Corrected version of a dodgy election graphic

donut chart of election results

I posted earlier today about a dodgy election graphic doing the rounds on social media (see left). I tweeted it out and put it about on Facebook. And wow – my blog stats went through the roof.

With yesterday’s post about another dodgy election graphic I’ve had over 1,000 hits yesterday and over 3,000 hits today. This is incredible for a niche vanity blog such as this, which does well to get 10 hits a day.

Anyway since the blog post got circulated I found a Facebook thread where a fella named Marrick Gaeafau actually modified the graph and added some extra context. He kindly agreed that I could share it on here.

I asked him where he got the data from and he said:

“The actual figures were gained from my own record of the elections. I’ve been keeping them since 1979”

Impressive data gathering from back in the day before it was available via the internet. Even if you don’t trust his data, it can be checked on Wikipedia here for 1995 and here for 2006.

The revised graphic is below along with a pro-Corbyn commentary that demonstrates once you have an accurate picture of what has gone how you interpret it matters too. The interpretation is where the debate should be.Proper possible election graphic

I’ve had over 2,500 hits on my two posts on dodgy graphics and I’ve had one person challenge me about why I’m doing this. He suggested that as someone on the left I should not be unpicking this as it damages the left. I disagreed with him and you can read the Twitter exchange here.

Before I provide the commentary on the graphic I want to outline why, despite being a public sector trade union official, and long time Labour Party member / activist I have critiqued the two dodgy graphics. My reasons:

  1. If you drive a car with ice on the windows you are likely to crash because you cannot get accurate feedback on your progress due to the distorted (or blocked) view you have because of the ice. If we want Labour to win the general election in 2020 we need to plot our course based on an undistorted view of progress, otherwise we will most likely crash. We do ourselves no favours if we allow ourselves to have a distorted picture of reality.
  2. I, like many other Labour Party activists rightly pour scorn on the Lib Dems for their leaflets with their legendary dodgy bar charts like this and these. We are stinking hypocrites if we indulge in our own version of dodgy graphics, and we lose the moral high ground.
  3. It damages Labour’s credibility, especially Corbyn supporters, to be associated with this. Corbyn supporters are already characterised by a hostile media as “unrealistic dreamers” and circulating such distorted graphics only serves to reinforce that characterisation.
  4. It is fundamentally dishonest to promote dodgy graphics. I’ve done my own fair share of trying to show the impact of this Tory government by using data sensibly such as an analysis (here and here) of council cuts that shows the most deprived councils are hit the hardest. Such woeful data representation from the left, such as shown on this post, gives the right evidence to say we cannot be trusted to tell the truth.
  5. We should never be afraid of the truth even if it tells us things we don’t want to hear. Telling small lies to ourselves just escalates into telling bigger and bigger lies to ourselves and we then end up somewhere very ugly.

Going back to the revised graphic with the more relevant info on it. Of course, this can be spun as a criticism of Jeremy Corbyn’s leadership. I don’t want to do that here, as there are plenty of people on social media and the mainstream media doing it and I don’t need to add to it; I’m going to give a pro-Corbyn spin on this instead.

For the record this is an academic exercise as I’m not sure I agree with all the arguments, but they are legitimate points of debate. My purpose for stating these points is not to argue for (or against) Corbyn – this post is more about the correct use of data – but I do want to demonstrate you can have a pro-Corbyn analysis of the results without resorting to distorting the way you use the data.

  1. Scotland was a mess well before Corbyn was leader and not of his making. It is far too soon to be expecting his leadership to fix these problems.
  2. Success in the mayoral elections in London and Bristol.
  3. Of course the council results were not good because the party was divided because of the public sniping from dissident MPs. That cost Labour votes. [Side nerd point: this is in fact a confounding variable that you’d want to control for. Therefore, if dissident MPs want to demonstrate Corbyn is electorally damaging they should stop their sniping and let him “fail on his own terms” to prove it is him and not their sniping that is the problem].
  4. The row over antisemitism lost Labour votes. [Side point: who is at fault for this becoming an issue is a whole other argument]
  5. In 2015 Labour suffered the worst election defeat since 1987. It is going to take a long time to turn the party’s fortunes around.
  6. They are actually a lot better than the loss of 100 – 150 councillors that were being predicted by the media before the elections.

