Saturday 4 January 2020

Were 88% of Tory Facebook adverts lies?

All election campaigns see politicians exaggerate, stretch the truth and make promises that they can't keep, but if a report issued in early December is anything to go by, the 2019 general election campaign was a particularly dishonest affair - and one party was particularly guilty.

On December 10th, Metro reported:

Investigation finds ‘88% of Tory ads misleading compared to 0% for Labour’

Similarly, the Independent reported:

Almost every Tory ad dishonest, compared with none of Labour’s, research finds

Websites which make no attempt to be impartial were more vociferous. Under the headline, The Tory war on truth – and how to fight back, Open Democracy reported...

Independent fact checkers have found that 88% of Tory Facebook adverts contain lies, while 0% of Labour’s do. 

After the election, the (admittedly risible) Canary asserted that...

The 2019 election was won on the back of lies – and Facebook helped

This was echoed by losing Labour candidate Faiza Shaheen when she appeared on Newsnight the day after the election, as well as by Richard Burgon on Sky two days later.

The BBC made itself unpopular with blowhards when it reported the statistic under the headline: 'General election 2019: Ads are 'indecent, dishonest and untruthful'. This resulted in the usual feeding frenzy on Twitter:

I remarked at the time that the percentage for neither party passed the smell test. I am not on Facebook, but I remember the Tory campaign being about little else than 'Get Brexit Done' while much of the Labour campaign revolved around the evidence-free fantasy that Boris Johnson plans to sell the NHS to Donald Trump.

Whilst Brexit may be a longer process than some people expect, the pledge to get Brexit done was not time limited and there is little doubt that the Conservatives will get it done sooner or later. The Tory slogan is not a lie, whereas the claim about NHS privatisation is the figment of somebody's imagination. As a prediction, it might not be fact-checkable in the way that last year's unemployment figures are fact-checkable, but nor would the claim that Jeremy Corbyn is going to kill the first born.

I hoped that someone might come back and review this factoid after the election, but since no one has bothered, I've done it myself. Going through Facebook's ad library to see which adverts were being assessed is a tedious process, but I got there. My conclusion is that the 88% vs 0% statistic was politically motivated fake news that should never have been reported (credit to the newspapers, including the Guardian, who didn't).

The figure comes from an organisation called First Draft News which describes itself as 'a global non-profit that supports journalists, academics and technologists working to address challenges relating to trust and truth in the digital age'. Its report (which is actually just a webpage) looked at all the Tory and Labour ads online between December 1st to 4th, a period that was explicitly chosen because it saw a large number of new Conservative ads.

According to First Draft, there were 6,749 live Conservative ads on Facebook in the four days studied. That sounds like a lot, but the vast majority were duplicates for each of the 650 constituencies. In practice - leaving aside minor variations - there were only about thirty different Tory ads live on Facebook during those four days. The Labour party had a similar number of unique ads but tended to have fewer duplicates and minor variations. 

To judge the veracity of the adverts, First Draft looked at blogposts from Full Fact, a respected organisation that is Facebook's official third party fact-checker. Full Fact weren't commissioned by First Draft to do anything and they didn't check all the adverts. They have since said that...

...we’ve seen claims on social media misinterpreting First Draft’s article to suggest that Full Fact has found that “0%” of Labour ads in this election were misleading. That’s not true: Full Fact wouldn’t make such a statement, and Labour definitely has released ads that contain claims we’ve disputed.

If Full Fact hadn't looked at an issue, First Draft couldn't judge it, and Full Fact only has so many hours in the day. And it generally doesn't fact-check predictions, such as the one about Donald Trump and the NHS, because anything is possible, although it did say that 'a trade deal is unlikely to fundamentally redesign the way the NHS is funded' and that Labour's estimate of a £500 million hike in the price of pharmaceuticals was 'an extreme and unrealistic rough calculation'. 

This could have been enough for First Draft to conclude that adverts like the four below were misleading (all of them were live between 1st and 4th December):

First Draft chose not to interpret them as misleading, however. Nor did it judge the ads below to be misleading (as we shall see in a moment, it didn't even look at them). These claimed that Labour would not raise taxes for 95% of the population. One of them claimed that Labour's extra spending would be paid for by billionaires. Full Fact disputed this, noting that a lot of people earning less than £80,000 a year would pay more tax under Labour as a result of changes to marriage allowance, inheritance tax, the sugar tax and corporation tax.

The rest of Labour's adverts do not contain any clear factual errors. They are either vague and general, like this...

Or are about specific policies like this...

Or are the kind of attack ads you expect in the cut and thrust of an election campaign...

The Tories might argue with some of those statements, but they cannot be considered factually inaccurate as such, but then nor can most of the statements in the Tory ads. As I expected, the majority of Conservative ads in this period were variations on 'Get Brexit Done'...

When I first saw the 88% claim, I thought perhaps that First Draft had classified any claim about getting Brexit done as inherently misleading. They didn't, and nor did Full Fact. While accepting that there will be a transition after January 31st, Full Fact says:

It’s not fraudulent of Boris Johnson to say his deal gets Brexit done. If approved by parliament, his withdrawal agreement—which is what he refers to as his “oven-ready” deal—would mean that the UK stops being a member of the EU.

As for the rest, most of their non-Brexit ads made defensible claims about policies such as the minimum wage and tax cuts:

And their negative ads generally made plausible, albeit contestable, claims and accusations:

However, there were two Tory claims that Full Fact challenged. With regards to the back-of-an-envelope calculation about households paying an extra £2,400 in tax under Labour, they said...

Labour’s calculations have their problems, but it’s wrong to suggest that every income tax payer in the UK will fund every penny of Labour’s additional spending, and in equal amount.

Although Labour's planned spending increases worked out at an extra £3,000 per income tax payer - and the Institute for Fiscal Studies said 'it is unlikely that one could raise the sums suggested by Labour from the tax policies they set out' - the advert below can be fairly described as misleading:

And then there is the question of hospital building. The Conservatives made several promises about the NHS (in addition to not selling it to Mr Trump) which all feature in the ads below:

The promise to employ 50,000 more nurses was argued about by nitwits after the government revealed that this target would be reached, in part, by getting some nurses who would have left the NHS to continue working for it. Nicky Morgan went viral on Twitter after Piers Morgan (no relation, presumably) affected to not understand the difference between 'more nurses' and 'new nurses'. If there are 300,000 nurses now and there will be 350,000 nurses in five years time, there will be 50,000 more nurses. It's not complicated. That's all Nicky Morgan had to say.

Leaving that nonsense to one side, the Tory pledge to build 40 new hospitals has been questioned by those who say that not enough money has been allocated, that most of the projects are upgrades rather than new builds, and that work has only begun on a handful of them. In their defence, the government has stood by its commitment, saying that the work will take place over two parliaments and that budgets will be allocated when planning permission has been granted. 

Full Fact say:

It’s correct that the government in this announcement has only allocated funding for six hospitals to receive building work by 2025. Up to 38 other hospitals will receive money to develop plans for upgrades between 2025 and 2030, but not to undertake any building work.

We cannot know for sure whether 40 hospitals will be built by 2030. It is obviously possible and, unlike selling the NHS to Trump, it would be politically feasible and popular. Full Fact do not call the pledge misleading or a lie, although they said in October that 'if governments want to issue major policy announcements with large headline figures, they should provide details of where the funding is coming from, or which institutions will receive it and when.' When they returned to the issue in November, Full Fact acknowledged that the Tory manifesto doesn't include costings for 40 hospitals because 'most of them are not intended to be built during the next parliament'.

First Draft assert that the hospital building pledge was 'labelled as misleading' by Full Fact, although I can find no evidence of them doing so, leaving us with two claims in the Tory set of ads that are potentially dodgy. First Draft also flag up two claims about NHS spending and income tax which failed to account for inflation. This is a fair criticism, even if Labour were guilty of the same mistake/trick.

But that is still only four adverts. So where does the 88% figure come from?

This is where it gets weird. First Draft count the '50,000 more nurses' advert as 'misleading', despite it not being misleading and Full Fact never describing it as such (they only noted that the Tory manifesto 'doesn’t say when it aims to deliver this by or whether they’ll all be full-time'). This seems harsh, to say the least.

But the bulk of the '88%' comes from the claim about new hospitals. Of the 6,749 Tory ads reviewed, First Draft say that...

A total of 5,132 ads push the claim that the Conservatives will build “40 new hospitals”, either in the caption, image or link.

5,132 ads! This seems implausible at first glance. The vast majority of the ads that were live at the start of December did not mention hospitals at all.

But the key is in the words 'or link', because...

Not every ad includes the misleading claim directly in its image or caption. At least 54% (3,646) of the total ads served link to a webpage carrying the misleading claims.

In other words, it doesn't matter what the advert says so long as there is a link to a website which contains something that Full Fact has ever taken issue with. Most of the Tory ads linked to this homepage. If you were one of the 1% of people who clicked on the link, you would have had the opportunity to scroll down and see a link to 'Strengthening our NHS'. If you clicked on that link, you might see the words '40 new hospitals will be built'. First Draft consider this to be a false claim and any advert that allowed a person to get to it was judged to be a lie.

