Tuesday, April 13, 2010

New England demography 2 - problems with projections

On 1 April in New England demography 1 - rise of the coast I began a review of New England's changing demography using the latest ABS regional population figures.

In it, I noted that between June 2001 and June 2009, the estimated residential population of New England grew by 107,875 or 8.5%. This growth was just fast enough to maintain New England's share of the total NSW population (19.22% vs 19.23%). However, this growth was entirely concentrated on the coast. 

On the coast including the Hunter, the estimated population grew from 1,052,430 to 1,159,970, up 107,520 or 10.22%. By contrast, New England's inland population grew from 211,306 to 211,641, up just 335 or 0.16%.

I also provided some historical material to indicate the scale of the demographic shifts that had taken place within New England over a relatively short period.

Since I completed that post, Sydney has released updated population projections for NSW through to 2036. These projections were completed in 2008 and are based on 2006 census data. You will find them here in both PDF and Excel format.

These projections show the population of what is called Northern falling from 180,000 in 2006 to 168,000 in 2036, down 12%. The population of what is called the North West is projected to decline from 139,000 to 123,000, down 16%. 

The projections have been greeted with chagrin. To quote the Northern Daily Leader:

NORTH West mayors have slammed population projections from the Department of Planning as inaccurate, misleading and potentially damaging to their towns’ reputations.

Before going on, the "North West" mayors referred to come from the North West as expressed in common geography, not the "North West" as defined in the projections. The "North West" as used in the projections covers a vast area in the west of NSW from the Queensland border as far south as and including Broken Hill. The mayors in question are in fact all from what is called "Northern" in the projections.

This is an example of what I referred to in passing in The fragmentation of NSW, the sometimes crazy and varying area definitions used in NSW for planning and administrative purposes, definitions that fragment.

Importance of Projections 

The mayors are right to be concerned.

The projections do make it clear that they are just that, projections, emphasising that they should not be used as forecasts. Yet the reality is that these projections will not just affect perceptions, but also actual resource allocation decisions.

The NSW Government has limited resources. Planners have to set priorities, taking current and projected needs into account. The fact that the "Northern" and "North West" areas now constitute such small proportions of the NSW population, the fact that the population is projected to decline, automatically affects priority ranking across every area of service delivery.

In theory, the projections could be adjusted to take varying local conditions into account. In practice, this can be quite hard to do. It's not just that the public servants involved lack local knowledge, a growing problem in NSW and Canberra, but also the desire to ensure that decisions are based in some way on "objective" criteria. This includes the increasing use of specific state wide "performance" indicators by both Sydney and Canberra, averages that then drive resource allocation decisions.

Derivation of the Projections

The NSW population projections include a detailed analysis of the assumptions used. I do not want to provide a detailed analysis of these, but we do need to understand some of the key assumptions that have to be made.

The first thing to note is that the projections include both population and household projections. I generally use population projections for the sake of simplicity, but household projections are very important for planning purposes.

If, for example, you are trying to estimate needs for social housing, then you need to know the future number and type of households. For example, you might want the number of households with people over 65 as a first step in calculating changing demand for older people's accommodation. By contrast, if you are estimating future school demand, then you want to know the changing number of people in different school age cohorts.

The second thing to note is that the area break-up for the household projections is different from that attached to population. The population projections use thirteen regions, whereas the household projections are based on four regions: Newcastle, Sydney, Wollongong and then the rest of the state. I am not sure that there is any statistical reason for this; I think that it just reflects planning priorities. However, outside Newcastle it's not much use from someone like me who is actually interested in New England.

However, there is a broader issue here. The smaller the area used in calculating projections, the harder to do the calculations, the more likely it is that the projections will be wrong. It's not just that specific local events may intervene, development of new mines is an example, but that the reliability of the techniques used to generate the projections increases with size.

This links back to my point about the use of the projections. The broader the geographic area you use, the more accurate the projections are likely to be. However, the larger the area used, the greater the variety likely to be found on-ground within an area. You then need to find other techniques for making more precise resource allocation decisions within areas.

Turning now to the generation of projections, there are various techniques that can be used.

For example, I often use fairly rough back-of-envelope calculations for test purposes. So I might take the difference between the census populations in 2001 and 2006 and then simply project this. I then take those numbers and look at the oddities, the things that stand out.

To illustrate, between 2001 and 2006, the population of Moree Plains Shire declined from 16,233 to 14,682, a decline of 1,541 or 9.5%. This is a very large fall, large enough to affect the total population numbers for inland New England. If you simply extrapolate, Moree begins to wither away.

Now I don't believe this. Moree has faced very particular problems over recent years connected, among other things, with drought. The town has faced real problems in attracting skilled workers - a number of businesses have vacancies - but even so, a continued rate of decline of this scale is just not credible. Indeed, the ABS estimates suggest that while between 2006 and 2008 the population fell further from 14,682 to 14,401,it then stabilised. The 2009 estimate shows an increase of five, not a lot but still a reversal of trend.

This is rough back of envelope stuff. The more sophisticated approaches take into account:

  • The structure of the population at start point. The 2006 census is generally used here because of its accuracy.
  • The projected birth rate. How many kids will be born taking into account the number of people in the relevant age cohorts.
  • The projected death rate, taking into account any expected changes in longevity.
  • emigration as locally born leave, immigration from new arrivals.

A major problem in making projections is that the total NSW population is strongly influenced not just be the natural birth rate, but also by:

  • emigration, with a continued flow of people from NSW overseas and to other states. Within this number, Sydney generally has a net outflow, the rest of the state a small positive inflow.
  • immigration, with a heavy flow of people especially to Sydney.

These two variables affect not just the total population of NSW, but also the distribution of people within NSW. To manage this, a top down approach is generally adopted. Estimates are made for the total State population that then set a frame for the derivation of regional estimates.

Expressed in this way, you can see just how many uncertainties are involved. You can also see just how important our total migrant numbers are to the projections.

The latest numbers

I have previously argued that I thought that for inland NSW as a whole, changing structures meant that the previous contribution to Sydney and the coast from internal migration had passed its peak.

Now when we look at the numbers, we can see that between 2006 and 2009 inland New England gained numbers.

Between 2001 and 2006, the population of inland New England declined from 211,306 to 207,401,  a decline of 3,905 or 1.8%. You can also see just how important the decline in Moree of 1,541 was in the total number.

Between 2006 and 2009, the estimated resident population of inland New England increased from 201,401 to 211,641, an increase of 4,241 or 2.04%. This is not a big increase, but a total reversal of trend.

Is this population increase likely to be maintained, at least for the short term? In my view, yes. The bigger population centres within inland New England all appear to be stable or growing. Indeed, there is something of a boom underway.


Given all this, the idea that the population of "Northern" could decline from 180,000 in 2006 to 168,000 in 2036 strikes me as absurd. Without having crunched all the numbers in detail, the only way it could happen would be some form of disaster such as the total closure of the University of New England or a huge and long prolonged drought. I just can't see it.

I would, I think, argue that this is a case where those generating the projections should have checked them with on-ground analysis.

In my next post in this series I will look as some other aspects of New England's population.

No comments: