Sex stats – size matters and a problem of approximation

Saturday morning. Test Match Special on the radio. Browsing the web.

My eye is caught by a BMJ press release (newsroom page) on holiday sex stats (pdf).

There are lots of percentages. The BMJ has helpfully interpreted those in words – something I strongly recommend in my training workshops.

Two interpretations have caught my eye:

BMJ-men-women BMJ-61.5-almost-two-thirds

In the first, almost 58.5% is described as over half, ie over 50%. True – but over half is a big range! And 58.5% is a long way over half.

In the second, 61.5% is described as almost two thirds. Two thirds as a percentage is 66.7% (to one decimal place).

The difference between 61.5% and 58.5% is just 3 percentage points. They are closer to each other than they are to the 66.7% and 50% to which they are respectively compared. Two thirds and one half differ by a whopping 16.7 percentage points.

Both 58.5% and 61.5% are 1.5 percentage points away from 60% aka three fifths.

So 61.5% could be described as “a little over three fifths”, and 58.5% could be described as “a little under three fifths”.

Does this matter?

Yes. When we convert percentages to words it makes it easier for others to understand, and to remember. A politician might glaze over at 61.5% and 58.%, but be comfortable with “almost two thirds” and “over half”.

A problem comes when approximations in words – easily recalled and reported onwards – get turned back to numbers. Things get lost in translation.

Let’s use our BMJ numbers: what if a debate becomes one of two thirds versus one half – of 66.7% to 50%. Not of 61.5% and 58.5%. How would any decisions be affected?

Alternatively, what if the words used had been “a little over three fifths” and “a little under three fifths”? Would recalling both as three fifths – a difference of zero percentage points – be less problematic than the 16.7 percentage point gap?

In the end, size matters. And so does approximate size. Choose your words carefully!


If you need to…

● make sense of the numbers, data and statistics you work with daily

● have greater impact with the data you collect and the surveys you commission

● develop confidence talking and writing about the numbers that matter

… I can help with training and consultancy.

Advertisements

A charity’s view on policy need not be a view on politics – they should enter the EU debate

Should charities speak out in the EU debate?

There is no doubt that Brexit carries with it big risks for the not-for-profit sector. Will post-Brexit Britain reproduce EU regulations, initiatives and partnerships that advance charities’ objectives.

And what of income? Most economists say Brexit is very likely to lead to a downturn in the UK economy. And a weaker economy means less money for government grants; less money for individuals to donate.

In an article on the CIPR’s Influence site, Kate Turner, discusses other risks:

“The risks of speaking out are arguably similar to that faced by any consumer-facing organisation or company whose fortunes depend on the goodwill and support of a considerable number of stakeholders. Alienating supporters who take a different view could have a significant impact on donations, volunteers and receptiveness to other key messages from a charity.”

It seems to me, that what is often overlooked is that entering a debate does not necessarily require taking a side.

Saying that an EU regulation, initiative or partnership has been a benefit to a charity’s work, is informative.

Saying that whatever the outcome of the referendum the charity would want to see a continuation of helpful circumstances is also informative.

Surely, a charity can voice a view on policy – regulations etc – without having to voice a view on politics – the Remain or Leave options?

And, by voicing such a policy view, charity-supporting voters (whether tending to Remain or to Leave) get extra information ahead of making their final decision.

Are PRs without data skills set to be left behind?

I doubt anyone imagines that a PR professional lacking skills with words is going to find their career anything but challenging.

But what about numbers?

I’ve been stressing for years that PRs need to embrace their numerate side. It’s not just me. The Global Communications Report 2016, launched at the World PR Forum in Toronto, is just the latest to identify the increasing use of data as a driver of change.

But while the future may see a specialist data analyst be embedded in your team, will you understand what they find?

Will you be comfortable – and, more importantly, competent – in presenting the numbers to the Board?

Could you explain a confidence interval? Or decide which average to use and why – arithmetic mean, or the median, or the mode? How about calculating a percentage with confidence?

