In the mid-1970s, when I was taking my Master’s degree at Queens University in Kingston, Ontario, I took a couple of courses from Nirmala Cherukupalle, who was, for a while, a contract lecturer with the School of Urban and Regional Planning. During the time Professor Cherukupalle was teaching us development economics, she published an article in Plan Canada entitled “Projections for Managed Growth Situation: Why Don’t They Teach Us About Declining Countries?” (Volume 16, Number 2, June 1976, pp. 102-109). It wasn’t a landmark article. I read it as a tacit act of brown nosing but it has stayed with me because it has resonated throughout my career as a planner in Atlantic Canada.
The June 1976 issue of Plan Canada was interesting in and of itself because it was a special edition on “population research.” In addition to Professor Cherukupalle’s article, it contains an interesting overview article by Larry Bourne called “Priorities for Demographic Policy Research” (pp. 83-88), which reflects not only an environment of growth but also a rising awareness of coming stabilization combined with concern about climate change (global cooling and related loss in Canada’s agricultural productivity, if you can believe it). Cherukupalle addressed the implications of this stabilization for practicing planners preparing projections and forecasts, a group that I was soon to join.
As demographic decllne began to dawn across the western world, Cherukupalle recognized that old approaches would not do any longer:
The contention of this [i.e., Cherukupalle's] article is that the context of the exercise is now quite different. The emphasis on the ‘management’ of growth, as opposed to that of the accommodation of growth calls for far more precise techniques of projection, or alternatively cybernetic determination, of both the size and the composition of future population.
She noted that practitioners to that time were accustomed to preparing liberal forecasts recognizing that accommodation of growth provided a wide margin of error given that overestimation would be absorbed quickly enough — if a projection or forecast for 15 years hence proved to be excessive, the expected population level was generally achieved in 20 or 25 years. In addition, she noted, the consequences of underestimation in terms of inadequate service and congestion, were far more serious than providing a little more than was needed.
She placed the challenges to precision in two categories: methodological and institutional. On the methodological side, she noted the challenge faced by planners dealing with smaller areas with open boundaries that increase the difficulty of estimating future migration. On the institutional side, she pointed to the ‘generalist’ bias of planners who were not well-equipped to apply complex mathematical models even when the models were what she quaintly called ‘mechanized.’ Planners she suggested did not have the sophisticated technical knowledge to address the underlying judgements needed to operationalize such models.
In this context Cherukupalle assessed the ‘planner’s tool kit,’ which she divided into three categories as follows:
- The Demographer’s Approach – Essentially, the cohort-survival method, which she noted had been supported by careful statistical record keeping dating from the 1920s.
- The ‘Your-guess-is-as-good-as-mine’ Approach – Projection methods, including apportionment methods through which a local projection was developed as a constant or changing share of a national level projection.
- The ‘Holding Capacity’ Approach — A variety of methods then popular with planners and still applied by which the ability of an area to absorb population was calculated and planned for.
Then and since I have ranked the merit of these techniques as Cherukupalle listed them. The holding capacity approach is certainly valid when projecting the population of a small area like a subdivision but it can be downright scary when applied to a municipal unit or a region. For the first half of my planning career, the Halifax-Dartmouth region in which I have practiced was governed by a 1976 land use plan designed to accommodate 435,000 people by 1991, the calculated capacity within a “regional development boundary” defined by the same plan. According to the 2006 Census, we are still 75,000 to 85,000 short of that number, although land was assembled and developed in the 1970s and 80s to provide for the explosion. The result was urban sprawl, fiscal imbalance, and the waste of money and non-renewable resources. While the mitigating influence of declining household size made some sense of huge suburban/exurban land assemblies, so many other options were available that could have provided more appropriate housing in a more cost-effective settlement pattern, if the real future had been addressed instead of a futuristic vision.
In another Atlantic Canadian instance with which I am familiar, a well-established engineering firm designed the water system for an utterly moribund coal mining town based on the assumption that the community would be developed at its then prevailing density across its entire vacant area. The resulting estimate of the town’s future population was ten times its current population and would have made the town one of the half dozen largest municipalities in the region. The community has lost population in every Census since and currently has about ten per cent fewer residents than at the time its future “population potential” was overestimated. I haven’t checked to see what they are paying per capita for their over-sized water supply.
Projection methods can be quite a bit more effective. They are however a bit distasteful as they are designed and applied to be easier rather than better, and because they do not give key details about the composition of population that are often more important than the total numbers. In addition, most projection methods are inherently flawed in situations in which circumstances are about to change, since they rest on the fundamental assumption that the future will be similar to the past.
The demographer’s approach or the cohort-survivial method has always seemed more attractive because it takes into account the individual components of growth (i.e., births and deaths) and provides important detail about sex and age composition that is normally absent from projections developed by other methods. Because the fundamental influences on cohort-survival are available and accurately recorded, it is also the perfect method to deal with changing demographic circumstances. The major problems presented by the technique are the estimation of migration and the relative effort required to apply the method.
These problems, I have always thought were well worth trying to overcome. The answer I believed was in mechanization or, nowadays, computerization. When calculations had to be done manually or even with a calculator, the cohort-survival method is not so much difficult to apply as tedious. Even in the 1970s when age groups were only broken down up to the age of 70, a single iteration of the cohort-survival method using five-year age cohorts required over 30 calculations, assuming the analyst wasn’t bothering to project birth or death rates, or add in the effects of migration. Developing even a 15-year projection must have taken a week’s work. At about that time, I had a job preparing 20-year projections of municipal populations in Eastern Ontario on which basis water and sewer rates were set. Suffice to say I never contemplated using the cohort-survival method for any aspect of this work.
In my first blog on my influences, about Bill James, I noted how little progress has been made in improving planning methods and models since the advent of the personal computer. The cohort-survival method is a leading example of this astonishing inertia. I remember a spreadsheet that was circulated in the United States in the 1980s, I think as a result of an article in what was then the Journal of the American Institute of Planners (now JAPA). I got a copy but, as I also recall, it was difficult to understand and employed different rates for black and Hispanic populations that were not relevant to Canada. Canada Mortgage and Housing Corporation also produced and, for awhile, marketed its Potential Housing Demand or PHD model, which still seems to be used by the Corporation and by some municipalities, but is hardly widespread. I’ve developed and refined cohort-survival models through several iterations in BASIC, QikBASIC, and several spreadsheet programs. I’m sure there are others out there, who have done the same but I am still startled by the crudeness of a lot of local demographic analysis and the mystery that still appears to surround application of the cohort-survival method.
The concerns that Cherukupalle raised, in other words, are with us still. Planners don’t really understand the best methods of demographic projection and forecasting, and aren’t comfortable applying them or interpreting the results. We have, however, fully entered the period of slow growth/decline that Cherukupalle saw coming and the challenges for predction that it presents are real. I don’t see it as a crisis but if we want a sustainable future, we have to avoid the waste inherent in over projection. For that we need sharper instruments more precisely applied. We are soon going to be in need of more considered methodologies for the developing world as well as the developed world, given that many countries in the Middle and Far East, and even Latin America are going through the same process of declining birth rates.
