John C. Kirk (johnckirk) wrote,
John C. Kirk

Political propaganda

Here's a bar chart that's been going around on social media today:


I can see why people like this; it looks like a neat rebuttal to UKIP's racism. Unfortunately, the chart is fundamentally flawed.

Just to be clear, I'm not a UKIP supporter: I've never voted for them, and I have no intention of voting for them in the future. However, I think it's important to be honest, rather than sinking to their level. It's also important to understand why this chart is misleading, so that you can recognise similar problems in other charts.

Problem 1: They don't cite their sources.

Where did those figures come from? Did someone just make them up? Just because a friend of a friend of a friend posted this on Facebook, that doesn't make it true!

Think about the various hoaxes and urban legends that keep going around, even when a quick glance at Snopes could debunk them. For instance:
"Microsoft said yesterday that this is the most destructive computer virus ever!"
"The police said yesterday that gang members will kill anyone who flashes their headlights, as part of their initiation!"
In situations like this, you need a specific date, e.g. "20-May-2014" rather than "yesterday", so that you can recognise the messages which have been circulating for months/years. You also need a link to the relevant authority's website, so that you can verify the announcement.

The Mirror (specifically ampp3d) have done some digging into this, and I have to say that I'm impressed. They've cited their sources, given some numbers (rather than just proportions), and explained why some people come up with different totals for the number of UKIP MEPs (1 MEP changed party after he was elected).

Based on that, it looks as if the raw data for the chart is correct. However, that doesn't justify publishing it without a citation; it should have included a link to a website where people could verify it.

According to the government's offender figures (Table 1.6 in that Excel file), there were 547 Romanian citizens in prison on 31-Dec-2013, out of 84,163 people in total. That's 0.6499% (to 4 decimal places) or 0.6% (to 1 decimal place).

There have been 22 UKIP MEPs (including the defector), and 2 of them went to prison:
* Ashley Mote was sentenced to 9 months in Sep 2007, so presumably he was out by Jun 2008. (Source: BBC News.)
* Tom Wise was sentenced to 2 years in Nov 2009, so presumably he was out by Nov 2011. (Source: The Telegraph.)
2 out of 22 is 9.1%.

Problem 2: The proportions are inconsistent.

The first bar shows the proportion of prisoners who are Romanian. The second bar shows the proportion of (former) UKIP MEPs who've been in prison. This may seem subtle, but it can make a big difference which way round you do it.

As a simple example, look at the British monarchy. According to the census data from 2012 (table 2), there were 57 million British citizens resident in the UK. 1 of them was monarch, i.e. Queen Elizabeth. So, the proportion of monarchs who were resident British citizens is 100%. However, the proportion of resident British citizens who were monarchs is 0.000002%.

There are 2 ways to make this a fair comparison:
1) "Proportion of Romanians who've been imprisoned" vs "Proportion of former UKIP MEPs who've been imprisoned".
2) "Proportion of prisoners who are Romanian" vs "Proportion of prisoners who have served as UKIP MEPs".

Which is better? This goes back to the coffee fallacy (which I wrote about in 2006). E.g. it wouldn't make sense to say "90% of prisoners previously drank coffee, so coffee causes crime". You'd need to look at the proportion of coffee drinkers who commit crimes, and compare that to the general population. So, in this case we need the first option, i.e. the first bar should be "Proportion of Romanians who've been imprisoned".

As a side note, if you did take the second option then 2 prisoners out of 84,163 is 0.002%; that's a lot less than 9.1%!

Problem 3: The time frames don't match.

The first bar is looking at the number of Romanians who were in prison on a specific date (31-Dec-2013). The second bar is looking at the number of former UKIP MEPs who have been in prison at any time. The first UKIP MEP was elected in 1999, so that's a 15 year window (1999-2014). To be fair, you need to look at the same period of time for both groups.

Let's take the single day option first, using the date that the original chart maker chose. Using ampp3d's figures, there were 547 Romanians in prison, out of the 105,273 who were resident in England and Wales, so that's 0.5% (to 1 decimal place). Meanwhile, both former UKIP MEPs had been released from prison by that point, so they had 0 out of 22, which is 0%. So, the Romanian figures were similar, but the UKIP figures have changed a lot!

Putting that into a new chart:


As a related point, ampp3d say that 0.15% of Britons were in prison on the same day, so there are a disproportionate number of Romanians in prison (for whatever reason).

Alternately, we can look at the 15 year period. In this case, the 9.1% figure for UKIP is correct. ampp3d didn't have any equivalent data for Romanians, and neither do I, so I'll leave that as an exercise for the reader. Here's an incomplete chart:


All in all, I don't think this is really going to annoy or embarrass UKIP. In fact, I think it's counter-productive. If I can tell at a glance that someone is talking out of their arse then I won't take their views seriously, and UKIP can legitimately claim that their opponents are misrepresenting the facts (which will attract more supporters).

So, a plea to my friends: please don't share charts like this, however tempting they may look.
Tags: charts, politics, statistics

  • Root canal surgery

    This morning I went to the dentist for root canal surgery. Proverbially this is supposed to be quite unpleasant, e.g. "that film sounds so bad that…

  • Knight Rider

    When I was a kid, I watched Knight Rider on TV. I still have a lot of nostalgia for it, and I think it was quite iconic: in the same way that it's…

  • Incremental changes

    The BBC recently posted an article: Tell loved ones they are overweight this Christmas. Other people have disagreed with this, e.g. Fausterella…

  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded