Some thoughts on uncertainty in politics

We do not easily understand uncertainty and probability. Something as basic as a “30% chance of rain” can cause problems when it comes to the various ways we can interpret such a statement (cf. Gigerenzer 2005). How can we deal with even more complicated uncertanties and probabilities in the domain of politics? I am not sure, and I do not have a unified framework to think about these questions. Instead, in this post, I bring a few (random) thoughts on uncertainty in relation to politics.

There are very few things, if any, we can be certain about in politics. However, the one thing politicians do in order to win elections is to convey certainty, at least in relative terms compared to the competition. It is difficult as a politician to both demonstrante competence and being uncertain at the same time. Accordingly, I do not find it interesting to discuss how politicians acknowledge and communicate certainty. Instead, I find it much more interesting to reflect about uncertainty in the domain of politics, and in particular different kinds of uncertainty.

I have been reading different studies on the effects of acknowleding uncertainty when communicating scientific findings, i.e., outside the domain of politics. In general, a core finding is that accepting uncertainty in predictions and estimates is not problematic when communicating scientific findings. On the contrary, it can increase trust in science and scientists.

Howe et al. (2019), for example, show how acknowleding uncertainty in climate predictions can increase the confidence in scientists and acceptance of the message being communicated. Joslyn and LeClerc (2013) showed how the provision of uncertainty estimates in weather forecasts increased trust in the forecasts (see van der Bles et al. 2019; 2020 for more on how to communicate uncertainty). A recent review by Gustafson and Rice (2020) shows that, in most cases, acknowleding the inevitable uncertainties of science will not harm the effectiveness of the message.

What about politics? Why do politicians not acknowledge the inherent uncertainties in whether implemented policies will work as intended? I believe there are both supply side and demand side explanations. First, for the supply side, politicians are, for the most part, not trained in thinking about uncertainty, risk, probability, etc. One might even expect that the people who are more likely to run for office are the people who are more certain that they can make a difference if and when they are in power (and maybe that is one of the many reasons why I would never run for office). Second, for the demand side, if two politicians differ in the level they can guarantee that a specific policy (or set of policies) will have the intended effect(s), why not vote for the less uncertain politician? My guess is that we pay more attention to the confidence in the promised (average) policy effects rather than the confidence intervals related to such effects when we evaluate politicians.

Politics might simply not a great place to acknowledge uncertainty. A recent study, for example, found that ideological extremists are more likely to be certain about their correctness of their political beliefs (Costello and Bowes 2022). In other words, it might even be seen as a weakness, even though we can make a strong case for why it should be a strength. When you have strong ideological beliefs, you are certain that you are not uncertain.

However, it is a bit more complicated to talk about uncertainty in politics. There are many ways to talk about different types of uncertainty; risks and uncertainty, epistemic uncertainty and aleatoric uncertainty, etc. In the book Radical Uncertainty: Decision-Making Beyond the Numbers, John Kay and Mervyn A. King operate with two specific kinds of uncertainty: resolvable uncertainty and radical uncertainty. Specifically, they write:

Resolvable uncertainty is uncertainty which can be removed by looking something up (I am uncertain which city is the capital of Pennsylvania) or which can be represented by a known probability distribution of outcomes (the spin of a roulette wheel). With radical uncertainty, however, there is no similar means of resolving the uncertainty – we simply do not know. Radical uncertainty has many dimensions: obscurity; ignorance; vagueness; ambiguity; ill-defined problems; and a lack of information that in some cases but not all we might hope to rectify at a future date. These aspects of uncertainty are the stuff of everyday experience.

My concern is that when we talk about uncertainty in politics, we talk more about resolvable uncertainty than radical uncertainty, i.e., issues and events that can be represented by a known probability distribution of outcomes. Or, actually, it is even worse: we treat radical uncertainty as if it is resolvable uncertainty. This is a feature, not a bug. Part of what politicians do is to take radical uncertainty and turn it into resolvable uncertainty. That is, uncertainty in politics is not about reducing the probability of a negative event from 30% to 28%, or the magnitude of such an event, but turning uncertainties (something that we cannot easily quantify) into risks (something we could, in principle, turn into a number).

