Metadata matters

Nobody forms a conclusion based on objective evidence alone. This is especially true when it's something related to a specialist field and we don't trust our own judgement. We don't want to start jabbing needles in our eyes - we prefer to go to an optician, to benefit from the experience of someone who's already stabbed lots of people in the eye, or better yet, has read all the medical journals describing in great detail why you probably shouldn't do that and should just use eye drops instead.

It makes perfectly good sense to trust an expert when you (a) recognise that you yourself are not an expert and (b) recognise that somebody else is. For example, dinosaurs are jolly interesting, but I don't want to go out and dig them up myself. I want the experts to do that for me. It saves a great deal of time and stops me getting all dusty.

This means that when I read about the latest argument over whether T-Rex was a terrifying super-predator or just an overgrown vulture, and I want to reach my own opinion, I have to heavily rely on metadata. Instead of looking at the bones myself through a high-powered microscope, I'm considering other evidence : specifically the qualifications and experience of the experts involved. Sure, I might well have my own preference based on the issues presented, but if the article describes the findings of Professor Sir Joe Blogs, Oxford Chair of Palaeontology, I place a lot more weight in it than if the exact same thing came from the mouth of Foul Ole Ron, professional drunk.

Buggrit, buggrit, millennium hand and shrimp.

This is all well and good when it comes to esoteric debates of little or no actual import. Jurassic Park not being a documentary, my opinion on dinosaur appetites is about as pointless a thing as it's possible to be interested in*. Less reassuringly, a very similar process affects politics.

* Just because it's pointless doesn't mean it's not worth doing. I'm a big fan of doing things just because they're cool. Hell, I research clouds of gas that don't do anything, so I know all about pointless research.

This press release describes an effort to measure how this works in the real world. Specifically, when a party changes policy, do voters shift allegiance away from that party, or change their opinion of that policy to agree with their preferred party ? Does policy drive opinion, or does opinion drive policy ?

It seems that policy drives opinion to no small degree. Party loyalty, it turns out, is not a small factor.

Slothuus and Bisgaard discovered that among those voters who backed the Liberal Party or the DPP, policy opinions overall “immediately moved by around 15 percentage points in response to their party’s new issue position,” while those of other partisan groups were mostly unchanged. By surveying the same individuals throughout the study, Slothuus said, the researchers could ensure the effect they observed at the party level wasn’t being caused by people shifting party allegiances in response to the parties’ shifts on the issues.

“We can see that these welfare programs were actually quite popular… and many of the voters of the centre-right party were in favour of these welfare programs,” Slothuus said, referring to unemployment benefits and early retirement. “Nevertheless, we can see that they reversed their opinion from supporting these welfare programs to [immediately] opposing these welfare programs.”

That they followed individuals is important. It's of course still possible that the party base has changed its overall opinion, so changing policy and bringing the rest on board, but here we see very clearly that it's not a small shift : tribal loyalty is important. Although it's certainly a popular anecdote that people oppose or support policies based on who's saying it, this is real data in support of that. And it's somewhat surprising, given how much effort politicians go to in formulating policies to win support. Why don't they just tell everyone to shut up and get on board instead ?

Presumably there are limits to how far this can be pushed. If Nicola Sturgeon decided to announce that Scotland was a dreary place and would be better off consigned to the dustbin of history, support for the SNP would crash. If Keir Starmer decided that Labour would officially endorse killing and eating poor people, Labour wouldn't last long. And for a real-world example, witness the collapse of the Liberal Democrats.

Of course, when it comes to trivial issues, it makes sense to support a party you generally align with. That is in part what representative democracy is for : having experts do the legwork. But that makes it all the more interesting that the policy shifts described in the article were major and of real importance. So I had to go and read the original paper.

It's a commendably thorough piece of work. The bottom line is that this shift in opinion is real and by far and away the most probable causal factor is the change in party stance. They were very careful to account for any confounding factors.

One thing that's key to this study, unlike previous similar cases that they cite, was that they were lucky. They didn't have to contrive an experiment or tease data from the archives. They weren't expecting any kind of policy shift of the issues they'd happened to be polling, but there was. And they sampled quite frequently, so they got to see the change in opinions happening very quickly. They also have a reasonable representative sample size (~1,200 individuals, though only 292 of which identified with a political party).

The issues they monitored were a policy to reduce unemployment benefit and abolish early retirement, both in Denmark. These were not issues of minor consequence or emotionless fiscal management :

When the new rules took effect in 2013, 34,000 individuals were falling out of the unemployment insurance system in the first year; two out of three left without any other government income support. This overnight shift in the parties’ policy position sparked intense criticism in the news media.

Due to its popularity, the early retirement program enjoyed a “taboo status in Danish politics,” and the “taboo had been kept intact by the Liberal-Conservative governments since 2001. The minor party in the center-right coalition, the Conservatives, had tried to push for this policy change from within the government, but the Liberals “had previously flatly rejected doing so,” and the Liberal Party “therefore found itself with a considerable problem explaining the party’s new position”.

So these are ideal cases. Two emotive, important issues on which official party policy was unambiguous and the shift was linear, with no indecisive flip-flopping from the leadership. And they noted that respondents correctly remembered how the party stance had changed and the policy implemented, so it was clearly communicated to voters. It wasn't that their memories had been clouded. They knew exactly what was happening.

One potential problem is that there might be some underlying event which causes a widespread change of opinion among the electorate at large, thus with policy change being a consequence of opinion change rather than a cause of it. But their sample is large enough to rule this out. True, a significant caveat is that it took place among the backdrop of recovering from the financial crisis. But there was no wider movement around either of these changes - they were sudden, spontaneous decisions. And, crucially, while those supporting the parties changed their stance, those not identifying with the political groups did not change opinion at all. Since they followed individuals, they can be certain that opinions really did change, and they're not just seeing a change in voters for each party.

Another concern is that the overall position of party members was either neutral or somewhat positive regarding the eventual change well before it happened. As I've noted many times, rallying the troops is quite a different prospect to recruiting new ones : to persuade someone to like something a bit more is a lot easier than getting them to start liking something in the first place. But even here the results are clear. The shift in preference was about the same across all party members, including those who were actively opposed to it previously. Those members didn't suddenly jump fully on the bandwagon, but they became more supportive of the policy all the same. There was no evidence of any backfire effect at all.

The shift in opinion was rapid. For the case of cutting unemployment benefit, their sampling constrains the shift to a two month period exactly coinciding with the policy change : that is, immediately after the party supported the new policy, opinions shifted. This rapid shift, coupled with the lack of change in opinion of voters not associating themselves with the parties concerned, makes it extremely unlikely that there was any other external event that could be a common cause. At most, I suppose the financial crisis could prime people for being more receptive towards the new policy than they might be at other times, but by itself, it clearly hadn't encouraged a shift in opinion whatsoever.

The change in stance was rapid, strong, and prolonged. If anything, they say, it's stronger than what's seen in more controlled conditions. And despite the criticism, opinions remained flatly constant after the switch for months afterwards (for as long as polling continued).

The only issue I can think of that isn't covered here is what steps, if any, the parties took to persuade supporters that the new policy was a good idea. But that two-month switch is astonishingly rapid given the emotive nature of political issues - again, see the Liberal Democrats after they decided to support higher tuition fees. Either the Danish political leaders must have had a truly outstandingly genius bit of argument, or party affiliation plays an enormous role in persuasion all by itself.

Of course, anecdotally this isn't all that surprising. But to demonstrate it objectively is important. As they say, it complicates the democratic process if parties quite literally, and however unintentionally, determine what people think. The ideal is that voters independently examine the evidence about major issues and then vote for the parties that best reflect their conclusions. But this shows the mirror goes both ways : the voters reflect their parties.

Why might people think like this ? One reason is the use of metadata mentioned earlier, with voters treating their preferred parties as experts to defer to. Another is identity-based politics, with political affiliation being such a strong part of people's identity that mentally it may be easier for them to switch opinions than switch parties (these two interpretations are not mutually exclusive). I would add a third prospect : that once some in party change their opinion, their behaviour towards other members changes, and a combination (usually subtle) of praise and shame can bring the rest of the group in line. But this is likely to a slight effect when it comes to such a small group influencing the opinions of a much larger one. It would have been nice if they'd gone back to at least a few individuals and simply asked them why they thought their opinions shifted.

This is all very interesting in terms of psychology. But it may not matter a fig in terms of policy : if party loyalists change their minds, that says nothing about how voting share changes. In these cases the answer was "very little", about 2-3%. So these quite radical policy shifts made little difference to party allegiance. But we perhaps shouldn't expect it to, since individual policies don't always have much impact on voting intention.

Whatever the exact cause of the shift, that political parties may such a large role in shaping opinion is interesting by itself. I cannot think of any issues where I've suddenly changed my mind as a direct result of politicians changing theirs... but I expect that's because the Labour party told me everything was fine and there was no need to worry. Phew, for a minute there I was really worried.

EDIT : A hypothesis. In accordance with the principle that you cannot reason people out of positions they haven't been reasoned into, perhaps these two examples where people took a stance on a more-or-less purely emotional basis, not a rational one. It follows that if they initially adopted (or opposed) the policies based largely on emotion, then they could be persuaded to change their stance based wholly on what other people were doing. If so, then the same would not hold for other policies on which they had formulated their opinion more carefully. The difficulty with this hypothesis is that it seems unlikely that so many people would happen to have adopted both favourable and unfavourable opinions of the policies in this way; one would have thought that at least some of them, especially given the sample size, ought to have been more rational. Yet I cannot escape the strong feeling that, though I've certainly had irrational factors play a major part in influencing my, I myself have never acted this irrationally about policy - and I don't think I'm exceptional in that regard either. Indeed, I don't think I know a single person who is utterly committed to every policy of their preferred party.