open Democracy News Analysis - The numbers game: statistics and politics , Andrew Dilnot Michael Blastland - Comments

harrystottle on "The numbers game: statistics and politics "

harrystottle — Sun, 21 Oct 2007 14:59:23 +0000

(replying to ai_1) and your comment highlights another aspect of analytical perspective. In fact ALL parties to that dispute (tagged versus untagged) are looking at the wrong parameters. The ONLY one that really matters is the global effect on commission of crime. And it is only that parameter upon which we can find consensus - nobody disagrees with the proposition that we want to see overall crime reduced. So all we should be looking at, when measuring the effectiveness of various anti-crime measures is the net effect on levels of crime. Tagging, for example, might turn out to have unexpected (and hopefully positive) effects on crime levels not because of what the technology does to the potential offender, but because of the perceived stigma of having to wear one. Being tagged provides nothing like the street cred of being an ex convict. It is more like a humane version of being placed in "Stocks".

harrystottle on "The numbers game: statistics and politics "

harrystottle — Sun, 21 Oct 2007 14:56:39 +0000

One of the biggest problems for anyone addressing a statistical argument is context. You get close to the problem when you talk about about "are they comparing like with like" but that is not sufficiently explanatory or comprehensive. Consider the question of which form of transportation is safest. Even reputable statisticians will frequently tell us that Air transport is by far the safest because they are calculating deaths per passenger mile. But this is clearly invalid as air journeys tend, for obvious reasons, to be much longer than other journeys (with the possible exception of journeys by sea). What I - as someone who is avowedly afraid of flying - want to know is not what my risk of dying PER MILE is but what is my risk PER JOURNEY. This is much more rational for a number of reasons, not least because by far the greatest danger to a flight occurs at take-off and landing - so the number of take-offs and landings is much more correlated to the accident rate than the number of miles flown. However, the comparison is further complicated by the fact that, when flying, 95% of the risk is in the take-off and landing - so the mileage is almost irrelevant; whereas, when driving, the risk scales very directly with mileage. So short car journeys are much safer than long ones (although a disproportionate number of accidents do happen close to the beginning or end of journeys) If you perform this comparison, you suddenly find that the previously apparently huge risk gap between car journeys and plane journeys narrows dramatically. Scientific American did some work on this in the context of Flying and Driving after the September 11 Attacks and showed that the risk is between 1 and 65 times greater for car journeys depending on the length of the journey. In other words, my gut feeling that I'm actually safer in my car (on a short journey) than I would be in a plane, turns out to be not so daft after all. All of which supports my plea for an International Table of Standardized Risks where all the risk data we ever collect is assimilated into a huge database which can present the data in ANY comparative context as determined by the user. We could choose first, for example, to see transport risks by passenger mile, and then by passenger journey. Or time of day, or country of origin of airline, etc etc. One of my pet peeves is the atrocious abuse of statistics used in the War on Drugs. You can see an example of one of my rants in my stumble comment on a recent BBC Story about Cannabis increasing the risk of psychosis. Yes it does, if you have one or two copies of the wrong variant of the COMT gene and you consume too much of it as an adolescent. But a) the risk if you don't fit those categories is almost non existent b) increased risk alone means nothing unless it is put in context. (If the risk was low to start with - which it was - then an increased risk can still be low) c) other behaviours increase other mortality risks (not just mental damage but fatal) by vastly more than the Cannabis example but aren't ever used as pathetic excuses for the moral panic represented by Prohibitionism. For instance, one of the examples I use in that rant is cycling - which is supposed be healthy and green; but it increases the risk of death in a road accident by 12,666%! (compared to bus journeys) This kind of abuse could be eliminated or, at least, neutralized by the Standardized Table of Risks. On the table of Risk Damage by Substance, for example, we would learn that it is easier and far more common to kill yourself by drinking too much water than smoking too much ganja; that aspirin causes more deaths than heroin (but also has far more users); that prescription drugs - even on a per user basis - are far more dangerous and deadly than any of the illegal drugs - and so on.

tonycurzonprice on "The numbers game: statistics and politics "

tonycurzonprice — Tue, 09 Oct 2007 17:57:58 +0000

e.c.apling - I think the fonts got you confused between GB and G8 ... Tony

e.c.apling on "The numbers game: statistics and politics "

e.c.apling — Tue, 09 Oct 2007 13:15:52 +0000

This, otherwise very good and timely, article is marred by a serious data error. The population of the UK is quoted as 750 million !!! It is of course, possibly, 75 million (is it really as high as that? - the last projection I have seen was for around 61 million by 2011) - but let us take 75 million as a rough figure for quick, rough, calculation - then this means the G8 proposed debt contribution to poorer counties was £10 per person (not £1), or around 20p per person per week (not 2p). If you are arguing about numbers and statistics iit is CRUCIAL to get your figures at least APPROXIMATELY correct......

ai_1 on "The numbers game: statistics and politics "

ai_1 — Mon, 08 Oct 2007 23:44:29 +0000

This is a very good article – up to a point. In order to be comfortable with numbers and to talk in numbers, you must to understand numbers. Sadly, British education is by-and-large a very poor means to understand and be comfortable with numbers. Most British children would encounter their last numerate lesson before the GCSE, and this lesson will be very basic indeed. The entire structure of British education (compared with French, or German, or Italian) militates against the understanding of what numbers mean and that numbers are not made of plasticine or putty in the hands (or mouth) or a politician or a journalist. For it is not just politicians. It is the overwhelming majority of "public intellectuals", and in particular journalists. Graduates of political science, English Lit, media studies, classics or Greats, bright, articulate, outspoken, at home with the cut and thrust of verbal argument and afraid of numbers as devil of the holy water. They speak truth unto power, as long as this truth doesn't require mathematics or scince or engineering. And in the modern world this means very little truth. Politicians get away with it because they are in tune with the politically-aware public: they speak the language of non-numbers. If we want politically-aware public, teach kids mathematics and statistics until 18. Let them be at home with numbers, understand baysian statistics, understand scales of numbers and, for goodness sake, understand the essentials of formal logic and rigourous causality.

tonycurzonprice on "The numbers game: statistics and politics "

tonycurzonprice — Mon, 08 Oct 2007 21:40:12 +0000

If JMK changed his mind when the facts changed, lucky him - it clearly gave him a good indicator of when the facts - rather than anything else - did change. Kuhn, Popper, Quine and Lakatos all developed a more sensible skepticism, allowing either one's mind or one's interpretation of what are facts to change under stress. what changes - facts or points of view - is an endemic problem to social science. Indeed, Dilnot and Blastland give an excellent example themselves: does tagging work? Should you compare tagged and un-tagged people who have recently left prison? or ex-tagged individuals with ex-incarcerates? or the behaviour in prison of those who would come to be tagged against those who would not? The authors imply it is obvious that the "proper" comparison is the second or third, whereas the misleading Jack Straw offers the first. Well ... that depends on what you think "likeness" amounts to in the case of prisoners --- there is nothing obvious about this. "Likeness" is determined by a theory: a belief that "proximity to time inside determines behaviour", for example, would make the first comparison one of "like with like". If you disagree with a claim in the social sciences and you can shoot it down with numbers, then why not go for the quick win. But even if you can't shoot down the numbers, you can usually maintain your disagreement, because what is presented as a number usually clothes a contestable piece of theory. Numbers are often just a way of delaying the time when you talk about _real_ disagreements. Tony

The numbers game: statistics and politics , Andrew Dilnot Michael Blastland

Mon, 08 Oct 2007 15:42:47 +0000

The economist John Maynard Keynes said that when the facts changed, he changed his mind. Politicians at loose with statistics are tempted to a different tack: changing the facts. "Lies, damned lies and statistics" expresses a cynical truth that citizens all understand.

And yet, despite sharing the scorn for statistical sleights of hand, we believe that this saying is one of the worst to infect public argument.

Its fault is to encourage disengagement. If we have nothing better to show than cynicism, we resign ourselves to powerlessness, and turn away, understanding nothing, griping about everything.

Andrew Dilnot is principal of St Hugh's College, Oxford and former director of the Institute of Fiscal Studies

Michael Blastland is a journalist and radio producer who makes programmes for BBC Radio 4. He is the author of Joe: The Only Boy in the World (Profile, 2007)

Andrew Dilnot & Michael Blastland are the co-authors of The Tiger That Isn't: Seeing Through a World of Numbers (Profile, 2007), available from Amazon here

The wiser alternative, the engaged response, is to find the means of seeing through what the politicians are up to, some way of challenging and understanding the numbers. If we fail, the debate on the big political issues becomes meaningless - on everything from health risks, foreign aid, government spending, social change, to targets, league-tables, comparisons, surveys and samples, about hospitals, schools, migrants, crime, and climate change. To dismiss numbers, whether through fear or loathing, is to give up the game on almost every cause you love or hate. For numbers are now, whether we like it or not, for good or ill, the dominant language of public argument.

So what can citizens do? The first task is to begin understand the tricks played on us. Fortunately, this is often easier than it looks, and acquiring a more accurate measure of a political statement can sometimes be preposterously easy. Here are three such examples of how politicians' rhetoric can deceive, and recommendations for a way of responding to it.

The scale of words

The first such ploy is over size. Faced with a few zeros on the end of a number, many of us are quickly confused, and even readily impressed. Politicians enjoy playing to that tendency, and throw large numbers about with bravado.

So pause a moment. Then offer a challenge that appears naïve, a question of banal simplicity, but one that touches the most underrated problem with the way numbers are produced and consumed: "is that a big number?"

Though it will strike some as painfully obvious, the problem is persistent, and the media frequently taken in.

