On Friday, June 20, a news item prepared by a local outlet had the headline “New Report Confirms Rapidly Declining Population.” The story indicated that in the first quarter of 2014, “1500 people left the province,” and referred to a local political figure, “If it continues, she says in this year alone it will translate to 6,000 people leaving.”
A closer look at the facts calls the headline and the quoted statements above into question.
Start with the numbers that caused all of the fuss. Each quarter Statistics Canada publishes population figures for each province and simple math can be used to obtain the change for that period. In this past quarter, the population at the start was 526896 and at the end the population was 525378, which represents a net loss of 1518, or roughly 1500, as was reported. Since there are four quarters in the year, multiplying this value by four yields an annual loss of 6000.
It’s easy enough, then, to show where the numbers came from and there’s no error in arithmetic. There are, however, at least three flaws in the reasoning behind the headlines and the statements in the story.
- Quarterly population values show very high variability in terms of both gains and losses.
- Trends cannot be identified from two data points owing to this high variability.
- Population trends are not linear; they are far more complex.
As is often the case, although people like to pretend it’s otherwise, the truth is by no means a simple thing. Let’s take a better look at the population values in Newfoundland Labrador. All of the data used here came from the same source–Statistics Canada.
First, let’s just put the data out there to see. The table below (source) lists the quarterly populations in NL by year since 1971. The arithmetic has also been done that shows the change from the previous quarterly value. There’s no need to read it in detail; it’s just listed so you can check the charts and conclusions that follow, if you so choose.
The news article title noted a “rapidly declining population” based on the last two data points in the above table and, while a decline of 1500 does look alarming, as is generally the case, it’s important to take a longer view. Larger fluctuations, which occur in short periods of time, do show both positive and negative trends but, over the longer period, tend to smooth themselves out.
Rapid Population Change?
When you look at the population over the whole period and not just the past 3 months, there does not seem to be a rapid change in population currently in effect. The graph below shows the quarterly population figures from 1971 to the present. It is common in such graphs to start the y-axis at the lowest value recorded, rather than at zero. This has the effect of accentuating differences in case you have an axe to grind. For now, let’s put that aside and look at the raw population. We will blow it up later, once we have had a chance to explore the random fluctuations that occuer quarter to quarter.
The graph can be summarized as follows:
- There was a steady rise in population up to around 1983.
- During the remainder of the 1980’s there was a slight dip followed by a rise again; or perhaps the whole period from 1983-1990 could be described as more or less no change.
- There was a sharp drop in the 1990’s, probably coinciding with the collapse of the traditional fish-based economy.
- The downward trend slowed and then reversed from 2000 to 2013.
- There may be a gentle downward trend just beginning now, in 2014, the question is: can we make valid predictions based on it?
It’s clear, though, from this graph that there’s no evidence of a RAPID population loss at present.
A Continuing Trend?
That last bullet clause above, “can we make valid predictions based on it,” needs justification. On the one hand we can just state that human interactions are far too complex and unpredictable to make prognostication anything other than a game. On the other hand, we could use the data to justify it. Lets take that course.
The graph below plots the quarterly change against the time. Those familiar with calculus will know that we are, in a crude way, divining the first derivative of the relationship; looking at the rate of change at various times. While the shape of the graph below does not exactly match that of the one above, there are features that coincide.
- For the change graph below, values above the time axis indicate population growth. This coincides with an upward trend (a positive slope) in the population graph above.
- Conversely, values below the time axis indicate population loss.
- The distance from the time axis indicates the amount of change; the further away the points are from the axis in the below graph, the steeper the slope in the above population graph.
When you look at figure 2 above, several observations can be noted:
- in any given year there is a huge variability; one point maybe significantly below the line (loss) while the adjacent ones are above it (gain).
- overall trends can only be spotted on a multi-year basis. From one quarterly period to another there is too much variability to be able to predict the location of the next point.
This leads to an important conclusion: you can’t use one quarter to predict the next one.
This means that it is invalid to say that a quarterly loss of 1500 translates to an annual loss of 6000.
Let’s look at it another way. We just had a drop of roughly 1500 people in the last quarter. Is that a typical amount? To answer that we can plot a histogram of the various changes we have had since 1971. For convenience we’ll group then by the hundred. The results are shown below. The green arrow indicates where the current value would be located.
When you plot histograms such as this one based on vary large data sets (which we certainly do not have) the results are often bell shaped with most of the results clustering around a mean value. An imaginative person could superimpose a bell shape on this one, frankly, that would be stretching things so let’s not go there and make any conclusions based on a normal distribution.
The quarterly drop is still somewhat atypical in that it lies some distance away from the middle cluster (values from around -700 to around +1300 do look more common). It is not, however, an “outlier” located at the extreme ends of the histogram. Let’s just say it’s a point of interest. Maybe it represents a trend and maybe not.