And, could you turn your numbers into compelling messages that sway opinions and influence behaviours among your audiences, publics and consumers?

As the Global Communications Report puts it: “writing… might be considered a ‘price of admission’ ability for a communications department”.

In years to come, lack of data skills may not see you refused entry to the profession. But it’ll be those who can handle the numbers who’ll get into the VIP lounge.


If you need to…

● make sense of the numbers, data and statistics you work with daily

● have greater impact with the data you collect and the surveys you commission

● develop confidence talking and writing about the numbers that matter

… I can help with training and consultancy.

Hope you’re well – when does the oil of communication become grit in its gears?

Dear reader,

I hope you are well.

Indeed, I do. After all, what type of person would I be if I hoped you’re not well?

It goes without saying, surely?

Its use by PR practitioners clearly frustrated the Economist’s Anne McElvoy according to her recent tweet…

Anne-McElvoy-tweet-hope-you-are-well

I have some sympathy – nearly every time the phone rings and it’s a sales, marketing or fund-raising call, the caller opens with a scripted “how are you today?” I hate it. I feel put on the back foot, politeness requiring me to respond to a faceless stranger about my personal wellbeing. I’ll be itching to put the phone down, whatever they have to say.

But this has got me thinking about the flipside.

Why do we pepper our communications, spoken and written, with phrases – comments and enquiries – that are arguably redundant?

It seems that, in many ways, they are the oil for the gears of human interaction; reducing friction and helping things run more smoothly. Without them the gears still work but they grind.

So, some attention to personal circumstances is a good thing. For example, what I do like from unsolicited calls is those which open by asking if it is a convenient time to call. That, to me, shows proper concern that I may be in the middle of something pressing.

Of course, it’s not one oil for all applications. Butter might help get a tight ring from a finger, but it’ll play havoc with your two-stroke, petrol-fuelled hedge-cutter.

What we may think is a friendly gesture, others find unnecessary, insincere, even creepy. It can also get tiresome in its overuse: note that Anne McElvoy tweets about seeing the same statement four times in one morning. The oil of communication has become grit in the gears.

It’s tough enough getting a journalist’s attention; worse still if your email triggers a negative reaction that has them reaching for the delete key before they’ve even finished the opening sentence!

I don’t think there any golden rules here. But, till a relationship is established, my feeling is less is better than more: keep to the point, and be politely formal.

Your thoughts? Are there phrases that you automatically use, or put your back up the moment you read them? Do you have your own rules of engagement?

In the meantime, dear reader, I hope this has found you well, and that you’re having a great weekend. Love to you and yours xxx !

The stats have it – PRs want to do stats!

abacus

That’s according to number-crunching by the Chartered Institute of Public Relations (CIPR) on the goals set by participants in its continuous professional development (CPD) scheme.

Number 4 in the top ten most popular goals is “use data and statistics to communicate better”.

It’s vital that PRs realise that high quality data, suitably analysed, can make campaigns more effective. As former CIPR president, Stephen Waddington, sets out in a linked post: “Public relations is becoming increasingly data driven. That’s a good thing.” And he usefully sets out ten ways PRs can use data.

I’d add that this goal is not just about making communications better with data and statistics. It’s also communicating the data and statistics better.

Percentages, averages, correlations, projections, trends, blips, margins of error, significance.

You might find your analysis compelling and convincing, but can you explain your conclusions and make your recommendations to non-expert audiences – that could be your Board! – in ways that won’t leave their lost and confused?

And if they get lost, how about your designers, and – most important – your target audience (journalists, general public)?

That’s why I’m hoping there’ll be a spike in members reading the CIPR skills guide I put together on using statistics in public relations, or reviewing the associated webinar. Each has five CPD points.

And a plug for me… get in touch if you or your colleagues would benefit from some ‘live-action’ training – whether that’s a tutorial session for  up to four people, a team’s lunchtime bite-size seminars, or tailored half- or full-day workshops.

Naked percentages – and other ways PRs slip up with numbers

I train PR practitioners in working with, and communicating, numerical information.

I’ve put together the following five ways that tell me someone isn’t necessarily comfortable with numbers, data and statistics.

1 – Naked percentages – look away now!

Percentages. No other topic can conjure the same look of terror on those who come to my workshops. And a naked percentage even more so!

And when I’m reading press releases – especially those reporting survey research – it’s the abundance of “naked percentages” that often suggest the writer wasn’t comfortable.

It’s this sort of thing (paraphrased from a real example):

“We found that 8% of women and 9% of men thought of a thing every day compared to only 5% of adults who thought of another thing entirely. Of the people who thought a lot about the thing, 77% stated it was because it made them happy – while 61% said it was because it was a nice shape.”

I call percentages that just stand there in all their glory, naked percentages. I recommend that most – if not all – percentage figures be “clothed”, ie in brackets following an interpretation in words. Something like this:

“We found that a little under one in twelve (8%) women and one in eleven men (9%) thought of a thing every day, compared to only one in twenty (5%) adults who thought of another thing entirely. Of the people who thought a lot about the thing, just over three quarters (77%) stated it was because it made them happy – while three fifths (61%) said it was because it was a nice shape.”

A couple of reasons for doing this … it’s easier for the reader to imagine twelve women and picture one of them in particular. And if you are working with an infographic designer you’ll have given them a head start on their work.

To make this easier for me and others, I’ve put together an Excel spreadsheet to generate phrases you might want to use in your writing. I’ve also developed simple tweaks for Word so that as you type the numbers it automatically inserts an appropriate phrase.

2 – Pi(e) in the sky precision

How many digits of pi can you recall? Here are a few … 3.1419526535879.

How precise do you need pi for practical purposes? If you’re guiding the International Space Station, 15 it seems. But for most of us, four is probably more than okay. And for rough approximations 3 or 3.1 could well be enough.

The same principle counts for PR.

You don’t need that many figures to make your point. Yet, so many people are tempted to go over the top with precision. A paraphrased example: “We’ve lost 45.01 per cent of our social referrals.”

What does that extra “.01” add? That’s down to a level of 1 in 10,000! There’s probably more noise and error in the measurement than that. My experience is that quoting more than three significant figures readily loses an audience – especially if it’s done for every figure you’re presenting.

I think people like the extra precision because they think it means accuracy. But it only implies it. The eagle-eyed among you may have spotted that the figure I’ve given for pi above is wrong. I’ve deliberately switched some figures. Other than for working in space, you’d be fine using it.

And this is another reason to avoid unnecessary precision – the more numbers you have to type the more chance you have of inadvertently typing them wrong. Stick with the 45 per cent not the 45.01 … cos one day you’ll type 40.51 per cent without noticing

3 – Average but untypical – do you mean to use the mean?

Anything in which an average is quoted grabs my attention. Sadly, usually for the wrong reasons.

By “average” most people don’t mean the sixteenth century custom duty. They mean the mean. Specifically the arithmetic mean – count the numbers; then add them up; divide the result of the latter by the result of the former.

Statisticians and other number-wranglers have come up with lots of different averages. All are intended to summarise the data in a way that gives insight into what is typical or expected – in jargon, a measure of central tendency.

Here are nine numbers: 500, 4, 1, 200, 3, 1, 5, 1, 5. What is the average? Here are three ways of giving an answer:

Type of average Result
The arithmetic mean – add all up and divide by 9 80
The median – line up in order and find the middle number 4
The mode – single most frequently occurring 1

The mean is 80, bigger than seven of the nine numbers. The median (4) is within 3 of seven numbers. The mode is within 4 of seven numbers, but is the lowest number of them all.

Which is the most appropriate? In real life it is often a question of what most of your audience will feel is typical of their own experience – get that wrong and no matter how good your message, you’ve probably lost their confidence already.

4 – Data butter or data marge? Spread matters!

The choice of average depends a lot on the distribution of your data – how the numbers are spread out. What’s highest? What’s lowest? Do most numbers bunch at one end or the other, or in the middle?

