Wednesday, August 28, 2013

Population Growth and Residential Construction Activity in BC

(This is the first in a three part series analysing population growth and construction activity in BC. See the second and third parts.)

There is an obvious correlation between housing activity and population growth. Since the 1980s one can see the ebbs and flows of population growth, alongside housing starts and completions, have been a normal occurrence in BC's history. The second graph puts population growth in a longer historical perspective for posterity.

The best correlation is between housing completions and population growth, however there are two behaviours, one pre-2000 and another post-2000. Pre-2000 there were approximately 0.22 completions per person of growth, with a 1 quarter lag (ie population growth rises and falls one quarter before completions). Post-2000 there were approximately 0.33 completions per person of growth with a 3 quarter lead (ie population growth rises and falls three quarters after completions).
Some clues as to why there is a difference should be evident by observing the fraction of detached completions of the total has been dropping since the early 1990s in Vancouver that makes up the majority of residential construction activity in the province.

  • Though it's not immediately evident from the data, by visual inspection, there is arguably a "baseline" 10,000 units built each year to maintain the existing stock of housing for the existing population. That is, if population growth were zero there would still be a turnover demand of around 10,000 units per year.
  • The number of completions per person of growth after 2000 corresponds to 0.33, or about 3 people per completion. Before 2000 this was 4.5 people per completion. It looks as if the shift to multi-unit construction at least partially accounts for this difference.
  • As a rough rule of thumb, if population growth is 30,000 this results in annual housing demand of 20,000. Over the course of BC's history since the 1960s, 30,000 is a typical nadir. Most recent population growth has been around 40,000, which corresponds to a demand of about 22,000-25,000 units. Current completion rate is near the high end of this range but the number of starts suggests the number of completions will not markedly deviate from current population growth.
  • Population projections typically do not account for the variability in year-to-year population growth but concentrate on averages. Population growth has averaged 60,000 and completions have averaged 28,000 since 1965. Growth projections are about 55,000–60,000 until 2035 based on BC Stats analysis. Population growth cycles with a period of about 10–15 years.
I track population growth closely as it aligns with residential construction activity, GDP growth, and new and used housing sales and for-sale inventory.


Unknown said...

I know this is a late comment, but I am curious to know why you have used one quarter lag versus three quarter lead for your correlation study.

jesse said...

Cross-correlations are used to find best fit. So long as the best correlation is within a few quarters it should be reasonable to assume the two are related.

If one activity leads another there is some evidence to support causality. I cannot comment too much on whether construction activity causes population growth or vice versa.

One thing I think should be obvious is that residential construction activity supports the housing of not only the construction workers themselves but also non-related industries. We can think of it simply as construction demand will be for construction workers (A) and non-contruction workers (B) for a total activity of A+B. Once A is completed there will be less demand: all that is left going forward is the "baseline" B. That means a surplus of residential construction workers. If construction workers start moving out of province as they did in the late 1990s, their previous demand A will become supply. I think this is in large part why residential activity is inherently cyclical. My bet is the ebbs and flows of population growth over the past two generations are not random occurrences.