extends past growth into the future, either by a fixed number of people per year or a fixed percentage. It is quick and transparent, but it assumes the conditions behind the past trend continue.
Cohort-component method. This is the standard approach in many places. Population is divided into age and sex groups (cohorts). Each year, the model ages everyone forward, applies assumed birth rates to women of childbearing age, applies assumed death rates, and then adds assumed net migration, which is people moving in minus people moving out. It is more detailed than trend extrapolation, and it shows how an aging population may change needs even if the total stays flat.
Housing-based or employment-based methods. Instead of projecting people directly, these start from the number of homes expected to be built or jobs expected to be created and work backward to population. They tie the forecast to land and economic plans, but they can become circular: if a plan assumes growth and allows housing to match, the forecast just confirms the plan.
The part that is hardest: migration
In most cities, births and deaths change slowly and can be estimated with reasonable confidence over a few years. Migration is different. People move for jobs, housing costs, family, schooling, disasters and policies, and these shift faster than demographic data can follow. A few years of unexpected moves can overturn a forecast.
Migration is also local. A neighborhood or small town can see swings that disappear in regional totals. The smaller the area and the longer the horizon, the less reliable the projection tends to be. Forecasts for a whole region are generally steadier than forecasts for one district.
A small hypothetical example
Imagine, purely for illustration, a town of 20,000 people. A planner considers three growth rates over ten years, compounding yearly.
- At 0.5 percent a year, the town would grow to about 21,000 (20,000 × 1.005 to the tenth power is roughly 1.051, giving about 21,020).
- At 1 percent a year, it would reach about 22,100 (1.01 to the tenth power is roughly 1.105, giving about 22,090).
- At 1.5 percent a year, it would reach about 23,200 (1.015 to the tenth power is roughly 1.161, giving about 23,210).
The gap between the low and high cases is roughly 2,200 people, or about 11 percent of today’s population. Even though the growth rates differ by only one percentage point, the spread is large enough to change how many classrooms or housing units would be needed. And that is before considering that real growth might be negative, or might come in a sudden burst in a single year. The numbers are made up for demonstration, but the point holds in real places: small changes in assumptions lead to meaningful differences after a decade.
Why forecasts miss
Several recurring causes explain errors:
- Turning points. Trends do not continue forever. Economic shifts, changes in birth rates, new employers, a plant closure or a change in housing supply can reverse a direction. Forecasts that simply extend history tend to miss these turns.
- Feedback from the plan itself. If a city builds homes and services, growth may increase; if it does not, growth may be limited. The forecast and the policy influence each other.
- Data lag. Official counts come from periodic censuses and estimates between them. Estimates may later be revised, and a forecast built on an old baseline carries that error forward.
- Boundary and definition changes. Annexations, redrawn boundaries or new definitions of households can make series hard to compare.
- Optimism and politics. There can be pressure to produce a number that supports a funding request or development goal. A forecast that is useful for advocacy is not necessarily a good estimate.
How planners cope
Because certainty is out of reach, the practical goal is to make plans that stay sensible across a range of futures.
Use ranges and scenarios. Instead of one number, planners present a low, middle and high case, and sometimes a case with unusual assumptions. The question becomes “what do we do in each?” Decisions that work reasonably in all cases are safer than ones that depend on a single outcome.
Prefer flexible, phased commitments. Buildings and networks that can be expanded in stages are less risky than a large facility sized for a single forecast. Spending can be timed to follow evidence of growth rather than anticipate it. Questions of whether existing systems can absorb growth are often examined through an infrastructure capacity check.
Track leading signals. Building permits, school enrollment, utility connections and employment changes can show whether reality is following the forecast. If it is not, the forecast can be revised early. Early school enrollment shifts, for example, often appear before they show up in official population estimates.
Update on a schedule. Forecasts should be revisited when new census data or other major information arrives, and the revision should be explained rather than quietly replaced.
Be open about the assumptions. Publishing the method and key assumptions allows others to challenge them. It also helps readers understand that a number such as “22,000 by 2036” is a conditional estimate.
Reading a forecast sensibly
When you meet a population projection in a planning document, a few questions will sharpen your reading. What is the base year and data source? How long is the horizon? Is it for the whole region or a small area? Are there alternative scenarios? What would make it wrong? Shorter horizons and larger areas generally deserve more confidence, and any single figure is best treated as a reasonable midpoint of a range.
Forecasts will always be imperfect, and that is not a reason to ignore them. The planning task is to use them as one input, stay alert for evidence that things are moving differently, and avoid locking in decisions that would be costly if the forecast turns out to be off.