Let’s not fool ourselves, compared to 1995 and 2006 these results were not good. But the above points do show it is still possible to make a pro-Corbyn argument around these results without using misleading graphics.

None of us should ever be frightened of the truth. Unless you have a clear and undistorted picture of where you are, you cannot make the right decisions to get to where you want to.

And before I sign off a repeat of my top tips to look out for when looking at graphics and data on social media:

  1. The graphic should have its provenance on it. That way people know who has done it and can track down the creator to ask questions. Not having the provenance on it doesn’t mean the graphic isn’t correct, but it should set alarm bells ringing.
  2. The graphic should say where the data was sourced from, with a link to the data if at all possible. If you cannot track down the data that created the graphic, then be very wary.
  3. If some significant calculations or data analysis is required then there should be a link to the spreadsheet or other analysis that was done so it can be checked. Remember the study by highly respected academics Reinhart and Rogoff, that purported to prove austerity worked; it turned out to have a spreadsheet error that made their conclusions invalid. It was only because they made their spreadsheet available that this error was spotted.
  4. If the graphic purports to compare things – ask yourself is it a fair or false comparison?
  5. If the graphic proves what you want it to, remember confirmation bias and ask yourself if you are just believing what you want to believe. Next, ask yourself if it proved the opposite of what you wanted, what would be your criticisms of the graphic? Finally, ask why these criticisms aren’t valid even though the graphic proves what you want.
  6. It if looks too good to be true, it might well be.

As the late, great physicist Richard Feynman once said

“the first principle is that you must not fool yourself – and you are the easiest person to fool”

Edit added few days after the original post: My claim in the blog title that this was as “corrected version of the dodgy graphic” may have had some issues. Several commenters have made some very valid comments of the revised graphic (read them in the post comments). So in the interests of fairness I think it is important to highlight this rather than let readers assume it is wholly correct and only find the criticisms if they dig down in the comments.  I believe it is better to be upfront about this. I am of course unhappy at having posted the graphic without casting a critical eye over it first. In my haste on a Sunday morning to get the item posted I didn’t check it. My bad. But there should be no shame in acknowledging this. I would however point out, that my criticisms of the original graphic still stand and this second graphic still has some merits, especially in the use of NEV, but it was a mistake for the graphic not to point out it was using NEV. I’m going to post more about this (and other political graphics) when I get time as I think it has thrown up some wider issues especially on whether or not the use of NEV in the revised graphic was reasonable (I think it was).

Advertisements

36 thoughts on “Corrected version of a dodgy election graphic

  1. I’m neither one side or the other. I like about three quarters of Corbyn’s policies, I’m not sure he’s the right person to sell them to the electorate though. My objective was to highlight the stupid use of statistics to present a false picture. This will not succeed in galvanising the troops, as I am sure the original meme intended, but rather it will lull them into a false sense of security. Let’s have the painful truth.

    Liked by 1 person

  2. Hi Ravi, interesting blog. I’ve been sharing these figures, so it’s interesting/potentially sobering to learn their provenance (or lack thereof). I was directed to your blog via Twitter.

    When I first saw these numbers quoted – not through the graphic, as it happens – they were framed as being for England rather than the whole UK; which may still not be accurate, but would I suppose alter the reading somewhat. Grateful for your thoughts/insights, best wishes, John

    Like

    • Hi John – actually I have no idea if the original graphic was the whole of the UK or just England as it did not cite its sources (which was one of my complaints). Whether or not the figures are for England or the UK does not alter the fact made in my original bog post about the graphic – it is making a false comparison across the years. It (the original graphic) is still meaningless.

      Like

      • I think the original graphic was for England, as it refers to “councillors”, who were only being elected in England. The percentages in the graphic represented percentages of councillors elected, rather than percentage of the vote, and as far as I could see they were correct on that basis. Of course, under FPTP you can achieve a similar or higher number of election gains with a smaller percentage of the vote, but this will also be true at the next general election.

        Like

  3. Fair comments but no context.
    Like the fact the first two were interchangeable and supported by Murdoch and his cronies. While Corbyn is monstered by the media and bitter losers in his party.

    Liked by 1 person

  4. I do agree that simple graphics don’t tell the whole story and I applaud what you’ve done here. Having said that I don’t think the revised graphic is doing a much better job. It doesn’t tell any of the story of the 1995, 2006 or 2016 elections and why it is difficult to draw conclusions from comparisons. Also I’m dubious about the worth of only measuring change in seats without context. Solely examining change ignores potentiality.

    For example in the 1995 election there were 10880 seats being contested. Labour started from a point of 3109 seats and ended up with 4916 seats. Therefore they contested 7771 new seats and won around 23% (1807) of them.

    In 2006 there were a total of 4310 seats being contested. The Conservatives started from a point of 1514 and ended up with 1830. Therefore they contested 2796 new seats and won 11% (316) of them

    In 2016 there were 2769 seats being contested. Labour started form a point of 1344 and ended up with 1326. Therefore they contested 1425 new seats and lost all of them AND lost 1.7% (23) of their starting point.

    This draws no conclusions about election success but it does begin to demonstrate the difficulty in winning seats when you’re contesting low numbers AND you’re starting from a higher proportion of them. The reduced potentiality is because you’re trying to change the vote of a committed core vote having already won the “easier” more transient votes.

    Anyway, as I say thanks for putting these posts up. It’s only through having an honest assessment of data that we can look for root causes in why people vote.

    Liked by 1 person

    • Little to disagree with in what you’ve said. One of the key points about this sort of analysis is what you compare with what; and what that actually means. Your comments address much of this – but there is more. I intend to blog more about this aspect if I get time.

      I absolutely agree with your sentiment “it’s only through having an honest assessment of data that we can look for root causes in why people vote.”

      Like

  5. Doesn’t this whole article miss the point? Yes, the shared graphic is wrong, but so is the revised one. 2016 council elections were a good showing because they were reflected against 2012, which was a very strong win by Milliband when the Tories and the LibDems were on an “omnishambles” low, and Corbyn pretty much kept all those conquests Milliband made. The problem with Milliband was never his performance in 2012, but that it plummeted after.

    I see this as the strongest pro-point and it’s missing. The explanations below just seem to be excuses for a poor result, which it wasn’t. It was a fine result in England.

    Like

    • You make some valid points about the 2012 baseline and I could and should have made that point. Thanks for pointing it out.

      But I don’t think you can call losing seats (even from a high baseline” and control of a council “a fine result.” That is stretching it.

      Like

      • It was barely any seats, and no councils (net) and the tories lost more. I think the idea that you can keep going up indefinitely is silly. If that was a general election it would have been a Labour victory in England (except the results wouldn’t carry over because, yes, it doesn’t work that way).

        I think to properly compare this you would need to dig out full results for all relative elections historically (what was the one before, 2008? 2004? etc etc) and then compare, preferably with some notes on story and poll conditions at the time.

        Like

  6. To quote Mark Twain, “Three kinds of lies – lies, damned lies and statistics”. Neither graphic gives “context” – to do so you would have to say what each of the results is being compared with: is the comparison with a “high point” – the MSM compared Corbyn’s performance with Miliband’s, but Miliband won a large number of seats which this time round were being defended so to “improve”, Corbyn would have had to have won seats where the Tories were successful in 2012; or is it with a previous “low” performance? We also need to know turnout and how this affected the result (this will be difficult, if not impossible) – eg in Labour’s “worst GE result since 1987” the total Labour vote was c200,000 short of its 2005 vote (9.3 million to 9.5 million) – when Blair won his third term with a 60-seat majority – and about a million up on its 2010 total.
    I agree with your general points regarding reservations, but even so both graphics are weak. As an accountant, when preparing a budget I must state the assumptions made and when showing comparative figures, must compare like with like. Not to do so would make the figures meaningless.

