Common Regression
Expert predictions are too simple
Idea: People make complex predictions about the future which they will be reluctant to replace with simple but potentially more accurate expert predictions
I’ve never liked the phrase ‘common sense’. I’m a scientist. I want to hear specifics about facts and hypotheses that I can independently verify. The idea there’s some magical shared knowledge that’s just obvious goes against how I believe we should understand ourselves and the world. But… I also don’t feel comfortable being completely dismissive, and the reason is because of the success of democracies.
Scientists think in terms of the number of repeats for an experiment, the ‘n number’, as important for how confident you can be you’ve observed a genuine phenomenon. So, what’s the n number for democracy? It’s dozens of countries, representing 100s of millions of people, voting over multiple generations. Democracy is a huge experiment. And what are the results? It’s self-evident that when the whole of a society is asked to choose who should govern and under what policies the trend is towards materially successful and stable countries that progress socially. I’ll say much more about democracy in separate posts, but for now I’ll just leave this observation here as evidence for some kind of shared and effective understanding about the world.
Is there a more formal, scientific way to form an ‘effective understanding about the world’? Regression analysis is a statistical technique for predicting future events based on correlations in data. For example, a typical application might be to predict the chance of getting cancer based on genetics, exposure to toxins, and age, etc. Regression analysis is widely used. It’s often behind many expert predictions.
An example of regression analysis was the predictions that were presented to the British public leading up to the vote to leave the European Union. An example from this analysis was the statement that following a vote to leave the EU, ‘Britain would be permanently poorer by the equivalent of £4,300 per household by 2030 and every year thereafter’. Very specific expert predictions about the economic consequences of Brexit played a major role in the incumbent government’s case for remaining in the EU.
I’m not interested in the accuracy of these predictions. The purpose of this blog post is to examine the weighting people gave these predictions as part of their own analysis about the Brexit vote. This analysis considered much more than just the economic consequences of Brexit. One of the most important topics was immigration from the EU.
People considered things such as: would they be happy with the social change that might come with further immigration? Would people within their social circle or wider community be happy with these changes, and how do they imagine they would react if unhappy? Would their local or wider physical environment sustain more immigration (roads, housing, schools, nature, etc)? Did they trust those with power, in particular politicians, to listen to any practical problems that might arise from further immigration? Would they materially benefit from more immigration, either directly or indirectly?
The questions above are only a small sample of the things people might have considered worth thinking about. And these questions would be relevant across different social scales (personal, family, community, national) and across different time-scales (their own life, children, grandchildren). One of the first things to notice is how incredibly complex this analysis is. Trying to understand and find correlations within the behaviour of dozens to 1000s of people over time and under different scenarios so that predictions can be made is clearly very difficult, no? But, and this is a key point in this blog post, people thinking about the Brexit vote wouldn’t have considered these questions optional. They had to have an answer. And short of an army of experts providing a regression analysis specifically tailored to each voter, these questions would have to be answered by people themselves.
Another thing to notice is how minor the expert regression analysis above becomes in the context of the wider analysis each person is making. Whether they believe it or not doesn't really matter. It’s just too simple to outweigh the more complex analysis each voter is making. And this is the key idea in this post. People will be reluctant to replace a complex analysis, whatever its quality, with a simple analysis, whatever its claimed accuracy.
An obvious rebuttal to the argument above is that those who voted to remain in the EU did accept the expert analysis of the economic consequences of leaving. To that I would say I don’t actually believe this was the reason they voted to remain in the EU. I think their analysis was just as complex and the expert predictions were just supplementary. In fact, what’s never considered is that both sides of the Brexit divide may have been correct in the actual analysis they were making. That is, those who voted Remain would have continued to be happy within the EU, and those that voted Leave would have become increasingly unhappy. It just turned out that on the day there was more of the latter than the former.
Scientists have suggested our brains are prediction machines. This is one of the reasons why I think it’s futile to expect people to replace complex predictions with simple ones in situations where a complex analysis is required. It’s just not how our brains work. People’s lived experience includes a huge amount more information about people and society than an expert prediction would ever include. It’s completely unrealistic to expect people to disregard this information and the predictions they make from it. And if you think these predictions are poor, explain the success of democracies.