Population revision – where have all the people gone?

On July 31st, the ABS released the first 2011-Census based population estimates for Local Government Areas and smaller areas.  These are based on the 2011Census results which are adjusted for the under count and people overseas. The are considered to be more accurate than the Census count and are the official population. They also provide the first opportunity to revise all the population estimates made since the 2006 Census. It was clear from the 2011 Census that previous ABS estimates were too high, but we weren’t prepared for was the scale and geographic distribution of the reduction.

The July ERP is a revision to the figures published in March. The March figures were extrapolations from the 2006 Census (updated annually based on births, deaths and estimates of migration). We have long warned that the interim figures released each year between Censuses are subject to review, and not to get too excited about them. We also knew from the state figures released in June that the ABS had revised the populations of most states downward from the pre-Census estimate, so that most areas could expect a reduction.

Some areas have seen very large revisions downwards, to the point of losing most of their five year population growth, while others have had the figures bumped up by the ABS.

Note that this isn’t about actual population change – this is about the change from what we were initially told the population was (the pre-Census estimate) to what the ABS now estimates it is (the post-Census revised figure).

This whole game is complicated by the fact that the ABS has also moved to a new geography for 2011 and onwards, and there were no pre-Census estimates based on the new geography. Fortunately, however they have also released estimates based on the old geography and this is what I’ve based the downward revision tables below on (so the figures for Sydney, Melbourne etc. are different to those used in this article, which is based on the new geography).

Across Australia, the biggest revisions downwards have been in regional areas, particularly in Queensland, NSW and Victoria. In regional Qld and Vic the reduction was quite savage, amounting to a loss of about 4% of the total population, but when you look at this in terms of the growth which we thought had occurred in those areas, it is a reduction of close to 50% of this apparent growth since 2006.

The next chart shows the Local Government Areas with the biggest downward revisions of their population in numerical terms.

The Gold Coast has apparently 22,500 fewer people than their pre-Census estimate would indicate, going from 536,000 down to 514,000. Remember that although this chart shows a negative, we’re not saying the Gold Coast has declined in population. It just apparently hasn’t grown as much as we thought. It still grew by 47,500 people in 5 years. Nevertheless this revision lopped 32% off the growth we thought was there when the figures came out last March.

Smaller in number but more significant in percentage terms are downward revisions in places like Latrobe (Vic), Mildura (Vic), Wagga Wagga (NSW), Albany (WA) and Gympie (Qld), all of which had revisions downward in excess of 2/3rds of the population growth we thought they had. It’s not confined to rural areas entirely. Several LGAs in Melbourne and Sydney had similar downward revisions knocking out most of their apparent growth, including Knox (Vic), Whittlesea (Vic) and Penrith (NSW). In some cases these revisions seem justified, and just bring the population growth back to what we had forecast in their forecast.id sites.

At the extreme end of this scale though are the most perplexing examples. Some areas which had shown significant growth from 2006-2011 in the preliminary estimates (which, remember, are based on births, migration and building happening in the area between the Censuses), now show up as having actual declines in population. These include Tea Tree Gully (SA), Wangaratta (Vic), Colac-Otway (Vic), and Leeton (NSW).

The most perplexing thing about these, is that in many cases the Census counts went up over 5 years. For instance, in Colac-Otway Shire in Victoria, the usual resident Census count went up from 19,867 in 2006 to 20,345 in 2011 – yet the official population estimate has gone DOWN in that time from 21,044 in 2006 to 20,578 in 2011.

The apparent reason is that the ABS seems to think there was virtually no undercount in regional Victoria and South Australia – ie. in 2011, virtually everyone resident in those regions was successfully counted in the Census (and hardly anyone was overseas on Census night), meaning that the adjustment to get from the Census count to the ERP is much lower. See my article A Quality Result for more info. In Regional Victoria the ABS reckons the undercount at only 0.1% of population, which is incredibly low (historically, the national undercount averages about 1.8%). So not many people get added to the Census count to produce the ERP.

Another factor is that the ABS probably overestimated the undercount in 2006, leading them to inflate the population figures at that time a bit. This causes a problem because the 2006 figures are final, and the ABS can only adjust back to 2007, meaning that apparent growth seems lower than expected.

Whether this is correct remains to be seen, and we believe that some of the estimates of the undercount this time (particularly that for Regional Victoria) are too low to be credible. While it’s great to have a quality Census result, it seems highly unlikely to us that it is as low in Regional Vic and SA as has been quoted, and hence we think that the population growth in these areas has been underestimated in the post-Census revision. We recommend using .id’s forecasts in these cases and we think our forecasts are actually much closer to the actual population  in places like Latrobe, Bendigo and Wagga Wagga.

Unfortunately this has significant implications for many regional centres in terms of funding, which is usually based on population size, as the ABS ERPs are the official population estimates used for this.

On the other hand, the ABS has revised upwards populations in many areas, predominantly in the inner parts of capital cities, many parts of Perth, and in the mining areas of WA.

Brisbane had the biggest upward revision – perhaps some of those people thought to be on the Gold Coast were actually in Brisbane! Most of the rest are in WA, and inner suburbs of Sydney and Melbourne. The population in the mining areas seems to have been particularly badly estimated before the Census, with areas like East Pilbara and Roebourne having large upward adjustments. This may be to do with the directive to fly-in/fly-out workers in 2011, to mark the mines as their usual address if they spend more than half their time there. This directive was not communicated clearly in 2006 and so many FIFO workers had their usual address recorded elsewhere.

The upshot of all of this is that the “intercensal” ERP estimates, while they are official populations at the time, are not a very good guide to population. Of the 568 LGAs in Australia, the ABS only got the 2011 preliminary ERP within 1% of the revised figure in 100 of them, or less than 20%. In many cases our own .id forecast numbers were closer, though in some cases these too were out, as councils asked us to try to match the interim ERPs coming from the ABS.

There are likely to be further developments on the ERPs over the next year. As stated above, at .id we do think that some of the populations are some way off the mark, but whether they will be changed, we’re not sure. Watch this space!
Access more information about the Australian Census 2011.