That is how First Draft manages to conclude that most Tory ads were 'misleading'. It's absurd.

Most Conservative ads linked to a general landing page which had links to various manifesto commitments. Most Labour ads did not, which is lucky for them because Full Fact found factual inaccuracies in their manifesto. The claim that 88% of Tory ads were 'misleading' is based on the fact that most Tory ads linked to a website where - if you clicked on another link and looked around a bit - you could find a specific promise about hospital building which Full Fact had asked questions about but had not actually 'labelled as misleading'.

This is ridiculous. You have to ask who is really being misleading here? Re-reading First Draft's 'report' after you understand their methodology, you notice the subtle use of weasel words to disguise the fact that there is nothing wrong with most of the ads themselves. They complain that certain adverts 'push the claim' of such-and-such, and they talk about certain claims being 'amplified by' a particular advert. For the most part, this is because the claim they are taking issue with isn't actually in the advert.

Their own claim is this...

Some 88% (5,952) of the most widely promoted ads featured claims about the NHS, income tax cuts, and the Labour Party which had already been labelled misleading by Full Fact.

And that is more misleading than any of the Tory ads.


On December 12th, the day of the election and two days after the 88% claim was all over social media, First Draft published an addendum in which it admitted that the proportion of misleading Labour ads was not zero, as they had previously said, but 7%. The offending adverts linked to webpages which hosted incorrect information about disability hate crime, cuts to the social care budget and the amount of money households would supposedly save under a Labour government.

First Draft said that three ads linked to this page (which has the wrong disability hate crime figures) and four ads linked to this page (which has the wrong figures on social care).

According to First Draft, Labour launched 104 new ads in the first four days of December, although I can only see 92. All but 26 of them were generic duplicate adverts telling people how to find their polling stations, as shown below.

Unsurprisingly, adverts telling people where to vote don't contain any 'misleading' information. The 7% figure is therefore rather flattering, but the addendum provided the first indication that First Draft had not looked at the adverts that were live in the first four days of December, but the adverts that were launched in that period.

As mentioned above, the vast majority of adverts counted by First Draft were duplicates. As far as I can tell Labour launched 22 distinct adverts during this period and the Tories launched 19. In total, the Tories launched over 6,000 while Labour launched around 100, but this reflects the Tories' fondness for duplicating every advert (presumably for different constituencies).

The Labour adverts launched between 1st and 4th December were:

Poll station finder
Green industrial revolution (x2)
You can't trust the Tories
Plan for young people
Plan to fix the economy (x2)
Free tuition
Labour will fix the economy
NHS (with claim that 'Boris Johnson wants to sell it off to the US')
Plan to make life more affordable
Free personal care
Invest in schools
Fix public services
£10 living wage
A million green jobs
Invest in every community
Labour will protect your pension
Cut rail fares
Scrap cashpoint fees
Cut season ticket prices by a third
Compensation for WASPIs

The Tory adverts were:

Get Brexit Done (two different backgrounds)
Let's Get Brexit Done (two different flags)
End the chaos
Boris selfie
Labour's taxes
Inheritance tax
Gisela Stuart backs Tories
Build 40 hospitals
Labour left no more money
Let's get Brexit done and move on (parents with child)
Let's get Brexit done and move on (old people)
Let's get Brexit done and move on (family)
We will cut taxes
Employ 50,000 more nurses
Upgrade 20 hospitals
End low pay work
The cost of Corbyn

On the Tory side, Full Fact said that the £2,400 'cost of Corbyn' statistic was misleading, and it asked questions about the funding of the hospital building programme. On the Labour side, the WASPI pledge was uncosted in the manifesto and Full Fact criticised the 'unrealistic predictions' made by Labour with regards to the NHS under the Boris Johnson. Full Fact also criticised Labour for wrongly claiming that only the richest 5% would pay more tax, although Labour did not launch any adverts on that theme in the narrow window of time examined by First Draft.

Neither party comes out smelling of roses, but the majority of adverts produced by both Labour and the Conservatives in this period contained either defensible pledges, or aspirations that are too vague to be fact-checked. As a package, you can query whether they are affordable and whether the parties were capable of carrying them out, but that's elections for you.

The idea that 88% of the Conservative adverts were 'lies' is itself a lie, seemingly designed to affect the result of the election as the campaign neared its end. It appears to have been designed to help Labour win and has subsequently been used as an excuse for why it lost.

No comments: