Welcome to European Tribune. It's gone a bit quiet around here these days, but it's still going.

Long Term Population Statistics = Bollocks

by nanne Mon Sep 1st, 2008 at 02:12:22 AM EST

So Eurostat has drafted a news release for its new population statistics, which has been picked up by the Beeb and several blogs, as they show the UK becoming the largest European country in terms of population in 2060.

The UK allegedly is to grow to a population of 77 million by then, and will be bigger than France (72 million in 'Metropolitan' France) and Germany (71 million).

Never mind Turkey...

European countries apparently want to have these numbers, as they have asked Eurostat in a Council resolution to produce them. Still, Eurostat's methodology leaves a lot to be desired.

(crossposted from the DJ Nozem blog)

Front-paged by afew


The data behind the story is basically contained in the report 'statistics in focus 72', where at the bottom you can get the estimates for the numbers that are supposed to drive population growth for the coming 52 years.

There are some big differences here.

Basically, the report works with only one scenario (dumb), which assumes convergence of socio-economic conditions, with 2150 as the convergence point. It is not stated which data was taken as the baseline for this scenario (nor is this stated in either of the notes on the metadata), but it seems to have been the data from the past few years.

A few countries like Italy and Spain have had large inflows of immigrants in the past few years. In Spain's case, this was partly due to the construction boom, partly due to policy. In Italy's case, it was mainly due to immigration from Albania and Romania, who are not going to send such a large number of immigrants to Italy in the medium term.

So we get the odd statistic that Italy will lose 12 million of its current 'natural' population, but will overcome this by adding 12 million net immigrants. Spain will lose 5 million population, but will similarly add 12 million through immigration. Considering the hell that is already being raised in Italy about immigrants, this is an absurd proposition.

By comparison, France is expected to get only 4 million net immigrants.

Long-term population statistic can change rapidly depending upon government policy and natural changes. Last year, the German fertility rate (number of live childs per woman) increased from 1.33 to 1.37, and it might continue to go up this year because of more family-friendly policies. Between 2001 and 2006, the rate increased from 1.63 to 1.84 in the United Kingdom.

Since fertility statistics have a cumulative effect upon total population and even go exponential over a 52 year period, those small differences behind the comma translate into millions by 2060.

Immigration flows can change even more rapidly.

You can see that the predictive power of these statistics is quite low. And it's just the predictions that get reported. The statistics do have a basic if-then quality, but this too is quite low, because they are linear (there's just a line towards a convergence point), and consider only one scenario.

So here's a general rule for longer-term planning: linear, single-scenario models are useless, what is useful for formulating policy are dynamic, multi-scenario models.

(It is quite amazing how many policies are informed by dumb models when you start noticing this. We can do so much better)

Considering that Eurostat only releases these statistics once every 3 or 4 years, there should be enough time and resources for a bit more imagination.

Display:
This topic was discussed in a few of the past Salons. Have at it.
by nanne (zwaerdenmaecker@gmail.com) on Wed Aug 27th, 2008 at 04:00:04 PM EST
Great to see the cross-post!

Having worked in/with various civil service stats depts in Wales it doesn't surprise me when stats are used in a linear and too simple way like you've outlined here. When carried out internally as an ongoing stats collection and analysis process, I've not seen great access to systems that could allow for more complex modelling of different scenarios to give better predictability. So it comes down to the capabilities and forethought of the stats people which is variable. A different matter if a big piece of research or trend analysis is commissioned.

I would have expected more to be put in at European level though.

by In Wales (inwales aaat eurotrib.com) on Wed Aug 27th, 2008 at 05:39:26 PM EST
[ Parent ]
I also highly praise the concept of the cross-post.
by Metatone (metatone [a|t] gmail (dot) com) on Wed Aug 27th, 2008 at 05:51:33 PM EST
[ Parent ]
In addition to her praise of the cross-post of the concept?

A vivid image of what should exist acts as a surrogate for reality. Pursuit of the image then prevents pursuit of the reality -- John K. Galbraith
by Carrie (migeru at eurotrib dot com) on Thu Aug 28th, 2008 at 05:43:08 AM EST
[ Parent ]
What do you mean by "systems?"

