Demography is a multi-layered business. Most of the attention in demography is devoted to international population growth and movement: the population boom, the fall of birth rates below replacement level in western societies, the impact of AIDS in Africa, and so on. Its fascinating stuff that is fundamental to the future of the human race but its not what really interests us at Demography+.
The focus of Demography+, on the contrary, is on local population changes. Most of the examples cited in these blogs are Canadian because I am Canadian. In fact, a lot, are likely to be Atlantic Canadian and/or Nova Scotian, because I am both of those also. If the other stuff interests you, search the site and you might find some references to the Baby Boom and the impact of immigration, but you will find a lot more to your liking on sites like the Population Reference Bureau or the blog site Demography Matters, both of which are listed on our Links page.
Our interest here is local populations and how to predict them. In many respects, its the more complicated issue because it requires us to deal with smaller numbers and greater volatility. Its important too. While it isn’t about the carrying capacity of the plant, it is what determines available markets and demand for local services. As I’ve noted in my Introduction, its fundamental to avoiding over and under investment and to generally right sizing services.
As a planner, its an issue that has interested me for a long time. One of the first articles I ever wrote, called “Apportionment of Regional Population Forecasts” (Plan Canada, vol. 26 no. 5, July 1986), was addressed to the subject. It explained a very simple method I came up with for projecting the populations of sub-areas within a region using a modified exponential function to project the share of each sub-area into the future. Essentially, the idea was that all sub-areas whose share was falling would be projected to 0 per cent, while those whose share was increasing were projected to 100 per cent. If these shares summed to more than 100 per cent in the future, they were pro-rated down. Not a lot to it but a technique that I still use occassionally and which addresses a pretty basic problem for urban planners of how population is distributing itself within a region.
I appreciated then and appreciate more now that this little model was at best quick and dirty. It drew a lovely letter to the editor of Plan Canada suggesting that anyone could do as well with a ruler. Thus encouraged, I gave a lot more thought to the problem. As I will explain elsewhere from time to time on this blog site, I place my ultimate confidence in the cohort-survival method as my preferred method of population projection. Statistics Canada generously (for the most part) provides the necessary inputs and, for my money, it comes by far the closest of any technique to replicating the real world dynamics of population change. The cohort-survival method, furthermore, provides a level of detail (i.e., predictions of age and sex composition) that can’t be matched by any other method.
From the late 1980s, I developed cohort-survival models through several iterations written in BASIC and for several spreadsheet platforms that I used to prepare projections and forecasts for clients in my work as a planning consultant. I always enjoyed the process because not only does the cohort-survival method generate accurate projections of total population, the detail provides abundant insight to the reasons behind population change. As time has gone by I’ve figured out how to calculate net migration by age group applying the residual method to cohort-survival projections and how to vary the parameters of chort-survival models to create forecasts.
Estimated Net Migration, Portage la Prairie, Manitoba, 2001-2006
In the late 1990s, I began to develop methods to create and reconcile chort-survival projections for multiple areas. It would have been a considerable task in the 1980s but by the end of the 1990s spreadsheets were more than capable of handling the calculations and generating the desirable graphics. I eventually wrote a second article for Plan Canada on this called “Beyond Survival,” (Plan Canada, Vol. 44 no.2, Summer 2004). Essentially, it is a hybrid of the apportionment model with the cohort-survival technique and residual method. It is based on a top down approach. It assumes that the projection for a larger area such as a province or region is more accurate than for its component sub-areas. The region and its sub-areas are all projected at once but the sub-areas are then prorated so that the sum of their predicted populations equals the predicted total population of the region. The result is very rich projections for the region that profile the population of each sub-area, illuminating gainers and losers, and providing insights to the reasons for gains and losses (i.e., age structure or migration patterns, or a combination of the two).
Lately, I’ve assembled data for every province and territory in Canada and for all of the Census Metropolitan Areas (CMAs) and Census Agglomerations (CAs) within each province and territory. I now have a model in which projections for Corner Brook, Newfoundland, are reconciled with projections for Newfoundland and Labrador, which in turn are reconciled with a projection for Canada. Its a lot of information and fascinating to sift through.
I’d like to divide the provinces further to develop projections for Census Subdivisions (CSDs) as well as CAs and CMAs. I have already done it for some areas in Nova Scotia and Ontario. The ultimate goal will be a “projection/ forecasting” machine for all of Canada through which areas can be delineated on a map so as to produce projections that can be manipulated by the user (i.e., adjusted to create forecasts reflecting future possibilities as well as past trends). We’re working on it.