    Like

  7. Thanks for your post – very good point about using correct data.
    My concern is that pretty much all data is quoted out of context – lies, damned lies and stats…

    In 1995 Blair inherited Labour mid-term on a failing Tory administration, after 2 years of policy review under John Smith, which in turn built on work by Neil Kinnock – they were very much in the ascendant.
    In 2006, Cameron inherited the Tories when Gordon Brown’s administration was rapidly running out of steam, again the tide was with him.
    Corbyn has inherited Labour when the Tories are (over)confident and, as other commentators have said, on the back of an electoral disaster. His policy review is only just underway, so local councillors could only really offer ‘socialist values’. He faces an uphill battle.

    In my view, none of the other candidates for Labour Leadership, collectively offering same-old same-old, stood any chance of being sufficiently distinctive and with policies that would really challenge Cameron. Get nearer to 2020, with a roll-out of Corbynite policy, and things could be very different. During the local elections, the unsubtle application of a “Tories can’t be trusted” mantra, smacks of a quality prelude to an electorally strong formula.

    Of course, if we are to make the “Tories can’t be trusted” mantra stick – we need to use trust-worthy data!

    Like

  8. Looking at the “corrected” graphic….

    “Won [n]% of the vote” – those look like calculated equivalent vote shares, not actual vote won.

    The last time an opposition lost councillors was 1992, which was followed by a *Labour* landslide. (Admittedly that election is generally considered an outlier as it was only a month after the general election; but still wrong.)

    Number of councillors lost in 2016 is now out of date (though sources seem to differ on what the correct number is – BBC has 18, Guardian has 11; with Conservatives losing either 48 or 49).

    The big red crosses seem a bit excessive when the numbers actually were correct insofar as they went. (Adding further graphs to add context might be more useful.)

    Which isn’t to say that these isn’t a more representative presentation of events than what it replaces; I just feel that memes complaining about others’ use of statistics have a bit of a responsibility to get their own statistics right, and that seems like quite a lot of errors for that many pixels.

    Like

  9. Ironically, the corrected graphic makes the same error as the original but reversed: it compares the Projected National Share in 2016 (31%) with the true vote in 1995 (48%) whereas the original compares the Projected National Share in 2016 (46%) with the true vote in 2016 (47%).

    Liked by 1 person

  10. Labour did not win 31% of the vote! That was a calculation of the vote share *if* every seat in the country had been up for election. And I think they are considering Labour’s bad results in Scotland when calculating that.
    The actual share of the vote in the 2016 English local council elections was:
    Labour 46%
    Tories 33%
    LibDems 8%
    UKIP 7%
    Greens 4%
    Ind 1%
    Others 1%
    Source: https://twitter.com/britainelects/status/729399561400295424

    Liked by 1 person

    • Thanks for pointing that out. I didn’t do the graphic but I will pass on the comments. I think what was done there was the “National Equivalent Vote Share” which is what would have happened if there had been a general election.

      Like

  11. I would also add that reading self-congratulatory figures (such as the false ones referred to here) can make one complacent: “look at our support! Everything’s going to be fine!”. The last thing the left needs at the moment is complacency; I believe this is one of the reasons why Labour lost the last general election.

    Liked by 1 person

  12. Can you explain why your statistics are ‘the real ones’ and those used in the original graphic are not? Also, is “not very good” a statistic or an opinion?

    Like

  13. Hi Ravi, I think you’ll find that the chart is pointing to the number of councillors won, rather than the share of the vote.

    This is Correct:
    – In 1995 Labour won 5647 of a possible 12153 councillors up for election (46%) (http://www.parliament.uk/briefing-papers/RP95-59.pdf);

    – In 2006 the Conservatives won 1830 of a possible 4312 seats up for election (42%)
    (http://researchbriefings.files.parliament.uk/documents/RP06-26/RP06-26.pdf)

    – In 2016 Labour won 1326 of 2771 sets up for election (48%)
    (http://www.theguardian.com/politics/ng-interactive/2016/may/05/uk-election-results-tracker-2016)

    Yes it probably isn’t the best set of data to use to explain the full results, but when talking simply about the amount of seats won (it is FPTP after all), it is correct!

    Like

    • Thanks for pointing that out. I didn’t do the graphic but I will pass on the comments. I think what was done there was the “National Equivalent Vote Share” which is what would have happened if there had been a general election.

      Thanks for taking the time to reply and put in some links to your source data. Can’t fault you.

      Like

      • It still isn’t a great comparison however as there are different numbers of seats up for election, different authorities (that have their own party bias) and different positions in each data set so you can’t truly compare like with like.

        While the raw data is correct it should probably be weighted to make it more relevant.

        Like

  14. The major flaw with the 46% percentage for 1995 in the original graphic is that it is comparing apples with oranges. This can be seen from the figures presented above by The Greek in which total council seats up for election were 12153 in 1995 as compared to 4312 in 2006 and 2771 in 2016. The 1995 elections included Wales plus numerous English district councils which no longer exist in that form or were not involved in the 2006 and 2016 elections.

    Like

    • Correct, you can’t truly compare like with like.

      While the raw data is correct it should probably be weighted to make it more relevant.

      Like

  15. Pingback: Local elections & Labour – SoylentDave.com

  16. The debate is definitely in the interpretation, and there is far to much interpretation going on which is essentially shallow and based in massive fallacies.

    1) Gamblers Fallacy : Previous result will have an influence on future results. Comparing Labour Landslides, and previous Tory gains, to predict how well Labour is currently doing is fallacious, as they do not have an influence. It’s like saying “England won the world cup in 1966, so we have a good chance now that it is 60 years on”. The previous teams are totally different, as is the political landscape.

    2) There is FAR more going on here, than a simple comparison of “Ooh seats lost” vs “ooh seats gained”, which have incredibly deep influential factors, that cannot be conveyed or explained through a simple election graphic. To name a but a few… EU / Brexit. Scottish independence. Voter apathy (for example, in my area the vote was not for councillors, but for Police Crime commissioners, and we had a voter turnout of about 14% – because honestly, who bloody cares?) and so much more…

    What does and will matter is the General Election. Everything else is just bluster and rhetoric for the media to toss crap around. Noone will be looking back at this in 3.5 years time, when the campaigns Labour/Tory are going head to head to score voters for the next ruling party.

    Like

    • Can’t work out how to make a general reply, so just replying to the last post, but not directly following on from it.

      There is a delightful irony that I’m now seeing this blog post shared on social media as proof of how bad Corbyn’s performance was and how misguided his supporters are. This despite, as has been pointed out repeatedly, that the new graphic does not correct the previous graphic. It may infact be less accurate in failing to distinguish between vote and NEV. There is still nothing in the post to confirm the initial graphic was actually factually wrong as far as I can see.

      I see others have repeatedly pointed out the flaw in a historical analysis based on seats won/ lost. I’ve no idea how what is otherwise a fairly decent analysis can still make such a basic mistake. (To be fair the Guardian article on the same topic also made this basic mistake).

      But it is worth looking at the new graphic in the context of your 6 point ‘test’ for a good graphic:
      1. Provenance – Not included. FAIL
      2. Data source – Not included. Even on being contacted, the answer seems to be ‘trust me I have data, but you can’t see it’. FAIL.
      3. Calculations – This graphic (seems to) include a comparison NEV, but doesn’t provide any working to show that it has been calculated using the same methodology each time. FAIL.
      4. Fair comparison – Compares seats gained/lost despite the fact there were less seats available in total for Corbyn to gain than Blair gained, so even 100% of the seats would have seen Corbyn do ‘worse’ than Blair. FAIL.
      5. Confirmation bias – Author wanted to show original graphic was flawed, this graphic claimed to show original graphic was wrong. Endorsed this graphic despite it failing tests 1 to 4. Avoided confirmation bias? FAIL.
      6. Too good to be true – Exactly proved the captions under each graph in the original that graph was meant to disprove. Graphic true? FAIL.

