Monday, 8 July 2013

A Cautious Estimate

Where are the missing 90 year olds? There's an interesting report by BBC's "More or Less" on thousands of elderly people missing. Apparently, the last UK census found far fewer people in their 90s than expected, and the same thing happened in the US with people over 100.  This are not huge numbers, however. There were 30,000 fewer people aged in their 90s than previously believed, giving a figure of 429,000 instead of 457,000. That works out at around 15% fewer men, 5% fewer women. And something similar has happened in the United States.
Before we put the X-Files hat on, and launch into Alien Abduction Conspiracy, we need to look more closely as what is happening with these figures.
"Between censuses, annual population estimates are made by taking the numbers from the last census (which are, in part, estimates), and applying mortality tables, information from registered deaths and immigration data."
The 2004 projection estimated that by 2010, there would be 114,000 people aged 100 or over. But in fact, there were less than half that actually counted by the census.
What has happened is that people are just not living as long as statisticians had predicted. They've just died between the projection and the census. The further you get from the census, the more likely the prediction is likely to be out. It is like weather forecasting - the more days ahead, the fuzzier the forecast.
In Jersey, we've just had some current figures. The recent population report estimates the population at around 99,000. It takes the 2011 census of 97,857 as a baseline, and updates it using numbers of births and deaths in Jersey, and from information on education, health and employment to estimate net migration.
This looks somewhat firmer, and we are talking about smaller numbers. But what makes me just a small bit suspicious is the lack of a range. I'd have thought that any estimations would come with a range which is likely rather than just one very round figure.
There is clearly a fuzzy area - some people might be Jersey resident, but die outside of Jersey, perhaps in hospitals in the UK where they are receiving treatment or palliative care, and it's not clear how these could be recorded. That's something which may well effect figures on the elderly more than the young, although sadly, I know of at least one young baby who died at a neonatal unit in the UK. If you die in the UK, your death will be recorded there and not by the Office of the Superintendent Registrar.  And some elderly people, or at any rate people in their 60s relocate to France - I know two or three myself - and these will again skip past the net. But I wouldn't think these are that significant.
What it would be interesting to know is how accurate estimates of 2011 - the census year - were in 2010, for example, but there isn't any report. We do have one which suggests that the resident population of Jersey at the end of 2009 is estimated as 92,500 giving an estimate rise of 5,357 in 2 years to 97,857 in 2011. At the end of 2011 - after the census, it was estimated as 98,000.
Interestingly, the 2011 report states that  "Figures for 2000 to 2009 are revised from those previously published" and these actually show a greater increase than had previously been estimates.
For example, we have
2005 estimate 88,400 revised to 91,000
2006 estimate 89,400 revised to 92,300
2007 estimate 90,900 revise to 94,000
2008 estimate 91,900 revised to 95,400
2009 estimate 92,500 revised to 96,200
This enables 2010 to be estimated at 97,100, bringing it in line with the census the following year of 97,857, and a much smoother ascent. Otherwise we have a big jump over the last two years.
Evidently the 2009 figures were quite a bit out - by 3,700 - and the divergence from the previous census was marked, although in this case by underestimating the amount of people in the Island.
This was significant enough to have an extra section devoted to it, called "Reconciliation of population measures" to give an explanation for the differences.
It concludes that "This discrepancy is due to modelling of migration patterns, leading to a cumulative under-estimation of net inward migration in the previous annual update methodology, which was based on those migration patterns seen at the time of the 2001 census."
Is the model better now? We will find out the next time there is a census!


Mark Forskitt said...

A quick observation - today's 90 year olds were born in the 1920's. Could the great depression, and associated poor start in life, have had an observable influence?

Nick Palmer said...

Soylent Green?