Evidence-based voting

I enjoyed Michael Plant’s article in The Conversation today, ‘Which party’s manifesto promises would make Britain happiest?‘ Plant attempts an evidence-based approach to choosing which party to vote for by reading their manifestos. Despite it being basic and limited, I’m very glad it exists, and I think there should be more attempts to select an objectively best party to vote for.

Call me a radical, but I think people should vote for the party that will do the most to increase happiness. If a party’s policies won’t reduce misery and help people have more pleasant, fulfilling lives, what are they good for?

You may recall that myself and Hugo Burgin attempted a similar analysis in 2016 for the Australian Federal election. As we said then, “We say ‘attempted’ because such analyses are incredibly complex (which is possibly why none exist), although we believe that some attempt at picking the best party is better than no attempt.”

Voting correctly is a lot more important than people often think it is. Again, in 2016 Burgin and I said:

“People often say that you’re unlikely to have any impact when voting, or that the impact of your vote is so small that it’s not worth thinking about, but this is only true if you only care about yourself. In Doing Good Better, Will MacAskill simplistically estimates that the expected value of voting for a US citizen, when spread out across all citizens in USA, is around $5,200 USD (~$7,000 AUD at the time of writing). That is to say, on average, $5,200 of the budget will be spent differently as a result of your vote (see the appendix for a more detailed explanation of why this is so). This means it’s very important to vote for the party that will spend the budget in the best way possible.”

I do have some concerns around Plant’s analysis. I want to stress first that I’m not necessarily criticising Plant for this. He was (presumably) operating alone, with limited time and resources, and also had a limited number of words on the article to play with. I’m just outlining what I would want in an ideal analysis. Having said that, my concerns are as follows.

Plant doesn’t account for non-human animals (edit – he did mention them briefly, I just missed it), which is a major gap, though he does try to account for non-British citizens. He also doesn’t seem to look at the future or far future. Needless to say, far future effects (e.g. 1,000 years plus) are extremely difficult to predict, so again, this is not a criticism of Plant. He relies on manifestos and promises, which won’t necessarily be kept. An ideal analysis would look at history and likelihood of individual parties meeting their promises.

One reservation around these types of analyses in general is that people might use them to come out with the answer they want, whether consciously or subconsciously, although there are ways around this with sufficient oversight.

My ideal outcome looks something like this: A group of benevolent individuals grants an organisation funding say 1 year prior to an election. This organisation can’t be a non-profit in many countries (e.g. Australia), because they are not legally able to support any one political party. This organisation then produces and releases the report shortly before the election. The majority of the population, being motivated by maximising wellbeing of all sentience over the course of the universe (I wish), votes accordingly.

There is a very real question as to how many people would trust such an analysis. There will probably be some people who will never change the party they support out of sheer mistrust that it didn’t pick their party. The trust may have to be built up slowly over several elections and with strong, impartial oversight. I have no idea how to do this, but I do think it is important and worth dedicating time and money to. People have $5,200 worth of impact every time they vote, and we surely want to see that impact being positive.