Should income disparity be a predictor of populist activity? As income disparity can reflect a middle class that is “left behind,” it’s suggestive. Certainly examples of populist anger over the last 100 years – from 1930’s Germany to the present – are associated with a disenfranchised, resentful, and struggling middle class.
However, a real populist movement also needs a target for resentment, a feature that goes beyond simple disparity numbers. For one thing, that target can be real or imagined. Look no further than Trump’s use of resentment against immigrants and minorities as an imaginary target. We might expect – I haven’t looked at the numbers systematically – disparity to align with populist movements, but not to stand alone.
For a moment though: let’s imagine we took a hard look at the data, and found that income disparity plus a target of resentment always associated with a populist movement. And we’ve built a nice stable “predictive” model to represent this fact. Would that mean resentment and income disparity “predict” populism? It’s hard to say no – after all, what more could we possibly want? But I (differing with some of my friends on this point) say the answer is still no – it’s not really a prediction.
For a true prediction – saying that one thing leads to another – comes from a theory or interpretation lying outside the data. In my hypothetical example, if we set aside what we think we know about the problem, what can we really say from the relationship alone? That income disparity + resentment gives us populism, or that populism (from some other cause) gives us income disparity + resentment? If we’re really careful about our assumptions, I don’t think we can pick between these options, without introducing facts or assumptions outside our data set. In addition, without a theory to work from, we can’t determine if historical data are relevant to predictions for the future.
From this perspective, “predictive” analytics is supportive, and purely empirical models are incomplete. In practice, we do know that predictions based on purely empirical models can be error prone. Even when we do a good job of crafting a “predictive” model in a manner that extrapolates our existing information, without the context of a theory rude surprises often lie in store. That partially explains the triumph, then tragedy, of US election polling and predictions over the last 12 years. One of our greatest challenges is not so much the mechanics of analytics, as appreciating the limitations of results and deciding when a potential conclusion should be avoided. There can be tremendous pressure to conclude something.
Returning to income disparity and populism: if we work with the thesis that income disparity is, at least, a motive force behind populist action, another question arises. How can the incoming US government succeed? Income disparity cannot be relieved overnight, particularly not by a ruling party that believes in minimizing taxes on the rich. It also cannot be relieved by a ruling party that has no firm plan of governance, other than presidential musings and a mandate to deconstruct current policy. And, it cannot easily be relieved in a government and country now more divided than it was even a year ago. The Republicans may come to live the age-old problem of power: the means, methods, and personalities seeking power are either incompatible with its wise administration, or make its administration impossible.