      The original graphic was poor, but this one is frankly awful in providing an equally poor analysis but using it to rubbish a previous analysis.

      Like

      • Martin your post made me smile – of all the replies yours was the best. I like how you use my own tests against me – can’t fault you.

        I am going to blog more about this as it has clearly rattled some cages and got some good discussion going. So I won’t respond to every point you’ve made as I can cover that in a blog post.

        I guess in posting this revised graphic I made the mistake of not critically appraising it myself before posting it. My bad. But as for the six point test, one points one and two whilst they were not on the graphic they were both given in the blog post. So I’d not give a fail there.

        What the second graphic (as with the first graphic) both failed on significantly, and is the crux of the problem is “fair comparison” – those are the points that many other commenters have made too.

        But essentially you are correct – and I’m humble enough to accept that. We should never be frighted of admitting when we’ve got something wrong. That’s how we learn. I will learn not to post someone eles’s graphic to refute a different poor graphic without carefully checking it first.

        What I perhaps did get right was the six point checklist that you used against me.

        What is interesting is on the substantive point of whether the election results were good or bad for Corbyn, there seems to be no decisive comment that has been made by anyone that clarifies this point.

        Like

      • Martin having slept on it I think your criticism of the use of NEV in the second graphic may be erroneous. NEV gives a better indication of the vote share nationally than the actual vote share in these elections as not all councils are up for election. In fact NEV is similar to but not identical to an opinion poll except of course it is a real election.

        Given NEV is an accepted metric for assessing party support and outcome at a general election I do believe it is a valid comparison to make.

        Like

  17. I found this in the Independent’s comment section, which is another way to crunch the data. Would love to hear thoughts on it as this is not my forte.

    “My nephew has carried out the following analysis of the results:-
    I have crunched some numbers; perhaps you could spread them around; they make different reading to what the BBC is reading.
    What annoys me is firstly they call these elections midterm when they are not and Laura whatsherface said Labour at this stage of the electoral cycle should be winning hundreds of seats.
    Well the figures from previous local elections do not support that assertion.
    The first local elections for incumbent governments are usually pretty good and the opposition makes no or little progress. Also the 2012 locals WERE midterm and so Labour did pretty well. Labour are bucking this trend but not the way the media is reporting it. See below.
    1980 – after winning the 1979 election the Tories gained 40% of the vote only 2% behind Labour.
    1984 – after winning the 1983 election the Tories won the local elections
    1988 – after winning the 1987 election the Tories won the local elections.
    1992 – after winning the GE the Tories won the local elections with 46% of the vote.
    1998 – after winning the 1997 election Labour won the local elections.
    2002 – After winning the 2001 GE Labour polled a mere 1% behind the Tories in share of national vote.
    2006 – After winning the 2005 GE the government does indeed take a pasting which has been put down to the anger over the Iraq War.
    2011 – After winning the largest share of the vote in the 2010 GE and set up the Tory led coalition support for the Tories drops only 1.1% from the GE.
    2016 – After winning the 2015 GE the Tories votes drops 6.9% from GE support but this is not widely reported.
    As you can see in this stage of the electoral cycle, the incumbent government has tended to do well in the following local elections and so the idea that the opposition should be making progress or grabbing hundreds of seats does not bear scrutiny and is completely wrong.
    In fact this is the worse Tory performance in the local elections since 1996 when John Major only got 29% which was an improvement from 1995 when they only got 25% of the vote. But again this is not being reported.”

    Like

    • As you will have seen I’ve made comments on other comments above where I say making a fair comparison seems to be crux of this issue. The comments you’ve provided add another dimension on what would be a fair comparison.

      Thanks for the comments as they have given another perspective, but it is not clear if they actually help us answer the question on whether the results were good or bad for Corbyn.

      Like

  18. Pingback: Another dodgy political graphic | More Known Than Proven

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s