Surely not a lack of software? You can do quite sophisticated models with GNUPLOT and a cheap FORTRAN compiler.

- Jake

Friends come and go. Enemies accumulate.

by JakeS (JangoSierra 'at' gmail 'dot' com) on Thu Aug 28th, 2008 at 04:12:23 AM EST
[ Parent ]
I don't know exactly what software that the Assembly stats people have access to but the two you mention I have never come across - what may be in common use within academia doesn't necessarily find it's way into the public sector.

When I worked as a data analyst we developed all of our own tools in Excel, Access and Snap since that was all we had!

by In Wales (inwales aaat eurotrib.com) on Thu Aug 28th, 2008 at 04:55:34 AM EST
[ Parent ]
Poor windows users...

A vivid image of what should exist acts as a surrogate for reality. Pursuit of the image then prevents pursuit of the reality -- John K. Galbraith
by Carrie (migeru at eurotrib dot com) on Thu Aug 28th, 2008 at 05:34:09 AM EST
[ Parent ]
You can model anything you want with a spreadsheet (even with Excel ;-) - they are Turing complete! And modern spreadsheets (such as Excel) are augmented with Visual Basic or some other bona-fide programming language.

If people don't do scenario analysis is because they can't or won't, not because the software won't support it.

A vivid image of what should exist acts as a surrogate for reality. Pursuit of the image then prevents pursuit of the reality -- John K. Galbraith

by Carrie (migeru at eurotrib dot com) on Thu Aug 28th, 2008 at 05:36:56 AM EST
[ Parent ]
Doing scenario analysis efficiently requires some degree of training. Whereas we'd have training on the software, eg VB, we weren't trained to apply it and nobody in the team I worked in was actually qualified to do so. It was all developed through experience, so you can see how oversights would arise in that example.

Organisations such as the Office for National Statistics would be a very different matter and entry to jobs there requires rigorous standards to be met eg stats/maths related degree.

by In Wales (inwales aaat eurotrib.com) on Thu Aug 28th, 2008 at 05:45:15 AM EST
[ Parent ]
So the people producing these statistics don't know anything about statistics?

Because these news items often sound as if someone has decided the graph starts here and their one and only data point goes there and if they draw a straight line - it's a news story!

But it's the media, so silliness abounds.

by ThatBritGuy (thatbritguy (at) googlemail.com) on Thu Aug 28th, 2008 at 06:34:57 AM EST
[ Parent ]
Reminds me of the astrophysics PhD candidate who wanted to fit a regression line to 4 points (the guy only had 4 gamma-ray bursts to work with, unfortunately).

A vivid image of what should exist acts as a surrogate for reality. Pursuit of the image then prevents pursuit of the reality -- John K. Galbraith
by Carrie (migeru at eurotrib dot com) on Thu Aug 28th, 2008 at 06:42:10 AM EST
[ Parent ]
Well, you know the joke about black sheep in Scotland :-P

- Jake

Friends come and go. Enemies accumulate.

by JakeS (JangoSierra 'at' gmail 'dot' com) on Thu Aug 28th, 2008 at 06:46:13 AM EST
[ Parent ]
Well, that's what you can expect from a press that's being fed ready-made "stories" from belief tanks. But I'm dumbstruck that a government institution would use Excel for anything more serious than the coffee club's accounting.

- Jake

Friends come and go. Enemies accumulate.

by JakeS (JangoSierra 'at' gmail 'dot' com) on Thu Aug 28th, 2008 at 06:45:42 AM EST
[ Parent ]
But I'm dumbstruck that a government institution would use Excel for anything more serious than the coffee club's accounting.

Why? Excel/VBA skills are in high demand for investment banking research jobs.

A vivid image of what should exist acts as a surrogate for reality. Pursuit of the image then prevents pursuit of the reality -- John K. Galbraith

by Carrie (migeru at eurotrib dot com) on Thu Aug 28th, 2008 at 06:53:22 AM EST
[ Parent ]
Why not use Excel?

The journal Computational Statistics & Data Analysis has a whole special issue on the subject.

I am teaching a graduate course of ecologists and forestry students this year, where Xcel will be outlawed. I expect much wailing and gnashing of teeth, but I won't have it in my lab. Plus, gnuplot produces better graphics. Better than R (ducks)

by PIGL (stevec@boreal.gmail@com) on Mon Sep 1st, 2008 at 06:51:55 AM EST
[ Parent ]
So the people producing these statistics don't know anything about statistics?

Sometimes yes.  Not all Govt institutions though but at least a small number I have come into contact with.

by In Wales (inwales aaat eurotrib.com) on Thu Aug 28th, 2008 at 10:16:00 AM EST
[ Parent ]
Different scenarios don't give better predictability, they provide stress-testing of assumptions and illuminate extreme cases (not somethign you actually predict) and blind spots in qualitative projections.

It always comes down to capabilities and forethought because of the GIGO problem.

A vivid image of what should exist acts as a surrogate for reality. Pursuit of the image then prevents pursuit of the reality -- John K. Galbraith

by Carrie (migeru at eurotrib dot com) on Thu Aug 28th, 2008 at 05:42:19 AM EST
[ Parent ]
We know how many people are around today, how long they are expected to live, on current trends, and how many kids they are likely to have (similarly, on current trends). So you can get some okay ballpark figures.

And for shorter periods (say, up to 25 years), you can expect to be pretty close to reality given the huge inertia of population aggregates.

In the long run, we're all dead. John Maynard Keynes

by Jerome a Paris (etg@eurotrib.com) on Wed Aug 27th, 2008 at 05:38:55 PM EST
Yes, though that would mainly go for endemic population growth / decline. Immigration policy is mainly a matter of the politics.

I'm sure that Germany could manage to get a few hundred thousand net immigrants a year if some effort were to be put in, but it is not desired (for a variety of reasons, some of which are reasonable, some of which aren't).

by nanne (zwaerdenmaecker@gmail.com) on Wed Aug 27th, 2008 at 06:00:32 PM EST
[ Parent ]
Unless you have some kind of black swan event.

And those are more common than one can think of.

Even the fact that we don't have a big war here since 45 is not enough: There was a demographic black swan in Portugal in 74 - The revolution made lots of people to return from Africa (here are talking hundreds of thousands in a country of 10 million - not negligible at all).

Outside of Europe we can still see big demographic movements caused by war (are you really sure we can ignore that happening again here?).

Also, resource constraints might make Europe unpalatable for more emigration (we are densely packed here).

And, by the way, Anglo-disease might the UK less competitive than other EU countries. I actually do believe that the forecast of the UK being the biggest country in population terms is utter bullocks precisely because of this.

by t-------------- on Thu Aug 28th, 2008 at 07:44:22 AM EST
[ Parent ]
There was a huge influx of ethnic Germans from Eastern Europe following 1989.
by MarekNYC on Thu Aug 28th, 2008 at 09:21:29 AM EST
[ Parent ]
You don't have to go for wars and outside Europe. The collapse of the economies (and, arguably, societies) of the post-'communist' countries in the nineties resulted in drastic drops of birth numbers, waves of emigration, and (though maybe this was the weakest effect) reductions in life expectancy.

*Lunatic*, n.
One whose delusions are out of fashion.
by DoDo on Mon Sep 1st, 2008 at 03:13:31 AM EST
[ Parent ]
Population projections may be bollocks, but the statistics?

A vivid image of what should exist acts as a surrogate for reality. Pursuit of the image then prevents pursuit of the reality -- John K. Galbraith
by Carrie (migeru at eurotrib dot com) on Thu Aug 28th, 2008 at 05:33:51 AM EST
You're right, of course. What are these, properly, statistical projections?
by nanne (zwaerdenmaecker@gmail.com) on Thu Aug 28th, 2008 at 07:17:19 AM EST
[ Parent ]


I've been accused of being a Marxist, yet while Harpo's my favourite, it's Groucho I'm always quoting. Odd, that.
by BruceMcF (agila61 at netscape dot net) on Thu Aug 28th, 2008 at 01:28:11 PM EST
[ Parent ]


Display:
Go to: [ European Tribune Homepage : Top of page : Top of comments ]