Now that we have explored the variability and, hopefully, have arrived at a more open-minded view of the data, let’s take a much closer look at the numbers. The graph below once again shows the population values for each year since 1971. This time, however, two changes have been made:
- The graph has been scaled so as to accentuate the differences. Now that it’s been established that whatever is there is likely small, let’s magnify it to take a closer look anyway.
- The quarterly values for any year have all been stacked so as to give a better indication of measurement uncertainty for any given year.
Notice the following:
- The trends pointed out earlier are all more clearly visible. In particular the decline in the 1990’s is rather stark.
- The recovery through the 2000’s is clearly visible as well.
- There is a SLIGHT, noticeable downward trend from 2013 to 2014, even when you take measurement uncertainty into account.
- It is impossible to state with any surety, though, whether this will continue.
So now the speculation starts. Some would claim that this is the start of a huge decline such as the one that started in 1990 or is this a minor trend, such as the one that occurred in 1983.
Best answer: nobody knows for sure. Those who like to speculate on the future, whether it be about sports, economics, politics or whatever, do tend to come loaded down with anecdotes and quasi-sensible explanations that explain their positions but, in the end they’re all just throwing darts.
Is Birth/Death Rate a Factor?
Here are a few extras. You may be wondering if, (A) now that the baby boomers are hitting 65, maybe the downward trend is due to increased deaths or (B) decreased fertility rates mean we are having less children and thus less people to compensate for those deaths. Table 2 (source: births and deaths) shows as much data on this as is available from Stats Can.
Table 2: Births and Deaths by Year
While, yes, the last period did show a net deficit, you can’t really spot any real trend in the data as of yet. Maybe once the new figures become available we will see one. For now, the numbers do not support any conclusion.
Is there Evidence of an Urban/Rural Divide?
You may also be thinking that this is a “townie vs. baymen” thing. The graph below (source) shows the population breakdown in the province
First, there’s something weird about the data for 1990. The dip of around 25000 in urban population is matched by a corresponding rise in the rural one, perhaps leading you to believe that, for one year, 25000 townies moved to the bay and then returned! I am speculating but do believe its more reasonable to assume that a single large community (maybe CBS, Mount Pearl or Corner Brook, perhaps) was mis-labelled for one year. I’m ignoring that point as an outlier at any rate.
You can see these from the graph:
- The urban population saw a steady rise until around 1980. It dropped during the 1990’s and now remains fairly steady.
- The rural population is showing no great change overall.
There’s room here for all sorts of speculation, of course. The urban rise may well have been a transfer of population from rural to urban, for example. Overall, though, we do seem to be something of a steady state at the moment. There are now more urban dwellers than rural ones but the gulf is not changing markedly.
Is this about Low Earnings?
In a word: no. The graph below (source) shows average weekly earnings by province and territory. The data for NL is shown using larger markers to make it easy to spot. Not surprisingly the largest earnings tend to be in (1) the Northwest Territories and the Yukon, which have a combined population of less than 80,000 and a huge cost of living and (2) steadily-booming Alberta.
Look closely at NL’s position within the rankings. Not only is it significantly ahead of most provinces but it seems to be climbing in status. No doubt this growth is due to an increasingly strong local oil and gas sector but also to the increased reliance on relatively high paying jobs that require residents to take temporary residence in Alberta as they commute to and from high paying “camp jobs” there.
With strong average wages it is difficult to support the notion that any decline–if it even exists–is about money.
The numbers do not seem to support any grandiose claims regarding a rapid decline in population figures. While there may be evidence of a slight decline the high degree of variability from quarter to quarter does not lend credence to any attempts to extrapolate this and to make predictions about the future. Perhaps even more to the point, the recent historical figures show periods of both growth and loss, none of which could have been predicted in advance.
While what follows should be treated with an even greater degree of skepticism, it is hard not to offer up a few possible causal factors that might be at play on quarterly population figures in the near future.
- The increased public confidence that was associated with the Williams administration is now over. This increased uncertainty may be translating to out-migration.
- Jobs in the oil and gas sector pay significantly more than all others. This is a push factor for those who cannot find local employment in that sector.
- The province has an increasing reliance on foreign temporary workers. Their comings and goings will only increase the quarterly variability.
- The province has an increasing reliance on the commuter jobs to northern Alberta. Their comings and goings will also increase the quarterly variability.
As is always the case, attempts to predict the future based on past events should be viewed with skepticism. The future will do it’s thing and we, as citizens, should work to try and spot any trends as best we can and then to try and mitigate any possible negative outcomes. At any rate, we still owe it to ourselves to do so armed with the facts and now just a bit of speculation based on unreliable recent data.
A final word: between the lines of the news article is the implication that a decline in population reflects a general failure, either on behalf of government administration, or on behalf of the province as a whole. One wonders, though, about the validity of this claim. Seen one way, yes, a net drop in population may be reflective of a society in decline. Seem another, though, is also may be a realistic reaction to booming conditions elsewhere. It may not be the case that things are getting worse here, but rather, that they are comparatively so much better in one other place, namely Alberta.