Two articles in The Economist

A bad error is in this, from 2007 (found when I was arguing that children ARE a public good):

First, it should be obvious that nations don't have to have pension systems highly sensitive to worker-to-retiree ratios. A shift to a system of mandatory personal retirement accounts immediately solves that problem.

It's an old joke that when The Economist is at its most certain it is at its most wrong, and this would back that up. It's not 'obvious' because such a shift does not solve that problem, either immediately or over the medium-term. If you think about it for a second it's obvious that it does not - imagine if no-one was born after you, and so when you reached 65 or 70 and retired there were no younger people in employment. What good would your personal retirement account be then? If it was in equities, those companies would be worthless, if it was in govt bonds, the govt would have defaulted, it was in cash that would be inflated away. Or simply think about it in terms of actual transactions - who would make you food, or provide healthcare?The only get out is if it was invested overseas, but that isn't the argument being made here.

Less of an error, but sloppy, is in this week's edition. In an article on French competitiveness, it says:

The Netherlands, with a fraction of its population, now exports more than France.

The Netherlands is an important trading nation, but a key reason for the Netherlands' remarkable export performance is its huge amount of "re-exports" (about 50% of exports). These are imports that come into a Dutch port via a trader, and are then rexported unchanged to another country, most commonly Germany. There is value-added by Dutch traders, but it is a small fraction of the nominal value of goods traded.

Now there might be good reasons to believe French competitiveness has declined and its export performance compared with the Netherlands might be one of them. But this particular statistic doesn't really tell us anything. After all the Netherlands also exports more than the UK, which has a similar population to France.

How to call bullsh*t on statistics part whatever

One of my first ever blog posts was noting a long-running error in the Economist about US car ownership. Here a similar version appears in The Atlantic. The claim is that US - so associated with car ownership - has fewer cars per 1,000 people than Europeans.

The claim is not wrong, as the figures provided show. But it is misleading, for the simple reason that in the US SUVs, or pickup trucks, which are predominately used for private transport, ie as a car, are counted as "light trucks". In Europe they are not. In both markets these vehicles have taken market share since the 1980s.

The first chart (apologies for the formatting and layout here) shows US car ownership per 1,000 people over time. It shows that car ownership peaked in the mid-1980s at about 550 cars per 1,000 citizens, and has steadily declined since to about 420.

The second chart explains why this has happened. It shows the number of "trucks" per 1,000 people, which has risen steadily to over 350 by 2010.

 Finally the third chart puts them together. The cars have declined, the trucks have risen, and the total has - as you might expect - steadily increased.

It should be said that some of these "trucks" are what we in Europe would call trucks - vehicles for carrying cargo. But the amount is relatively small - each year about 500,000 heavy trucks are sold, compared to 6 million or more light trucks.

The reason I spotted this one was simply because I have researched this area quite a bit over the years. But if there is a lesson, it's probably that statistics that don't seem to fit in with what you know about the world around you should be checked thorougly.

UK GDP per capita lower than in Q1 2004

UK GDP per capita, the most common measure of economic well-being, is now the lowest it has been since Q4 2003. I calculate in Q1 2012 it was £5,577 per head, just below the £5,584 per head it was in Q1 2004, eight years ago. It is 7% lower than at its peak in Q4 2007, but 1% higher than its low, at the peak of the recession in early 2009. 

Note these are my own calculations using real GDP and population, of which 2011/2012 is projection. 

Unemployment by Parliamentary Constituency

Duncan Weldon has an interesting piece in The Guardian on how unemployment is affecting different parts of the country differently, and how this could affect the election. This bit caught my attention:

The national rate of people claiming jobseekers' allowance is currently 5%. In Labour-held seats, the rate is an average 5.2%, while in Conservative-held seats it is considerably lower at 2.9%. In the 50 most marginal Conservative-held seats it is 3.6%, well below the national average and that of Labour-held seats.

The first thing to note is something seems a little wrong. If the national % unemployed (let's call it that for now but it is a narrow measure) is 5%, and the Conservative seats, by far the most numerous, only suffer 2.9%, then one would think the others would have much higher unemployment to bring the average up. But Labour's seats, the second most numerous, see only slightly higher unemployment than the national average at 5.2%. A quick calculation suggests this means the Lib Dems and other parties must have an unemployment rate of something like 11%, which isn't credible.

In fact although the UK claimant count is nationally said to be 5.0%, the individual constituency data  averages to 4.2% nationally, a number which fits the party figures much better [1].

The following table shows this national figure of unemployment in % for May 2010 and Feb 2012 and split out by constituency according to the party affiliation of its MP. The figures I calculate are slightly different than Duncan's, perhaps because I'm using old constituencies. I would use his if you want the accurate numbers.

Party	Unemployment May 2010	Unemployment Feb 2012	Change
Conservative	2.8	3.1	0.30
Labour	4.9	5.5	0.66
Lib Dems	2.9	3.3	0.45
SNP	3.0	3.5	0.45
PC	2.7	3.3	0.45
GB	3.7	4.2	0.50

It's no surprise that unemployment is higher in Labour constituencies than others. What might matter more in for the next election is the change in unemployment since the last election, which is shown in the final column. It is again true that since the last election unemployment has risen by more in Labour seats (0.66% points) than Lib Dems' (0.45%) and especially Tories' (0.30%)  - although note the Lib Dems and nationalist party seats have seen the biggest proportional rise.

With respect to the election though, what matters is the marginal constituencies. Here the situation is less clear-cut. In the 50 most marginal Conservative seats to Labour (on 2010 boundaries [2]) the unemployment rate has risen from 3.7% to 4.1%, an increase of 0.42% points. This is very similar to the national average. By comparison the 50 most marginal Labour seats to the Tories have seen a rise from 4.4% to 4.9%, an increase of 0.57% points. So I don't really see any obvious difference in the marginal seats than the national picture.

Here's a chart which shows this is more detail. Each dot represents one of the 100 most marginal Con/Labour seats. On the horizontal axis is plotted the Conservatives' lead over Labour in % points at the 2010 election. On the vertical axis the change in unemployment. There isn't a huge difference between them, except the trend for Conservative seats to have lower unemployment. Note the two labelled seats have presumably seen such a large increase because (I think) the data is not seasonally adjusted [3].

So my tentative conclusion is the marginal seats are seeing no better unemployment situation than any other, although Conservative seats in general are having a better time of it.

[1]The reason for this discrepancy is unknown to me, maybe a difference between claimants and the jobseeker's allowance or perhaps seasonal adjustment?)
[2] The new boundaries perhaps render this analysis pointless, but I think the unemployment changes will largely be the same given they are in the same area. But I'm not sure.
[3] Only one seat has seen a larger increase in unemployment than Blackpool South, and that is Bradford West (+2.4%)...

How to call bullshit on statistics. e.g 2

This article has got a lot of press coverage as well. It contains a startling claim:

Goodall says, "our total use of materials [in 2007] was almost the same as it was in 1989, despite the economy having tripled in size in the intervening years

This falls into the A Second's Thought Should Tell You It's Wrong category.

A tripling of the economy in the years from 1989 to 2007 would require a growth rate on average of 3^(1/18)-1 which equals 6.4% per year, every year. I'm not going to look up the data, but I'm pretty confident the UK hasn't grown that fast in a year in the past 30 (and hardly ever - perhaps the Barber boom peak?). In fact between those years UK GDP rose by 59%, a rate of 2.6%.

So what does he mean? I know where the figure comes from - it is true that UK nominal GDP rose by 3 times between 1989 and 2007. Clearly this was mostly inflation, ie a rise in prices. There's no reason why a rise in prices should require more 'stuff', and of course to use such a figure is completely wrong.

To be fair to the academic, it might be a journalistic or sub-editing error as the original paper doesn't contain the statistic.

How to call bulls**t on statistics. e.g 1

I'm good at spotting dodgy statistics, and I thought it might help me and others in listing the ways you can spot them (which I'll do as they occur to me.

This has been doing the rounds:

A couple of years ago, there were more [Porsche] Cayennes circulating in Greece than individuals who declared and paid taxes on an annual income of more than €50,000, a figure only slightly above the vehicle’s list price

.

Now, what rings the alarm bells here? It's not the source, which seems respectable enough. It's also not the overall feel of it, as we know Greece has a problem with richer taxpayers declaring income. It's not even the slightly odd comparison of a stock (Cayennes in Greece) with a flow (income per year). There is the niggling concern that this allows us to laugh at foreigners, but the author is a foreigner so we'll let that pass.

No, what makes it require further checking is the specificity of it. For unless Cayennes' have an extremely high market share of the luxury sector, which would in itself be unusual*, if there are more Cayennes in Greece than €50,000/year taxpayers, there must be loads more other luxury cars: Porsches, Ferraris, Mercedes, BMWs, and so on. So why not say that to make the point about tax evasion?

So let's do some checking. We find there are statistics on Greek Cayenne sales. For example in 2005 there were 278 sold. In 2007 265.

In total since its launch in 2003 (in Greece, 2002 worldwide) a rough (I've not used a calculator) figure is about 1,700 sales. This is less than 0.1% of the market in this period, I think, possibly less than their UK share (but I can't find figures for those).

So on the face of it, unless there are exceptionally few taxpayers over €50k**, the statistic looks false (and if there are, why use Cayennes as your comparison?). There might of course be factors I do not know about - secondhand imports, non-official sales, or maybe I've made a mistake. After all the author of the piece is Greek. But it was the Cayenne bit that seemed wrong.

* Also Greece is surely hilly, and in places will have bad roads, so the idea of buying a Cayenne is not as ludicrous as the author suggests. But then again he is from the country.
** Of course this is the other way to look at it, how many taxpayers are there? I could only find statistics of 5,000 over €100,000, which is obviously a low figure but more than the Cayennes. And there would be many times (10? 20?) more taxpayers over €50,000 than €100,000.
Nb: Author also mentions a certain town, the region of which it is in reported only a handful of Cayenne sales in those years.

Govt deficts do not equal country deficits

The BBC reports:

1450: BBC's Mark Gregory has been looking at the economic impact of the quake. Early estimates have put the cost at £112bn ($180bn), equal to 3% of national output. "That's a large burden for a nation that even before last week's disaster had the highest public debt - 200% of GDP - of any advanced economy," he says.

Well not really. It does mean rebuilding will be quite a burden for the Japanese government, which is heavily indebted. But almost all of the debt is owned by Japanese households, and then some - the net external assets of Japan are something like $3 trillion, about 55% of Japan's GDP.

So the amount to spend on reconstruction should not really be seen in the context of the government's deficit, but the assets of the country, its GDP and also its claim on foreigners. All of which are huge.