Tuesday, June 27, 2006

Normal service resumed

In today's column, Polly writes that:

The government only just failed to pass the "incitement to religious hatred" bill because Blair himself accidentally failed to turn up to vote.

Actually, the Racial and Religious Hatred Act 2006 did receive royal assent on the 16 February 2006 (source).  The government lost a vote on a proposed compromise between the government's version of the bill and some Lords amendments on 31 January by a single vote.  An amendment, not the bill (source).

Of the UK, Polly writes:

the very small number of religious practitioners in this most secular of nations.

Social Trends 36, on page 200, says this:

Attendance at religious services varies across Europe. Figure 13.19 shows the percentage of individuals who attended a religious service irrespective of faith at least once a month for the EU nations surveyed. In 2002 the highest attendance was by people resident in Poland (75 per cent) and the lowest by people of Denmark (9 per cent). The countries with the highest rates of attendance all followed the Catholic or Orthodox religion, while the Protestant Scandinavian countries recorded the lowest rates. The United Kingdom is placed 13th with 19 per cent of residents attending religious services at least once a month.

Not even in Europe is the UK the "most secular of nations".

Friday, June 23, 2006


I clearly need to take more holidays.

As an anonymous commenter pointed out here, there was much fuss during the week about Polly's inaccurate attempt to criticise a critic -- see http://www.stephenpollard.net/002642.html.  Today brings a correction from the Guardian here:

In a column headed Britain is smiling, but it looks daggers at Labour, page 31, June 20, we accurately reported a blog posting by Janan Ganesh but incorrectly attributed to him the heading on the blog: "Let the Sudanese die - it's none of our business". The heading was not written by Mr Ganesh.

Like I said, maybe I should take more holidays?

Friday, June 16, 2006


FactcheckingPollyanna is on holiday until next Friday.

Tuesday, June 13, 2006

Stange factoids, indeed

In today's column, Polly Toynbee writes:

The whole sector spends £25bn a year (the state spends £400bn).

Actually, at £457bn, it's closer to £500bn (see Public sector finances, April 2006.  pdf link here).  Being right to the closest £100bn is quite a low bar for accuracy...

She also says:

Is it the same nostalgic delusion that led John Major to dismantle the railways, forgetting that they would always need taxpayers' cash?

These clearly cannot be the same, apparently dismantled railways which Social Trends 36 means on page 182:

The number of journeys made on Great Britain’s railway network (including underground and metro systems) rose by 114 million between 2003/04 and 2004/05, to 2.2 billion. There were around 1.3 billion passenger journeys per year in the early 1980s and, apart from a period in the early 1990s, these numbers have generally increased. Between 1993/94 and 2004/05 passenger numbers rose by 44 per cent

Friday, June 09, 2006

Talking of re-offending...

Today's column offers a wide selection of misquoted and misinterpreted statistics. Let's start with:

Crime is in long-term decline, down 43% since 1995, according to the British Crime Survey (BCS)

According to the Home Office, (pdf link here to Crime in England and Wales 2004/2005 -- oh, yes, by the way, it is worth noting, though Polly doesn't, that this only covers England and Wales):

Since peaking in 1995, BCS crime has fallen by 44 per cent, representing 8.5 million fewer crimes, with vehicle crime and burglary falling by over a half (both by 57%) and violent crime falling by 43 per cent during this period. [emphasis added, p.1]

But this is a minor quibble. It gets worse. Polly writes:

More criminals reoffend. The Prison Reform Trust points to official figures showing how prison overcrowding raises the reoffending rate. In 1995 56% reoffended within two years of release - now it is 67% (53% of those given community sentences reoffend).

Actually, the Home Office is quite careful to explain that they can't actually track reoffending accurately, merely reconviction. So, for example, they write in the excellent Prison statistics England and Wales 2002 (pdf link here, p.150):

Reconviction rates are the proportion of prisoners discharged from prison that are convicted on a further occasion within a given time period (usually 2 years). They only give a minimum indication of the proportion of offenders who re-offend because not all offenders who re-offend will be caught or prosecuted.

And so therefore, an improved detection rate would increase the statistic, giving the appearance of an increase in re-offending. Have detection rates improved? Well, apparently so. Polly writes about:

...improved crime detection, the best in five years.

Polly also writes that:

...courts have 40% fewer "ineffective" trials abandoned through bungling.

We're not done the courtesy of being given a time period here, so let's assume it is compared to the Public Service Agreement baseline date of June-August 2002. The figures show that in the Crown Court, ineffective trials were only 13.3% of the total in October-December 2005, compared to the baseline rate of 23.7%. This is a cut of 44% in the rate, but not to be sniffed at.

The Crown Court is only half the story, though (well, actually more like a fifth, but we'll come on to that). In the Magistrates' Courts, the rate has again fallen, but sadly only from 30.9% to 21.2% over the same time period (figures are here). This is only a 31% cut in the rate.

How to weigh the two rates? According to Criminal Statistics 2004, England and Wales (pdf link here), in 2002 434,500 people were committed for trial at the Crown Court, and in 2004 (the latest year which they cite) it was 382,000 (Table 2.6) In Magistrates' Courts, the figures were 1,925,000 and 2,023,000 respectively. If we're generous and weigh the two rates (44% and 31%) at 1:4, we get a weighted average of 34% and not 40%. Note that this is being as generous as possible. I am also ignoring the fact that the overall number of trials is increasing -- so the absolute number of ineffective trials will be falling slower than the rate -- and also that the balance between the Crown Court (with the higher rate) and the Magistrates' Court (with the lower rate) is shifting in favour of the latter, so the rate is falling slower than the crude weighted average suggests.

UPDATE: The following quote from the column has been worrying me:

How tragically revealing that half of all prisoners are completely illiterate and another 20% have a reading age under eight.

If it were true that 70% of prisoners had a reading age below eight, it would truly be a cause for urgent concern. Luckily, it's not true. According to Prison statistics England and Wales 2002, referred to above, 36.9% of admitted prisoners fall below "Level 1" -- defined as roughly "about GCSE standard" and "11 year old" (the amiguity is in the source. See table 11.4 in Prison Statistics, as well as the notes below it and the summary of Chapter 11 for the figures and quotes).

Now, of course 37% of prisoners having literacy levels below that of an 11 year old is too many (even if it is not "half of all prisoners are completely illiterate"), though note that this compares to about 16% for the population as a whole (see page 13 of Reducing Re-Offending Through Skills and Employment, HMG's Green paper on Offender Learning, available here).

Tuesday, June 06, 2006

Putting the poll into Polly

In today's column, Polly Toynbee writes:

The latest Ipsos Mori poll yesterday put the Conservatives 10 points ahead, at 41%; only six months ago they were 10 points behind.

And yet, when we look at, oh, I don't know, say the, er, Observer on the 18 December 2005, in a piece headlined:

Tories seize nine-point poll lead

This is from a MORI poll, conducted between 9 and 12 December 2005, which is about as close to six months ago as we are going to get. The piece goes on to say:

Current voting intentions among the 53 per cent of the electorate who say they are 'absolutely certain' to vote in a general election show 40 per cent for the Conservatives (up 7 points), 31 per cent for Labour (down 5) and 21 per cent for the Liberal Democrats (down 2).

So even in the previous poll, longer than six months ago, the Labour position would have been 36% (31+5) and the Tories would have been on 33% (40-7), or three points behind.

And not ten.

How does this happen?

UPDATE:Interestingly, the ten-point Labour lead in a Mori poll is from November 2005 (source here), or seven months ago as I count.

UPDATE 2:The poll Polly Toynbee cites today was carried out between 25-30 May. The one showing a ten point Labour lead was carried out between 17-22 November. I therefore concede that saying Labour was ten points ahead six months ago is an acceptable approximation, and withdraw the challenge to the statement.

Note, though, that by the same standard one could equally well say that the Tories were nine points ahead six months ago.

Sunday, June 04, 2006

Better or Worse? p.209

Today, my copy of Better or Worse? fell open to page 209.  On it, Polly Toynbee and David Walker write:

In February 2003, Tony Blair stepped in to make another of his eye-catching promises.  He pledged to halve the number of asylum seekers within six months.

Actually, no.  This is how a Guardian report from 8 February 2003 characterised what he said:

"I would like to see us reduce it by 30% to 40% in the next few months and I think by September of this year we should have it halved. I think we can get below that then, in the years to come," he said on BBC2's Newsnight.

February to September is of course seven and not six months, but even that is a relatively minor inaccuracy.  Consider what the Guardian report goes on to say:

The Home Office made clear that the baseline which will officially be used to judge the success or failure of the prediction will be the as yet unpublished monthly figure for asylum applications in October 2002

So actually, the (clarified) promise was actually to halve the number between October 2002 and September 2003 -- what those of who are not numerically fragile think of as eleven rather than six months.

Toynbee and Walker go on to say in their book that:

But Blair hit his target.  Asylum numbers were cut in half in six months.

The truth is quite ironic.  Blair didn't meet what Polly says his target was -- a halving between February 2003 and September 2003 -- in both months the number of asylum applications was about 4,250 (forgive the imprecision -- I am reading this numbers from a graph on page 3 of Asylum Statistics United Kingdom 2004, prepared by the Immigration Research and Statistics Service, Research Development and Statistics Directorate of the Home Office.  They don't provide the numbers behind the chart, but my reading should be accurate to within +/- 100.  pdf link to the report here).

OK, that bit wasn't ironic, but this is.  Blair did hit his actual (and not Polly's mangled) target -- asylum applications fell from 8,770 in October 2002 to some 4,250 in September 2003.  And asylum applications did fall by half in six months -- 8,770 in October 2002 to about 2,500 in April 2003 -- just not the six months Polly is talking about.

Looking back on the figures, it is interesting to note, parenthetically, that Blair promised to halve the number of asylum seekers (compared to October 2002, as the Home Office rapidly 'clarified') in February 2003, in which month the number of asylum applicants was fewer than half that for October 2002 already.  Committing to deliver the recent past seems a little unimpressive.

Toynbee and Walker go on to claim that:

The number who succeeded in getting refugee status fell fast too.  In 2002 34 per cent were successful, but only 11 per cent by 2004.

Compare that to this, again from Asylum Statistics United Kingdom 2004, but this time page 14:

Of the initial decisions made in 2004, 1,565 (3 per cent) recognised the applicant as a refugee and granted asylum, 160 (0.3 per cent) granted HP and 3,835 (8 per cent) granted DL and 40,465 (88 per cent) were refusals.

[emphasis added.  'HP' means humanitarian protection and 'DL' means discretionary leave]

So that'll be 3%, and not 11%.

Friday, June 02, 2006

Numerical fragility

Today's Polly Toynbee column says of the tax credit system:

That means 20% will be overpaid and 10% will be underpaid. How bad is that? Not all that bad. The standard error rate of the social security system for means-tested benefits under all governments has always been around 10%.

Actually, if you are overpaying 20% and underpaying 10%, that means an error rate of 30%, considerably higher than the already generous 10% error rate Polly cites.

She then goes on to say:

But most of these over- and underpayments are not errors. The whole point of the system is to make sure families can keep on an even keel when their incomes change. Overpayment happens if they fail to declare when their household incomes rise, when earning more, moving in with an earning partner or stopping childcare, or when a child reaches 18. Families are underpaid if they fail to claim when their income drops or they have another child.

Overpayments also happen because of government error, of course. The IFS, cited approvingly later on the article in a different context, has some interesting research here on this. To save you reading the entire thing, consider just the headline:

Government paying tax credits and benefits to 200,000 more lone parents than live in the UK

When Polly does cite the IFS, she says:

according to the Institute for Fiscal Studies, tax credits have delivered "the longest sustained fall in child poverty since records began" - 700,000 fewer children now live below the poverty line, back to levels of 20 years ago

I think a reasonable person would think from this inaccurate précis that the IFS were claiming that the fall of 700,000 children in child poverty was due to the tax credits. In fact, their research (pdf link here) says on page 46 that:

The size of the population effect depends on the overall change in the number of children and on the average poverty risk. The table shows that the fall in the number of children reduced child poverty by around 42,000

On page 49, they go on to say that:

To summarise a complicated set of changes, child poverty fell primarily because

  • there were large falls in the risk of poverty for children in workless families, those with part-time working lone parents and those in couple families with one full-time parent and one non-working parent

  • there was a substantial decline in the proportion of children living in workless families.

So part of the fall was attributable to a fall in the number of children, and part was due to a 'substantial decline' in children living in workless families. Which is not to say that the tax credits haven't helped, or that a fall in unemployment is a bad thing. It is to say that there is more at work than the tax credit system.

And finally, in the very paragraph where she talk of Oborne's "numerical fragility", she misses the point of an average, as represented by the idea of the so-called 'Tax Freedom Day'. In a nutshell, it represents the proportion of total national income which is paid to the government in the form of tax as a day on the calendar. This is an average. Polly rightly implies that the day will occur sooner for people earning less, however she wrongly states that it will occur on that precise day for the "very.very rich, of course -- the Notting Hill people." Averages rarely represent the true figure for individuals (the old cliché is that no-one has the average number of legs).