For a lot of PR work, these questions should be asked before any calculations are done, and definitely before anything gets written up.

And the best way to do that is to visualise the numbers. You don’t have to be a whizz with Excel charts to do this. Grab a pencil or pen and a piece of paper – and sketch. For the numbers above, this is my ten-second effort:

Data-sketch

Anything which is skewed – the bunching is not in the middle but at an end – strongly suggests that the mean is not going to be a good average. The overly high (or low) numbers will ‘dominate’ the calculation and the result will send the wrong message.

5 – A picture paints a thousand words – but does it have a bit of the Jackson Pollocks?

You might not want to present your one-minute sketch. But a good graph can make all the difference to gaining understanding. And a strong infographic can readily go viral.

If you’re using a traditional chart or graph there are some basic rules that should never (well, very rarely) be broken. It’s a long list so I won’t go into everything here.

But here are a few …

Percentages in pie charts should add up to 100 … unless you’re Fox!

Fox-pie-chart

Even the smallest table needs the numbers to make sense … CNN take note!

CNN-table

Bar charts should start at zero … this one looks like it’s been booted over the bar!

Football-bar-chart

And with infographics there is so much more to get right – and wrong. Not least that you’ll get your numbers right but still send a damaging message. For example, be careful that your stylisation of typical people isn’t stereotyping … is this perpetuating notions of what are ‘male’ and ‘female’ jobs?

JRF-figures-stereotypical

Thank you for reading. If you want to know more come on one of my open workshops or courses – or book a bespoke session for your organisation.

Adverts and percentages – is statistical beauty only skin deep?

I have developed a curious obsession.

Lately, I’ve been taking photographs of adverts for health and beauty products.

Most of them are from television. It’s not the products that interest me. It’s those captions at the bottom used to back up some voiceover assertion. Like this:

72-97

(Though it has Helen Mirren, one of my favourite actors, so what’s not to like?)

Some large – or even not so large percentage – is quoted of the number of women (no men’s products yet spotted) who agreed with some statement.

Some more of my photos are at the bottom of this post. Here’s a table of the figures from them:

Total women %age Total times %age Rounded Rounded divided by Total In words
241 74% 178.34 178 73.9% 178 out of 241 women agreed
207 70% 144.90 145 70.0% 145 out of 207 women agreed
194 72% 139.68 140 72.2% 140 out of 194 women agreed
193 73% 140.89 141 73.1% 141 out of 193 women agreed
182 78% 141.96 142 78.0% 142 out of 182 women agreed
114 88% 100.32 100 87.7% 100 out of 114 women agreed
97 72% 69.84 70 72.2% 70 out of 97 women agreed
53 70% 37.10 37 69.8% 37 out of 53 women agreed
30 96% 28.80 29 96.7% 29 out of 30 women agreed

I’ve set out to two decimal places what the percentage means of the total number surveyed. I’ve also done the calculation using the rounded number of women. It looks like most advertisers round down on their percentages.

So, are these impressive figures? Who knows? The problem is that without knowing how the survey was set up, there are just too many ways in which figures like this might actually be mediocre.

These don’t seem to be like the surveys that you might be used to in political opinion polling – usually of around 1,000 voters with a lot of care taken to get a representative sample.

Nor do they seem to be the results you might associate with drug development with their extensive randomised controlled trials.

Cognitive bias

Psychological research tells us a lot about cognitive biases that can undermine any survey. Most people like to agree to questions put to them, for a start.

Even when we agree, we may disagree! It all depends on how the question is asked. I tweeted this screengrab from Twitter:

Fair-questions-paradox

Different wording, different results. Different conclusions? Now I remember my quantum mechanics from my uni days and how something can be in two states at the same time – think Schrödinger’s Cat. But this result ain’t quantum. It’s just how phrasing affects people’s considerations.

So, back to the TV adverts. These companies make big profits. To paraphrase one, aren’t you worth it for them to spend a little more on their numbers?

 

72-194

70-53

70-207

72-97

73-193

74-241

78-182

88-114