Thursday 4 January 2018

Statistics are altogether too bourgeois for the public sector elite


I've been thinking for a while about Graeme Archer's article on how 'getting' statistics is as important as 'getting' Shakespeare yet people still, in that giggling way, exclaim how they were never any good at maths.
There is no “either/or” between words and numbers; at the same time, of course, there is. CP Snow’s two cultures exist, most clearly in the words/numbers dichotomy, but only because we structure them into existence through conscious decisions about how to educate our young. It’s a human choice, not a law of the universe. Good with words or good with numbers is a story we tell about ourselves, and like all stories it has its heroes, and villains.
Now I'm a social scientist (if I'm anything properly academic) who once gloried in the title of Research and Planning Director and I recall a number - see what I did there - of stories about folk dumbfounded by maths and stats. We once prepared an analysis of customer geography for a knitting wool company (sexy, eh!) based on a simple index and had to spend the first twenty minutes of the presentation explaining what indices were. And when I did my 'Research Methods' module for my MSc, we were told to avoid quantitative research because (I summarise) maths is hard and you don't need to do it.

Such is the reality of a world where my personal experience of buying a railway ticket is a better guide to policy than statistics explaining why I am an outlier in buying such an expensive ticket. Or where - as I put it in a debate at last year's Battle of Ideas:
We’d reckon on uplift in response of around 2X or maybe 3X compared to a random selection. Great until you realise that the response to random was around 0.2% - all that clever technology means that, instead of getting ignored by 998 out of 1000 people, you only get ignored by 994.
Only the 2X or 3X gets reported not the actual numbers even though they tell us we needn't fret so much.

Statistics matter yet the extent to which we understand them or appreciate them is troubling. I recall being told - in very definite terms - that my comments on the relative poverty line were wrong because I was using the median household income not the mean and "median isn't an average". This is despite the fact that median is the right measure where there's a lower bound (i.e. you can't have a negative household income - which might come as a surprise to bankrupts).

The thing with Graeme's article, however, is not that people don't know enough maths and statistics but that they are proud of this fact. And, you know, I might have just found a little glimpse at the reasons why this is the case - numbers, counting, accounting are terribly bourgeois things. Here's a chunk from a chapter in Deirdre McCloskey's Bourgeois Equality (the chapter is entitled 'Aristocrats Scorned Measurement'):
In England before its bourgeois time Roman numerals prevailed. Shakespeare's opening chorus in Henry V...apologises for showing battles without Cecil B. DeMillean numbers of extras: yet "a crooked figure may/ Attest in little place a million/ And let us, ciphers to this great accompt, / On your imaginary forces work." The crooked figure he has in mind is not Arabic "1,000,000" but merely a scrawled Roman M with a bar over it to signify "multiplied by 1,000": 1,000 times 1,000 is a million.
McCloskey - like Graeme Archer - quotes Samuel Johnson (albeit a different reference):
"To count is a modern practice, the ancient method was to guess; and when numbers are guessed they are always magnified."
And that modernity was, in Johnson's time, the advent of Bourgeois England - Capitalist England (although McCloskey would be cross at the sloppiness of this latter designation). Before this time, accounting was a grubby thing of merchants, pursers and bursers not something the great and good should be bothered with. Indeed, McCloskey goes on to note that, prior to 1803, students arriving at Yale were expected to be fluent in Latin and Greek but not to be qualified in arithmetic. Such mundane things as statistics were beneath such grand folk.

We appear to have come full circle with quoting poetry and qualitative (for which, usually, read anecdotal) analysis given greater credence than any sort of proper measurement, analysis and appraisal based on statistics, probabilities and other scary number-crunching things. As Simon Jenkins disdainfully put it (when it seemed his basic statistical knowledge was found wanting):
I seriously doubt this poll, since it implies that two-thirds did know the answers. All on whom I tested it failed, including myself. Nor could they see the point. With one voice they replied: “That’s the sort of thing you learned at school.” So what is the point?
What were the sorts of question? Calculating the mean, median and mode - terms used every day in the Guardian where Jenkins writes. The area of a circle. And even long division and multiplication. Not even GCSE maths, not much advanced on what we'd like an eleven year old to know. We don't need to know this stuff, say the bookish elite, it's smelly maths.

Yet how can you know whether government policies - reducing poverty, improving health and so forth - are effective if you've no idea how the things we're reducing or improving are measured? I'll tell you how - we use anecdote, stories. So, rather than look at the statistics we relate the tale of someone's grandma or how a single mum is struggling on benefits. We sometimes gather together several of such stories and embellish them with commentary from 'experts' relating their own understanding based on further anecdote - or 'qualitative research' as they like to call it.

Statistics, maths, measurement, accounting, calculation - these are the tools of the capitalist, the merchant, the city trader. Thing to be disdained and dismissed in favour of our, often pre-judged, opinion founded on at best a reasoned assessment but more commonly on a collection of tales. We might then hold our noses and wander into the world of maths seeking out some numbers that seem to match what we've found from our tales. People compare apples with pears, use percentages when the numbers are too small for this to make sense, and calculate means when the bounds make the mean meaningless. Journalists - often with the title Science Editor or Health Correspondent - present headlines that bear no relationship to the research they report on. Not because they want to mislead but because they've no idea what 'statistical significance' means let alone what a p value is.

Hardly a day passes without us witnessing people in powerful and influential positions dismissing the evidence of statistics in favour of anecdote and conjecture. And, so long as people can respond by quoting Disraeli - lies, damned lies and statistics - without being laughed at, there is little chance that we'll see the proper use of measurement in the formulation of public policy. Meanwhile, out in the world of markets, capitalism, trade and investment, people have no choice but to use measurement and statistics, without them they would fail. For the public sector elite, however, such concerns are just too bourgeois and instead they make policy inspired by stories in the Guardian or Daily Mail.

.....

2 comments:

Anonymous said...

As some of us learnt in the corporate world, "If you can't measure it, you can't manage it".

But 'measuring it' can produce as many, if not more, uncomfortable answers as positive ones, hence the preference amongst the bullshitting classes for the qualitative, hence more malleable, approach.

Not forgetting that there are two types of statistics - those you look up and those you make up. 95% of statistics are the latter (see what I did there?).

Anonymous said...

Yet the same people will lambast Google, for collecting statistics.