Take the agreement reached by the G8 on developing-country debt at its Gleneagles summit in July 2005. At a face-value of $50 billion, it initially appeared a gesture of startling magnitude, and was indeed helpful to some countries. But ask the question: was it a big number?

$50 billion was the total stock of debt, not the annual cost to the G8, so a better measure would be to say that the G8 wrote off annual repayments of about $1.5 billion a year, divided between eight of the richest countries of the world. Put that into pounds sterling and it equals about £750 million. Conveniently, the population of the G8 is also about 750 million, making Gleneagles equivalent to £1 per person, per year... or 2p a week. Though since at least some of the money came from existing planned aid budgets, it was probably less than that.

The principle here is that numbers relating to the developing world, the G8, or even a single country, will all, most likely, be in the millions, billions, or even trillions. They trail zeros simply because there are an awful lot of people about in such huge economies or large sweeps of the globe. The simple, sensible way to get a number into proportion, indeed the only way to begin to know if it is genuinely big or small, is to divide it by the whole population it affects or represents. On a global or national scale, it's hard to make sense of the numbers. Convert them to a personal one, and it's easier.

The reality gap

A second signal to watch for is how well politicians' grand rhetorical promises are supported by the data. This simple check should not be the kind to escape media attention, but it often does. The question, "does the data match the promise?", may seem in one sense trivial, but it can be surprisingly powerful.

So when Gordon Brown, as Britain's chancellor of the exchequer, said in March 2006 that the government's intention was to raise annual education spending on state-school pupils (about £5,000 per head) to the average level of spending for private-school pupils (about £8,000 per head), it was greeted as an act of bold significance as he announced the first stage in that process - an additional £600 per head of already planned and genuinely new spending.

Also in openDemocracy on statistics, media and politics:

* Jean Seaton, "The numbers game: death, media, and the public" (6 October 2005)

* Mary Dejevsky, "Russia: what demographic crisis?" (27 September 2006)

* Michel Thieren, "Deaths in Iraq: how many, and why it matters" (18 October 2006)

* Phil Gunson, "Bolivarian myths and legends" (1 December 2006)

* Gunnar Heinsohn, "Islamism and war: the demographics of rage" (16 July 2007)

Check the ambition against the declared long-term spending plans, however, and it turns out the government is beginning to slow rapidly from the high rates of public-spending growth of the last few years. At a rate implied by current plans, it would take until about 2018 for the government's aim to be realised, assuming - and this is the nub - that private-school spending per head stayed still. Unsurprisingly, it doesn't. Even at the rapid rates of rising state- education spending since 2001 in Britain, the gap between state and private schools has not closed. Yet in future, this rate of increase on the state side is likely to slow down. The words do not match the numbers.

To compare rhetoric with data is a task we should be able to expect of any self-respecting newspaper or broadcaster. If they don't do it, then we must withhold judgment on the politics, however great the promises.

A respect agenda

The third trick to look for is one of comparison. The question to ask is: "are politicians comparing like with like?" This is a question for whenever we are offered a league table or international comparison. And we are offered a growing multitude of them. Being top, or bottom, is a boast for any government and ammunition for every opposition. But it is seldom what it seems. A quick example makes the point

Christopher Pollitt of Erasmus University, on a visit to Finland, was surprised to discover a category of prison - open prisons - with no escapes, ever. "You never have anyone escape from an open prison?" he asked an official. "Oh no, but because they are open prisons we don't call it an escape, we classify them as absent without leave."

Imagine such differences of definition or value when we are offered, as we are, rankings of countries' whole healthcare systems, and where what constitutes a good system is in many ways a political judgment, not at all a simple metric.

A recent home secretary in Britain was responsible for another example of bogus comparison: he managed to compare released former prisoners with those currently electronically tagged, to claim that since the tagged commit fewer crimes, measured over a shorter period, tagging is better at preventing reoffending.

But either you are let out early wearing a tag or you are not let out until later. So the alternative to tagging is jail, not liberty. That means the proper comparison group for the tagged is the incarcerated, either when both are serving sentence, or when both are fully free, not when one is still serving sentence and monitored, while the other is out and about. Like with like?

Often, such maneouvres are accompanied by a huffing-and-puffing insistence from politicians about their right to interpret the figures as they see them.

But this is more serious than the expression of a point of view. It is about having respect for data over wishful thinking, respect for the public, and for themselves. If you'd prefer to be flattered by bogus numbers, to believe that the world changes when you play statistical games, or at least to act as if it does, you are, let's be blunt, delusional and dangerous.

The citizen trying to come to terms with the blizzard of numbers that often now passes for political argument can easily feel daunted. But the principles we need in order to begin to see through the tricks politicians and others play are often relatively straightforward, if we can bear to apply them.

The great advantage of a more engaged reaction to numbers is that it raises the political cost of abusing them. So the first improvement to public argument with statistics is most likely to come from us, citizens and voters. Let's ask these three questions, consistently and more often.

Average rating

(3 votes)