This is why we tend to underestimate the role of significant unpredictable events in politics (or black swans, to pay tribute to Nassim Nicholas Taleb). This is not to say that we should not make predictions (predictions are difficult, but not impossible, to make in complex social systems, see e.g. Hofman et al. 2017). However, even if we can put numbers on uncertanties and risks and design a model allowing us to make predictions, what a model can predict the world of today might not be able to predict the world of tomorrow. One reason is that the sources of a problem can change over time. Bowlsby et al. (2020), for example, show that the factors that can predict political instability in one time period are not necessarily good predictors in other time periods. And this is not merely a data issue that will be solved by big data (the infamous Google Flu Trends is a good example on the limitations of big data to predict the future, cf. Lazer et al. 2014).

Uncertainty in politcs is also heavily affected by the fact that political time horizons tend to be short. As described by Tyler Cowen in Stubborn Attachments: A Vision for a Society of Free, Prosperous, and Responsible Individuals: “Voters are keen to receive more government spending now and postpone the required taxes to the more distant future. Few governments do everything they can to promote economic growth for the more distant future.” We have a lot of uncertainty in politics, especially when it comes to the future.

Furthermore, the problems we deal with in politics will often have more uncertainty than, say, problems in economics. That is, fixing problems in politics is, at the end of the day, a much more challenging endeavour than fixing problems in economics. Or as Abba P. Lerner described it more than 50 years ago: “An economic transaction is a solved political problem.”

There is something about uncertainty in the domain of politics that is different from uncertainty in other domains of life (as always, this is not an original point, but something that first-year Political Science students know all about). In politics, even defining what uncertainty is, and what we are uncertain about, is political. A simple example if climate change where climate change deniers clearly rely on a different notion of uncertainty and try to question every single fact. That is, we need to have a certain level of confidence in what we do in order to deal with political issues.

However, our level of certainty will in and by also shape the level of uncertainty we should operate with. The COVID-19 pandemic, for example, is a self-defeating prophecy. The more serious you take the pandemic, the less likely it is that you encounter the virus. The more you you don’t believe in the pandemic, the more likely it is that the pandemic is real. We are all in the treatment group in a global pandemic, but some people believe they are in the control group. Consider this conclusion written by Maarten Boudry in April 2020:

As the novelist Frank Herbert once said: “The function of sci-fi is not to predict the future, but to prevent it.” That leads us into a strange paradox: the better we manage to contain this pandemic, the less we will learn from it. Because there is one thing you can bet on for sure: as soon as this whole crisis blows over, the same minimalists will come forward and claim that it wasn’t as bad as the “fearmongers” had told us. Indeed, some of them are already busy committing that very fallacy.

Similarly, the more certainty we approach certain political issues with, the more uncertainty we will face. That is, the probability distribution of outcomes can in and by itself be affected by our perceived probability distribution of outcomes. The reason why we do not care a lot about whether we get a “30% chance of rain” right or wrong is that it will not affect the likelihood of rain.

In addition, what we understand as “rain” can be affected by the amount of rain we get. One of the reasons is that our definition of a problem will change as we address it. To understand why, it is good to think about prevalence-induced concept change. That is, as there is a decrease in the prevalence of a stimulus (e.g., a specific problem), people will expand their concept of the problem. In experiments, for example, when people are presented with less blue dots, people are more likely to see purple dots as blue (Levari et al. 2018).

Last year, I wrote a post on global risks, climate change and COVID-19. If we look at the The Global Risks Report 2022, we see that infectious diseases are now number 6th on the list of the most severe risks over the next 10 years. My guess is that it will be further down the list in the next report. Not because I have any reason to believe that the risk of infectious diseases have decreased since last year (even in relative terms compared to other risks), but because we are more certain about how to deal with a pandemic. For that reason, we might face even greater uncertainty.

There is a lot to be said about uncertainty in politics, even a lot more than the few thoughts I have outlined here. That is one thing we can be certain about.